Magnetic Bead‐Based Ligand Fishing Identifies SARS‐CoV‐2 Main Protease Inhibitors From Reynoutria japonica

ABSTRACT

Traditional Chinese medicines (TCMs) are rich sources of structurally diverse bioactive compounds, yet rapid identification of protease inhibitors from complex botanical matrices remains challenging. Here, a magnetic bead‐based affinity‐fishing platform coupled with UPLC‐Q‐Exactive‐Orbitrap‐MS/MS was applied to identify SARS‐CoV‐2 main protease (M^pro) inhibitors from Reynoutria japonica. The crude ethanolic extract inhibited M^pro with an IC₅₀ of 44.92 ± 2.31 µg/mL. M^pro‐functionalized magnetic beads enriched eight constituents, which were identified by high‐resolution mass spectrometry and evaluated using a FRET‐based inhibition assay. Six compounds showed concentration‐dependent activity, with resveratrol being the most potent (IC₅₀ = 34.01 ± 1.67 µmol/L), followed by glycitein, genistein, emodin, apigenin‐7‐O‐glucoside, and biochanin A. HPLC‐DAD further quantified the principal active constituents. Molecular docking suggested interactions with residues in the Mpro substrate‐binding pocket, while network pharmacology predicted associations with host‐related targets involving IL6, TNF, and AKT1. Overall, this workflow enables efficient enrichment, structural identification, and functional prioritization of M^pro‐inhibitory constituents from complex botanical extracts, providing a practical strategy for target‐directed natural product discovery.

A magnetic bead‐based ligand fishing workflow using immobilized SARS‐CoV‐2 M^pro was coupled with UPLC‐Q‐Exactive‐Orbitrap‐MS/MS to screen potential M^pro‐binding constituents directly from a crude Reynoutria japonica extract. Eight compounds were captured and annotated, of which six showed concentration‐dependent M^pro inhibition, with resveratrol displaying the strongest activity. The workflow provides a target‐directed route from complex botanical extracts to prioritized enzyme‐inhibitory constituents.

1. Introduction

The coronavirus disease 2019 (COVID‐19) pandemic has reinforced the need to identify antiviral agents that act on well‐defined and conserved viral targets. Severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) translates its genomic RNA into the large polyproteins pp1a and pp1ab, which must be proteolytically processed to generate the nonstructural proteins required for viral replication and transcription [1, 2]. The SARS‐CoV‐2 main protease (M^pro), also known as the 3C‐like protease (3CL^pro) or nonstructural protein 5 (NSP5), mediates most of these cleavage events and is therefore indispensable for the viral life cycle [2, 3]. M^pro contains a conserved catalytic dyad formed by His41 and Cys145 and has no closely related human homologue with the same substrate specificity, making it an attractive target for selective antiviral intervention [3, 4, 5]. Structural and biochemical studies have established the druggability of the M^pro active site and enabled the discovery of inhibitors with cellular antiviral activity [4, 5, 6]. Nevertheless, continued exploration of chemically diverse Mpro‐binding scaffolds remains important for broadening the range of potential inhibitors and for developing tools with which to investigate protease–ligand interactions.

Medicinal plants represent a rich source of structurally diverse natural products, including stilbenes, anthraquinones, flavonoids, phenolic glycosides, and terpenoids. Reynoutria japonica Houtt., also known by the synonyms Polygonum cuspidatum Siebold and Zucc. and Fallopia japonica (Houtt.) Ronse Decr., is the botanical source of the traditional Chinese medicinal material Polygoni Cuspidati Rhizoma et Radix. The herb has traditionally been used to clear heat and resolve toxicity, and it is included as one component of several multi‐herb preparations used for acute respiratory tract infections [7]. Phytochemical investigations have revealed a complex composition dominated by stilbenes, anthraquinones, flavonoids, and related phenolic constituents, among which resveratrol, polydatin, emodin, and their derivatives are frequently reported [8]. Previous work evaluated Reynoutria rhizome extracts, fractions, and selected constituents using molecular docking and an in vitro fluorescence‐based M^pro inhibition assay [9]. That study showed that Reynoutria‐derived materials contain compounds capable of inhibiting M^pro, although activity varied substantially among extracts, fractions, and individual constituents. Natural‐product screening has also yielded experimentally characterized M^pro inhibitors from other botanical sources, further supporting medicinal plants as useful reservoirs of target‐directed chemical diversity [10]. However, the chemical complexity of R. japonica means that additional M^pro‐binding constituents may remain unresolved in unfractionated extracts.

Conventional bioactivity‐guided isolation commonly requires repeated extraction, fractionation, activity testing, and purification, making it labor‐intensive and potentially unfavorable for the detection of low‐abundance constituents. Affinity‐based ligand fishing provides a complementary target‐directed strategy by using immobilized biomacromolecules to enrich candidate ligands directly from complex mixtures [11]. Magnetic beads are particularly suitable supports because they allow rapid separation and convenient washing and elution. Protein‐functionalized magnetic beads have been successfully coupled with LC–MS in both offline and online affinity‐selection workflows [12, 13], while oriented immobilization strategies have been developed to improve target accessibility and protein functionality [14].

Magnetic bead‐based ligand fishing has increasingly been applied to botanical extracts. Immobilized enzymes have been used to capture candidate ligands from Houttuynia cordata, Hunteria zeylanica, and other medicinal plants, followed by high‐resolution mass spectrometric characterization and enzymatic validation [15, 16, 17]. More directly related to the present study, our group previously developed an NSP5/M^pro‐functionalized magnetic bead ligand‐fishing method coupled with UPLC‐Q‐Exactive‐Orbitrap‐MS and demonstrated its application to Rosmarinus officinalis [18]. Building on that work, the present study applies the platform to R. japonica, a chemically distinct botanical matrix, and further integrates HPLC‐DAD quantification, authentic‐standard comparison, and FRET‐based M^pro inhibition assays.

The high resolving power and mass accuracy of Q‐Exactive Orbitrap instruments facilitate molecular‐formula assignment and MS/MS‐based annotation of constituents in complex extracts [17, 18, 19]. However, affinity capture alone does not establish functional inhibition, because retained compounds may bind weakly, interact with non‐catalytic regions, or undergo residual nonspecific adsorption. Enzymatic validation is therefore required. Molecular docking and network pharmacology can provide additional mechanistic hypotheses, but their predictions should be interpreted cautiously and require experimental confirmation [20].

In this study, we applied and further evaluated an M^pro‐functionalized magnetic bead ligand‐fishing workflow coupled with UPLC‐Q‐Exactive‐Orbitrap‐MS/MS to screen potential M^pro‐binding constituents from a 75% ethanol extract of R. japonica. The captured compounds were structurally annotated, quantified by HPLC‐DAD, and evaluated using an in vitro FRET‐based M^pro inhibition assay. Molecular docking and network pharmacology were further used to explore potential binding interactions and broader target–pathway associations.

2. Results and Discussion

2.1. Preliminary Screening of 20 Chinese Medicinal Herb Extracts for M^pro Inhibition

To identify botanical extracts with inhibitory activity against SARS‐CoV‐2 M^pro, an initial in vitro screening was performed using twenty traditional Chinese medicinal herbs. As shown in Table 1, the ethanolic extract of R. japonica exhibited the lowest IC₅₀ value among the tested extracts (44.92 ± 2.31 µg/mL; Figure 5A). Astragalus membranaceus (58.34 ± 11.11 µg/mL) and Crataegus pinnatifida (84.17 ± 2.98 µg/mL) also showed measurable inhibitory activity. Previous studies have reported potential antiviral activities of phytochemicals associated with these herbs, including astragalin and flavonoid constituents from C. pinnatifida [21, 22].

TABLE 1.

Inhibitory activity of herbal extracts against M^pro.

Herb	Extracting solvent	IC50 (µg/mL)
Reynoutria japonica	75%EtOH	44.92 ± 2.31
Astragalus membranaceus	95%EtOH	58.34 ± 11.11
Hydnocarpus anthelmintica	95%EtOH	83.28 ± 4.33
Crataegus pinnatifida	75%EtOH	84.17 ± 2.98
Eclipta prostrata	75%EtOH	89.16 ± 6.35
Agrimonia pilosa	70%EtOH	102.10 ± 6.69
Ilex cornuta	90%EtOH	116.10 ± 18.83
Amomum villosum	80%EtOH	118.50 ± 5.51
Gardenia jasminoides	70%EtOH	170.20 ± 13.06
Nelumbinis plumula	95%EtOH	193.70 ± 11.84
Glechoma longituba	75%EtOH	224.70 ± 18.67
Scleromitrion diffusum	80%EtOH	233.40 ± 18.06
Andrographis paniculata	70%EtOH	433.40 ± 45.00
Centipeda minima	70%EtOH	475.30 ± 18.16
Patrinia scabiosaefolia	75%EtOH	508.10 ± 47.76
Lycium chinense	75%EtOH	729.10 ± 91.07
Phragmites rhizoma	95%EtOH	> 1000
Glehnia littoralis	95%EtOH	> 1000
Codonopsis pilosula	95%EtOH	> 1000
Actaea cimicifuga	95%EtOH	> 1000

FIGURE 5.

Concentration‐dependent inhibitory effects of the R. japonica extract and selected compounds against SARS‐CoV‐2 M^pro. (A) Inhibition curve and calculated IC₅₀ value of the 75% ethanol extract of R. japonica. (B) Inhibition curve and calculated IC₅₀ value of the positive control baicalein. (C–H) Inhibition curves and calculated IC₅₀ values of resveratrol, glycitein, genistein, emodin, biochanin A, and apigenin‐7‐O‐glucoside, respectively. Data are presented as the mean ± SD (n = 3).

Among the twenty extracts tested, R. japonica showed the strongest inhibition of recombinant Mpro under the present assay conditions and was therefore selected for further investigation. Previous studies have reported antiviral activities for several constituents of R. japonica. For example, resveratrol has been shown to inhibit norovirus replication [23], while reynoudiol has been reported to exhibit activity against oseltamivir‐resistant influenza virus strains [24]. These findings, together with the Mpro inhibitory activity observed in the present screening, supported further investigation of the M^pro‐targeting constituents of R. japonica.

2.2. Ligand Fishing Validation via HPLC Profiling

The performance of the magnetic bead‐based ligand fishing assay was evaluated using a synthetic mixture containing compounds with known differences in M^pro‐binding activity, followed by HPLC analysis (Figure 1). The three successive wash fractions (W ₁–W ₃) showed a progressive decrease in chromatographic signals, indicating the gradual removal of unbound and weakly or nonspecifically retained compounds. In the final wash fraction (W ₃), most signals were close to the chromatographic baseline, suggesting that the majority of the unretained components had been removed.

FIGURE 1.

HPLC evaluation of ligand capture by the magnetic bead‐based ligand fishing assay. The chromatograms show the progressive removal of unbound and weakly retained compounds during the washing steps (W ₁–W ₃), followed by enrichment of ampelopsin (A) and baicalein (C) in the elution fractions (E ₁–E ₄). In contrast, only minimal retention of the low‐affinity control compound coumarin (B) was observed. The chromatograms are representative of three independent ligand fishing experiments (n = 3).

Following four successive elution cycles with 70% (v/v) aqueous methanol (E ₁–E ₄), pronounced peaks corresponding to ampelopsin (Peak A) and baicalein (Peak C), which were used as positive M^pro‐binding controls, were observed in the elution fractions. In contrast, coumarin (Peak B), which was included as a low‐affinity negative control, showed only a weak signal in these fractions. The differential retention and elution profiles support the ability of the magnetic bead‐based assay to preferentially retain M^pro‐binding compounds over a compound with low binding affinity in a multicomponent mixture.

2.3. Optimal Conditions for Fishing Potential Inhibitors of M^pro

Key experimental parameters of the ligand‐fishing assay were evaluated using baicalein as the reference ligand (Figure 2). The HPLC peak area of eluted baicalein was highest when an M^pro concentration of 1.0 mg/mL and an immobilization time of 30 min were used. For the ligand‐binding step, the highest baicalein response was observed after 5 min of incubation, whereas longer incubation times resulted in lower peak areas. Evaluation of the methanol concentration showed that 70% methanol produced the highest baicalein signal among the tested elution conditions. Based on these results, an M^pro concentration of 1.0 mg/mL, an immobilization time of 30 min, a 5 min binding time, and 70% methanol were selected as the preferred conditions for the reference‐ligand optimization assay.

FIGURE 2.

Optimization of key experimental parameters for the magnetic bead‐based ligand fishing assay based on the HPLC peak area of eluted baicalein. (A) Effect of M^pro protein concentration; (B) effect of M^pro immobilization time; (C) effect of baicalein binding time; (D) effect of methanol concentration used for elution. Data are presented as the mean  ±  SD from three independent experiments (n = 3). Differences among conditions were analyzed by one‐way ANOVA followed by Tukey's multiple‐comparisons test, with α = 0.05.

2.4. Sequential Validation and Structural Annotation of M^pro‐Targeted Ligands via HPLC and UPLC‐Q‐Exactive‐Orbitrap‐MS/MS

To systematically profile the captured ligands, HPLC analysis was performed on the initial R. japonica extract solution (P ₀) incubated with the M^pro‐immobilized magnetic beads, the subsequent washing solutions (W ₁–W ₃), and the final 70% methanol eluates (E ₁–E ₄). As depicted in Figure 3, the post‐wash chromatograms revealed a progressive and substantial reduction in background signals, with the final wash (W ₃) demonstrating negligible analyte presence. This confirmed the effective clearance of the complex, unbound botanical matrix. However, upon elution with 70% aqueous methanol, several enriched peaks emerged in the E ₁–E ₄ fractions. The recovery of these peaks was consistent with preferential retention by the M^pro‐functionalized magnetic bead system; however, residual nonspecific retention cannot be completely excluded.

FIGURE 3.

HPLC chromatograms of compounds recovered from R. japonica extract using the M^pro‐immobilized magnetic bead ligand‐fishing system. The chromatographic profiles show the original extract before incubation with the beads (P ₀), the progressive removal of unbound and weakly or nonspecifically retained components during the washing steps (W ₁–W ₃), and the enrichment of retained compounds in the successive 70% aqueous methanol eluates (E ₁–E ₄). The chromatograms are representative of three independent ligand‐fishing experiments (n = 3).

To characterize these enriched candidate M^pro‐binding constituents, the four 70% aqueous methanol eluates were pooled and subjected to high‐resolution LC–MS/MS analysis. Through comprehensive analysis of the base peak ion chromatograms (Figures S1–S2) and corresponding MS/MS fragment ions (Figures S7–S14), eight primary compounds (C‐1 to C‐8) were structurally annotated (Figure 4, Table 2).

FIGURE 4.

Chemical structures of the eight annotated M^pro‐binding compounds (C‐1 to C‐8) captured from the R. japonica extract.

TABLE 2.

UPLC‐Q‐Exactive‐Orbitrap‐MS/MS information of compounds recovered from the M^pro ligand‐fishing eluates of R. japonica extract.

No.	RT (min)	Ion mode	Fragment ions	MW (ppm)	Formula	Identification
C‐1	32.218	[M−H]−	269.0456,431.0984,225.0559,311.5502	432.1056 (0.0002)	C21H20O10	Apigenin‐7‐O‐glucoside
C‐2	15.165	[M+H]+	229.0857, 107.0491, 135.0440, 91.0542, 119.0490, 165.0696, 183.0807, 211.0752	228.0786 (0.07)	C14H12O3	Resveratrol
C‐3	44.398	[M−H]−	283.0613, 240.0429, 268.0377, 212.0480	284.0685 (0.61)	C16H12O5	Glycitein
C‐4	32.222	[M−H]−	269.0456, 225.0557, 241.0508	270.0528 (−0.61)	C15H10O5	Genistein
C‐5	35.960	[M+H]+	285.0754, 242.0572, 211.0753, 253.0492	284.0685 (−0.18)	C16H12O5	Biochanin A
C‐6	36.659	[M+H]+	271.0598, 229.0493, 197.0598	270.0528 (−0.15)	C15H10O5	Emodin
C‐7	21.797	[M+H]+	265.1219, 296.1641, 250.0984, 235.0750, 219.0805	295.1572 (−0.26)	C19H21NO2	Nuciferine
C‐8	36.39	[M−H]−	221.1548, 236.1057, 293.1763, 193.1598, 136.0893, 177.0921, 89.0244	294.1831 (0.43)	C17H26O4	6‐Gingerol

As shown in Table 2, UPLC‐Q‐Exactive‐Orbitrap‐MS/MS analysis enabled the annotation of eight structurally diverse compounds recovered from the M^pro ligand‐fishing eluates. Polyphenolic and flavonoid structures were predominant among the annotated compounds. Compounds C‐1, C‐3, C‐4, and C‐5 were assigned as flavonoids or flavonoid glycosides. Compound C‐1, with a retention time (RT) of 32.218 min and a molecular weight of 432.1056, was annotated as apigenin‐7‐O‐glucoside. Glycitein (C‐3, RT = 44.398 min) and biochanin A (C‐5, RT = 35.960 min) had the same molecular weight of 284.0685 but exhibited different chromatographic retention and fragmentation patterns. Genistein (C‐4, RT = 32.222 min) was annotated with a molecular weight of 270.0528. Their characteristic fragment ions supported the assignment of the corresponding flavonoid and isoflavone structures.

Among the other phenolic compounds, C‐2 (RT = 15.165 min; molecular weight, 228.0786) was annotated as the stilbene derivative resveratrol, whereas C‐8 (RT = 36.390 min; molecular weight, 294.1831) was assigned as 6‐gingerol. Structurally distinct constituents were also recovered from the ligand‐fishing eluates. Compound C‐6 (RT = 36.659 min; molecular weight, 270.0528) was annotated as the anthraquinone emodin, while C‐7 (RT = 21.797 min; molecular weight, 295.1572) was assigned as the alkaloid nuciferine. Collectively, the high‐resolution mass spectrometric data revealed the structural diversity of the compounds recovered from the R. japonica extract. Their specific interactions with M^pro and functional effects, however, required further evaluation using the subsequent enzymatic inhibition assays.

2.5. In Vitro Assessment of M^pro Inhibitory Activity

The inhibitory activities of the 75% ethanol extract of R. japonica and the compounds identified by ligand fishing were evaluated using an in vitro FRET‐based enzymatic assay. The crude R. japonica extract exhibited concentration‐dependent inhibition of SARS‐CoV‐2 M^pro, with an IC₅₀ value of 44.92 ± 2.31 µg/mL (Figure 5A). Under the same assay conditions, the positive control baicalein showed an IC₅₀ value of 85.53 ± 2.55 µmol/L (Figure 5B).

Following the structural identification of the eight captured compounds, their inhibitory activities against M^pro were individually evaluated. Nuciferine (C‐7) and 6‐gingerol (C‐8) exhibited negligible inhibition within the tested concentration range, and reliable IC₅₀ values could therefore not be determined. The potential M^pro‐independent antiviral relevance of nuciferine is discussed in Section 2.9. In contrast, the other six captured compounds showed concentration‐dependent inhibition of M^pro (Figure 5C–H).

Among these compounds, resveratrol (C‐2) exhibited the lowest IC₅₀ value of 34.01 ± 1.67 µmol/L (Figure 5C), followed by glycitein (C‐3), with an IC₅₀ value of 57.18 ± 1.73 µmol/L (Figure 5D). Genistein (C‐4) and emodin (C‐6) showed IC₅₀ values of 88.56 ± 2.87 and 123.20 ± 3.78 µmol/L, respectively (Figure 5E,F). Biochanin A (C‐5) and apigenin‐7‐O‐glucoside (C‐1) exhibited comparatively weaker inhibition, with IC₅₀ values of 226.60 ± 6.22 and 221.00 ± 6.33 µmol/L, respectively (Figure 5G,H).

The differences in inhibitory activity may be associated with variations in the chemical scaffolds and substitution patterns of these compounds. However, the limited number of compounds examined does not permit a definitive structure–activity relationship to be established. Overall, six compounds recovered by the ligand‐fishing procedure inhibited recombinant M^pro under the present in vitro assay conditions, supporting the utility of the workflow for prioritizing target‐binding constituents from the crude extract.

2.6. Quantitative Analysis of Captured Compounds by HPLC

To correlate the in vitro inhibitory potency of these six active compounds with their natural abundance, quantitative HPLC analysis was performed on the crude R. japonica extract (20 mg/mL). Quantitative calibration curves for the six reference compounds (emodin, resveratrol, apigenin‐7‐O‐glucoside, genistein, glycitein, and biochanin A) were established, demonstrating excellent linearity (R ² > 0.997) across their respective evaluated concentration ranges (Figure S3).

Quantitative analysis revealed that the anthraquinone emodin (C‐6) dominated the 20 mg/mL R. japonica extract at a concentration of 416.19 µg/mL (representing 2.08% w/w of the dry extract). This was followed by resveratrol (C‐2, 194.01 µg/mL, 0.97% w/w) and apigenin‐7‐O‐glucoside (C‐1, 107.09 µg/mL, 0.54% w/w). Minor constituents included genistein (7.78 µg/mL, 0.0389% w/w), glycitein (5.74 µg/mL, 0.0287% w/w), and biochanin A (4.60 µg/mL, 0.0230% w/w).

These quantitative findings suggest that resveratrol and emodin may contribute to the overall M^pro‐inhibitory activity of the R. japonica extract because of their relatively high abundance and measurable inhibitory activity. Notably, low‐abundance compounds such as genistein and glycitein, which accounted for less than 0.04% of the extract, were successfully enriched and showed measurable concentration‐dependent inhibitory activity. This finding illustrates the ability of the ligand‐fishing approach to enrich low‐abundance constituents for subsequent functional evaluation.

2.7. Molecular Docking

To explore the possible structural basis of the observed M^pro inhibition, molecular docking was performed for three of the more active compounds, namely resveratrol, glycitein, and genistein, using the SARS‐CoV‐2 M^pro crystal structure (Figure 6). The predicted binding poses placed all three compounds within or adjacent to the substrate‐binding region of M^pro, where they formed hydrogen bonds and additional nonbonded contacts with residues surrounding the catalytic pocket.

FIGURE 6.

Predicted binding poses of resveratrol, glycitein, and genistein within SARS‐CoV‐2 M^pro obtained by molecular docking. (A) Overall view of the ligand‐binding region of Mpro. Enlarged views of the predicted binding poses of (B) resveratrol, (C) glycitein, and (D) genistein. Selected residues surrounding the docked ligands are labeled. Detailed two‐dimensional interaction maps are shown in Figures S4–S6.

Resveratrol adopted a predicted binding pose stabilized by four hydrogen bonds involving Gly143 (2.70 Å), Ser144 (2.78 Å), Leu141 (2.97 Å), and Met49 (3.22 Å) (Figures 6B and S4). Additional nonbonded contacts were observed with residues including His163, Glu166, Met165, Cys145, Asp187, Arg188, and Gln189. In this predicted pose, the aromatic scaffold of resveratrol was positioned close to the catalytic region containing Cys145 and His41, suggesting a possible steric contribution to interference with substrate access. However, because molecular docking provides a static computational model, this binding orientation should be regarded as a structural hypothesis rather than a demonstrated mechanism of inhibition. Glycitein was predicted to form three hydrogen bonds with Leu141 (3.07 Å), Glu166 (2.55 Å), and Thr190 (2.96 Å) (Figures 6C and S5). The docked complex also showed additional contacts with residues surrounding the binding pocket, including His163, His172, Phe140, Gln189, Cys145, Met165, and Arg188. These interactions positioned the isoflavone scaffold within the M^pro binding cavity and may contribute to stabilization of the predicted complex.

Genistein was predicted to form hydrogen bonds with Cys145 (3.07 Å), Tyr54 (2.60 Å), and Asp187 (2.65 Å) (Figures 6D and S6). Additional nonbonded contacts involved Leu141, Glu166, Met165, His41, Met49, His164, and Arg188. Notably, the predicted interaction with Cys145 places genistein in close proximity to the catalytic region of M^pro, although the functional significance of this interaction requires experimental confirmation.

Overall, the docking results indicate that the three compounds can adopt poses involving multiple hydrogen bonds and contacts with residues surrounding the M^pro substrate‐binding pocket. Differences in scaffold geometry and hydroxylation pattern may contribute to their distinct predicted interaction profiles and inhibitory activities. Nevertheless, the limited number of compounds examined does not permit a definitive structure–activity relationship to be established, and the docking results should be interpreted as supportive computational evidence rather than proof of a specific binding mechanism.

2.8. Network Pharmacology and Pathway Enrichment Analysis

In addition to the M^pro inhibitory activity observed for six of the captured compounds, COVID‐19 involves complex host inflammatory and immune responses. Network pharmacology was therefore used as a hypothesis‐generating approach to explore the potential host target and pathway associations of the eight compounds recovered by ligand fishing.

Target prediction identified 842 putative human targets for the eight compounds. Comparison with 400 COVID‐19‐associated targets retrieved from the GeneCards database yielded 89 overlapping targets (Figure 7A). Construction of the compound–target–disease network showed that resveratrol and genistein were associated with the largest numbers of overlapping targets, with 51 and 41 predicted target connections, respectively. Protein–protein interaction (PPI) analysis of the 89 overlapping targets further identified several highly connected nodes within the network (Figure 7B). Among these, IL6 and TNF were associated with inflammatory and immune‐response pathways, whereas AKT1 and VEGFA were linked to signaling processes relevant to vascular and cellular responses. TLR4 and NLRP3 were also included in the predicted network, suggesting possible associations with innate immune and inflammatory signaling. These associations are based solely on computational prediction, and no gene‐expression, cytokine, or pathway‐level experiments were performed in the present study to validate them. Accordingly, the network results should be interpreted as hypotheses regarding the broader biological context of the identified compounds rather than as evidence of direct modulation of these targets or pathways.

FIGURE 7.

Network pharmacology and functional enrichment analyses of the eight compounds recovered by ligand fishing. (A) Compound‐target network of the eight captured compounds constructed using Cytoscape. (B) Protein–Protein Interaction (PPI) network of the core targets in treating COVID‐19. (C) Top Gene Ontology (GO) terms enriched among the overlapping targets. (D) Top 17 KEGG pathways enriched among the overlapping targets.

Gene Ontology (GO) enrichment analysis further characterized the predicted biological functions of the overlapping targets (Figure 7C). Enriched cellular‐component terms were primarily associated with the cell surface and extracellular region, while molecular‐function terms included cytokine activity and receptor binding. The enriched biological‐process terms were mainly related to inflammatory responses and regulation of gene expression. These results suggest that the predicted targets of the identified compounds are associated with extracellular signaling and immune‐ and inflammation‐related biological processes, although these associations remain to be experimentally validated.

KEGG pathway enrichment analysis identified several pathways potentially relevant to COVID‐19 and host inflammatory responses (Figure 7D). Among the enriched pathways were “Coronavirus disease–COVID‐19”, “Toll‐like receptor signaling pathway”, “Cytokine–cytokine receptor interaction”, and “IL‐17 signaling pathway”, indicating potential associations with antiviral and inflammatory signaling. In addition, “Lipid and atherosclerosis” and “AGE–RAGE signaling pathway in diabetic complications” were also enriched, suggesting possible links between the predicted targets and pathways relevant to COVID‐19‐associated metabolic and cardiovascular complications. Collectively, these network‐based results generate the hypothesis that some R. japonica constituents may influence host immune and inflammatory pathways in addition to the M^pro inhibitory activity observed for selected compounds. However, these predicted host‐directed effects require confirmation in cellular and in vivo models.

2.9. Discussion

The integration of magnetic bead‐based ligand fishing with LC–MS/MS and HPLC enabled the systematic identification and quantification of potential M^pro‐binding constituents from R. japonica. Previous studies have reported the potential antiviral activities of Reynoutria species and, more specifically, the in vitro M^pro‐inhibitory effects of R. japonica rhizome extracts and constituents such as resveratrol and emodin [9, 25, 26]. Our quantitative results are consistent with these previous findings and indicate that resveratrol and emodin are relatively abundant M^pro‐inhibitory constituents in the extract. In contrast to conventional fraction‐by‐fraction screening, our affinity‐based platform facilitated the direct, target‐driven capture of potential ligands from the complex ethanolic extract. This targeted approach led to the identification and functional evaluation of several additional flavonoids, including genistein, glycitein, and apigenin‐7‐O‐glucoside, as in vitro M^pro inhibitors, thereby expanding the range of R. japonica‐derived constituents shown to inhibit M^pro under the present assay conditions.

To further contextualize these in vitro results, molecular docking and network pharmacology analyses were performed as hypothesis‐generating computational approaches. Molecular docking suggested that the identified compounds may adopt different binding poses within M^pro, potentially reflecting differences in their structural scaffolds. For instance, resveratrol was predicted to occupy the substrate‐binding cleft and form hydrogen‐bond interactions with Gly143, Ser144, and Leu141, whereas genistein was predicted to interact with Cys145 and Tyr54. However, these predicted poses and residue contacts do not confirm direct binding, binding stability, or a definitive mechanism of inhibition. No molecular dynamics simulations, binding free‐energy calculations, or orthogonal biophysical measurements were performed to validate the docking predictions. Similarly, network pharmacology identified IL6, TNF, AKT1, and other host‐related proteins as predicted central targets associated with inflammatory and disease‐related pathways. Because no gene expression analysis, cytokine measurement, target‐validation assay, or biological pathway confirmation was conducted, these network pharmacology results should be interpreted solely as computational predictions that generate hypotheses for future experimental investigation.

Despite these insights, the current screening paradigm presents certain inherent limitations. Compounds such as nuciferine and 6‐gingerol were enriched in the ligand‐fishing eluates but exhibited negligible inhibitory activity in the in vitro M^pro assay. This discrepancy does not, by itself, confirm direct or specific binding to M^pro. Possible explanations include weak or non‐inhibitory interactions with the protein, interactions at non‐catalytic or surface sites, or residual nonspecific retention by the immobilized protein, magnetic beads, or linker chemistry. The present data cannot distinguish among these possibilities. Moreover, no orthogonal biophysical methods, such as surface plasmon resonance (SPR), isothermal titration calorimetry (ITC), microscale thermophoresis (MST), or differential scanning fluorimetry (DSF), were performed. Therefore, the binding affinities, kinetics, thermodynamic parameters, and specific binding sites of the enriched compounds remain unconfirmed. Future studies incorporating these complementary biophysical techniques will be required to determine whether these compounds interact directly and specifically with M^pro.

Furthermore, although LC–MS/MS provides high‐resolution structural annotation, definitive identification of unknown constituents generally requires comparison with authentic standards or complementary spectroscopic evidence. In the present study, authentic reference standards were evaluated for the prioritized active compounds. Agreement in HPLC RTs supported their structural assignments, while their dose‐dependent inhibition in the FRET assay independently confirmed their functional activity against recombinant M^pro. Future studies incorporating nuclear magnetic resonance spectroscopy will be necessary for the definitive structural elucidation of constituents for which authentic standards are unavailable.

A fundamental limitation of this study is that the biological activities of the identified compounds were evaluated only using an isolated recombinant M^pro enzyme assay. No SARS‐CoV‐2‐infected cell‐based antiviral assay or cytotoxicity evaluation was performed. Therefore, the cellular permeability, antiviral efficacy, cytotoxicity, and selectivity indices of these compounds remain unknown. On the basis of the present evidence, the identified compounds should be regarded as in vitro M^pro inhibitors rather than confirmed antiviral agents. In addition, the reported IC₅₀ values were obtained under a single set of assay conditions, and the modes of inhibition and potential assay‐interference mechanisms were not systematically investigated. Cell‐based antiviral and cytotoxicity evaluations, together with mechanistic and orthogonal binding studies, are therefore required to determine whether the observed enzyme inhibition can be translated into cellular antiviral activity.

The inhibitory potency of the constituents identified in this study should be considered within a realistic pharmacological context. Resveratrol, which showed the lowest IC50 value among the compounds recovered from the ligand‐fishing eluates, inhibited recombinant M^pro with an IC₅₀ value of 34.01 ± 1.67 µmol/L. In comparison, nirmatrelvir, and ensitrelvir have been reported to inhibit SARS‐CoV‐2 M^pro with a K _i value of 3.11 nM and an IC50 value of approximately 13 nM, respectively [27, 28]. GC376 has also shown substantially greater potency, with a reported IC₅₀ value of 0.03 µmol/L [29], while the natural flavonoid baicalein has exhibited submicromolar to low‐micromolar inhibition of M^pro in previous biochemical assays [30]. Although direct numerical comparisons across studies should be interpreted cautiously because of differences in enzyme constructs, substrate concentrations, incubation conditions, assay procedures, and the use of both K _i and IC₅₀ measurements, the available data indicate that resveratrol and the other constituents identified here are considerably less potent than established M^pro inhibitors. These compounds should therefore be regarded as weak, preliminary enzyme‐inhibitory hits rather than drug candidates or confirmed antiviral agents. Their principal significance in the present study lies in demonstrating that the affinity‐fishing workflow can recover constituents from an unfractionated botanical extract that exhibit measurable inhibition of recombinant M^pro under the tested conditions.

It is equally important to delineate the specific contribution of the present study. Resveratrol and emodin have previously been evaluated as SARS‐CoV‐2 M^pro inhibitors, and their inhibitory activities are therefore not claimed as new findings [9]. In addition, our group previously developed an NSP5/M^pro‐functionalized magnetic bead ligand‐fishing method and demonstrated its application to Rosmarinus officinalis [18]. The contribution of the present work therefore lies in applying and further evaluating this platform in the chemically distinct matrix of R. japonica and in integrating ligand fishing with UPLC‐Q‐Exactive‐Orbitrap‐MS/MS, HPLC‐DAD quantification, authentic‐standard comparison, FRET‐based enzyme inhibition assays, and computational analyses. To our knowledge, the M^pro‐directed recovery and functional evaluation of glycitein, genistein, apigenin‐7‐O‐glucoside, and biochanin A from R. japonica have not previously been described. Accordingly, this study represents an extension and application of the affinity‐fishing strategy to identify and prioritize M^pro‐binding constituents from an unfractionated botanical extract, rather than the discovery of entirely new chemical entities or the de novo development of the magnetic bead‐based method.

Magnetic bead‐based ligand‐fishing methods have been applied to various enzyme targets, including neuraminidase, monoamine oxidase A, α‐glucosidase, and SARS‐CoV‐2 3CLpro, for the enrichment of potential ligands from complex natural‐product extracts [16, 31, 32, 33, 34]. These studies generally combined target immobilization with LC–MS identification and subsequent enzymatic validation, although the numbers of recovered compounds and the extent of analytical validation varied considerably. Our previous study also applied an M^pro/NSP5‐functionalized magnetic bead platform to identify 15 constituents from Rosmarinus officinalis [18]. In the present study, this strategy was extended to the chemically distinct matrix of R. japonica by integrating high‐resolution LC–MS/MS, authentic‐standard comparison, HPLC‐DAD quantification, and FRET‐based enzymatic validation. Because previous studies employed different targets, magnetic supports, sample matrices, and evaluation criteria, direct comparison of sensitivity, throughput, and overall analytical performance is not appropriate. The contribution of the present work therefore lies in the integrated application and evaluation of M^pro‐directed ligand fishing in R. japonica, rather than in claiming superiority over existing methods.

The case of nuciferine merits particular consideration because its negligible inhibition of M^pro in the present study does not necessarily indicate an absence of antiviral potential. Yang et al. recently reported that nuciferine and its hydrochloride salt exhibited both virucidal and post‐infection antiviral activities against SARS‐CoV‐2 in cell‐based assays [35]. These findings are not inconsistent with our results, because antiviral activity at the virion or cellular level does not necessarily depend on inhibition of M^pro. The observed virucidal activity may involve a direct effect on viral particles, whereas the post‐infection antiviral activity may involve another viral target or a host‐dependent process. However, the molecular target responsible for these effects has not yet been established.

Moreover, the recovery of nuciferine in the ligand‐fishing eluates, together with its negligible M^pro inhibition, cannot by itself establish binding to a specific non‐catalytic site. This result may reflect non‐inhibitory interaction with M^pro, weak binding without a measurable functional consequence, or residual nonspecific retention by the immobilized‐protein system. Nuciferine should therefore not be regarded as inactive against SARS‐CoV‐2 solely on the basis of its lack of M^pro inhibition; rather, its reported antiviral activity may arise through an M^pro‐independent mechanism that requires further investigation.

Overall, the present workflow should be regarded as a qualitative, target‐directed approach for enriching and prioritizing potential M^pro‐binding constituents from complex botanical extracts. Further validation of capture performance and experimental reproducibility would be required before the platform is extended to comparative or quantitative screening applications.

3. Conclusions

This study applied an integrated workflow combining M^pro‐functionalized magnetic bead ligand fishing, UPLC‐Q‐Exactive‐Orbitrap‐MS/MS, HPLC‐DAD, and enzymatic assays to investigate potential M^pro‐binding constituents from R. japonica. Eight compounds were recovered and structurally annotated, of which six showed concentration‐dependent inhibition of recombinant SARS‐CoV‐2 M^pro, with resveratrol exhibiting the strongest activity (IC₅₀ = 34.01 ± 1.67 µmol/L). Molecular docking provided plausible binding poses for selected active compounds, while network pharmacology suggested potential associations with host immune‐ and inflammation‐related pathways. However, these computational results remain predictive, and no SARS‐CoV‐2‐infected cell‐based antiviral or cytotoxicity assays were performed. Therefore, the identified compounds should be regarded as preliminary M^pro‐inhibitory hits rather than confirmed antiviral agents. Overall, the study supports the utility of the ligand‐fishing workflow for prioritizing target‐associated constituents from complex botanical extracts and provides a basis for further mechanistic and cell‐based validation.

4. Experimental Section

4.1. Chemicals

BeyoGold His‐tag Purification Resin and BeyoMag Carboxyl Magnetic Beads (1 µm) were purchased from Beyotime Biotechnology (China). The 20 traditional Chinese medicinal herbs were obtained from the First Affiliated Hospital of Jinan University. Analytical reference standards of baicalein, biochanin A, apigenin‐7‐O‐glucoside, nuciferine, 6‐gingerol, coumarin, resveratrol, genistein, ampelopsin, and glycitein were purchased from Shanghai Macklin Biochemical Technology Co., Ltd. and Jiangsu Yongjian Pharmaceutical Technology Co., Ltd. (China). According to the certificates of analysis supplied with the respective products, the stated purities were ≥ 98% for baicalein, biochanin A, apigenin‐7‐O‐glucoside, nuciferine, 6‐gingerol, resveratrol, genistein, ampelopsin, and glycitein, and ≥ 99% for coumarin. All standards were used as received without further purification.

4.2. Sample Preparation

For ethanol extraction, 50 g of each herbal material was mixed with 1000 mL of aqueous ethanol at the concentration specified for that material and ultrasonically extracted for 1 h. The ethanol concentration was selected according to the corresponding monograph in the Chinese Pharmacopoeia or previously published extraction procedures for the individual herbal material. Because the characteristic constituents of the tested herbs differ in polarity, herb‐specific ethanol concentrations were used to facilitate the extraction of their representative phytochemical constituents.

After filtration, the residue was re‐extracted with an additional 1000 mL of fresh extraction solvent at the same ethanol concentration for 1 h and filtered again. The two filtrates were combined, concentrated under reduced pressure using an N‐1100 rotary evaporator (Tokyo Rikakikai Co., Ltd., Japan), and dried in a DZF‐6090 vacuum drying oven (Shanghai Yiheng Scientific Instrument Co., Ltd., China). The resulting crude extracts were stored at −20°C until further analysis. The initial screening was performed to identify a promising herbal material for subsequent investigation. Following this screening, all further experiments were conducted using the R. japonica extract prepared with 75% aqueous ethanol under a fixed extraction protocol.

4.3. M^pro Expression and Purification

The His‐tagged SARS‐CoV‐2 M^pro was heterologously expressed in E. coli BL21(DE3) using the pET28a(+)‐M^pro expression vector. Bacterial cultures were grown overnight in LB medium supplemented with 50 µg/mL kanamycin at 37°C with agitation (180 rpm), followed by induction with 0.5 mM IPTG for 12 h at 18°C to enhance soluble protein expression [36]. Cells were harvested by centrifugation (6000 × g, 20 min, 4°C), resuspended in lysis buffer (20 mM Tris‐HCl, 300 mM NaCl, 10% glycerol, 1 mM PMSF, pH 7.4), and lysed by ultrasonication (8 cycles of 30 s pulse/30 s rest, 60% amplitude). The lysate was clarified via centrifugation (12 000 × g, 45 min, 4°C) and loaded onto BeyoGold His‐Tag Purification Resin pre‐equilibrated with binding buffer (20 mM Tris‐HCl, 500 mM NaCl, 10 mM imidazole, pH 7.9). Weakly bound contaminants were removed by washing with 20 mM imidazole, and the target protein was eluted using 250 mM imidazole in an elution buffer. Purified M^pro was dialyzed against storage buffer (50 mM Tris‐HCl, 150 mM NaCl, 1 mM DTT, pH 7.5) and confirmed by SDS‐PAGE (> 95% purity).

4.4. Immobilization and Validation of Potential Inhibitors on Magnetic Beads

M^pro was covalently conjugated to carboxyl‐functionalized magnetic beads (MBs) via EDC/NHS chemistry. Briefly, 1 mL of activated MBs was incubated with 1 mL of purified M^pro (1 mg/mL) for 30 min at 25°C under continuous rotation at 20 rpm. Following magnetic separation, the unreacted carboxyl sites on the MBs were blocked by incubation with 5% BSA (1 mL) for 2 h at 25°C.

4.5. Ligand Fishing Assay Validation Using Standard Compounds

To validate the specificity of the ligand fishing assay, a synthetic mixture comprising known high‐affinity M^pro ligands (baicalein [37], ampelopsin [38]) and a low‐affinity compound (coumarin) was prepared. A 1 mL aliquot of this mixed solution (1 mg/mL of each compound) was incubated with the M^pro‐immobilized MBs for 5 min at 25°C with gentle shaking to allow for binding. Subsequently, the MBs were magnetically separated and washed three times with 1 mL of PBS containing 0.1% Tween‐20 for 3 min per wash to remove nonspecifically bound molecules. The specifically bound ligands were then eluted by incubating the MBs with 1 mL of 70% methanol four times for 10 min each.

4.6. Optimization of Ligand Fishing Parameters

To establish the optimal conditions for the ligand fishing assay, baicalein (10 µm) was utilized as a reference standard to systematically evaluate four key operational parameters: M^pro concentration (0.5–2.0 mg/mL), enzyme immobilization time (15–120 min), ligand‐binding incubation period (5–60 min), and methanol elution concentration (10%–90% v/v). During each optimization trial, the M^pro‐immobilized magnetic beads were incubated with the baicalein solution under the specified conditions. Following incubation, the beads were magnetically separated and subjected to three stringent washes with PBS containing 0.1% Tween‐20 to eliminate nonspecific interactions. The specifically bound baicalein was subsequently released using the designated methanol eluent. The resulting eluates were collected, filtered through a 0.22 µm membrane, and analyzed by HPLC‐DAD using the peak area of eluted baicalein as the analytical response. All optimization experiments were conducted in independent triplicates. Comprehensive chromatographic conditions, column specifications, and the corresponding calibration curves are detailed in the Supporting Information.

4.7. Ligand Fishing From R. japonica Extract

The ligand‐fishing procedure was subsequently applied to screen the R. japonica extract. Briefly, 1 mL of magnetic beads was incubated with 1 mL of M^pro solution (1.0 mg/mL) for 30 min to immobilize the enzyme. After magnetic separation and removal of the supernatant, the M^pro‐conjugated magnetic beads were blocked with 1 mL of 5% BSA for 2 h at 25°C. The blocked beads were then magnetically collected and incubated with 1 mL of the R. japonica extract (20 mg/mL) for 1 h at 25°C with continuous shaking at 30 rpm. After incubation, the beads were magnetically separated and subjected to three successive washing cycles. In each cycle, the beads were resuspended in 1 mL of washing buffer consisting of PBS containing 0.05% Tween‐20, shaken for 3 min at 25°C, and magnetically separated. The supernatants collected after the first, second, and third washing cycles were designated W ₁, W ₂, and W ₃, respectively.

Following the washing steps, the bead‐bound compounds were eluted through four successive elution cycles. In each cycle, the beads were resuspended in 1 mL of 70% (v/v) aqueous methanol, shaken for 10 min at 25°C, and magnetically separated. The supernatants collected after the first, second, third, and fourth elution cycles were designated E ₁, E ₂, E ₃, and E ₄, respectively. The original R. japonica extract solution before incubation with the beads was designated P ₀. Each P ₀, W ₁–W ₃, and E ₁–E ₄ fraction was filtered through a 0.22 µm membrane and analyzed individually by HPLC‐DAD. The E ₁–E ₄ fractions were also combined and dried before LC–MS/MS analysis.

4.8. LC‐MS/MS Analysis and Structural Annotation

Ligand profiling and structural annotation were performed using a UPLC‐Q‐Exactive‐Orbitrap‐MS/MS system. Chromatographic separation was achieved on a Hypersil GOLD C18 column (100 × 2.1 mm, 1.9 µm; Thermo Fisher Scientific) utilizing a binary solvent system of 0.1% (v/v) aqueous formic acid (A) and acetonitrile (B). The gradient program initiated at 5% B (0.01–5.00 min), linearly increased to 25% B at 40.00 min, rose steeply to 75% B at 50.00 min, and was maintained at 100% B until 60.00 min. The injection volume was 2 µL.

Mass spectrometry was conducted in both positive and negative electrospray ionization (ESI±) modes. The capillary temperature was set to 320°C, spray voltage at 3.5 kV, and sheath gas flow rate at 35 arb. Full‐scan MS (100–1500 m/z) combined with data‐dependent MS/MS fragmentation was acquired. Data processing and structural annotation were performed using Compound Discoverer 3.1 software. Structures were annotated by matching accurate precursor mass (mass error < 5 ppm) and MS/MS fragmentation patterns against mzCloud and mzVault databases. While LC‐MS/MS provided high‐resolution structural annotation, the identities of the prioritized active ligands were further corroborated by comparing their HPLC RTs with authentic reference standards during the quantitative analysis.

4.9. In Vitro M^pro Inhibition Assay

In vitro inhibition assays were conducted in 96‐well plates using a FRET‐based method with the synthetic fluorogenic substrate Dabcyl‐KTSAVLQSGFRKME‐Edans. The reaction mixtures (100 µL total volume) comprised 80 µL of assay buffer, 15 µL of substrate (20 µm), and 5 µL of the test compound (or vehicle). Crucially, the assay buffer (50 mM Tris‐HCl, 150 mM NaCl, 1 mM EDTA, pH 7.4) was supplemented with 0.01% Tween‐20 to prevent the nonspecific aggregation of polyphenolic compounds and mitigate the risk of false‐positive target inhibition.

After a 30 min incubation at 37°C, fluorescence was quantified (λ _ex = 340 nm/λ _em = 490 nm) using a microplate reader. All trials included baicalein and dihydromyricetin as positive controls and comprised four experimental groups: Blank (RFU₁: substrate + buffer, no enzyme), Compound background (RFU₂: inhibitor + buffer, no enzyme), Full enzymatic activity (RFU₄: enzyme + substrate, no inhibitor), and Test sample (RFU₃: enzyme + substrate + inhibitor). The compound inhibition rate (%) was calculated using the following equation:

$$\mathrm{Inhibition}\mathrm{rate}\left(\%\right)=\frac{\left(\mathrm{RF}{\mathrm{U}}_4-\mathrm{RF}{\mathrm{U}}_1\right)-\left(\mathrm{RF}{\mathrm{U}}_3-\mathrm{RF}{\mathrm{U}}_2\right)}{\left(\mathrm{RF}{\mathrm{U}}_4-\mathrm{RF}{\mathrm{U}}_1\right)}\times 100\%$$ (1)

The IC₅₀ values were derived from dose–response curves obtained from three independent experiments.

4.10. Quantitative Analysis of Captured Compounds by HPLC

Quantitative HPLC analysis of the R. japonica extract (20 mg/mL) was performed on a Hypersil GOLD C18 column (100 × 2.1 mm, 1.9 µm) maintained at 40°C. The mobile phase consisted of 0.1% (v/v) aqueous formic acid (A) and acetonitrile (B) at a flow rate of 1.0 mL/min. The gradient elution program was set as follows: 0–0.01 min, 5% B; 0.01–5.00 min, 5%–25% B; 5.00–40.00 min, 25%–75% B; 40.00–50.00 min, 75%–100% B; and 50.00–60.00 min, 100% B. Analytes were monitored at 254 nm using a diode array detector (DAD) with a 2 µL injection volume. Calibration curves for the reference standards were established across their respective concentration ranges, with all calibration curves showing coefficients of determination (R²) greater than 0.997.

4.11. Molecular Docking

Molecular docking was performed to evaluate the binding mechanisms of the annotated R. japonica ligands against the SARS‐CoV‐2 M^pro crystal structure (PDB: 6LU7). Ligand 3D structures were generated and energy‐minimized using ChemOffice 2019. The M^pro structure was prepared in PyMOL 2.5 by removing crystallographic water molecules and adding polar hydrogens. Docking was conducted using AutoDock 4.2.6 with a grid box (60 × 60 × 60 Å, 0.375 Å spacing; centered at x = −9.253, y = 16.459, z = 65.388) encompassing the catalytic dyad (Cys145‐His41) [39]. The top‐ranked docking conformations were visualized in PyMOL 2.5. Critical intermolecular interactions, specifically hydrogen bonds (1.5–3.2 Å) and hydrophobic contacts, were subsequently mapped and analyzed using LigPlot+.

4.12. Network Pharmacology Analysis

A network pharmacology approach was employed to explore the potential modulation of COVID‐19‐associated host pathways by the identified M^pro ligands. Compound structural data were retrieved from PubChem and processed via SwissTargetPrediction. To prioritize high‐confidence interactions, only target predictions with probability scores > 0.1 were retained [40, 41]. Disease‐associated targets linked to “COVID‐19” and “SARS‐CoV‐2” were collated from GeneCards to establish an integrated ligand–disease target repository [42]. Functional annotation, including GO classification and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, was performed using the clusterProfiler package in R (Bioconductor). Statistical significance was defined as an FDR‐adjusted p < 0.05. Visualizations of the enriched pathways and target densities were generated using the ggplot2 and pathview packages based on their −log10(p value) metrics.

4.13. Statistical Analysis

Unless otherwise stated in the corresponding figure legends, quantitative experiments were performed in three independent experiments (n = 3), and data are presented as mean ± standard deviation (SD). For the FRET‐based M^pro inhibition assay, each independent experiment was measured in technical triplicate wells. Fluorescence signals were corrected for blank and compound‐background signals as described in Equation (1). No additional data transformation or normalization was applied, and no data points were excluded as outliers.

Dose–response curves were fitted by four‐parameter nonlinear regression using a variable‐slope log(inhibitor)‐versus‐response model. IC₅₀ values obtained from three independent experiments are presented as mean ± SD. For the ligand–fishing optimization experiments, differences among experimental conditions were analyzed by one‐way analysis of variance (ANOVA) followed by Tukey's multiple‐comparison test. All statistical tests were two‐sided, and p < 0.05 was considered statistically significant.

Statistical analyses and dose–response curve fitting were performed using GraphPad Prism 9.5.1. Network enrichment analyses were conducted using R 4.3.1 and clusterProfiler 4.8.1. P values from the enrichment analyses were adjusted using the Benjamini–Hochberg false‐discovery‐rate procedure, with an adjusted p value < 0.05 considered statistically significant.

Author Contributions

Yujing Huang: writing – review and editing, methodology, and visualization. Chenjing Tian: writing – original draft, validation, methodology, visualization, and data curation. Zhipeng Gu: software. Jinlian Liang: software. Duntao Zu: methodology and data curation. Jianni Yang: methodology. Wanrong Zhang: writing – review and editing and investigation. Nina Ren: writing – review and editing and Supervision. Junxia Zheng: writing – review and editing, resources, and funding acquisition.

Funding

This study was supported by Guangdong Basic and Applied Basic Research Foundation Natural Science Foundation Project (2023A1515011445), Guangdong Basic and Applied Basic Research Foundation Enterprise Joint Foundation (2022A1515220142), and Guangzhou Basic and Applied Basic Research Foundation—Science and Technology Elite “Piloting” Project (2024A04J6491).

Conflicts of Interest

The authors declare no conflicts of interest.

Use of Generative AI and AI‐Assisted Technologies in the Writing Process

AI‐assisted tools were used solely for language polishing, including grammar, wording, and readability improvements. No AI tools were used for study design, data generation, data analysis, or scientific interpretation. All language edits were reviewed and approved by the authors.

Supporting information

Supporting Information for this article is available on the WWW under https://doi.org/10.1002/MS‐number. Supporting File 1: cbdv71722‐sup‐0001‐SuppMat.docx

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.