An Antiviral Drug Screening Platform with a FRET Biosensor for Measurement of Arenavirus Z Assembly

The smallest arenavirus gene product, Z protein, plays critical roles in the virus life cycle. Z is the major driving force of budding and particle production because of a unique property that defines self-assembly. In addition to the roles in budding, Z also participates in the suppression of type I interferon production to evade host antiviral immunity. Therefore, Z and its assembled form are an attractive drug target for arenaviral hemorrhagic fever, such as Lassa fever. Here, we developed a biosensor that enabled the evaluation of the prototype arenavirus, lymphocytic choriomeningitis virus (LCMV), Z assembly using the principle of Förster resonance energy transfer (FRET). This FRET biosensor consisted of three tandem Z that were sandwiched between super-enhanced cyan-emitting fluorescent protein and variant of a yellow-emitting mutant of green fluorescent protein so that Z-Z intermolecular binding via the really interesting new gene finger domain increased the emission ratio. To identify novel anti-arenavirus compounds, the FRET biosensor was employed to screen the PathogenBox400 for inhibitors of Z assembly in a 96-well plate format. The assay performed well, with a Z’-factor of 0.89, and identified two compounds that decreased the emission ratio of the FRET biosensor in a dose-dependent manner. Of them, the compound, 5,6,7,8-tetrahydro-7-(benzyl)-pyrido[4',3':4,5]thieno[2,3-d]pyrimidin-2,4-diamine, was found to significantly inhibit LCMV propagation in infected cells. Thereby, the present study demonstrated that a novel FRET biosensor incorporating Z assembly built on FRET and named Zabton, was a valuable screening tool to identify anti-arenavirus compounds in the context of inhibition of Z assembly.


Introduction
Several arenaviruses are responsible for severe hemorrhagic fever (HF) in humans, such as Lassa fever (LF) and the South American HF (Buchmeier et al., 2013). Although the off-label nucleoside analog drug ribavirin (1-β-D-ribofuranosyl-1,2,4-triazole-3-carboxamide; Rib), is used for the treatment of LF, the mechanism of action of Rib against the viruses is currently not entirely understood (Parker, 2005). In addition, the anti-arenaviral effect of Rib is limited and is associated with a significant side effect (Buchmeier et al., 2013). The lack of FDA-approved drugs represents a severe public health problem, even if these viral HF diseases are endemic in certain countries of West Africa and South America, due to increased travel to and from these endemic regions (Isaacson, 2001).
All arenaviruses encode five mature products: a nucleoprotein (NP), an RNA-dependent RNA polymerase (RdRp or L), the matrix protein (Z), and two envelope glycoproteins (GP1 and GP2) (Buchmeier et al., 2013). Of these viral proteins, the Z protein is composed of three domains, the N-terminal domain, central-domain, and C-terminal domain (Urata and Yasuda, 2012). The glycine at position 2 in the N-terminal domain is conserved among the arenavirus Z proteins and is responsible for the myristoylation required to anchor the host membrane at the budding stage (Perez et al., 2004;Strecker et al., 2006;Urata and Yasuda, 2015). The N-terminal domain is important for antagonizing retinoic acid-inducible gene-I (RIG-I) like receptor-dependent interferon (IFN) production (Xing et al., 2015). The central domain is a really interesting new gene (RING) finger motif, which is a zinc-finger type structural domain and is responsible for assembly (Hastie et al., 2016;Kentsis et al., 2001). This central RING finger domain is essential for the inhibition of viral RNA synthesis (Cornu and de la Torre, 2002;Kranzusch and Whelan, 2011) and the interaction with several host factors (Kentsis et al., 2001). Through a variety of interactions with host molecules, including the promyelocytic leukemia protein (PML) and the eukaryotic translation initiation factor 4E (eIF4E), Z protein regulates host and virus RNA synthesis, as well as protein translation (Kentsis et al., 2001). Importantly, the RING finger domain is highly conserved across all arenavirus species, and contributes to both Z self-assembly and recruitment of cellular proteins (Urata and Yasuda, 2012). The C-terminal of Z protein possesses a late (L)-domain, which is a short amino acid motif critical for budding via interaction with specific host factors (Urata et al., 2006;Urata and Yasuda, 2012;Urata et al., 2009). When arenavirus Z protein is transiently expressed in mammalian cells, Z protein is spontaneously transported to the cell membrane, assembled, attached to the host membrane, and facilitates the formation of virus-like particles (Perez et al., 2003; protein, with many vital functions in viral replication, was envisaged as a promising target for anti-arenavirus infection. Over the past decades, several trials have been reported to acquire novel anti-arenavirus drugs targeting Z protein functions (Garcia et al., 2006;Garcia et al., 2010;Lu et al., 2014). To the best of our knowledge,  (Urata et al., 2006) and Marburg virus VP40 (Urata et al., 2007). An anti-FLAG mouse monoclonal antibody was purchased from Sigma-Aldrich. The mouse immunoglobulin (Ig) G1 isotype control was obtained from BioLegend (San Diego, CA, USA). Anti-green fluorescent protein, anti-β actin rabbit polyclonal antibodies, and the horseradish peroxidase (HRP)-conjugated anti-rabbit IgG secondary antibody were purchased from Cell Signaling Technology (Danvers, MA, USA). Anti-LCMV NP rat-IgG antibody (clone VL-4, BioXCell, West Lebanon, NH, USA) and Alexa Fluor 488-labeled anti rat-IgG antibody (Abcam, Cambridge, MA, USA) were used for LCMV titration.

Expression vector for the FRET biosensor
Expression vectors for Venus, a variant of a yellow-emitting mutant of green fluorescent protein, and SECFP, a super-enhanced cyan-emitting fluorescent protein, were kindly provided by A. Miyawaki (RIKEN). Zabton was generated using the pCAGGS eukaryotic expression vector; it encoded a chimeric protein that consisted of Venus, the 5× GS linker (Gly-Gly-Gly-Gly-Ser), multiple LCMV-Z proteins, the 5× GS linker, and SECFP from the amino terminus (see Fig. 1B Lipofectamine 3000 transfection reagents (Gibco, Carlsbad, CA, USA).

Fluorescence measurement
The fluorescence of Zabton probes expressing cell lysates were measured using a

Screening of the Pathogen Box compounds
The 400

FRET imaging
HeLa cells expressing Zabton were cultured in phenol red-free DMEM (Invitrogen) buffered with 15 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) (HEPES, pH 7.4 to avoid CO2 control) and plated on a collagen-coated glass base plate (Asahi Techno Glass, Tokyo, Japan). Cell image acquisition was performed, as previously described (Mizutani et al., 2010). Briefly, the cells were imaged with an IX81 inverted microscope (Olympus, Tokyo, Japan) equipped with an automated XY-stage (Chuo Redmond, WA, USA), and used to calculate the emission ratio.

Immunoblotting
The cell lysates were diluted in 6× Laemmli sample buffer (Nacalai Tesque Inc.), incubated at 95 °C for 5 min, and then, the samples were subjected to sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE). The separated proteins were transferred to a polyvinylidene difluoride membrane (Millipore, Darmstadt, Germany) and subjected to immunoblot analysis using antibodies indicated in the figure legends.

Viral titration
An immunofocus assay was performed to determine LCMV titers. Vero 76 cells (

Evaluation of the selected compounds on LCMV propagation
The Armstrong strain of recombinant LCMV (rLCMV) was rescued using the reverse genetics method and propagated in Vero 76 cells (Flatz et al., 2006). HEK293T cells were infected with rLCMV at 0.5 multiplicity of infection (MOI). At 1.5 hours post-infection, the viral inoculum was removed and fresh media containing different concentrations of compounds, or vehicle, were added. At 24 hours post-infection, culture supernatant was collected to measure viral titration, as described above.

Cytotoxicity assay
The effect of hit compounds on cell viability was determined using the CellTiter-Glo

Quantification and Statistical Analysis
Excel and GraphPad Prism 5 (GraphPad Software, Inc., San Diego, CA. USA) software were used for all statistical analyses. Quantitative data are presented as the mean ±S.D.
from at least three independent experiments (unless otherwise indicated). For all calculations, p < 0.05 was considered significant and was represented using an asterisk.
Group comparisons were performed using one-way analysis of variance (ANOVA), followed by Dunnett's multiple comparisons test. Welch's t-test was used as a comparison between two groups.

Development of a FRET-based biosensor to monitor Z protein assembly
Previous reports have demonstrated that LCMV Z proteins are self-assembled in mammalian cells after transfection with LCMV-Z protein-encoding plasmid DNA alone (Kentsis et al., 2002;Urata et al., 2006). Accordingly, we designed a protein in which several Z proteins were sandwiched between Venus and SECFP to develop a new FRET-based biosensor for monitoring Z protein assembly. This biosensor was expected to increase the FRET efficiency upon Z-Z intermolecular assembly, and was named Zabton: Z assembly based on FRET for screening ( Fig. 1A and B). To confirm that the increased emission ratio was caused by FRET, a Zabton series was digested with proteinase K. Since green fluorescence protein mutants, Venus and SECFP, are exceptionally resistant to proteinase K, this treatment will generate free forms of Venus and SECFP (Miyawaki and Tsien, 2000). After treatment with proteinase K, the fluorescence intensity at 530 nm excited at 475 nm remarkably decreased in all Zabton series, demonstrating that these signals were caused by FRET (Fig. 1C). Among the four biosensors constructed using multiple Z proteins, Zabton-3 (trimeric tandem Z protein) exhibited the highest FRET/CFP emission ratio ( Fig. 1C and D). We also noticed that some numeric Zabtons exhibited higher S.D. than Zabton-3, which was most likely due to the aggregation. Totally, Zabton-3 was superior to the other Zabton series in S/N ratio and statistical variability. As shown previously, RING finger domains of the LCMV-Z protein is required for the assembly of the protein (Kentsis et al., 2002). Indeed, destabilization of the zinc-binding sites in the RING finger domain of LCMV-Z protein influences the stability and folding of Z protein, resulting to decrease in the capacity of self-assembly (Kentsis et al., 2001;Kentsis et al., 2002). Thus, we next examined if the emission ratio of Zabton-3 depended on the RING finger domain. First, we treated Zabton-3 with the metal ion chelators, EDTA or TPEN, to disrupt the ion coordination in the RING finger domain. We found that the emission ratio of Zabton was modestly reduced in the presence of either chelator, in a dose-dependent manner (Fig. 1E). A partial inhibition of the RING finger domain did not significantly affect the Z assembly, as previously reported (Wang et al., 2012). To further examine the effect of the RING finger domains on FRET, we created a mutant of Zabton-3 harboring an alanine substitution at the central Cys3HisCys4 motif in the RING finger domain, referred to as Zabton-3 C7H1/A8 . The emission ratio of Zabton-3 C7H1/A8 was significantly reduced compared to that of the wild-type (WT) Zabton-3 (Fig. 1F). Immunoblotting experiments revealed that the expression levels of Zabton-3 and Zabton-3 C7H1/A8 were comparable in the experimental conditions (Fig. 1G). These observations highlighted that the assembly of Z protein via the RING finger domain was critically important for the FRET observed with Zabton-3.

Screening of compounds using the Zabton-based FRET assay
To examine if Zabton-3 was suitable for identifying compounds that inhibit Z protein assembly, we calculated the Z' factor between Zabton-3 and Zabton-3 C7H1/A8 , which was 0.89, with a signal-to-background of 3.21, and a signal-to-noise of 246.1 in a 96-well format (Fig. 1E, and Materials and Methods). Accordingly, we determined that Zabton-3 (hereafter, referred to as Zabton) was a valuable tool to monitor LCMV Z protein assembly via RING finger domains in a high-throughput screening assay. Using the Zabton-based FRET assay, we screened the Pathogen Box library of 400 small molecules. The library contained drug-like molecules active against neglected diseases and structurally diverse compounds, including a set of currently marketed drugs (for details, see Materials and Methods). Compounds were screened at a final concentration of 10 μM in a volume of 100 μL Zabton-expressing cell lysate per well ( Fig. 2A). Hits (#1-#4) were defined as compounds that decreased the emission ratio of Zabton by greater than three times the standard deviation (SD) for compound responses in each plate, with reproducibility ( Fig. 2B). Although four hits were identified from the first screen, we noticed that two compounds (#2 and #3) significantly decreased the FRET ratio of the assembly deficient Zabton C7H1/A8 . Notably, the inhibitory efficacies on Zabton C7H1/A8 with compound #3 (51.9 ± 0.2%) were comparable to that of WT-Zabton (53.1 ± 0.4%) (Fig. 2B). Additionally, we noticed that compound #2 had significant auto-fluorescence at 414 nm, and that #3 exhibited a strong emission at 480 nm with 405 nm excitation (Supplementary Tables 1 and 2). Unlike these false-positive compounds (#2 and #3), compounds #1 and #4 did not exhibit obvious auto-fluorescence detected in the CFP and FRET channels (Supplementary Tables 1 and   2). With regard to the remaining two compounds, #1 and #4, we found both of them decreased the FRET ratio in a concentration-dependent manner and compound #1 exhibited higher inhibition magnitudes than compound #4 in terms of the FRET/CFP emission ratio. Interestingly, these hit compounds are known to have different antibacterial activities as follows, MMV688547 (#1) has demonstrated activity against Trypanosoma brucei (Duffy et al., 2017) and MMV688345 (#4) has an inhibitory effect on Plasmodium falciparum (Rosowsky et al., 1973;Valenciano et al., 2019). Based on the results of FRET inhibition using cell lysates, we next monitored the effect of supplementation with each hit compound on live cells. As shown in Fig. 2D, live imaging with time-lapse dual-emission fluorescence microscopy highlighted that the extracellular addition of the compound continuously decreased the FRET/CFP ratio of Zabton in HeLa cells. Therefore, Zabton seemed to be a valuable tool for alternative screening of anti-arenavirus drugs through live-cell imaging.

Hit compounds inhibited LCMV replication.
Next, we evaluated the effect of the hit compounds from the FRET-based screening assay on LCMV propagation. While the treatment of compound #1 on LCMV-infected cells did not show any inhibitory effect of virus propagation, the production of infectious viruses from compound #4 treated LCMV-infected cells (both 20 and 50 μM) was reduced compared to that of DMSO-treated LCMV-infected cells (Fig. 3A). Since the reduction of LCMV production by the treatment of compound #4 was statistically significant, we also verified the compound's cell toxicity. As shown in Fig 3B, the treatment of compound #4 exhibited a slight decrease in cell viability without statistical significance. Thus, we concluded that #4 had an inhibitory effect on LCMV propagation.

Discussion
Novel screening of 10 μM compounds using a FRET biosensor named Zabton identified two hits after a cut-off of three SDs from the average of the negative control and the exclusion of compounds through a validation assay. Out of the two selected compounds, one compound #4, (5,6,7,8-tetrahydro-7-(benzyl)-pyrido[4',3':4,5]thieno [2,3-d] pyrimidin-2,4-diamine) exhibited inhibitory activity against LCMV propagation at 20 μM. In the present study, we thus demonstrated that the screening method using the novel FRET sensor Zabton was a promising platform for the identification of effective drugs against arenavirus replication and propagation.
Since the arenaviral Z protein is not only a structural component of the virus, but also interacts with cellular proteins to disorder the host immune system, the Z protein is an attractive drug target for arenavirus infectious diseases (Urata and Yasuda, 2012). Of note, several active compounds have demonstrated the involvement of the reactive compound with the RING finger domain, which is a highly conserved domain among arenaviruses (Garcia et al., 2006;Garcia et al., 2010). Although some compounds exhibited evidence for a remarkable effect on virus replication, the mechanism of action in those compounds frequently affected the zinc-coordination of host cellular proteins, which were essential for host homeostasis. Our present study suggested that compound #4 was most likely not affecting zinc coordination with the Z protein, because the emission ratio from other Zabton types, in which FRET was most likely dependent on zinc-binding, was not decreased in the presence of compound #4 (data not shown).
Compound #4, thereby, had a unique mechanism of action that might disrupt the precise structure of the Z protein-trimer, which would have much less side effects on the host than previous zinc-finger targeting prodrugs. Arenavirus Z protein has been shown to interact with several host protein partners, including eIF4E and PML (Kentsis et al., 2001). Through the interaction with eIF4E, Z protein selectively represses protein expression involved in host innate immunity, such as IFN regulatory factor 7, which is a master regulator of type-I IFNs (Colina et al., 2008;Honda et al., 2005). As the assembled formation of the Z protein was critical for the interaction with these host proteins, the trimeric conformation might form a favorable motif to associate with the individual host protein. Identification of the partners that specifically interact with trimeric Z protein may reveal the mechanism of action of compound #4, which significantly inhibited LCMV production.
Zabton-based screening identified two-hit compounds (#1 and #4), but the virus propagation was only inhibited by compound #4, not by #1. Although further experiment will be required to clarify the different outcome, there are several possibilities which might explain this controversial result. Arenaviral Z protein has been known to interact with the other viral components, including NP, L, and GPC, during the virus infection (Capul et al., 2007;Loureiro et al., 2011;Ortiz-Riano et al., 2011).
Such complicated structures may establish the steric barriers that block the compound's recruitment into the inhibitory site of Z. Another possibility was raised from the spatial regulation of the Z assembly during the virus infection. Junin virus, one of the arenavirus species, replicates in the characteristic membrane structures called the Replication-Transcription complex (RTC) (Baird et al., 2012). Since the RTC detected in puncta in the cytoplasm, the compound should correctly reach to where Z assemble.
The elegant crystal structure of the Lassa virus Z protein provides evidence that the dodecameric form is built based on dimer blocks (Hastie et al., 2016). Hence, disruption of the Z protein dimer structure would be a more convincing drug target with profound anti-assembly activity, resulting in the suppression of virus production. Indeed, we developed several Zabton series other than Zabton-3. Zabton -2, -4, and -6 all increased the emission ratio, and these are most likely dependent on the dimerized self-assembly property of Z protein. Interestingly, we found that the emission ratio of these Zabtons was not influenced by compound #4 treatment (data not shown). These "even-numeric" Zabton probes are expected to facilitate the identification of other useful compounds 28 with an inhibitory effect on Z protein dimer formation. The detailed binding mode between the compound #4 and the Z-oligomer is not fully explored and should be revealed in the future studies.
As the present study was performed using a preliminary small-scale screen, we found only one hit compound with a pyrimidine ring ( Table 1) that is found in a variety of natural products, including nucleotides, vitamin B1 and drugs (Li, 2013). Many pyrimidine-related compounds exhibit striking biological activities, including anti-malarial, anti-cancer, and anti-HIV activities (Ajmal et al., 2014;Li, 2013