We retrieved pertinent information regarding the active ingredients of Zhe-Ba-Wei with anticancer properties from PubMed (https://pubmed.ncbi.nlm.gov), using Atractylodes macrocephala Koidz., Paeonia lactiflora Pall., Fritillaria thunbergii Miq., Chrysanthemum morifolium Ramat., Corydalis yanhusuo W., Scrophularia ningpoensis Hemsl., Ophiopogon japonicus (Thunb.) Ker-Gawl., and Curcuma wenyujin Y.H. Chen et C. Ling as keywords.

SwissTargetPrediction (http://www.wisstargetprediction.ch) is a free database based on compound structures to predict the potential targets for active ingredients. OMIM (http://www.omim.org/) and GeneCards (http://www.genecards.org) are searchable and integrated human genes database that provide concise genomic information on all known and predicted human genes. Discovery Studio 2019 is a professional life science molecular simulation software applied to probe the binding affinity between active ingredients and potential disease targets through semiflexible docking. Discovery Studio 2019 receptors and ligands were obtained from the PDB database (https://www.rcsb.org/) and PubChem database (https://pubchem.ncbi.nlm.nih.gov/), respectively. The PDB codes of the receptor proteins are MET:5EOB and EGFR:1M17. These receptor proteins were treated in Discovery Studio 2019 as follows:1) delete water, 2) clean protein, 3) define receptor, 4) define binding sites, and 5) delete ligand.

Ligands were also processed in Discovery Studio 2019. We need to select “prepare ligands” and then “apply forcefield” operation on the prepared ligands. “CDOCKER Energy” was used for assessing the docking efficiency of these active ingredients, estimating their ability to form stable interactions.

The Discovery Studio 2019 software was used to predict the ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties of active ingredients including aqueous solubility, blood–brain barrier penetration, human intestinal absorption, hepatotoxicity, cytochrome P450 2D6 inhibition, and plasma protein binding. Similarly, the Discovery Studio 2019 was employed to predict the toxicity of active ingredients. The “TOPKAT” module was utilized to forecast toxicity parameters such as rodent carcinogenicity, mutagenicity, aerobic biodegradability, and rat oral LD₅₀.

Human NSCLC cell lines (A549, PC9, and H1975) were from the ATCC and authenticated by short tandem repeat DNA profiling analysis. All cells were cultured with RPMI 1640 medium (Gibco BRL, USA) containing 10% fetal bovine serum (FBS) and incubated in a cell incubator at 37°C, 5% CO₂. For the cell proliferation inhibition assays, 3,000 NSCLC cells in a 100-μL medium were inoculated into each well of a 96-well plate, together with an equal amount of PBS in the marginal wells to avoid edge effect. After treating with different concentrations of drugs for the indicated hours, 100 μL of the diluted CCK-8 solution was added into each well and incubated for 1–2 h according to the manufacturer’s instructions. Finally, the OD value was evaluated by the Multiskan FC microplate reader (Thermo Scientific, USA), and GraphPad 7.0 was used to calculate and analyze the IC₅₀ of the drug (37). All experiments were repeated at least three times.

The cells were collected and lysed with RIPA buffer (#P0013B, Beyotime, China). Twenty micrograms of total protein was used for SDS-PAGE after concentration measurement and then transferred to a PVDF membrane at 250 mA for an appropriate time. After blocking with 5% skim milk for 1 h, the membrane was incubated overnight with primary antibody (1:1,000). Subsequently, it was washed with PBST buffer and incubated with HRP-conjugated secondary antibody (1:1,000) for 1 h. Finally, the ECL kit (#1705061, Bio-Rad, USA) was used and visualized with ChemiDoc Imaging System (Bio-Rad, USA) (38).

Numerous active ingredients associated with antitumor properties have been discovered in Zhe-Ba-Wei as reported in the literature. Taking cancer-related ingredient as the criterion, five active ingredients from A. macrocephala Koidz., one from P. lactiflora Pall., five from C. morifolium Ramat., five from C. yanhusuo W., and one from C. wenyujin Y.H. Chen et C. Ling were selected ( Table 1 ). It is worth noting that no literature has reported any active compounds with anticancer potential in F. thunbergii Miq., S. ningpoensis Hemsl., and O. japonicus (Thunb.) Ker-Gawl. Subsequently, the SMILES number of these 17 active ingredients were identified in PubChem, and the potential targets for various active ingredients were obtained through the online server SwissTargetPrediction (39) ( Supplementary Table S1 ).

To investigate the overlapping targets of active ingredients, these 17 active ingredients were paired together to select the overlapping targets influenced by every two active ingredients. Subsequently, the overlapping targets of every two active ingredients were amalgamated, resulting in the formation of a merged group encompassing every three active ingredients’ overlapping targets. Similarly, the subset comprising the three active components with a greater number of overlapping targets underwent pairwise pairing to obtain a cohort of four active ingredients with the highest degree of shared targets. However, there were limited overlapping targets for every five active ingredients. Thus, based on the screening process, a group of four active ingredients, namely, ferulic acid, rutin, quercetin, and luteolin (referred to as group 1), characterized by a substantial number of overlapping targets, has been successfully identified ( Table 2 ). Interestingly, all four active ingredients come from the same herbal C. morifolium Ramat. Meanwhile, an additional group, denoted as group 2 (such as palmatine, coptisine, tetrahydrocoptisine, and berberine), derived from the same herbal C. yanhusuo W., displayed a notable intersection of targets and possessed a considerable number of overlapping targets ( Table 2 ).

Based on the literature reported, the antitumor properties of diverse active ingredients primarily target the following 12 types of cancers: lung cancer, cervical cancer, colorectal cancer, bladder cancer, breast cancer, oral squamous cell carcinoma, nasopharyngeal carcinoma, hepatocellular carcinoma, leukemia, pancreatic cancer, melanoma, and glioma. Henceforth, all the reported gene targets that correlated with each of these 12 types of cancer were identified from the OMIM (40) and GeneCards databases (41), respectively ( Supplementary Table S2 ). Next, we analyzed the overlapping targets between the four active ingredients (group 1 or group 2) and cancer disease using Venn diagrams. As a result, the numbers of overlapping targets between group 1 and different cancers are shown in Figures 1A, B , but the numbers of overlapping targets between group 2 and cancer disease were limited for further analysis. Thus, we focus our study on the overlapping targets between group 1 and cancers.

Through analysis of the Venn diagrams, it was observed that when the active ingredients of group 1 acted in concert, they shared nine common therapeutic targets among nine types of cancers: lung cancer, cervical cancer, colorectal cancer, bladder cancer, breast cancer, oral squamous cell carcinoma, nasopharyngeal carcinoma, hepatocellular carcinoma, and pancreatic cancer. These targets included arachidonate 5-lipoxygenase (ALOX5), glycogen synthase kinase 3 beta (GSK3B), matrix metalloproteinase-9 (MMP9), matrix metalloproteinase-2 (MMP2), recombinant cytochrome P450 1B1 (CYP1B1), ATP-binding cassette superfamily G member 2 (ABCG2), carbonic anhydrase IX (CA9), epidermal growth factor receptor (EGFR), and mesenchymal to epithelial transition factor (MET). These nine targets have been extensively studied and are known to play pivotal roles in tumor proliferation, drug resistance, and metastasis. Therefore, the potential of these active ingredients in treating tumors warrants further exploration in future studies. Nevertheless, melanoma and glioblastoma share only two overlapping targets (EGFR and MMP2). Furthermore, the active ingredients of group 1 exhibited seven additional targets—F2, ALOX15, NOX4, MAOA, APP, CA2, and CA12—that are distinct from the nine aforementioned cancer types.

Consequently, the complete array of target genes of these four active ingredients was charted into the KEGG pathways and subjected to analysis utilizing Gene Ontology (GO) terms through the DAVID tool ( Figure 1C ). This approach was undertaken to explore the signaling pathways and biological processes through which these four active ingredients exert their anticancer effects.

A false discovery rate threshold of less than or equal to 0.05 was employed to sieve out the signaling pathways and biological processes of higher reliability. Interestingly, we solely acquired the nitrogen metabolism signaling pathway along with 12 additional GO terms ( Figure 1C ). Among them, the GO term exhibiting the strongest correlation with these targets was involved in the plasma membrane. Twelve among the 17 total potential targets (namely, ABCG2, MET, NOX4, APP, ALOX15, CA12, CA2, CA9, F2, EGFR, GSK3B, and MMP2) were associated with this particular biological function. It suggested that the active ingredients we screened potentially exert their anticancer effects by selectively targeting proteins situated on the cellular plasma membrane. In addition, the interconnections among target proteins were investigated using STRING (https://cn.string-db.org/), a highly convenient web platform that facilitates the retrieval of known and predicted protein interactions. The outcomes obtained from the aforementioned website unveil a notable level of interplay between the target proteins, excluding monoamine oxidase A (MAOA) ( Figure 1D ), which displays relatively lower interaction levels.

Considering that both EGFR (42) and MET (43) reside on the cellular membrane, we have opted to pursue further investigation into these two proteins as potential drug targets. Their localization on the cell membrane underscores their significance and motivates our continued research efforts in this direction. It is well-established that EGFR and its associated pathways significantly contribute to tumor progression. EGFR participates in the ERBB signaling pathway, whereas MET is expressed in its alternative pathway, and GSK3B is situated in its downstream pathway known as the PI3K–Akt signaling pathway ( Figure 2 ) (44), which plays critical roles in NSCLC, breast cancer, glioma, etc. According to previous reports, EGFR on the plasma membrane is the receptor for epithelial growth factor (EGF), and mutation or overexpression of EGFR is associated with tumor cell proliferation, angiogenesis, invasion, metastasis, and apoptosis (45). MET is present in the bypass pathway of the ERBB signaling pathway and is a proto-oncogene that interacts with EGFR, thereby promoting drug resistance of the tumor (46). GSK3B is involved in the PI3K–Akt signaling pathway, which can promote the proliferation, differentiation, and angiogenesis of tumor cells (47, 48). In summary, the three proteins mentioned above are most likely the important targets of the four active ingredients we screened; therefore, we speculate that these four active ingredients may have expected therapeutic effects on NSCLC and breast cancer by inhibiting the EGFR and MET (49, 50) in the following text.

To enhance our understanding of the potential interactions between small molecule active components and protein receptor targets, we employed molecular docking techniques utilizing Discovery Studio 2019. The results from this analysis revealed that ferulic acid ( Figure 3A ), quercetin ( Figure 3B ), and luteolin ( Figure 3D ) exhibited remarkable binding affinity toward the EGFR protein, as evidenced by their -CDOCKER energy values of 25.3315 kcal/mol, 36.8433 kcal/mol, and 36.801 kcal/mol, respectively. These findings highlighted the significant potential of EGFR as a target for these compounds. In addition, rutin ( Figure 3C ) exhibited relatively low binding affinity to EGFR protein with -CDOCKER energies of 3.6186 kcal/mol.

Additionally, it was worth noting that ferulic acid ( Figure 4A ), quercetin ( Figure 4B ), and luteolin ( Figure 4D ) also exhibited notable binding affinity to MET, as supported by the theoretical analysis utilizing -CDOCKER energy values of 23.2231 kcal/mol, 33.6036 kcal/mol, and 35.1918 kcal/mol, respectively. In addition, rutin ( Figure 4C ) exhibited relatively low binding affinity to MET protein with -CDOCKER energies of 1.55349 kcal/mol.

The in-silico prediction of ADMET properties stands as a pivotal stage in lead compound discovery and drug development, enabling substantial savings in time, manpower, material resources, and financial resources. The ADMET properties of the four active ingredients were individually predicted using the four most used anticancer drugs in clinics (namely, pemetrexed, osimertinib, gemcitabine, and taxol) as control, employing ADMET descriptors and the Lipinski filter tools provided by Discovery Studio 2019. The analysis provided essential physical parameters, such as aqueous solubility, blood–brain barrier penetration, human intestinal absorption, hepatotoxicity, cytochrome P450 2D6 inhibition, and plasma protein binding.

As shown in Figure 5 , ferulic acid exhibited the most favorable ADMET characteristics among the four bioactive constituents. The parameters were evaluated according to Lipinski’s rule of five (ROF), which serves to assess the drug-like characteristics of the compounds. Furthermore, ferulic acid, luteolin, and quercetin exhibited reasonably favorable absorption parameters. However, the analysis results indicated that rutin’s absorption and blood–brain barrier permeability were not ideal.

Among the four active ingredients, rutin stood out with an AlogP value below 0, indicating its inferior blood–brain barrier capacity and favorable water solubility ( Table 3 ). Except rutin, the other three active ingredients had good absorption in the body ( Table 4 ). Meanwhile, we predicted the hepatotoxicity, cytochrome P450 2D6 inhibition, and plasma protein binding properties of these four active ingredients. Similar to many other anticancer drugs, most of them showed no inhibitory effects to cytochrome P450 2D6, except luteolin ( Table 5 ). Moreover, quercetin, rutin, and luteolin exhibited low rates of plasma protein binding, indicating their favorable in-vivo activity and ability to undergo transmembrane transport, metabolism, and excretion ( Table 6 ). Furthermore, quercetin, rutin, and luteolin demonstrated certain levels of hepatotoxicity, but comparatively lower than that of osimertinib and pemetrexed ( Table 7 ). In addition, ferulic acid exhibited low hepatotoxicity similar to taxol ( Table 7 ).

Next, the toxicity of the four ingredients was calculated through the “TOPKAT” module in Discovery Studio 2019. Predicted data on rodent carcinogenicity, mutagenicity, and rat oral LD₅₀ toxicity parameters were obtained. Quercetin exhibited the lowest levels of rodent carcinogenicity, but its rat oral LD₅₀ was recorded as 175.2 mg/kg. On the other hand, ferulic acid exhibited the highest level of rodent carcinogenicity although its rat oral LD₅₀ stood at 651.1 mg/kg, which is considered remarkable performance. In summary, ferulic acid had the most favorable parameters for drug development among these four ingredients because it exhibited minimal hepatotoxicity, the largest rat oral LD₅₀ value, and diminished first-pass effect due to its potent inhibition on cytochrome P450 2D6.

Since rutin has poor absorption in the body, we chose ferulic acid, luteolin, and quercetin to verify their anticancer activity in NSCLC cells such as A549 (EGFR wild type), PC9 (EGFR mutant without T790M), and H1975 (EGFR mutant with T790M). After the concentration drug treatment was indicated, cell proliferation was detected using CCK-8. The results showed that all of them have good inhibitory effects on NSCLC cells. Notably, ferulic acid ( Figure 6A ), luteolin ( Figure 6B ), and quercetin ( Figure 6C ) showed lower IC₅₀ in the EGFR wild-type cell line A549, compared with that of PC9 (EGFR mutant without T790M) and H1975 (EGFR mutant with T790M). However, there was no significant difference between PC9 (EGFR mutant without T790M) and H1975 (EGFR mutant with T790M), suggesting that these drugs may specifically target the EGFR protein. We further detected the expression of the EGFR protein with the indicated drug treatment by Western blot. The results indicated that all of these three active ingredients, namely, ferulic acid ( Figure 7A ), luteolin ( Figure 7B ), and quercetin ( Figure 7C ), had an inhibitory effect on the expression of the EGFR protein in a concentration-dependent manner.