2.1 Ligand preparation

The structures of 53 flavan-3-ols and 889 flavanones were obtained from the COCONUT database (https://coconut.naturalproducts.net/) (Listing in Supporting data files). The 3D structures were minimized energy using ChemOffice v.16.0 software based on the molecular mechanics-2 (MM2) method until the root mean square (RMS) gradient value lower than 0.001 kcal/mol Å (El Aissouq et al., 2020). These ligands with *.sdf file format were converted to *.pdb file format using OpenBabel v.2.4.1 software.

2.2 Protein preparation

3D structures of target proteins, including N1-H274Y-oseltamivir complex (PDB ID: 3CL0 – oseltamivir-resistant NA) (Ha et al., 2024); 30S ribosome-binding factor (PDB ID: 1PA4 − RBfA) (Rodrigues et al., 2019) and immunodominant protein P40/P90 complex (PDB ID: 6TLZ) (from M. pneumoniae) (Osigbemhe et al., 2022; Rubin et al., 2003), were obtained from the Protein Data Bank (https://www.pdb.org). These structures were removed the residues, undesirable chains, and water using Discovery Studio Visualizer v.2.5 software. They were then optimized based on the AMBER ff14SB force field with 500 steps of steepest descent and 100 steps of the conjugate gradient using Chimera v.1.16 software. The grid of the docking site of these proteins was provided in Table S1 of Supporting data.

2.3 Molecular docking screening for 3CL0 protein

The binding affinities between the ligands and 3CL0 protein were predicted using AutoDock Vina parameters based on the PyRx software (Trott and Olson, 2010). Firstly, the protein was input into the PyRx software and over 900 ligands were then docked into 3CL0 protein at the catalytic active site (Table S1 – S4) with the exhaustiveness and number of modes of 8, and scoring function based on the X-Score function. The candidates were selected when binding affinity values were equal to or lower than the binding energy at −8.4 (kcal/mol) of laninamivir, positive control of 3CL0 protein successfully identified from a previous study (Ha et al., 2024). The interaction effects between these structures and 3CL0 protein were visualized and analyzed to investigate the bifunctional NA inhibitors using Discovery Studio Visualizer v.2.5 and Chimera v.1.16 software.

2.4 Prediction of drug-likeness and ADMET values

Drug-likeness properties and ADMET parameters of the candidate compounds were predicted using the Canonical SMILES files, which were submitted to SwissADME (https://www.swissadme.ch) and pkCSM online server (https://biosig.lab.uq.edu.au/pkcsm/prediction).

2.5 Based-virtual screening for potential ligand

The ligand was applied for based-virtual screening in the ChEMBL database using the SwissSimilarity online tool and based on the combined method for the search algorithm (Zoete et al., 2016). The hits were selected by combining the results of molecular docking screening and similarity scoring values greater than 0.6. Molecular docking screening using AutoDock Vina (Trott and Olson, 2010) and SwissADME prediction (https://www.swissadme.ch) were performed to obtain the most suitable hit.

2.6 Evaluation of the docking methodology for 1PA4 and 6TLZ proteins

Decoys were generated using the Directory of Useful Decoys-Enhanced (DUD-E) online tool (https://dude.docking.org/) to evaluate the docking methodology for protein RBfA, and the active ligands were designed in the ChEMBL database using the SwissSimilarity online tool based on ZINC04259381. These ligands were randomly selected, and molecular docking with RBfA (PDB ID: 1PA4) (Table S1) was performed using AutoDock Vina (Trott and Olson, 2010).

To determine the reliability of the docking methodology for the P40/P90 complex, two decoys (laninamivir and oseltamivir) were selected. The two compounds exhibited comparable pharmacophores, featuring the N-acetylneuraminyl moiety of 3′-sialyllactose and a robust interaction with the NA of the influenza virus. However, no clinical studies have demonstrated these NA inhibitors' ability to reduce the growth of M. pneumoniae in children with influenza (Hume-Nixon et al., 2022). Meanwhile, the active ligands were generated from 3′-sialyllactose using the SwissSimilarity online tool based on the ChEMBL database. These compounds were randomly chosen, and molecular docking was performed with the P40/P90 complex at the active site (Table S1) using AutoDock Vina (Trott and Olson, 2010).

2.7 Molecular docking study of candidates with 1PA4 and 6TLZ proteins

The potential candidates were docked into RBfA (PDB ID: 1PA4) and P40/P90 complex (PDB ID: 6TLZ) at the active site (Table S1) using AutoDock Vina parameters assisted with the PyRx software (Trott and Olson, 2010). The interaction results were visualized and analyzed using Discovery Studio Visualizer v.2.5 and Chimera v.1.16 software.

2.8 Molecular dynamics simulations

The ligand–protein complexes of ChEMBL1779463 with PDB ID 3CL0 or 6TLZ were performed in the molecular dynamics simulations based on the NAnoscale Molecular Dynamics (NAMD) software (Phillips et al., 2005). Topologies and parameter files of ligands and proteins were parameterized using CHARMM36m force field. The solvated and ionized complexes were carried out as in the previous study (Ha et al., 2024). The simulation steps were set at 500 ns using the NPT ensemble and a constant temperature of 310 K. The number of hydrogen bonds (HBs) and the backbone root mean square deviation (RMSD) of these complexes were analyzed using Visual Molecular Dynamics (VMD) v.1.93 software.

3.1 Molecular docking screening for 3CL0 protein

The molecular docking method's reliability was evaluated in a previous study (Ha et al., 2024). The workflow of this study is illustrated in Fig. 1. Since the NA-resistant oseltamivir protein (PDB ID: 3CL0) was used as the target protein, the binding energy of oseltamivir with the catalytic active site (−6.7 kcal/mol) exhibited a significantly higher value than its laninamivir (−8.1 kcal/mol) (Table S2). Therefore, our recent study carried out the molecular docking screening for over 900 flavan-3-ols and flavanones collected from the COCONUT database, and the candidates were selected when binding energies were lower than the threshold of −8.1 kcal/mol (Ki value < 1.145 µM). After this step, the remaining 9 flavan-3-ols and 122 flavanones (Tables S3 and S4) showed suitable properties for further visualization and analysis.

Close

Fig. 1

Procedure of molecular docking screening, based-virtual screening, ADMET, and synthesis prediction of flavan-3-ols and flavanones from the COCONUT database.

Export to PPT

In the flavan-3-ols group (abbreviated as FL) (Fig. 2 and Tables S3, S5), only 3 compounds (FL13, FL22, and FL26) interact with both the active site and the adjacent 430-cavity. FL22 showed the lowest binding energy (−9.9 kcal/mol) with a Ki value of approximately 54.8 nM (Table 1). Several key interactions between FL22 with the catalytic active site and adjacent 430-cavity were obtained, such as Asp151, Arg224, Glu227, Glu277, Arg371, Ile427, Arg430, Pro431, and Lys432 through various interaction types (conventional hydrogen bond, π-alkyl, π-anion,…). Besides, FL22 also has two types of interactions (Van der Waals and conventional hydrogen bond) with the triad of arginine residues Arg118-Arg292-Arg371.

Close

Fig. 2

3D molecular docking binding into the sialic acid binding cavity and the 430-cavity, and 2D interaction diagrams of N1-H274Y-oseltamivir protein (PDB ID: 3CL0) with a flavan-3-ol FL22 (Fig. A) and three flavanones FN166, FN316, and ChEMBL 1,779,463 (Fig. B-D, respectively).

Export to PPT

Other 2 flavan-3-ols, FL13 and FL26, exhibited weak binding energies (−8.5 kcal/mol and −8.4 kcal/mol, respectively) with Ki values around 583.0–690.3 nM (Table 1). Similar to FL22, these compounds also have several key interactions such as conventional HBs, π-alkyl, π-anion/cation, carbon-hydrogen bond, and Van der Waals, with some essential amino acids Arg118, Glu119, Asp151, Arg152, Glu227, Ser246, Glu277, Arg292, Arg371, Ile427, Arg430, and Pro431 (Tables S3, S5). Our docking screening results for 53 flavan-3-ols from the COCONUT database suggested that FL22 was the most suitable candidate compared to FL13 and FL26. Even though only some previous studies suggested that FL22, theaflavin, and their derivatives had the in vitro and in silico anti-NA of different influenza virus strains (Mohamed et al., 2022), there is no report about the role of theaflavin as a potential for interaction with both the active site and the adjacent 430-cavity through the triad of arginine residues. Our results suggested the new mechanism of FL22 and proposed it as a potential bifunctional NA inhibitor (Fig. 2).

Regarding the flavanone group (abbreviated as FN), the binding energies of this skeleton were significantly lower than those of flavan-3-ols. During the analysis, 46 compounds showed binding interactions with both active site and adjacent 430-cavity through several critical amino acids (Tables S4, S6). Among those, two compounds, FN166 and FN316, exhibited the lowest binding energies following −11.4 kcal/mol and −11.7 kcal/mol, respectively, with Ki values around 2.6–4.4 nM (Fig. 2 and Tables 1, S4, S6). Besides the most suitable skeleton (biflavonoid) for bifunctional NA inhibitory effect, the results also revealed that flavanone diglycosides (FN150, FN193, and FN330) showed binding energies better than flavanone glycoside (FN106, FN270, and FN334). Because of their high similarity pharmacophore, FN166 and FN316 were focused for further analysis. As shown in Fig. 2, FN316 and FN166 exhibited key interactions with the catalytic active site and adjacent 430-cavity through the triad of arginine residues.

3.2 Based-virtual screening for ligand FN316

During the above screening procedures for over 900 flavan-3-ols and flavanones, three anti-NA candidates (FL22, FN166, and FN316) were selected for further evaluation for the “drug-likeness” properties based on the SwissADME online tool. As illustrated in Fig. 3 and Table S7, FN166 and FN316 showed fewer violations of Lipinski's rule than FL22. However, FN166 and FN316 still presented unsuitable values, including the number of hydrogen bond donors of more than 5 and molecular weights of over 500 Da. Hence, this study examined further similarities with these flavanones that mitigate the violations of Lipinski's rule. Previous reports intimated that the ChEMBL database contains NPs and NP derivatives, NP mimetics, and NP pharmacophores. Hence, this database was used for based-virtual screening for ligand FN316 using the SwissSimilarity online tool and combined method for the search algorithm with similarity scoring values greater than 0.6. Five similarities were selected, with at least one flavanone moiety not present in the previous list of 889 flavanones. Among those five, four compounds (ChEMBL3763296, ChEMBL3765207, ChEMBL4075132, and ChEMBL4086291) contain alkylamines, while ChEMBL1779463 is a biflavanone structure (Fig.S1). The “drug-likeness” properties of five similarities showed that ChEMBL3763296, ChEMBL4086291, and ChEMBL1779463 had only a violation of Lipinski's rule, while other compounds (ChEMBL3765207 and ChEMBL4075132) did not exhibit any violation of this rule.

Close

Fig. 3

The “drug-likeness” properties based on the SwissADME online tool for four potential anti-NA candidates (FL22, FN166, FN316, and ChEMBL1779463).

Export to PPT

After performing molecular docking screening of the five similarities for 3CL0 protein using the AutoDock Vina tool, ChEMBL1779463 showed the lowest binding energy (−11.4 kcal/mol) with a Ki value of around 4.4 nM (Table S8). In addition, ChEMBL1779463 showed several key interactions with both the catalytic active site and adjacent 430-cavity (Fig. 2). It was suggested that the biflavanone structure could become a more suitable skeleton than flavanone linked with alkylamines.

3.3 ADMET study

Four potential anti-NA candidates (FL22, FN166, FN316, and ChEMBL1779463) were evaluated for ADMET parameters (Table 2). Several ADMET parameters were analyzed in detail, as follows: (1) aqueous solubility of testing compounds was in accepted ranges from −6.5 to 0.5; (2) absorption in the human intestine of ChEMBL1779463 was highest than others; (3) the blood–brain barrier (BBB) permeability of four compounds are lower than 0.3, which considered to un-cross the BBB (Muehlbacher et al., 2011); (4) ChEMBL1779463 showed a better bioavailability score than the remaining other candidates; (5) other indicators (such as P-glycoprotein substrate, total clearance,…) suggested that those testing compounds could be used as safety drugs with less toxicity for human, especially ChEMBL1779463 may become the most potential than others.

3.4 Molecular docking between potential candidates with m. Pneumoniae proteins

RBfA is a cold shock response protein required for bacterial growth at low temperatures and efficient processing of the 5′ end of 16S rRNA during assembly of the 30S subunit (Maksimova et al., 2021). Previous studies revealed that aminoglycosides and ZINC04259381 could bind to the 30S subunit, leading to transcriptional errors that interfere with its structural function and inhibit the protein synthesis of bacteria (Rodrigues et al., 2019). Therefore, in this study, ZINC04259381 was used as a positive control for screening the potential of the four above candidates against 30S RBfA. In comparison with a previous report, several key interactions of ZINC04259381 with target protein were similar in binding to the active site through some essential amino acids such as Arg15, Asn18, Val35 (Fig.S2) (Rodrigues et al., 2019). In addition, this protein's docking method was successfully evaluated for 10 random decoys and 05 random active ligands (Table S9). These results suggested that our docking methodology was suitable for further investigation of testing candidates. As summarized in Table 3, the binding energies of the four candidates were from −8.6 kcal/mol to −9.2 kcal/mol, similar to its positive control (−9.2 kcal/mol). Several key interactions between four candidates with the active site were also analyzed. Fig. 4A demonstrated the interaction of FL22 with RBfA via several amino acids such as Lys6, Ile17, Asn18, His34, and Val35 by various interaction types. Compounds FN166, FN316, and ChEMBL1779463 did not have π-cation likely FL22, but they still bound into the active site through several critical amino acids (Fig. 4B-D).

Close

Fig. 4

3D molecular docking binding into the active site and 2D interaction diagrams of 30S RBfA (PDB ID: 1PA4) with a flavan-3-ol FL22 (A) and three flavanones FN166, FN316, and ChEMBL 1779463 (B-D, respectively).

Export to PPT

The effect of four anti-NA candidates were investigated the binding effect on the immunodominant protein P40/P90 complex against M. pneumoniae adhesion to host cells. A previous study reported the interaction of P40/P90 with 3′-sialyllactose through several important amino acids, such as Pro235, Pro237, Lys238, Phe631, and Thr632 (Vizarraga et al., 2020). Therefore, the site containing these amino acids was described as the binding site for sialic acid-based molecular mimicry. In comparison with the previous report, key interactions of sialyllactose with protein P40/P90 were bound to the sialic acid binding site (Fig.S3) (Vizarraga et al., 2020). In addition, results in Table S10 indicated that 02 decoy compounds showed binding energies from −4.9 to −5.3 kcal/mol, significantly higher than 3′-sialyllactose (−6.1 kcal/mol). However, 07 random active ligands exhibited binding energies similar to 3′-sialyllactose. These results revealed that this molecular docking methodology could be applied for further study. Four anti-NA candidates showed binding energies significantly lower than 3′-sialyllactose, with values from −7.7 to −8.7 kcal/mol (Table 3). In a detailed analysis, three compounds (FN166, FN316, and ChEMBL1779463) exhibited great interactions with protein P40/P90 via several amino acids, while FL22 exhibited fewer interactions than others (Fig. 5).

Close

Fig. 5

3D molecular docking binding into sialic acid cell receptor and 2D interaction diagrams of the immunodominant protein P40/P90 complex (PDB ID: 6TLZ) with a flavan-3-ol FL22 (Fig. A) and three flavanones FN166, FN316, and ChEMBL 1,779,463 (Fig. B-D, respectively).

Export to PPT

Biflavanone could become a potential structure for inhibiting the M. pneumoniae via (1) binding to the 30S subunit, leading to transcriptional errors and interfering with the protein synthesis; (2) inhibiting the bacteria attachment to host cells through binding to the sialic acid cell receptor P40/P90. According to the above molecular docking results, molecular docking combined with other computer techniques allowed for inexpensive, highly effective, and easy-to-perform, leading to advantages for further drug development. However, several drawbacks need to be improved for this computer tool, including the conformational flexibility of target proteins and weak or transient binding of ligand–protein complexes. Therefore, further testing in vitro experiments is suggested for potential candidates.

3.5 Molecular dynamics simulations

3CL0 and 6TLZ were chosen to further validate the ligand–protein complexes' stability since they showed at least one hydrogen bond with ChEMBL1779463. As shown in Fig.S4, RMSD values of these complexes showed less than 2Å, suggesting that the predicting crystallographic binding orientation was a good solution (Ha et al., 2024). The molecular dynamics simulation of ChEMBL1779463 with protein 3CL0 showed the number of HBs from 2 to 4 bonds in the range from 0 to 500 ns, suggesting that this ligand has proper interaction with several types of amino acids (Arg118, Glu119, Val149, Arg156, Glu227, Arg292, Tyr347, and Arg371) (Table S11 and Fig.S4). The complexes of ligand–protein 6TLZ exhibited HBs from 2 to 4 during the molecular dynamics simulation period of 0–400 ns, significantly reduced after the 400–500 ns. Some amino acids with the HBs interactions with ChEMBL1779463 were listed as Asn223, Arg226, Asn233, Leu630, and Thr632 (Table S11 and Fig.S4 D).

3.6 Proposed the total synthesis strategy and biological pathways

Even though NPs have substantive roles in the research and development of medicine, the limited plant sources encourage researchers to investigate the synthesis strategy. So, this study also suggests a scheme for the total synthesis of the second metabolites of natural resources. Among four candidates, FL22 (theaflavin) is a major agent from black tea, a commercial product with an antioxidant effect, while compounds FN166 and FN316had been successfully proposed for the total synthesis (Sagrera and Seoane, 2010). Therefore, this study focuses on the ChEMBL1779463 to propose a total-synthesis strategy using commercial and affordable reagents with some key steps, such as using a green procedure and protecting the phenol group using tetrafluoropyridyl (TFP) ether moiety attachment (Brittain and Cobb, 2019; Nishimura et al., 2010). First, 4-hydroxybenzaldehyde was reacted with horseradish peroxidase enzyme in the presence of hydrogen peroxide (3% in water) at pH∼4 to form a dimerization structure (1) (Fig. 6) (Nishimura et al., 2010). Intermediate compound (1), or 2′,4′-dihydroxyacetophenone, were mixed with pentafluoropyridine (PFP) and K₂CO₃ in the acetonitrile (MeCN) at room temperature for 16h to obtain readily cleavable protecting group for phenols (2) or (3), respectively (Brittain and Cobb, 2019). The intermediate compounds (2) and (3) were then linked together through the reaction with NaH in the tetrahydrofuran (THF) at room temperature for 24h (Sagrera and Seoane, 2010). Compound (4) was then created with acetic acid with the assistance of the reflux method for 96h to form the compound (5). Finally, the tetrafluoropyridyl (TFP) ether groups of compound (5) were removed to obtain ChEMBL1779463 following the procedure: mixing the intermediate compound with potassium fluoride, 18-crown-6, and methyl thioglycolate at 50°C for 24h (Sagrera and Seoane, 2010).

Close

Fig. 6

The proposed total synthesis of ChEMBL1779463. (a) Horseradish Peroxidase, H₂O₂, pH 4, 5–10 min. (b and c) PFP, K₂CO₃, MeCN, rt, 16 h. (d) NaH, THF, rt, 24 h. (e) AcOH, reflux, 96 h. (f) KF, 18-crown-6, methyl thioglycolate, CD₃CN, D₂O, 50 °C, 24 h.

Export to PPT

The noteworthy points of this study were the suggestion of the biological pathways of biflavanone, the most potential candidate for simultaneously inhibiting bifunctional NA protein and anti-M. pneumoniae. Biflavanone (1) inhibited the final step of influenza A viral life cycle (virion release); (2) could bind to the 30S subunit, leading to transcription mistakes and bacteria's protein synthesis interference (Maksimova et al., 2021); and (3) could interact with the sialic acid receptor P40/P90, preventing bacteria from attaching to host cells (Vizarraga et al., 2020) (Fig. 7).

Close

Fig. 7

The proposed biological pathways of biflavanone for inhibiting the final step (virion release) of influenza A virus via NA receptor (A), binding to the 30S subunit (B) and the sialic acid cell receptor P40/P90 (C) of M. pneumoniae.

Export to PPT