Tumor-suppressive miR148a is silenced by CpG island hypermethylation in IDH1-mutant gliomas

Abstract

Purpose: IDH1/2-mutant gliomas harbor a distinct glioma-CpG island methylation phenotype (G-CIMP) that may promote the initiation and progression of secondary pathway gliomas by silencing tumor-suppressive genes. The potential role of tumor-suppressive microRNAs (miRNA; miR) in this process is not understood.

Experimental design: To identify potential tumor-suppressive miRNA hypermethylated in glioma, the methylation profiles of IDH1/2(WT) gliomas (n = 11) and IDH1(MUT) glioma (n = 20) were compared by using massively parallel reduced representation bisulfite sequencing (RRBS). The methylation status of selected miRNA was validated by using targeted bisulfite sequencing (BiSEQ) in a large cohort of glioma tissue samples including 219 IDH1(WT) and 72 IDH1/2(MUT) samples. The expression of selected miRNAs was determined by using the TaqMan qPCR. Functional analyses of miR148a were conducted and target genes were identified.

Results: We identify miR148a as a novel, G-CIMP-associated miRNA whose methylation is tightly correlated with IDH1 mutation and associated with improved survival in patients with malignant glioma. We confirm that downregulation of miR148a can occur via DNA methylation. We demonstrate that IDH1 mutation provides a mechanism of miR148a methylation and downregulation, and that restoration of miR148a reduced tumorigenic properties of glioma cells, possibly by targeting DNMT1.

Conclusions: We identify miR148a as a novel G-CIMP-associated miRNA, and provide results suggesting that miR148a restoration may have therapeutic implications.

Figure 1. Discovery of hypermethylated miRNAs in glioma CpG Island Methylator Phenotype (G-CIMP)

A, schematic strategy used to identify target miRNAs in G-CIMP. B, unsupervised hierarchical clustering of differentially methylated CpG islands (P<0.05, unpaired t-test) that were identified by comparing IDH1^MUT (MUT) and IDH1^WT (WT) glioma patient samples using Reduced Representative Bisulfite Sequencing (RRBS). All CpG islands within 5000 bp of pre-miRNA regions are shown. C, methylation profile of the miR-148a associated CpG island via RRBS. Top panel, map of the miR-148a CpG island (chr7:25990013-25991319), position of pre-miR-148a region (black box, chr7:25989539-25989606), predicted putative transcription start sites (TSSs) and PCR products used for bisulfite sequencing in region 1 and 2 (Black arrow); bottom panel, representative CpG site methylation pattern of IDH1^WT gliomas or IDH1^MUT gliomas determined by RRBS. D, differential expression data for miR-148a in IDH1^WT and IDH1^MUT GBM from The Cancer Genome Atlas (TCGA) dataset. E, TaqMan qPCR analysis of relative miR-148a expression in IDH1^WT and IDH1^MUT glioma tissue samples from validation cohort. Data are standardized to the mean value for IDH1^WT samples, which was set as 100%.

Figure 2. The methylation status of miR-148a is prognostic in GBM and Grade III gliomas

A, Kaplan–Meier overall survival analysis of GBM patients in the validation cohort (n=224). Survival among miR-148a-Methylated (n=29, solid line) and miR-148a-Unmethylated (n=195, dotted line) patients is shown. B, Kaplan–Meier overall survival analysis of Grade III glioma patients in the validation cohort (n=42, solid line). Survival among miR-148a-Methylated (n=30) and miR-148a-Unmethylated (n=12, dotted line) patients is shown.

Figure 3. miR-148a is methylated and downregulated in glioma cell lines and upregulated following treatment with DNA demethylation agent or DNMT1-siRNA

A and B, TaqMan RT-PCR analysis of miR-148a expression in HEK293T cells, human neural stem cells (hNSC), glioma cell lines (U251, LN18, U87 and T98G) and fibrosarcoma cell line (HT1080) without (A) or with (B) 5-aza-CdR treatment for 72 hours. miR-148a expression was determined by TaqMan qPCR. Data were standardized to the mean value of hNSC (A) or DMSO (B), which were set as 100%. C, miR-148a expression level in U251 cells transfected with DNMT1-siRNA was determined by TaqMan qPCR. D, left, 1.623 kb DNA fragment localized at the upstream of predicted miR-148a TSSs (−1038~−2661) was cloned into pGL4.17 vector as shown in the scheme; right, promoter activity of miR-148a CpG island was measured as the mean value of Firefly/Renilla luciferase activity ± SD (n=4). E, total number of methylated CpG sites determined by targeted BiSEQ of region 1 and 2 of miR-148a CpG island in parent HEK293T cells (parental 293T), empty vector control HEK293T cells (293T-EV), IDH1^WT expressing HEK293T cells (293T-IDH1^WT) or IDH1^MUT expressing 293T cells (293T-IDH1^MUT). Data shown represents mean value from 2 independent clones for each construct. F and G, miR-148a expression levels were measured in parent 293T cells, 293T-EV cells, 293T-IDH1^WT and 293T-IDH1^MUT. Results were standardized to the mean value for parent 293T cells in each passage (F) or the mean value for 293T-EV treated with DMSO (G), which was set as 100%. Each bar compared to HEK293T-EV treated with DMSO.*, P<0.05; **, P<0.01.

Figure 4. miR-148a inhibits the proliferation and migration of glioma cells

In vitro cell growth of miR-148a overexpressing glioma cells (U151 and T98G) are determined by soft agar assay (A), colony formation assay (B) and MTT assay (C); cell cycle fraction and apoptosis index are determined by FACS assay (D); cell migration ability is determined by monolayer cell migration scratch assay in a non-FBS culture condition as shown in representative phase micrographs (E); in vivo tumorigenesis of miR-148a overexpressing glioma cells (U151) is determined in NOD-scid mouse subcutaneous xenograft model (F) by measuring of tumor size (G) and weight (H), or in intracranial xenograft model by using Kaplan-Meier overall survival (OS) analysis (I).*, P<0.05; **, P<0.01; ***, P<0.001.

Figure 5. DNMT1 is the direct target of miR-148a in glioma

A and B, the expression level of DNMT1 in IDH1^WT and IDH1^MUT GBM patients from the TCGA data set (A) and our validation patient cohort (B). C–F, DNMT1 expression was determined by qPCR (C and E) and Western Blot (D and F) in U251 cells transiently (C and D) and stably (E and F) overexpressing miR-148a. G, human DNMT1 3′-UTR region contain miR-148a binding sites and dural luciferase assays were conducted using DNMT1 wild type (WT) and mutant (MUT) 3′-UTR reporter constructs. H and I, gene expression of hypermethylated RBP1, CIDEB and DLC1 in glioma cell lines U251 (H) and T98G (I) overexpressing miR-148a was determined by qPCR. Data were standardized to EV which was set as 100%. J and K, qPCR analysis of gene expression level in U251 (J) or T98G (K) cells treated with DMSO or 5-aza-CdR for 72 hours. Data were standardized to vehicle control (DMSO) which was set as 100%. *, P<0.05; **, P<0.01; ***, P<0.001.