Integrative network analysis reveals subtype-specific long non-coding RNA regulatory mechanisms in head and neck squamous cell carcinoma

Jiang Li; Tan Wu; Kai Song; Lina Zhu; Yijuan Wang; Tingting Chen; Xin Wang

doi:10.1016/j.csbj.2022.12.030

Integrative network analysis reveals subtype-specific long non-coding RNA regulatory mechanisms in head and neck squamous cell carcinoma

Comput Struct Biotechnol J. 2022 Dec 21:21:535-549. doi: 10.1016/j.csbj.2022.12.030. eCollection 2023.

Authors

Jiang Li^{1

2}, Tan Wu^{1

2}, Kai Song¹, Lina Zhu¹, Yijuan Wang¹, Tingting Chen¹, Xin Wang^{1

2

3}

Affiliations

¹ Department of Surgery, The Chinese University of Hong Kong, Shatin, Hong Kong SAR, China.
² Department of Biomedical Sciences, City University of Hong Kong, Hong Kong SAR, China.
³ Shenzhen Research Institute, City University of Hong Kong, Hong Kong SAR, China.

Abstract

Head and neck squamous cell carcinoma (HNSC) is one of most common malignancies with high mortality worldwide. Importantly, the molecular heterogeneity of HNSC complicates the clinical diagnosis and treatment, leading to poor overall survival outcomes. To dissect the complex heterogeneity, recent studies have reported multiple molecular subtyping systems. For instance, HNSC can be subdivided to four distinct molecular subtypes: atypical, basal, classical, and mesenchymal, of which the mesenchymal subtype is characterized by upregulated epithelial-mesenchymal transition (EMT) and associated with poorer survival outcomes. Despite a wealth of studies into the complex molecular heterogeneity, the regulatory mechanism specific to this aggressive subtype remain largely unclear. Herein, we developed a network-based bioinformatics framework that integrates lncRNA and mRNA expression profiles to elucidate the subtype-specific regulatory mechanisms. Applying the framework to HNSC, we identified a clinically relevant lncRNA LNCOG as a key master regulator mediating EMT underlying the mesenchymal subtype. Five genes with strong prognostic values, namely ANXA5, ITGA5, CCBE1, P4HA2, and EPHX3, were predicted to be the putative targets of LNCOG and subsequently validated in other independent datasets. By integrative analysis of the miRNA expression profiles, we found that LNCOG may act as a ceRNA to sponge miR-148a-3p thereby upregulating ITGA5 to promote HNSC progression. Furthermore, our drug sensitivity analysis demonstrated that the five putative targets of LNCOG were also predictive of the sensitivities of multiple FDA-approved drugs. In summary, our bioinformatics framework facilitates the dissection of cancer subtype-specific lncRNA regulatory mechanisms, providing potential novel biomarkers for more optimized treatment of HNSC.

Keywords: AUC, area under the curve; BH, Benjamini-Hochberg; CI, confidence interval; CTRP, The Cancer Therapeutics Response Portal; Competitive endogenous RNA; DEG, differentially expressed gene; DEX, dexamethasone; DFS, disease-free survival; EMT, epithelial-mesenchymal transition; FPKM, fragments per kilobase million; GEO, Gene Expression Omnibus; GO, Gene Ontology; GSEA, gene set enrichment analysis; HNSC, head and neck squamous cell carcinoma; HR, hazard ratio; Head and neck cancer; ICGC, The International Cancer Genome Consortium; KEGG, Kyoto Encyclopedia of Genes and Genomes; LASSO, least absolute shrinkage and selection operator; Long non-coding RNAs; Network inference; OS, overall survival; ROC, receiver operating characteristic curve; Subtype-specific; TCGA, The Cancer Genome Atlas; TPM, transcripts per million; UCSC, the University of California Santa Cruz; ceRNA, the competitive endogenous RNA; lncRNA, long non-coding RNA; miRNA, microRNA.