Characterization of long non-coding RNA transcriptome in clear-cell renal cell carcinoma by next-generation deep sequencing

Abstract

Introduction: Long non-coding RNA (lncRNA) have proven to play key roles in cell physiology from nuclear organization and epigenetic remodeling to post-transcriptional regulation. Last decade, gene expression based-classifications have been developed in clear-cell renal cell carcinoma (ccRCC) to identify distinct subtypes of disease and predict patient's outcome. However, there are no current lncRNA comprehensive characterizations in ccRCC.

Patients and methods: RNA-sequencing profiles of 475 primary ccRCC samples from the Cancer Genome Atlas (TCGA) were used to assess expressed lncRNA and identify lncRNA-based classification. In addition, integrative analysis was performed to correlate tumor subtypes with copy-number alterations and somatic mutations.

Results: Using stringent criteria, we identified 1934 expressed lncRNA and assessed their chromatin marks. Unsupervised clustering unravels four lncRNA subclasses in ccRCC associated with distinct clinicopathological and genomic features of this disease. Cluster C2 (23.4%) defines the most aggressive tumours, with the highest Fuhrman grade and stage and the worst overall survival time. Furthermore, cluster C2 is enriched for 9p deletion and chromatin remodeler BAP1 somatic mutations. Interestingly, cluster C4 (7.8%) is related to a tumor subtype arising from the distal tubules of the nephron. Consistent with its distinct ontogeny, cluster C4 is devoid of classical alterations seen in ccRCC, bears frequent 1p deletion and 17q gain, and is enriched for MiTF/TFE translocations. In addition, reexaminations of copy-number data from one side and tumor histology by pathologists from the other side reveal misclassified tumors within C4 cluster including chromophobe RCC and clear cell papillary RCC.

Conclusion: This study establishes a foundation for categorizing lncRNA subclasses, which may contribute to understand tumor ontogeny and help predicting patients' outcome in ccRCC.

Keywords: Expression profiling; Genomic aberrations; Histone markers; Long non-coding RNA; RNA-Seq; ccRCC.

Figure 1

Unsupervised consensus clustering of long non‐coding RNA showing four subclasses of clear cell renal cell carcinoma associated with different clinicopathological features.

Figure 2

Box‐plots showing genes expressed differentially between cluster C4 and other clusters. A,B) Compared to the expression in the other clusters, overexpression of FOXI1 and LHX1 genes with the associated cis‐acting long non‐coding RNA in cluster C4. C,D) Compared to the expression in the other clusters, downregulated expression of RNF186 and SLC6A13 genes with the associated cis‐acting long non‐coding RNA in cluster C4. The p‐values in boxplot were calculated from one‐way ANOVA.

Figure 3

Kaplan–Meier curves showing that patients belonging to C2 cluster display the worst overall survival as compared to those belonging to other clusters.

Figure 4

Box‐plots for expression of genes associated with renal lineage specificity and showing that cluster C4 is associated with CDH1 and CDH16 overexpression, and VIM and MME downregulated expression. The p‐values in boxplot were calculated from one‐way ANOVA.

Figure 5

Comparisons among the 4 long non‐coding RNA clusters. A) Percent alterations of the genomic regions that are differentially altered between the 4 long non‐coding RNA clusters. B) Gene set enrichment analysis (GSEA) showing overexpression of genes belonging to 3p25 region in cluster C4 as compared to others. C) GSEA showing downregulation of genes belonging to HIF1A and HIF2A targets in cluster C4 as compared to others. D) Percent mutations of BAP1 and PBRM1 chromatin remodeling genes within the 4 long non‐coding RNA clusters.

Figure 6

Box‐plots showing alterations associated with the 4 subclasses of long non‐coding RNA (lncRNA) in clear cell renal cell carcinoma. A) Three lncRNAs belonging to 1p36.23 and which are underexpressed in cluster C4 relative to the other clusters. B) Three lncRNAs belonging to 8q24.22 and which are overexpressed in cluster C2 relative to the other clusters. The p‐values in boxplot were calculated from one‐way ANOVA.