Introduction
Lung cancer is the second most common cancer with high mortality in both men and women (1). More than 80% of lung cancer are non-small cell lung cancer (NSCLC) (2,3). In China, the incidence of lung cancer in the past few decades has doubled due to the aging population, smoking and decline in air quality (4). Although some treatments have achieved certain therapeutic effect, the overall 5-year survival rates of NSCLC patients is still below 15% (5). Most NSCLC patients are in mid-term and advanced stage when diagnosed, which could contribute its poor prognosis. The progression of NSCLC is related to many genetic factors, such as abnormal gene expression (6). Gu et al found that NSCLC, comparing to other cancers, had a significantly higher ratio of VEGF-A protein/mRNA and a significantly lower level of miR-497, suggesting the presence of the post-transcriptional control of VEGF-A in NSCLC which was obviously different from other cancers (7). It demonstrated that miR-497 played an important role in suppression of VEGF-A-mediated NSCLC by inhibiting the proliferation and invasion of cancer cells (7). Therefore, it is very important to identify potential biomarkers of NSCLC for its prevention and therapy.
With the rapid development of molecular biology and bioinformatics, we can explore the mechanisms of the development and progression of cancers at the molecular level. It also provides some significant references for the diagnosis, prevention and treatment of cancer (8). However, the mechanism and development of new target-based therapies related to lung carcinogens are still unknown. Gene microarray, as an efficient technology, has been used to detect the expression levels of some genes in cells and tissues at different stages of cancer. It is also widely used for the genome-scale detections, which may help discover the key genes associated with tumorigenesis (9).
In this study, we aimed to explore the key genes in the early and mid-stages of NSCLC via gene microarray data analysis. Differentially expressed genes (DEGs), Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways which were associated with NSCLC were identified. A transcription factor (TF) network was also constructed. Our study aimed to provide a new idea and method for a better understanding of NSCLC, so as to improve the diagnosis and treatment of early stage NSCLC.
Materials and methods
Affymetrix microarray data
Microarray data set GSE50081, submitted by Der et al (10), was obtained from the Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) database. A total of 181 gene chips were available, including 48 samples in IA stage, 79 samples in IB stage, 9 samples in IIA stage and 45 samples in IIB stage of NSCLC, respectively. Microarray data were analyzed using GPL570 platform (Affymetrix Human Genome U133 Plus 2.0 Array), which included 54,675 probes to detect the expression of gene transcription levels.
Data preprocessing and screening of DEGs
The raw data were preprocessed. CEL files were normalized and converted to expression profiles using Affy package of R (11). If multiple probes corresponded to one given gene, the mean expression value was defined as the expression value. Limma package of R was used to analyze the DEGs of stage IA samples, IB samples, IIA samples and IIB samples (12). The samples were divided into three groups: IA vs. IB, IIA vs. IB and IIA vs. IIB. The gene expression values were normalized followed by the t-test of the non-paired samples (13). The multiple testing corrections were performed through Beniamini-Hochberg. FDR <0.05 and |logFC (fold change)| >2 were used as the threshold for identifying the DEGs.
Functional enrichment analysis
GO and KEGG pathway analysis of DEGs were conducted via ToppGene Suite (http://toppgene.cchmc.org) (14). ToppGene Suite is a one-stop portal for gene list functional enrichment, candidate gene prioritization using either functional annotations or network analysis and identification and prioritization of novel disease candidate genes in the interactome (14). Adjusted P≤0.05 was used as cut-off criterion.
Construction of transcriptional regulatory network
With the cut-offs of adjusted P<0.05, the TFs and miRNAs which interacted with target DEGs were identified by the online database GeneCoDis (http://genecodis.cnb.csic.es/) (15). These pairs were used to construct transcriptional regulatory network. Visualization of transcriptional regulatory network was performed by Cytoscape software (16).
Results
Identification of DEGs
A total of 19,851 genes were obtained after the processes of background calibration, standardization and the transformation from probe symbols to the gene symbol, which were used for the further analysis. The total 56 DEGs, consisted of 25 in IB vs. with IA, 17 in IIA vs. IB and 14 in IIB vs. IIA, as detailed in Table I. The variation in trends of the DEGs expression values in the three groups are shown in Fig. 1.
Functional enrichment analysis of DEGs
A total of 110 enriched GO terms and 8 KEGG pathways for DEGs were identified. The top 10 enriched GO terms are listed in Table II, and 8 enriched KEGG pathways are shown in Table III. The top three pathways were bioactivation via cytochrome P450 (P=1.01E-04), Arachidonic acid metabolism (P=3.19E-04), Cytochrome P450, arranged by substrate type (P=3.57E-04).
Analysis of regulatory network between TFs and miRNAs
Following the screening of TFs, we constructed a regulatory network between TFs and miRNAs including 13 TFs, 120 miRNAs and 563 edges (Fig. 2). Based on this regulatory network, TFs (e.g., hsa-miR-653, hsa-miR-548a-5p, hsa-miR-491-3p, hsa-miR-518d-3p, hsa-miR-518c, hsa-miR-330-3p and hsa-miR-374a) were obtained which closely connected with others. Each of them interacted with seven genes, respectively. ID1 was regulated by the most number of the genes including 35 TFs and miRNA. The top 10 genes which closely connected with others in the NSCLC samples of these four stages are shown in Fig. 3.
Discussion
Substantial research efforts has been made in exploring the mechanisms of NSCLC, however, the current understanding of the genetic alterations associated with the progressing NSCLC has not been elucidated. In our present study, 52 DEGs of NSCLC were identified using gene microarray analysis. These DEGs may influence the occurrence and development of NSCLC, thus, they will provide important references for the diagnosis and treatment of NSCLC.
Gene microarray data in the early and mid-stage NSCLC were obtained from the GEO database. An important coinciding gene (S100P) was found differentially expressed in IA vs. IB and IIA vs. IB, respectively. S100P is a 95-amino acid member of S100 protein family. It contains 2 EF-hand calcium-binding motifs (17) and involves in a series of cellular regulation processes, such as cell cycle progressions and differentiation (18). S100P was found to be one of the target genes studied in many cancers including NSCLC (19). The expression of S100P was associated with drug resistance, metastasis, and poor clinical outcome. According to our results, S100P was upregulated in IB vs. IA and downregulated in IIA vs. IB. It could be speculated that the expression of S100P increased in the processes of cancer migration and invasion in the early stage of NSCLC, and then its expression was inhibited by certain factors, such as the immune system.
Besides, 4 overlapped genes (ALOX15B, CCL11, NLRP2 and SERPINA3) were identified in IIA vs. IB and IIA vs. IIB. The knockdown of the human arachidonate 15-lipoxygenase type B (ALOX15B) was reported to reduce the inflammation and lipid accumulation, suggesting its active pro-inflammatory and proatherogenic role (20). Pyrin domain-containing protein 2 (NALP2) was characterized by an N-terminal pyrin domain (PYD). It was involved in the activation of caspase-1 by Toll-like receptors (21). The expression of NALP2 was regulated by inflammatory mediators which were closely related to cancer (22,23). As a member of the CC chemokine family, eotaxin-1 (CCL11) was initially regarded as an eosinophil chemoattractant, and involved in the recruitment of inflammatory cells, such as eosinophils and neutrophils (24). Overexpression CCL11 was found in various inflammatory diseases, such as allergic asthma. CCL11 was also reported as diagnostic marker for some cancers, such as prostate cancer (25,26). Collectively, the dysregulation of the 4 overlapped genes in different paired comparisons (Table I) revealed that the key genes may play different roles at different stages of NSCLC.
The regulatory network revealed that FoxO4 was closely connected with other nodes. It participated in the processes of cell cycle, aging, apoptosis, stress response, tumorigenesis and metabolisms. The expression of FoxO4 was confirmed in the studies of 8 cases of NSCLC patients. The immunohistochemistry method was used to characterize the expression of FoxO4 (27). Many studies have demonstrated the relationships between FoxO4 and a variety of cancers. According to Su et al, the transcription factor FoxO4 was downregulated and inhibited tumor proliferation and metastasis in gastric cancer (28). The expression of FoxO4 was significantly decreased in colorectal cancer, indicating that FoxO4 acted as a tumor suppressor in the development and progression of colorectal cancer (29).
miRNA is a non-coding RNA of about 20–25 nt in length, which can affect the phase of post-transcriptional process to regulate the expression levels of genes (30). The development of cancer is associated with the reduction of tumor suppressor genes which are regulated by miRNAs. In this study, hsa-miR-491 was regulated by the most number of TFs serving as modulators and biomarkers for the invasion and metastasis of oral squamous cell carcinoma (31). According to other reports, hsa-miR-491 was found to be involved in metastasis of hepatocellular carcinoma by inhibition of matrix metalloproteinase and epithelial to mesenchymal transition based on the the methods of miRNA microarray analysis, RT-PCR, western blotting, Transwell invasion and immunohistochemistry (32), and it may even play a role in stemness of stem cells and cancer stem cells in lung tumors (33). Moreover, the roles of other target genes of hsa-miR-491 have also been reported in many studies, so they may be the novel biomarkers of NSCLC.
In conclusion, a set of DEGs were screened and considered to be related to the development of NSCLC. S100P, ALOX15B, CCL11, NLRP2, SERPINA3, FoxO4 and hsa-miR-491 may act as candidate diagnostic markers for NSCLC patients. However, further experiments are still needed to confirm the results.