Luteolin inhibits A549 cells proliferation and migration by down-regulating androgen receptors

Lung cancer-related genes were collected from GeneCards. The chemical constituents and target genes of 13 Chinese herbal medicines in YQG were screened using the TCM System Pharmacology Database (TCMSP) (https://​tcmsp-e.​com/​). The screening conditions were oral bioavailability (OB) of at least 30% and drug-like (DL) degree of at least 0.18. In addition, UniProt was used to verify the gene symbols of the genes screened in TCMSP and to construct a dataset of Chinese medicine chemical components and targets.

R software (version 3.6.3) was used to calculate the overlapping genes between lung cancer-related genes and known YQG lung cancer treatment target genes to construct a protein–protein interaction (PPI) network. The STRING database (version 11.0) was then used to analyze the interaction of target genes in the network. To ensure the robustness of the results, when the interaction score was greater than or equal to 0.4, the network type was selected as a physical subnetwork, and the Cytoscape software program (version 3.7.1) was used for analysis. The top ten proteins with three-dimensional structure with the strongest interaction between proteins were selected as the main target proteins (hub proteins).

We downloaded the corresponding three-dimensional structure diagram of the active compounds and the proteins related to the hub genes from PubChem and the Protein Data Bank, respectively, and then used Discovery Studio (DS) 2019 software program (Dassault Systèmes, Vélizy-Villacoublay, France) for molecular docking and absorption, distribution, metabolism, and excretion (ADME) prediction.

Using the method of machine learning, the hub proteins and the top 10 multi-target active components screened were analyzed based on QSAR. Download data: Use the ChEMBL (https://​www.​ebi.​ac.​uk/​chembl/​) database to obtain the biological activity data of drug molecules that act on hub proteins as a training data set for modeling. The SMILES sequences of the top 10 multi-target active ingredients were obtained from the PubChem (https://​pubchem.​ncbi.​nlm.​nih.​gov/​) database. Descriptor calculation and feature selection: Use the PaDEL-Descriptor library in python to calculate the molecular descriptors of compounds. The input to the PaDEL-Descriptor software is the SMILES sequence of the molecule, and then features with low variance are removed using the python sklearn library feature selection method Variance Threshold. Model selection: Use the LazyRegressor function in python’s lazypredict library for automatic training, compare 41 machine learning models, and use the Adjusted R-squared, R-squared, and root mean square error (RMSE), time taken four criteria to evaluate the effect of the model, select the model with the best evaluation results to establish the QSAR model of the effective Chinese medicine monomer in the treatment of lung cancer. Use the four indicators of explained_variance_score, mean_absolute_error (MAE), mean_squared_error (MSE), and R-squared value (r2_score) provided by the sklearn library metrics to judge the reliability of the regression model. Finally, the descriptors of TCM monomer molecules are input into the model to predict the biological activity value.

The human lung cancer cell lines PC9, A549, and H460 were provided by the Clinical Research Center of Jiading District Central Hospital. They were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (Gibco Fisher Scientific, Beijing, China) containing 10% fetal bovine serum (Gibco) at 37 °C, with 5% CO₂ and 95% air humidity, plus 100 U/mL of penicillin and 100 U/mL of streptomycin (Gibco).

The cell counting kit 8 (Dongren Chemical Technology Co., Ltd., Japan) colorimetric method was used to detect the effect of luteolin (Chengdu Herbpurify Co., Ltd., Chengdu, China) on the proliferation of A549, PC9, and H460 cells and the effect of luteolin on A549 cells after mutation of androgen receptor phosphorylation sites at the binding site. The cells were inoculated in a 96-well plate (1 × 10⁴ cells/well), incubated alone at 37 °C for 24 h, then incubated with different doses of luteolin in a 5% CO₂ incubator at 37 °C for 48 h. Luteolin (80 µM) was applied to A549 cells after transfection of AR wild-type and mutant-type plasmids for 24 h. Next, 10 µL of CCK-8 was added to each well, followed by incubation for one hour. The optical density value was measured at 450 nm using a microplate reader (Thermo Fisher Scientific, MA, USA). Each experiment was repeated three times.

A549 cells were inoculated in a six-well plate (3 × 10⁵ cells/well), and a scratch test was performed after 24 h of attachment, where we added 80 μM luteolin and took pictures of scratches at 0, 24, and 48 h, respectively. Each experiment was repeated three times. Luteolin (80 μM) was applied to untransfected A549 cells (Control), A549 cells transfected with empty vector plasmids (NC), AR wild-type (AR-WT) and mutant-type plasmids (AR-MU) for 24 h, and scratch pictures were taken at 0, 24, and 48 h. The cell migration rate was calculated using Image J (version 1.52a; National Institutes of Health, Bethesda, MD, USA), and the GraphPad Prism software program was used for statistics and graphing.

A lactate dehydrogenase (LDH) cytotoxicity detection kit was purchased from Beyotime Biotechnology (Shanghai, China). The 2 × 10³ A549 cells per well were inoculated in a 96-well plate. After 24 h, the cells adhered to the wall and the medium was removed. Cells were washed once with PBS and serum-free medium containing different concentrations of luteolin was added to the cells for 47 h. Then, LDH release reagent was added, everything was mixed by pipetting several times, and further incubation was completed in the presence of luteolin effect lasting for 48 h. Next, 120 μL of the supernatant was collected from each well and added to a new 96-well plate, and the sample was measured at an optical density of 490 nm.

A549 cells were seeded on a 6-well culture plate for 12 h, treated with different concentrations of luteolin for 48 h and collected. The cells were resuspended with PBS (4 °C) and centrifuged at 2000 rpm for 5–10 min more, and then washed. According to the instructions for the production of the Annexin V-fluorescein isothiocyanate/propidium iodide (Annexin V‐FITC/PI) Apoptosis Kit (BD Biosciences, CA, USA), the cells were suspended with 500 μL of binding buffer, and 5 μL of Annexin V-FITC and PI were added and incubated at room temperature and in the dark for 15 min. Cell apoptosis was detected by flow cytometer (BD FACS Canto II).

AR wild-type plasmid was constructed by primers F and R to amplify the target gene. The primer sequences are listed in Table 1. PCR amplification conditions were as follows: 66 °C annealing temperature, 68 °C extension temperature for 1 min and 30 cycles. The DNA fragment was collected and digested by the EocR I and BamH I (Takara Bio, Shiga, Japan) simultaneously. The vector pLenO-GTP-C-3X flag (Zorin Biological Technology Co., Ltd., Shanghai, China) was digested with the same condition and ligated with the target fragment by Fusion DNA ligase.

AR mutant expression plasmids (AR-MU) were constructed using the point mutation PCR method. Using primers F and R-MUT to amplify the region fragments, and primers F-MUT and R to amplify the region fragments, both DNA fragments were mixed and used as templates for PCR amplification, and primers F and R are used to amplify the fragments of the target gene. The PCR amplification conditions were the same as with the AR wild-type plasmid.

A549 cells were inoculated in a six-well plate (3.5 × 10⁵ cells/well). Then, we mixed the empty vector, wild-type (WT) AR, and the mutant AR plasmids with lipofectamine 2000 (Thermo Fisher Scientific) at a ratio of 5 μg to 8 μL, and transferred them into different wells of the six-well plate. Next, 2 mL of serum-free and dual-antibody-free DMEM was added to each well. After 4 h, the medium with serum and non-double antibody was replaced. Twenty-four hours after transfection, transfection was observed under a fluorescence microscope (Leica Camera, Wetzlar, Germany).

A549 (7 × 10⁵/well) cells were seeded in a six-well plate overnight. After the cells adhered, DMEM complete medium containing luteolin 40 μM or 80 μM was added to three wells of the six-well plate, and treated for 48 h. Cells were collected after rinsing and scraping with PBS. Cells were lysed on ice with lysis buffer (Beyotime Institute of Biotechnology, Shanghai, China). The concentration of protein was quantified using a BCA kit (Tiangen Biotech, Beijing, China). An equivalent amount of protein was loaded on 10% SDS-PAGE gels and then transferred to nitrocellulose membranes. The membranes were blocked with 5% non-fat milk for an hour, and then incubated with primary antibodies (1:1000) at 4 ℃ overnight. The antibodies were as follows: GAPDH (YOBIBIO, Shanghai, China), AR (Cell Signaling Technology, MA, USA). On the next day, the membranes were washed three times (5 min each time) with TBST solution, and incubated with secondary antibody (1:5000) for 2 h. After three washes with TBST (10 min each time), images were acquired using a gel imaging system (Thermo Fisher Scientific, Waltham, MA, USA).

All animal studies were approved by the Ethics Committee of Shanghai University of Medicine and Health Sciences (Shanghai, China). Male Balb/c nude mice 6–7 weeks of age were purchased from Daoke Medical Technology Co., LTD (Shanghai, China). 1 × 10⁷ A549 cells in PBS were injected subcutaneously to the right side of the back of each nude mouse. When the tumor volume reached ∼84 mm³, the mice were randomly divided into four groups according tumor volume (ten mice per group). In the control group, 10 μl/g 10% DMSO/corn oil (vehicle) was given through intragastric administration. Mice in the experimental groups were given 150 mg/kg luteolin oral gavage, 1 mg/kg 5α-Dihydrotestosterone (DHT, Selleck, USA) (0.2 mg/ml) subcutaneous injection, and 150 mg/kg luteolin combined with 1 mg/kg DHT subcutaneous injection). All treatments were observed every 2 days for 28 days. The weight of the experimental animals was measured about once a week after inoculation until before grouping, and the tumor volume was measured once a week when the tumor was visible. After grouping, the weight of experimental animals and the tumor volume of experimental animals were measured twice a week. The mice were euthanized 1 days after the last dose and tumors were removed and photographed. Tumors were fixed in 4% paraformaldehyde for 12 h, embedded in paraffin, and cut into 5-μm sections. The sections were assigned for hematoxylin–eosin (HE) stain and immunohistochemistry (IHC) assay and examined by light microscopy. For IHC, anti-AR (1:500; Cell Signaling Technology, MA, USA) antibodies were used.

All of the experiments were repeated three times. Results are expressed as mean and standard deviation (SD) values, and all of the statistical comparisons were evaluated using one-way or two-way analysis of variance. Significant differences between means were measured using Duncan’s multiple range test. Statistical analysis was performed using GraphPad Prism software (version 8.0.2) and R-studio (R version 3.6.3). P < 0.05indicated that the difference was statistically significant.

182 compounds of 13 Chinese herbal medicines were identified as “qualified” by a database search and literature search. A total of 232 genes were predicted as possible targets for YQG. Meanwhile, 314 lung cancer genes with lung cancer-related gene scores above 40 points were screened by Gene Cards to create a lung cancer gene dataset. In total, there were 61 interacting genes that overlapped with each other (Fig. 1A), and an interaction network of 61 proteins was constructed by using STRING version 11.0 (Fig. 1B). The interaction diagram of these 61 interacting genes with YQG active components was drawn by using the Cytoscape and R software programs (Fig. 1C). The ten genes with the strongest protein interaction associations (STAT3, EGFR, ESR1, JUN, CCND1, STAT1, AKT1, AR, MYC, and VEGFA) were selected as hub genes. Since CAV1 has no three-dimensional protein structure, DS was unable to conduct molecular docking, so it was temporarily excluded (Fig. 1D). In order to explore the interaction between the ten YQG-related hub genes, another PPI network was constructed (Fig. 1E), and the ten hub gene expression proteins were associated with the active components related to YQG (Fig. 1F). In this network, the red outer circle seen in Fig. 1F represents the compound, the green represents the proteins corresponding to the 10 hub genes, the yellow represents the disease, and the blue represents the drug. An ADME prediction of the 91 hub gene-related compounds was performed using DS. Following the above screening, we found that quercetin, luteolin, kaempferol could dock with six, five and four proteins expressed by hub genes, respectively (Table 2), and are multi-target active compounds. Meanwhile, ADME predicted that luteolin had better oral absorption and utilization efficiency. Bioactivity values predicted by QSAR model (Table 3) showed that luteolin had higher bioactivity values among the top ten multi-target compounds. Therefore, luteolin was selected as the active compound for subsequent experimental verification. Combined with the results of molecular docking, QSAR model validation and literature review, we chose luteolin as the active compound for the next step of molecular biology experiment validation.

A CCK-8 assay was performed to examine the inhibition of A549, PC9, and H460 cell proliferation by luteolin (0 μM, 5 μM, 10 μM, 20 μM, 40 μM, and 80 μM). As seen from the line chart (Fig. 2A–F), the action of the drug is not stable at 24 h, but the drug level reached an apparent steady state by 48 h. The CCK-8 results suggested that luteolin demonstrated strong inhibitory activity on the proliferation of A549 cells in a dose-dependent manner at 48 h. The concentration of dimethyl sulfoxide (DMSO) was 2.29 μL/mL, which was required for the dissolution of luteolin. As shown in Fig. 2A–F, DMSO had no effect on cell proliferation at this concentration. Among the three lung cancer cell lines, compared with the gefitinib mutant lung cancer cell line PC9 and the large cell lung cancer cell line H460, the wild-type lung adenocarcinoma cell line A549 showed a better concentration dependence on luteolin. As A549 cells have a broader representative significance in lung cancer, follow-up experiments mainly focusing on A549 to screen the docking site and mechanism of luteolin in A549 are warranted.

The LDH release experiment was used to detect the effect of luteolin at different concentrations on the lung cancer cell line A549 (because A549 cells have a better concentration dependence on luteolin inhibition of their proliferation), and the results showed that there was no significant difference in the release of luteolin at different concentrations (0 μM, 5 μM, 10 μM, 20 μM, 40 μM, and 80 μM) (Fig. 2G). The results indicated that the toxicity of luteolin to lung cancer cells was not obvious at these concentrations. (C/NP represented the negative control and C/NP2 represented the positive control.)

Cell scratch test was used to observe the migration of A549 cells with 80 μM Luteolin. We observed that after treatment with luteolin at 80 µM, the migration of A549 began to be inhibited after 24 h, and the inhibition effect was more significant at 48 h (Fig. 2H).

FCM analysis was used to further verify the pro-apoptotic effect of luteolin on A549 cells. It can be seen from Fig. 3 that, when luteolin was used for 48 h, with concentrations of 40 μM and 80 μM, no obvious apoptosis-inducing effect was seen. When luteolin was used in combination with DDP (10 μg/mL), however, the concentration of 80 μM could induce a significant increase in lung cancer cell apoptosis (**P < 0.01) (vs. 10 μg/mL of DDP alone and 10 μg/mL of DDP with 40 μM of luteolin, respectively).

We used the DS proteomic analysis software to calculate the three-dimensional structure diagram of luteolin and AR binding as well as the sites of their stable interaction (Fig. 4A and B). In this context, THR 877, the binding site of luteolin to AR protein, mutates to ALA, resulting in phosphorylation inactivation (Fig. 4C). A549 cells were transfected with the plasmids containing green fluorescent protein, and the transfection efficiency was observed 24 h later (Fig. 5). At the same time, 80 μM of luteolin administered to A549 cells was transfected with AR wild-type and mutant-type plasmids. The results of scratch images at zero, 24, and 48 h revealed that the inhibitory effect of luteolin on cell migration was weakened following phosphorylation site mutation and inactivation (Fig. 6A and B). Meanwhile, CCK-8 experiment showed that luteolin had a certain inhibitory effect on cell proliferation following the transfection of AR mutation (phosphorylation site inactivation) plasmid, but its inhibitory ability was weaker than that of the control group (Fig. 6C).

Western Blot showed that AR protein expression was different after treating A549 cells with different concentrations of luteolin for 48 h: When luteolin was 80 μM, AR protein expression was significantly decreased (Fig. 7A and B).

To further evaluate the anti-tumor effects of luteolin, we engrafted A549 cells subcutaneously into nude mice. After the establishment of solid tumors, luteolin, DHT, luteolin + DHT or vehicle was administered every 2 days, respectively. We found that, compared with the control group, luteolin significantly inhibited the growth of A549 xenogeneic tumors at day 25 (P < 0.037) and 28 (P < 0.046) of administration, and compared with the DHT group, luteolin group also showed significant differences at day 25 (P < 0.046) and 28 (P < 0.035) of administration (Fig. 7C and D, Table 4). In addition, we examined the expression of AR in subcutaneous transplanted tumor of mice by immunohistochemistry. The level of AR positive staining was significantly reduced in the luteolin-treated group compared to the control group (Fig. 7E and F). Our results showed that luteolin suppresses tumor proliferation and AR expression.