Two crude drugs of HMMCR were obtained from the Pharmacy Department of Dongfang Hospital, Beijing University of Chinese Medicine (Beijing, China). Hedysarum Multijugum Maxim (Huangqi, batch No. 20210412) were collected from the province of Neimenggu, China; Curcumae Rhizoma (Ezhu, batch No. 19062001) were collected from the province of Guangxi, China. The quality of each crude drugs was strictly ensured according to Chinese Pharmacopoeia to guarantee quality control (20). HMMCR aqueous extract was prepared according to the following experimental steps: 30 g of HMM and 15 g of CR were soaked in 10 volumes of distilled water for 30 min; then, the mixture was decocted for 1 h, and the resulting supernatant was collected. This extraction procedure was repeated twice, the supernatants obtained from two decoctions were mixed and evaporated to 45 ml, and the HMMCR with a concentration of 1 g/ml was prepared. Finally, the solution was passed through a filter with a 0.22-μm pore diameter and stored at −20°C.

Compounds of HMMCR were analyzed using liquid chromatography–mass spectrometry (LC-ESI-MS/MS) system (UHPLC, Thermo Fisher U3000 ultrahigh performance liquid chromatograph; MS, Thermo Scientific Q Exactive Plus™ Orbitrap MS system). The identification of unknown compound was performed by Compound discover 3.1, with mzcloud and mzVault databases. SwissTargetPrediction (http://www.swisstargetprediction.ch/), a public database for ligand-based target prediction of small biologically active molecule, was used to predict the potential targets of HMMCR. We transformed each compound identified by LC-ESI-MS/MS into canonical SMILES through the PubChem (https://pubchem.ncbi.nlm.nih.gov/). Subsequently, these SMILES were imported into SwissTargetPrediction to predict targets of compounds. Species were selected as “Homo sapiens” with probability >0 as the screening condition.

The keyword “non-small cell lung cancer” was utilized to collect disease-related targets from the following databases: the GeneCards database (http://www.genecards.org/), Online Mendelian Inheritance in Man (OMIM, http://www.omim.org/) databases, Therapeutic Target Database (TTD, https://db.idrblab.org/ttd/), and PharmGkb (https://www.pharmgkb.org/). The overlapped targets among putative targets of the chemical compounds of HMMCR and NSCLC-related targets were considered as the potential targets of HMMCR for non-small cell lung cancer, which was visualized with a Venn diagram. Cytoscape 3.9.0 (https://cytoscape.org/) was applied to establish HMMCR active compounds–NSCLC targets network.

PPI network was constructed by the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING; https://string-db.org/), using the overlapping targets among targets of HMMCR active ingredients and NSCLC-related targets, with the species limited to “Homo sapiens” and minimum required interaction score selected as a highest confidence score (0.900). The topological property of each node in the PPI network was evaluated by calculating six parameters with a Cytoscape plugin CytoNCA: “betweenness centrality (BC)”, “closeness centrality (CC)”, “degree centrality (DC)”, “eigenvector centrality (EC)”, “network centrality (NC)”, and “local average connectivity (LAC)”. Subsequently, nodes above twofold median values were selected as key targets and the hub nodes of HMMCR on NSCLC were constructed.

Gene ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were conducted in R (version: 3.6.3) using the “ClusterProfiler” package, and p. adjust (FDR) <0.05 was considered statistically significant.

The following reagents were used: Cell Counting Kit (CCK-8) assay solution and Cell Cycle Detection Kit (KeyGEN BioTECH, Beijing, China); Bicinchoninic Acid (BCA) Protein Assay Kit and bovine serum albumin (BSA), RIPA cell lysis buffer, and penicillin-streptomycin solution (Beyotime Institute of Biotechnology, Shanghai, China); Fetal bovine serum (FBS, Gibco, Grand Island, USA); SuperSignal Chemiluminescent HRP Substrate (Thermo Fisher scientific Inc., Rockford, USA); Trypsin-EDTA (Invitrogen, Rockville, USA); Dulbecco’s minimum essential medium (DMEM, Corning Incorporated, New York, USA); antibodies against AKT1, p-AKT, PI3K, E-cadherin, N-cadherin, and β-actin (ProteinTech Group, Chicago, USA).

Human NSCLC cell lines NCI-A549 and NCI-H460 were kindly gifted by Professor Zongfang Zheng from Peking University Health Science Center. Cells were cultured in DMEM medium supplemented with 10% fetal bovine serum (FBS) and 100 U/ml of penicillin–streptomycin solution and maintained at 37°C in a humidified incubator with 5% CO₂.

Cell viability was explored using CCK-8 solution. Briefly, A549 cells and H460 cells (5×10³ cells/100μl/well) in 96-well plates were exposed to HMMCR at different concentrations (0, 3, 6, 12, 24, or 48 mg/ml) for 24, 48, and 72 h. Ten microliters of CCK-8 reagent was added to each well, and cells were incubated for 2 h at 37°C and 5% CO₂. A microplate reader was used to measure the optical density (OD) at 450 nm of each well.

Cells (2,000 cells/well) in six-well plate were cultured overnight before being exposed to different treatment conditions for 48 h and cultured in complete medium for additional 12 days. Colonies were fixed for 15 min at room temperature in methanol and stained for 15 min in crystal violet.

Wound healing assay was used to determine the migration of A549 and H460 cells. Briefly, a straight line was marked using a marker pen on the back of the well before cells (4×10⁵ cells/2 ml/well) were seeded in six-well plates. After cells were grown to 90% confluence, the cell was scraped lightly with the 200-μl tip of the spear to draw another straight line, perpendicular to the marking line on the back of each well at the center. Subsequently, cells were then rinsed with sterile phosphate-buffered saline (PBS) for three times and cultured with serum-free DMEM medium. The wound healing was observed and photographed at the time point of 0 and 24 h with different concentrations (0, 6, 12, or 24 mg/ml) of HMMCR. Image J Software was used to calculate the migrated distance.

Total protein was extracted from A549 and H460 cells challenged with different concentrations (0, 6, 12, or 24 mg/ml) of HMMCR for 48 h. RIPA Cell Lysis Buffer was used to lyse cells on ice for 30 min, followed by centrifuged at 15,000 g at 4°C for 20 min. Bicinchoninic acid (BCA) method was used to detect the protein concentration. Proteins (25 μg/sample) were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene fluoride (PVDF) membranes. After blocking in 5% BSA for 1 h at room temperature, the membranes were blotted with primary and secondary antibodies, respectively, then detected by the Enhanced Chemiluminescence Detection Kit. ChemiDoc MP Imaging System and the software Image Lab version were applied to acquire images of the Western blot. Image J Software was used to calculate the grayscale value.

All results shown in the current study are expressed as mean ± standard deviation (SD). SPSS 21.0 software was used for statistical analysis. t-Test was used for differences between the two groups. Differences between multi-groups were assessed by one-way ANOVA. *p < 0.05, **p < 0.01, and ***p < 0.001 indicated statistically significant difference.

HMMCR aquatic extract samples were analyzed by LC-ESI-MS/MS. A total of 181 compounds of HMMCR were detected ( Supplementary Table S1 ). The secondary mass spectra of each compound are shown in Supplementary Table S2 . SwissTargetPrediction was used to predict the targets of each compounds. Twenty-eight components (MOL3, MOL7, MOL42, MOL47, MOL49, MOL57, MOL63, MOL68, MOL74, MOL81, MOL84, MOL92, MOL95, MOL98, MOL99, MOL109, MOL114, MOL115, MOL141, MOL156, MOL157, MOL158, MOL161, MOL164, MOL165, MOL166, MOL170, and MOL178) without canonical SMILES or targets were eliminated. SwissTargetPrediction predicted a total of 13,896 potential targets of the rest of 153 compounds, and we obtained 1,255 after removing duplicate targets ( Supplementary Table S3 ).

A total of 5,977 NSCLC-related targets were identified after comprehensive searching from GeneCards, OMIM, TTD, and PharmGkb databases ( Supplementary Table S4 ), and the Venn diagram of NSCLC targets is shown in Figure 2A . Furthermore, a Venn plot showing the intersections of HMMCR and NSCLC-related targets is shown in Figure 2B . The result shows that a total of 756 targets of HMMCR were overlapped with those of non-small cell lung cancer, which indicated that HMMCR has a strong therapeutic effect on NSCLC, and its anti-cancer mechanism is worthy of further exploration. In addition, the HMMCR active compounds–NSCLC targets network was further constructed, which included 909 nodes and 9,186 edges, indicating that HMMCR exerts anti-NSCLC effects through multiple ingredients and targets ( Figure 3 ). The green circle represents the ingredients of HMMCR, and the red triangle represents 756 targets. Six key compounds were identified with degree ≥80, namely, hispidulin (MOL155), 5-hydroxy-6,7-dimethoxylflavone (MOL113), genistein (MOL108), (15Z)-9,12,13-trihydroxy-15-octadecenoic acid (MOL14), isokaempferide (MOL55), and naringenin (MOL106). The top 20 components by degree are shown in Table 1 .

The 756 compound–disease co-targets were uploaded to the STRING database. A PPI network contains 629 nodes, and 4,312 edges were established, which indicated that there are high connectivity relationships between these targets ( Figure 4A ). To further explore the hub targets of HMMCR on NSCLC, topological analysis was performed on the PPI network according to six parameters of “BC”, “CC”, “DC”, “EC”, “NC”, and “LAC” and select nodes above twofold median values as hub targets. The threshold values of the first screening were BC > 213.4, CC > 0.05, DC > 8, EC > 0.008, LAC > 2.8, and NC > 3.9, and the results were 184 hub nodes and 2,238 edges. Subsequently, the second screening threshold values were BC > 70.95, CC > 0.49, DC > 19, EC > 0.045, LAC > 8, and NC > 9.11. The results were 62 hub nodes and 789 edges ( Figure 4B ). Therefore, the results of hub targets network indicated that 62 hub targets may account for the crucial therapeutic effects of HMMCR on NSCLC ( Table 2 ).

To further explore the potential therapeutic mechanisms of HMMCR on NSCLC, the 62 hub targets were performed by GO and KEGG pathway enrichment analyses. A total of 2,150 biological process (BP), 101 cellular component (CC), and 171 molecular function (MF) terms were enriched ( Supplementary Table S5 ). The top 10 strongly enriched GO terms in BP, CC, and MF are shown in Figure 5A . The results showed that the terms of BP mainly contained regulation of cell–cell adhesion, positive regulation of cell adhesion, and peptidyl-tyrosine phosphorylation; the terms of CC mainly contained focal adhesion, cell–substrate junction, and extrinsic component of membrane; the terms of MF mainly contained DNA-binding transcription factor binding, RNA polymerase II-specific DNA-binding transcription factor binding, and phosphatase binding. Furthermore, 172 pathways were enriched ( Supplementary Table S5 ). The top 30 significantly enriched pathways of HMMCR on NSCLC are shown in Figure 5B . The results indicated that PI3K/Akt signaling pathway may be the critical pathway of HMMCR in the treatment of NSCLC ( Figure 6 ).

To explore the effect of HMMCR on NSCLC cell proliferation, viable cell numbers were continuously detected by CCK-8 for 72 h when A549 and H460 cells were challenged with different concentrations of HMMCR (0, 3, 6, 12, 24, or 48 mg/ml). Results showed that HMMCR dose- and time-dependently inhibited the viabilities of A549 and H460 cells ( Figures 7A, B ). Meanwhile, the result of colony formation assay demonstrated that HMMCR significantly reduced the colony formation of A549 and H460 cells ( Figures 7C, D ). These above results indicated that HMMCR can significantly inhibit the proliferation of NSCLC cells in vitro.

To further investigate the effect of HMMCR on the NSCLC cells migratory ability, A549 and H460 cells were treated for 24 h with 0, 6, 12, or 24 mg/ml of HMMCR and the effect determined using wound-healing migration assay. Our results indicated that migration of A549 and H460 cells was inhibited by HMMCR in a dose-dependent manner ( Figure 8 ). Therefore, the results indicated that HMMCR can significantly inhibit the migration of NSCLC cells in vitro.

To verify the molecular mechanism of HMMCR for NSCLC, the key protein of PI3K/Akt pathway that predicted by network pharmacological analysis was detected by Western blotting. As shown in Figure 9 , our results showed that PI3K, AKT1, and p-AKT protein levels were significantly reduced in a dose-dependent manner after treatment of A549 and H460 cells with HMMCR. Meanwhile, cell-migratory-related protein (E-cadherin and N-cadherin) were further analyzed. Our data indicated that expression of N-cadherin was decreased and E-cadherin was increased in a dose-dependent manner. The results demonstrated that HMMCR inhibited the proliferation and migration of NSCLC cells at least through the PI3K/Akt pathway, thereby achieving an anti-cancer effect.