Deregulated expression of the imprinted DLK1-DIO3 region in glioblastoma stemlike cells: tumor suppressor role of lncRNA MEG3

Abstract

Background: Glioblastoma (GBM) stemlike cells (GSCs) are thought to be responsible for the maintenance and aggressiveness of GBM, the most common primary brain tumor in adults. This study aims at elucidating the involvement of deregulations within the imprinted delta-like homolog 1 gene‒type III iodothyronine deiodinase gene (DLK-DIO3) region on chromosome 14q32 in GBM pathogenesis.

Methods: Real-time PCR analyses were performed on GSCs and GBM tissues. Methylation analyses, gene expression, and reverse-phase protein array profiles were used to investigate the tumor suppressor function of the maternally expressed 3 gene (MEG3).

Results: Loss of expression of genes and noncoding RNAs within the DLK1-DIO3 region was observed in GSCs and GBM tissues compared with normal brain. This downregulation is mainly mediated by epigenetic silencing. Kaplan-Meier analysis indicated that low expression of MEG3 and MEG8 long noncoding (lnc)RNAs significantly correlated with short survival in GBM patients. MEG3 restoration impairs tumorigenic abilities of GSCs in vitro by inhibiting cell growth, migration, and colony formation and decreases in vivo tumor growth, reducing infiltrative growth. These effects were associated with modulation of genes involved in cell adhesion and epithelial-to-mesenchymal transition (EMT).

Conclusion: In GBM, MEG3 acts as a tumor suppressor mainly regulating cell adhesion, EMT, and cell proliferation, thus providing a potential candidate for novel GBM therapies.

Keywords: MEG3 lncRNA; cancer stem cells; chromosome 14q32; glioblastoma.

Fig. 1

Expression of gene transcripts from DLK1-DIO3 region: MEG3 and MEG8 downregulation correlates with patient overall survival. (A) Volcano-plot representing the expression level of miRNAs on chromosome 14q32 from GSCs (n = 9) compared with normal neural stem cells (NSCs) from both adult (olfactory bulb) and fetal origin (n = 3). The bigger plot represents the 8 miRNA samples with the same value (miRs -433, -432*, -380*, 323–5p, -300, -453, -496, -412). P-values are based on Student’s t-test and adjusted by Bonferroni correction. (B) Real time RT-PCR analysis performed on normal brain samples (NB), GBM tissues (GBM), NSCs, and GSCs. The colored points in MEG3 and MEG8 panels represent GBM tissue and GSCs derived from the same patient. Samples were run in duplicate and normalized with glyceraldehyde 3-phosphate dehydrogenase. No statistically significant differences (P = 0.30) were observed among paired samples. (C-D) Kaplan–Meier survival curves: MEG3 and MEG8 downregulation significantly correlates with a poor clinical outcome in our cohort of patients (C; P = 0.0002) and in patients from the glioblastoma database of TCGA (D; P = 0.0066).

Fig. 2

Analysis of methylation pattern of DLK1-DIO3 region. (A) Annotated heatmap of chromosome 7 and 14 probes informative for methylation status. MS-MLPA analysis was performed on HhaI digested DNA from 20 GBM tissue specimens, 14 GSC, 6 normal brains (NB) and 6 peripheral blood mononuclear cell (PBMC) samples from normal controls (CTRL). For NBs, the region of origin of the sample is shown: F = frontal, O = occipital, P = parietal, T = temporal; for some samples, multiple regions were examined. Probes are named by gene and position (Intr = intron, ex = exon). For further information about probe genomic positions, please refer to MRC-Holland Salsa MS-MLPA probemix ME032-A1 UPD7/UPD14—description version 05;10–12–2014. Hypermethylated samples are marked with an asterisk (*). (B) Heatmap generated starting from the beta values of the CpGs associated to the DLK1-DIO3 region, considering Illumina array 450k methylation data. The samples (A to F: Normal Brains; #1, #30p, #61, #76, #83 and #163: GSC lines) have been clustered using the Euclidean distance metric. (C) Integrative Genomics Viewer (IGV) screenshot of all significantly differentially methylated Infinium probes (P < 0.01) in the DLK1-DIO3 genomic region. In red are highlighted the 12 CpG probes within MEG3-DMR, while in blue is highlighted one CpG probe (cg16126137) within the putative MEG8-DMR.

Fig. 3

MEG3 tuning modulates cell growth, migration and clonogenic abilities of GSCs. Functional in vitro assays on (A) GSC#1 (B) GSC#61, and (C) GSC#83 cell lines transduced with either empty or MEG3 vectors. Growth curves of GSCs (left panels). Points and range lines at each day represent mean and SD of at least 2 independent experiments in triplicate. Two-way analysis of variance for repeated measures was performed on the whole set of data. BrdU incorporation after 72 h pulse in empty vector and MEG3 transduced GSC lines (center-left panels). Absorbance (450–550 nm) has been shown as ratio between BrdU + and BrdU− wells. Values are reported as mean ± SD from 2 independent experiments in triplicate. Analysis of efficiency in colony formation of GSCs after transduction with MEG3 (center-right panels). Percent colony number values from 2 independent experiments in duplicate were calculated over the correspondent empty vector and are shown as mean ± SD for each GSC line. Analysis of variance demonstrated a significant effect of the restoration of MEG3 on the colony-forming ability of GSCs. Analysis of migration efficiency in GSCs transduced with MEG3 48 h after induction (right panels). Values are reported as percent relative to control vector and shown as mean ± SD from 2 independent experiments in duplicate. (D) Functional in vitro assays on GSC#83 transduced with sh-NTC, sh-MEG3_B or sh-MEG3_D vectors.

Fig. 4

Effect of MEG3 over-expression on the growth of brain xenografts of GFP expressing GSC #1. (A) Kaplan–Meier analysis of mice with brain grafts of GSCs (left). Mice grafted with MEG3 GSCs showed weight loss (>20% of initial weight) or neurological signs later than those grafted with control GSCs (n = 12; P = 0.0133, log rank test). Coronal sections of brain across the grafting site in a control and MEG3 mouse (right). (B) The density of tumor cells spreading in the striatum, amygdala, anterior commissure, and optic chiasm is highly reduced in xenografts of MEG3 overexpressing cells. (C) Graphs showing results of tumor cell counts in the brain regions analyzed; **P < 0.01; ***P < 0.001.

Fig. 5

Effect of MEG3 overexpression on gene expression and signaling pathways. (A) Linear regression equation between the log expression level in control GSCs (ln-GFP) with MEG3-GSC (ln-MEG3) cells evaluated for GSC #1 (left) and for GSC #61 (right) lines. (B) Pathway enrichment analysis examined by GSEA of highly modulated genes in both GSC#1 and #61 overexpressing MEG3.

Fig. 6

Targeted pathway analysis on GSCs. Normalized, relative protein expression levels as measured by RPPA analysis on control and MEG3-overexpressing GSC#1 and GSC#61. Barplots of selected, functional protein endpoints are grouped by pathway, as reported in Supplementary Table 2.