Evaluation of broad-spectrum antiviral compounds against chikungunya infection using a phenotypic screening strategy

Chikungunya fever is an emerging disease and a significant public health problem in tropical countries. Recently reported outbreaks in Brazil in 2015 drew attention to the need to develop prevention and treatment options, as no antiviral chemotherapy or vaccines are currently available for this disease. Two strategies have been proved to accelerate the discovery of new anti-infectives: phenotypic screening and drug repurposing. Phenotypic screening can support the fast interrogation of compounds without the need for a pre-validated drug target, which is not available for the chikungunya virus (CHIKV) and has the additional advantage of facilitating the discovery of antiviral with novel mechanism of action. Drug repurposing can save time and resources in drug development by enabling secondary uses for drugs that are already approved for human treatment, thus precluding the need for several of the mandatory preclinical and clinical studies necessary for drug approval. A phenotypic screening assay was developed by infecting the human hepatoma Huh-7 cells with CHIKV 181/25 and quantifying infection through indirect immunofluorescence. The compound 6-azauridine was used as a positive control drug. The screening assay was validated by testing a commercial library of 1,280 compounds, including FDA-approved drugs, and used to screen a panel of broad-spectrum antiviral compounds for anti-CHIKV activity. A high content assay was set up in Huh-7 cells-infected with CHIKV. The maximum rate of infection peaked at 48 hours post-infection, after which the host cell number was greatly reduced due to a strong cytopathic effect. Assay robustness was confirmed with Z’-factor values >0.8 and high correlation coefficient between independent runs, demonstrating that the assay is reliable, consistent and reproducible. Among tested compounds, sofosbuvir, an anti-hepatitis C virus drug, exhibited good selectivity against CHIKV with an EC of 11 μM, suggesting it is a promising candidate for repurposing.


Introduction
Chikungunya virus (CHIKV) is an arthropod-borne virus that belongs to the Alphavirus genus of the Togaviridae family. Alphaviruses are positive-sense, single-stranded RNA viruses that can produce severe encephalitis, such as in the infections caused by Ross River virus (RRV), Western-(WEE), Eastern-(EEE) and Venezuelan-equine encephalitis (VEE) virus. Alphaviruses can also be arthritogenic, such as in the case of CHIKV, Mayaro virus (MAYV), and O'nyong'nyong virus (ONNV) 1 . CHIKV was responsible for several recent (re)emerging outbreaks in humans 2-4 . Nowadays, approximately one billion people around the globe, especially in the tropics, are estimated to live in risk areas of CHIKV outbreaks 4,5 . In the Americas, CHIKV was first detected in 2013, in St. Martin, an island in the Caribbean, and quickly spread to other countries, including Brazil 6 . CHIKV produces an acute disease with high fever, headache, nausea, vomiting and conjunctivitis. Patients also develop severe joint pain, which eventually evolves into an arthritogenic syndrome that can last from weeks to years 7 . Recently, CHIKV infection has also been associated with neurological complications 8 . There are no antiviral drugs or vaccines available for CHIKV, and the supportive care treatment aims at reducing symptoms and include analgesics, anti-inflammatory and antipyretic drugs.
Some anti-CHIKV molecules have been discovered as a result of antiviral screening campaigns, such as a harringtonine, a plant alkaloid that reduced CHIKV replication by interfering with protein translation in vitro 9 ; D-N4-hydroxycytidine (NHC), a nucleoside analogue, that inhibits RNA synthesis by targeting replication complex 10 ; and barberine, abamectin and ivermectin, which all also reduce viral RNA synthesis 11 . Most assays were based on replicon systems, a classic way to evaluate drugs that interfere with the viral replication phase, but which cannot account for drugs that might inhibit other steps of the viral cycle, such as cell entry or virion assembly and release. Thus, alternative assays that deploy infectious viral particles, such as those that are based on measurement of cellular infection by high-content screening (HCS) 12,13 , enable the investigation of compounds that may interact with different stages of infection and lead to the to discover of new classes of antivirals.

Virus
The CHIKV 181/25 vaccine strain was used in the present study 14 . The strain was obtained from the World Reference Center for Emerging Viruses and Arboviruses (WRCEVA) of the University of Texas Medical Branch. The viral stock was propagated in Vero cells. Supernatant of infected tissue cultures was harvested and titrated by plaque assay 15 . Briefly, Vero cells were seeded in 24-well plate and incubated at 37°C, 5% CO 2 for 24 hours. Virus suspension was diluted 10-fold in DEMEN high glucose, and 0.2 µl from each virus dilution was added to infect Vero cells. After 1 hour in 37°C, 5% CO 2, inoculum was removed and cells were washed twice with Dulbecco's Phosphate-buffer saline (DPBS pH 7.4, Sigma-Aldrich). An overlay was added with High-glucose DMEM (GIBCO), 10% FBS (Life Technologies) and 3.5% carboxymethylcellulose (CMC, Sigma-Aldrich) prepared in distillated water. At 3 days after, the overlay was removed. Then, cells were fixed with 4% paraformaldehyde (PFA) diluted in DPBS and stained with 0.5% crystal violet to enable plaque visualization and counting. Virus titers were expressed as plaque forming units (PFU) per milliliter.

Production of mouse hyperimmune sera (MHS)
Mouse hyperimmune sera was obtained from previously prepared stocks 16 . Briefly, to prepare these stocks mice (Mus musculus) received 4 weekly inoculations of 0.2 ml of brain macerate suspensions from newborn mice infected with CHIKV in PBS, by the intraperitoneal route. At 5 days after the last immunization the animals were anesthetized and underwent intracardiac puncture for blood collection. CHIKV-MHS was obtained from this blood.

Assay development
Huh-7 cells were seeded in black polystyrene 384-well assay plates (Greiner Bio-One) at 3,000 cells/well in 40 µl DMEM-F12 supplemented with 10% FBS and incubated overnight. Cells were infected with 10 µl of inoculum of CHIKV 181/25 at different multiplicities of infection (MOIs) of 0.5, 0.05 and 0.01. Plates were fixed at different periods of time (36, 48 and 72 hours) and submitted to the immunofluorescence assay (described below) and images are acquired using an InCell Analyzer 2200 (GE Life Sciences).
Primary screening and assay validation A library stock plate containing the aforementioned compounds at 2 mM in DMSO was used to prepare the intermediate plate by a 16.6-fold dilution in DPBS, to a concentration of 60 µM and 3% DMSO. Then, 10 µl of the intermediate plate content was transferred onto the cell-containing plate. The final concentration of library compounds in the assay plate was 10 µM, with 0.5% DMSO. Controls were placed in lateral columns in all plates. Positive controls were infected cells treated with 50 µM of 6-azauridine as well as non-infected cells treated with vehicle (0.5% DMSO in DPBS). Negative controls were infected cells treated with vehicle. Cells were infected by 10 µl of CHIKV 181/25 at MOI 0.05. Plates were incubated for 48 h at 37°C, 5% CO 2 under humidified atmosphere, and then fixed with 4% (w/v) PFA for 15 min at room temperature and washed twice with DPBS. Then, plates were incubated with CHIKV-MHS (mouse hyperimmune sera) diluted 1:1500 (v/v) prepared in blocking buffer (DPBS containing 5% FBS) for 30 min. Each plate was washed twice with DPBS, followed by incubation at room temperature for 30 min with the AlexaFluor488-conjugated goat anti-mouse IgG (Cat No. A-11001, Thermo-Scientific) diluted 1:2000 (v/v), and 5 µg/ml of 4',6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich) in DPBS. Each plate was washed twice with DPBS. All plates were filled up with 50 µl of PBS/well. Images were acquired using a confocal microscope High-Content System InCell Analyzer 2200 (GE Life Sciences) and processed by InCell Investigator v.1.6.1 software (GE, USA). Four different images were acquired from each well at x20 magnification. Automated image analysis was performed cellby-cell through a defined mask based on fluorescence signal measurement and cell morphology. Cell segmentation parameters defined the analysis performed by the Investigator software. The nuclei were segmented from the DAPI staining images and each nucleus was determined as a minimum area of 50 µM. Total cells were filtered from the AlexaFlour488 channel and were defined as minimum area of 100 µM. The mean AlexaFluor488 florescence signal of infected cells were defined from the cytoplasm mask, with values based on means signal from fluorescence of wells from infected cells showing a six-times higher value compared to the mean of Alexa488 fluorescence signal of non-infected cells. Images were treated using image analysis program ImageJ v1.51 to set up colors and merge image channels to final visualization. The validation of screening was conducted in two independent experiment.

Data normalization
Infection ratio (IR) was defined as the ratio between (i) the total number of infected cells, and (ii) the total number of cells. Data were normalized with the negative (DMSO-treated, infected cells) and positive (infected cells treated with 50 µM 6-azauridine) controls. Normalized activity was calculated as described by Pascoalino et al. 12 . Cell survival was expressed as the percentage of the total cell number from test sample divided by the average total cell number from the positive control wells: Cell number test sample/Avg. cell number of positive control) × 100. Normalized activity and cell survival values were processed with the GraphPad Prism software version 7. Representative graphs of sample distribution were obtained by plot data at TIBCO Spotfire 7.0 software. Plates were also submitted to quality control measurement of Z'-factor as described by Zhang et al. 17 .

Dose-response assay
For dose-response curves, drugs were prepared as described 12 . The initial test concentrations were 100 nM for IFN-α2A, 50 nM for bafilomycin, 120 µM for chloroquine and mycophenolic acid, and 100 µM for sofosbuvir, daclatasvir, ledispavir and 5-fluorouracil. Dose response assay were calculated based on percentage of normalized activity and cells survival for each concentration tested. Data were plotted with GraphPad Prism software version 7. The sigmodal dose-response curve (variable slope) function were used to calculate the effective concentration that inhibited 50% of infection (EC 50 ), and the concentration of compound that presented a 50% reduction in cell number in comparison to the controls (CC 50 ). The ratio between CC 50 and EC 50 determines the selective index (SI).

Statistical analysis
Two-way analysis of variance (ANOVA) with Sidak's test, a multiple comparison test, was conducted to calculate statistical significance (P < 0.05) of cell numbers from non-infected and infected CHIKV 181/25 at different multiplicity of infection and incubation time experiment. The coefficient of determination (R 2 ) test were using to determinate statically coefficient of variation between screening replicates from normalize activity data. All data were plotted using GraphPad Prism Software version 7.

Assay development
A high-content screening assay was developed to evaluate compounds activity against CHIKV infection in vitro. The first step consisted on defining the cell model to support viral infection. A range of cell lines has been reported as being susceptible to CHIKV infection, such as Vero, human fetal lung fibroblast (MRC-5), baby hamster kidney (BHK), human embryonic kidney 293 (HEK-239T) and Huh-7 18,19 . The Huh-7 cell line was selected as it has desirable features for high content imaging, such as adherent monolayer growth, and is human cell line, meaning it is a more representative in vitro model than would be cells of another species. The second step was determining the optimal multiplicity of infection (MOI) and the necessary period of time for the efficient viral infection in 384-well plates. Cells were infected at three different MOI (0.5, 0.05 and 0.01) and incubated for different periods of time (36, 48 and 72 hours). The total cell number and the IR were determined. When cells were plated and infected concomitantly, even the lowest MOI tested showed high cytopathic effect (data not shown). Thus, cells were plated 24 h before infection ( Figure 1A). For the highest MOI, CHIKV infection decreased cell number by 70% at 48 hours and by almost 100% at 72 hours compared with non-infected cells. There was no significant difference in cell number between non-infected cells and infected cells for both 0.05 and 0.01 MOIs at 36 hours. Compared to non-infected cells, a decrease in cell number by 42% and 22% at MOIs 0.05 and 0.01, respectively, was observed at 48 hours ( Figure 1A). After 36 hours of incubation, the IR showed an association with the MOI: 0.95±0.04 (MOI 0.5), 0.40±0.19 (MOI 0.05) and 0.12±0.20 (MOI 0.01) ( Figure 1A), demonstrating that the assay endpoint was within the dynamic range of the infection. For 48 hours of incubation, the IR reached 0.99 for all MOIs, but the lowest MOI gave high variation in IR between replicate wells. Therefore, with the aim of testing drugs, a 0.05 MOI at 48 hours of incubation was selected for further experiments to achieve the longer time of exposure to drug treatment possible under these conditions, a high ratio of infection with minor variability, and a cell survival rate greater than 50% (compared with noninfected) ( Figure 1A). Figure 1B describes the established general scheme of CHIKV high-content assay.

Assay validation
To validate the assay, high-content screening was run using a commercial library of compounds. Cell infection was determined by indirect CHIKV immunofluorescence detection. Figure 2A shows a raw image and software segmentation analysis of the same image. The 6-azauridine compound was previously reported to have activity against CHIKV 20 , and was chosen as the reference compound in this assay. The activity of 6-azauridine was assessed using a dose-response curve ( Figure 2B). The EC 50 of 0.65 µM 6-azauridine and EC 100 of 50 µM 6-azauridine were determined against CHIKV. In order to validate the assay reproducibility and robustness, a commercial library composed of 1,280 compounds was tested at a single concentration (10 µM). A good window between positive and negative controls was observed ( Figure 2C). As a result, the mean for all plates Z'-factor values were 0.86±0.09, indicating that the established assay is reliable. Additionally, there was a high correlation coefficient between runs (Coefficient of determination R 2 : 0.86), which was determined using normalized activity of each single well between the first (R1) and the second (R2) screens, including compounds and controls ( Figure 2D).

Evaluation of known antivirals against CHIKV
A set of compounds with known antiviral activity were evaluated against CHIKV. A total of 9 compounds were tested in dose-response curves. Figure 3 lists the name, molecular structures and dose-response curve plots for all compounds. Interferon α2A (IFN-α2A) and mycophenolic acid have reported activity against CHIKV 19,20 . The HCS assay confirmed their reported activity, demonstrated by EC 50 values of 0.7 nM and 0.8 µM and high SI of >14 and 8.25, for IFN-α2A and mycophenolic acid, respectively. In order to evaluate compounds with previously reported activity against CHIKV, bafilomycin A1 21 and chloroquine 22 were tested, giving an EC 50 of 0.01 µM and 21 µM, respectively; however, they were cytotoxic in Huh-7 cells, with low SI values (5 and 0.3, respectively). The antiviral activity of daclatasvir, an anti-hepatitis C virus (HCV) drug 23 , against CHIKV was associated with high cytotoxicity, and it had a low SI value of 1.3. Ledispavir, also an anti-HCV compound 24 , and 5-fluorouracil, which has reported activity against ZIKV 12 , did not present anti-CHIKV activity in the HCS assay. Sofosbuvir is an FDA-approved compound against HCV, and has been recently described as an active compound against other flaviviruses, including dengue and Zika 25,26 . Sofosbuvir demonstrated dose-dependent activity against CHIKV (EC 50 , 11 µM) with no cytotoxicity in Huh-7 cells. Representative images of sofosbuvir activity, alongside 6-azauridine, are displayed in Figure 4.  . Z'-factor value of 0.87 was obtained from total data controls from two independent runs for screening validation, between inter-replicates and intra-replicates plates. The continuous line represents the mean of each control, the dotted line represents 3 standard deviations from the mean of the negative and positive controls. Right graphic: Data correlation (normalized activity) from two screening runs of a small library; dots represent each single tested well and colors represent different treatments, where: 0.5% DMSO negative control (yellow); non-infected cells (red); 50 µM 6-azauridine positive control (green); and tested compounds (gray). The coefficient of determination of R square (R 2 0.86) was calculated using GraphPad Prism software.

Discussion
Currently, most assays available for drug screening against CHIKV are based on cell viability methodologies, which evaluate the compounds capacity to prevent cell lysis 27-30 . Such approaches have the advantage of being of lower in cost and higher-throughput than image-based phenotypic assays. However, background noise interference in quality and usage of counterscreening assays to assess compound cytotoxicity and support conclusions should be considered. Conversely, HCS assays provide multi-parametric evaluation of both viral infection and cytotoxicity in same assay 28,31 . Therefore, in the present study we propose the development of a reproducible, phenotypic HCS for CHIKV, in order to trial drugs with antiviral activity. Different approaches have been described to assess drugs in a high-throughput screening (HTS) format against CHIKV, including measurement of cell viability using a resazurin assay 27 , replicon-based assay using a Renilla luciferase reporter 10,11,32,33 , which targets only replication-process-interfering compounds, or a HCS assay using BHK cells 9 . In this study, we opted for the Huh-7 cell line as this has been used for HCS for antiviral discovery by our group and others for hepatitis C 34 , dengue 13 and Zika 12,35 . Moreover, it is reported that Huh-7 cells are permissive to CHIKV infection. The CHIKV viral cycle usually happens in a short period of time, between 8 and 16 hours, following high cytopathic effect 36 . In this manner, we opted to use a relatively low MOI (0.05) to prevent high cell lysis ( Figure 1A), and it can be expected that multiple infection cycles happen during the assay duration (48 h). Thus, all potential targets during the whole viral cycle can be exposed to the compounds. The developed assay also proved to be robust and reproducible.
IFN-2αA, bafilomicyn A1, chloroquine and mycophenolic acid had all been previously reported as active against CHIKV in vitro 21,22,37 , and their antiviral activity was confirmed under the conditions used in this study. However, bafilomycin A1 and chloroquine were cytotoxic, resulting in low SIs (<5). Bafilomicyin A1, an inhibitor of mammalian vacuolar-type H(+)-ATPase, prevents the acidification of the endosomal compartment, where the low pH allows the fusion of viral capsid followed by the entrance in the cytoplasm, thus preventing a crucial early step of the CHIKV virus cycle 38 . The same inhibition mechanism was observed in vitro during for the infection of sindbis virus, a prototype alphavirus 39 . Previous studies have reported the cytotoxicity of bafilomicyin A1 in HEK123 cells 21 , as was also observed here in Huh-7 (Figure 3). The activity of chloroquine, an antimalarial compound, was extensively investigated against CHIKV, although studies in vivo with infected mice showed inefficient activity 22,40 . In addition, clinical trials comparing doubleblinded placebo groups and patients with CHIKV infection group did not present convincing data regarding chloroquine treatment efficacy 41 . Studies in CHIKV-infected Vero cells suggested that chloroquine exerts antiviral activity by preventing CHIKV internalization. The chloroquine EC 50 values observed in this study (21 µM) in Huh-7-infected cells are in accordance with values previously reported for Vero cells (17 µM) 22 . However, the cytotoxicity of chloroquine seems to vary depending on the cell type or assay conditions, as chloroquine showed greater cytotoxicity in Huh-7 cells (with values of CC 50 of 56 µM) than Vero cells (CC 50 >100 µM) 22 . Mycophenolic acid inhibits inosine monophosphate dehydrogenase (IMPDH), an essential enzyme in de novo biosynthesis of guanine, and has been reported to have antiviral activity for both single strand RNA negative and positive viruses, for instance, against influenza virus 42 , Nevertheless, a human cell model, such as the Huh-7 cell line, should be preferred to screen antiviral candidates to promote a more representative values of activity and cytotoxicity, which may diverge when compared to non-human cells lines 45 .
Sofosbuvir, 5-fluorouracil, daclatasvir and ledispavir are all FDA-approved drugs with reported activity against flaviviruses. The nucleoside analog 5-flurouracil is used to treat neoplastic disease 46 , and we have recently shown its antiviral activity in vitro against ZIKV infection 12 . However, 5-flurouracil presented no activity against CHIKV in our assay, suggesting selectivity for flaviviruses. Comparable results were obtained for ledispavir, which did not show inhibition against CHIKV, even at the highest concentration tested. Ledispavir, daclatasvir and sofosbuvir are direct-acting antiviral agents and have been successfully used to treat HCV-infected subjects. Those compounds target NS5A and NS5B, two HCV non-structural proteins 47 . NS5A presents three domains, which are responsible for genome replication, virus assembly through production of infection virus particles, and regulation of viral genome replication, from direct interaction of NS5A domain II with NS5B. NS5B is an RNA-dependent RNA-polymerase (RdRp) that directs the RNA synthesis in the HCV replication cycle 48 . Ledispavir and daclatasvir are NS5A inhibitors, while sofosbuvir, an uridine nucleoside analog that targets NS5B that is usually administrated in combination with daclatasvir 49 . Besides HCV, recent studies have demonstrated that sofosbuvir can inhibit infection by other flavivirus in vitro and in mice 25,26 . Our results demonstrated sofosbuvir elicited a concentration-dependent inhibition of CHIKV infection (Figure 3 and Figure 4), suggesting that this drug might have a broader antiviral spectrum than previously known.
Differently from HCV, which the genome organization consisted in five non-structural proteins (NS2, NS3, NS4A, NS4B, NS5A 50 ), CHIKV possess four non-structural protein (nsP1, nsP2, nsP3 and nsP4), being the RdRp domain localized at nsP4. Alignment sequence of RdRp has demonstrated highly conserved regions between CHIKV and other flaviviruses. More specifically, the motif B region, which is a functional domain of viral RdRp coding region, and the R1 motif, which has a role in nucleoside triphosphate binding during viral RNA synthesis, are highly conserved. Besides, CHIKV RdRp forms similar structures to the RdRp of other RNA viruses 51 . The search for direct-target compounds against CHIKV have focused on nsP2, due to its multifunctioning domains, which acts as helicases to form RNA secondary structures, as triphosphates responsible for RNA capping enzyme and removing terminal phosphate from new RNA template, and as proteases responsible for processing non-structural polyproteins 52 . In addition, its well-known structure makes nsP2 a suitable target for drug design 53 . However, few studies have focused on the search for compounds that target RdRp for CHIKV 54,55 . A compound that targets RdRp would be attractive, as RdRp acts on a viral process, is essential for replication of the viral genome and does not affect host cells 55 .
In conclusion, the phenotypic high content analysis established herein revealed that sofosbuvir is a promising candidate for use against CHIKV infection. Further studies should be performed in order to elucidate the exact mechanism related to CHIKV RdRp inhibition by sofosbuvir.

Data availability
Dataset 1. All raw data from the present study. Raw data are separated according to the figure in which they are presented; a guide to the data is available as a .docx file. DOI: https://doi. org/10.5256/f1000research.16498.d221905 56 .

Grant information
This work has been supported by the Sao Paulo State Research Foundation -FAPESP (Process no. 2016/03780-5).
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Mukesh Kumar
Georgia State University, Atlanta, GA, USA The manuscript entitled "Evaluation of broad-spectrum antiviral compounds against chikungunya infection using a phenotypic screening strategy" by Bonotto and colleagues evaluated an imagebased phenotypic assay for high-throughput screening of anti-CHIKV compounds. The screening assay was validated by testing a commercial library of 1,280 compounds, including FDA-approved drugs. The research topic is interesting and has potential significance. However, there are major gaps in the depth of the information reported in this manuscript that make publication of the findings in its current form problematic. Moreover, results in this manuscript are poorly presented and inadequate information is provided.
In this manuscript, the screening assay was validated by testing a commercial library of 1,280 The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias • You can publish traditional articles, null/negative results, case reports, data notes and more • The peer review process is transparent and collaborative • Your article is indexed in PubMed after passing peer review • Dedicated customer support at every stage • For pre-submission enquiries, contact research@f1000.com