Establishment of a screening platform based on human coronavirus OC43 for the identification of microbial natural products with antiviral activity

ABSTRACT

Human coronaviruses cause respiratory tract infections and are of great importance due to the recent severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic. Human betacoronavirus OC43 (HCoV-OC43) is an adequate surrogate for SARS-CoV-2 because it infects the human respiratory system, presents a comparable biology, and is transmitted in a similar way. Its use is advantageous since it only requires biosafety level 2 infrastructure that minimizes costs and biosafety-associated limitations. In this report, we describe a high-throughput screening platform to identify compounds that inhibit the propagation of HCoV-OC43. Optimization of assays, based on inhibition of the cytopathic effect and virus immunodetection with a specific antibody, has provided a robust methodology for the screening of a selection of microbial natural product extracts from the Fundación MEDINA collection. Using this approach, a subset of 1,280 extracts has been explored. Of these, upon hit confirmation and early liquid chromatography mass spectrometry (LC-MS) dereplication, 10 extracts that contain potential new compounds were identified. In addition, we report on the novel antiviral activity of some previously described natural products whose presence in bioactive extracts was confirmed by LC-MS analysis.

IMPORTANCE

The COVID-19 pandemic has revealed the lack of effective treatments against betacoronaviruses and the urgent need for new broad-spectrum antivirals. Natural products are a valuable source of bioactive compounds with pharmaceutical potential that may lead to the discovery of new antiviral agents. Specifically, compared to conventional synthetic molecules, microbial natural extracts possess a unique and vast chemical diversity and are amenable to large-scale production. The implementation of a high-throughput screening platform using the betacoronavirus OC43 in a human cell line infection model has provided proof of concept of the approach and has allowed for the rapid and efficient evaluation of 1,280 microbial extracts. The identification of several active compounds validates the potential of the platform for the search for new compounds with antiviral capacity.

INTRODUCTION

Coronaviruses (CoVs) have the ability to propagate and generate new species that cause epidemic diseases. The emergence of CoV infections has persistently caused serious public health alarms over the years. Severe acute CoV infections, including the respiratory syndrome-related coronavirus (SARS-CoV), the Middle East respiratory syndrome-related coronavirus (MERS-CoV), and the pandemic virus SARS-CoV-2, have become a rising and long-lasting global risk. The impact of the SARS-CoV-2 pandemic has revealed the urgent need for broad-spectrum antivirals to prevent future viral pandemics of unknown origin.

To date, efficient vaccines are lacking for many viruses, and novel antiviral drugs with clinical efficacy are called for. The situation is further aggravated by the potential development of mutants that are drug resistant, thus limiting drug efficacy (1 – 4).

Natural products represent a rich source of diverse, new, and inexpensive chemical starting points for drug discovery in the area of antivirals and serve as an outstanding source of biodiversity for the discovery of novel compounds. Indeed, approximately 65% of successfully approved drugs are either natural products or derivatives (5). Many natural products derived from microbes have been reported to have robust antiviral activity, with anti-hepatitis, anti-herpes simplex, anti-HIV, anti-influenza, anti-respiratory syncytial virus, and anti-SARS-CoV-2 properties (6 – 12). However, when performing the search for active compounds from whole crude microbial extracts, the process is both time consuming and capital intensive. For that reason, it is very critical to develop quick, efficient, and high-quality high-throughput primary screening assays integrated with chemistry platforms for early dereplication of known compounds to guarantee that only extracts with relevant pharmacological properties are selected for further development downstream.

Antiviral drug discovery using SARS-CoV-2 for quantitative screening requires a biosafety level (BSL) 3 facility, which is costly and labor intensive. Hence, the availability of model systems of HCoV for high-throughput screening (HTS) assays that involve BSL-2 is highly advantageous. Human coronavirus OC43 (HCoV-OC43) (order: Nidovirales, family: Coronaviridae, genus: Betacoronavirus, subgenus: Embecovirus, species: Betacoronavirus1) exhibits several common features with SARS-CoV-2 regarding structure and biology. Thus, despite some existing differences between the host receptor implicated in viral attachment and entry for both viruses (13, 14), we believe that HCoV-OC43 is a well-matched surrogate for SARS-CoV-2 (15). As members of the same genus, HCoV-OC43 and SARS-CoV-2 are closely related genetically (16), and they are transmitted by aerosols and droplets and replicate in human respiratory epithelium (17).

A wide variety of methods to evaluate HCoV-OC43 infection using different cell lines have been described (15). With regard to host cells, HRT-18 (human colon cancer cells) (18, 19), MRC-5 (human lung fibroblasts) (20), and Vero E6 expressing TMPRSS2 (21), among others, have been utilized. In addition, different detection methods (plaque assay, indirect immunostaining, and resazurin reduction) (15, 22) and infection conditions have been explored. Here, we report the standardization of methodology for the quantification of infection by HCoV-OC43 and describe the development of a robust experimental approach using an HTS platform to identify compounds that exhibit anti-OC43 activity. The experimental design has been successfully applied to analyze 1,280 microbial extracts from actinomycetes and fungi of the Fundación MEDINA’s library (Fig. S1). We have combined the use of determination of the cytopathic effect (CPE) with antibody-based detection bioimaging assays in the analysis that revealed the presence of novel antiviral compounds. To our knowledge, this is the first report of HTS using microbial natural product extracts against HCoV-OC43.

MATERIALS AND METHODS

Reagents and antibodies

Dimethyl sulfoxide (DMSO), chloroquine diphosphate salt, mycophenolate mofetil, linoleic acid, destruxin A, bovine serum albumin (BSA), and resazurin were purchased from Sigma-Aldrich. Tamoxifen, ribavirin, destruxin B, and gymnoascolide A were purchased from Tocris, Santa Cruz Biotechnology, Adooq, and Cayman, respectively. All compounds were dissolved in 100% DMSO. The mouse monoclonal anti-HCoV-OC43 antibody (MAB2012) was purchased from Millipore, and the secondary antibody AlexaFluor 488-conjugated anti-mouse and Hoechst 33342 were purchased from Thermo-Scientific.

Cell culture

The human lung fibroblast cell line MRC-5 (CCL-171, ATCC) and cell line BHK-21 (CCL-10, ATCC) were cultured in Minimum Essential Medium (MEM) (Life Technologies) supplemented with 10% fetal bovine serum (FBS) (Life Technologies), 100 units/mL penicillin, and 100 µg/mL streptomycin (Pen/Strep) (Life Technologies). Cells were incubated at 37°C in a humidified atmosphere of 5% CO₂ and were periodically analyzed and confirmed to be mycoplasma negative.

Virus production

The human betacoronavirus HCoV-OC43 (VR-1588, ATCC) was propagated in MRC-5 cells. MRC-5 cells at a 90% confluence were inoculated with HCoV-OC43 in infection media (MEM, 2% inactivated FBS, and Pen/Strep) and incubated for 2 h at 33°C, rocking the flask every 15 min for virus adsorption. After virus adsorption, infection media was added, and infected cells were incubated at 33°C for 5–7 days until more than 50% of the cells presented a CPE resulting in cell death. For virus recovery, an infected culture was subjected to three freeze-thaw cycles, centrifuged at 500 × g, 10 min at 4°C to spin down cells and cell debris. Viral particles were recovered from the supernatant and aliquoted in cryotubes. Viral stock aliquots were then rapidly frozen in a dry-ice/ethanol bath and stored at −80°C.

Batch infection with HCoV-OC43

MRC-5 and BHK-21 cells were seeded 24 h prior to infection for a 90% confluence at the time of infection. Different multiplicities of infection (MOI) were tested for HCoV-OC43 infection of MRC-5 and BHK-21 cells. The virus adsorption was performed for 2 h at 33°C, rocking the cells every 15 min, and then infected cells were incubated for 24 h at 33°C before seeding into 96- or 384-well plates.

Resazurin-based assay

Assay plates were previously prepared with microbial extracts to be screened at a final concentration of 1/100 dilution of whole broth equivalent (WBE) and at a final 0.2% DMSO concentration in 96-well plates. Infected cells were washed, trypsinized, and seeded in assay plates at a cellular concentration of 2 × 10⁴ cells/well in infection media. The plates were incubated at 37°C for 96 h in the presence of the extracts, 5 days after infection media was withdrawn, and 120 µL of infection media containing 20% resazurin (Sigma-Aldrich) was added per well. Plates were incubated for 2 h at 37°C, and fluorescence was determined at 550–590 nm in a TECAN infinite F200 plate reader. Non-infected cells and infected cells with 0.2% DMSO were used as the positive and negative controls, respectively. Extracts showing a CPE inhibition superior to 40% were selected as hits.

The resazurin assay was also used to determine the cytotoxicity of extracts or compounds in MRC-5 cells. Briefly, 2 × 10⁴ cells/well were seeded in 96-well plates containing the microbial extracts or pure compounds. After 96 h, cell culture media was withdrawn, and 120 µL of infection media containing 20% resazurin (Sigma-Aldrich) was added per well. Plates were incubated for 2 h at 37°C, and fluorescence was determined at 550–590 nm in a TECAN infinite F200 plate reader. MRC-5 cells treated with 50 µM tamoxifen were used as a negative control, while positive controls corresponded to MRC-5 cells incubated in the presence of 0.2% DMSO for 96 h.

High-content screening assay

Assay plates were previously prepared with microbial extracts to be screened at a final concentration of 1/100 dilution of WBE and at a final DMSO concentration of 0.2% in 384-well black bottom-transparent plates (Greiner Bio-One). MRC-5-infected cells were washed, trypsinized, and seeded in assay plates at a cellular concentration of 3,000 cells/well in infection media. The plates were incubated at 37°C for 72 h in the presence of the extracts. Four days post-infection, cells were subjected to indirect immunostaining by using an anti-OC43 monoclonal antibody (MAB2012). Briefly, cells were washed with 1× phosphate-buffered saline, fixed with 4% p-formaldehyde, and permeabilized with 0.4% Triton X-100. Next, cells were incubated in a 5% BSA blocking solution for 2 h and then incubated with the anti-HCoV-OC43 monoclonal antibody at a 1:2,000 dilution O/N at 4°C. Cells were washed, incubated with the antibody AlexaFluor 488-conjugated anti-mouse, and washed again; nuclei were stained with 4 µM Hoechst 33342. Digital images were captured using the Operetta CLS High Content Analysis System (PerkinElmer) with a 5× air objective. Images were analyzed using the Harmony software (PerkinElmer), and the following parameters were determined: number of nuclei per well, number of infected cells per well, and total 488-fluorescence signal per well. Based on the 488 signal and cell nuclei, cells are segmented for the calculation of the total 488-fluorescence signal.

LC-MS dereplication and database matching of known secondary metabolites

The active extracts of interest were subjected to liquid chromatography mass spectrometry (LC-MS) dereplication and database matching to identify possible known components using an Agilent (Santa Clara, CA) 1260 Infinity II single Quadrupole LC-MS system and LC-MS analytical conditions and methodology previously described (23 – 25). Dereplication refers to the analytical method of rapidly identifying already known natural products in samples from natural sources (26).

Bioassay-guided isolation of gymnoascolide A to confirm LC-MS identification

The culture from which gymnoascolide A was identified by LC-MS was subsequently re-fermented at a scale of 100 mL and extracted by addition of 100 mL of acetone and shaking (Kühner shaker) at 200 rpm, 24°C, for 2 h. The extract was centrifuged and filtered, and the acetone was evaporated under a N₂ stream to obtain the concentrated crude sample, an aliquot of which was tested and found to retain the antiviral activity. The crude extract was loaded onto an HP-20 column, and the flow through was collected and saved. The column was washed with 1 vol milliQ water, and the flow through was discarded. Next, the column was eluted with 2 vol acetone and 1 vol methanol, and the organic phases were pooled together. The combined organic extract was evaporated to dryness (under a N₂ stream) and weighed to yield 41.0 mg. This mass was dissolved in 300 µL DMSO, filtered, and fractionated by a semi-preparative reversed-phase HPLC on an Agilent Zorbax SB-C18 column (9.4 × 250 mm; 5 µm) at 3.6 mL/min flow with a linear gradient of 5–100% CH₃CN/H₂O in 45 min with UV detections at 210 and 280 nm. Out of the 80 fractions generated, fraction F53-54 showed antiviral activity, and from this fraction, 0.9 mg of pure gymnoascolide A was obtained, and its identity was confirmed by both high-resolution mass spectrometry and nuclear magnetic resonance (NMR) (¹H and HSQC) spectroscopy.

RNA isolation and real-time RT-PCR

Viral RNA from supernatants was purified using the Macherey-Nagel Nucleospin RNA Kit. RT-qPCR was performed in a single step using the One-Step TB Green PrimeScript RT-PCR Kit II (Takara Bio). The HCoV-OC43 nucleocapsid gene was amplified with the following primers: the forward primer 5′ AGCAACCAGGCTGATGTCAATACC-3′ and the reverse primer 5′ AGCAGACCTTCCTGAGCCTTCAAT-3′ (27). A standard curve was generated with purified HCoV-RNA (Vircell).

Data analysis

CPE inhibition

Extract activities were calculated automatically using the Genedata Screener software (Genedata AG, Basel, Switzerland), and the percentage of CPE inhibition of each extract was determined by equation 1 integrated in the Genedata Screener software:

$${\displaystyle CPEInhibition\left(\mathrm{\%}\right)={\displaystyle \frac{\left(Flu{o}_{well}-Flu{o}_{neg}\right)}{\left(Flu{o}_{pos}-Flu{o}_{neg}\right)}}\times 100}$$ (1)

where Fluo_well is the measured fluorescence for each well, Fluo_pos is the average fluorescence for the positive control (infected MRC-5 cells treated with 400 µM ribavirin), and Fluo_neg is the average fluorescence for the negative control (infected MRC-5 cells with 0.2% DMSO).

Cytotoxicity

Cellular cytotoxicity was determined by equation 2:

$${\displaystyle Viability\left(\mathrm{\%}\right)={\displaystyle \frac{\left(Flu{o}_{well}-Flu{o}_{neg}\right)}{\left(Flu{o}_{pos}-Flu{o}_{neg}\right)}}\times 100}$$ (2)

where Fluo_well is the measured fluorescence for each well, Fluo_neg is the average fluorescence for the negative control (MRC-5 cells treated with 50 µM tamoxifen), and Fluo_pos is the average fluorescence for the positive control (MRC-5 cells with 0.2% DMSO).

HCoV-OC43 infection

The sum intensity (SumInt) per well corresponding to the Alexa 488 fluorescence signal was automatically given by the Operetta Analysis System software. HCoV-OC43 infection percentages were determined by equation 3:

$${\displaystyle Infection\left(\mathrm{\%}\right)={\displaystyle \frac{\left(SumIn{t}_{well}-SumIn{t}_{neg}\right)}{\left(SumIn{t}_{pos}-SumIn{t}_{neg}\right)}}\times 100}$$ (3)

SumInt_well corresponds to the sum of 488-fluorescence signal in wells with extract, and SumInt_neg and SumInt_pos correspond to the 488-fluorescence signal of negative and positive control wells, respectively. Positive control corresponds to MRC-5-infected cells with 0.2% DMSO, and the negative control refers to MRC-5-infected cells treated with 400 µM ribavirin.

The 50% cytotoxic concentrations (CC₅₀) and the half-maximal effective concentration (EC₅₀) were determined using SigmaPlot 15.0 software.

Evaluation of quality control of the screening assay

Z′ factor and signal to background (S/B) are quality control parameters that reflect the robustness and reproducibility of the assay. Z′ factor is a statistic parameter used to evaluate the quality of an assay, and it is calculated by relating the sum of the standard deviations for the reference wells to the signal range given by the difference in their mean values:

$${\displaystyle Z^{\prime }=1-{\displaystyle \frac{3\times SD\left(N\right)+3\times SD\left(SR\right)}{\left(mean\left(N\right)-mean\left(SR\right)\right)}}}$$

being SD as the standard deviation, N as the neutral controls, and SR as the scale reference. A value of 0.4–1 characterizes a robust assay. S/B is the ratio of the mean value of the raw data for positive and negative controls.

RESULTS

Set up of a resazurin-based assay for the detection of antivirals that inhibit CPE

In order to optimize the protocol of batch infection with HCoV-OC43, different MOIs were tested in two different cell lines that had been previously described as host models for HCoV-OC43: MRC-5 and BHK-21 (19, 28). Both cell lines were infected with increasing HCoV-OC43 MOIs (0.01, 0.1, and 0.5), and 24 h post-infection cells were seeded in 96-well plates. CPE-induced cell death was assessed 5 days post-infection using a resazurin-based approach. In this assay, the fluorescence signal is directly proportional to cell survival and accordingly decreases when the viral CPE causes cell death. For both cell lines, we obtained a reproducible and homogenous infection in the 96-well plate format at MOIs higher than 0.1. A decrease in viability of BHK-21 and MRC-5 cells was not observed at the lower MOI of 0.01 (Fig. 1A).

Fig 1.

CPE evaluation. (A) Cell viability of MRC-5 and BHK-21 cell lines 5 days post-infection with HCoV-OC43 at different MOIs, 0.01, 0.1, and 0.5. (B) Average Z′ factor and signal to background (S/B) values for BHK-21 and MRC-5 cells infected with HCoV-OC43.

MRC-5 cells with infection at a 0.1 MOI were selected as the most suitable model, since the decrease in cell viability 5 days post-infection was higher than 90%, while in the case of BHK-21 cells with the same MOI, the loss of cell viability was less pronounced (80% decrease). Z′ factors were 0.88 for MRC-5 cells and 0.84 for BHK-21 emphasizing the robustness of the assay in both cases (Fig. 1B).

For the selected model, the progression of the CPE up to 5 days post-infection gives rise to a cell death of MRC-5-infected cells higher than 90%. This CPE will be inhibited in the presence of active antiviral components, and thus, the resazurin-based fluorescence signal obtained will increase and be compared to the signal corresponding to non-infected cells (100% viability control). Pharmacological validation of the assay was carried out using three compounds that had been previously described as antivirals against SARS-CoV-2 and HCoV-OC43: ribavirin, chloroquine, and mycophenolate mofetil (28 – 31).

The EC₅₀ values were determined for the three compounds together with the corresponding CC₅₀ values in non-infected MRC-5 control cells (Fig. 2). Chloroquine showed a low EC₅₀ of 6.22 µM, although it was also the most cytotoxic with a CC₅₀ of 14.8 µM. Ribavirin and mycophenolate mofetil showed no cytotoxicity at concentrations below 800 µM and 200 µM, respectively. As for the antiviral activity, mycophenolate mofetil exhibited a lower EC₅₀ (20.66 µM) compared to ribavirin (118.38 µM).

Fig 2.

Evaluation of reference antivirals ribavirin, mycophenolate mofetil, and chloroquine. Determination of EC₅₀ and CC₅₀ values after 96 h incubation with the drugs of infected cells (after 5 days of infection) and non-infected cells, respectively. EC₅₀ and CC₅₀ values were calculated using the SigmaPlot software.

Ribavirin was selected as the antiviral control compound to be routinely included in the assay. A concentration of 400 µM ribavirin was used, since it showed a good selectivity index, close to 8, and it has been widely described and used as a broad-spectrum antiviral (32).

Antiviral primary screening and hit identification

For the primary screening of 1,280 (320 fungal and 960 actinomycete) microbial extracts from the MEDINA microbial library, the resazurin-based assay for CPE inhibition determination was used (general workflow is shown in Fig. 3). The average Z′ factor for the assay was 0.971 ± 0.04, and the fluorescence signal of infected cells was five times lower than the positive control (Fig. 4A and B). We selected as hits those extracts that inhibited more than 40% of the CPE (Fig. 4C). Out of the 1,280 microbial extracts, 40 were selected as hits (a hit rate of 3.12%) (Table 1); 13 corresponded to actinomycete extracts, and 27 corresponded to fungal extracts.

Fig 3.

Schematic representation of the platform for the discovery of antiviral drugs from microbial extracts.

Fig 4.

Method validation and primary screening of a subset of 1,280 microbial extracts. (A) Distribution of the Z′ factor along the different plates. (B) Fluorescence distribution of the positive control (MRC-5 infected with 0.1 MOI HCoV-OC43 and treated with 400 µM ribavirin) and the negative control (MRC-5 infected with 0.1 MOI HCoV-OC43). (C) Representation of 1,280 extracts from primary screening. The green line represents the cut-off for hit selection (40% inhibition of CPE), and the circles in green correspond to the 40 selected hits presenting a CPE inhibition higher than 40%.

TABLE 1.

Number of extracts identified as hits at different stages of the screening process

Screening stages	Number of active extracts (%)
Primary screening	40 (3.12)
Cherry-picking for confirmation HCS	32 (2.50)
Dose-response curves for EC50 calculation	21 (1.64)
Prioritized for regrowth in 100 mL volume	15 (1.17)
Activity reconfirmation after 100 mL regrowth	12 (0.93)
Dose-response curves for EC50 calculation after 100 mL regrowth	10 (0.78)

Implementation of a high-content screening in 384-well format

In the CPE inhibition screening, cytotoxicity can interfere with the assay since the readout is the same for CPE-induced death and toxicity caused by the extracts. High-content screening (HCS) is an approach that combines automated imaging and quantitative data analysis in a high-throughput format and allows to simultaneously evaluate different parameters. We successfully optimized HCS methodology in a 384-well format for the parallel determination of antiviral activity and toxicity of microbial extracts by a dual labeling of HCoV-OC43 and cell nuclei (Fig. 5A). The infection conditions were maintained between the two methodologies, with the exception of the experiment endpoint that was reduced to 4 days post-infection for HCS since we are no longer quantifying CPE/death inhibition but viral presence in viable cells. To optimize the methodology, we assayed three different densities of MRC-5 batch-infected cells (MOI of 0.1) in the 384-well format. A density of 3,000 infected cells per well was selected. Z′ factor for this assay was 0.88, and the S/B window was 21.79 compared to the S/B of 5.11 from the CPE inhibition assay.

Fig 5.

Set up of the high-content screening assay with reference antivirals. (A) Immunofluorescence of infected cells treated with increasing concentration of the reference compound ribavirin. (B) Determination of EC₅₀ values for ribavirin, mycophenolate mofetil, and chloroquine after 72 h incubation with the drugs and 4 days of infection. EC₅₀ values were calculated using the SigmaPlot software.

We validated the HCS model with the three reference compounds previously assayed: ribavirin, chloroquine, and mycophenolate mofetil. EC₅₀ values were mostly consistent between both methodologies, except for mycophenolate mofetil that showed a much lower EC₅₀, 1.21 µM, compared to 20.66 µM determined in the CPE inhibition assay (Fig. 5B). The EC₅₀ obtained this way for chloroquine was 1.23 µM and that for ribavirin was 92.08 µM.

Hit confirmation

We assessed the activities of the extracts selected in the CPE inhibition assay and confirmed their antiviral activity by cherry-picking. The average Z′ factor was 0.780 ± 0.09, and the fluorescent signal in infected cells when compared with infected cells treated with 400 µM ribavirin was 30 times higher, improving the difference observed in the CPE assay (Fig. 6A and B). We established a cut-off value of less than 70% infection for hit confirmation (Fig. 6C). Out of the 40 hits, 32 (9 actinomycetes and 23 fungi) confirmed their antiviral activity (Table 1) using the immunodetection bioimaging assay with a specific anti-HCoV-OC43 antibody. The high rate of hit confirmation using two independent assays further reinforces the robustness of the approach. Dose-response experiments were subsequently performed to evaluate extract potency. Dose-response curves consisting of five points of serial twofold dilutions were assayed in duplicate (Fig. 7). Twenty extracts presented an EC₅₀ value lower than 0.7 WBEs and were selected and progressed for LC-MS dereplication.

Fig 6.

High-content screening assay validation. (A) Distribution of the Z′ factor along the different plates. (B) Distribution of the fluorescence signal of the negative control (infected cells treated with 400 µM ribavirin) and the positive control (infected cells with 0.2% DMSO). (C) Representation of the 40 hits from the primary screening. The green line represents the cut-off for hit selection (70% of infection), and the circles in green correspond to the 32 extracts that presented a percentage of infection, compared to controls, lower than 70%.

Fig 7.

Dose-response curves for potency determination. (A) Visualization of an HCS 384-well plate. Column 1 (from A2 to P2) corresponds to the 100% infection control: MRC-5 cells infected with HCoV-OC43 at 0.2% DMSO. Column 22 (from A23 to P23) corresponds to the 0% infection control: MRC-5 cells infected with HCoV-OC43 treated with 400 µM ribavirin. Columns 2–21 contain 5 points of twofold dose-response curves of extracts. Example of one of the dose-response curves in duplicate is squared in orange. (B) Heatmap of the infection percentage in dose-response curves of the 32 hits. EC₅₀ values are expressed as WBEs. (C) Examples of dose-response curves of extracts. ID-1 and ID-15 correspond to active potent extracts (EC₅₀ < 0.7 WBE) while ID-32 corresponds to a non-potent extract (EC₅₀ > 0.7 WBE). EC₅₀ values were calculated using the SigmaPlot software.

Early LC-MS dereplication and bioassay-guided fractionation

LC-MS analysis allowed for the detection and identification of the main components of active extracts. A MEDINA in-house application was used to search for matches of the LC-UV-MS data of the metabolites found in the active extracts to known metabolites already registered in our database (24) that was used. Although this application makes an identification of known compounds present in extracts based on their UV spectrum, positive and negative mass spectra, and retention time, it cannot be ruled out that isomers of compounds included in the database might produce identical sets of data, and therefore, for a full confirmation or the identity of the molecule, it is advisable to perform its isolation and structural confirmation using NMR approaches. While extract composition may vary depending on culture conditions and batches, the reproducibility of the composition of extracts under different culture conditions was monitored to ensure robustness of the procedure. We identified 19 known compounds (Fig. 8A) present in the bioactive extracts and purchased those commercially available: destruxin A (extract 15), destruxin B (extract 15), linoleic acid (extract 7), and gymnoascolide A (extract 3). The four compounds were tested in both assays (Fig. 8B and C) for antiviral activity, and their cytotoxicity was also evaluated. While destruxins A and B resulted as non-active against HCoV-OC43 infection in MRC-5 cells and showed cytotoxicity in the range of 10 µM, linoleic acid and gymnoascolide A exhibited EC₅₀ values of 18.31 and 20.62 µM, respectively, with no evident toxicity against MRC-5 host cells (Fig. 8B and C). Gymnoascolide A was later isolated from the corresponding scaled-up crude extract, and its identity was confirmed by ¹H and HSQC NMR (Fig. S2). Additionally, we evaluated by RT-PCR the presence of HCoV-OC43 in culture supernatants after 72 h of incubation with gymnoascolide A and linoleic acid (Fig. 8D). Gymnoascolide A at 80 µM reduced the viral load in the supernatant below 5%, and at 40 µM, HCoV-OC43 RNA levels remained under 20%. For linoleic acid, the HCoV-OC43 RNA load was below 50% at a 15 µM concentration.

Fig 8.

Antiviral activity and cytotoxicity evaluation of the known metabolites identified by LC-MS. (A) List of the 19 identified known compounds in LC-MS analysis. The ID number of the extracts containing the known compounds is indicated. (B) Dose-response curves for CPE inhibition at 5 days post-infection and 96 h in the presence of each drug. EC₅₀ values for the CPE inhibition assay and CC₅₀ values for MRC-5 cells. (C) Dose-response curves obtained with the HCS assay at 4 days post-infection and 72 h in the presence of the drug. EC₅₀ value for the HCS assay. (D) Extracellular HCoV-OC43 RNA levels after 72 h exposure to gymnoascolide A and linoleic acid. EC₅₀ and CC₅₀ values were calculated using the SigmaPlot software.

After LC-MS dereplication, 15 extracts containing potentially novel components and showing well-resolved metabolic profiles were prioritized for small-scale regrowth. The antiviral activity was confirmed for 10 of them.

DISCUSSION

The outbreaks of CoV infections that cause high case-fatality rates in the last decade since the SARS-CoV-1 outbreak in 2003 followed by MERS in 2012 and SARS-CoV-2 in 2020 have highlighted the inadequacy of existing antiviral treatments. Together with vaccines, drugs are essential to battle future epidemics making necessary the discovery of new antivirals. Microbial natural products are an interesting alternative for drug discovery due to their chemical diversity. The main purpose of this work was to establish a robust HTS platform for the discovery of new antivirals from microbial extracts that may serve as broad-spectrum antivirals for future outbreaks. Fundación MEDINA has been available for screening the world’s largest collection of microbial natural product extracts harboring over 200,000 microbial natural product extracts generated from over 190,000 strains of filamentous fungi, actinomycetes, and unicellular bacteria cultured in diverse formats to ensure a rich metabolite-producing spectrum. This collection has been thoroughly annotated with both chemical and biological data for tailor-made drug discovery purposes. In the current proof-of-concept (pilot) campaign, we screened 1,280 extracts, 390 of them from fungi and 960 from actinomycetes. The use of natural product collections requires expertise to isolate and identify the pure compound responsible for the antiviral activity. The workflow (Fig. 3) used in this platform guarantees that only those extracts tested and rigorously proven to contain activity and potentially new metabolites are progressed.

We have successfully implemented two different methodologies to assess the antiviral activity of microbial crude extracts against HCoV-OC43. The host selected for HCoV-OC43 propagation is a human lung fibroblast cell line, MRC-5. HCoV-OC43 infection in MRC-5 cells produced a more pronounced CPE after 5 days of infection when compared with hamster cell line BHK-21. In a recent work, Schirtzinger et al. also found MRC-5 to be the most suitable host for HCoV-OC43 infection when compared to other cell lines commonly used as viral hosts (15). The plaque assay, widely used in viral studies, was not suitable for HCoV-OC43 infection of MRC-5; despite HCoV-OC43 being a lytic virus, the cell morphology made the plaques grow into each other (15). In addition to the plaque and antibody-based assays for TCID50 determination, resazurin reduction is a fluorimetric assay that has also been used in antiviral and neutralization assays when evaluating lytic virus infection (33, 34).

Fluorimetric-based assays have some limitations. In the case of resazurin, since its conversion depends upon an enzymatic reaction, certain compounds may interfere by directly inhibiting this process. Likewise, colored compounds may interfere with the assay (33). To overcome the possible limitations, we set up an HCS assay that allows us to simultaneously evaluate both antiviral activity and possible toxicity of different sets of compounds by quantifying indirect immunostaining of the HCoV-OC43 nucleocapsid and nuclei staining of MRC-5 cells.

We therefore have established a new and standardized methodology for the evaluation and quantification of HCoV-OC43 infection upon treatment with either pure compounds or microbial natural extracts that may have potential antiviral activity. The methodology was first validated with three compounds that had already been described as antivirals: chloroquine, mycophenolate mofetil, and ribavirin (28 – 31). All three compounds showed activity against HCoV-OC43 in both assays. Although mycophenolate mofetil showed promising results, in a previous study, severe side effects in an in vivo model in marmosets infected with MERS (35) and no antiviral activity against HCoV-OC43 in a mouse model were described (29).

Our efficient LC-MS dereplication methodology enables us to deprioritize active extracts containing known metabolites, thus allowing us to focus on potentially novel-metabolite(s)-containing extracts. The identification of novel compounds is key to develop new antiviral treatments; however, we should not overlook known compounds that could provide a starting point in the development of new therapies. Destruxins A and B were previously described as antivirals against hepatitis B virus, although they did not have an effect on HCoV-OC43 infection (7, 36). On the other hand, linoleic acid and gymnoascolide A showed activity against HCoV-OC43. While this is the first time that gymnoascolide A is described to have anticoronavirus activity, different in silico studies have related linoleic acid to interact and interfere with SARS-CoV-2. Toeltzer et al. were the first to describe the interaction of linoleic acid with the RBD of the spike protein of SARS-CoV-2 (37). In a recent study, linoleic acid was reported to interact with and inhibit RNA-dependent RNA polymerase in silico. They also performed in vitro and in vivo studies, where they exposed the cells to linoleic acid at different time points of the infection. In both models, they observed an antiviral effect of linoleic acid with the reduction of viral RNA load in the media and viral RNA present in tissues, respectively (38).

The confirmation of the antiviral activity of linoleic acid and gymnoascolide A validates the strategy used in this screening approach. From the LC-MS dereplication data, we can conclude that our screening platform is able to identify known compounds from crude extracts with antiviral activity, proving that we can dereplicate extracts containing cytotoxic metabolites or known compounds that have not been previously described as antivirals. The drawback of this methodology is that we may miss active metabolites that are less abundant than the cytotoxic ones.

This is the first report of an established HCS methodology for the discovery of new anti-HCoV-OC43 antivirals present in complex microbial extracts. From a collection of 1,280 extracts, we have identified and confirmed 10 different fungal extracts with new potent anti-HCoV-OC43 activity. These hits are currently undergoing further studies to identify the metabolites responsible for the antiviral activity. We have shown that this platform can be used to validate already pure compounds with previously described antiviral activity such as ribavirin or linoleic acid but can also be used for the discovery of novel antiviral activity of known natural products, as in the case of gymnoascolide A.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/spectrum.01679-23.

Supplemental material. spectrum.01679-23-s0001.pdf.

Fig. S1 and S2.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.