A Mechanism Exploration for the Yi-Fei-San-Jie Formula against Non-Small-Cell Lung Cancer Based on UPLC-MS/MS, Network Pharmacology, and In Silico Verification

Non-small-cell lung cancer (NSCLC) is one of the most prevalent cancers worldwide. A Yi-Fei-San-Jie formula (YFSJF), widely used in NSCLC treatment in south China, has been validated in clinical studies. However, the pharmacological mechanism behind it remains unclear. In this study, 73 compounds were identified using ultraperformance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS), with 58 enrolled in network pharmacology. The protein-protein interaction network, functional enrichment analysis, and compound-target-pathway network were constructed using 74 overlapping targets from 58 drugs and NSCLC. YFSJF has many targets and pathways in the fight against NSCLC. PIK3R1, PIK3CA, and AKT1 were identified as key targets, and the PI3K/AKT pathway was identified as the key pathway. According to the Human Protein Atlas (THPA) database and the Kaplan–Meier Online website, the three key targets had varying expression levels in normal and abnormal tissues and were linked to prognosis. Molecular docking and dynamics simulations verified that hub compounds have a strong affinity with three critical targets. This study revealed multiple compounds, targets, and pathways for YFSJF against NSCLC and suggested that YFSJF might inhibit PIK3R1, PIK3CA, and AKT1 to suppress the PI3K/AKT pathway and play its pharmacological role.


Introduction
Lung cancer is the most frequent and leading cause of cancer death, with NSCLC accounting for most cases. Fighting against lung cancer is still a major issue. According to the World Cancer Report 2020 published by the International Agency for Research on Cancer (IARC), 2.1 million new lung cancer cases and 1.8 million deaths worldwide were reported in 2018, accounting for 11.6% of cancer cases and 18.4% of cancer deaths [1].
Te most common treatments for lung cancer include surgery, chemotherapy, radiotherapy, targeted therapy, and immunotherapy [2]. With the development of clinical studies, a combination of diferent therapies has been a trend for present and future treatment, especially for advanced lung cancer patients [3][4][5][6][7][8]. Many hospitals in China use a combination of contemporary pharmacological medications and traditional Chinese medicine (TCM). Te Yi-Fei-San-Jie formula (YFSJF), formerly known as the Yi-Qi-Chu-Tan formula, has been used in NSCLC treatment in South China for years. Years of clinical studies have shown that YFSJF improves NSCLC patients' survival and quality of life while also reducing the harmful efects of contemporary therapies. A multicenter, prospective, clinical cohort study with 315 elderly advanced NSCLC patients showed no signifcant diference in the median overall survival (mOS) and the median time to progression (mTTP) between the Yi-Qi-Chu-Tan group and the chemotherapy group [9]. A casecontrol study of 62 patients with advanced NSCLC demonstrated that the Yi-Qi-Chu-Tan formula could improve the mOS and the median progression-free survival (mPFS) [10]. In a retrospective study of 376 patients with epidermal growth factor receptor (EGFR) wild-type advanced NSCLC, it was discovered that combining YFSJF with chemotherapy could improve the mOS compared to chemotherapy alone [11]. Another retrospective analysis of 90 patients with EGFR mutant-type advanced NSCLC showed that combining targeted therapy with YFSJF could extend PFS and reduce the side efects of targeted therapy [12]. Other clinical studies reported that combining YFSJF with chemotherapy could reduce fatigue and enhance the quality of life [13][14][15]. Te mechanism of YFSJF for treating NSCLC deserves more attention because of its valid clinical proof.
Scientists are working on multiple aspects of studying TCM, including clinical efcacy, pharmaceutical mechanisms, efective components, extraction separation, and preparation development. Te multiple components and targets of TCM make it challenging to fgure out its mechanism. With the development of analysis technology, a combination of chromatography and mass spectrometry was invented and applied in various felds, providing the opportunity to separate and analyze complex compositions simultaneously. UPLC-MS/MS has been extensively applied in TCM for chemical analysis, quality control, pharmacokinetics, serum pharmacochemistry, and rapid screening of active components [16][17][18][19][20][21]. In recent years, network pharmacology has been used to explore the mechanism of TCM, especially for a formula with multiple components [22]. From a holistic standpoint, it can explain the pharmacological impact of a complicated object. Tis study aimed to use UPLC-MS/MS, network pharmacology, and in silico methods to investigate the mechanism of YFSJF against NSCLC. Figure 1 illustrates the fowchart.

Reagents and Materials.
All herbs were purchased from Kangmei pharmacy (Kangmei Pharmaceutical Co., China). HPLC-grade acetonitrile and ethanol were purchased from Fisher Scientifc (Pittsburgh, USA), and HPLC-grade formic acid was purchased from Sigma (St. Louis, USA). Water was purifed via a purifcation system (Millipore, Bedford, MA, UAS).

Preparation of YFSJF.
Te prescribed amount of YFSJF was soaked for 30 min, followed by the addition of 8 and 6 times the volume of water, respectively, for refux extraction (twice), each for one hour. Te fltrate was concentrated to 100 mL, and ethanol was added to the cooled fltrate for a fnal concentration of 85%. Te liquid was fltered, concentrated, and diluted in 100 mL water after 24 hours. Te fnal crude drug concentration was 1.45 g/mL. Te sample used in high-performance liquid chromatography-mass spectrometry (HPLC-MS/MS) was a tenfold diluent with ethanol after fltering with a 0.22 μm membrane. gradient, 5-60% A at 0-60 min, 60-95% at 60-90 min,  95-5% at 90-95 min, and 95-95% at 95-100 min.  A triple TOF-TM 5600 plus hybrid triple quadrupole  time-of-fight mass spectrometer (AB SCIEX, USA) equipped with Analyst ® TF 1.6 software was utilized for mass spectrometry. Te following were the mass spectrometry conditions: electrospray ionization (ESI) source; ion spray voltage, 5500 V in a positive mode and −4500 V in a negative mode; ion source gas 1/2, 55 psi; curtain gas, 35 psi; temperature, 550°C; collision energy, 10 eV; collision energy spread, 20 eV; declustering potential, 80 eV; and scanning range, 50-1500 m/z. Positive and negative ion modes were used for detection, respectively. Te PeakView software equipped with the Chinese medicine mass spectrometry database (AB SCIEX, USA) was used to evaluate the data. By comparing the mass fragments in the database and analyzing the fragmentation pattern, the ingredients of YFSJF were identifed. . Molecules with anticancer efects or oral bioavailability (OB) less than 20% and drug-likeness (DL) less than 0.15 were enrolled for target screening [23]. As a backup, the PubChem database was used to retrieve the canonical SMILES strings and 3D structures of the specifed components (https:// pubchem.ncbi.nlm.nih.gov/). Te canonical SMILES strings of each selected molecule were imported into the Swiss Target Prediction database to obtain related targets when species was set as Homo sapiens (https://www. swisstargetprediction.ch/). Te target with a probability of 0.1 or higher was maintained. UniProt was used to correct the target name (https://www.uniprot.org/).

Potential Target for YFSJF in Treating NSCLC.
Te overlapping targets of YFSJF and NSCLC were shown in a Venn diagram and were the potential target in treating NSCLC.

Protein-Protein Interaction (PPI) Network Construction.
A PPI network without independent nodes was used to elucidate the molecular mechanisms of YFSJF's anti-NSCLC activities by the STRING database (version 10.0, https:// www.string-db.org/), with a required confdence level >0.9. Te Cytoscape software (version 3.6.1; https://www. cytoscape.org/) was used to examine the network. Te key treatment targets were those whose betweenness, closeness, and degree were larger or equivalent to the median, and they were preserved for future visualization. Te top 10 degrees in the PPI network were hub targets.
2.6. GO and KEGG Pathway Analysis. Te Database for Annotation, Visualization, and Integrated Discovery (DA-VID, https://david.ncifcrf.gov/) was employed for functional analysis of the key prospective therapeutic targets. Te DAVID online tool analyzed target functions using gene ontology (GO) analysis and the Kyoto Encyclopedia of Genes and Genomics (KEGG) pathway analysis.

C-T-P Network Construction.
Te compounds, main therapeutic targets, and top 20 KEGG pathways were collected in an MS Excel fle as a backup. Te C-T-P network was created using the Cytoscape software (version 3.6.1; https://www.cytoscape.org/). Hub compounds were among the top ten compounds.

Relationship between Objective Hub Targets and NSCLC
Patients. Te expression of hub targets in normal and NSCLC tissues was compared using the THPA database (https://www.proteinatlas.org/). Te Kaplan-Meier Online website (https://kmplot.com/analysis) was used to investigate the connection between the expression level of hub genes and prognosis in NSCLC. Te result was depicted through a survival curve with a hazard ratio (HR).

In Silico Verifcation
2.9.1. Molecular Docking of the Key Targets. Molecular docking technology was used to study the binding interaction between hub compounds and objective hub targets. Compounds and targets were constructed using the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and the PDB website (https://www.rcsb.org/), respectively. Te AutoDock software was used to remove water molecules from the protein structure and hydrogenate it, after which docking pockets were constructed using the protein's ligand. Te CB-Dock website (https://cao.labshare.cn/cb-dock/) was used to predict docking pockets for proteins without their ligand. After determining docking pocket parameters, the AutoDock Vina software (1.1.2 version) was used for molecular docking and conformation grading. A 2D combination conformation was generated through Discovery Studio (2019 version).

Molecular Dynamics Simulation of the Key Targets.
A molecular dynamics simulation through GROMACS 2021 was used to verify the interaction between hub compounds and objective hub targets. Te unconstrained dynamics simulation was run with the CHARMM36 force feld and the TIP3P water model. Molecular MOL2 fles were converted to str fles through the CHARMM General Force Field (CGenFF) website (https://cgenf.umaryland.edu/initguess), and then, str fles were converted to topology fles through Evidence-Based Complementary and Alternative Medicine Python 2.7 (script: cgenf_charmm2gmx.py) (https:// mackerell.umaryland.edu/charmm_f.shtml#gromacs). Te simulating system employed a dodecahedral solvent box with a 1 ns periodic boundary condition. To balance the system, 500 ps NVT and 1000 ps NPT were used, with a temperature of 300 K and a pressure of 1 bar, followed by a 20∼30 ns simulation. Te time increment was 2 fs, and the confrmation was saved every 10 ps.

Compounds of YFSJF. Te complicated ingredients of
Chinese decoction make it hard to fgure out efective compounds against a specifc disease. Compounds were detected using UPLC-MS/MS. Te total ion chromatography (TIC) of UPLC-MS/MS is shown in Figure 2. Tose with anticancer efects published, OB less than 20%, and DL less than 0.15 were enrolled in target screening. As shown in Table 1, 73 compounds were identifed with their herbal attributions indicated by corner markers and 58 compounds (M1-M58) were enrolled for further targeting collection.

Overlapping Targets of YFSJF and NSCLC and the Proteinto-Protein Interaction (PPI) Network.
According to the Swiss Target Prediction database, 522 targets were predicted for 58 compounds. NSCLC was estimated to have 1668 targets according to various disease target databases. As shown in Figure 3, 254 targets overlapped and were potential targets for YFSJF to treat NSCLC.
After eliminating independent nodes, the Cytoscape software created a PPI network with 223 nodes and 1288 edges based on 254 overlapping targets. To obtain additional parameters, CytoNCA was used to analyze the network. We screened out 74 main targets whose betweenness, closeness, and degree were greater than or equal to the median (≥123.0941842, ≥0.358642973, ≥8, respectively). Based on key 74 targets, a PPI network with 74 nodes and 640 edges was generated, as shown in Figure 2(b). Te degree is determined by the depth of the color or the size of the node. Te top 10 in terms of degrees were hub targets, whose interrelation is shown in Figure 2

Efects of Key Targets on NSCLC Patients.
Te expression of the three major targets in tissue was investigated using the Human Protein Atlas (THPA) database. As shown in Figure   7, the expression levels of the three targets were distinguished between normal and NSCLC tissues. PIK3R1 and PIK3CA were not detected in normal alveolar cells but were found in lung adenocarcinoma and squamous cell carcinoma with lower or medium expression. AKT1 expression was found to be high in normal alveolar cells but average in lung adenocarcinoma and squamous cell carcinoma. Different expressions suggest that PIK3R1, PIK3CA, and AKT1 may play a role in the occurrence and progression of NSCLC.
Te Kaplan-Meier plotter online database was used to analyze the correlation between three key targets and the prognosis of NSCLC (Figure 8). Te survival curve indicated that the expression of PIK3R1, PIK3CA, and AKT1 correlated with the mOS of lung adenocarcinoma (p < 0.05). Te mOS of the group with high PIK3R1 and PIK3CA expression was better than that of the group with low PIK3R1 and PIK3CA (HR � 0.4, HR � 0.61, respectively). Te mOS result was reversed for diferent expressions of AKT1 (HR � 1.67). In the case of lung squamous cell carcinoma, the curves for diferent expressions in 3 key targets appeared to separate, particularly for AKT1. Te separating tendency indicated a correlation between the target and the mOS in lung squamous cell carcinoma, but there was no statistical signifcance. Evidence-Based Complementary and Alternative Medicine

Molecular Docking and Molecular Dynamics Simulation.
Te binding afnity of ten hub compounds and three major targets was investigated using molecular docking. Generally, binding afnities of less than −5 kcal/mol are considered, indicating a good confrmation intersection. Te result was visualized through a heatmap (Figure 9(a)), in which a deeper blue color indicated a stronger binding afnity. All 10 hub compounds were signifcant for PIK3R1, PIK3CA, and AKT1 docking, with M32 (glycyrrhetinic acid), M57 (imperialine), and M39 (ganoderic acid A) possessing the strongest binding afnity, respectively. Figure 9(b) shows the structure of the most potent binding molecules with three important targets. It suggests that the molecule may combine with protein through a hydrogen bond, carbon-hydrogen bond, alkyl, and pi-alkyl. Te molecular dynamics simulation showed that the three systems became stable between 20∼30 ns and that all proteins' root mean square error (RMSD) fuctuated around 0.1 nm. Te radius of gyration (Rg) gradually increased more or less as the simulation progressed, indicating that the molecule may hinder protein by uncoiling it.

Discussion
TCM's abundance of herbs and diverse formulas create a blue ocean for medical study, attracting an increasing number of researchers. Tis study discovered multiple ingredients, targets, and signal pathways for YFSJF against NSCLC. Te discovery of 73 YFSJF components using UPLC-MS/ MS established the groundwork for additional research. Fifty-eight fltered compounds were enrolled in network pharmacology analysis. Te PPI network was built using the intersection of compound targets and NSCLC. PTPN11, FYN, AKT1, PIK3CA, PIK3R1, MAPK3, STAT3, MAPK1, HSP90AA1, and Src were promising targets for YFSJF against NSCLC by the PPI. Tey are mainly involved in PI3K-AKT signaling and MAPK signaling. Src and Fyn are Evidence-Based Complementary and Alternative Medicine  Evidence-Based Complementary and Alternative Medicine members of the Src family kinases of nonreceptor tyrosine kinases, mediating diferent intracellular signaling pathways. Src family kinases have been identifed as important in terms of progression, invasion, metastasis, bone pain, and drug resistance in human cancer, including NSCLC [24][25][26]. Src and Fyn are activators of PI3K and can mediate the PI3K/ AKT/mTOR pathway [27]. PIK3R1, PIK3CA, and AKT1 play a critical role in the PI3K/AKT pathway. Type I PI3K is a heterodimer, consisting of PIK3R1 and PIK3CA, closely associated with cancer. PI3K activation can phosphorylate PIP2 into PIP3, then, PIP3 recruits PDK1 and AKT1 to the plasma membrane, and PIP3 activates AKT1. Te activation of AKT1 further stimulated downstream pathways to promote cancer cell proliferation, invasion, and metastasis angiogenesis [28,29]. MAPK1 and MAPK3 (also known as ERK2 and ERK1, respectively) are serine-threonine protein kinases, which are found downstream of classical MAPK signaling, and can promote cell proliferation, cell cycle, and adhesion [30]. Besides, there is cross talk between the two signaling pathways. Inhibition of the PI3K/AKT signaling pathway leads to the compensatory generation of the MAPK/ERK signaling pathway and vice versa [31,32]. PI3K inhibitors inhibit AKT and activate B-Raf (upstream of the MAPK pathway), boosting MAPK pathway activation [33]. By inhibiting mTOR and P70S6K (downstream of the PI3K-AKT pathway), mTOR inhibitors can activate the MAPK pathway via a feedback loop [34]. In NSCLC patients, aberrant activation of PI3K/AKT and MAPK pathways regulates tumor occurrence, development, and treatment resistance. Hence, multiple inhibitors of two pathways are being developed and clinically tested for NSCLC therapy [35,36]. For instance, HH2710, an ERK1/2 inhibitor, was approved by the FDA in September 2019 to treat cancer, including NSCLC with a gene mutant of the MAPK pathway, and subsequently approved by the CDE in March 2020. Te gene ontology biological process (GO-BP) analysis showed that YFSJF is primarily involved in protein/amino acid phosphorylation and autophosphorylation. Protein phosphorylation is one of the most prevalent post-translational modifcations. It regulates protein function and    [37]. Te GO-CC analysis result showed that YFSJF is mainly related to intracellular constituents, such as cytosol, nucleus, and nucleoplasm. It was consistent with hub target-related pathways, as PI3K/AKT and MAPK pathways are intracellular [37]. Te GO-MF analysis result showed that hub targets were uniform. Te KEGG analysis result revealed that YFSJF produced therapeutic efects on NSCLC by regulating multiple pathways (e.g., pathways in cancer, PI3K-AKT signaling pathway, proteoglycans in cancer, HIF-1 signaling pathway, ErbB signaling pathway, and VEGF signaling pathway). Tey were closely related to the tumor microenvironment (e.g., extracellular matrix stability and hypoxia) and tumor behavior (e.g., invasion and metastasis), and the PI3K-AKT pathway was involved in other pathways [38][39][40][41][42][43][44][45]. Ten YFSJF hub chemicals were identifed using the C-T-P network. Combining PPI and KEGG pathway results, we infer that YFSJF may mainly treat NSCLC by targeting PIK3R1, PIK3CA, and AKT1 and regulating the PI3K/AKT pathway. As a result, prognostic analysis and molecular docking were applied to the three primary targets. Te results showed that PIK3R1, PIK3CA, and AKT1 expressions could distinguish between normal and NSCLC tissues and might be used as prognostic markers in NSCLC. Overexpression of PIK3R1 was shown to be signifcantly associated with a poor prognosis in adenocarcinoma [46]. Studies have shown that PIK3CA mutations may confer a survival advantage [47,48]. Other studies have found that PIK3CA gene mutations are associated with poor OS and reduced PFS of EGFR-TKI treatment in diferent subtypes of NSCLC [49][50][51][52]. Overexpression of p-AKT may help cancer patients live longer [53]. Finally, molecular docking and dynamics simulations validated the hypothesis that YFSJF inhibits PIK3R1, PIK3CA, and AKT1 to suppress the PI3K/ AKT pathway and fght NSCLC. After further literature review, it was found that the antitumor role of hub chemicals has been reported by many researchers, especially for lung cancer [54][55][56][57][58][59]. For instance, ginsenoside compound K, the major intestinal bacterial metabolite of ginsenoside Rb1, can induce apoptosis and autophagy in NSCLC cells by regulating the AMPK/mTOR pathway [54]. Naringenin can induce Bax-mediated mitochondrial apoptosis in human lung adenocarcinoma A549 cells [55]. Liquiritigenin can suppress lung adenocarcinoma A549 cell migration via the PI3K/AKT pathway [56]. 18β-Glycyrrhetinic acid can induce apoptosis and G2/M cell cycle arrest and inhibit migration via ROS/MAPK/STAT3/NF-κB signaling pathways in A549 lung cancer cells [57].

Conclusion
Te study verifed the characteristics of multiple ingredients, targets, and pathways for YFSJF against NSCLC, with PIK3R1, PIK3CA, and AKT1 identifed as the major targets associated with NSCLC patients' prognosis. Te PI3K/AKT pathway is identifed as the key pathway. Hub compounds demonstrated a good afnity for the three key targets. To combat NSCLC, YFSJF may inhibit PIK3R1, PIK3CA, and AKT1 to suppress the PI3K/AKT pathway. Tis study provides a solid foundation for future YFSJF mechanism research.

Data Availability
All the results used to support this work are included in the article. More details were deposited in the Science Data Bank (DOI: 10.11922/sciencedb.01534).

Conflicts of Interest
Te authors declare that there are no conficts of interest.