RNA and DNA Extractions

Total RNA was isolated from snap frozen gliomas or cell cultures using the TRI Reagent^® (Sigma-Aldrich). Each RNA sample used for the analysis was composed of a pool of numerous extractions (3 to 7), each obtained from individual tumor tissue, different cell preparations, or normal mouse brains. The RNA isolated from adult brain was a commercial mix pooled from multiple donors and several brain regions (Ambion Inc.).

Tumor DNA was extracted from paraffin-embedded tissue using the Paraffix kit (Syntezza Bioscience, Jerusalem, Israel), according to the manufacturer's instructions.

miRNA Quantification

A total of 180 miRNA genes were quantified in the miRNA array using the TaqMan^® miRNA Assays Panel Early Access Kit (Applied Biosystems), according to the manufacturer's instructions. Briefly, the assay included two steps: RT and PCR. Each RT reaction was incubated for 30 minutes at 16°C and then at 42°C. Real-time PCR was performed on an ABI 7000 sequence detection system in triplicate for each sample. The fold-change normalization was based on cycle threshold (CT) changes of the mean calculated values obtained by subtracting individual CTs for each tissue from a median CT (ΔCT). For the individual miRNA quantification, the fold change was normalized to the RNU6B transcript (ΔCT). Following normalization, the fold change of each miRNA was calculated between the analyzed tissue (tumor, cell line or SC) and normal brain reference (ΔΔCT).

LOH Analysis

For the microsatellite PCR-based LOH analysis, primer pairs for the microsatellite loci D14S65 at 14q32.2 and D14s292 at 14q32.31 were labelled with NED and FAM, respectively (Applied Biosystems). PCR was performed on each sample in a final volume of 25 µl, containing the two primer pairs (3 pmoles of each primer), 25 ng of DNA, 10 mM Tris–HCl (pH 8.3), 50 mM KCl, 0.2 mM deoxyribonucleoside triphosphates, 2.5 mM MgCl₂, and 0.6 unit of AmpliTaq Gold DNA polymerase (Applied Biosystems). PCR cycling conditions were 95°C for 9 minutes once, 42 cycles at 94°C for 45 seconds, 55°C for 45 seconds, and 72°C for 60 seconds, with a final elongation step of 45 minutes at 60°C. Amplified PCR products were electrophoresed in denaturing 5% polyacrylamide gels on an ABI Prism 310 automated DNA sequencer (Applied Biosystems). Data analysis was performed with GeneScan Analysis software version 3.1 (Applied Biosystems).

Gliomas Display an miRNA Expression Profile That Is Reminiscent of the Profile From SCs

As described above, miRNA expression was quantified by RT real-time PCR, using the expression profile of the normal adult brains as a reference for all other tissues, separately for each species. It is interesting to note that we observed 80% identity between the miRNA expression signature of human gliomas and the miRNA expression profile of methylcholanthrene-induced mouse glioma GL-261. In the human glioma panel, 71 of the 180 miRNAs showed a distinct expression pattern (defined as either less than 0.5-fold or more than 2-fold compared with adult brain). The expression pattern of these 71 miRNAs was remarkably reminiscent of the expression pattern observed in ESCs and NPCs (Fig. 1A, Table 1).

Open in a separate window

Fig. 1.

Heat maps displaying the expression of the miRNAs that showed distinct expression in gliomas, ESCs and NPCs, when compared with normal adult brains. (A) Hierarchical clustering of the 71 miRNAs showed distinct expression (less than 0.5-fold and more than 2-fold) in human gliomas, hESCs, and NPCs, when compared with normal brains. The clustergram was generated using MATLAB Version 7.5.0.342 (R2007b). The expression values ranged between + 20 log₂ to −20 log₂. (B) Heat map of the clustered miRNAs that showed significantly different expression when compared with normal adult brains. The chromosomal location of the miRNAs is specified in the right row of each species panel: human (right) and murine (left) panels. This figure also demonstrates the similarity in miRNA expression between human and murine gliomas. The expression values ranged between + 10 log₂ to −10 log.

Clustering of the miRNAs That Constitute the Shared Profile of SCs and Gliomas

About half of the 71 miRNAs exhibiting similar expression profiles in gliomas, hESCs, and NPCs clustered in seven genomic regions: (1–3) the miR17 family, containing the miRNA clusters miR17-92, miR106b-25, and miR106a-363 and encoded by three gene clusters on chromosomes 13q31.3, 7q22.1, and Xq21.1, respectively; (4) mir183-182 on chromosome 7q32.1; (5) mir367-302 on chromosome 4q25; (6) mir371-373 on chromosome 19q13.42; and (7) the miRNA cluster in the Dlk1-Dio3 genomic region at 14q32.31 (Fig. 1B). Five of these clusters were recently linked to SCs (Table 2).⁸

The miRNA Cluster on Chromosome 14q32.31 is Downregulated in All of the Investigated Gliomas

The large 7 + 46 bipartite Dlk1-Dio3 miRNA cluster on chromosome 14q32.31 was of special interest. This cluster was downregulated in the initial panel in all gliomas, ESCs, and NPCs. This suggests that this cluster might represent the largest tumor-suppressor miRNA cluster. For this reason, we further explored this region and quantified the expression of 13 miRNAs located along the Dlk1-Dio3 miRNA cluster. For this evaluation, we used 12 astrocytic tumors and 3 Os that did not contain 1p/19q LOH. All of the miRNAs that were tested appeared to be downregulated in hESCs, hNPCs, and in all of the gliomas (Fig. 3A). Because the miRNAs from this large miRNA cluster are expressed only from the maternally inherited allele,⁹ deletion of the active allele may result in complete silencing of these miRNAs. To examine whether a chromosomal deletion was the likely reason for the downregulation of these miRNAs, LOH analysis was performed on paraffin-embedded samples of other gliomas, specifically 20 Os and 20 GBMs. D14S65 at 14q32.2 and D14s252 at 14q32.31 were used as markers for this analysis. Only 5% (1/20) of the Os and 35% (7/20) of the GBMs showed LOH of 14q in this region (data not shown).

Open in a separate window

Fig. 3.

Heat maps displaying the expression levels of miRNAs of 2 clusters in gliomas, hESCs and NPCs, relative to human brains. (A) Thirteen miRNAs within the miRNA cluster in the Dlk1-Dio3 genomic region 14q32.31 in 15 gliomas, hNPCs and hESCs (white cell represents untested). The expression values ranged between +10 log₂ to −10 log₂. (B) The ES-specific cluster, mir367-302, on chromosome 4q25 (fold change is specified in each cell) in human (left panel) and murine (right panel) tissues comprises gliomas and precursor cells.

Unclustered miRNAs in the Shared Profile

In addition to the previously discussed miRNA clusters, we also identified unclustered miRNAs that shared similar expression in gliomas and SCs (Fig. 1A, Table 1). This group included mir21, which has been previously found to be related to the pathogenesis of gliomas,¹⁰ as well as both mir124 and mir137, which both demonstrate a significantly reduced expression in gliomas in our study as well as in a previous study.⁷

Both Human and Mouse Gliomas have NPC-like miRNA Expression Signatures

Our results demonstrated that human and mouse glioma cell lines, as well as primary human gliomas, had an NPC-like miRNA expression signature (Figs 1A and ​and2A).2A). Although our cohort was relatively small, our data is supported by previous miRNA profiling studies performed separately on gliomas^7,12 or SCs (including NPCs) in diverse stages of differentiation.^8,11,13 Several miRNAs, mir21, mir124, and mir137, shared similar expression in gliomas and SCs (Fig. 1A, Table 1) and have been previously implicated in the pathogenesis of gliomas.^7,10 Furthermore, analysis of Sibler's⁷ raw data demonstrated similar results to those observed in our data (Supplementary Material, Fig. S1). These findings of miRNA expression signature, based on data produced independently in 2 different laboratories, strongly supports the results of our relatively small group of samples. (Supplementary Material, Fig. S1).

We found a similar NPC-like expression signature in cell lines that represent a relatively unified cell population, as well as in unsorted human tumors. This might indicate that most tumor cells, not only cancer SCs, exhibit an SC-like miRNA expression signature. The rationale for this conclusion is based on the previous observation that the fraction of cancer SCs within the tumor bulk is extremely small.¹⁴ To conclusively resolve this issue, further evaluation, comparing the miRNA profiles of different cell populations contained within gliomas, is required. We also demonstrated an across-species similarity in the miRNA expression signature of gliomas, independent of whether the tumor arose spontaneously in humans or was induced in mice by the carcinogenic chemical methylcholanthrene. Furthermore, both low-grade and high-grade human gliomas exhibit an NPC-like expression profile. This finding is unexpected considering the phenotypical and genotypical differences between low-grade and high-grade gliomas and it might arise from the small sample size that served for the analysis of each tumor grade, which limits the ability to detect the anticipated differences. It should be noticed that the study was not designed to evaluate differences between tumors of various grades. However, 2 other studies that examined this issue found some differential expressions between GBM and tumors of lower grades.^7,15 The first study performed an miRNA array on 8 high-grade gliomas (4 AAs and 4 GBMs) and the other study analysed a single miRNA (mir10b) in 43 gliomas of both low and high-grades. Further studies with larger number of gliomas of each grade should clarify this point in the future.

Half of the miRNAs That Show Similar Expression Patterns in Gliomas and SCs Cluster in Seven Genomic Regions Susceptible to Genetic/Epigenetic Alternations in Various Cancers

This study is the first to demonstrate that half of the miRNAs that are expressed distinctly in gliomas, and exhibit similar expression profiles to hESCs and NPCs, cluster in 7 genomic regions. Most of these miRNA clusters are either involved in cancer development or have been previously connected to SCs. Furthermore, the same 7 genomic regions are frequently susceptible to genetic or epigenetic alternations in different types of malignancies (see Table 2).^16,17

We thank Prof. Haim Ovadia for reviewing the manuscript and providing useful comments.

Conflict of interest statement. None declared.