MicroRNA Expression Profile Distinguishes Glioblastoma Stem Cells from Differentiated Tumor Cells

Abstract

Glioblastoma (GBM) represents the most common and aggressive tumor of the brain. Despite the fact that several studies have recently addressed the molecular mechanisms underlying the disease, its etiology and pathogenesis are still poorly understood. GBM displays poor prognosis and its resistance to common therapeutic approaches makes it a highly recurrent tumor. Several studies have identified a subpopulation of tumor cells, known as GBM cancer stem cells (CSCs) characterized by the ability of self-renewal, tumor initiation and propagation. GBM CSCs have been shown to survive GBM chemotherapy and radiotherapy. Thus, targeting CSCs represents a promising approach to treat GBM. Recent evidence has shown that GBM is characterized by a dysregulated expression of microRNA (miRNAs). In this study we have investigated the difference between human GBM CSCs and their paired autologous differentiated tumor cells. Array-based profiling and quantitative Real-Time PCR (qRT-PCR) were performed to identify miRNAs differentially expressed in CSCs. The Cancer Genome Atlas (TCGA) data were also interrogated, and functional interpretation analysis was performed. We have identified 14 miRNAs significantly differentially expressed in GBM CSCs (p < 0.005). MiR-21 and miR-95 were among the most significantly deregulated miRNAs, and their expression was also associated to patient survival. We believe that the data provided here carry important implications for future studies aiming at elucidating the molecular mechanisms underlying GBM.

Keywords: cancer; cancer stem cells; glioblastoma; microRNAs; qPCR.

Figure 1

Principal Component Analysis (PCA) of CSC (in red) and FBS (in blue) samples based on the complete miRNA expression data set (A). Principal Component Analysis (PCA) of the samples according to their individual pair (B). Hierarchical clustering of the 14 significantly differentially expressed miRNAs (p < 0.005); miRNAs marked with the asterisk (*) correspond to a less abundant form (C). Functional Interpretation analysis of 67 differentially expressed miRNAs at a statistical level of p < 0.05 (D). List of the 6 most significant differentially expressed miRNAs between CSC and FBS samples (E).

Figure 2

Venn diagrams of the target genes identified by TargetScan and miRDB for the six most significant miRNAs (A). Top canonical pathway identified on IPA from the list of the target genes in the intersections of the Venn diagram of the six most significant miRNAs (B). The significance values (p-value of overlap) for the canonical pathways are calculated by the right-tailed Fisher’s Exact Test. The x-axis displays the -log of the p-value. The orange line indicates the significance threshold. Gray bars indicate pathways for which no prediction of activation or inhibition can be made due to insufficient evidence in the Knowledge Base for confident activity predictions across datasets. White bars indicate pathways with z-scores at or very close to 0.

Figure 3

Boxplots of the six most significant miRNAs expression distribution in four subtypes of Glioblastoma (GBM) of The Cancer Genome Atlas (TCGA) patients.

Figure 4

Kaplan–Meier analysis of miR-21 (A) and miR-95 (B) in the four subtypes of GBM of the TCGA patients.