The human lung cancer cell lines A549 (lung adenocarcinoma) and NCI-H460 (large cell lung carcinoma) were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA). Cells were plated in 96-well plates (5,000 cells/well) and were incubated overnight (37°C, 5% CO₂), followed by treatment with EGCG (20, 40, 80, 160, 320 µM) and 0 µM EGCG (control group) for 24 or 48 h respectively. The MTT-based assay (Sangon Biotech Co., Ltd., Shanghai, China) was performed to determine viable cell numbers at 37°C and DMSO was used to dissolve the formazan. Absorbance at 570 nm was measured and the 50% maximal inhibitory concentration (IC₅₀) of EGCG for A549 and NCI-H460 cells were calculated by Graphpad Prism version 6.0 (GraphPad Software, Inc., CA, USA).

Total RNA was extracted from 1×10⁶ A549 or NCI-H460 cells treated with different concentration of EGCG and control group respectively, using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.), purified with a mirVana miRNA Isolation kit (Ambion; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol, and was quantified using a spectrophotometer (NanoDrop ND-1000; NanoDrop, Wilmington, DE, USA). The RNA integrity of each sample was assessed by capillary electrophoresis using the RNA 6000 Nano Lab-on-a-Chip kit and Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Only RNA extracts with RNA integrity values >6 were subjected to further analysis.

complementary (c)DNA labeled with a fluorescent dye (Cy5 or Cy3-deoxycytidine triphosphate; CapitalBio Technology Co., Ltd., Beijing, China) was produced by Eberwine's linear RNA amplification method as previously described (23), and subsequently RNase H enzymatic reaction (37°C for 45 min, and followed by 95°C for 5 min). The labeled cDNAs were purified using a Capital Bioc RNA Amplification and Labeling kit (CapitalBio Corporation, Beijing, China) and then hybridized with specific probes (CapitalBio Technology Co., Ltd.) in a hybridization oven (Xinghua Analytical Instrument Factory, Jiangsu, China) overnight at 45°C.

The expression levels of lncRNAs and mRNAs were determined using Gene Spring software V13.0 (Agilent Technologies). The differentially expressed lncRNAs and mRNAs between the EGCG-treated and control groups were identified based on the threshold values of ≥2 and ≤-2 fold-change and paired Student's t-test P<0.05. Data were log2-transformed and median-centered using genes and Adjust Data function of CLUSTER 3.0 software (http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm) and analyzed by hierarchical clustering with average linkages. Tree visualization was performed using the Java TreeView (Stanford University School of Medicine, Stanford, CA, USA; http://sourceforge.net/projects/jtreeview/files). Gene-lung cancer associations were investigated by using the DisGeNET database (http://www.disgenet.org/web/DisGeNET/menu), which records disease-associated genes and provides literature support, as in the previous reference (24).

GO analysis was performed using an online database (www geneontology.org). The GO database provides a network of three structured definition terms that describe the properties of a gene product (25). Pathway analysis of the differentially expressed genes was performed according to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (www.genome.jp/kegg/). Specifically, Fisher's exact test and χ² tests were performed to classify the GO category and select the significant pathway. The false discovery rate (FDR) was calculated and the threshold of significance was defined as P<0.05. FunDO database (http://fundo.nubic.northwestern.edu/) was used for disease analysis.

Search Tool for Interactions of Chemicals (STITCH; http://stitch.embl.de/) is a database of interactions and correlations between compounds and genes. The EGCG direct target genes were searched and analyzed by using STITCH according to the reference (26). Search Tool for the Retrieval of Interacting Genes (STRING, http://string-db.org/) is a database of protein-protein interactions (27). The interaction of EGCG direct target gene and differential expression oncogene affected by EGCG was determined in the STRING database, and only the interactions with a combined score > 0.4 were considered as significant. Finally, the network of compound and gene interactions was mapped.

We downloaded the transcriptional factors for each of the 20 differential expression oncogenes most affected by EGCG treatment from the UCSC Genome Browser (http://genome.ucsc.edu/cgi-bin/hgTracks?hgsid=698152999_7PLacukNAMsjS2GSMIFtIxAic9jW). Then, the lncRNA-transcriptional factor-oncogene regulatory network was constructed by using the LncPro database (http://bioinfo.bjmu.edu.cn/lncpro/) as previously described (28).

The Total RNA of 1×10⁶ A549 cells treated with 80 µM EGCG for 48 h and the control group cells was extracted as described above respectively. The PrimeScript^™ RT Reagent kit (Takara Bio, Inc., Otsu, Japan) was used for cDNA synthesis and genomic DNA removal. qPCR was performed using SYBR^® Premix Ex Taq™ (Takara Bio, Inc., Otsu, Japan) and in an ABI 7000 instrument (Applied Biosystems; Thermo Fisher Scientific, Inc.) for 40 cycles (95°C for 15 sec, 60°C for 1 min) after an initial 3 min degeneration at 95°C, and β-actin was used as an internal control (primers for qPCR listed in Table I). The relative gene expression data was analyzed by 2^−ΔΔCq method as previously described (29).

The experiments were performed three times. All results are expressed as the mean ± standard deviation. A paired Student's t test was performed to compare the two groups in the microarray analysis. P<0.05 was considered to indicate a statistically significant difference.

Treatment with EGCG inhibited lung cancer cell proliferation in dose-dependent manner (Fig. 1A). EGCG inhibited the growth of A549 cells, with IC₅₀ values of 122.4 and 108.2 µM at 24 and 48 h, respectively. Additionally, the IC₅₀ values of EGCG in NCI-H460 cells were 693.2 and 485.1 µM at 24 and 48 h, respectively. These results demonstrated that A549 cells are more sensitive to EGCG than NCI-H460 cells. Therefore, A549 cells treated with 80 µM (≈IC₃₀) for 48 h were used for microarray analysis.

A total of 19,446 lncRNAs were detected by Gene Spring software V13.0. Based on the results of the lncRNA expression profile, differentially expressed lncRNAs between the EGCG-treated and control groups were identified. According to their expression levels in the samples, the clusters were hierarchically clustered (Fig. 1B). The threshold was set to a fold-change ≥2, P<0.05, and FDR <0.05 and a total of 960 differential expressed lncRNAs, including 450 and 510 up- and downregulated lncRNAs, respectively were identified (Fig. 1C).

Among the differentially expressed lncRNAs, there were 416 intergenic, 302 antisense, 141 unknown, 44 divergent, 38 intronic, and 19 sense lncRNAs (Fig. 1D). The length of the differentially expressed lncRNAs ranged from 200 base pairs (bp) to 8 kb, with the lengths of 660 lncRNAs (68.75%) ranging between 200–1,000 bp (Fig. 1E). Additionally, these differentially expressed lncRNAs were distributed in nearly all human chromosomes (Fig. 1F).

A total of 1,434 mRNAs were detected whose expression significantly differed between lung cancer cells treated with EGCG and the control group (fold-change ≥2.0, P<0.05, and FDR <0.05). Of these, 804 mRNAs were downregulated and 630 were upregulated. Through hierarchical clustering analysis, different expression patterns were predicted (Fig. 2A and B). The top 10 upregulated and 10 downregulated mRNAs between the two groups are listed in Table II. The top 5 upregulated and 5 downregulated mRNAs associated with lung cancer are listed in Table III. Genes associated with lung cancer in the DisGeNET database were investigated. By comparing the results of the database with the results obtained upon EGCG treatment, 168 known lung cancer genes in the EGCG-treated groups were identified, and the top 50- changed genes were plotted as a heat map (Fig. 2C).

A total of 10 differentially expressed lncRNAs were randomly selected, including 5 upregulated (ENSG00000272796.1, ENSG00000254054.2, ENSG00000260630.2, ENSG00000235142.2 and ENSG00000224063.1) and 5 downregulated lncRNAs (ENSG00000251018.2, ENSG00000226403.1, PSMC3IP, ENSG00000230109.1 and SG00000130600.10), for verification in these lung cancer cells. The RT-qPCR results of these 10 selected lncRNAs were consistent with those from the microarray analysis (Fig. 3A). Additionally, 10 differentially expressed mRNAs were also randomly selected, including 5 upregulated (aldolase C, hyaluronan-binding protein 2, complement factor B, interleukin (IL) 11, and secreted modular calcium-binding protein 1) and 5 downregulated mRNAs (amphiregulin, epiregulin, humanin-like protein 6, polymerase (RNA) III (DNA Directed) Polypeptide G, and Ly1 antibody reactive). The results for all 10 selected mRNAs were consistent with those from the microarray analysis (Fig. 3B).

GO analysis consists of three parts: Biological processes, cellular components, and molecular functions. Through GO analysis, it was demonstrated that the differentially expressed mRNAs were mainly enriched in the GO terms ‘cell cycle’, ‘mitotic cell cycle’, ‘cell cycle process’ and ‘mitotic cell cycle process’ for biological processes (Fig. 4A); ‘chromosome’, ‘centromeric region’, ‘chromosomal region’, ‘condensed chromosome kinetochore’ and ‘chromosome’ for cellular components (Fig. 4B); and DNA-dependent ATPase activity’, ‘DNA helicase activity’ and ‘iron ion binding’ for molecular functions (Fig. 4C). Pathway analysis was based on the KEGG database. The differentially expressed mRNAs were associated with DNA replication, complement and coagulation cascades, cell cycle and other pathways (Fig. 4D). The dysregulated mRNAs were associated with a number of diseases, of which cancer was the most relevant (Fig. 4E).

A total of 20 EGCG target genes were detected using the STITCH database; these genes were AKT1, Caspase 3, IL-6, prostaglandin-endoperoxide synthase 2 (PTGS2), tumor protein 53 (TP53), vascular endothelial growth factor A (VEGFA), FOS, mitogen-activated protein kinase 8 (MAPK8), nitric oxide synthase 3 (NOS3), signal transducer and activator of transcription 3 (STAT3), matrix metalloproteinase 9 (MMP9), C-C motif chemokine ligand 2 (CCL2), transthyretin (TTR), MMP2, DNA (cytosine-5)-methyltransferase 1 (DNMT1), sirtuin 1 (SIRT1), nuclear receptor subfamily 1 group H member 4 (NR1H4), cytochrome P450 family 1 subfamily A member 2 (CYP1A2), cyclin dependent kinase inhibitor 2A (CDKN2A), and IL-8. The genes are annotated as being directly targeted by EGCG (26). According to the lung cancer tissues’ gene expression levels, obtained from The Cancer Genome Atlas database, it was demonstrated that the expression levels of these 20 genes were closely associated to the lncRNAs regulated by EGCG, among which DNMT1 is also a well-known lung cancer-associated gene. All the 20 EGCG direct target genes and 20 oncogenes mostly affected by EGCG were entered into the STRING database for network construction. The network of gene interactions as well as compound-gene interactions were illustrated in Fig. 5. A broad linkage between EGCG target genes and known lung cancer-associated genes suggested that EGCG was very likely to affect NSCLC by acting on these genes.

LncPro was used to predict the interaction of the top 10 lncRNAs and top 20 oncogenes affected by EGCG. A total of 8 of the 10 lncRNAs were demonstrated to bind the transcriptional factors (TF) of relevant oncogenes with relatively high binding scores. The network of lncRNA-TF-oncogenes is demonstrated in Fig. 6, where the lncRNAs can bind to nearly all oncogene-associated TFs, including B-cell lymphoma-2-modifying factor and collagen type IV α3 chain, which are regulated by additional lncRNA-bound transcriptional factors.