To evaluate the expression levels of EZH2 in GBM cells and non-neoplastic brain (NNB) we performed immunohistochemistry for EZH2 protein expression on tissue microarrays containing GBM and NNB samples. Most of the GBM samples showed fields of strong nuclear staining for EZH2 while none of the NNB samples did (Fig. ). Increased EZH2 expression correlated with glioma grade and glioma recurrence (Fig. ), suggesting that EZH2 could be a marker for glioma aggressiveness. In addition, the Rembrandt database was used to show that EZH2 expression correlated with decreased GBM patient survival (Fig. ). EZH2 protein was strongly expressed in human GBM cell lines, including U251 and U87, but not in NNB (Fig. ).
EZH2 expression is associated with high grade glioblastoma
In order to determine whether potential GBM-expressed miRNAs could affect EZH2 expression we first determined which miRNAs expressed in NNB are differentially expressed in GBM (Supplemental Table S1A
). Next, we used miRbase [28
] to identify 63 miRNAs predicted to target EZH2. Upon integration of the list of miRNAs predicted to target EZH2 and the differential GBM/NNB miRNA expression ratios, we found that miR-101, miR-98, miR-137, and miR-139 were down-regulated in GBM tissue as compared to NNB and have the potential to regulate EZH2 (Supplemental Table S1B
). Another miRNA which was previously found to target EZH2, miR-26a [29
], was not included in our subset of miRNAs expressed in the brain, and therefore not part of our study.
We were particularly interested in miR-101 since it was confirmed to bind the EZH2 3’-UTR at two sites (Fig. ), and was recently shown to interact with EZH2 in other types of cancer [30
]. Previous analysis showed that genomic loss of miR-101-1 and miR-101-2 alleles was observed in 18.7% of GBM cases [30
]. Based on these findings, we decided to further analyze miR-101/EZH2 functionality in GBM. First, down-regulation of miR-101 was confirmed in primary glioma samples of different WHO grades by quantitive PCR (qRT-PCR) analysis (Fig. ). To establish that miR-101 affects EZH2 protein expression and histone methyltransferase activity in GBM, we transfected human U87 GBM cells with pre-miR-101 molecules and determined the levels of EZH2 protein and H3K27me3. In addition to pre-miR-101, we included EZH2 siRNA, non-related control oligonucleotides, and the S-adenosylhomocysteine hydrolase inhibitor DZNep, a potent EZH2 inhibitor [33
]. DZNep, EZH2 siRNA, and pre-miR-101 all notably repressed EZH2 protein expression and reduced the levels of trimethylation of H3K27 (Fig. ), indicating inhibition of EZH2 function.
miR-101 is down-regulated in GBM and targets EZH2
To determine the effects of EZH2 on GBM cell proliferation we first analyzed which genes associated with cell proliferation correlated with EZH2 expression in GBM and NNB [35
]. First, EZH2 was found overexpressed in most GBM samples as compared to NNB. However, in few samples the EZH2 mRNA expression was found to be in the same range as in the NNB (Fig. ). Out of the 1419 genes that were linked to the proliferation gene ontology as determined by AmiGO [36
], 214 genes showed a clear correlation (>67%) with EZH2 expression in GBM (Fig. ). Interestingly, the GBM samples with normal EZH2 expression levels also showed less expression of the genes associated with cell proliferation. Next, cellular proliferation was studied in GBM cell cultures to determine if EZH2 influences the proliferation of GBM cells. miR-101 induction, EZH2 knock down by siRNA and treatment with DZNep significantly reduced cellular proliferation in U87-GFP GBM cells (Fig. ). The effect of DZNep treatment on proliferation inhibition was confirmed in other GBM cell lines in vitro
miR-101-regulated EZH2 affects GBM proliferation in vitro
To determine the effects of EZH2 on GBM cell migration we analyzed which genes belonging to the migration gene ontology correlated with EZH2 expression in GBM and NNB. A significant correlation between the expression of 28 out of 279 genes associated with cell migration and EZH2 expression was observed (Fig. ). In order to determine whether miR-101 up-regulation or EZH2 inhibition also affected GBM cell migration, scratch assays were performed. Up-regulation of miR-101 by pre-miR-101 resulted in a significant decrease in U87 migration. The EZH2 inhibitors DZNep and EZH2 siRNA showed a similar decrease in migration (Fig. ). To further evaluate the effects of miR-101/EZH2 modulation on in vitro migration and invasion, a Boyden chamber assay was used. U87 cells that were transfected with pre-miR-101 showed a significant decrease in ability to invade, as visualized by Hoechst staining (Fig. and quantitated in ). Again, similar results were observed after treatment with the EZH2 inhibitor DZNep or EZH2 knock down by siRNA (Fig. ).
In silico analysis of EZH2 mRNA expression and the correlation to migration-related mRNAs
To determine the effects of EZH2 on GBM-induced angiogenesis we also analyzed which genes belonging to the angiogenesis gene ontology correlated with EZH2 expression in GBM. Again, a significant correlation between the expression of 33 out of 308 genes associated with angiogenesis and EZH2 expression was observed (Fig. ). Next, HBMVECs were cultured in EBM, EGM, or EBM supplemented with U87 human GBM cells expressing GFP (U87-GFP), all on a Matrigel substratum to promote tubule network formation. Tubules were visualized by a combination of light and fluorescence microscopy. After pre-treatment of the GBM cells with DZNep, or transfection with pre-miR-101, EZH2 siRNA, or non-related oligonucleotides of similar chemistry, and subsequent co-culturing with HBMVECs on Matrigel, we analyzed tubule length and tubule branching. Up-regulation of miR-101 in U87-GFP cells resulted in a substantial decrease in total tubule length and tubule branching (Fig. ). In addition to pre-miR-101, treatment of the U87-GFP cells with EZH2 siRNA and DZNep also inhibited U87-induced tubule network formation (Fig. ).
miR-101/EZH2 affects GBM angiogenesis in vitro
Finally, to study the effects of modulation of EZH2 on GBM growth in vivo, we implanted U87 human GBM cells stably expressing Fluc and the fluorescent protein mCherry (U87-Fluc-mCherry) into the flanks of nude mice. Tumor growth was monitored over time by intravenous injection of the Fluc substrate D-luciferin and in vivo bioluminescence imaging using a CCD camera. After tumor cell implantation, we injected one set of mice (n = 5) intravenously with the EZH2 inhibitor DZNep (0.07 mg/kg) and a parallel control set (n = 5) with PBS only, at day 3, 5, and 7, followed by weekly injection. CCD camera imaging of Fluc bioluminescence activity in the tumor allowed us to monitor tumor growth over time. The tumor volume of the mice treated with PBS increased logarithmically over time, while the tumor volume in the mice treated with DZNep showed reduced growth (Fig. ).
Inhibition of EZH2 affects GBM development in vivo