Aggressive serous ovarian cancer subtype defined by high centrality lncRNA profiles and master transcription factors

Abstract

Long non-coding RNAs (lncRNAs) regulate the progression and metastasis of high-grade serous carcinoma ovarian cancer (HGSC). However, HGSC is yet to be classified based on these transcripts. In addition, the crosstalk between master transcriptional factors (MTFs) and lncRNAs remains unclear. Therefore, we aimed to classify HGSC based on lncRNA expression and identify the integrated MTFs for highly correlated mRNAs and lncRNAs. Unsupervised clustering was conducted using highly expressed lncRNAs derived from 367 HGSC samples obtained from The Cancer Genome Atlas. DNA mutations, somatic copy number alterations, microRNA expression, and DNA methylome were analyzed to identify the genetic and epigenetic factors affecting unsupervised clustering. Multiple Sample Virtual Inference of Protein-activity by Enriched Regulon analysis (msViper) was conducted to identify transcription factors simultaneously exhibiting positive correlation with lncRNAs and mRNAs in each cluster. In vitro analyses were performed to determine if these lncRNAs regulate both the MTFs and target genes. Functional analysis enabled the lncRNA-based classification of HGSC into five groups: "Immune," "EMT," "Estrogen response," "EMT-Androgen response," and "Differentiation" groups. The EMT-Androgen response group showed poor prognosis in the oncologic outcome. Of the transcription factors selected in this group, three MTFs with the highest eigenvector centrality scores were identified (MSC, AEBP1, CREB3L1). However, seven lncRNAs exerted a higher centrality than the selected MTFs. Our results suggest that HGSC can be classified based on lncRNA expression and characterized using molecular features. Therefore, lncRNAs and MTFs may synergistically contribute to molecular features of HGSC that could be indicators for personalized medicine.