Development of FRET‐based high‐throughput screening for viral RNase III inhibitors

Abstract The class 1 ribonuclease III (RNase III) encoded by Sweet potato chlorotic stunt virus (CSR3) suppresses RNA silencing in plant cells and thereby counters the host antiviral response by cleaving host small interfering RNAs, which are indispensable components of the plant RNA interference (RNAi) pathway. The synergy between sweet potato chlorotic stunt virus and sweet potato feathery mottle virus can reduce crop yields by 90%. Inhibitors of CSR3 might prove efficacious to counter this viral threat, yet no screen has been carried out to identify such inhibitors. Here, we report a novel high‐throughput screening (HTS) assay based on fluorescence resonance energy transfer (FRET) for identifying inhibitors of CSR3. For monitoring CSR3 activity via HTS, we used a small interfering RNA substrate that was labelled with a FRET‐compatible dye. The optimized HTS assay yielded 109 potential inhibitors of CSR3 out of 6,620 compounds tested from different small‐molecule libraries. The three best inhibitor candidates were validated with a dose–response assay. In addition, a parallel screen of the selected candidates was carried out for a similar class 1 RNase III enzyme from Escherichia coli (EcR3), and this screen yielded a different set of inhibitors. Thus, our results show that the CSR3 and EcR3 enzymes were inhibited by distinct types of molecules, indicating that this HTS assay could be widely applied in drug discovery of class 1 RNase III enzymes.


| INTRODUC TI ON
RNA interference (RNAi) is an important defence-response system of eukaryotic cells that results in the silencing of viral gene transcripts (Ratcliff et al., 1997;Fire et al., 1998). In the process, small interfering RNAs (siRNAs) play a crucial role in amplification and maintenance of the silencing response through loading onto Argonaute proteins, which uses siRNAs to target and cleave homologous viral RNAs (Chiu and Rana, 2003). As such, RNAi has become a powerful research tool for studies of gene function and siRNA-based antiviral therapies (Tuschl and Borkhardt, 2002;Small, 2007;Badia et al., 2017;Ghosh et al., 2017;Qureshi et al., 2018).
Viruses are some of the most damaging pathogens in sweet potato, and more than 30 viruses in nine families have been reported to infect the crop. Among them, the crinivirus sweet potato chlorotic stunt virus (SPCSV) is one of the most harmful viruses because of its key role in synergistic infections (Clark et al., 2012;Valkonen et al., 2015). Studies have shown that SPCSV can build synergistic interactions with RNA viruses (potyviruses, ipomoviruses, carlaviruses, and cucumoviruses) as well as DNA viruses (cavemoviruses, begomoviruses, and solendoviruses), increasing plant susceptibility to those viruses and causing severe symptoms (Mukasa et al., 2006;Untiveros et al., 2007;Kreuze et al., 2008;Cuellar et al., 2011Cuellar et al., , 2015. Among them, the disease caused by the synergistic interaction of SPCSV and the potyvirus sweet potato feathery mottle virus (SPFMV) is the most common and devastating disease in sweet potato (Gutierrez et al., 2003;Tairo et al., 2005;Clark et al., 2012). Furthermore, it has been proved that the synergistic interaction was mediated by a class 1 RNase III protein encoded by the RNA1 genome segment of SPCSV (abbreviated as CSR3 in this study), and its endonuclease function was necessary for its role as an RNA silencing suppressor in inducing synergistic disease (Kreuze et al., 2005;Cuellar et al., 2009;Weinheimer et al., 2015). Therefore, considering the key role of CSR3, identification of CSR3 inhibitors could be an effective strategy to defend against synergistic virus disease in sweet potato.
Enzymes of the ribonuclease III (RNase III) family are expressed in a variety of organisms including bacteria and multicellular eukaryotes, with their enzyme size ranging from 140 to 1,900 amino acid residues (Court et al., 2013). Among them, RNase III of SPCSV (CSR3) can cleave siRNAs, leading to the suppression of the RNA silencing pathway in sweet potato (Kreuze et al., 2005). In addition, in viruses, another RNase III with similar function has only been found in the animal virus Pike-perch iridovirus up to now (Weinheimer et al., 2015).
Studies have demonstrated that CSR3 sequences are highly conserved in SPCSV that infect wild species of sweet potato, indicating that CSR3-mediated host RNAi suppression is a conserved function that is necessary for infectivity (Tugume et al., 2013). Despite the important role of CSR3 in suppressing the host RNA silencing, no inhibitors for RNase III of SPCSV and Pike-perch iridovirus have been reported, and no inhibitor screen has been applied for RNase III family enzymes according to the Binding Database (https://www.bindi ngdb.org/bind/index.jsp) or Web of Science (https://apps.webof knowl edge.com). At present, several viral RNA silencing suppressors (RSSs) have been reported in plants, such as P19 of tombusviruses, HcPro of potyviruses, 2b of cucumoviruses, and P15 of pecluviruses, which interfere with different components of the RNA silencing pathway (Moissiard and Voinnet, 2004;Burgyan and Havelda, 2011).
However, chemical inhibitor screening has been mainly carried out with proteins P19 and 2b, for example the screening of 5,000 chemicals for their ability to prevent siRNA binding to viral RSS of cucumber mosaic virus (CMV) (2b) and tomato bushy stunt virus (TBSV) (p19) led to the identification of strong inhibitors (Shimura et al., 2008). Interestingly, this screening led to the identification of efficient antiviral agents against turnip mosaic virus (TuMV) later on (Fujiwara et al., 2013). Inhibitor screening for the RSS was mainly carried out by methods such as surface plasmon resonance and electrophoretic mobility shift assay (Sagan et al., 2007;Danielson et al., 2015;Hu et al., 2018).
Fluorescence resonance energy transfer (FRET), which is the nonradioactive transfer of energy between two molecules, has been used for studies of protein-protein interactions as well as high-throughput screening (HTS) assays for drug discovery (Boute et al., 2002;Benz et al., 2011). FRET-based assays are suitable for HTS because the method is sensitive and compatible with small volumes (Klostermeier et al., 2004). FRET has also been used as a probe for tracking the stability, formation, delivery, and degradation of siRNAs in gene-silencing studies by spectral imaging and a fluorescence detection system (Jarve et al., 2007;Alabi et al., 2012).
An effective HTS assay is necessary for screening libraries of compounds to identify candidate inhibitors of enzyme activities (Macarron et al., 2011). Therefore, we developed a cost-effective, FRET-based HTS assay for identifying inhibitors of CSR3 using siRNA as the substrate. The coefficient Z prime (Z′), measuring statistical effect size as defined by Zhang et al. (1999), was employed to evaluate the suitability of the HTS assay by taking into account information from dynamic range and variation of signal measurements.
Moreover, the decrease in reaction rate was defined in terms of the percentage of inhibition (PI) (Schultz, 2006). We used an siRNA tagged with a fluorescein reporter and a quencher as the substrate to measure the PI of compounds. Finally, our success with identifying CSR3 inhibitors with this HTS assay and its successful application on RNase III of Escherichia coli (EcR3) suggests that the HTS can be used for screening inhibitors of other CSR3-like enzymes.

| CSR3 characteristics
Enzymes of high activity are necessary for the success of any HTS assay. For the development of our HTS assay, His-tagged CSR3 and its double-mutant CSR3-A (D37A, D44A) were expressed in E. coli and purified with Ni-NTA agarose. The recombinant CSR3 and its mutant were analysed with sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE), revealing a predominant band at c.26 kDa ( Figure 1a). A second round of elution ( Figure 1a) yielded the majority of CSR3, and aliquots of this fraction were made. In addition, western blotting showed that both proteins can exist as mixed monomer, dimer, and tetramer in storage buffer (Figure 1b).
We also characterized the oligomerization of CSR3 by size-exclusion chromatography with detection using multi-angle light scattering.
The only detectable protein peak at molecular mass 68.93 kDa was larger than that of the theoretical dimer of molecular mass 52 kDa, which could be explained either because most of our CSR3 preparation comprised a mixture of dimers and tetramers in phosphatebuffered saline (PBS) running buffer, or by the nonspherical nature of the dimer, which could cause a disruption during size-exclusion elution ( Figure 1c). In general, the molecule state of CSR3 was consistent with previous characterization of CSR3 by native western blotting (Weinheimer et al., 2014) and with functional and structural studies of another class 1 RNase III from Aquifex aeolicus, which demonstrated that the dimer is the catalytically active form of class 1 RNase III enzymes (Blaszczyk et al., 2001). The CSR3 activity was evaluated using a 200-bp double-stranded RNA (dsRNA) substrate.
This substrate was cleaved to smaller dsRNA fragments in the presence of CSR3 ( Figure 1d) but remained intact in the presence of CSR3-A, as was also the case for the control reaction lacking any endoribonuclease (Ctl). These results validated the activity of the purified CSR3 enzyme.

| FRET assay setup
Our HTS assay was based on FRET, in which CSR3 cleaves a labelled siRNA and generates a fluorescent signal (Figure 2a

| CSR3 titration assay
The rate of increase in fluorescence was dependent on the amount of enzyme and labelled siRNA in the reaction. Fluorescence measurements ( Figure 3a) showed that, compared with a low concentration, a high concentration of CSR3 led to a rapid increase and higher fluorescence at the beginning of the kinetic measurement. To quantify the increase in fluorescence, we calculated the slopes (fluorescence changes over time in seconds depicting the reaction rate of CSR3) between all neighbouring detection cycles. The slopes were used to select the optimal detection time for each CSR3 concentration tested. The results suggested that the maximal initial slope for the concentrations 144, 72, and 36 nM of CSR3 occurred after 5,  (Table 1) suggested that an enzyme concentration ranging from 72 to 144 nM could be used because the resulting Z′ values were >0.9. Therefore, the final concentration of enzyme 100 nM CSR3 and substrate 375 nM siRNA were used in HTS assay.
Moreover, these two control conditions indicated that both the labelled siRNA and enzyme were stable at room temperature (25 °C), an important aspect for HTS given that 20 assay plates were required for our screening and each of them has to be run with 12 detection cycles.

F I G U R E 2
Overview of the fluorescence resonance energy transfer (FRET)-based assay with CSR3 and CSR3-A. Schematic representation of the FRET-based assay. The labelled small interfering RNA (siRNA) was incubated with CSR3 (a FRET-absent condition) or with either CSR3-A or no enzyme (b FRET-present condition). (c) A representation of fluorescence signal curve of the FRET-absent and FRET-present conditions, indicating by real-time relative fluorescence units (RFU) in function of c.17 min total (12 cycles), measured at 37 °C by an optic module with excitation at 485 ± 6 nm and detection at 520 ± 5 nm. Data represent the mean ± SD (n = 24). (d) Agarose gel (2%) electrophoresis of labelled siRNA incubated for 30 min at 37 °C with CSR3, CSR3-A, or without any enzyme (Ctl). All reactions contained 15 µl of 375 nM labelled siRNA. L, DNA ladder

| Determination of the K d value for CSR3
To monitor the kinetics of CSR3 endoribonuclease activity, twofold dilutions (yielding 50-800 nM) of the labelled siRNA substrate were tested with 50 nM CSR3. As expected, start points of raw fluorescence increased with the FRET-siRNA concentration ( Figure 4a). Moreover, the slope of the raw fluorescence between all neighbouring detection cycles was calculated for all cycle numbers.
These data revealed that the maximal slope (2.2-17.1) increased with both FRET-siRNA concentration (50-800 nM) and reaction cycles (2-10) (Figure 4b). Finally, the maximal slope for every la- A similar kinetic study of an endonuclease activity demonstrated that FRET-based methods were more sensitive and reproducible than gel electrophoresis-based methods using radiolabelled substrates (Ghosh et al., 1994).

| Whole-plate validation
Homogeneity within and consistency between plates is crucial for HTS (Ahsen and Bomer, 2005). To assess these aspects of our assay, we used three replicate assay plates containing only positive F I G U R E 3 Titration assay with CSR3 and labelled small interfering RNA (siRNA). CSR3 (two-fold dilution 575 to 36 nM, plus 0 nM control) and labelled siRNA (375 nM) were used to determine the optimal ratio of enzyme-to-substrate concentrations. Three replicate plates (z01, z02, z03) containing labelled siRNA were prepared with a dispenser, and enzyme was added to initiate the reaction. (a) Raw fluorescence expressed in RFU as a function of detection cycle. In total, 25 cycles were acquired (38 min total). Data represent the mean ± SD (n = 48). (b) Slope between neighbouring cycles obtained from raw fluorescence data (a). The highest CSR3 concentration (575 nM) was excluded. Data represent the mean ± SD (n = 48). (c) A linear correlation was found between maximal slopes obtained from neighbouring cycles (b) and CSR3 concentration TA B L E 1 Calculated Z′ values for the five different CSR3 concentrations used in the three replicates over 12 cycles

| Primary HTS screen and a dose-response screen of CSR3
The FRET-based assay was used in a primary screen of 6,620 small molecules of diverse structure. Of these 6,620 compounds, 109 (1.66%) had a PI > 30% (for PI distribution of those 109 compounds, see Table 2). The 12 compounds with a PI value >90% had diverse structures, and no common scaffold was readily apparent ( Figure 6).
A dose-response assay was carried out with the top 109 compounds  Table 3 presents data for the top three compounds based on DSS values. The

| Screening of potential inhibitors for the EcR3
The aforementioned 109 compounds were screened for their ability to inhibit the activity of the EcR3.  (Figure 7). For EcR3, 32 of the 109 compounds had PI ≥ 30%; this was expected because the 109 compounds were selected based on the CSR3 screen. The differences in PI values might be explained by differences in the active sites of the two enzymes, in that one amino acid is different between EcR3 and CSR3, which might be critical (amino acid sequence alignment and active site, Figure S1a,b, respectively).
Specifically, the active site of CSR3 comprises four residues, namely E40, D44, N126, and E129 (black arrows, Figure S1a). The lone difference is D114 of EcR3, which corresponds to N126 of CSR3, and previous studies have shown that, at this position, aspartic acid (D) is prevalent in most RNase III family enzymes (Nicholson, 2014).   (Court et al., 2013;Liang et al., 2014;Nicholson, 2014). Thus, it was expected that the class 1 RNase III enzymes would not have any size selectivity in vitro and can process any size of dsRNA. In addition, because the amino acid identity between CSR3 and Ipomoea endogenous RNase III is very low (<5%), we believe there is a high probability of identifying CSR3specific inhibitors. We selected a 22 bp siRNA with a two-nucleotide overhang along with a specific FRET reporter (FAM) and quencher Based on this, the identified potential inhibitors could bind within the active site of RNase III or bind to a specific location essential for siRNA loading into CSR3.

| D ISCUSS I ON
FRET is a distance-sensitive method requiring a Förster interaction radius of less than 10 nm (Piston and Kremers, 2007).

Moreover, FRET-based methods allow measurements of fluores-
cence changes in real time because the FRET signal decays exponentially in nanoseconds without excitation (Piston and Kremers, 2007). To use FRET in HTS, reproducibility is an important factor owing to the rate of FRET decay, which is highly sensitive to experimental factors such as pH, buffer reagents, and fluorescence quenchers (Piston and Kremers, 2007). To ensure the assay performed well and consistently within and between plates, we carried out a three-plate test with only negative and positive controls.
This test was necessary because small molecules are usually tested at a single concentration for primary-inhibitor HTS (Ahsen and Bomer, 2005;Campagnola et al., 2011;Ilouga and Hesterkamp, 2012). We report here an HTS assay of 6,620 compounds cherry-picked from different compound libraries provided by the

Institute for Molecular Medicine Finland (FIMM). The HTS screen-
ing was carried out for 7 hr using 384-well plates and a 20 µl volume. Considering the sensitivity of the FRET-based assay, it could be possible to use the 1,536-well format and miniaturize the assay to 1-5 µl per well (Rodems et al., 2002). Therefore, our method could be easily adapted to screen larger compound libraries to identify potential new drugs. On the contrary, detection time is influenced by many factors, for example the ratio of the concentration of the enzyme to substrate, reaction buffer, and reaction conditions. In our primary assays, measuring kinetics for eight cycles (11 min total) was sufficient for obtaining good results, yet, we monitored the reaction for 12 cycles (17 min total) to obtain more consistent results. It is noteworthy that the measurement to carry out HTS at c.25 °C, but the catalytic efficiency was rather low, and therefore HTS assay required more enzyme or time to achieve similar results than the recommended 37 °C condition.
We identified 109 CSR3 inhibitors in the assays with a PI value of 30% as a threshold. By raising the threshold, one is able to effectively reduce the number of potential false-positives, and induce false-negatives (Ilouga and Hesterkamp, 2012). The best three com- In the future, the promising candidates could be transferred to plants. Two compounds were tested at several concentrations (0.1 nM to 100 µM) on sweet potato plants coinfected with both SPCSV and SPFMV grown in medium. Preliminary results of the two compounds FIMM027749 and FIMM072436 with different PI values (88% and 41%, respectively) showed that FIMM027749 was able to down-regulate virus accumulation of both SPCSV and SPFMV by two to three times, but without displaying a typical dose-response curve ( Figure S2a,b). Compound FIMM072436, on the other hand, had a mild impact on SPFMV but no impact on SPCSV ( Figure S2a,b).
Their effects on virus accumulation in plants were consistent with the PI values obtained by in vitro assay. At the same time, plant height over time was estimated and both compounds did not have any effect on plant growth (see plant picture in Figure S2c), suggesting that they did not interfere with plant development through nonspecific interaction with host factors including endogenous RNase III. Therefore, the chance of identifying inhibitors that could work in vivo should be high. Nonetheless, these data are still preliminary and more in vivo assays or biochemistry tests will be needed to develop the candidates as antiviral agents in the future.
The identification in a fish DNA virus, Pike-perch iridovirus, of an RNase III protein with a similar RNAi suppression function (Weinheimer et al., 2015) indicates that RNase III-based viral suppression exists in both plant RNA virus (SPCSV) and animal DNA virus (Pike-perch iridovirus), thus it is likely that more viral RNase III acting as RSSs are present in nature waiting for discovery. In addition, our HTS assay was also successfully applied to the screening of bacterial EcR3, validating that this assay could be adapted to other similar RNase IIIs. Moreover, the different screening results between CSR3 to EcR3 demonstrate not only the specificity of the screening method but also the possibility of identifying broad-spectrum inhibitors for class 1 RNase III. Taken together, our findings provide possibilities for developing new antiviral strategies for sweet potato virus disease, and our HTS assay could be used to identify inhibitors of various class 1 RNase III enzymes.

| Size-exclusion chromatography coupled with multi-angle light scattering
Size-exclusion chromatography coupled with multi-angle light scattering was used for characterizing oligomeric states of CSR3.

| FRET assay development
The HTS assay was designed to maximize siRNA cleavage by CSR3.
We with the same conditions. Plates were sealed, centrifuged briefly, and then, immediately analysed with a PHERAstar FS (BMG Labtech) with a fluorescence-intensity optic module (excitation at 485 ± 6 nm, detection at 520 ± 5 nm) for 12 cycles (c.17 min total) at 37 °C. All dispensing was done using the BioTek MultiFlo FX. All screening plates contained positive-control wells (lacking CSR3) and negative-control wells (25 nl dimethyl sulphoxide, vehicle), which were used as standards to calculate the PI of compounds.

| Data analysis
The endoribonuclease activity of CSR3 was calculated with slope, representing fluorescence changes in function of assay time (s), using MARS Data Analysis software (BMG Labtech). In addition, the Z′ of the assay was calculated according to Equation 1 (Zhang et al., 1999), which was used to evaluate the suitability of the method during assay development, optimization, and screening. Moreover, the signal-to-noise ratio and signal-to-background ratio were also used to evaluate the quality of each assay: signalto-noise ratio = c− − c+ ∕ c+ and the signal-to-background ratio = c− ∕ c+ .
where μ c+ is the mean of the positive control, μ c− is the mean of the negative control, σ c+ is the standard deviation of positive control, and σ c− is the standard deviation of the negative control.
The PI of each compound for CSR3 was calculated according to Equation 2. A PI threshold of 30% was used as the cut-off value for one concentration of HTS. Furthermore, the dose-response curves of PI in function of compound concentration were evaluated with DSS according to Yadav et al. (2014).
where S c+ is the mean of the slope of the positive control, S c− is the mean of the slope of the negative control, and S s is the mean of the slope of samples.
The kinetic constant (K d ) for CSR3 was calculated using the three-parameter Michaelis-Menten model (MM.3) included in the R package dcr (Ritz et al., 2015). The statistical significance of differences between values was assessed with one-way ANOVA using the aov function in the R package.

CO N FLI C T O F I NTE R E S T
Patent application FI 20205392.

AUTH O R CO NTR I B UTI O N S
L.W. contributed to planning the research, experiment performance, interpreting the data, and writing the draft. J.S. and S.P. were involved in experiment performance and review. J.V. was involved in reviewing and ensuring infrastructure. All authors read and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.