2.1 Sydowiol B and violaceol I are dual-targeting antivirals for SARS-CoV-2

Based on high-throughput screening, we identified two natural product hits, sydowiol B and its analogue violaceol I (Fig. 1A), from a subset of our in-house natural product library (approximately 2000 compounds). Both compounds exhibited dual inhibitory activity against the SARS-CoV-2 Mpro and PLpro. For Mpro and PLpro, the IC50 values of sydowiol B were 2.91 ± 1.09 μM and 5.77 ± 1.16 μM, respectively (Fig. 1B), while those of violaceol I were 8.27 ± 1.15 μM and 42.35 ± 1.02 μM, respectively (Supplementary Fig. S1). Notably, sydowiol B exhibited comparable activity to the reference compounds GC376 and GRL0617. We further confirmed the binding affinity between the two compounds and Mpro or PLpro. Sydowiol B and violaceol I both exhibited micromolar-scale interactions with the two proteases, comparable to GC376 and significantly better than GRL0617 (Fig. 1B, Supplementary Fig. S1).

Fig. 1

Sydowiol B and violaceol I are dual inhibitors of SARS-CoV-2 Mpro and PLpro. A Schematic representation of the high-throughput screening process that identified sydowiol B and violaceol I from an in-house natural product library (approximately 2000 compounds). The chemical structures of sydowiol B and violaceol I are shown. B Inhibitory activity (IC50 values) and binding affinity of sydowiol B against Mpro and PLpro. GC376 and GRL0617 were used as positive controls for Mpro and PLpro, respectively

2.2 Molecular docking of sydowiol B with SARS-CoV-2 Mpro

To elucidate the possible binding site and interactions involved, we performed molecular docking of sydowiol B with SARS-CoV-2 Mpro. Previous research has revealed multiple conformations of SARS-CoV-2 Mpro and the equilibrium among these conformations, as well as possible intermediate states. Additionally, substrate or inhibitor binding can induce structural changes in the protein [11,12,13,14]. Considering these factors that could affect docking results, we selected multiple PDB structures as docking models, including 5R80, 5RE4, 7ALH, 7K0G, 7LMD, 7LZT, 6WTM, 6XHU, 7B3E, 7BB2, 7BE7, 7BGP, 7C2Q, and 7C2Y. These structures covered both the apo and complex states of Mpro, as well as different conformational changes induced by diverse inhibitors. According to the docking results, sydowiol B might bind to SARS-CoV-2 Mpro at two different sites: one in the active site where substrates and inhibitors typically bind, and the other at the nano-channel located between the two subunits of the Mpro dimer (Fig. 2A).

Fig. 2

Molecular docking of sydowiol B with SARS-CoV-2 Mpro. A The surface of the structure 6WTM complexed with sydowiol B, highlighting two potential binding sites: the active site (blue circle) and the nano-channel (red circle). B The representative interaction and involved residues between sydowiol B and the active site of Mpro are illustrated using the PDB model 7LZT. The conformation of the ligand belongs to mono_group 2. C The representative interaction and involved residues between sydowiol B and the nano-channel of Mpro are illustrated using the PDB model 6WTM. The conformation of the ligand belongs to dimer_group 1. D The enzymatic activity and thermal stability (Tm value) of Mpro mutants (P values were calculated by comparing the mutants with WT). E The inhibition of sydowiol B against Mpro mutants derived from the nano-channel. F The effect of sydowiol B on the thermal stability (Tm value) of Mpro mutants derived from the nano-channel

By comparing the conformations of the ligand, we further subgrouped the docking results within the two binding sites. For the active site, the docking conformations of sydowiol B could be sorted into three subgroups based on their similarity: mono_group 1, mono_group 2, and mono_group 3 (Fig. 2B, Supplementary Fig. S2A-C). Between mono_group 1 and mono_group 2, four residues were common: Asn142, Gly143, Cys145, and Met165. The third subgroup, mono_group 3, shared limited overlap with the other two subgroups, involving only the catalytic dyad (His41 and Cys145). Combining this information, we selected five residues for further mutation analysis: His41, Asn142, Gly143, Cys145, and Met165 (Fig. 2B).

Similarly, for the nano-channel binding site, the docking conformations of sydowiol B could be sorted into two subgroups: dimer_group 1 and dimer_group 2 (Fig. 2C, Supplementary Fig. S2D-E). By comparing the overlaps between dimer_group 1 and dimer_group 2, six residues were selected for further analysis: Phe3, Arg4, Lys5, Tyr126, Glu288, and Phe291 (Fig. 2C).

In total, 11 residues were selected for mutation analysis. As expected, most mutants lost considerable enzymatic activity (Fig. 2D), since the activity of Mpro is regulated by numerous residues in the substrate-binding site and extra domains [15,16,17]. Only five mutants (R4A, N142A, M165A, E288A, and F291A) retained measurable enzymatic activity, with N142A displaying a slight increase in enzymatic activity (approximately 20%) compared to the wild-type (WT) enzyme. Moreover, these mutants exhibited variable thermal stability. R4A and N142A had comparable activity and similar thermal stability to WT, while F291A exhibited significantly lower activity than WT but had a slightly increased melting temperature (Tm) value. Despite their unmeasurable enzymatic activity, the remaining eight mutants displayed similar or lower thermal stability compared to WT. These results suggest that there is no clear correlation between thermal stability and enzymatic activity, at least in SARS-CoV-2 Mpro.

To verify the possible interaction between sydowiol B and Mpro, we measured the IC50 values of sydowiol B against Mpro mutants and Tm values of Mpro mutants with or without sydowiol B interference. Of the five mutants with measurable activity, N142A and M165A showed minor increases in the IC50 values of sydowiol B. In contrast, R4A, E288A, and F291A exhibited remarkably decreased inhibition from sydowiol B, with 4.52-fold, 6.19-fold, and 12.91-fold higher IC50 values compared to WT, respectively (Fig. 2E). In the thermal shift assay, four mutants from the nano-channel (R4A, K5A, E288A, and F291A) showed decreased ΔTm values (Fig. 2E, Supplementary Table S1), especially F291A, indicating impaired interaction of sydowiol B with the corresponding mutants and suggesting that these residues possibly participated in the interaction between sydowiol B and Mpro WT. Additionally, there was a notable increase in the ΔTm value of N142A, partly excluding the possibility of interaction between sydowiol B and Asn142. Combined with the results of the IC50 measurements, where sydowiol B exhibited nearly unchanged inhibition against mutants derived from the active site (N142A and M165A), we concluded that the binding site of sydowiol B with Mpro would be the nano-channel.

2.3 Sydowiol B bound to SARS-CoV-2 Mpro at the nano-channel

To inspect the binding situation of sydowiol B with SARS-CoV-2 Mpro in more detail, we conducted a 100 ns molecular dynamics (MD) simulation for both the apo and complex (com, Mpro complexed with sydowiol B) states, using the 6WTM model, which exhibited better structural integrity and quality as assessed by the PDB database. The corresponding docking result was used as the initial pose. The root-mean-square deviation (RMSD) plot revealed that after a short fluctuation (approximately 6 ns), the ligand sydowiol B was consistently held at the nano-channel, which could also be observed from the trajectory of the ligand after aligning Mpro (Fig. 3A). Additionally, we measured the distance between the ligand and three loops located in the nano-channel from different directions: residues 3–5 above the ligand, residues 284–286 beneath it, and residues 288–291 on the flank. The distances from all three directions showed that the ligand was held steady in the nano-channel (Supplementary Fig. S3A). Meanwhile, we discovered that the ligand progressively drifted away from chain A and approached chain B in all three directions. The radius of gyration (Rg) plot indicated that the structures of both apo and com remained compact during the MD simulation (Supplementary Fig. S3B). Further decomposition of Rg into three independent axes showed no significant difference in trends between apo and com (Supplementary Fig. S3C). The root-mean-square fluctuation (RMSF) plot and the distance between His41 and Cys145 revealed the asymmetry between the two chains of Mpro (Fig. 3B, Supplementary Fig. S3D), consistent with previous reports [18, 19]. It seemed that chain B adopted the "right conformation" necessary for catalysis, considering the distance between His41 and Cys145 [19], and sydowiol B significantly stabilized chain B with minimal influence on chain A (which showed only a momentary conformation transition after ligand binding but mostly remained in the unsuitable conformation). Particularly, residues around Asp48 (located at the S2 helix and forming the flank of the active site) and Tyr154 (located in a loop connecting the two β-sheets around the S1 and S4 loops) of chain B displayed significantly reduced fluctuation after sydowiol B binding.

Fig. 3

Sydowiol B bound to SARS-CoV-2 Mpro at the nano-channel. A The RMSD plot illustrates the stability of Mpro and the ligand sydowiol B in the apo (unbound) and com (complexed with sydowiol B) states, along with the trajectory of the ligand during the simulation after aligning the protein. B The RMSF plot represents the flexibility of residues in chain A and chain B of Mpro in the apo and com states. C This panel compares the size of the two active sites and the nano-channel in the docked conformation, apo state, and com state (the latter two represent the lowest energy conformations from principal component analysis): (i) the active site from chain A, (ii) the active site from chain B, (iii) the nano-channel, and (iv) a bar graph depicting the volume of these three sites in the respective states. D This panel illustrates the conformational changes induced by the binding of sydowiol B in the nano-channel, including the ligand itself and the surrounding residues

We then calculated the size of the active site by measuring the distances between the S1' loop (residues 21–26) and the S4 loop (residues 167–170), as well as between the S1 loop (residues 139–144) and the S2 helix (residues 46–50). During the simulation process, the distance between S1' and S4 remained stable in both chains and both states. However, the distance between S1 and S2 in both chains of com displayed a remarkable shrinkage compared to apo (Supplementary Fig. S3E). It has been reported that the active site of Mpro is malleable, especially the S2 helix, and upon ligand binding, it gradually expands to accommodate the side chains of the ligands [20,21,22]. In contrast, the changes observed upon sydowiol B binding caused the active site to contract and restricted its plasticity for substrate binding, which might result in the limitation of catalytic activity.

Additionally, we measured the size of the nano-channel. Unsurprisingly, the nano-channel exhibited considerable expansion in three different measurements, including the distances between residues 288–291 (chain A) and residues 288–291 (chain B) (this distance was positively correlated with the width of the nano-channel), between residues 3–5 (chain A) and residues 284–286 (chain B), and between residues 3–5 (chain B) and residues 284–286 (chain A) (the latter two distances were positively correlated with the length and height of the nano-channel, respectively) (Supplementary Fig. S3F).

We further conducted principal component analysis (PCA) on both states. The sizes of the active site and the nano-channel were calculated in the resulting lowest energy conformations from PCA (Supplementary Fig. S3G) and the docking conformation. The volumes of the two active sites from the docking conformation seemed identical, as it was a crystal structure of Mpro in the apo state. The two active sites of apo, which were relatively free in solution compared to the tightly compact docking conformation, tended to expand in both chains, while the binding of sydowiol B inhibited this progress, particularly in chain B (Fig. 3C i, 3C ii, and 3C iv). The measurement of active site volume also supported the asymmetry of the two chains in Mpro, with chain B adopting the "right conformation" and showing more flexibility in the active site, which might be related to substrate accommodation. Further inspection of several structural markers of protomer activity, including the salt bridge between Glu166 and His72 and the π-π stacking between Phe140 and His163, supported that chain B was the active protomer. Chain A in apo retained one of the two interactions (Glu166-His172) and collapsed to a certain extent, while both interactions of chain A in com were disrupted after sydowiol B binding (however, Glu166 was not oriented towards His163 to form an interaction; it just rotated away from both His172 and His163) (Supplementary Fig. S3K). The two structural markers of the docking conformation remained and were nearly identical, along with the previously mentioned two identical active site volumes, which were similar to a previously reported co-crystal [23]. Notably, the active site of com still maintained the two interactions (Glu166-His172 and Phe140-His163), the same as apo. It seemed that sydowiol B did not induce the collapse of active sites but limited their expansion and accommodation of substrates. On the other hand, the size of the nano-channel in apo seemed almost uniform with the docking conformation, but that in com displayed slight expansion, which might be a result of sydowiol B binding (Fig. 3C iv).

Gmx_MMPBSA decomposition analysis was conducted for both states between the two chains and between Mpro and the ligand sydowiol B. Compared to the apo state, the binding energy between the two chains was reduced upon sydowiol B binding (ΔGbind = 18 kcal/mol), particularly at Glu290 (chain A) and Arg4 (chain B), which formed a salt bridge crucial for maintaining the dimerization of Mpro. In contrast, the other pair of interactions between Glu290 (chain B) and Arg4 (chain A) seemed unaffected (Supplementary Fig. S3H, S3I). Sydowiol B formed a stable complex with Mpro, with a binding energy of -30.6 ± 0.09 kcal/mol (Supplementary Table S2), which was mainly contributed by residues from chain B, namely Phe3, Arg4, Lys5, Ser284, Glu288, and Phe291 (Supplementary Fig. S3J), consistent with experimental results. Further analysis of the apo and com conformations from PCA revealed the torsion of sydowiol B, which resulted in π-π stacking within the ligand (Fig. 3D i). Compared to the docking conformation, this torsion led to the loss of a hydrophobic interaction with Arg4 but an additional hydrogen bond with Lys5. The two hydrogen bonds with Lys5 in com were both inaccessible in apo (Fig. 3D ii). A similar situation was observed with Ser284, where the binding of the ligand pulled the residue inwards (Fig. 3D iii). The torsion of sydowiol B might cause a steric clash with Glu288, forcing it to rotate slightly more than in apo, resulting in the loss of one of the hydrogen bonds with Glu288 in the docking conformation (Fig. 3D iv).

To further understand the regulation of sydowiol B from the nano-channel to the active site, we used the webPSN (http://webpsn.hpc.unimore.it) webserver tool to investigate the structural communication within the sydowiol B-bound Mpro. As reported in the literature, there was an intense network present in Mpro (Supplementary Fig. S4A). When we filtered the network by the requirement "midway through the ligand", we found that sydowiol B regulated the active site of chain B perhaps through two directly interacted residues (Glu288 and Lys5) from chain B, which was consistent with the gmx_MMPBSA results, and a downstream pathway Glu290(B)-Arg4(A)-Tyr126(B)-Phe140(B)-His163(B) (Supplementary Fig. S4B). As for the active site of chain A, the allosteric modulation of sydowiol B was more complex, possibly through Arg4(B)-Lys137(A)-Arg131(A)-Asn133(A)-Ala194(A)-Phe185(A)-Phe181(A) and thereafter spreading. The directly interacted residue Arg4 (B) was still consistent with the gmx_MMPBSA results. Apart from the active site, sydowiol B also displayed regulation of residues from the dimer interface, such as Phe3(B), Arg4(A/B), Lys5(B), Ala7(A/B), Val125(B), Tyr126(B), Phe8(A), Lys137(A), and more, which might thus weaken the binding between the two chains (Supplementary Fig. S4A).

2.4 Molecular docking of sydowiol B with SARS-CoV-2 PLpro

Similar to the approach used for Mpro, we performed multiple molecular docking simulations of sydowiol B with SARS-CoV-2 PLpro using various PDB structure models. In contrast to the results of Mpro, the predicted binding sites of sydowiol B on PLpro were diverse and scattered. Consequently, we selected the two most promising sites for further analysis: the active site and allosteric site (Fig. 4A).

Fig. 4

Docking results of sydowiol B with multiple PDB models of SARS-CoV-2 PLpro. A The surface representation of the model 8G62 complexed with sydowiol B highlights the two binding sites: the active site (blue circle) and the allosteric site (red circle). B The representative interactions and involved residues between sydowiol B and the active site of PLpro are illustrated using the PDB model 8G62. C The representative interactions and involved residues between sydowiol B and the allosteric site of PLpro are depicted using the PDB model 7JIW. D The enzymatic activity and thermal stability (melting temperature, Tm) of PLpro mutants are shown (P-values were calculated by comparing with the wild-type enzyme). E The inhibitory effect of sydowiol B against PLpro mutants derived from the active site is presented. F The impact of sydowiol B on the thermal stability (Tm) of PLpro mutants derived from the active site is illustrated

Twelve out of the 25 docking results showed sydowiol B binding at the active site of PLpro, accounting for nearly half of the total results. Six residues (Gly163, Asp164, Arg166, Tyr264, Tyr268, and Tyr273) were consistently involved in the interaction across different docking models (Fig. 4B, Supplementary Fig. S5A). As Gly163 was predicted to interact with sydowiol B through its main chain carbonyl oxygen, it was excluded from further mutation analysis. The second binding site of sydowiol B on PLpro was located beneath the active site, resembling a U-shaped valley (Fig. 4A). Among the four docking results in this site, eight residues (Glu214, Tyr251, Glu252, Leu253, Lys254, Thr257, Phe258, and Val303) were consistently present and selected for further analysis (Fig. 4C, Supplementary Fig. S5B).

Among the 13 analyzed mutants, most exhibited impaired enzymatic activity, with Y264A, Y273A, L253A, F258A, and V303A displaying nearly unmeasurable activity levels. In contrast, the T257A mutant seemed unaffected, exhibiting a slightly increased activity (Fig. 4D). Regarding thermal stability, the majority of the mutants exhibited either unaffected or decreased stability compared to the wild-type enzyme. However, R166A, Y273A, and V303A showed a slight increase in Tm values, despite their compromised enzymatic activity. Notably, L253A and F258A exhibited a clear decrease in thermal stability, with their melting curves being unfittable (Fig. 4D).

The enzymatic activity assay results suggested that residues from both binding sites (i.e., Arg166 and Tyr251) might be involved in the interaction between sydowiol B and PLpro, as the corresponding mutants exhibited increased IC50 values (Fig. 4E, Supplementary Fig. S5C). However, these two mutants also displayed impaired enzymatic activity compared to the wild-type enzyme, which could partially contribute to the increased inhibition by the compound. To identify the authentic binding site, we employed a thermal shift assay. All five mutants derived from the substrate-binding site (D164A, R166A, Y264A, Y268A, and Y273A) showed decreased ΔTm values, with Y264A and Y273A exhibiting the most significant reductions (Fig. 4F, Supplementary Table S3). These results collectively support the notion that sydowiol B interacts with PLpro at the active site. Some mutants from the U-shaped binding site also exhibited slightly decreased ΔTm values (Supplementary Fig. S5D, Table S3), which might be attributed to overall structural effects induced by sydowiol B.

2.5 Sydowiol B bound to SARS-CoV-2 PLpro at the active site

To further confirm the interaction between sydowiol B and PLpro, we conducted a 100 ns MD simulation in the apo (unbound) and com (complexed with sydowiol B) forms, respectively. The model 8G62 was selected due to its better structural integrity and quality, as assessed by the Protein Data Bank (PDB). The corresponding docking result was used as the initial pose. The RMSD plot revealed that the protein reached equilibrium quickly (around 2 ns) in both apo and com forms. Similarly, after approximately 25 ns of fluctuation, the ligand sydowiol B also reached a steady state (Fig. 5A). The distance between sydowiol B and Tyr264 (approximately 2.5 Å) was well-sustained (Supplementary Fig. S6A), and the ligand trajectory showed that it remained inside the active site pocket throughout the entire simulation (Fig. 5A). Beginning from 25 ns, we calculated the RMSF of residues, which revealed that apart from the terminal residues, four loops displayed particularly pronounced fluctuations: two loops coordinating the structural Zn2+ (one loop centered on Cys189/Cys192, and the other centered on Cys224/Cys226), the BL2 loop (centered on Tyr268), and the loop centered on Glu280 (Fig. 5B). Furthermore, the former two loops exhibited more fluctuation when sydowiol B bound, while the other two loops seemed unaffected. The Rg plot of apo and com displayed steady and similar movements, confirming that the protein fold was well-sustained (Fig. 5C). However, when we decomposed the total Rg into three independent axes, we found that although the fluctuation level was similar between apo and com, apo displayed a more compact stacking in the X-axis but a looser stacking in the Z-axis compared to com (Supplementary Fig. S6B), which might result from the accommodation of the ligand in the X-axis and the inward contraction of the BL2 loop in the Z-axis in the com form. We further compared the size of the active pocket by measuring the distance between the BL2 loop (represented by Tyr268 and Gln269) and three residues (Leu162, Gly163, and Asp164) that formed a narrow tunnel leading to the active Cys111. In the apo form, the pocket preferred to adopt an open conformation, while in the com form, after a brief period of expansion (approximately 25 ns), it was well-sustained in a closed conformation (Supplementary Fig. S6C). However, this period of expansion did not result in pulling the ligand into the pocket (as can be seen from the distance between the ligand and residues 162–164, Supplementary Fig. S6A) but mainly arose from the conformational change of the ligand.

Fig. 5

Sydowiol B bound to SARS-CoV-2 PLpro at the active site. A The RMSD plot depicts the stability of PLpro and the ligand sydowiol B in the apo (unbound) and com (complex with sydowiol B) states, along with the trajectory of the ligand during the simulation after aligning the protein. B The RMSF plot represents the flexibility of residues of PLpro in the apo and com states. C The Rg plot illustrates the compactness of PLpro in the apo and com states. D This panel shows the conformational changes induced by the binding of sydowiol B in the active site, including the ligand itself and the surrounding residues. E The distances between the Zn2+ ion and its four coordinating cysteine residues (Cys189, Cys192, Cys224, and Cys226) are presented. F The distance between the Zn2+ ion and the S2 helix in the apo and com states is depicted. G The trajectory of the Zn2+ ion during the simulation after aligning PLpro in the apo and com states is shown. H This panel illustrates the conformational changes of residues from the S2 helix induced by the binding of sydowiol B

We conducted principal component analysis (PCA) and gmx_MMPBSA decomposition analysis to further inspect the detailed conformational changes induced by sydowiol B (Supplementary Fig. S6D). The gmx_MMPBSA analysis revealed that the binding of sydowiol B to PLpro was stable (ΔGibbs = -25.29 ± 0.07 kcal/mol, Supplementary Table S4), and the primary residues participating in the interaction included Leu162, Asp164, Ala246, Pro247, Tyr264, Asn267, Tyr268, and Thr301 (Supplementary Fig. S6E), which corresponded to the results of the thermal shift assay. The lowest energy conformation resulting from PCA was extracted and compared between apo and com, as well as the docking conformation. The PDB model (8G62) in the docking conformation was a co-crystal structure of PLpro with inhibitors, where the inhibitors adopted a relatively planar configuration. Compared to the docking pose, sydowiol B in com rotated approximately 90 degrees around the Y-axis, forming a π-π stacking between two benzene rings. The BL2 loop preferred a closed position in both the docking conformation and com. However, affected by the rotation of sydowiol B, the BL2 loop in com was forced to move away to avoid steric clash with the ligand, whereas in apo, it adopted an expanded conformation (Fig. 5D i). Compared to the docking conformation, the carbonyl of Asp164 flipped to sustain the hydrogen bond with the phenolic hydroxyl group of sydowiol B but from a different benzene ring, while the carboxyl group of the apo might induce a spatial clash with sydowiol B (Fig. 5D ii), reflecting the induced accommodation of the ligand. Leu162 of apo and com both flipped towards Cys111, consistent with previous reports about its role in blocking access to the catalytic Cys111. However, influenced by sydowiol B, the hydrophobic side chain of Leu162 moved slightly towards the methyl group of the ligand (Fig. 5D ii), similar to the main-chain carbonyl of Ala246, which formed a hydrogen bond with sydowiol B (Fig. 5D iii), and the side-chain benzene ring of Tyr264, which formed a π-π stacking with sydowiol B (Fig. 5D iv). Affected by the 90-degree rotation of sydowiol B, the side chain of Asn267 in com flipped towards the ligand to form an additional hydrogen bond, whereas in apo, it continued moving away from the docking conformation. The π-π stacking between sydowiol B and the side chain of Tyr268 was well-sustained in com (Fig. 5D v). The side chain of Thr301 between the docking conformation and com was practically unchanged, but considering the rotation of sydowiol B, it formed an additional hydrophobic interaction with the ligand, while in apo, it further moved away (Fig. 5D vi).

On the other hand, we found that the structural Zn2+ in com became unstable and could not settle in the zinc finger domain. Correspondingly, the Zn2+ in apo never moved outside the zinc finger, although the coordinate bonds between Zn2+ and Cys189/Cys192 were disrupted, and only the connection between Zn2+ and Cys224/Cys226 was sustained (Fig. 5E, 5G). In com, after approximately 36 ns of fluctuation, Zn2+ settled well in the S2 helix (Fig. 5F, 5G). After calculating the vacuum electrostatics of PLpro (in both apo and com forms), we found that the S2 helix displayed a significantly negatively charged surface (Supplementary Fig. S6H), which might be the reason for Zn2+ settling there. We further measured the distance between Zn2+ and individual residues of the S2 helix, and the negatively charged residues (Asp62, Glu67, and Glu70) and their surroundings exhibited more contacts and shorter distances than others (Supplementary Fig. S6G). Considering its role in binding ISG15 and di-ubiquitin, we further investigated the conformational changes in the S2 helix induced by sydowiol B and the resultant unstable Zn2+. First, the negatively charged residues (Asp62, Glu67, and Glu70) displayed visible rotation or flip towards Zn2+, accompanied by their adjacent residues (Fig. 5H i). It has been reported that some residues (Val66, Phe69, Glu70, His73, and Thr75) play an important role in mediating the interaction of PLpro with either ubiquitin or ISG15 [24, 25], and we inspected the influence of sydowiol B on these residues. There were also clear shifts in the positions of these residues compared to the ISG15-bound PLpro (Fig. 5H ii). Considering that sydowiol B already occupied the substrate-binding pocket, the resultant migration of Zn2+ and conformational changes of residues in the S2 helix might further create an unfavorable environment for ISG15 or ubiquitin binding.

To confirm that the migration of Zn2+ was induced by sydowiol B binding, we conducted a 100 ns MD simulation of PLpro complexed with GRL0617 (com2). It was revealed that similar to sydowiol B, GRL0617 also induced more fluctuation of the two loops around Cys189/Cys192 and Cys224/Cys226 compared to apo (Supplementary Fig. S7B). Differently, the structural Zn2+ was well settled in the zinc finger domain, similar to apo (Supplementary Fig. S7G, S7H), which proved the specific influence of sydowiol B on Zn2+. Another notable point was that GRL0617 was not well held in the pocket (Supplementary Fig. S7A). Even though the overall fold was well sustained during the simulation, the protein in com2 exhibited more fluctuation in three independent coordinate axes than apo or the sydowiol B-bound com (Supplementary Fig. S7C, S7D). The active pocket was expanded similarly to that in apo instead of being closed as in the sydowiol B-bound com (Supplementary Fig. S7E).

It has been reported that the zinc finger domain is important for structural stability and proteolytic activity [26]. Apart from competitively occupying the active site, the allosteric effect of sydowiol B on the zinc finger and subsequent S2 helix might further intensify inhibition against PLpro. We also investigated the structural communication inside PLpro after sydowiol B binding using the webPSN server. It can be seen that the global network inside PLpro was less intense, but the network induced by sydowiol B was more interlaced compared to Mpro (Supplementary Fig. S8A, S8B). The allosteric effect was mediated through residues in the palm domain and reached the loop around Cys224/Cys226 (namely Gln237) and the β-sheet around the Cys189/Cys192 loop (namely Arg183). The ligand also showed regulation in residues from the thumb domain and UBL domain, such as Leu64, Thr72, Val11, and His17.

2.6 The broad-spectrum antiviral activity of sydowiol B against homologous coronaviruses

To explore the broad-spectrum antiviral activity of sydowiol B, we tested its inhibition on Mpro and PLpro from three homologous coronaviruses: SARS-CoV, MERS-CoV, and BtCoV Rp3/2004 (also known as SARS-like coronavirus Rp3).

Compared to SARS-CoV-2 Mpro, sydowiol B showed similar inhibition on SARS-CoV Mpro and even better outcomes against BtCoV Rp3 Mpro in the enzymatic activity assay (Fig. 6A). Its inhibition on MERS-CoV Mpro was not verified since the activity was unmeasurable under experimental conditions (Fig. 6B). Additionally, sydowiol B showed a similar effect on the thermal stability of MERS-CoV Mpro and BtCoV Rp3 Mpro, but to a slightly lesser extent on SARS-CoV Mpro. Meanwhile, we found that SARS-CoV Mpro exhibited comparable enzymatic activity to SARS-CoV-2 Mpro, but BtCoV Rp3 Mpro harbored only 37% activity of SARS-CoV-2 Mpro (Fig. 6B), which has just one mutation at residue 170 (G170E) compared to SARS-CoV Mpro. Moreover, the thermal stability of BtCoV Rp3 Mpro also presented a notable decrease, from 55.92 °C of SARS-CoV Mpro to 45.75 °C of BtCoV Rp3 Mpro. The less homologous MERS-CoV Mpro displayed clear decreases in both enzymatic activity and thermal stability (Fig. 6B, Supplementary Table S5). In the surface plasmon resonance assay, sydowiol B displayed a micromolar-level affinity with the three homologous Mpro, comparable to GC376, with KD values of 2.54 μM for SARS-CoV, 9.35 μM for MERS-CoV, and 1.93 μM for BtCoV Rp3, respectively (Fig. 6E, Supplementary Table S6).

Fig. 6

The broad-spectrum antiviral activity of sydowiol B. A The effect of sydowiol B on the enzymatic activity and thermal stability of Mpro from homologous coronaviruses is depicted. B The enzymatic activity and thermal stability of Mpro from homologous coronaviruses are shown (P-values were calculated by comparing with SARS-CoV-2 Mpro). C The effect of sydowiol B on the enzymatic activity and thermal stability of PLpro from homologous coronaviruses is illustrated. D The enzymatic activity and thermal stability of PLpro from homologous coronaviruses are presented (P-values were calculated by comparing with SARS-CoV-2 PLpro). E The binding affinity of sydowiol B with Mpro and PLpro from homologous coronaviruses is shown

Regarding PLpro, sydowiol B exhibited more potent inhibition on PLpro from the three homologous coronaviruses compared to SARS-CoV-2 (Fig. 6C). Additionally, MERS-CoV PLpro showed clearly impaired activity compared to other coronaviruses, similar to Mpro, but with slightly increased thermal stability (Fig. 6D). Both SARS-CoV PLpro and BtCoV Rp3 PLpro showed increased activity and thermal stability compared to SARS-CoV-2 PLpro. Moreover, sydowiol B showed a similar influence on the thermal stability of SARS-CoV-2 PLpro and BtCoV Rp3 PLpro, but to a lesser extent on SARS-CoV PLpro and MERS-CoV PLpro (Fig. 6C, Supplementary Table S7). In the surface plasmon resonance assay, sydowiol B displayed a more potent affinity than GRL0617 against PLpro from the three coronaviruses. KD values were 14.2 μM for SARS-CoV PLpro, 0.45 μM for MERS-CoV PLpro, and 11.8 μM for BtCoV Rp3 PLpro, respectively (Fig. 6E, Supplementary Table S8).

Violaceol I also exhibited broad-spectrum antiviral activity against Mpro and PLpro from the three homologous coronaviruses, with comparable inhibition to sydowiol B and even more potent affinity (Supplementary Fig. S9, Tables S6 and S8).

2.7 The antiviral activity of sydowiol B and violaceol I in vitro

To further verify the antiviral activity of sydowiol B and violaceol I, we tested their inhibition on a homologous coronavirus, HCoV-OC43, in vitro. qRT-PCR analysis revealed that both compounds displayed potent inhibition against HCoV-OC43, with EC50 values of 0.69 ± 0.10 µM for sydowiol B and 2.38 ± 0.02 µM for violaceol I (Fig. 7A). Moreover, the two compounds significantly decreased the expression of HCoV-OC43 nucleocapsid (N) protein, as revealed by immunofluorescence assay (IFA) (Fig. 7B). Sydowiol B and violaceol I showed no influence on cell viability at concentrations up to 100 µM (Fig. 7A, Supplementary Table S9).

Fig. 7

The in vitro antiviral activity of sydowiol B and violaceol I. A The inhibitory effect of sydowiol B and violaceol I against the human coronavirus OC43 (HCoV-OC43) is demonstrated through quantitative reverse transcription-polymerase chain reaction (qRT-PCR) analysis. B The inhibition of HCoV-OC43 by these compounds is confirmed by immunofluorescence assay (IFA)