Inhibition of the SARS-CoV-2 3CLpro main protease by plant polyphenols

Graphical abstract


Introduction
Coronavirus disease 2019  is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which was first reported in December 2019. In the early months of 2020, COVID-19 spread worldwide, to cause a pandemic . This first became a global health problem, and then later a socioeconomic problem, and it has claimed more than 4 million lives to date. Importantly, together with the relevant authorities, the medical and academic communities responded quickly, with intensive research campaigns that were accompanied by proposals for social guidelines, hygiene practices and the dissemination of quality information (Fry et al., 2020).
The SARS-CoV-2 virus belongs to the Coronaviridae family, and is named after its 'crown' serrations (Kahn & McIntosh, 2005). It is a single-stranded positive-sense RNA virus , and its virulence ranges from a common cold to serious pathogenic potential, as demonstrated by the outbreaks of SARS in mainland China and Hong Kong in 2003, and of Middle East respiratory syndrome in Saudi Arabia, mainland China, United Arab Emirates and the Republic of Korea in 2012 (de Wit et al., 2016). In addition, swine acute diarrhoea syndrome coronavirus has been described in association with the plethora of animal diseases caused by coronaviruses, and it causes highly pathogenic severe acute porcine diarrhoea syndrome (Yang, Yu et al., 2020). Therefore, the exploration of bioactive constituents in foods and the development of novel therapeutic options and drugs are essential for the future control of coronavirus infections (Sarkar et al., 2020;Khan, Umbreen et al., 2021).
Polyphenols are a structurally diverse group of aromatic compounds that contain multiple hydroxyl groups (Abbas et al., 2017). These naturally occurring compounds are produced through secondary metabolism in plants, where they are involved in defence mechanisms against pathogens and UV radiation (Treutter, 2006). Due to the abundance of polyphenols in edible plants, they constitute an important part of the human diet (Williamson, 2017). In the past, several studies have suggested positive effects of plant polyphenols on human health, as different polyphenols have demonstrated antioxidant, antiinflammatory, antitumour, anti-allergic and antimicrobial activities (Rasouli et al., 2017).
Polyphenols have also been shown to have antiviral activities and even studied in the context of SARS-CoV-2 main protease (Khan, Heng et al., 2021). Different mechanisms of attenuation of viral infections by polyphenols have been demonstrated, such as impairment of the viral replication cycle  and prevention of virus entry through blocking attachment of viral particles to host cells (Calland et al., 2015). In the light of the current COVID-19 pandemic, a number of studies have considered the polyphenols as possible natural remedies to combat SARS-CoV-2 infection (Mehany et al., 2021, Paraiso et al., 2020, El-Missiry et al., 2021. However, the vast majority of studies that have investigated the interactions of polyphenols with SARS-CoV-2 proteins have only applied in-silico methods. Nevertheless, selected polyphenols have been shown to interact with different SARS-CoV-2 proteins in-vitro, including the receptor binding domain of the Spike glycoprotein (Goc et al., 2021), and the 3CL pro which are among the most studied SARS-CoV-2 drug targets.
3CL pro (cysteine protease; EC 3.4.22.69) in particular is crucial for the cleavage of coronavirus polyproteins to form mature non-structural proteins that are themselves essential for viral replication mechanisms. Proteolysis itself occurs via a catalytic dyad defined by Cys145 and His41 and this enzyme has been successfully targeted by a multitude of medicinal chemistry approaches (Dai et al., 2020). In this work we have conducted a multitude of molecular dynamics experiments on a complete set of in-house available polyphenols supported by in-vitro biological evaluation.

Fig. 1.
Structures of the polyphenols used in this study. The polyphenols are divided into seven groups, according to their structures: phenolic acids, hydroxycinnamic acids, phenylacetic acids, flavonoids, curcuminoids, stilbenes and other polyphenols.

Polyphenol library preparation and molecular docking
A selection of 19 plant polyphenols that we study in-house and have available as physical samples ( Fig. 1) was obtained via the PubChem Compound Database as SMILES strings. They were checked for structural errors, and their three-dimensional (3D) structures were calculated using the Schrödinger ligand preparation workflow and the Ligprep software (release 2021-1; Schrödinger LLC, New York, NY, USA). The final 3D conformations included enumeration of chiral centres and tautomeric structures, ionisation at pH 7.4, and minimisation using the OPLS3e force field. In the end, a focused library of 39 polyphenols was obtained, with these expanded in the subsequent docking and consensus scoring to increase the ranking power and to gain information on the predicted binding modes (Houston & Walkinshaw, 2013).
The docking software GlideXP (Friesner et al., 2006) was used for the consensus scoring (release 2021-1; Schrödinger LLC, New York, NY, USA), along with Auto Dock VINA (Trott & Olson, 2010) implemented in the Yasara structure (release 20.12.24) and CmDock (version 0.1.3; https://gitlab.com/Jukic/cmdock). The 3D structure of the SARS-CoV-2 3CL pro was obtained from the Protein Data Bank (PDB ID: 6M2N). The structure of the 3D protein complex was solved at 2.20 Å using X-ray diffraction, as a homodimer with an inhibitor (3WL; polyphenolic; 5,6,7trihydroxy-2-phenyl-4H-chromen-4-one) bound to the active sites (one binding site per subunit). The crystal structure active site residue positioning has been carefully checked against all available similar structures and the HOLO structure was chosen in order to provide accessible "open" conformation of the catalytic active site. The multistep Schrödinger protein preparation workflow (release 2021-1; Schrödinger LLC, New York, NY, USA) was used for the final preparation of the molecular docking with the receptor. The prepared protein structure was free of any ligand and water molecules, hydrogens were added, the H-bonding network was optimised, and the final structure was edited with restrained minimisation of heavy atoms towards 0.3 Å in the OPLS3e force field. For the CmDock docking package, a docking receptor was generated using CmCavity (Ruiz-Carmona et al., 2014).
Initially, to evaluate all possible binding sites, additional blind docking was performed using AutoDock VINA implemented in the Yasara structure (release 20.12.24) (200 runs). As expected, the regions of the catalytic site in the vicinity of the Cys145 residue of the SARS-CoV-2 3CL pro were most favoured for the ligand binding. Therefore, a cubic simulation cell was constructed with 10 Å extensions along each axis from the centre of the catalytic site, and docking into the catalytic site was repeated (200 runs, 32 exhaustiveness parameters). Next, molecular docking was performed using both the GlideXP software (Gli-deScore version XP5.0; 3 poses per ligand) and the CmDock software (htt ps://gitlab.com/Jukic/cmdock; 100 runs in exhaustive docking mode).

Molecular dynamics
The most favoured complexes from the molecular docking were prepared using the Yasara structure software (release 20.12.24). The proteins were capped (N-terminal acetyl group; C-terminal N-methyl amide group) and residue ionisation was assigned at pH 7.4 (Krieger & Vriend, 2015). A cubic box (10 Å away from all atoms) was solvated with the TIP3P water model and 0.9% NaCl to define a physiological solution. After removing bumps/clashes using steepest descent minimisation, annealing minimisation was used to reach a stable local energy minimum. From here, the simulation was started by assigning random initial velocities and slowly heating the system to 298 K. A molecular dynamics simulation was started using the AMBER14 force field for the solute, GAFF and AM1BCC charges for the ligands, and TIP3P for the water. The temperature was set to 298 K at a pressure of 1 atm (NPT ensemble, Berendsen barostat and thermostat, coupled to time-averaged temperature and density), with snapshots saved every 100 ps. Non-bonded long-range interactions were calculated using the Particle Mesh Ewald algorithm. The Shake algorithm was not used, and the energy of the system was stable throughout the production run, as was the root-mean-square deviation of the protein backbone for all of the systems (details in Supplementary Materials Fig. S1. The production run was performed over a period of 110 ns, with the first 10 ns serving to equilibrate the system.

Cloning, expression and purification of recombinant 3CL pro
The codon-optimised gene encoding SARS-CoV-2 3CL pro with an Nterminal 6 × His tag and a TEV protease cleavage site was obtained (Integrated DNA Technologies, USA) and cloned into the pET-28c(+) plasmid between the NcoI and NotI restriction sites. The resulting expression vector was used to transform chemically competent Escherichia coli NiCo21(DE3) (New England Biolabs, USA). The transformed bacteria were cultured at 37 • C in lysogeny broth supplemented with 50 µg/mL kanamycin, with agitation at 250 rpm. When an OD 600 of ~ 1.8 was reached, the cultures were cooled to 16 • C and 3CL pro expression was induced with 0.2 mM isopropyl β-D-1-thiogalactopyranoside. The cultures were then incubated at 16 • C with agitation at 250 rpm for 24 h. The cells were harvested by centrifugation (3000×g for 20 min at 4 • C), and the cell pellets were stored at − 80 • C until purification of the 3CL pro .
For purification of the recombinant product, the cell pellets were resuspended in buffer A (20 mM Tris, pH 7.5, 0.05 mM ethylenediaminetetraacetic acid [EDTA], 2.5 mM dithiothreitol [DTT], 10% glycerol). The cells were lysed by sonication on ice, and the insoluble cellular debris was removed by centrifugation (15000×g for 10 min at 4 • C; repeated twice). The clarified supernatant was then filtered to a 100-kDa molecular weight cut-off (Amicon Ultra-15 centrifugal filter units; Merck, Germany). The filtrate was combined with 0.5 M (NH 4 ) 2 SO 4 (final concentration) and loaded onto a 1 mL HiTrap Phenyl HP column (Cytiva, USA) that was pre-equilibrated with buffer B (50 mM Tris, pH 7.5, 0.5 M (NH 4 ) 2 SO 4 , 0.05 mM EDTA, 2.5 mM DTT, 10% glycerol). The column was then washed with 15 mL buffer B and the 3CL pro protein was eluted from the column using a linear gradient of 0% to 100% buffer A in buffer B. The 3CL pro protein eluted as a broad peak at ~ 70% buffer A. The eluted 3CL pro was concentrated with a 30-kDa molecular weight cutoff (Ultra-4 centrifugal filter units; Amicon), snap-frozen in liquid nitrogen, and stored at − 80 • C. The purity of the isolated 3CL pro was assessed by SDS-PAGE using 12% polyacrylamide gels. The concentration of 3CL pro was determined using absorbance at 280 nm and an extinction coefficient of 34380 M − 1 cm − 1 (using the ProtParam bioinformatic tool).

3CL pro inhibition assays
The inhibitory activities of the polyphenols against 3CL pro were determined using the fluorescence resonance energy transfer (FRET) peptide substrate MCA-AVLQSGFR-Lys(Dnp)-Lys-NH 2 (Isca Biochemicals, UK). Stock solutions of 8 mM polyphenols were prepared in 100% dimethylsulphoxide (DMSO), and were diluted in 50 mM Tris (pH 7.3), 1 mM EDTA for the analyses. The same buffer was used for dilution of the substrate and the 3CL pro .
For the initial screening, 10 µL of the different polyphenols at 40 μM were pre-mixed with 10 µL of the FRET substrate at 64 μM. The reactions were initiated by addition of 20 µL 3CL pro at 0.2 μM, to the final reaction volume of 40 µL. The final concentrations in the reaction mixtures were thus 0.1 µM 3CL pro , 16 µM substrate and 10 µM polyphenol. For the blanks, 20 µL buffer was added to the reaction mixtures instead of the 3CL pro . All of the reactions contained 1.25% (v/v) DMSO, including the control reactions without any polyphenols added. The reactions were carried out in black 384-well plates. Fluorescence signals were recorded continuously every 15 s for 10 min after addition of 3CL pro (Spark microplate reader; Tecan, Switzerland), at 320 nm excitation and 405 nm emission.
To determine the concentrations of the selected polyphenols required to inhibit 50% of the 3CL pro activity (IC 50 ), 10 µL polyphenols at different concentrations were mixed with 20 µL 3CL pro and incubated at 25 • C for 30 min, followed by addition of 10 µL substrate. The final concentrations in the reaction mixtures were again 0.1 µM 3CL pro and 16 µM substrate. For the blanks, the polyphenols were mixed with 20 µL buffer instead of 3CL pro before addition of the substrate. These reaction mixtures were incubated in black 384-well plates at 25 • C. Two hours after addition of the substrate, the fluorescence signals were recorded, as above.

Surface plasmon resonance
For the surface plasmon resonance (SPR) assays, 1.45 mg/mL 3CL pro was dialysed against 10 mM HEPES (pH 7.3), 1 mM DTT, and then diluted in 10 mM sodium acetate buffer (pH 5.0) to the final concentration of 100 µg/mL 3CL pro . The assays were performed on a SPR system (Biacore T200; GE Healthcare, USA) at 25 • C. The 3CL pro protein was immobilised via the free amino groups using CM5 sensor chip coated with carboxymethylated dextran (Cytiva, USA). Before the covalent immobilisation of 3CL pro , the chip surface was activated with 0.4 M 1-ethyl-3-(3-dimethylaminopropyl)-carboiimide hydrochloride and 0.1 M N-hydroxysuccinimide. The injection of 3CL pro was adjusted to obtain 9000 response units of immobilised 3CL pro on the chip surface, and the uncoupled carboxymethyl groups were blocked with ethanolamine. The running buffer for the SPR assays was 10 mM HEPES (pH 7.4), 1 mM EDTA, 150 mM NaCl, 0.005% P20 and 5% DMSO. The polyphenol samples were prepared by diluting the polyphenol stock solutions in running buffer. The final concentration of DMSO in all of the polyphenol samples was 5%. Different concentrations of the polyphenols were injected over the chip surface for 60 s (250 s for ellagic acid). Dissociation of the polyphenols from the chip surface was then monitored for 60 s (200 s for ellagic acid). The polyphenols that remained bound to the chip surface were removed after the dissociation phase by a 6 s pulse of 2 mM NaOH. The buffer flow rate was set at 30 µL/min for all of the phases of the SPR assays. Sensorgrams were corrected for the corresponding responses of flow-cell 1 without the immobilised 3CL pro . The Biacore T200 evaluation software was used to determine the equilibrium dissociation constants (K D ), by fitting the data to a steady-state affinity model or to two-state kinetics. All of the experiments were performed in duplicate.

Circular dichroism
The 5 mM stock solutions of polyphenols for the circular dichroism (CD) measurements were prepared in 96% ethanol. Tris buffer (50 mM, pH 7.3) with 1 mM EDTA was used to dilute the 3CL pro and polyphenol stocks to the final concentrations of 0.1 mg/mL and 100 µM, respectively. For the CD measurements, quartz cuvettes with an optical path of 1 mm were used. The CD spectra were scanned (J-1500CD spectrometer; Jasco, Japan), from 250 nm to 200 nm. The scanning speed was set at 20 nm/min, with a bandwidth of 1.0 nm. All of the CD spectra were collected at 20 • C. The spectra of the blank samples (buffer with polyphenols) were subtracted from the corresponding 3CL pro spectra. The mean residue weight of 111 g/mol was used to calculate the molar ellipticity of 3CL pro .

In-silico screening of a polyphenol focused library
The top scoring 18 plant polyphenols obtained were subjected to cluster analysis to obtain representative top scoring binding conformations, and the results of the consensus docking experiment are shown in Table 1.
Polyphenols are often known as 'promiscuous' binders, and we fully acknowledge their inclusion in lists of pan-assay interference compounds (Bael & Holloway, 2010). Nevertheless, we felt compelled to investigate whether they have the potential to bind to the active site of SARS-CoV-2 3CL pro . Furthermore, we advocate their potential use as scaffolds or hits ideal for scaffold hopping, similarity searches and further molecular optimisation. To this end, we performed additional blind pre-docking analyses in which the polyphenols were docked as potential ligands to the entire 3CL pro protein without prior knowledge of the binding sites. Fig. 2A shows that ligands can generally interact with several different regions on the 3CL pro dimer structure, although the catalytic site regions produced the most favourably scored binding poses. If focusing on the active site of 3CL pro , Fig. 2B demonstrates the Glu166. Further π-π interaction can be observed between ligand aromatic system and His41 where ligand occupies P1 ′ and P1 subsites extending towards P2 pocket of the active site (Fig. 2). After calculation of the consensus docking scores as the mean of the AutoDock VINA, Glide XP and CmDock scores, additional re-scoring was carried out using the Prime molecular mechanics/generalised Born model and solvent accessibility (MM/GBSA) methodology, as implemented in Schrödinger SMD (release 2021-1; Schrödinger, LLC, New York, NY, USA). Here, MM/GBSA dG binding was estimated via the Prime energy of the optimised complex minus the Prime energy of the free ligand and the free receptor, all in the OPLS3e force field with implicit solvent. The consensus docking scores correlated well with the MM/GBSA dG binding estimates, as well as with the Glide XP scores, which suggests that our top scorers might be promising binders of SARS-CoV-2 3CL pro . Upon inspection of the top-scoring binding poses, we observed a general similar binding motif to the crystal complex (PDB ID: 6M2N) of the polyphenolic ligand (5,6,7-trihydroxy-2-phenyl-4H-chromen-4-one), where all of the screened polyphenols were positioned at the active site or at the active site gorge entrance.

Molecular dynamics contact analysis
To gain further insight into the compound interaction with the 3CL pro protein, we performed a plethora of molecular dynamics simulations on calculated binding poses of examined polyphenols. We were focusing on the stability of the complexes (see Supplementary Materials for details of all of the complexes examined) and we therefore examined the conformational stability of the examined polyphenols during the molecular dynamics production runs. We prepared an in-house script which recorded the contact maps of each polyphenol during the production runs, with a total of 12 polyphenols at 100 ns each (1.2 µs production simulation time). From the contact maps, it was observed that the major contacts that persevered through the simulation production time of all examined polyphenols with the 3CL pro protein were towards the Met49, Glu166, Asp187 and Gln189, placing the compounds at the entrance or in the P2 pocket of the 3CL pro active site. Further observed major contacts were towards His41, Met165 and Glu166 at the P1 and P1 ′ 3CL pro active site pockets. The identified contacts (contact maps can be found in Supplementary Materials) Fig. 2. (A) Blind docking to identify binding sites, where the 3CL pro protomers are shown (blue, red cartoon models), with the binding sites (green) shown as volume representations; the purple arrows indicate the 3CL pro active sites. Each protomer is formed by three domains: I and II (residues 8-101 and 102-184, respectively) that host the substrate-binding site and domain III (residues 201-303) that regulates protein dimerization (B) Ellagic acid (purple stick model) at the 3CL pro active site with protein depicted in green cartoon model, binding site surface emphasised in grey and showing neighbouring residues with additional description of pocket subsites and catalytic Cys145 emphasised in red stick model. (C) Two-dimensional contact map of ellagic acid in the 3CL pro active site. (D) Interaction diagram of ellagic acid during the 100 ns production run (plus first 10 ns for equilibration), showing the binding mode stability at the 3CL pro active site. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) therefore place the polyphenols at the active site during the majority of the production runs, mainly at the P1-P1 ′ pockets, reaching also towards P2 subsite. Only two polyphenols: ferulic acid and resveratrol dissociated at the beginning of molecular dynamics simulations at 10 and 12 ns respectively (see Supplementary Materials for further details).
Epigallocatechin gallate, ellagic acid, curcumin, resveratrol and quercetin were subjected to further dose-response studies, to determine the concentrations required to inhibit 3CL pro activity by 50% (IC 50 ). Those polyphenols were selected on the basis of their relatively high inhibitory activities in the initial in-vitro screening (Fig. 3). For determination of the IC 50 values, increasing concentrations of these polyphenols were incubated with 3CL pro before the addition of the FRET substrate, and the fluorescence signals were determined 2 h later. The IC 50 values for EGCG, ellagic acid, curcumin, resveratrol and quercetin were 13.9 µM, 11.8 µM, 11.9 µM, 16.9 µM and 23.4 µM, respectively (Fig. 4). Thus, the potencies of these selected polyphenols for 3CL pro inhibition were similar in this study. Of note, none of these polyphenols perturbed the secondary structure of 3CL pro to any great extent (see Supplementary Materials Fig. S3 for further details). Therefore, the inhibition of 3CL pro by these polyphenols appears to be due to their block of the 3CL pro active site, and not due to disruption of the conformation of 3CL pro . These IC 50 values are comparable to, or lower than, those of the compounds that were identified as the top five SARS-CoV-2 3CL pro inhibitors among the 3987 US Food and Drug Administration approved drugs in a study by Mody et al. (2021).
Of these five polyphenols that showed the greatest inhibition of SARS-CoV-2 3CL pro in the present study, only quercetin and EGCG had been examined previously in in-vitro inhibition analyses of this protease. Quercetin was previously identified as the most potent inhibitor of SARS-CoV-2 3CL pro among 150 different compounds screened in-vitro, with an inhibition constant (K i ) estimated at 7.4 µM (Abian et al., 2020). This is in a similar range to the quercetin IC 50 determined in the present study (23.4 µM) (Fig. 4). An IC 50 for EGCG that was reported recently (0.056 µM) (Du et al., 2021) was considerably lower than that in the present study (13.9 µM) (Fig. 4), and in other recent studies, at 4.2 µM (Chiou et al., 2021) and8.8 µM (Ni et al., 2021). Interestingly, EGCG appears to inhibit SARS-CoV-2 3CL pro more strongly than SARS-CoV 3CL pro , which was demonstrated previously (Chiou et al., 2021).
Resveratrol and curcumin were among these five strongest SARS-CoV-2 3CL pro inhibitors in the present study (Fig. 3), and these are also widely known natural compounds with antiviral activities (Campagna & Rivas, 2010, Mathew & Hsu, 2018. In the two recent studies, it was demonstrated that resveratrol inhibited proliferation of SARS-CoV-2 in mammalian cell cultures, with EC 50 values from 5 µM to 10 µM (Yang, Wei et al., 2020, Pasquereau et al., 2021. Curcumin was previously shown to inhibit SARS-CoV 3CL pro in-vitro (Ryu et al., 2010), and was also predicted to bind stably into the SARS-CoV-2 3CL pro active site in-silico (Huynh et al., 2020). To the best of our knowledge, no previous studies have investigated the inhibition of SARS-CoV-2 3CL pro by curcumin and resveratrol in-vitro. Ellagic acid is a tetracyclic polyphenol that is abundant in berries and nuts (Daniel et al., 1989), and it was also previously predicted to inhibit SARS-CoV-2 3CL pro , although only insilico (Khalifa et al., 2020, Ni et al., 2021. The present study confirms the strong inhibitory activity of ellagic acid against this protease in-vitro, with an IC 50 in the micromolar range. Fig. 3. 3CL pro inhibitory activities of the polyphenols, ordered according to increasing inhibitory activities. Data are means ± standard deviation of three experiments. The dotted line marks 50% 3CL pro activity. HNPA, 4-hydroxy-3-nitrophenylacetic acid; EGCG, epigallocatechin gallate.

Interactions of polyphenols with 3CL pro
The binding of the selected polyphenols to 3CL pro was further studied using SPR spectroscopy, which allowed measurement of their interactions in real time. Here, the interactions were examined for three polyphenols that showed high 3CL pro inhibitory activities, as EGCG, ellagic acid, and resveratrol, and for two polyphenols that showed weak or no 3CL pro inhibitory activities in the initial in-vitro screening (Fig. 3), as syringic acid and caffeic acid. Although curcumin and quercetin were also among the most potent 3CL pro inhibitors identified here, they were not examined by SPR due to their low solubilities.
Epigallocatechin gallate, ellagic acid and resveratrol bound to the 3CL pro immobilised on the chip (Fig. 5). Sensorgrams of EGCG and resveratrol with 3CL pro indicated rapid association and dissociation of these polyphenols. The responses of EGCG and resveratrol did not reach plateaus even with these polyphenols at 200 µM. Together with the fitting of the sensorgrams to steady-state kinetics (see Supplementary Materials Fig. S4 for more details), this SPR analysis suggested nonspecific binding of EGCG and resveratrol to 3CL pro . In contrast, the two-state kinetics model provided a fit for the ellagic acid-3CL pro interaction, which indicated a K D of 311 ± 69 µM. The sensorgrams suggested that the association of ellagic acid with 3CL pro is slower than that of EGCG and resveratrol, and it can be noted that the dissociation of ellagic acid from 3CL pro appeared to occur in two steps, as an initial rapid decrease in the SPR response that was followed by a slow dissociation. This is likely to be the consequence of multiple ellagic acid binding sites on 3CL pro with different affinities. Syringic and caffeic acids only bound weakly to 3CL pro , with EGCG, ellagic acid and resveratrol showing 10-fold higher RU values (Fig. 5). This is consistent with the low inhibitory activity of syringic and caffeic acids towards 3CL pro .

Conclusion
We initially developed a consensus docking approach that resulted in a library enrichment of 2.5-fold, which was determined by the biological evaluation in the following step. IC 50 values were obtained for EGCG, curcumin, resveratrol, quercetin and ellagic acid of 13.9 µM, 11.9 µM, 16.9 µM, 23.4 µM and 11.8 µM, respectively. Further in-silico molecular dynamics studies indicated labile interactions for resveratrol, while they produced stable binding conformations at the active site of 3CL pro for the other four polyphenols. EGCG showed hydrophobic interactions with 3CL pro residues Met49, Met165 and Gln189, and hydrogen bonds with residues Asp187 and Glu166. Curcumin formed hydrogen bonds with 3CL pro residues Gln192 and Arg188. Ellagic acid occupied a binding mode at the active site of 3CL pro , with hydrogen bonds with residues Cys44, Asn142, His164 and Gln189, and it formed hydrophobic interactions with residues Asn142, Cys145, Met165 and Met166, and π-π stacking with His41. The binding of ellagic acid to 3CL pro was confirmed by SPR using a two-state kinetics model, with a measured K D of 311 ± 69 µM. We have therefore demonstrated that ellagic acid as a low micromolar SARS-CoV-2 3CL pro inhibitor that is suitable for further crystallographic studies. This study also enables the use of ellagic acid and similar polyphenols for scaffold hopping and similarity searches and optimisation, with a view to new potential SARS-CoV-2 inhibitors or novel molecular probes.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.