Comprehensive characterization of cancer subtype associated long non-coding RNAs and their clinical implications

Abstract

Long non-coding RNAs (lncRNAs) are a kind of RNAs with regulation that participate fundamental cellular processes via diverse mechanisms. Despite the potential importance of lncRNAs in multiple kinds of cancer has been well studied, no comprehensive survey of cancer subtype associated lncRNAs. Here, we performed an array-based transcriptional survey of lncRNAs across 150 lung cancer samples comprising both adenocarcinoma and squamous cell carcinoma, and 306 breast cancer patients with clear clinical information. In lung cancer, 72 lncRNAs are identified to be associated with tumor subtypes and their functions as well as the associated proteins are predicted by constructing coding-non-coding co-expression network. The results suggest that they are mostly related with epidermis development, cell adhesion and response to stimulus. The validation results show the high concordance and confirmed the robust of the identification results. In breast cancer, we found 3 lncRNA genes are associated with estrogen receptor α (ER) positive and ER negative subtypes and tumor histologic grade. Survival (Kaplan-Meier) analysis results suggest that the expression pattern of the 3 lncRNAs is significantly correlated with clinical outcomes. The current study provides the first large-scale survey of lncRNAs within cancer subtypes, and may offer new targets for their diagnosis, therapy and prognosis.

Figure 1. Expression profile and functional enrichment of cancer subtype associated genes.

(a) and (b) denote the expression profiles of lncRNA and protein-coding genes respectively. SCC: squamous cell carcinoma; AC: adenocarcinoma; (c) The functional enrichment results of differentially expressed protein-coding genes. (d) The functional enrichment map of GO terms with each node represents an GO term and an edge represents existing genes shared between connecting GO terms. Node size denotes the number of gene in the GO term. The main functional processes are marked for each group of GO terms.

Figure 2. Genomic context of NKX2-1-AS1 (a) and DSCAM-AS1 (b).

Evolutionary conservation status as shown under the gene is measured by multiple alignments of 100 and 46 vertebrate species respectively.

Figure 3. The SFTA1P-centered subnetwork and the functional enrichment results of its neighboring coding genes.

Red edges represent the connections between SFTA1P and coding genes, blue edges represent the connections between coding genes.

Figure 4. Validation study of cancer subtype associated lncRNAs.

(a). The number of cancer subtype associated lncRNAs (intersection) confirmed by an independent validation set (pink). (b). Principal-component analysis (PCA) of cancer subtype associated lncRNAs for lung adenocarcinoma and squamous cell carcinoma samples. PCA1, PCA2 and PCA3 represent the top three dimensions of cancer subtype associated lncRNA genes.

Figure 5. Characterization of breast cancer subtype associated genes.

(a). The overview of selecting lncRNA genes based on tumor subtype and grade. (b). The functional enrichment results of filtered protein-coding genes. (c). Kaplan-Meier survival curves for relapse free survival of two groups that clustered based on lncRNA expression pattern.