A prognostic long non-coding RNA-associated competing endogenous RNA network in head and neck squamous cell carcinoma

Abstract

Background This study aimed to develop multi-RNA-based models using a competing endogenous RNA (ceRNA) regulatory network to provide survival risk prediction in head and neck squamous cell carcinoma (HNSCC). Methods All long non-coding RNA (lncRNA), microRNA (miRNA), and mRNA expression data and clinicopathological features related to HNSCC were derived from The Cancer Genome Atlas. Differentially expressed RNAs were calculated using R. Prognostic factors were identified using univariate Cox regression analysis. Functional analysis was performed using GO, KEGG pathways, and PPI network. Based on the results, we derived a risk signature and compared high- and low-risk subgroups using LASSO regression analysis. Survival analysis and the relationship between risk signature and clinicopathological features were performed using log-rank tests and Cox regression analysis. A ceRNA regulatory network was constructed, and prognostic lncRNAs and miRNA expression levels were validated in vitro and in vivo. Results A list of 207 lncRNAs, 18 miRNAs and 362 mRNAs related to overall survival was established. Five lncRNAs (HOTTIP, LINC00460, RMST, SFTA1P, and TM4SF19-AS1), one miRNA (hsa-miR-206), and one mRNA (STC2) were used to construct the ceRNA network. Three prognostic models contained 13 lncRNAs, eight miRNAs, and 17 mRNAs, which correlated with the patient status, disease-free survival (DFS), stage, grade, T stage, N stage, TP53 mutation status, angiolymphatic invasion, HPV status, and extracapsular spread. KEGG pathway analysis revealed significant enrichment of “Transcriptional misregulation in cancer” and “Neuroactive ligand-receptor interaction.” In addition, HOTTIP, LINC00460, miR-206 and STC2 were validated in GTEx data, GEO microarrays and six HNSCC cell lines. Conclusions Our findings clarify the interaction of ceRNA regulatory networks and crucial clinicopathological features. These results show that prognostic biomarkers can be identified by constructing multi-RNA-based prognostic models, which can be used for survival risk prediction in patients with HNSCC.