MicroRNA profiling in pediatric pilocytic astrocytoma reveals biologically relevant targets, including PBX3, NFIB, and METAP2

Abstract

Pilocytic astrocytoma (PA) is a World Health Organization grade I glioma that occurs most commonly in children and young adults. Specific genetic alterations have been described in PA, but the pathogenesis remains poorly understood. We studied microRNA (miRNA) alterations in a large cohort of patients with PA. A total of 43 PA, including 35 sporadic grade I PA, 4 neurofibromatosis-1 (NF1)-associated PA, and 4 PA with pilomyxoid features, as well as 5 nonneoplastic brain controls were examined. BRAF fusion status was assessed in most cases. RNA was examined using the Agilent Human miRNA Microarray V3 platform. Expression of miRNA subsets was validated using quantitative real-time PCR (qRT-PCR) with Taqman probes. Validation of predicted protein targets was performed on tissue microarrays with the use of immunohistochemistry. We identified a subset of miRNAs that were differentially expressed in pediatric PAs versus normal brain tissue: 13 miRNAs were underexpressed, and 20 miRNAs were overexpressed in tumors. Differences were validated by qRT-PCR in a subset, with mean fold change in tumor versus brain of -17 (miR-124), -15 (miR-129), and 19.8 (miR-21). Searching for predicted protein targets in Targetscan, we identified a number of known and putative oncogenes that were predicted targets of miRNA sets relatively underexpressed in PA. Predicted targets with increased expression at the mRNA and/or protein level in PA included PBX3, METAP2, and NFIB. A unique miRNA profile exists in PA, compared with brain tissue. These miRNAs and their targets may play a role in the pathogenesis of PA.

Fig. 1.

A subset of microRNAs is differentially expressed between pilocytic astrocytomas (PAs) and nonneoplastic brain. Heatmap illustrates expression differences between PAs and nonneoplastic brain, as shown by hierarchical clustering. The nonneoplastic brain controls, representing cortex (C1-3), and fetal (C4) and pediatric cerebellum (C5) (box). Data were obtained using the Agilent miRNA Microarray V3 kit platform.

Fig. 2.

Differential microRNA expression in PAs, compared with nonneoplastic brain. Volcano plot shows microRNA expression differences between PA and nonneoplastic brain based on a cutoff value of 2-fold differential expression and adjusted P <.05.

Fig. 3.

Differential expression of microRNAs between PAs and nonneoplastic brain confirmed by qRT-PCR. miR-124 (A) and miR-129 (B) were relatively underexpressed in sporadic PA (n = 9), compared with nonneoplastic brain (pediatric cerebellum [n = 1], cortex [n = 3], and fetal cerebellum [n = 1]), and miR-21 was relatively overexpressed (C) in an internal validation experiment. In addition, independent nonneoplastic cerebellar samples and the 4 NF1-associated PAs were analyzed. The lowest expression of miR-124 (D) and miR-129 (E) was evident in NF1-PA (n = 4), compared with sporadic PA (n = 9) and nonneoplastic brain controls (cortex [n = 3], cerebellum [n = 11]). An opposite effect was observed with miR-21 (F). Each dot represents the mean of 3 replicates. Error bars are shown. P-values were obtained using the Wilcoxon rank sum test.

Fig. 4.

microRNA predicted mRNA target differences in PAs and nonneoplastic brain samples. Unsupervised hierarchical clustering of normal brain and PA cases according to the expression of the genes targeted by the identified microRNAs down-regulated in PA (profiling by Affymetrix HG-U133 Plus 2.0 chips). Target genes are shown by row, and the different samples are shown by column. A total of 64 PAs was studied. Nonneoplastic controls included cerebral cortex (CC; n = 9), hypothalamus (n = 8), and cerebellum (Cbll; n = 3). The centered Pearson's distance and the Ward's clustering methods were used. The color scale represents increased (red) or decreased (green) gene expression.

Fig. 5.

microRNA protein target differences in PAs and nonneoplastic brain samples. Proteins that were predicted to be targets of differentially underexpressed microRNAs in PAs included putative oncogenes, such as PBX3, METAP2, and NFIB. With use of immunohistochemistry on tissue microarray sections, PBX3 and NFIB demonstrated strong nuclear staining in PAs, compared with brain, and METAP2 demonstrated modest cytoplasmic staining in a subset of PAs and was strongest in NF1-associated PAs.