Forty-seven natural products, sourced from TCM essential oils with anti-upper respiratory tract virus activity, were collected (Supplementary Table 1). To identify compounds that potentially inhibit the binding of ACE2 to the SARS-CoV-2 spike (Spike) protein, virtual screening was conducted using various cryo-electron microscopy structures of the S-protein and ACE2 during the SARS-CoV-2 entry process. These structures included 7TL9 (with the receptor-binding domain (RBD) in chain A in the “up” open conformation), 8C5R (with RBDs in chains A and B in the “up” open conformation), 7TF8 (with all RBDs in the “down” closed conformation), and 7XWA (a complex structure where RBDs are bound to ACE2). Based on the molecular docking binding energies (Supplementary Table 2), hierarchical clustering analysis was performed on a heatmap (Figure 1A) to identify the top 10 compounds with the best binding performance. These compounds included four alcohols, three terpenoids, and one compound each of ethers, ketones, and esters (Figure 1B). These compounds were further analyzed for potential binding pockets in the target proteins.

In the Spike-RBD-1up conformation (Figure 1C), the compounds primarily stabilize a protein pocket located below the receptor-binding motif (RBM). These compounds may allosterically affect the RBD’s conformational changes, potentially decreasing its stability and reducing binding to ACE2. In the Spike-RBD-2up conformation, the compounds mainly stabilize the protein pocket formed by the CT1 (528–590) region of S1, S2’ (686–815) of S2, and Heptad Repeat 1 (HR1) (912–984), located near the fusion peptide (FP) region (816–855). These compounds may influence the release of the HR1 hinge, thereby inhibiting viral fusion (Xing et al., 2025). In the Spike-RBD-3down conformation, the compounds occupy a protein pocket between two RBDs in the closed conformation. This pocket contains a fatty acid (FA) binding site that can be stabilized by linoleic acid (LA), potentially locking the Spike protein in the closed conformation, reducing its interaction with ACE2, and inhibiting viral binding (Toelzer et al., 2020). In the ACE2-RBD complex, two binding sites are identified, one at the interface between RBD and ACE2. This may affect the stability of the ACE2-RBD interaction. These data suggest that the 10 identified compounds have the potential to block SARS-CoV-2 infection.

To further screen for effective compounds, the inhibitory activities of ten potential compounds on Spike-ACE2 binding were initially evaluated through Enzyme-Linked Immunosorbent Assay (ELISA) experiments. The outcomes revealed that Patchouli alcohol (PA), Caryophyllene oxide (CO), or Dibutyl phthalate (DBP) significantly inhibited the binding of Spike with ACE2 (Figure 2A). Subsequently, the cytotoxicity of these three compounds was evaluated using the Cell Counting Kit-8 (CCK8) assay (Figures 2B–D). The results demonstrated that the 50% cytotoxic concentration (CC₅₀) of PA was 172.7 μM, CO exhibited a CC₅₀ of 174.0 μM, and DBP showed a CC₅₀ of 303.3 μM. Next, to quantify Spike-mediated membrane fusion, a Cre-LoxP firefly luciferase (Stop-Luciferase) system was employed to detect DNA recombination events during cell-cell fusion, as previously reported. Through the cell membrane fusion assays (Figures 2E, F), the inhibitory effects of the compounds at concentrations of 50 and 100 μM were tested. Eventually, DBP was identified as the most effective inhibitor of cell membrane fusion, with an inhibitory rate of 66.727 ± 5.590% at a concentration of 100 μM.

To assess DBP’s inhibitory effect on SARS-CoV-2 Spike-mediated cell membrane fusion and SARS-CoV-2 PsV infection, we initially performed a cell membrane fusion experiment. In this experiment, effector cells expressing the SARS-CoV-2 S-WT protein and Cre recombinase (293T/SARS-CoV-2-S-WT/Cre) were co-cultured with target cells expressing the hACE2 receptor and the Stop-luc reporter gene (293T/hACE2/Stop-luc). The results indicated that DBP had a 50% inhibitory concentration (IC₅₀) of 64.53 μM for cell membrane fusion (Figure 3A). Furthermore, we assessed Spike protein cleavage using Western blot (Figure 3B). The results demonstrated that the DBP downregulated the expression of the S2’ subunit in a dose-dependent manner, suggesting that DBP significantly inhibited the fusion between effector cells and target cells. Figure 3C presents the gray values of the S2’ subunit based on three replicated experiments. In the 100 μM DBP treatment group, the gray value of the S2’ subunit was 0.520 ± 0.029, which was lower than the Neg group. Subsequently, we evaluated the inhibitory effect on viral infection with a SARS-CoV-2 pseudovirus experiment. The IC₅₀ of DBP on SARS-CoV-2 pseudovirus was 73.06 μM (Figure 3D). we also tested DBP’s effect VSVΔG/G pseudovirus infection. The results indicated that DBP did not significantly inhibit VSVΔG/G pseudovirus infection, further confirming the specificity of DBP for SARS-CoV-2 pseudovirus.

To observe the inhibitory effect of DBP on cell membrane fusion, a fluorescence colocalization assay was employed to evaluate the quantity of membrane-fused cells after DBP treatment. In this experiment, 293T cells in the Mock group and the experimental group were co-transfected with mCherry and ACE2 plasmids (293T/hACE2/mCherry). mCherry serving as a marker for ACE2, appearing red. Simultaneously, another group of 293T cells was singly transfected with Spike plasmids containing a GFP tag (293T/SARS-CoV-2-Spike/GFP), appearing green. When ACE2 and Spike bind, membrane fusion occurs, forming yellow fused cells (Figure 4A). In contrast, the positive control group lacked hACE2 plasmid, so no yellow fused cells were observed. Hoechst 33,528 dye was utilized to label the cell nuclei as blue for cell enumeration. The results revealed that in comparison with the Mock group, the positive control group had almost no yellow fused cells. In the experimental group, as DBP concentration increased, the number of membrane-fused cells decreased significantly. Subsequently, the ratio of the yellow to the blue regions (yellow/blue) in the fluorescence images was calculated. Results from three replicate experiments were analyzed statistically. The percentage of fused cells and the inhibition rate of membrane fusion are depicted in Figures 4B, C. At 100 μM, DBP inhibited 53.201 ± 1.712% of Spike-mediated membrane fusion.

SPR experiments were conducted to examine DBP’s affinity for both the SARS-CoV-2 spike protein trimer (S trimer) and ACE2, as well as its effect on the Spike-ACE2 interaction. DBP demonstrated high affinity for both proteins, with equilibrium dissociation constant (K_D) values of 0.640 μM for the S trimer and 0.887 μM for ACE2 (Figures 5A,B). Without DBP, the K_D of ACE2 for the S trimer was 8.28 nM. However, when 100 μM DBP was added, this value increased significantly to 86.7 nM. Additionally, the maximum response (Rmax) decreased from 108 to 3.6 at 250 nM ACE2 in the presence of DBP (Figures 5C, D). These findings suggest that DBP competitively inhibits the S trimer-ACE2 interaction, which is important for viral attachment and host cell entry. A summary of the SPR kinetic parameters is shown in Table 1. The proposed mechanism of DBP action is summarized in Figure 5E.

The SARS-CoV-2 S trimer binds to the host ACE2 receptor through its RBD. To assess the inhibitory effect of DBP on this binding, we used ELISA, which showed an IC₅₀ of 379.3 μM for DBP (Figures 6A, B). To determine whether DBP acts on ACE2 or RBD, we designed three preincubation conditions in the ELISA assay (Fu et al., 2021): no preincubation (DBP No Premix), preincubation of DBP with ACE2 (DBP-ACE2 Premix, 1 h), and preincubation of DBP with RBD (DBP-RBD Premix, 1 h) (Figure 6A). Specifically, preincubating DBP with RBD resulted in a more pronounced inhibitory effect on ACE2-RBD binding compared to preincubation with ACE2 or no preincubation (Figure 6C), indicating that DBP primarily targets RBD to disrupt the interaction.

Given ACE2’s roles in regulating blood pressure, reducing inflammation, and protecting heart function (Li et al., 2024), we further examined DBP’s impact on ACE2 enzyme activity. DBP had no significant effect on ACE2 activity at concentrations between 12.5 and 200 μM, while the positive control, MLN4760, significantly decreased ACE2 activity (Figure 6D). Molecular docking revealed that DBP stably binds at the RBD-ACE2 interface (Figure 6E), where the “head” (benzene ring) forms a π-cation bond with SER-496, and the oxygen atoms of the “legs” (ester groups) form hydrogen bonds with TYR-501, Tyr495, and SER-496. Additionally, one “hand” (acyl group) forms a hydrogen bond with Tyr453. We further analyzed the ACE2/RBD interaction, both with and without DBP, using the HDOCK server (Figures 6F, G). The results showed a decrease in hydrogen bonds from 18 to 7 and the loss of the salt bridge (GLU-35, ARG-493). The binding score increased from –368.24 to –270.71 (Table 2). Protein residues marked with an asterisk (*) in Table 2 indicate those that lost their interaction after DBP binding. In general, a lower (more negative) docking score reflects a stronger binding affinity. Overall, DBP effectively disrupts the RBD-ACE2 interaction without affecting ACE2’s enzymatic activity.

To investigate DBP’s inhibitory effect on various SARS-CoV-2 mutant strains, membrane fusion experiments were performed using spike proteins from the Delta and Omicron XBB1.5 mutant strains (Figures 7A–C). The results showed that DBP had an IC₅₀ of 49.22 μM for the Delta variant and 53.70 μM for the Omicron XBB1.5 variant. Notably, DBP exhibited stronger inhibitory activity against both variants compared to the WT strain (Figure 7D), suggesting that DBP’s antiviral efficacy is mutation-dependent, consistent with the molecular docking results. Western blot analysis demonstrated that 100 μM DBP effectively inhibited Spike-mediated membrane fusion in both Delta and Omicron XBB1.5 variants. S-protein cleavage was absent in ACE2-negative conditions, confirming that DBP specifically blocks Spike-mediated membrane fusion (Figures 7E–G). These results indicate that DBP has broad-spectrum antiviral activity against multiple SARS-CoV-2 variants.

To elucidate the critical residues mediating DBP-RBD interaction, we performed site-directed mutagenesis targeting Tyr453 and Tyr495, two conserved residues identified from molecular docking analyses (Figures 8A–C; Supplementary Table 3). Structural modeling revealed that Tyr453 and Tyr495 are pivotal for DBP binding across WT, Delta, and Omicron RBDs: WT-RBD forms two hydrophobic interactions with DBP at these sites, Delta-RBD engages in hydrogen bond, π-stacking and hydrophobic interactions, while Omicron-RBD relies on hydrogen bonding networks involving these residues (Table 3). Sequence alignment (Figure 8D) confirmed the conservation of Tyr453 and Tyr495 in SARS-CoV-2 Spike, and we engineered mutants (WT-Y453G Y495G and Delta-Y453G Y495G) to substitute these tyrosines with glycine. Spike-mediated membrane fusion assay (Figures 8E, F) demonstrated that Y453G/Y495G mutations significantly abrogated DBP’s inhibitory efficacy. In the WT background, 100 μM DBP inhibited membrane fusion by 54.831 ± 1.465%, whereas the mutant (WT-Y453G/Y495G) showed a significantly reduced inhibition of only 18.375 ± 14.124% (Figure 8E). A similar trend was observed in the Delta variant: DBP’s inhibition decreased from 69.490 ± 2.365% in the Delta strain to 13.112 ± 7.211% in the Delta-Y453G/Y495G mutant (Figure 8F). Notably, the Y453G/Y495G mutant exerted a more pronounced effect on attenuating DBP’s inhibitory activity in the Delta variant compared to the WT background (Figures 8E, F). These functional results corroborate the structural predictions, confirming that Tyr453 and Tyr495 are essential for DBP’s binding to the RBD and its consequent suppression of Spike-mediated membrane fusion.

10 mM/mL Caryophyllene oxide (CO) (HY-N3544), β-caryophyllene (HY-N1415), α-caryophyllene (HY-N6968), Cedarol (HY-N2071), Bisabolol (HY-121222), Nerolidol (HY-N1944), Dibutyl phthalate (DBP) (HY-Y0304), Patchouli alcohol (PA) (HY-N0207) and Pogostone (HY-N1416) were obtained from MedChemExpress and stored at –20 °C for all subsequent experiments. Bergamotene (WKQ-0007133) was purchased from weikeqibiotech. Polybrene and Cell Counting Kit-8 (CCK-8) (GK10001) were purchased from Good Laboratory Practice Blosclence (GLPBIO). Luciferase Assay System with Reporter Lysis Buffer (E4030) was purchased from Promega. NB Transfection Reagent was purchased from bioscien. Hoechst 33,258 was purchased from Solarbio (C0021). SARS-CoV-2 Spike trimer protein (40589-V08H4), ACE2 protein (10108-H08H), Recombinant Anti-ACE2 Antibody (10108-R003), SARS-CoV-2 (2019-nCoV) Spike Antibody (Rabbit PAb) (40592-T62) were purchased from sino biological (Beijing, China). SARS-CoV-2 Spike-ACE2 Interaction Inhibitor Screening Assay Kit (502050) was purchased from Cayman. ACE2 Inhibitor Screening Kit (P0320S) was purchased from Beyotime.

The 293T cell line overexpressing ACE2 (ACE2-293T) (CTCC-009-091) was obtained from MEISENCTCC. The human primary embryonic kidney cell line (293T) (CL-0005) was obtained from procell. The cells were cultured in the Dulbecco’s Modified Eagle Medium (DMEM) media supplemented with 10% FBS at 37 °C in a 55 ± 5% humidity with 5% CO₂ atmosphere. The SARSCoV-2 spike plasmids of WT, Delta and Omicron XBB.1.5 for cell-cell membrane fusion assay was provided from Guangzhou Laboratory, guangzhou, China. The SARS-CoV-2 Spike pseudovirus and pLV-Spike-C-GFPSpark (VG40589-ACGLN) were purchased from sino biological (Beijing, China). The NL4-3-mCherry-Luciferase and pMD2.G plasmids were purchased from HedgehogBio Science and Technology Ltd.

Based on the active components of anti-respiratory virus essential oils reported in the literature (Gao et al., 2023), the 3D molecular structure files of each component were retrieved from the PubChem database² and were first converted to PDB format using OpenBabel (O’Boyle et al., 2011). The converted ligand structures were then prepared for docking using AutoDock Tools-1.5.7 (Goodsell et al., 2021). In this step, their rotatable bonds were defined, and hydrogen atoms and Gasteiger charges were added to establish the correct protonation states at physiological pH (7.4). The final prepared structures were saved in PDBQT format. Crystal structures of various conformations of the Spike-ACE2 interaction (PDB IDs: 7TL9, 8C5R, 7TF8, 7XWA) were obtained from the PDB database.³ Water molecules and sulfate ions were removed using PyMOL,⁴ followed by hydrogenation and charge assignment for the receptor using the prepare_receptor4 command. Binding sites were predicted using the DoGSiteScorer algorithm available on the Proteins Plus platform.⁵ DoGSiteScorer (Volkamer et al., 2012) is a grid-based method which uses a Difference of Gaussian filter to detect potential binding pockets and splits them into subpockets. Each small molecule was then docked to the protein subpockets with AutoDock Vina-1.2.0 (Eberhardt et al., 2021). The conformations with the strongest binding affinity were selected, and a heatmap was generated using the ComplexHeatmap package in R. Hierarchical clustering was applied to rank the compounds, and the top ten were identified as potential active ingredients that inhibit the interaction between Spike and ACE2.

The ACE2-293T cells were plated into 96-well plates at a density of 1.2 × 10⁴ cells/well and incubated overnight. Subsequently, the CO, DBP, or PA solution (at concentrations of 25, 50, 100, 200, 400, or 800μM) was introduced for 24 h. At the conclusion of each experiment, 100 μL of CCK-8 reagent was added to each well, and the cells were further cultured for 2 h at 37 °C. The optical density (OD450) was assessed using a Multiskan microplate reader.

In brief, ACE2-His-HRP was added to rFc-RBD-coated test wells containing 50–800 μM of the compounds, as well as to negative control wells without any compounds. Blank wells were left free of both compounds and ACE2-His-HRP. Afterward, HRP-conjugated anti-His antibody and TMB Substrate Solution were added to each well, followed by the addition of HRP Stop Solution. Finally, the plate was read at a wavelength of 450 nm.

The ACE2 Inhibitor Screening Kit utilizes fluorescence resonance energy transfer (FRET) technology, which operates on the following principle: MCA serves as a fluorescence donor (Donor), while Dnp functions as a fluorescence acceptor (Acceptor) or quenching group (Quencher). The absorption spectra of these two fluorescent groups partially overlap, and when the distance between them is optimal (typically 7–10 nm), the donor’s fluorescence energy is transferred to the acceptor, resulting in a reduction in the donor’s fluorescence intensity. The ACE2 substrate binds both MCA and Dnp. Upon cleavage of the substrate by ACE2, MCA and Dnp are separated, causing MCA to emit fluorescence. The ACE2-containing assay solution, containing DBP (at concentrations of 0, 12.5, 25, 50, 100, and 200 μM) or the positive control MLN4760 (at concentrations of 20 nM), was prepared. Substrate was then added, followed by mixing. The reaction was incubated at 37°C for 30–60 min, protected from light, and subsequently analyzed by zymography (Ex325nm/Em393nm).

As previously reported (Yu et al., 2022), Human Embryonic Kidney 293T cells (HEK293T) cells were transfected with the SARS-CoV-2 spike (S) protein (WT/Delta/XBB1.5) and Cre plasmid (293T/SARS-CoV-2-S/Cre). Another set of HEK293T cells were transfected with hACE2 and Stop-Luc plasmid (293T/hACE2/Stop-Luc), while only Stop-luc plasmid was transfected in positive control group (293T/Stop-Luc). The cells were cultured in DMEM medium supplemented with 10% FBS at 37°C for 24 h. Following incubation, both cell lines were treated with 0.25% trypsin and then seeded at 80% confluence in a 1:1 ratio. The cells were co-cultured in the presence or absence of DBP at 37°C for 24 h. To assess the inhibition rate, cells were lysed using passive lysis buffer, and luciferase activity was measured using a luciferase reporter system (Promega), following the manufacturer’s protocol. Western blot analysis was performed to evaluate the cleavage of the Spike protein.

For fluorescence co-localization experiments involving cell-cell membrane fusion, HEK293T cells were transfected with both the mCherry and hACE2 plasmids (293T/hACE2/mCherry), while the positive control group was established by transfecting cells with only the mCherry plasmid (293T/mCherry). A separate set of cells was transfected with the Spike-GFP plasmid (293T/Spike-GFP). Cell nucleus were stained with Hoechst 33,258, and subsequent procedures were performed as previously described. Cells were observed and photographed under a fluorescence microscope at 24 h, and the areas of red and yellow overlap were quantified.

To generate VSVΔG/G Pseudotyped lentiviral particles, 6 × 10⁶ HEK293T cells were plated in a 10 cm dish and cultured overnight. The NB Transfection Reagent (30 μL) was diluted in 250 μL of Opti-MEM. This solution was then combined with a second mixture containing the plasmids NL4-3-mCherry-Luciferase (7.5 μg) and pMD2.G (2.5 μg) in 250 μL of Opti-MEM, according to the manufacturer’s instructions. The combined mixture was incubated for 15 min at room temperature and subsequently added to the plated HEK293T cells. After 6 h, the culture medium was re-placed with fresh complete medium, and the supernatant containing the virus was collected at 48 and 72 h.

A mixture of 25 μL of SARS-CoV-2 PsV or VSVΔG/G PsV in medium was incubated with DBP at concentrations of 0, 25, 50, 75, and 100 μM for 1 h at 37°C. The resulting mixtures (100 μL) were then used to infect ACE2-293T cells. Polybrene with a final concentration of 5 μg/mL was added. After an additional 72 h, cells were lysed using passive lysis buffer, and luciferase activity was measured using a luciferase re-porter system (Promega), following the manufacturer’s protocol.

Surface plasmon resonance (SPR) assays were conducted using a CM5 sensing chip. The stationary phase was activated with EDC/NHS (1:1 ratio) for 420 s. To characterize the binding of DBP, SARS-CoV-2 Spike S1 + S2 trimer protein and human ACE2 protein were separately immobilized on the experimental channel, while the reference channel was used as a baseline control. Binding and dissociation were recorded at a flow rate of 50 μL/min with a gradient dilution of DBP (starting at 3.125 μM, followed by 4-fold dilutions to ≥ 5 concentrations), for 60 s each. Kinetic parameters, including K_D, Kon, and Koff, were calculated using either the Kinetic 1:1 Binding model or the Steady-State model. To assess DBP’s inhibitory effect, after im-mobilizing the Spike trimer on the experimental channel, the binding curves for the inhibitor-free group (ACE2 at 250 nM, with 2-fold gradient dilution, 30 μL/min flow rate, 120 s binding, and 80 s dissociation) and the inhibition group (with 100 μM DBP and ACE2-small molecule pre-incubation, 60 s binding and 140 s dissociation) were compared. The difference in binding curves was assessed after regenerating the microarrays with 10 mM glycine-HCl (pH 1.5) for 30 s. DBP’s blocking efficacy on Spike-ACE2 interactions was quantified based on the change in affinity. All experiments were conducted in a 1 × PBS-T buffer system, and data were processed by subtracting the reference channel’s background.

The 3D structures of ligands were obtained from PubChem² in SDF format and were first converted to PDB format using OpenBabel (O’Boyle et al., 2011). The converted ligand structures were then prepared for docking using AutoDock Tools-1.5.7 (Goodsell et al., 2021), where their rotatable bonds were defined. Hydrogen atoms and Gasteiger charges were added to establish the correct protonation states at physiological pH. The final prepared structures were saved in PDBQT format. Protein as receptors had their PDB files sourced from RCSB Protein Data Bank.³ In PyMOL-3.0.0, water and ligands were removed from the protein structure, and PDBQT files were generated. The DoGSiteScorer algorithm (Volkamer et al., 2012) in Proteins Plus⁵ was employed to predict protein pockets. Among the predicted pockets, the one located at the ACE2-RBD binding interface was selected for subsequent docking studies. The center coordinates and dimensions of the selected binding pocket were defined as the docking box. Finally, each protein pocket parameter and the three AutoDock Vina run parameters (energy_range = 5, exhaustiveness = 8, num_modes = 8) were saved in a “config.txt” file. The Lamarckian genetic algorithm ran AutoDock Vina-1.2.0 (Eberhardt et al., 2021) to dock ligands with protein pockets, recording scores and visualizing modes in PyMOL-3.0.0. To determine the docking score of S protein with ACE2, the HDOCK server⁶ was used (Yan et al., 2020). The crystal structures of RBD with and without DBP were uploaded to the server separately. Then, the crystal structure of ACE2 was uploaded for docking analysis. Using the default parameters of the server, the binding scores of the complexes with and without DBP were compared, and the results were visualized in PyMOL.

SARS-CoV-2 spike (Wuhan-Hu-1, GenBank: QHD43419.1) was homo sapiens codon-optimized by genescripts and generated de novo into pcDNA3.1 vectors by polymerase chain reaction (PCR). Spike WT, Delta and Omicron XBB.1.5 variants containing point and deletion mutations were generated using stepwise mutagenesis using spike construct containing the truncated 19 amino acids at the C-terminal (CTΔ19). Site-directed mutagenesis and deletions were performed using customized primers (synthesized by Sangon) and KOD PLUS neo (TOYOBO) high fidelity polymerase. Parental methylated DNA was digested using DpnI and the resultant PCR products were then transformed into DH5α ultra-competent E. coli (KTSMCC600, AlpaLifeBio). Plasmids were then extracted using DNA extraction miniprep kits (AP-MN-P-250, AxyPrep) and DNA concentrations were adjusted using Nanodrop (Thermo). All mutant plasmids were validated by sequencing service provided by Sangon.

GraphPad Prism version 8.0 was used to graph. Statistical analyses were con-ducted using IBM SPSS Statistics software version 24.0. One-way analysis of variance (ANOVA) was performed, followed by the Least Significant Difference (LSD) post hoc test for homogeneous variances or the Dunnett T3 test for heterogeneous variances, to assess inter-group differences. Statistical significance is denoted as follows: *P < 0.05, **P < 0.01, ***P < 0.001 relative to the Neg or Mock group; #P < 0.05, ##P < 0.01, ###P < 0.001 relative to the positive control group. The data were presented as mean ± SD, n = 3.