MicroRNA-128 coordinately targets Polycomb Repressor Complexes in glioma stem cells

Abstract

Background

The Polycomb Repressor Complex (PRC) is an epigenetic regulator of transcription whose action is mediated by 2 protein complexes, PRC1 and PRC2. PRC is oncogenic in glioblastoma, where it is involved in cancer stem cell maintenance and radioresistance.

Methods

We used a set of glioblastoma patient samples, glioma stem cells, and neural stem cells from a mouse model of glioblastoma. We characterized gene/protein expression and cellular phenotypes by quantitative PCR/Western blotting and clonogenic, cell-cycle, and DNA damage assays. We performed overexpression/knockdown studies by lentiviral infection and microRNA/small interfering RNA oligonucleotide transfection.

Results

We show that microRNA-128 (miR-128) directly targets mRNA of SUZ12, a key component of PRC2, in addition to BMI1, a component of PRC1 that we previously showed as a target as well. This blocks the partially redundant functions of PRC1/PRC2, thereby significantly reducing PRC activity and its associated histone modifications. MiR-128 and SUZ12/BMI1 show opposite expression in human glioblastomas versus normal brain and in glioma stemlike versus neural stem cells. Furthermore, miR-128 renders glioma stemlike cells less radioresistant by preventing the radiation-induced expression of both PRC components. Finally, miR-128 expression is significantly reduced in neural stem cells from the brain of young, presymptomatic mice in our mouse model of glioblastoma. This suggests that loss of miR-128 expression in brain is an early event in gliomagenesis. Moreover, knockdown of miR-128 expression in nonmalignant mouse and human neural stem cells led to elevated expression of PRC components and increased clonogenicity.

Conclusions

MiR-128 is an important suppressor of PRC activity, and its absence is an early event in gliomagenesis.

Glioblastoma multiforme is the most common and aggressive intrinsic primary brain tumor in adults. Approximately 10 000 new cases are diagnosed in the United States every year, and patients' median survival is ∼14.6 months, even with surgery, radiation, and chemotherapy.¹ This tumor presents unique challenges to therapy due to its location, aggressive biological behavior, and diffuse infiltrative growth. Despite the development of new surgical and radiation techniques and the use of multiple anti-neoplastic drugs, effective long-term treatments remain elusive.² Numerous recent studies have demonstrated that glioblastoma cells retain many of the features of neural progenitor cells, including the ability to grow as neurospheres in culture and self-renew. These cells, known as glioma stemlike cells (GSCs), are a subpopulation of cells in the tumor with self-renewal capacity whose asymmetric division gives rise to the heterogeneous nature of cancer cells that make up the neoplastic mass. When implanted in immunodeficient mice, as few as 100 GSCs can give rise to tumors that recapitulate the histological features and global gene expression patterns of the parental patient tumors.^3–6 These observations have created novel opportunities for developing anti-glioblastoma therapeutics based on targeting the GSC component of glioblastoma.⁷

The discovery of microRNAs and the increasing appreciation of their importance in glioblastoma and other cancers^8–12 have led to improved understanding of key mechanisms underlying brain tumors and other cancer types. MicroRNAs are approximately 20- to 23-nucleotide noncoding RNAs that act as important regulators of gene expression by downregulating expression of target mRNAs through regions of partial complementarity in their 3′ untranslated regions (UTRs).¹³ Each microRNA may have hundreds of targets, and many genes are targeted by multiple microRNAs, thus leading to potentially highly complex regulatory networks. Distinct patterns of microRNA expression have been observed in many cancers, including glioblastomas.^8,10,14

In glioblastoma, several deregulated microRNAs have been strongly implicated in the maintenance of stemness. Work from our laboratory has established a link between microRNA-128 (miR-128), which is strongly downregulated in glioblastoma, and loss of GSC self-renewal, through direct regulation of the neural stem cell (NSC) self-renewal factor B lymphoma Mo-MLV insertion region 1 homolog (BMI1). We demonstrated that downregulation of BMI1 is dependent on miR-128 expression, leading to a decrease in proliferation of glioma cells in vitro and in vivo and to a reduction of the self-renewal ability of GSCs.¹⁵ BMI1 is a component of the Polycomb Repressor Complex (PRC), which suppresses expression of key target genes through chromatin modification. PRC components are commonly upregulated in glioblastoma and play an important role in normal development,^16,17 organogenesis,¹⁸ and embryonic stem cell biology.¹⁹ The entire PRC is composed of 2 distinct protein complexes: PRC1 ubiquitinates histone H2A, and PRC2 methylates histone H3.²⁰ These functions generally lead to transcriptional repression and, in neoplasia, are known to be important in cancer stem cell self-renewal^21–23 and epithelial-mesenchymal transition²⁴ by repression of tumor suppressor gene expression. Many of the protein constituents of PRCs are highly expressed in glioblastoma compared with normal brain, including BMI1,²⁵ suppressor of zeste 12 homolog–Drosophila (SUZ12),²⁶ and others.²⁷ Both PRC1 and PRC2 have been implicated in GSC maintenance,^28,29 and it has also been suggested that GSCs have acquired an oncogenic addiction to PRC activity.³⁰ Several reports have demonstrated a crucial role for both Polycomb complexes in maintaining the radioresistance of numerous cancer models and cancer stem cells,^31,32 including glioblastoma.^33,34 It is believed that the Polycomb proteins recruit the DNA damage response machinery and promote double-strand break repair.^35–37

Previously, we have shown that BMI1 mRNA expression is downregulated by miR-128 in glioma and GSCs. However, there have been reports that BMI1 upregulation by itself does not lead to tumorigenesis.³⁸ This could be potentially explained by the redundant and complementary functions of PRC components where downregulation of one component leads to upregulation of other components of PRC. We thus hypothesized that miR-128–mediated repression of BMI1, and consequently PRC1 activity, may be only a partial picture of a larger and more complex role in PRC regulation. In fact, here we report that miR-128 targets a component of PRC2 as well: SUZ12. It has been shown that the functionality of PRC2 as a complex depends on the presence of SUZ12.³⁹ Moreover, upregulation of SUZ12 was associated with the development of cancer stem cells.⁴⁰

Reestablishment of miR-128 expression blocks PRC activity, impairs GSC self-renewal, and increases radiosensitivity of GSCs, while loss of miR-128 expression appears to be an early event in gliomagenesis. These findings emphasize the role of microRNAs as molecules able to target activity of entire pathways by the simultaneous inhibition of multiple components, thus preventing their redundant functions.

Materials and Methods

Human Specimens

All tumor samples were obtained as approved by the institutional review board at The Ohio State University. Surgery was performed by E.A.C., I.N., and P.P. For each patient, samples of both tumor and brain devoid of gross tumor were resected, aliquoted, and processed for either extraction of total RNA (Trizol, Invitrogen) and protein (Cell Lysis Buffer, Cell Signaling) or isolation and establishment of patient-derived GSCs.

Cell Culture

U87 malignant glioma (MG) and U251MG glioblastoma cells were obtained from the American Type Culture Collection and cultured in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) supplemented with 10% fetal calf serum (Sigma). Primary human glioma stemlike cells GSC84, GSC528, GSC718, GSC816, GSC1123, AC17, and AC20 were obtained from freshly resected human MG specimens (GSC84, -528, -718, and -816 are World Health Organization [WHO] grade IV and GSC1123, AC17, and AC20 are WHO grade III) and grown, as previously reported,⁴¹ in stem cell medium, consisting of DMEM/F12/Glutamax (Invitrogen) supplemented with 1% N2 and 2% B27 (Invitrogen) and 20 ng/mL epidermal growth factor and fibroblast growth factor–2 (PeproTech). Cells were grown as adherent monolayers in poly-l-ornithine and laminin-covered dishes (Invitrogen) as previously described.⁴² The isolation and characterization of the human fetal NSCs SCP27⁴³ (a kind gift of Dr Kaspar from Columbus Nationwide Children's Hospital) and 16WF⁴⁴ were reported previously. They were grown as monolayer cultures as we have described. Mouse subventricular zone (SVZ) cells were obtained by harvesting mouse brain after euthanasia, isolating the SVZ region, dissociating tissue with TrypLE Express (Invitrogen), and culturing cells as neurospheres in stem cell medium.

Vector Construction

The lentiviral phosphorylated cadherin (pCDH; System Biosciences) construct carrying miR-128 was previously described.¹⁵ Transducing virus particles were obtained by co-transfecting human embryonic kidney 293 cells with pCDH/miR-128–1 plasmid DNA along with Lentiviral Plasmid Packaging Mix (System Biosciences) in the presence of Lipofectamine (Invitrogen) according to the manufacturer's recommendations. For the reporter assay, pMirTarget-BMI1, a luciferase vector containing the full-length (1.9 kb) BMI1 3′UTR was purchased from Origene. The control vector was obtained by removing the BMI1 sequence by restriction digestion. To obtain pMirTarget-SUZ12, the full-length SUZ12 3′UTR (1.7 kb) was amplified from U87MG genomic DNA and cloned into pMirTarget control vector. All the vectors produced in this study were sequenced to confirm sequence identity and in-frame accuracy. Similarly, a 468-bp fragment of the SUZ12 3′UTR, amplified from U87 genomic DNA and containing both target sequences for miR-128, was cloned into pMirTarget control vector to obtain the part of the 3′UTR containing pMirTarget-SUZ12. The construct was then mutated using the QuickChange kit (Stratagene), in which the predicted miR-128 target sequences actgtgg and actgtga were substituted with agtcagg and aaaggct, respectively. The anti–miR-128 vector (miRZip–miR-128 green fluorescent protein [GFP]) was purchased from System Biosciences. MiRZip-GFP (negative control) was obtained by removing the anti–miR-128 sequence from the parental construct by high fidelity PCR-mediated amplification. Sequences of all primers are shown in Supplementary Table S1.

Nanostring, Real-time PCR, and Database Analysis

Total RNA was extracted using Trizol (Invitrogen) and treated with RNase-free DNase (Qiagen) as previously described.¹⁵ In order to identify microRNAs that are specifically deregulated in glioblastoma stem cells, we compared the microRNA expression patterns of 10 samples (2 nonmalignant NSCs and 8 GSCs). The Nanostring microRNA technology was used to search for unique gene signatures linked to glioblastoma stem cells. Total RNA was used for the nCounter microRNA platform. All sample preparation and hybridization were performed according to the manufacturer's instructions. All hybridization reactions were incubated at 65°C for a minimum of 12 h. Hybridized probes were purified and counted on the nCounter Prep Station and Digital Analyzer (Nanostring) following the manufacturer's instructions. For each assay, a high-density scan was performed. For platform validation using synthetic oligonucleotides, Nanostring nCounter microRNA raw data were normalized for lane-to-lane variation with a dilution series of 6 spikes in positive controls. The sum of the 6 positive controls for a given lane was divided by the average sum across lanes to yield a normalization factor, which was then multiplied by the raw counts in each lane to give normalized values. All significantly deregulated microRNAs were used for visualization in a heatmap and analyzed by principal component analysis⁴⁵ using dChip software with the Statistical R package. The array was performed at The Ohio State University Comprehensive Cancer Center Microarray, Nucleic Acids, and Proteomic Shared Facilities with their technical assistance. Mature miR-128 expression analysis was carried out using a microRNA real-time (RT) PCR detection kit (Applied Biosystems), following manufacturer's protocol. Quantitative RT-PCR was performed using the Applied Biosystems Step One Plus PCR apparatus. U6 small nuclear RNA was used as the endogenous control. Of note, miR-128 and U6 probe sequences were identical for both human and mouse transcripts. Complementary DNA for RT-PCR was synthesized using iScript (BioRad). Analysis of mRNA expression was carried out using Power SYBR Green (Applied Biosystems) with human BMI1, SUZ12, and 18SrRNA as endogenous control. Sequences of all primers are shown in Supplementary Table S1. The collection of the data from The Cancer Genome Atlas (TCGA)⁴⁶ was compliant with all applicable laws, regulations, and policies for the protection of human subjects, and necessary ethical approvals were obtained. Analysis of all data was done in R project (http://www.r-project.org/). Agilent 8×15k microRNA expression and gene expression data for 206 glioma samples were downloaded along with clinical information from TCGA (level 2 normalized data, November 2012, http://tcga-data.nci.nih.gov/tcga/dataAccessMatrix.htm). Spearman's rank correlations were calculated for 2 probes for miR-128–1 and miR-128–2 against probes for SUZ12.

Cell Studies

Transfection of pre–miR-128 oligonucleotides (Ambion) at a final concentration of 50 nmol/L and of reporter plasmids was carried out with Lipofectamine 2000 (Invitrogen), following the manufacturer's instructions. To create stable glioblastoma cell lines expressing miR-128, U87MG and GSC528 cells were infected with pCDH-128-1-cGFP or pCDH-cGFP and allowed to grow for 7 days. Stably transfected cells were then sorted by GFP expression through flow cytometry (FACSAria III cell sorter, Becton Dickinson). Western blot analysis was carried out with whole cell protein lysate. Antibodies used were: BMI1 (Millipore), SUZ12, p21^Waf1/Cip1, H2a-K119^ubi, H3-K27^me3 and H3 (Cell Signaling), α-tubulin (Sigma), and CD133 (Amersham). Horseradish peroxidase–conjugated secondary antibodies against rabbit or mouse immunoglobulin G (Amersham) were then used. Cell proliferation following miR-128 transfection/transduction was quantified either by manual counting after cell dissociation or by the Click-it ethynyl-labeled (deoxy)uridine (EdU) cell proliferation assay kit (Invitrogen). Briefly, for cell counting, 150 000 cells were seeded on day 0 onto 6-cm well plates, transfected on day 1, and counted 4 days after transfection. For EdU studies, 10 000 cells were seeded onto 48-well plates and allowed to recover overnight. The following day, EdU was added to the medium to a final concentration of 10 μM. Cells were returned to the incubator for 3 h before they were fixed with 4% paraformaldehyde, and EdU incorporation was assessed following the manufacturer's protocol. For the clonogenic assay, cells were dissociated with TrypLE Express and plated with complete stem cell medium mixed with 0.4% low gelling temperature agarose (Sigma) pre-warmed at 37°C onto 6-well dishes over a layer of solidified 0.8% agarose. After allowing cells to settle down to the 0.8%/0.4% agarose interface, the dishes were kept at room temperature for 10 min to allow the medium to gel before placing the plates back at 37°C for culturing. Fresh stem cell medium was then used to cover the agarose and changed every 3 days. For DNA damage and response experiments, cells were irradiated with 2–4 Gy x-rays from an RS-2000 Biological Irradiator (Rad Source). Quantification of apoptosis and necrosis was carried out with the FITC (fluorescein isothiocyanate) Annexin V/Dead Cell Apoptosis Kit (Invitrogen) using a FACSCalibur machine (Becton Dickinson). Briefly, 150 000 cells were plated onto 6-cm dishes and allowed to recover overnight. The following day, cells were transfected with 50 nmol/L miR-128 or negative control oligonucleotides. After 36 h, cells were irradiated with 2 Gy or 4 Gy (except negative control samples), returned to the incubator for an additional 48 h, then collected for analysis following manufacturer's protocol. Evaluation of DNA repair was performed using the CometAssay Kit (Trevigen) according to the manufacturer's protocol. Briefly, 100 000 cells were plated onto 6-well plate dishes and allowed to recover overnight. The following day, cells were transfected with 50 nmol/L miR-128 or negative control oligonucleotides and allowed to grow for a further 36 h. Cells were then either left untreated or irradiated with 2 Gy and then harvested at 30 min and 15 h after exposure to ionizing radiation. For each sample, 500 dissociated cells were then embedded into agarose and fixed onto glass coverslips for quantification of DNA damage.

Immunohistochemistry

Human surgical specimens were fixed in formalin and embedded in paraffin before being sectioned at 4 μM thickness. Ventana BenchMark (Ventana Medical Systems) was used for staining. Expressions of BMI1 and SUZ12 were optimized at 1:500 and 1:250 dilutions, respectively. Each sample was counterstained with hematoxylin/eosin.

Statistical Analysis

Results are expressed as mean ± SD, and differences were compared using the 2-tailed Student's t-test. Statistical analyses were performed using Microsoft Office Excel 2010 or GraphPad Prism software. P < .05 was considered statistically significant (indicated by single asterisks in the Figures), and P < .01 was strongly significant (indicated by double asterisks).

Results

MiR-128 Targets mRNA Encoding SUZ12, a PRC2 Component

We previously demonstrated that miR-128 targets BMI1¹⁵ mRNA, a component of PRC1. Based on in silico analysis, we hypothesized that miR-128 might also target the 3′UTR of SUZ12 mRNA, a component of PRC2 (Fig. 1A). To test this hypothesis, we performed Western blotting using human glioblastoma cell lines (U87 and U251) and human GSCs (GSC528). As shown in Fig. 1B (left panels), transfection of miR-128 precursors led to a significant reduction in SUZ12 expression in all 3 cells. This was also evident when GSC528 cells were stably transfected with a lentiviral vector expressing pri–miR-128 (Fig. 1B, right panel). As a control, miR-128 also led to the expected reduction of BMI1. To demonstrate specific targeting, we used constructs containing a short fragment of SUZ12 3′UTR encompassing miR-128 binding sites and showed significantly reduced luciferase expression upon transfection with miR-128 precursors. Additionally, mutations introduced into both predicted miR-128 binding sites of the SUZ12 3′UTR led to the restoration of luciferase levels, demonstrating the specificity of miR-128 for these binding sites (Fig. 1C). Interestingly, mutation of only 1 out of 2 miR-128 binding sites led to only partial restoration of luciferase activity, while mutation of both sites led to full restoration. Therefore, miR-128 suppression of SUZ12 expression is most effective when both target sites of the SUZ12 mRNA are targeted. We also demonstrated targeting of both SUZ12 and BMI1 mRNAs by miR-128 using full-length 3′UTRs from both genes (Supplementary Fig. S1A).

Fig. 1.

Identification and characterization of SUZ12 as target of miR-128 in established glioma cell lines and glioma stem cells. (A) Schematic representation of miR-128 3′UTR seeding regions. Sequence of both miR-128 target sites within human SUZ12 mRNA 3′UTR. (B) MiR-128 regulation of BMI1 and SUZ12 protein levels. U87 and U251 glioma cell lines and the GSC528 (528) glioma stem cell line were transiently transfected with microRNA negative control (NC) or miR-128 precursor (128) or stably infected with either pCDH vector or pCDH vector expressing miR-128 (pCDH128) (GSC528 cells). Cell lysates were blotted with anti-BMI1 or anti-SUZ12 antibodies; anti–α-tubulin antibody (αTUB) was used as loading control. (C) Direct targeting of SUZ12 3′UTR by miR-128. Human embryonic kidney 293 cells were transfected (n = 3 experiments, each performed in triplicate) with a luciferase reporter vector containing a fragment of SUZ12 wild type (wt) or mutated in position 1141–1147 (m1) or in both positions 1141–1147 and 1389–1395 (m1m2) and co-transfected with either microRNA NC or pre–miR-128 (128). Vector without 3′UTR insert (no) was used as an additional control. Luciferase levels were quantified as luciferase activity per milligram of protein. Data are shown as mean relative to controls ± SD. RLU, relative luciferase activity. **P < .01. (D) Effect of SUZ12 knockdown on expression of BMI1. U87, U251, and GSC528 cell lines were transiently transfected with either siRNA NC or siRNA against BMI1 or SUZ12. Cell lysates were blotted with anti-BMI1 or anti-SUZ12 antibody; anti–α-tubulin antibody (αTUB) was used as loading control. Data were quantified by densitometric analysis using ImageJ software, normalized to loading control. Relative expression to averaged NC level is shown.

We then compared the effect of miR-128 on both genes with their specific knockdown using small interfering (si)RNAs (Fig. 1D). Interestingly, siRNA-mediated knockdown of SUZ12 led to a significant increase in BMI1 levels in all tested glioblastoma cells, including GSC528. This result confirmed the previously described phenomenon of an increase in BMI1 expression after selective downregulation of PRC2,⁴⁷ suggesting the existence of a compensatory mechanism that leads to elevation in the levels of one PRC component if another PRC component is “knocked down” by siRNA. MiR-128 overrides this mechanism as it simultaneously knocks down both PRCs (Fig. 1B).

Clinical Significance of PRC Expression

Previously, we had shown that there was opposite expression of miR-128 and BMI1 in human glioblastoma samples.¹⁵ To examine the clinical significance of SUZ12 in glioblastoma, we screened for the expression of SUZ12 in matched (ie, from the same individual) glioblastoma and adjacent brain samples from human subjects. Quantitative RT-PCR showed that the expression of SUZ12 was significantly upregulated in tumor compared with its adjacent brain tissue that was devoid of gross tumor (Fig. 2A). The result was confirmed by Western blot analysis of protein lysates from 3 of these samples (Fig. 2B) and immunohistochemistry (Fig. 2C). As a control, BMI1 also showed the same pattern of expression as previously reported by us (Supplementary Fig. S2A–C). Therefore, PRC components are highly expressed in human glioma specimens, similar to levels that had been detected in GSCs and established glioma cancer cell lines such as U87 and U251 (Fig. 1B).

Fig. 2.

SUZ12 deregulation in glioblastoma patients. (A) SUZ12 mRNA level in brain tumor tissue (BT) in comparison with matched (ie, from the same patient) brain adjacent to the tumor (BAT) from glioblastoma patients. Expression of SUZ12 was validated by quantitative (q)RT-PCR (n = 9). Mean relative gene expression level ± SD is shown. (B) SUZ12 protein levels in BT in comparison with matched (from the same patient) BAT from glioblastoma patients. Cell lysates were blotted with anti-SUZ12 antibody; anti–α-tubulin antibody (αTUB) was used as loading control (n = 3). (C) Number of SUZ12-positive cells in BT in comparison with matched (from the same patient) BAT from glioblastoma patients. The percentage of positive cells per random field was determined (left panel) in BT and BAT sections from 3 matched patient samples; results are shown as mean ± SD on the graph, *P < .05. Representative images of SUZ12 in immunohistochemistry staining in paraffin-embedded tissue. Scale bars, 50 µm. (D) Levels of miR-128 in BT in comparison with matched (from the same patient) BAT from glioblastoma patients. Expression of miR-128 was validated by qRT-PCR (n = 9). Mean relative microRNA expression level ± SD is shown. **P < .01. (E) Association of miR-128 expression with patient survival (Kaplan–Meier [KM] plot) in glioblastoma. Data obtained from The Cancer Genome Atlas. Upregulated (n = 21), downregulated (n = 20), and intermediary samples (n = 155) were analyzed. Up- vs downregulated P = .0308; upregulated vs intermediary P = .0018; downregulated vs intermediary P = .8322. (F) A miR-128–based signature in NSCs and GSCs. Heatmap showing the top 4 most significantly downregulated (fold <0.5, P < .05) microRNAs based on a Nano String array. (G) MiR-128 expression in GSCs derived from glioblastoma patients. Expression of miR-128 was validated by qRT-PCR. Mean relative gene expression level ± SD is shown. NSC n = 2, GSC n = 6.

In contrast, the levels of mature miR-128 in glioblastomas were significantly (∼8-fold) reduced compared with those from matched adjacent brain (Fig. 2D). We attempted to correlate miR-128 downregulation with SUZ12 protein upregulation in brain tumor tissue compared with their matched adjacent tissue. Only 3 samples had sufficient tissue to extract both mRNA and protein. We observed a significant inverse correlation between miR-128 and SUZ12 (Supplementary Fig. S2D). Interestingly, when we queried TCGA⁴⁶ for human glioblastoma patient survival linked to the level of miR-128 expression, we found that the group with upregulated miR-128 had a median survival of ∼620 days, whereas groups with intermediary and low expression of miR-128 had a significantly shorter median survival of only ∼350 days (Fig. 2E). These results thus validated miR-128 expression as significantly downregulated and SUZ12 and BMI1 as significantly upregulated in human glioblastoma compared with matched surrounding brain tissues from the same patient and associated miR-128 expression with improved survivorship. Additionally, we correlated the expression of SUZ12 and miR-128 based on the database of TCGA. Spearman's rank correlations showed that SUZ12 expression inversely correlated with expression of both miR-128–1 and miR-128–2 in a significant fashion (Supplementary Fig. S2E). Moreover, we queried the Repository of Molecular Brain Neoplasia Data and TCGA to determine the prognostic relevance of SUZ12 expression and whether there was an inverse correlation between SUZ12 and miR-128 expression. We found that human subjects whose gliomas expressed upregulated SUZ12 had a median survival of only ∼390 days, whereas those with intermediate expression of SUZ12 had a significantly longer median survival of ∼580 days (Supplementary Fig. S2F). To further establish the importance and selectivity of miR-128 downregulation in the glioblastoma stem cell compartment, we performed a global analysis of microRNA expression in 2 nonmalignant fetal NSC lines and 8 GSC lines derived from freshly resected human glioblastoma specimens. As shown in Fig. 2F, miR-128 was among the top 4 most significant differentially expressed microRNAs when comparing nonmalignant NSCs with GSCs. Furthermore, the signature of these 4 microRNAs (including miR-128) was sufficient to discriminate between NSCs and GSCs (Supplementary Fig. S2G). We validated this by quantitative RT-PCR analysis, underlining the significant difference in miR-128 expression between NSCs and GSCs (Fig. 2G). These results show that miR-128 is downregulated not only in glioma nonstem cells—a finding that was expected because in normal brain, miR-128 is expressed preferentially in neurons⁴⁸—but also in the GSC compartment.

Phenotypic Changes in GSCs Upon MiR-128 Expression

In order to study the phenotypic effect of miR-128 on GSCs, we overexpressed it either transiently, by introducing miR-128 precursor oligonucleotides, or stably, by using a lentiviral construct expressing the primary miR-128–1 sequence (overexpression of miR-128 upon lentiviral infection in GSC528 and U87 cell lines is shown in Supplementary Fig. S3A). Both delivery strategies resulted in upregulation of the tumor suppressor protein p21^CIP1 as expected⁴⁹ (Fig. 3A). PRC1 instigates ubiquitination of histone 2A at Lys119 (H2A K119^ubi), while PRC2 prompts trimethylation of histone 3 at Lys27 (H3 K27^me3).²¹ Both PRC-dependent histone modifications were significantly diminished upon enforced miR-128 expression (Fig. 3A). These molecular changes also occurred in the established glioblastoma cell lines U87 and U251 (Supplementary Fig. S3B). Additionally, stable expression of miR-128 led to a decreased level of the GSC marker CD133, suggesting reduced “stemness” of GSC528. As expected, siRNA-mediated knockdown of SUZ12 resulted in elevation of the PRC1-dependent H2A K119^ubi marker (Fig. 3B), in agreement with specific knockdown of SUZ12 leading to BMI1 overexpression (Fig. 1D). Interestingly, siRNA-mediated knockdown of BMI1 led to an increase in the PRC2-dependent H3 K27^me3 marker (Fig. 3B), as shown for another component of PRC1,⁵⁰ although SUZ12 expression levels did not seem to change (Fig. 1D). This suggests that there was functional redundancy or complementation between PRC1 and PRC2 function. Incorporation of bromodeoxyuridine in U87 and U251 glioblastoma cells and cell counting assays in GSC528 showed a significant reduction in proliferative potential when miR-128 expression was reestablished (Fig. 3C). SiRNA-mediated knockdown of SUZ12 and BMI1 also led to decreased proliferative potential of GSC528 (Supplementary Fig. S3C). To further examine changes in colony formation of GSCs, we utilized a soft-agar colony formation assay with GSC528 stably transduced with miR-128. MiR-128 expression caused a 2-fold reduction in colony forming ability compared with controls (Fig. 3D). There was also a different distribution of colony size. Cells overexpressing miR-128 produced predominantly small and medium-sized colonies, while control cells produced larger colonies (Supplementary Fig. S3D). These results show that miR-128 expression diminished PRC-dependent chromatin modifications, thus underscoring the significance of miR-128 on overall activity of PRCs, in contrast to siRNA-mediated knockdown of only 1 component leading to functional complementation. This was also associated with reduced glioblastoma cell proliferation and GSC colony formation.

Fig. 3.

Expression of miR-128 in GSCs alters histone modifications and reduces proliferation and self-renewal. (A) MiR-128 effects on levels of p21, on histone modifications, and on stemness marker. GSC528 cells (528) were transiently transfected with microRNA negative control (NC) or miR-128 precursor (128) or stably infected with either pCDH vector or pCDH vector expressing miR-128 (pCDH128). Cell lysates were blotted with anti-p21 WAF1/CIP1, anti-H2A K119^ubi, anti-H3 K27^me3, or anti-CD133 antibody; anti–total H3 or anti–α-tubulin (αTUB) antibody was used as loading control (n = 3). (B) Effect of specific knockdown of SUZ12 and BMI1 on PRC-dependent histone modification. GSC528 cell line was transiently transfected with either siRNA NC or siRNA against SUZ12 or BMI1. Cell lysates were blotted with anti-H2A K119^ubi, anti-H3 K27^me3, and anti-SUZ12 and anti-BMI1 antibodies; anti–total H3 or anti–α-tubulin antibodies (αTUB) were used as loading control. Data were quantified by densitometric analysis using ImageJ software, normalized to loading control. Relative expression to averaged NC level is shown. (C) MiR-128 effects on the proliferative potential of glioma cells. U251 and U87 cells were transiently transfected with microRNA NC or miR-128 precursor (128), or GSC528 cells (528) were stably infected with either pCDH vector or pCDH vector expressing miR-128 (pCDH128). Bromodeoxyuridine (BrdU) incorporation in U251 and U87 (top panel) and GSC528 cell number (bottom panel) were quantified in 3 independent experiments; data are expressed as mean ± SD. *P < .05, **P < .01. (D) MiR-128 effects on the self-renewal capabilities of GSCs. GSC528 cells (528) were stably infected with either pCDH vector or pCDH vector expressing miR-128 (pCDH128). Representative images of soft-agar colony formation (n = 3) are shown. Results expressed as mean number of colonies per well ± SD are shown on the right. **P < .01.

Reexpression of MiR-128 in GSC and Glioblastoma Lines Increases Radiosensitivity

PRCs have been shown to be activated by radiation, as they play a significant role in DNA repair.^31,33,35,37 First, we established that the response of glioblastoma cells to radiation involves a dose-dependent increase in the expression of the Polycomb genes SUZ12 and BMI1 by quantitative RT-PCR (Fig. 4A). The observed increase in mRNA level in GSC528 cells was accompanied by increases in SUZ12 and BMI1 total protein levels (Fig. 4B), and miR-128 expression reduced this increase. To test whether miR-128–driven downregulation of PRC complexes conferred increased radiosensitivity to GSCs, we determined the effect of radiation on the clonal growth of miR-128–expressing GSCs. The results in Fig. 4C show that miR-128 expression significantly increased the radiosensitivity of GSC528 cells. A dose of 4 Gy in combination with forced expression of miR-128 eradicated GSC528 cells. MiR-128–expressing GSC528 plated on laminin-coated dishes irradiated with 2 Gy showed significant decrease in cell number when we combined both radiation and miR-128 expression compared with radiation or miR-128 alone (Fig. 4D).

Fig. 4.

Effect of miR-128 expression on radiosensitivity. (A) SUZ12 and BMI1 expression after radiation exposure. U87, U251, and GSC528 (528) cell lines were untreated or radiated with 2-Gy or 4-Gy doses and harvested 1 h after radiation. Expression of SUZ12 and BMI1was validated by quantitative RT-PCR (n = 3). Mean relative gene expression level ± SD is shown. RQ, relative quantification. (B) MiR-128 effects on SUZ12 and BMI1 overexpression induced by radiation. GSC528 cells (528) were transiently transfected with microRNA negative control (NC) or miR-128 precursor (128), and after 36 h cells were left untreated or were irradiated using a 2-Gy or 4-Gy dose and harvested 36 h after radiation. Cell lysates were blotted with anti-SUZ12 and anti-BMI1 antibodies, and anti–α-tubulin antibody (αTUB) was used as loading control (n = 3). Data were quantified by densitometric analysis using ImageJ software, normalized to loading control. Average of relative expression level is shown. (C) MiR-128 effects on the clonogenic potential of GSC cells after radiation. GSC528 cells (528) were stably infected with either pCDH vector or pCDH vector expressing miR-128 (pCDH128), plated in soft agar, and left untreated or irradiated using a 2-Gy or 4-Gy dose. Representative images of soft-agar colony formation (n = 3) after 14 days are shown (left panels). Data are expressed as mean number of colonies per well (right panel) ± SD. *P < .05. (D) MiR-128 effects on the proliferation potential of GSC cells after radiation. GSC528 cells (528) were placed on laminin-coated dishes, transiently transfected with microRNA NC or miR-128 precursor (128), and after 36 h cells were left untreated or irradiated using a 2-Gy dose. Representative images of growing cells (n = 3) after 48 h are shown (left panels). Data are expressed as mean number of cells per well × 10⁵ (right panel) ± SD. *P < .05, **P < .01. (E) MiR-128 and radiation effects on glioma cell viability. GSC528 cells (528) were transiently transfected with microRNA NC or miR-128 precursor (128), and after 36 h cells were left untreated or radiated using a 2-Gy dose. After 48 h cell flow cytometry was performed. Data are shown as percentages of propidium iodide–positive (top) and –negative (bottom) cells staining. (F) MiR-128 effects on the DNA repair capability of radiated cells. GSC528 cells (528) were transiently transfected with microRNA NC or miR-128 precursor (128), and after 36 h cells were left untreated or radiated using a 2-Gy dose. A comet assay was performed either with nonirradiated cells or after 30 min and 15 h post-irradiation. Representative images of cells with fragmented DNA tails are shown (left) under low power and high power (inset) magnification. Data are expressed as mean number of cells with tails (right) ± SD. **P < .01.

As determined by propidium iodide staining and flow cytometry, miR-128 alone or radiation alone had little effect on the percentage of propidium iodide–positive cells, but combining miR-128 and radiation led to significant cytotoxicity (Fig. 4E). We attempted to determine the reason for the observed increase in miR-128–mediated cytotoxicity upon radiation. We demonstrated by a comet assay that cells expressing miR-128 displayed weaker DNA repair capability compared with control cells: 15 h after radiation ∼50% of miR-128–expressing cells had fragmented DNA tails (indicative of damaged DNA) compared with 25% of control cells (Fig. 4F). In conclusion, miR-128 appears to prevent irradiated cells from switching on a DNA repair program mediated by increased expression of Polycomb genes, rendering them more radiation sensitive.

Loss of Expression of MiR-128 Occurs Early in Gliomagenesis

To investigate further the role of miR-128 in glioblastoma biology, we utilized a mouse genetic model of glioma. We used mut3 mice (hGFAP-cre; Nf1^flox/+; Trp53^−/+) and their wild-type counterparts. Mut3 mice represent a genetic model in which initially healthy mice eventually develop malignant astrocytomas with 100% penetrance by 20–30 weeks of age.^51,52 We collected RNA and protein from GSCs obtained from the tumors of glioma-symptomatic, adult mut3 mice, as well as from the same brain region of wild-type mice of corresponding age. We were able to show significant downregulation of MIR128 expression in GSCs derived from the SVZs of mut3 mice (Fig. 5A). The loss of MIR128 expression in mut3 GSCs corresponded with elevated levels of SUZ12 and BMI1 in the same cells (Fig. 5B, quantification shown in Supplementary Fig. S4A). This confirmed that the same relationship between MIR128 and PRCs existed in this mouse genetic model of glioma as existed in human glioblastoma samples and in human GSCs.

Fig. 5.

Effect of miR-128 downregulation on clonogenic potential of nonmalignant NSCs. (A) MIR128 expression in GSCs derived from SVZs of mut3 mice. RNAs were extracted from the cells isolated from SVZs of wild-type (wt) mice (wt NSC, n = 4) or symptomatic (20–24 wk old) mut3 mice (mut3 GSC, n = 3). Expression of MIR128 was validated by quantitative (q)RT-PCR. Mean relative gene expression level ± SD is shown. (B) SUZ12 and BMI1 protein level expression in GSCs derived from SVZs of mut3 mice. Protein lysates were extracted from the cells isolated from SVZs of wt mice (wt NSC, n = 3) or symptomatic (20–24 wk old) mut3 mice (mut3 GSC, n = 3). Cell lysates were blotted with anti-SUZ12 or anti-BMI1 antibodies; anti–α-tubulin antibody (αTUB) was used as loading control. (C) MIR128 expression in NSCs from SVZs of mut3 mice. RNA was extracted from the cells isolated from SVZs of wt mice (wt NSC, n = 4) or young, presymptomatic (4 wk old) mut3 mice (mut3 NSC, n = 3). Expression of MIR128 was validated by qRT-PCR. Mean relative gene expression level ± SD is shown. (D) BMI1 and SUZ12 protein level expression in NSCs from SVZs of mut3 mice. Protein lysates were extracted from the cells isolated from SVZs of young wt mice (wt NSC, n = 3) or young, presymptomatic (4 wk old) mut3 mice (mut3 NSC, n = 3). Cell lysates were blotted with anti-SUZ12 or anti-BMI1 antibodies; anti–α-tubulin antibody (αTUB) was used as loading control (n = 3). (E) Effects of MIR128 knockdown on target protein levels in wt NSCs and on clonogenic potential. Protein lysates were extracted from the cells isolated from SVZs of young wt mice. Cells were stably infected with miR-ZIP control vector (ZIP) or miR-ZIP vector coding hairpin for miR-128 (128-ZIP). Cell lysates were blotted with anti-SUZ12 or anti-BMI1 antibodies; anti–α-tubulin antibody (αTUB) was used as loading control (n = 3) (left panel). Representative images of soft-agar colony formation (n = 3) after 14 days are shown (middle panels). Results expressed as mean number of colonies per well ± SD are shown on the right. *P < .05. (F) Effects of miR-128 knockdown on target protein levels in human SCP27 NSCs and on clonogenic potential. Cells were stably infected with miR-ZIP control vector (ZIP) or miR-ZIP vector coding hairpin for miR-128 (128-ZIP). Cell lysates were blotted with anti-SUZ12 or anti-BMI1 antibodies; anti–α-tubulin antibody (αTUB) was used as loading control (n = 3) (left panel). Representative images of soft-agar colony formation (n = 3) after 14 days are shown (middle panels). Results expressed as mean number of colonies per well ± SD are shown on the right. *P < .05.

We then tested whether MIR128 expression was altered in mice before the onset of glioma formation in the NSC niche, shown to be the glioblastoma cell of origin in this model.⁵³ We harvested NSCs from the SVZs of wild-type and mut3 mice at 4 weeks of age, well before the onset of any glioma-related symptoms or the appearance of tumors (which usually occur at the age of 20–30 wk), and harvested RNA and protein. As shown in Fig. 5C, MIR128 was significantly downregulated in the NSCs from the young, presymptomatic mut3 mice. This downregulation was accompanied by significantly increased levels of SUZ12 and BMI1 proteins (Fig. 5D, quantification shown in Supplementary Fig. S4B) and significantly elevated proliferation compared with wild-type, age-matched NSCs from littermates (Supplementary Fig. S4C). When we knocked down the expression of MIR128 in wild-type NSCs by introduction of a lentiviral miR-128–ZIP vector (data not shown), we observed a significant increase of SUZ12 and BMI1 levels. Moreover, such cells displayed increased clonogenicity as measured by increased colony number in a soft-agar colony formation assay (Fig. 5E). Also, large colonies were formed almost exclusively by cells infected with a miR-128–ZIP vector (Supplementary Fig. S4D).

Finally, to test the relevance of these findings in a human cell model, we performed similar experiments using the human fetal NSC line SCP27. Similar to mouse cells, introduction of a lentiviral miR-128–ZIP vector to SCP27 fetal NSCs brought about significant reduction of miR-128 expression (data not shown) and increased SUZ12 and BMI1 levels, followed by a hyperproliferative phenotype as measured in a soft-agar colony formation assay (Fig. 5F). These findings thus indicate that loss of miR-128 expression is an early event in the course of gliomagenesis in the mouse model, and similar findings in human fetal stem cells suggest that this conclusion may also be valid for human glioblastoma.

Discussion

The regulation of PRCs in tumor and glioma pathogenesis is not well known. Our previously published work uncovered that miR-128 is a regulator of BMI1 gene expression in gliomas. A number of microRNAs function by repressing the expression of multiple genes in a signaling pathway. In this instance, we wondered whether miR-128 repressed the expression of other components of PRC, as predicted by target recognition software. Here we report, for the first time, that (i) miR-128 simultaneously targets important constituents of Polycomb Repressor Complexes 1 and 2 and that its downregulation in glioblastoma contributes to their high level of expression compared with normal brain; (ii) specific knockdown of SUZ12 (PRC2) resulted in elevated expression of the PRC1 component BMI1 and consequently elevated PRC1 activity, while specific knockdown of BMI1 (PRC1) resulted in increased activity of PRC2, suggesting redundancy between both PRCs as postulated previously;⁵⁴ (iii) miR-128 expression correlates clinically with improved survivorship in subjects with glioblastoma; (iv) miR-128 is 1 of 4 components of microRNAs that are sufficient to discriminate between NSCs and GSCs; (v) miR-128 expression increases radiosensitivity of GSCs; and (vi) miR-128 downregulation with corresponding SUZ12 and BMI1 upregulation occurs in SVZ-derived NSCs in 4-week-old mice genetically engineered to develop gliomas by weeks 20–30. Taken in conjunction, these novel findings provide evidence of the important regulatory role of miR-128 in PRC function and point to deregulation of this role as an early event in glioma.

Restoration of miR-128 expression in GSCs is biologically consistent with our understanding of cancer as a disease of multiple aberrant signaling pathways that would thus require intervention at multiple levels. The fact that a single microRNA (such as miR-128) simultaneously downregulates a broad set of mRNA targets with potential pro-oncogenic properties provides a broad modulatory mechanism that affects cancers, such as glioblastomas. Our results also show that miR-128–mediated targeting of both PRCs prevents the PRC compensatory mechanism whereby specific knockdown of a PRC2 component leads to upregulation of PRC1 activity and vice versa, underlining the unique properties of microRNA-based interventions and providing the opportunity for the discovery of new therapeutic approaches. Biologically, evidence for specific inhibitory targeting between 2 molecules is carried out by rescuing the phenotype inhibited by 1 molecule by overexpression of the targeted molecule. This experiment is unlikely to work for microRNAs, since each microRNA generally targets multiple mRNAs often linked in several signaling pathways that all affect a phenotype. Rescue in this case would involve reexpressing all of the inhibited mRNAs. This is also true for miR-128: trying to rescue the antiproliferative phenotype observed with miR-128 expression by overexpression of SUZ12 is unlikely to work because miR-128 expression also targets multiple other mRNAs linked to proliferation (such as BMI1 and others). We thus believe that the functional readout that we have used (levels of well-established PRC-specific histone modifications) is more suited to prove a direct link between miR-128 and the activity of both the PRC1 and PRC2 complexes.

The expression of miR-128 and the PRC components SUZ12 and BMI1 is inversely correlated in human glioblastoma patient samples, as we demonstrated on multiple levels (microRNA/mRNA, proteins, and immunohistochemistry), and relatively higher expression of miR-128 in patient samples correlated with better survival, while higher expression of SUZ12 correlated with poorer survival. Additionally, analysis of expression of miR-128 (along with 3 other microRNAs) was sufficient to discriminate between human NSCs and GSCs. We also demonstrated that miR-128–mediated downregulation of PRC components led to a significant inhibition of PRC-dependent histone modifications, decreased expression of CD133, derepression of the tumor suppressor p21, decreased glioma cell proliferation, and decreased GSC self-renewal. Interestingly, specific, siRNA-mediated downregulation of SUZ12 (PRC2) or BMI1 (PRC1) led to increased activity of PRC1 or PRC2, respectively.

GSCs are thought to be characterized by a high degree of chemo- and radiotherapy resistance.^55–59 This causes evasion of these therapeutic modalities, leading to inevitable recurrence and increased resistance to conventional therapy. We demonstrated that ectopic expression of miR-128 in GSCs significantly increases their radiosensitivity by preventing the radiation-induced increase of expression of PRC components, possibly by impaired DNA repair. Other researchers demonstrated that DNA damage led to redistribution of PRC components toward altered chromatin rather than to their transcriptional activation.³³ Although miR-128 stops proliferation, a phenotype that is linked to increased radiosensitivity, its action on inhibition of PRC expression and function known to be linked to DNA repair after radiation is likely responsible for this otherwise paradoxical effect. MicroRNAs have attracted intense research that points toward therapeutic potential and immediate targeting for sensitization of glioblastoma cells to chemo- and/or radiotherapy.⁶⁰ Our results present a strong rationale for microRNA-mediated molecular targeting of oncogenic pathways to overcome resistance to conventional therapy.

Recently published findings have provided compelling evidence that expression of MIR128 is severely impaired in a mouse genetic model of glioma (H-RasV12, Trp53^+/−) and that its reexpression prevents glioma development.⁶¹ Our results confirmed loss of MIR128 in a mouse model of glioma, mut3 (Nf1^flox/+; Trp53^−/+), which is accompanied by elevated levels of SUZ12 and BMI1. Moreover, we demonstrated that loss of MIR128 expression and the concomitant increase in PRC activity are observed in young, presymptomatic animals, suggesting that this loss is an early event in brain cells that are poised to become tumorigenic but have not originated a tumor yet. We also showed that knockdown of miR-128 in mouse and human nonmalignant NSCs led to increased clonality. Whether this increase leads to enhanced tumorigenesis in vivo will be the subject of future studies. Additionally, miR-128 has the ability to target a broad range of oncogenic targets (epidermal growth factor receptor and platelet derived growth factor receptor signaling,⁶¹ E2F3,⁶² and p70S6K⁶³ among others). Recently published work has identified the oncoprotein SNAIL as a direct negative regulator of miR-128 in breast cancer.⁶⁴ Interestingly, it was demonstrated that both p53 and neurofibromatosis type 1 repress SNAIL,^65,66 providing a plausible explanation for miR-128 downregulation in mut3 cells. In ovarian cancer, SNAIL was shown to induce a stemlike phenotype and radioresistance.⁶⁷ Finally, SNAIL is targeted by p53-inducible miR-34,⁶⁸ which was identified as a tumor suppressor in glioma cells.^69,70 It remains unclear whether the mechanism of regulation of miR-128 in glioma cells is the same as in breast cancer.

It has been postulated that microRNA replacement approaches have strong therapeutic potential because of the fact that single microRNAs can regulate multiple oncogenic pathways that are commonly deregulated in cancer.⁷¹ Using GSCs in this study allowed us to demonstrate that miR-128 is functional both in this subpopulation and in more differentiated glioma cells.

Our report, providing novel evidence that miR-128 is inactivated early in the course of gliomagenesis, simultaneously targets major components of oncogenic PRC signaling and renders GSCs more susceptible to conventional therapy, pointing toward miR-128 replacement as an attractive candidate for this therapeutic modality.

Supplementary Material

Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/).

Funding

This work was supported by an Neurosurgery Research and Education Foundation Research Fellowship (American Association of Neurological Surgeons) awarded to P.P. and E.A.C.