First, alkaloids from the literature were compared in the Traditional Chinese Medicine Systems Pharmacology Database (tcmsp-e.com/index.php). Alkaloids with an OB value of ≥30% were selected for the next prediction and their structures were retrieved from the PubChem (pubchem.ncbi.nlm.nih.gov/) database. The Simplified Molecular Input Line Entry System (SMILES) representations of these alkaloids were subsequently imported into the Swiss Target Prediction database (swisstargetprediction.ch), where only targets with prediction probability scores >0.1 were retained as potential drug targets. Subsequently, the key words ‘non-small cell lung cancer’ were searched in the Genecards database (https://www.genecards.org) to identify relevant targets. Based on the binary relationships between NSCLC and its targets, a target network containing 5,388 nodes was constructed. Finally, the Cytoscape (cytoscape.org/, version 3.8.2) was applied to create a combined ‘drug-component-target-disease’ network diagram.

In total, 493 potential targets of 139 alkaloids and 5,387 NSCLC-related targets were imported into the Venny (bioinfogp.cnb.csic.es/tools/venny/, version 2.1.0) software, and the intersection of the two datasets was evaluated, which led to the identification of 304 common targets. These targets were then input into the STRING database (https://cn.string-db.org/) for analysis, with the organism restricted to ‘Homo sapiens’. The minimum interaction threshold was established at 0.4, resulting in the generation of target protein networks. The data were subsequently imported into Cytoscape (version 3.8.2) to construct a PPI network map.

The PPI subnetwork was analyzed using the MCODE (version 2.0.3) clustering algorithm (apps.cytoscape.org/apps/mcode); Cytoscape (version 3.8.2) and the MCODE plug-in were used for network topology analysis with the following parameters: Degree cut-off=2, a node score cut-off=0.2, a K-core of 2 and a maximum depth=100. Network topology parameters (degree, betweenness and closeness centrality) for these targets were obtained using the Analysis Network tool in Cytoscape. The CytoNCA (version 2.1.6; cytoscape.org/apps/cytonca) and CytoHubba (version 0.1) plugins (apps.cytoscape.org/apps/cytohubba) were used to identify HUB genes. Finally, a Venn diagram was constructed to identify the overlapping targets.

The core subnetworks derived from the clustering analysis of the PPI network were analyzed using the STRING database and DAVID databases (david.ncifcrf.gov/summary.jsp), Reactome (reactome.org/) and WikiPathways (wikipathways.org/) databases for the enrichment analysis (Gene Ontology (GO; geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG; kegg.jp/)). The enrichment results were organized by intensity, gene count (≥15) and P≤0.05, ranked from high to low scores, with the top 20 selected for visualization.

Imperialine, verticinone, verticine and peimisine were purchased from Chengdu Push Bio-Technology Co., Ltd., imperialine-3-β-D-glucoside and delavine were obtained from Chengdu Herbpurify Co., Ltd., and ebeiedinone and delavinone from Chengdu Must Bio-Technology Co., Ltd. The purity of all compounds was >98%. The Annexin V-FITC/propidium iodide (PI) apoptosis detection kit was acquired from Dojindo Laboratories, Inc. The BCA protein kit and RIPA buffer were from Beyotime Institute of Biotechnology and the protease inhibitor cocktail was from Roche Diagnostics. Phenylmethylsulfonyl fluoride and phosphatase inhibitor were purchased from Beijing Solarbio Science & Technology Co., Ltd. Chloroquine (CQ) and dimethyl sulfoxide were obtained from MilliporeSigma, and the Cell Counting Kit-8 (CCK-8) from Beijing Solarbio Science & Technology Co., Ltd. Antibodies specific for PARP (cat. no. CY5347) and cleaved PARP (cat. no. CY5265) were obtained from Shanghai Abways Biotechnology Co., Ltd.; antibodies for cleaved caspase-3 (cat. no. 9664T), LC3A/B (cat. no. 12741), GAPDH (cat. no. 2118) and HRP-conjugated goat anti-rabbit IgG secondary antibody (cat. no. ab6721) were obtained from Cell Signaling Technology, Inc. All other chemical reagents used were of analytical grade.

In total, three non-small cell lung cancer cell lines, NCI-H1975 (CRL-5908), NCI-H1299 (CRL-5803) and A549 (CCL-185), were obtained from the American Type Culture Collection. The cell lines were maintained in RPMI-1640 medium supplemented with 10% (v/v) fetal bovine serum (FBS), 100 U/ml penicillin and 100 µg/ml streptomycin (all Gibco; Thermo Fisher Scientific, Inc. at 37°C in a 5% CO₂ atmosphere, and the cell culture medium was replaced daily.

A single-cell suspension in the logarithmic growth phase was prepared and counted. Subsequently, 100 µl of the single-cell suspension were added to each well of a 96-well plate, with a concentration of 4×10³ cells/well for the A549 cells and 5×10³ cells/well for the NCI-H1299 cells, diluted with complete medium. Control groups included blank control group, negative control group and experimental groups, with each group containing ≥4 replicate wells. Following 48 h of drug treatment, 10 µl CCK-8 regent were added to each well. The wells were then incubated in the dark for 4 h, and the absorbance was measured at 450 nm using a microplate reader (Bio-Rad Laboratories, Inc.).

Early and late apoptosis cells were evaluated using flow cytometry. In brief, NCI-H1299 and A549 cells were treated with ebeiedinone (10, 15, 20 and 30 µM), verticinone (15, 30 µM) and gefitinib (15, 30 µM) for 48 h in a 37°C incubator. A blank control without medication and a 5-fluorouracil group (15, 30 µM) were set up. 5-Fluorouracil was used as a positive control drug for apoptosis to determine whether cells can undergo apoptosis. Gefitinib is used as a positive drug for comparison with Verticinone and ebeiedinone, to determine the ability to induce apoptosis. For the flow cytometry experiments, the cells were stained with 5 µl annexin V-FITC and 5 µl PI at room temperature for 15 min. The fluorescence intensity was measured using a BD FACSCalibur flow cytometer (Becton Dickinson and Company), and the apoptotic rates were analyzed using the NOVAexpress software (version 1.6.2; Agilent).

A549 cells were harvested and lysed using a nuclear and cytoplasmic protein extraction kit (Beyotime Biotechnology; cat. no. P0028,) according to the manufacturer's instructions. Protein concentrations were determined using the BCA protein assay kit. A total of 20 µg of protein sample to each well. Cytoplasmic proteins were separated by 7 and 15% SDS-PAGE and transferred to a polyvinylidene difluoride membrane (MilliporeSigma). The membranes were blocked with 5% skimmed milk at room temperature for 1 h, followed by incubation with primary antibodies at 4°C overnight. This was followed by incubation with goat anti-rabbit IgG (H&L) (cat. no. ab6721; 1:10,000; Abcam, Cambridge, UK) at 37°C for 2 h. Protein bands were visualized using the Immobilon Western HPR Substrate (MilliporeSigma) and images were captured with the GelView 6000Plus Chemiluminescence Imaging System (Guangdong Biolight Meditech Co., Ltd.). The Image J software (version 1.54; National Institutes of Health) was used for protein quantification. The following are the primary antibodies used in this experiment: GAPDH (14C10) Rabbit Monoclonal Antibody (cat. no. 2118; 1:1,000; Cell Signaling Technology); β-tubulin (9F3) Rabbit Monoclonal Antibody (cat. no. 2128; 1:1,000; Cell Signaling Technology, Inc.); PARP 1 Ab (cat. no. CY5347; 1:1,000); Cleaved PARP (cat. no. CY5265; 1:1,000; both Shanghai Abways Biotechnology Co., Ltd.); Cleaved caspase-3 (cat. no. 9664T; 1:1,000; Cell Signaling Technology), LC3A/B (D3U4C) Rabbit Monoclonal Antibody (cat. no. 12741; 1:1,000; Cell Signaling Technology; The United States). p62 (D1Q5S) Rabbit Monoclonal Antibody (cat. no. 39749; 1:1,000; Cell Signaling Technology; The United States)

All data are presented as the mean ± standard deviation. All statistical analyses were performed using GraphPad Prism software (version 7.0; Dotmatics). Statistical analyses were performed as follows: a one-way ANOVA was used for multi-group comparisons, with Tukey's HSD post hoc test applied for further pairwise analysis. The Student's t-test (unpaired) was used for direct comparisons between two groups. P<0.05 was considered to indicate a statistically significant difference.

Following a comprehensive literature search and retrieval of data from PubChem records, the SMILES representations of alkaloids in Fritillaria were imported into the Swiss Target Prediction program to identify drug targets that met the specified criteria (Table SI). Target genes for NSCLC were obtained following retrieval from the GeneCards database. These data were integrated to construct a ‘drug-component-target-disease’ network using Cytoscape (Fig. 1). The degree-based ranking of the top 20 pharmacologically active compounds is summarized in Table I.

A Venn diagram (Fig. 2A) was generated using Venny (version 2.1.0) to identify the intersection between 493 potential targets from 139 alkaloids in Fritillaria and 5,387 disease targets associated with NSCLC. In total, 304 common targets were identified within the NSCLC-related and potential alkaloid target sets. These common targets were subsequently input into the STRING database to retrieve target protein networks, and the results were imported into the Cytoscape software (version 3.8.2) to create a drug target PPI network diagram (Fig. 2B). Cytoscape (version 3.8.2) and the MCODE plug-in were used to analyze the topology of the PPI network. The results revealed several targets with high degree values within this PPI network, indicating a greater number of interactions with other proteins. The top 20 targets, with degree values ranging from 80 to 170, are illustrated in Fig. 2C.

In the MCODE cluster analysis, the PPI network was divided into 11 subnetworks (Table II), of which four contained >10 nodes (Fig. 3A and B). In addition, further CytoHubba analysis was performed on the PPI network, yielding the top 20 targets with the highest scores (Fig. 3C) and CytoNCA analysis was also conducted. Finally, the intersection of the enriched Cluster 1, CytoNCA and CytoHubba-related targets was identified using a Venn diagram, yielding 17 common targets (Fig. 3D), namely: MDM2, EP300, PIK3CA, STAT3, ERBB2, EZH2, GSK3B, HSP90AA1, MTOR, PARP1, JUN, AKT1, IGF1R, ESR1, SRC, BCL2L1 and JAK2. These targets are associated with tumor apoptosis, tumor invasion and immune evasion (26–29). Among these, BCL2 is a typical anti-apoptotic protein in the apoptotic pathway, providing insights for subsequent experimental validation (30).

The core subnetworks derived from the PPI network clustering analysis were examined using the STRING database for enrichment using GO, KEGG, Reactome and WikiPathways (Fig. 4). The results were organized in descending order based on the enrichment intensity and P-value.

In the GO enrichment analysis, cluster 1 was found to be enriched in 1,026 biological process terms and 104 molecular function terms (P≤0.05). Additionally, a total of 58 cellular component terms and the top 20 enriched pathways are presented in Fig. 4A-C. KEGG pathway enrichment analysis revealed the enrichment of 151 pathways in cluster 1, showing association with ‘Pathways in cancer’, ‘PI3K-Akt signaling pathway’, ‘EGFR tyrosine kinase inhibitor resistance’ and other tumor-related pathways ranked highest in the enrichment index (Fig. 4D).

Reactome pathway enrichment analysis demonstrated the enrichment of 281 pathways in cluster 1, with numerous tumor-associated pathways ranked highly in the enrichment index. These include pathways related to ‘AKT-mediated inactivation of FOXO1A’, ‘PTK6 Regulates cell cycle’, ‘signalling to STAT3’, ‘MET activated STAT3’ and other tumor signal transduction pathways (Fig. 4E). The WikiPathway enrichment analysis revealed the enrichment of 263 pathways in cluster 1. Among these, the ‘Robo4 and VEGF signaling pathways crosstalk’, ‘TCA cycle nutrient use and invasiveness of ovarian cancer’, ‘serotonin receptor 2 and STAT3 signaling’ and various other tumor-related pathways were highly ranked (Fig. 4F).

Based on the results of pathway enrichment analysis using different methods, it was preliminarily hypothesized that the active components in Fritillaria cirrhosa may exert certain effects on NSCLC. Combined with the results of core target enrichment analysis, it was considered that these effects may be achieved by promoting tumor cell apoptosis and autophagy and inhibiting immune escape, providing a basis for further verification.

The enrichment analysis suggested that the anticancer activity of alkaloids is closely related to the regulation of autophagy and apoptosis. Previous reports have indicated that alkaloids exert their anticancer effects through anti-inflammatory feedback (31,32). Furthermore, our previous research has shown that imperialine, verticinone, verticine, imperialine-3-β-D-glucoside, delavine, and peimisine can inhibit cigarette-induced oxidative stress and inflammatory responses in RAW264.7 cells (33). Based on the aforementioned network pharmacology results, this study ultimately selected imperialine, verticinone, verticine, imperialine-3-β-D-glucoside, delavine, peimisine, ebeiedinone, and delavinone as the subjects of investigation (Fig. 5). The effects of these eight distinct isosteroid alkaloids (Fig. 5A-H) derived from Fritillaria on the viability of three types of NSCLC cells were evaluated using CCK-8 assay. No significant effects on the viability of NCI-H1975 cells were observed for the majority of the alkaloids (Fig. 5I). By contrast, the survival rates of the NCI-H1299 and A549 cells treated with verticinone and ebeiedinone decreased to 75.57±4.80% and 68.9±3.06% (at 15 µM), and 65.75±6.44% and 67±3.20% (at 30 µM), respectively (Table SII, Table SII, Table SIII, Table SIV). Based on these findings, verticinone and ebeiedinone were selected for further analyses.

To investigate the potential mechanisms underlying the alkaloid-induced decrease in cell viability, cell death was detected using the Annexin V-FITC/PI dual labeling assay. The number of Annexin V-FITC-positive NCI-H1299 cells was comparable between the groups (Fig. 6). The group treated with gefitinib exhibited rates of 10.42 and 19.79% at concentrations of 15 and 30 µM, respectively (Fig. 6A-G). The number of Annexin V-FITC-positive A549 cells increased following treatment with ebeiedinone (Fig. 7A-G). Subsequently, experiments with a range of ebeiedinone concentrations were conducted. The results indicated that ebeiedinone promoted apoptosis in a concentration-dependent manner (Fig. 8A-H). Furthermore, analysis revealed that cells treated with dehydro-eudobelin (30 μM) exhibited significant differences compared to the blank control group, indicating that dehydro-eudobelin has the ability to induce apoptosis (Fig. 8I).

Apoptosis is a key biological process in multicellular organisms, characterized in part by the activation of caspases and the induction of caspase-dependent cell death. Caspases are regarded as the central mediators of apoptosis (34). PARP, along with genomic stability factors and DNA repair enzymes, was originally identified as a key proteolytic substrate that is degraded by caspase-3 and other cysteine proteases during apoptosis (35). Increased levels of cleaved PARP and cleaved caspase-3 in the cells exposed to ebeiedinone were demonstrated in the present study (Fig. 9). The increase in the level of cleaved caspase-3 was observed across various drug concentrations and was comparable to the effect induced by the positive control drug, gefitinib (Fig. 9A and C). Protein expression levels of cleaved PARP/PARP and cleaved capase-3/GAPDH significantly increased (Fig. 9B and D). In summary, these findings demonstrated that ebeiedinone significantly stimulated the apoptosis of NSCLC cells.

Subsequently, it was investigated whether ebeiedinone induces autophagy in A549 cells by examining the expression of LC3-II, a protein marker of autophagy (Fig. 10). Ebeiedinone induced LC3-II protein expression in a concentration-dependent manner in A549 cells. Furthermore, when ebeiedinone was used in combination with the autophagy inhibitor, CQ (10 µM, 1 h), which disrupts lysosomal function to block autophagy (36), LC3-II expression was further upregulated (Fig. 10A and C). Additionally, the effects of ebeiedinone on p62 protein expression levels were assessed, which demonstrated that the ebeiedinone reduces p62 expression, thereby promoting cellular autophagy (Fig. 10B and D). These results further indicated that ebeiedinone could promote autophagy in A549 cells.