In this work, we performed a comprehensive analysis of the full transcriptome sequencing dataset of Fujian Cancer Hospital and six microarray datasets downloaded from the GEO repository to uncover DEGs between NPC tissues and normal nasopharyngeal tissues. We identified DEmRNAs, DEmiRNAs, and DElncRNAs among the seven datasets, and constructed a lncRNA-miRNA-mRNA network of NPC. GO enrichment analysis, KEGG enrichment analysis, and GSEA proved that the enriched components and pathways among the DEGs associated with NPC were inseparable from the chromosome dysfunction discovered in NPC. We also identified the top 20 hub genes in the PPI network related to NPC, and the results of the enrichment analysis of the hub genes were similar to those of the DEGs.
Studies have shown that lncRNA-miRNA-mRNA networks play significant roles in the development and progression of tumors[36]. The lncRNA-miRNA-mRNA network constructed in our study indicated that in NPC, 2654 mRNAs could be regulated by 132 lncRNAs via 565 corresponding miRNAs. Li et al. identified 37 miRNAs, including 19 highly expressed miRNAs and 18 lowly expressed miRNAs, from the serum of 12 NPC patients with different radiosensitivity; these miRNAs were found to have remarkable differences between the patients (fold change ≥ 2 or ≤ 0.5 and P < 0.05)[37]. The highly expressed miRNA hsa-miR-6088 and the lowly expressed miRNA hsa-let-7f-1-3p from the above study were also found in our ceRNA network. We also identified hsa-miR-29a-3p and hsa-miR-103a-3p as DEmiRNAs, which were recently found to act as circulating biomarkers of NPC, with fairly good diagnostic accuracy for detecting NPC as compared with controls (area under the curve > 0.7)[38]. The radioresistant NPC CNE2-IR cell line has been shown to overexpress JUN; Guo et al. identified 35 JUN-related miRNAs by using mirDIP software, including hsa-miR-200b-3p, hsa-miR-139-5p, hsa-miR-200c-3p, hsa-miR-9-5p, and hsa-miR-92b-3p[39]. Thus, JUN could promote tumorigenesis and tumor development. Qing et al. found that inhibiting c-JUN expression could enhance radiosensitivity, and induce cell cycle arrest and apoptosis[40]. The above results show that ceRNA networks can offer insights into the complex regulation patterns of NPC and potentially facilitate the individualized treatment of NPC.
GO analysis of the BPs of the DEGs associated with NPC showed that the negative regulation of chromosome segregation, nuclear chromosome segregation, sister chromatid segregation, mitotic sister chromatid segregation, and negative regulation of chromosome organization were closely associated with the oncogenesis of NPC. Among the CC annotations, chromosomal region, condensed chromosome, chromosome (centromeric region), condensed chromosome (centromeric region), condensed chromosome kinetochore, and nuclear chromosome (telomeric region) were notably related to NPC. Several studies have reported on the chromosomal aberrations involved in the carcinogenesis of NPC, including chromosomal gains or losses[41], loss of heterozygosity[42], and homozygous deletions[43]. One study performed comparative genomic hybridization analysis on 51 NPC samples (25 primary tumor samples and 26 relapsed tumor samples), and found 36 informative samples (17 primary NPC samples and 19 recurrent NPC samples), with 249 significant chromosomal changes[41]. In one study, alterations in chromosome 3 were revealed using cytogenetic analysis of NPC specimens and three NPC xenografts[44]. Allelic loss on the short arm of chromosome 3 is one of the most common molecular genetic alterations in NPC[45]. Indeed, in one study, loss of heterozygosity on 3p was observed in 95%–100% of primary NPC specimens and almost 75% of pre-cancerous lesions[46]. A recent study has indicated that the novel 3p26.3 tumor suppressor gene CHL1 plays a role in the development of NPC[47]. Enrichment analyses of the 20 hub genes identified in our study were greatly compliant with the results of the enrichment analyses of the DEGs. Thus, the above findings clearly implicate chromosomal dysfunction as an important contributor to the carcinogenesis of NPC.
GSEA showed that the MAPK signaling pathway, PI3K-Akt signaling pathway, apoptotic pathway, and TNF signaling pathway were the top four pathways associated with NPC. The enriched pathways identified in our investigation are related to tumor progression, metastasis, and apoptosis. The MAPK/extracellular signal-regulated kinase (ERK) pathway has been reported to be closely related to cell proliferation, differentiation, migration, senescence, and apoptosis[48]. In prostate carcinoma, intracellular chloride channels have been proved to influence cell multiplication and migration via the MAPK/ERK pathway[49]. In hepatocellular cancer, lysyl oxidase propeptide could induce apoptosis via downregulation of the MAPK/ERK pathway[50]. Recent studies have found that knockdown of amyloid β precursor protein is closely associated with the downregulation of the MAPK pathway, and this could greatly impede the neoplasia of NPC[51]. The PI3K/Akt pathway is vital to NPC progression, metastasis, and invasion[52]. After the EBV genome is introduced into the NPC cells, this pathway could be activated[53]. A study showed that inhibiting the PI3K/Akt pathway could suppress NPC cell metastasis by inducing mesenchymal-epithelial reverting transition[54]. Studies have shown that asiatic acid notably reduces the viability of cisplatin-resistant NPC cells by inducing apoptosis via the internal and external apoptotic pathways[55]. The proto-oncogene JUN is associated with the cis-regulatory lncRNA RP4-794H19.1 and is very commonly found in cancers; JUN has been linked to the TNF signaling pathway, and may be a vital gene in NPC[56]. TNF-α may be a tumor-promoting factor in NPC, as TNF-α expression has been observed in both primary NPC specimens and serum derived from NPC patients, and can significantly predict the risk of distant metastasis in NPC patients[57].
In the PPI network constructed in our study, the following 20 DEGs with many interactions were selected as vital hub genes (NUSAP1, RACGAP1, PRC1, KIF4A, TOP2A, PBK, KIF2C, TPX2, CENPU, OIP5, TTK, MAD2L1, NDC80, BIRC5, MELK, CENPF, FOXM1, TYMS, CDK1, and CEP55). More observations of these genes may lead to further insights into the carcinogenesis of NPC. For example, PRC1 levels significantly decrease after cells exit mitosis and enter G1, and are high during the S and G2/M stages[58]. PRC1 has been associated with many malignant carcinomas, such as NPC[59], prostate cancer[60], and lung cancer[61]. In vitro experiments have shown that PRC1 depletion inhibits the multiplication and invasiveness of NPC, while in vivo studies have found that PRC1 inhibits the neoplasia and radioresistance of NPC[59]. Increased PBK expression in NPC patients has been positively correlated with clinical severity, specifically advanced T stage, and disease progression. Indeed, PBK overexpression is an independent prognostic factor that shortens overall survival; the malignant phenotype of NPC requires PBK, as downregulation of PBK expression can inhibit NPC cell proliferation[62]. One study found that the overexpression of miR‐372 via the downregulation of PBK not only enhances the radiosensitivity of NPC but also reduces its aggressiveness and inhibits metastasis[63]. TOP2A is closely related to cell division by selective cleavage, rearrangement, and reconnection of DNA strands. This gene is highly expressed in NPC subpopulations. Its enhanced immune expression is markedly related to advanced cancer and the invasiveness of NPC[64].
Our study has certain limitations. First, we did not perform any further verification using molecular experiments, such as western blot analysis, quantitative real‐time PCR, and immunohistochemistry, to fully clarify the roles of the predicted hub genes and signaling pathways and uncover the potential mechanisms underlying NPC. It is also necessary to conduct loss‐of‐function and gain‐of‐function studies with tissue‐type specificity and cell‐type specificity. Second, we lack corresponding clinical relevance studies and analyses based on clinical information. Third, we used seven datasets to reduce the high false-positive rate related to single microarray analysis. However, the use of many datasets may lead to inter-batch differences that cannot be avoided or removed during the analyses. Finally, it is not clear whether the prediction of biomarkers using the background-corrected matrix file was reliable, as each GEO dataset was obtained using a different correction method[65].
In summary, we performed integrated bioinformatics analyses on seven datasets (GSE12452, GSE13597, GSE95166, GSE126683, GSE70970, GSE43039, and FJCH) to identify DEGs involved in the pathogenesis of NPC. Altogether, we uncovered 3664 DEmRNAs, 4068 DElncRNAs, 265 DEmiRNAs, and 20 hub genes that may serve as biomarkers for the diagnosis, prognosis, and therapy of NPC. GO analysis, KEGG analysis and GSEA indicated that chromosome dysfunction could underlie the pathogenesis of NPC. We also constructed a lncRNA–miRNA–mRNA network to better understand the potential biological mechanisms among the identified genes. Our results may provide new targets for understanding the molecular mechanisms of NPC.