AG was purchased from the Tokyo Chemical Industry. A 100 mM stock solution of AG was dissolved in dimethyl sulfoxide (DMSO) and stored at −20°C. AG was diluted to the appropriate concentration with a final DMSO concentration of less than 0.1%. All other chemicals, including DMSO, sodium dodecyl sulfate (SDS), Tween 20, and 3-(4,5-dimethyl-thiazole-2-yl)2,5-diphenyl tetrazolium bromide (MTT), were purchased from MilliporeSigma. All antibodies used in the present study for western blotting are listed in Table I. Roswell Park Memorial Institute (RPMI)-1640 medium, 1X phosphate-buffered saline (PBS; pH 7.4), and 0.25% trypsin-ethylenediaminetetraacetic acid (EDTA) were obtained from Welgene, Inc. Fetal bovine serum (FBS) and penicillin/streptomycin were obtained from Gibco; Thermo Fisher Scientific, Inc.

Human non-small cell lung cancer (NSCLC) cell lines, namely A549 (adenocarcinoma), NCI-H292 (H292; mucoepidermoid carcinoma) and NCI-H522 (H522; adenocarcinoma), were obtained from the Korean Cell Line Bank (Seoul, Korea). Cells were cultured in RPMI-1640 medium supplemented with 10% FBS and 1% penicillin/streptomycin in a 5% CO₂ incubator at 37°C under 98% humidified conditions.

Cell viability was determined via the MTT (dissolved in DMSO) assay. Cells (1×10⁴ cells/well) were initially seeded into a 96-well culture plate in triplicate, followed by treatment with various concentrations of AG (10, 20, 50 and 100 µM) the next day in a fresh medium. After 24 h of incubation, MTT solution (5 mg/ml) was diluted in culture medium and added to each well of the plate to a final concentration of 0.5 mg/ml, and the culture plate was further incubated at 37°C for 4 h. Absorbance was measured at 540 nm using the Spectramax M2 microplate reader (Molecular Devices, LLC).

Lactate production in H292 cells was measured using the lactate fluorometric assay kit (BioVision, Inc.). Briefly, 1×10⁴ cells/well were seeded into a 96-well plate and incubated overnight at 37°C. Prior to pretreatment with various concentrations of AG (20, 50 and 75 µM), the medium was replaced with phenol red and serum-free RPMI medium and incubated at 37°C for 1 h. Finally, 1 µl of the medium from each well was received to measure the absorbance at 570 nm using the Spectramax M2 microplate reader (Molecular Devices, LLC).

After incubation, the cells were harvested and lysed in ice-cold 1% NP-40 lysis buffer containing 150 mM NaCl, 10 mM HEPES (pH 7.45), 5 mM sodium pyrophosphate, 5 mM sodium fluoride, 2 mM sodium orthovanadate, and a protease inhibitor cocktail (Roche Applied Science). Whole cell extracts were obtained from the supernatants after centrifugation of the cell lysate at 13,000 × g for 15 min at 4°C. Protein concentration was quantitated using the Bradford assay. Proteins (50 µg) were separated via SDS-polyacrylamide gel electrophoresis (PAGE; 8–13%) and transferred onto 0.45 µM nitrocellulose membranes (Amersham; Cytiva). The membranes were blocked for 60 min at room temperature using 5% (w/v) BD Difco skim milk powder (Thermo Fisher Scientific, Inc.) and immunoblotted with primary antibodies at 4°C overnight at optimal dilution (Table I). After washing three with Tris-buffered saline with 0.1% Tween 20 buffer, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies at a dilution of 1:2,000 for 1 h at room temperature. Immunoreactivity was detected using ECL Plus (Amersham; Cytiva) and digitalized using Image Quant LAS 4000 (GE Healthcare).

Kinase activity of PDK was slightly modified as previously described (19,20). Briefly, 50 ng of recombinant PDK1 (Abcam) was incubated with 100 ng of recombinant pyruvate dehydrogenase E1 subunit alpha 1 (PDHA1; Abcam) for 30 min at 37°C in PDK1 buffer containing 20 mM Tris buffer (pH 7.5), 0.1 mM EDTA, 1 mM MgCl₂, 2 mM dithiothreitol (DTT) and 250 µM ATP. The samples were subjected to SDS-PAGE and immunoblotted using antibodies against PDK1, p-PDHA1, and PDHA1 (Table I).

Total RNA was isolated from H292 cells using the TRIZOL^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) and quantified at 260 nm wavelength using the Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Inc.), according to the manufacturers' instructions (21). An equal amount of RNA (1 µg) from each sample was used to synthesize cDNA, which was reverse-transcribed using RevertAid reverse transcriptase (cat. no. EP0441; Thermo Fisher Scientific, Inc.). RT was performed according to the manufacturer's instructions. qPCR was performed using the StepOnePlus real-time PCR system (Applied Biosystems; Thermo Fisher Scientific, Inc.) with the RealHelix qPCR kit (cat. no. QP2-S500; NanoHelix Co., Ltd.). PCR was performed with initial denaturation at 95°C for 15 min, followed by 40 cycles of the reaction at 95°C for 20 sec, 60°C for 30 sec, and 72°C for 30 sec. Relative mRNA levels were normalized to actin levels. All the primers used in the present study are listed in Table II.

Apoptotic cells were examined using the Annexin V-fluorescein isothiocyanate (FITC) Apoptosis Detection Kit (Thermo Fisher Scientific, Inc.). Cells (3×10⁵ cells/well) were seeded into a six-well plate and incubated overnight at 37°C. The next day, cells were treated with various concentrations of AG (20, 50, and 75 µM) for 24 h and resuspended in 500 µl of binding buffer (1X) to obtain a cell density of 2×10⁵ cells/ml. The cells were then incubated with 5 µl of Annexin V-FITC and 10 µl propidium iodide (PI; 20 µg/ml) in cell suspension for 15 min at room temperature. Fluorescence intensities of the samples were determined using the FACS Canto II flow cytometer (BD Biosciences). In the histogram of FACS analysis, apoptotic cells were represented the combination of Q2 (PI⁺/Annexin V⁺, late apoptotic cell) and Q3 (PI⁻/Annexin V⁺, early apoptotic cell), and dead cells were represented the combination of Q1 (PI⁺/Annexin V⁻, necrotic cell) and Q2 (PI⁺/Annexin V⁺, late apoptotic cell).

The production of intracellular mitochondrial ROS was determined using MitoSOX Red (Invitrogen; Thermo Fisher Scientific, Inc.). Briefly, 1 µM MitoSOX Red was added to the cultured cells in conditioned medium and incubated at 37°C for 10 min. Fluorescence intensity was determined using a fluorescence microscope (AX10 Imager M1; Zeiss AG) and calculated using the ImageJ software (version 1.53; National Institutes of Health).

Cells (3×10⁵ cells/well) were seeded into a six-well plate and incubated overnight at 37°C. The next day, the cells were treated with various concentrations of AG (20, 50 and 75 µM) for 12 h and incubated with 250 nM tetramethylrhodamine methyl ester (TMRM; Thermo Fisher Scientific, Inc.) for 30 min. The cells were washed thrice with PBS and observed under a fluorescence microscope (AX10 Imager M1). Membrane depolarization was determined by analyzing the captured images using the ImageJ software (version 1.53; National Institutes of Health).

PCR product of PDK1 was ligated into the pMX–IRES-puromycin vector containing EcoRI and XhoI restriction enzyme sites. The primers used to clone the PDK1 construct are listed in Table II. Plat-A cells, a retroviral packaging cell line, were transfected with 1 µg of pMX–IRES empty vector (EV) and pMX–IRES-PDK1 vector (PDK1-OE) using Lipofectamine 2000 (Invitrogen; Thermo Fisher Scientific, Inc.). The supernatant was collected after 24 h, and the culture medium of H292 cells was replaced with filtered retroviral supernatant containing 5 µg/ml polybren (Santa Cruz Biotechnology, Inc.). The cells were selected using 2 µg/ml puromycin after viral infection for 24 h and cultured for an additional 2 weeks.

The known target gene list of AG was downloaded from PubChem (https://pubchem.ncbi.nlm.nih.gov/#query=andrographolide) and analyzed using Cytoscape's JEPPETTO plugin (https://apps.cytoscape.org/apps/jepetto). Kyoto Encyclopedia of Genes and Genomes analysis (https://www.genome.jp/kegg/) identified several cancerous, metabolic and p53 signaling pathways as AG-related pathways. Interaction between genes and chemicals by AG were analyzed using BioCarta Pathways Dataset (https://maayanlab.cloud/Harmonizome/search?t=all&q=andrographolide). The GSE74769 microarray dataset (obtained from the Gene Expression Omnibus database; http://www.ncbi.nlm.nih.gov/geo/) was examined to identify a potential link between glycolysis and AG. The correlation between PDK mRNA expression and AG cytotoxicity in the cells was performed using Spearman's correlation analysis.

All data are from at least three independent experiments and expressed as the mean ± standard deviation. Statistical differences were calculated via one-way analysis of variance with Dunnett's post hoc test or unpaired Student's t-test. GraphPad Prism software (version 5.0; GraphPad Software, Inc.) was used for all statistical analyses, and P<0.05 was considered to indicate a statistically significant difference.

To validate the molecular pathways related to AG, the known target gene list was downloaded from PubChem (https://pubchem.ncbi.nlm.nih.gov/#query=andrographolide) and analyzed using Cytoscape's JEPPETTO plugin (https://apps.cytoscape.org/apps/jepetto). Kyoto Encyclopedia of Genes and Genomes analysis (https://www.genome.jp/kegg/) identified several cancerous, metabolic and p53 signaling pathways as AG-related pathways. BioCarta analysis (https://maayanlab.cloud/Harmonizome/search?t=all&q=andrographolide) revealed that the p53, DNA damage, cell cycle, apoptosis, and HIF pathways are involved in AG treatment (Table SI, Table SII, Table SIII). Pathway analysis revealed no direct evidence of a connection between glucose metabolism and AG. Next, the GSE74769 microarray dataset (obtained from the Gene Expression Omnibus database; http://www.ncbi.nlm.nih.gov/geo/) was examined to identify a potential link between glycolysis and AG. The results clearly revealed that AG treatment downregulated the expression levels of numerous genes involved in pyruvate metabolism. Among them, focus was addressed on PDK1, a key regulator of pyruvate metabolism (22), as its levels were reduced by >50%, similar to the previously reported genes, including HK2, PFK1 and LDHA (Fig. S1) (17,18).

Next, the cytotoxic effects of AG were confirmed on several NSCLC cell lines, namely the A549, H292 and H522 cell lines. It was found that AG caused more severe cytotoxicity in H292 and H522 cells than that in A549 cells (Fig. 1). The GI₅₀ (concentration for 50% of maximal inhibition of cell proliferation) of each cell line was 102.28 µM for A459, 46.17 µM for H292, and 43.59 µM for H522. Examination of the Spearman's correlation between PDK mRNA expression and AG cytotoxicity in these cell lines revealed that PDK1 was most closely related to AG cytotoxicity (Fig. S2), whereas PDK4 expression was lacking in these cell lines. H292 and H522 cells demonstrated higher PDK1 expression than A549 cells at the protein level (Fig. S3A) as well as the mRNA level. Furthermore, AG exhibited strong cytotoxicity in HCT116 and DLD-1 cells with high PDK1 expression (Fig. S3). These findings suggested that PDK1 may be involved in AG-induced cytotoxicity in cancer cells.

Pyruvate, the end product of glycolysis, can be converted to lactate in the cytosol or to acetyl CoA to enter the TCA cycle in mitochondria. Therefore, the effects of AG on metabolic pathways following pyruvate synthesis were verified. Similar to a recent study (17), AG decreased lactate generation in a dose-dependent manner in H292 cells (Fig. 2A). The protein expression of LDHA, an enzyme that catalyzes the conversion of pyruvate to lactate, was unchanged (Fig. 2B). AG significantly decreased the phosphorylation of PDHA1, an enzyme that converts pyruvate to acetyl-CoA (Fig. 2B). Because the phosphorylation of PDHA1 is regulated by PDKs, the protein expression levels of PDKs were examined. As expected, AG induced a decrease in the protein expression levels of PDK1, 2 and 3 in a dose-dependent manner in H292 cells (Fig. 2C). In particular, the expression of PDK1 was markedly reduced. To confirm that AG directly controlled the activity of PDK1, an in vitro PDK activity assay was performed using recombinant proteins for protein binding. However, AG did not directly bind to PDK1 (Fig. 2D). Some binding inhibition was observed at a high concentration of 10 µM or more, which was comparable to the intracellular working concentration (20 µM). The positive control, dichloroacetate, demonstrated a binding inhibitory effect, even at a concentration of 1 mM, which was substantially lower than the intracellular working concentration (10 mM). AG not only decreased the protein expression levels of PDK1, 2 and 3 but also decreased their mRNA levels in a dose-dependent manner in H292 cells (Fig. 2E-G). These findings suggested that AG was involved in the regulation of the mitochondrial metabolic pathways by inhibiting the conversion of pyruvate to acetyl-CoA via a decrease in PDK1 expression.

Metabolic reprogramming of cancer cells from aerobic glycolysis to OXPHOS promotes mitochondrial ROS-dependent cell death (23,24). Thus, it was investigated whether AG induced apoptotic cell death and mitochondrial dysfunction in H292 and A549 cells, wherein different PDKs were expressed and different cytotoxicities were observed (Fig. 1). When the number of AG-induced apoptotic cells after Annexin V-FITC and PI staining was measured via FACS analysis, it was found that the number of A549 cells was weakly increased, whereas the number of H292 cells was significantly increased during apoptosis (Fig. 3A and B). Similar results were obtained in the histogram analyzed using the ImageJ software (Fig. 3C). Levels of cleaved caspase-3, cleaved caspase-9 and PARP, which are proteins involved in the apoptotic signaling pathway, were also increased in AG-treated H292 cells, but not in A549 cells, in a dose-dependent manner (Fig. 3D).

AG remarkably increased the generation of mitochondrial ROS in H292 cells, as determined via MitoSOX staining (Fig. 4A). As revealed in the histogram of the relative amount of AG-induced mitochondrial ROS production (Fig. 4C), a significant increase in ROS levels was observed in H292 cells. TMRM staining assay revealed that AG reduced the MMP in H292 cells in a dose-dependent manner (Fig. 4B). Histogram analysis revealed that AG decreased the MMP in H292 cells but not in A549 cells (Fig. 4D). These results suggested that AG-induced metabolic shift can lead to apoptotic cell death in H292 cells by increasing mitochondrial ROS levels and depleting the MMP.

To determine whether the inhibition of PDK1 expression by AG is critical for cancer cell death, a PDK1-overexpressing (PDK1-OE) cell line was constructed by transfecting a virus with its expression vector. As revealed in Fig. 5A, the PDK1 protein was overexpressed in H292 cells, as determined by western blotting. PDK1 mRNA was overexpressed in H292 cells, as determined by RT-qPCR (Fig. 5B). In addition, AG-induced cytotoxicity in the MTT assay was significantly suppressed in PDK1-overexpressed H292 cells than that in the EV cells (Fig. 5C). At the highest concentration of AG (75 µM), only 7.68% of empty vector cells survived, and 31.7% of PDK1-overexpressed cells survived. The morphological changes in PDK1-overexpressing H292 and EV cells after treatment with AG are demonstrated in Fig. 5D. In PDK1-overexpressing H292 cells, AG-induced cytotoxicity was relatively weak compared with that in the control (EV) cells. Histogram also revealed that the number of AG-induced dead cells was slightly reduced in PDK-overexpressing H292 cells compared with that in the control cells (Fig. 5E). Expression of apoptotic marker proteins, including cleaved-PARP and cleaved-caspase 3, was suppressed in PDK1-overexpressing H292 cells (Fig. 5F). These results indicated that PDK1 expression was partially involved in the cytotoxic effects of AG. By suppressing PDK1 expression with AG, the metabolic shift from glycolysis to OXPHOS caused a significant increase in mitochondrial ROS generation, which is another key factor for AG-induced cancer cell death (Fig. 5G).