Forsythia suspensa leaves were purchased from Chengdu Pi Du Tianyitang Traditional Chinese Medicine Clinic Co., Ltd. The identity of the herbs as genuine was confirmed by Professor Jianhua Huang, Chengdu University of Traditional Chinese Medicine. The FSL leaves were dissolved in ethanol to prepare a sample according to Zhou et al. (2021) and were concentrated to 1/3 volume using a vacuum rotatory evaporator at 90°C. This extraction was repeated five times, then the extract was filtered to obtain the test solution. All stock solutions were stored at 4°C until use. The analysis of FSL was carried out by using a Thermo Hypersil GOLD C18 column (100 mm × 2.1 mm i.d., 1.9 μm) maintained at a temperature of 30°C. The mobile phase was comprised of acetonitrile (A) and an aqueous solution of 0.1% formic acid (B). The gradient elution profile was as follows: from 0 to 5 min, the composition changed from 30% to 65% A; from 5 to 16 min, it increased from 65% to 80% A; then from 16 to 25 min, it rose from 80% to 90% A; and finally, from 25 to 30 min, it transitioned completely to 100% A. The flow rate was established at a rate of 0.3 mL/min, with an injection volume set at 5 μL. High-resolution mass spectrometry detection utilized a Q-Exactive Orbitrap MS system (Thermo Scientific, Waltham, MA, USA), which featured a heated electrospray ionization source operating in both positive (ESI+) and negative (ESI−) ion modes. Data acquisition occurred within the m/z range of 100–1,500 Da under full MS scan conditions. For ESI-MS analysis, specific parameters were configured: the spray voltage was adjusted to 3,000 V for positive ions and 3,500 V for negative ions; the sheath gas flow and auxiliary gas flow rates were set at 35 and 10 arbitrary units, respectively; the capillary temperature remained constant at 320°C, while the auxiliary gas heater temperature was kept at 350°C; stepped normalized collision energies were defined as 20, 40, and 60 eV; and scanning mode operated in full MS/dd-MS2 with resolutions recorded at 70,000 FWHM for full MS scans and 17,500 FWHM for dd-MS2 scans. Data collection and processing utilized Xcalibur version 4.1 along with Compound Discover version 3.0 software tools.

The Simplified Molecular Input Line Entry System (SMILES) formulas of the active components were then entered into the Swiss Target Prediction database (http://www.swisstargetprediction.ch/) to identify their target proteins. Subsequently, the target names were standardized using the UniProt database (https://www.uniprot.org/). The GeneCards database (https://www.genecards.org/) was also screened for targets related to the active ingredients. Venn diagrams were created to visualize the intersection of FSL target genes with the target of HAdV infection.

The mRNA dataset GSE68004 from the GEO database was downloaded, and it contained 37 healthy individuals and 19 HAdV-infected patients (Jaggi et al., 2018). The “limma” package in R4.3.3 software was according to the standard of log2FC >1 and P <0.05. Differentially expressed genes between healthy individuals and HAdV-infected patients were screened out.

The STRING database (https://cn.string-db.org/) constructed the PPI network, and the tsv file was downloaded. Cytoscape was used to construct the “drug-ingredient–target–KEGG pathway” network. The CytoNCA plug-in of Cytoscape software was used to analyze the degree centrality (DC), etc., and it was made as an index for screening and predicting therapeutic targets to topological analyses.

Predicted therapeutic targets were subjected to functional annotation and functional enrichment analysis to investigate the biological significance of the genes or targets. DAVID (https://david.ncifcrf.gov/) database was performed to screen and draw biological processes and signaling pathways.

The 3D structure of flavonoids components from FSL was obtained through the RCSB database (https://www.rcsb.org/). The protein structure was dehydrated, hydrogenated, and charged by Pymol, and then Autodock Vina was used to simulate the molecular docking of the active ingredient of FSL and the core target. The binding strength of the two was evaluated by binding energy value, and the results were visualized by Pymol.

MD simulations were performed using Gromacs 2022.3 with the Amber99sb-ildn force field. Small molecules were parameterized with GAFF via AmberTools22. After energy minimization and NVT/NPT equilibration (100 ps each), production runs were conducted for 100 ns with a 2-fs timestep at 300 K. Additionally, hydrogen atoms were introduced, and the RESP potential was computed using Gaussian 16W software. This potential data was subsequently integrated into the MD system’s topology file. During the simulations, a constant temperature of 300 K and an atmospheric pressure of 1 bar were maintained. The Amber99sb-ildn force field was employed, with Tip3p water molecules acting as the solvent. To maintain system charge neutrality, an adequate number of Na⁺ ions were introduced. Energy minimization was achieved through the steepest descent approach. The system was then equilibrated under isothermal isovolumic (NVT) and isothermal isobaric (NPT) conditions for 100,000 steps each, with a coupling constant of 0.1 ps over a period of 100 ps. Following equilibration, a free MD simulation was conducted, comprising 5,000,000 steps at a step length of 2 fs, totaling 100 ns. The results of the MD simulations, including RMSD, RMSF, Rg value, SASA, and H-bonds, were visualized using GraphPad Prism Version 9.5. The Gibbs free energy was calculated from RMSD and Rg values using a custom “bash” script and was visualized in both 3D and 2D landscapes through Origin (2021) software.

A549 cells at a density of 4–6 × 10⁴ cells/mL were used to inoculate two 24-well plates. On the next day, the cells were grown to be a monolayer of cells with a density of approximately 90%. A total of 10 concentrations were obtained by twofold ratio for serial dilution in the cell culture medium (DMEM + 10% FBS) with cell maintenance liquid. The waste liquid was discarded out of the 24-well plates, and they were washed gently with PBS solution for three times. Furthermore, 1 ml of each concentration gradient drug (drug diluent water) was added to the 24-well plates, and each same-concentration gradient drug was added to four wells. Four wells of normal cells and four wells of 5% concentration of PBS were set up as control groups. They were next cultured under the condition of 5% CO₂ and 37°C for 96 h and observed for any cell changes. The normal cell group and 5% DMSO group served as a control standard, and cytopathic method was used to detect the cellular minimal effective dose of various experimental drugs.

We tested the FSL concentrations (150–750 μg/mL) via CCK-8 assay and confirmed >90% cell viability in all samples. We test the viral particles according to Li et al (Li X. et al., 2021). Briefly, the A549 cell monolayers in 24-well plates were infected with 100 TCID₅₀ of HAdV-2 for 1 h at 37°C. Then, the monolayers were washed twice with MEM and incubated for 7 days. The infected cells were then observed daily under microscopy to check the cytopathic effect (CPE). So, we used different concentrations (containing 0, 150, 250, and 750 μg/mL) to test the virus gene and protein. After A549 cells were infected with HAdV 2 (MOI = 1) for 2 h, they were treated with FSL of different concentrations (containing 0, 150, 250, and 750 μg/mL) for 24 h. HAdV2 was a gift from professor Erguang Li. Subsequently, RNA was isolated from these cells using the RNA isolator Total RNA extraction reagent (Vazyme, R401-01-AA). Then, the RNA was extracted with chloroform. The supernatant fluid was mixed with the same volume of isopropanol to extract the RNA. Finally, the RNA was washed three times with anhydrous ethanol and dissolved in DNase-free water. The RNA concentration of each sample was measured using Nano800 Plus (Jiapeng, Shanghai, China). Then, 1 µg of RNA from each sample was reverse-transcribed to DNA templates with a HiScript II Q RT SuperMix for qPCR (+gDNA wiper) kit (Vazyme, R223-01). The cDNA template was double-diluted with RNase water. The qPCR master mix was followed according to the manufacturer’s instructions (Vazyme, Q712-02). The primers used were designed as presented in Table 1 . The results were processed using Archimed X6 (ROCGENE) and analyzed. This experiment was conducted in duplicate and repeated three times.

Seeding of A549 5 × 10⁵ cells/well in was carried out on a six-well plate, and the cells were incubated under the condition of 5% CO₂ and 37°C for 24 h. The next day, we treated with different concentration of FSL extract (150-750 μg/mL) after infection with HAdV2 (MOI=1). Then, the cell medium was discarded, and the cells were washed twice with PBS. Then, the cells were lysed with NP-40 lysis buffer (Beyotime, P0013F) on ice to obtain the protein, and the protein concentration of each well was measured by the Bradford method. Furthermore, 20 μg protein of each sample was denatured by using 5× SDS-PAGE sample loading buffer (Beyotime, P0015). Then, the denatured proteins were separated using 10% SDS-PAGE gel electrophoresis (Beyotime, P0012AC) and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, USA). The membranes were blocked by using 5% skimmed milk and incubated with Adenovirus-2/5 E1A antibody (1:1,000, Santa Cruz, sc-25, lot #K1022), CDC25A (1:1,000, Beyotime, AF6459), CCNB2 (1:1,000, zenbio, R381384), AURKA (1:1,000, Zenbio, 222073), CHEK1 (1:1,000, Beyotime, AF6498), and CCNA2 (1:1,000, Beyotime, AF6624), with GAPDH (1:2,000, Beyotime, AF0006) serving as the loading control. Then, the membranes were incubated with the HRP-labeled goat anti-mouse secondary antibody (1:2,000, Beyotime, A0216) or HRP-labeled goat anti-rabbit secondary antibody (1:2,000, Beyotime, A0208), and the images were exposed by using Tanon 5200 (Tanon, Shanghai, China).

Following digestion, quenching, and centrifugation, the A549 cells (1 × 10⁶ cells/mL) were resuspended in cell culture medium and seeded into a six-well culture plate for 24 h. For cell cycle analysis, the cells were pretreated with FSL at concentrations ranging from 150 to 750 µg/mL, followed by HAdV infection at an MOI of 1. After infection, the cells were incubated for 24 h before being harvested for analysis. The harvested cell suspensions were subjected to dual PBS washing cycles (4°C, 1,000 ×g, 10 min). These were fixed via dropwise addition of 4.5 mL ice-cold 70% ethanol during vortex mixing, followed by 18 h of fixation at -20°C. The pelleted cells underwent PBS rehydration (2 mL, 1,000 ×g, 10 min) and RNA digestion with 500 μL RNase A (100 μg/mL, 37°C, 30 min, dark). Nuclear staining was carried out using PI solution (25 μL of 1 mg/mL stock, 10 min, dark). The cell cycle profiles were ultimately quantified using a LongCyte™3 flow cytometer (Challenbio, Beijing, China) with FlowJo_v10.8.1 analysis software.

All data were presented as mean ± SEM from three independent trials. Statistical evaluation was conducted via one-way analysis of variance (ANOVA) with Tukey’s post-hoc test analysis using GraphPad Prism version 9.5.0.

The total ion chromatogram (TIC) in both positive and negative ion modes is displayed in Figure 1 . The structures of the compounds were confirmed by comparing their retention times, accurate and high-resolution masses, as well as MS/MS fragments with standards or related literature. A total of 39 components were identified ( Table 2 ; Figure 1 ), comprising 11 organic acids, four terpenoids, two lignans, four flavonoids, and 18 other substances.

To convert the target protein names into gene names, the UniProt database was utilized. After removing duplicates, a total of 556 corresponding gene targets were obtained.

A total of 37 normal and 19 HAdV-infected peripheral blood RNA-seq samples were obtained from the GEO database. By employing DESeq2 and setting a significance threshold at a P-value of less than 0.05 along with an absolute log2 fold change greater than 1, we detected 990 genes that were differentially expressed (DEGs) when comparing the normal group to the HAdV-infected group. To illustrate the data, a volcano plot was created, highlighting 467 genes that were upregulated and 523 genes that were downregulated ( Figure 2A ). Additionally, a heatmap was generated to depict the expression patterns of these DEGs ( Figure 2B ), and the top 10 upregulated and top 10 downregulated DEGs are presented in Table 3 . The 990 DEGs were selected as potential targets of HAdV infection and were further analyzed in subsequent studies.

To get the therapeutic targets of FSL against HAdV infection, we made the intersection of disease targets and component target. A total of 990 disease targets were matched with 556 drug targets, resulting in 30 common targets that were selected for further investigation ( Figure 3A ). The “drug ingredient–targets” network was constructed by using Cytoscape ( Figure 3B ). By utilizing the STRING database, interactions between these targets were constructed and then visualized by using Cytoscape. The resulting network diagram displayed 27 nodes and 96 edges, with nodes of higher degree values appearing with darker colors and larger sizes at the center of the protein–protein interaction (PPI) network ( Figure 3C ). The core genes contained CCNA2, CHEK1, and AURKA (degree was 14 throughout), CDC25A (degree = 13), CCNB2, KIF11, TOP2A, NELK, AURKB, PBK, and TYMS (degree was 12 throughout), as well as other PLK4 and TTK (degree was both 11) of FSL for anti-HAdV infection.

The analysis of GO enrichment for 30 key genes was conducted using “R,” with a significance threshold established at P <0.01. The biological process (BP) enrichments were mainly cell cycle G2/M phase transition, mitotic cell cycle process, nuclear chromosome segregation, mitotic cell cycle, nuclear division regulation, cell cycle phase transition, spindle tissue, etc. Cellular component (CC) enrichments were mainly intracellular components related to the cell cycle. Molecular function (MF) enrichment was involved in various protein kinase and lyase activities ( Figure 4A ). The KEGG signaling pathway, the target gene of FSL against HAdV infection, was involved in a total of 12 pathways, including cell cycle, progesterone-mediated oocyte maturation, acute leukemia, chemocarcinogenicity-receptor activation, human immunodeficiency virus type 1 infection, and human T-cell leukemia virus type 1 infection. The heatmap shows the top 10 pathway ( Figure 4B ). A chord diagram was generated to illustrate the connections between target genes and enriched pathways ( Figure 4C ). The KEGG pathway–key gene–active ingredient relationship network was constructed by using Cytoscape ( Figure 4D ). The core genes CCNA2, AURKA, CHEK1, CCNB2, and CDC25A were involved in five or more virus-related signaling pathways.

In order to validate the findings that four flavonoids’ components bind with their key targets, molecular docking techniques were used to assess it. The spatial coordinates of the docking configurations and comprehensive details relating to the targets and compounds used in the analysis can be found in Table 4 . To provide a visual depiction of the interactions between compounds and targets, the binding modes were illustrated using PyMOL-1.7.2.1 and Discovery Studio 2020, with the results showcased in Figures 5A–D . The binding energy outcomes from the docking simulations varied, ranging from -6.7 to -9.2 kcal/mol, which suggests that these molecular complexes exhibit stable binding characteristics, indicating promising interactions between the compounds and the selected targets. Additionally, we observed that kaempferol, luteolin, and quercetin exhibit a strong binding activity with CHEK1 and CCNB2 ( Figures 5A–D ). Therefore, these compounds are considered candidate drug molecules.

Furthermore, a 100-ns molecular dynamics (MD) simulation was conducted to explore the dynamic characteristics of the CCNB2–wogonin, CCNB2–kaempferol, and CHEK1–luteolin complexes obtained through molecular docking. The MD simulation outcomes, particularly those related to the root mean square deviation (MSD), root mean square fluctuation (RMSF), radius of gyration (Rg), solvent accessible surface area (SASA), and hydrogen bonds (H-bonds), are essential to assess the stability of both chemical compounds and protein complexes. Additionally, these simulations facilitate the evaluation of the stability of protein tertiary structures after binding with small molecules, the hydrophobicity of amino acid residues, and other pertinent factors.

Adenovirus is a double-stranded DNA virus that includes early genes (E1A, E1B, E2, E3, and E4) and late genes (L1, L2, L3, L4, and L5). To confirm the impact of FSL treatment for HAdV infection, A549 cells were treated with FSL to assess the expression of HAdV early genes (E1A, E2, E3, and E4) and E1A proteins. FSL concentrations (150–750 μg/mL) were tested via CCK-8 assay and >90% cell viability was confirmed in all samples ( Supplementary Figure S1 ). A549 cells were exposed to 24 h of varying different concentrations of FSL (150–750 μg/mL) prior to 2 h of HAdV infection. The expression of the E1A, E2, E3, and E4 genes was examined using real-time PCR. FSL led to a reduction in HAdV gene expression in a dose-dependent manner compared to that in the control group ( Figures 9A–D ). FSL also reduced the expression of HAdV E1A protein in a manner that depended on the dose when compared to the control group ( Figures 9E, F ). These results imply that FSL may be effective in treating HAdV infections.

To evaluate the impact of FSL on HAdV-induced cell cycle alterations, A549 cells or BEAS-2B cells were pretreated with escalating FSL concentrations for 24 h prior to infection with HAdV (MOI = 1). Flow cytometric analysis revealed a dose-dependent shift in cell cycle distribution. Compared with untreated controls, FSL-exposed A549 cells exhibited significantly reduced G1-phase populations (from 31.7% to 5.62%) with concomitant G2/M-phase accumulation (increasing from 47.8% to 76%) ( Figure 10 ). Moreover, we found that FSL-exposed BEAS-2B cells increased the G2-phase population ( Supplementary Figure S2 ). The distinct G2/M phase arrest pattern suggests that FSL may disrupt normal cell cycle progression during HAdV infection.

To evaluate the effect of FSL treatment on the expression of hub genes during HAdV infection, A549 cells were pretreated with FSL before being infected with HAdV, allowing for the assessment of hub gene expression. Following 24 h of pretreatment with different concentrations of FSL prior to HAdV infection, qRT-PCR and Western blot analyses revealed distinct regulatory effects on cell cycle-related targets. Compared to controls, FSL treatment significantly downregulated the expression of the hub gene CDC25A while dose-dependently increasing the mRNA levels of CCNA2, AURKA, CHEK1, and CCNB2 ( Figure 11 ). Consistent with the qRT-PCR results, the CDC25A protein levels decreased significantly, whereas the CCNB2, CCNA2, CHEK1, and AURKA levels were markedly increased following FSL treatment ( Supplementary Figure S3 ). These findings suggest that FSL treatment may influence the hub genes associated with HAdV infection.