Targeting Glioma Stem Cells by Functional Inhibition of a Prosurvival OncomiR-138 in Malignant Gliomas

Cell Reports

Article Targeting Glioma Stem Cells by Functional Inhibition of a Prosurvival OncomiR-138 in Malignant Gliomas Xin Hui Derryn Chan,1,5 Srikanth Nama,1,5 Felicia Gopal,1 Pamela Rizk,1 Srinivas Ramasamy,1 Gopinath Sundaram,1 Ghim Siong Ow,2 Ivshina Anna Vladimirovna,2 Vivek Tanavde,2 Johannes Haybaeck,3 Vladimir Kuznetsov,2 and Prabha Sampath1,4,* 1Institute

of Medical Biology, Agency for Science Technology and Research, Singapore 138648, Singapore Institute, Agency for Science Technology and Research, Singapore 138671, Singapore 3Department of Neuropathology, Medical University of Graz, 8036 Graz, Austria 4Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 117597, Singapore 5These authors contributed equally to this work *Correspondence: [email protected] http://dx.doi.org/10.1016/j.celrep.2012.07.012 2Bioinformatics

SUMMARY

Malignant gliomas are the most aggressive forms of brain tumors, associated with high rates of morbidity and mortality. Recurrence and tumorigenesis are attributed to a subpopulation of tumor-initiating glioma stem cells (GSCs) that are intrinsically resistant to therapy. Initiation and progression of gliomas have been linked to alterations in microRNA expression. Here, we report the identification of microRNA-138 (miR-138) as a molecular signature of GSCs and demonstrate a vital role for miR-138 in promoting growth and survival of bona fide tumor-initiating cells with self-renewal potential. Sequence-specific functional inhibition of miR-138 prevents tumorsphere formation in vitro and impedes tumorigenesis in vivo. We delineate the components of the miR-138 regulatory network by loss-of-function analysis to identify specific regulators of apoptosis. Finally, the higher expression of miR-138 in GSCs compared to non-neoplastic tissue and association with tumor recurrence and survival highlights the clinical significance of miR-138 as a prognostic biomarker and a therapeutic target for treatment of malignant gliomas. INTRODUCTION Malignant gliomas are comprised of anaplastic astrocytoma (WHO grade III) and glioblastoma multiforme (GBM) (WHO grade IV) lesions that are highly invasive and display histological evidence of malignancy. Patients with such highly aggressive, often lethal intracranial malignancy have a median survival of less than 12 months (Legler et al., 1999). Composed of a heterogeneous mixture of poorly differentiated neoplastic astrocytes, malignant gliomas primarily affect adults and preferentially occur

in the cerebral hemispheres. Despite the current multimodality of treatment strategies (Garden et al., 1991), frequent recurrence of the tumors has been attributed to intrinsically resistant glioma stem cells (GSCs) critical for gliomagenesis (Bao et al., 2006a; Galli et al., 2004; Hemmati et al., 2003; Singh et al., 2004). They are chemo- and radio-resistant and, therefore, responsible for tumor progression and recurrence (Bao et al., 2006a). The infiltrative nature of GSCs coupled with their resistance to standard therapy has restricted the development of successful treatment. The molecular mechanisms that contribute to gliomagenesis are poorly understood. Alterations in the expression of endogenous small noncoding RNAs, the microRNAs (miRNAs), may play a crucial role in cancer initiation, progression, and metastasis (Calin and Croce, 2006; Ma et al., 2007). miRNAs can function as tumor suppressors or oncogenes by binding to complementary sequences on the respective target mRNA (EsquelaKerscher and Slack, 2006; Slack and Weidhaas, 2006). The reduced/loss of expression of a miRNA may lead to aberrant expression of its target proteins, resulting in an altered phenotype. Impaired miRNA regulatory network is known to be one of the key mechanisms in brain tumor pathogenesis (Bottoni et al., 2005; Chan et al., 2005). Comparison of the miRNA profiles of glial tumors, embryonic stem cells (ESCs), neural precursor cells (NPCs), and normal adult brain tissues has revealed that primary gliomas have NPC-like miRNA expression (Lavon et al., 2010). Specific role of miRNA-21 (miR21), miR-451, and miR-9/9* in GBM suggests that they are potential targets for therapeutic intervention (Chan et al., 2005; Corsten et al., 2007; Godlewski et al., 2010; Schraivogel et al., 2011). Overexpression of miR-34a, miR-137, and miR-451 inhibits proliferation, disrupts neurosphere formation, or induces differentiation of GSCs, respectively, but does not lead to apoptotic death of GSCs (Gal et al., 2008; Guessous et al., 2010; Li et al., 2009; Silber et al., 2008). In this study we demonstrate the role of miR-138 as a prosurvival oncomiR for GSCs and a regulator of GSC growth and survival. We further demonstrate the association of miR-138 with tumor recurrence and survival.

Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors 591

Figure 1. GSCs Display a Unique miRNA Signature (A) Graphic representation of differential miRNA expression. The scatterplot represents the transcript abundance of miRNAs in GSCs versus NSCs. The vertical axis shows the transcript abundance in NSCs, whereas the horizontal axis shows the transcript abundance in GSCs. (B) miR-138 expression analysis in NSCs and GSCs by quantitative stem-loop real-time reversetranscription PCR Three NSC lines are represented as SVZ, FL, Lon (Lonza), and five GSC lines as NNI-1, NNI-4, NNI-8, NNI-11, and NNI-12. (C) Expression of miR-138 in NSCs and GSCs validated by northern blot analysis. (D) Expression of pre-miR-138-1 and pre-miR138-2 in NSCs and GSCs. (E) miR-138 expression analysis by quantitative stem-loop real-time reverse-transcription PCR in GSCs on differentiation. N.S., not significant; **p < 0.001. See also Figure S1.

RESULTS GSCs Display a Unique miRNA Signature GSCs derived from five patients with malignant glioma, NNI-1, NNI-4, NNI-8, NNI-11, and NNI-12 (Ng et al., 2012), were first characterized for their self-renewal, multipotency, and tumorigenic properties. Similar to the normal human neural stem cells (NSCs), all GSC lines form spheres in culture and express markers like CD133 (Prominin-1) and Nestin (Figures S1A and S1B). GSCs not only self-renew as confirmed by single-cell

(clonal) sphere-formation assays (Reynolds and Rietze, 2005), they also undergo multipotent differentiation (Figures S1C and S1D). However, unlike NSCs, GSCs are tumorigenic. Tumor mass could be observed 4–6 months after intracranial transplantation of GSCs in NOD-SCID mice (Figure S1E) (Bao et al., 2006b; Galli et al., 2004), confirming the tumorigenic potential of GSCs. To address the tumorigenic properties of GSCs, we compared the miRNA expression profiles of GSCs with NSCs. Using ANOVA on normalized ChIP data, we identified differentially expressed miRNAs. The overall miRNA expression pattern in NSCs is very similar to that of GSCs, as revealed by scatterplot of miRNA expression profile. Among the few differentially expressed transcripts, miR-138 emerged as the most significant differentially expressed miRNA (Figure 1A). Expression of miR-138 was validated by quantitative stem-loop realtime reverse-transcription PCR (Figure 1B). These results were confirmed by northern blot analysis using specific LNA probes that recognize miR-138 (Figure 1C). Because miR-138 has two putative precursors, pre-miR-138-1 and pre-miR-138-2, and the synthesis of mature miR-138 is regulated by stringent transcriptional and posttranscriptional control mechanisms (Obernosterer et al., 2006), we analyzed the expression of precursor transcripts in GSCs and NSCs. The detection of elevated levels of pre-miR138-2 only in GSCs, not in NSCs, suggests that miR-138 is under tight transcriptional control in NSCs (Figure 1D). Next, we analyzed the expression of miR-138 when GSCs were subjected

592 Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors

Figure 2. Targeting miR-138 in GSCs Impedes Formation of Tumorspheres (A) Representative images of tumorspheres from GSCs transduced with scramble control (top panel) and antimiR-138 (bottom panel). Note a significant reduction in tumorsphere formation by day (D) 9 in GSCs transduced with antimiR-138 relative to the scramble control. (B) The histogram represents average diameter of tumorspheres at indicated time points in GSCs transduced with antimiR-138 or scramble control. The error bars are the relative mean ± SD of 30 spheres. (C) The line graph indicates that the number of viable cells increased with time in GSCs transduced with scramble control, but not in GSCs transduced with antimiR-138. (D) Formation of tumorspheres is hampered by the loss of miR-138. Of the wells that were confirmed to contain one viable cell, 26% ± 1.5% (n = 3 independent cultures) formed tumorspheres in scramble control, whereas targeting miR-138 in GSCs prevented tumorsphere formation. A clone-derived tumorsphere (representative image below) can be observed only in scramble control, not in GSCs transduced with antimiR-138. (E) Expression levels of miR-138, miR-106a, miR-21, miR-886, and miR-1274a in GSCs transduced with scramble control or antimiR-138. (F) Representative neurosphere images of NSCs transduced with scramble control (top panel) and antimiR-138 (bottom panel). Scale bars, 100 mm. **p < 0.001. See also Figure S2.

to differentiation. Clearly, transcript levels of miR-138 were significantly downregulated when GSCs differentiated (Figure 1E). Together, these results indicate that GSCs consistently and reliably express elevated levels of miR-138 relative to NSCs. Functional Characterization of miR-138 To understand the functional significance of elevated expression levels of miR-138 in GSCs, we used lentiviral-based antagomirs to obtain stable loss-of-function phenotype (Kru¨tzfeldt et al., 2005; Scherr et al., 2007). To control for potential off-target effects, a nontargeting-scrambled (scramble) control was used. Lentivirus-transduced cells were subjected to puromycin selection. Sequence-specific inhibition of miRNA-138 by lentivirus encoded antagomiR-138 (antimiR-138), affected growth and survival of GSCs (Figure 2A), and this is apparent by a decrease in size of tumorspheres compared to the scramble control (Figure 2B). To confirm that the effect observed was not due to differences in lentivirus transduction or integration

efficiency, transduction efficiency was assayed by FACS and qPCR quantification of LTRGAG and GFP (vector-derived sequences) (Figures S2A and S2B). Cell titer-growth curve assay demonstrated that the total number of viable scramble control GSCs increased by
10-fold over 7 days, whereas GSCs depleted of miR-138 decreased in number, indicating that miR-138 preferentially contributes to sustained growth of GSCs (Figure 2C). Clonogenic assays on GSCs transduced either with antimiR-138 or scramble control confirmed the same, where in 26% ± 1.5% (n = 3 independent culture preparations) of wells formed tumorspheres in scramble control, whereas targeting miR-138 in GSCs prevented tumorsphere formation (Figure 2D). Supporting this observation, when single-cell suspension of GSCs transduced with antimiR-138 or scramble control was plated in soft agar, macroscopically visible colonies were observed only in GSCs transduced with scramble control, but not with antimiR-138 (Figure S2C). Collectively, these results suggest that miR-138 is essential for the growth and

Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors 593

Figure 3. miR-138 Is a Prosurvival OncomiR for GSCs (A) Cell-cycle analysis indicates a decrease in percentage of cells in S phase, coupled with an increase in the percentage of cells in the sub-G1 phase in antimiR138-transduced GSCs. (B) Immunoblot analysis for phosphoprotein histone 3 and cleaved PARP. (C) Caspase-3/7 activity relative to the cell number increased in antimiR-138 transduced GSCs relative to the scramble control. (D) Annexin V staining demonstrates increased apoptosis in GSCs transduced with antimiR-138 compared to the scramble control. (E) TUNEL staining exhibits elevated apoptosis in antimiR-138-transduced GSCs relative to the scramble control. Scale bars, 50 mm. **p < 0.001.

survival of bona fide stem cells with self-renewal potential. A significant decrease in the level of only miR-138, and no change in the levels of miR-886, miR-106, miR-21, and miR-1274a in GSCs transduced with antimiR-138 relative to the scramble control, confirms that antimiR-138 is highly transcript specific (Figure 2E). Neurosphere-formation capacity of NSCs was not affected when transduced with either antimiR-138 or scramble control, indicating cell specificity of antimiR-138 and highlights that antimiR-138 has no toxic effect on normal cells (Figure 2F).

Targeting miR-138 Inhibits Proliferation and Promotes Apoptosis of GSCs To determine the specific role of miR-138 in regulating proliferation and growth, a dual-color flow cytometric analysis with DNA content determination was performed on GSCs transduced with scramble control or antimiR-138. A reduction in cellular entry into S phase (Figure 3A), coupled with reduced expression of phosphohistone H3 (Figure 3B), a proliferation marker of cells in late G2 and M phase, in GSCs transduced with antimiR-138 but not

594 Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors

Figure 4. miR-138 Is an Essential Prosurvival OncomiR for GSCs (A) Representative real-time images of xenografts tracked by bioluminescence at indicated time points after engraftment. Bioluminescence detected on days 2 and 6 in antimiR-138-bearing GSCs and scramble control-bearing GSCs, indicating survival of transplanted cells (n = 10). (B) Representative light micrograph showing H&E for GSC-derived tumor showing characteristic intracranial tumor location in immunocompromised mice. Tumor mass detected only in mice bearing nontargeting scramble-expressing GSCs. (C) Survival plot indicates development of intracranial tumors and neurological symptoms only in mice implanted with GSCs expressing scrambled control. (D) ISH on xenograft tumor sections or control normal tissue section probed with miR-138 probe or scramble probe. Scale bar, 100 mm. (E) ISH on xenograft tumor sections to detect miR138 expression along with coimmunostaining for CD133 (right panel). Higher magnification of the same (left panel). Note that all CD133-positive GSCs do not express miR-138 (white arrowhead); only a subpopulation of CD133-positive GSCs expresses miR-138 (red arrowhead). The box represents the region magnified in the next panel. Scale bar, 20 mm.

when compared to the scramble control, suggesting increased apoptosis (Figure 3D). Finally, increased terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining that detects DNA fragments in late phase of apoptosis, only in antimiR-138 targeted GSCs, not with scramble control (Figure 3E), suggests that targeting miR-138 results in increased apoptotic death of GSCs, highlighting that miR-138 is a prosurvival oncomiR for GSCs.

scramble control indicates attenuation of proliferation. A robust increase in sub-G1 population (Figure 3A), accompanied by detection of 89 kDa poly-ADP-ribose polymerase (PARP) cleavage fraction in GSCs transduced with antimiR-138, but not with scramble control, suggests that knockdown of miR138 leads to apoptosis (Figure 3B). Functional inhibition of miR-138 prevents tumorsphere formation partly due to attenuation of proliferation, coupled with apoptotic cell death of GSCs. Supporting this result, Caspase-3/7 activity normalized to cell number increased in miR-138-targeted GSCs compared to the scramble control (Figure 3C). Further targeting miR-138 in GSCs increased the percentage of annexin V-positive cells

miR-138 Is an Essential Prosurvival OncomiR for GSCs To evaluate the role of miR-138 expression in GSCs and tumorigenicity, we intracranially implanted luciferase-expressing GSCs bearing antimiR-138 or nontargeting scramble into the forebrains of immunocompromised mice (n = 10). The viability of the engrafted GSCs was measured by tracking the bioluminescence after implantation (Figure 4A). The bioluminescence measured on days 2 and 6 after implantation indicates that the antimiR-138-expressing GSCs survive transplantation like the scramble control. However, tumorigenesis only in mice bearing GSCs expressing scramble control not in mice bearing GSCs expressing antimiR-138 suggests that functional inhibition of miR-138 leads to apoptotic death of GSCs and impedes tumorigenesis (Figures 4B and 4C). The xenograft tumor sections, but not the control normal brain sections, exhibited

Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors 595

Figure 5. miR-138 Impedes Apoptosis and Promotes Proliferation (A) Expression values of selected genes from microarray data represented as a heatmap, displayed in shades of red or blue relative to the individual mean value of the gene in a linear scale. (B) Scatterplot represents the fold changes (Scramble/AntimiR-138) of selected genes obtained from microarray and quantitative stem-loop real-time reversetranscription PCR (qRT-PCR). A trend line was plotted by linear regression using the least-squares method and R2 = 0.96. (C) Immunoblot analysis of selected genes in GSCs transduced with antimiR-138 or scramble control. (D) Validation of direct targets of miR-138 by reporter assays. Cells cotransfected with indicated 30 UTR luciferase reporter constructs either with miR-138 or control vector. Normalized relative luciferase activity is shown as a bar diagram. Luciferase activity is expressed as mean relative to controls ±SD (**p < 0.001). See also Figure S3.

positive staining for miR-138 as observed by in situ hybridization (ISH) (Figure 4D). GSCs expressing miR-138 also express CD133 (Figure 4E); however, not all CD133-positive cells express miR138, suggesting that a subpopulation of CD133-positive cells is tumor-initiating GSCs. Taken together, miR-138 activity in GSCs relates primarily to the maintenance and survival of GSCs, and miR-138-positive GSCs are tumorigenic. miR-138 Targets Tumor Suppressors and Proapoptotic Genes Furthermore, to delineate the underlying mechanism of miR-138 as a prosurvival oncomiR, we sought to identify the targets by comparing gene expression profiles from GSCs transduced with either antimiR-138 or scramble control. Representation of selected array data as a heatmap displayed multiple differentially

expressed genes (Figure 5A). A linear correlation between array and quantitative stem-loop real-time reverse-transcription PCR indicates that the array profiles can be used to ascertain global and specific properties of miR-138 (Figure 5B). Several of the candidate gene products were analyzed by immunoblotting, and expected changes in protein levels were observed. Increased expression of specific genes on functional inhibition of miR-138 suggests that these genes are potential targets of miR-138 (Figure 5C). Interrogation of TargetScan v5.1 database and correlation of the predicted targets with microarray data revealed potential direct targets of miR-138. These include Caspase-3 (CASP3), Bladder cap-associated protein (BLCAP), and MAX dimerization protein 1 (MXD1), each with at least one conserved binding site for miR-138. To determine the ability of miR-138 to

596 Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors

Figure 6. miR-138 Promotes Gliomagenesis (A) Model depicting miR-138 network. As a negative regulator, miR-138 targets proapoptotic genes and blocks apoptosis. As a positive regulator, miR-138 targets tumor suppressors, transcriptional repressors, and indirectly facilitates expression genes that promote survival and proliferation of GSCs. (B) Log-log plot of the normalized expression levels of miR-138 versus miR-21 from samples derived from patients with secondary and primary GBM. Expression levels of the miRNAs were normalized against normal non-neoplastic brain tissue. Data were obtained from GEO database (accession number GSE13030).

negatively regulate CASP3, MXD1, and BLCAP expression, chimeric 30 UTR luciferase reporter assays were performed. Expression of miR-138 resulted in significant downregulation of reporter activity with CASP3, MXD1, or BLCAP wild-type 30 UTR, whereas the reporter vector with a mutation in miR-138 binding site was not affected, confirming that CASP3, MXD1, and BLCAP are direct targets of miR-138 (Figure 5D). Repressing CASP3, a proapoptotic gene, and BLCAP, a tumor suppressor, miR-138 prevents apoptosis and S phase arrest. Targeting MXD1 (MAD), a transcriptional repressor that competes with MYC for binding to MAX (Amati et al., 1993; Ayer et al., 1993), the oncomiR-138 facilitates formation of a stable MYCMAX complex and may indirectly enhance oncogenic activity of MYC. Not surprisingly, targeting tumor suppressors, inhibitors of proliferation, and transcriptional repressors, miR-138 indirectly activates several genes involved in tumorigenesis. This is confirmed by the downregulation of Cyclin D1 (CCND1), Cyclin A2 (CCNA2), Aurora kinase A (AURKA), and proto-oncogene c-Myc (cMYC) on miR-138 inhibition (Figure 5A). Concomitant repression of these genes using gene-specific shRNAs results in a decrease in S phase population coupled with an increase in sub-G1 population, suggesting that targeting these genes partially phenocopy the effect of antimiR-138 in GSCs (Figure S3A). In conclusion, targeting specific inhibitors of proliferation, miR-138 indirectly modulates the expression of several genes to promote proliferation, underscoring its role as an oncomiR. Finally, to examine whether the oncomiR-138 alone can confer a transformed phenotype, we transduced normal NSCs (Lonza) with lentivirus expressing miR-138. Ectopic expression of miR-138 in NSCs results in a moderate increase in size of neurospheres formed when compared with the scrambled control (Figure S3B). Time course gene expression analysis by quantitative stem-loop real-time reverse-transcription PCR further confirmed that overexpression of miR-138 in NSCs results in a significant downregulation of a few targets accompanied by a moderate increased expression of a few genes influenced indirectly by miR-138 (Figure S3C). These results highlight that miRNA expression is highly context specific, and ectopic expression of miR-138 alone may not transform normal NSCs to GSCs.

miR-138 Is a Potential Prognostic Biomarker Elevated expression of miR-138 in secondary GBMs compared to the primary GBMs (D’Urso et al., 2012) (GSE13030) (Figure 6B) and expression in GSCs prompted us to study the expression of miR-138 in tumor recurrence. Increased miR-138 expression in tumor sections from patients with recurrent GBM compared to patients with first diagnosis and control normal human brain sections suggests a role of miR-138 in tumor recurrence (Figures 7A–7C). To determine the clinical significance of miR-138 expression, we used data-driven grouping (DDg) method to classify patients into high-risk and low-risk groups, based on an optimal expression cutoff that provides the best separation of their survival curves. We focused on subgroups of patients with tumor progression or recurrence. In TCGA analysis of 197 GBM patients with tumor progression, we found a potentially oncogenic pattern for miR-138 (p = 0.129; Figure 7D). On the other hand, for the group of 80 patients with tumor recurrence, DDg method can provide significant separation of patients into two risk groups based on miR-138 expression (p < 0.05; Figure 7E). This suggests a potentially oncogenic role of miR-138 in tumor recurrence, which leads to higher risk of mortality. As a negative control, we investigated miR-21 to stratify patients into high- and low-risk groups based on tumor recurrence using Mann-Whitney U test. GBMs display high levels of miR-21 (Chan et al., 2005; Corsten et al., 2007; Papagiannakopoulos et al., 2008). Results show that miR-21 is not differentially expressed in these groups (p = 0.736), suggesting that miR-21’s expression is unlikely to contribute to grouping of patients according to their risk. Together, this highlights the clinical significance of miR-138 as a potential prognostic biomarker and its association with tumor recurrence and survival. DISCUSSION The discovery of GSCs, highly tumorigenic, self-renewing subpopulation of GBM cells, suggests that therapeutic approaches to effectively eradicate these cells may improve patient outcome. We report the identification of oncomiR-138 as a molecular signature of GSCs. Targeting miR-138 expression with antimiRs significantly impaired the growth and survival of GSCs.

Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors 597

Figure 7. miR-138 Is a Potential Prognostic Biomarker (A) ISH with miR-138 probe or scramble probe on patient-derived GBM tumor sections and control non-neoplastic normal human brain sections (n = 12). Note increased staining with miR-138 probe in patients with recurrent GBM compared to patients with first-diagnosis GBM. The normal brain tissue exhibits little or no staining. Scale bar, 100 mm. (B) Quantification of miR-138 expression in patients with first diagnosis compared to patients with tumor recurrence normalized to nuclear DAPI staining (*p < 0.05). (C) Quantification of miR-138 transcript abundance by quantitative stem-loop real-time reverse-transcription PCR in patients with first diagnosis compared to patients with tumor recurrence (*p < 0.05). (D) Kaplan-Meier survival curves representing the classification of 197 patients with tumor progression by DDg method (p = 0.129). (E) Kaplan-Meier survival curves representing 80 patients with tumor recurrence (p < 0.05). Black and red curves denote patients with low and high miR-138 expression, respectively.

The aberrant expression of an oncomiR can repress the expression of its target tumor suppressor genes resulting in oncogenesis. GSCs express elevated levels of miR-138 not miR-21, and knockdown of miR-138 has a deleterious effect on GSCs. Apparently, for GSCs, miR-138 is a prosurvival oncomiR, and miR-138-mediated altered protein synthesis may promote cellular transformation. MYC, indirectly activated by miR-138, is known to trigger an antiapoptotic signaling cascade in GSCs (Wang et al., 2008; Zheng et al., 2008). The oncogenic activity of MYC requires dimerization with MAX. Targeting MXD1 (MAD) that competes with MYC for binding to MAX, the oncomiR-138 facilitates formation of a stable MYC-MAX complex and indirectly enhances oncogenic activity of MYC. We therefore hypothesize that targeting tumor suppressors like MAD, miR-138 facilitates proliferation of GSCs. A role for miR-138 in evasion of apoptosis is strengthened by validating the proapoptotic gene, CASP3, as its direct target. Microarray profiling of antimiR-138-treated cells showed a wide range of alterations including many predicted target genes. This includes tumor suppressor candidate 2 (TUSC2), which induces apoptosis by initiating G1 cell-cycle arrest (Kondo et al., 2001), and B cell translocation gene 2 (BTG2), an antiproliferative protein

(Rouault et al., 1996; Yao et al., 2007). In addition the master regulator miR-138 may modulate the expression of downstream effectors of the p53-driven signaling pathway that includes BTG2, and growth arrest and DNA damageinducible a (GADD45a). Furthermore, directly targeting and downregulating BLCAP, a stimulator of apoptosis and an inhibitor of the antiapoptotic oncogene B cell CLL/lymphoma 2 (BCL2), miR-138 may impede apoptosis by a p53 and NFkB-independent pathway (Fan et al., 2011; Yao et al., 2007). The oncomiR-138-mediated inhibition of proapoptotic genes and multiple tumor suppressor pathways simultaneously may facilitate GSCs to overcome programmed apoptotic cell death (Figure 6A). miRNA-138 may modulate cell proliferation pathways through indirect interaction with critical cell-cycle regulators. Targeting inhibitors of proliferation like Retinoic acid receptor-a (RARA) that blocks CCND1 (Ball et al., 1997; Wang et al., 2010) and metastasis suppressors like Thioredoxin-interacting protein (TXNIP), a transcriptional repressor of Cyclin A2 (Han et al., 2003), miR-138 promotes cell proliferation. The oncomiR-138 indirectly enhances aurora kinase activity, a regulator of Cyclin B1, by targeting GADD45a (Shao et al., 2006). This leads to the hypothesis that by blocking antiproliferative signals, miR-138 may indirectly control cell proliferation and alter cellular homeostasis in GSCs. As potent regulators of transcript stability, miRNAs directly regulate the cell’s proteome and in this process determine its phenotype and plasticity. miR-138 targets and downregulates negative regulators of gliomagenesis, such as Ephrin-A1

598 Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors

(EFNA1), that exerts an antioncogenic effect in tumor cells (Liu et al., 2007), and Pannexin2 (PANX2), a brain-specific gapjunction protein with tumor suppressor function. Ectopic expression of gap-junction proteins in tumor cells is known to reverse the neoplastic phenotype (Lai et al., 2009). Controlling transcript stability of these targets, miR-138 potentially facilitates gliomagenesis. GSCs support nonstem tumor cells through a paracrine effect by inducing neovascularization and play a critical role in tumor growth (Bao et al., 2006b). The oncomiR-138 may promote tumor vascularization by targeting Collagen type IV a 1 (COL4A1), an efficient inhibitor of tumor angiogenesis and neovascularization (Colorado et al., 2000; Nyberg et al., 2008). It is evident by the loss-of-function phenotype that a strikingly large number of genes are regulated by miR-138 either directly or indirectly, and the prosurvival oncomiR potentially plays a vital role in tumor invasion by coordinating some of the intricate gene expression programs. The pattern of a miRNA expression can reflect tumor origin, stage, and other pathological variables (Lu et al., 2005), and miRNA can serve as a biomarker. Specific miRNA signatures can contribute to the phenotypic diversity of glioblastoma subclasses and predict survival in glioblastoma (Kim et al., 2011; Malzkorn et al., 2010). miR-138 was found to be differentially expressed in tumors from patients with good prognosis versus patients with poor prognosis. High-expression miRNA138 was associated with poor prognosis for survival, highlighting the role of the oncomiR as a potential prognostic biomarker for malignant gliomas. EXPERIMENTAL PROCEDURES Cell Culture GSCs from samples from five patients with malignant glioma (NNI-1, NNI-4, NNI-8, NNI-11, and NNI-12; obtained after institutional review board-approved protocol) were cultured as described by Chong et al. (2009), Foong et al. (2011), and Ng et al. (2012). Cells were cultured in chemically defined serum-free selection growth medium with 20 ng/ml basic fibroblast growth factor (bFGF), 20 ng/ml epidermal growth factor (EGF), 20 ng/ml human recombinant leukemia inhibitory factor (LIF), 5 mg/ml heparin, serum-free supplement (B27) in a 3:1 mix of Dulbecco’s modified Eagle’s medium (DMEM) and Ham’s F-12 Nutrient Mixture (F12). Cultures were incubated at 37 C and 5% CO2, with growth factor replenishment. Neurosphere cultures were expanded by mechanical trituration. Normal human NSCs/ progenitor cells were obtained from Lonza. In addition, NSCs were also obtained from subventricular zone (SVZ) and the frontal lobe (FL) of the cortex of human brain (Kobayashi et al., 2010). NSCs were cultured as freefloating neurospheres in a serum-free DMEM/F12 medium (Invitrogen) supplemented with bFGF (20 ng/ml; PeproTech), endothelial growth factor (EGF, 10 ng/ml; PeproTech), 2 mg/ml heparin (Sigma-Aldrich), and 2% B27 (Invitrogen). Immunocytochemistry on Differentiated GSCs GSCs (NNI-1, NNI-4, and NNI-8 [passage #36–40]) on coverslips coated with Laminin and poly-L-ornithine (Sigma-Aldrich) were grown in differentiation medium. Cells were fixed in 4% PFA, and incubated overnight with the following primary antibodies: anti-GFAP (Sigma-Aldrich), anti-b-Tubulin (Covance), anti-O4 (Millipore). After washing, cells were incubated with secondary antibodies: anti-mouse Alexa 568, anti-rabbit Alexa 488, or antimouse Alexa 488 IgM from Invitrogen at room temperature. Nuclei were counterstained with DAPI. Image acquisition was done with Olympus FV1000 confocal microscope.

Tumorsphere-Formation Assay Harvested spheres (NNI-8, passage #36) were dissociated and plated as single cells in 96-well plates by serial dilution or FACS. The number of wells with tumorspheres was scored after 21 days. miRNA Profiling Total RNA from GSCs (NNI-8, passage 34) and NSCs (Lonza, passage 3) was subjected to miRCURY LNA Array microRNA profiling (Exiqon). After background subtraction and normalization, differential gene expression analysis was performed. A multiple ANOVA model was used to remove batch effects from replicate samples. Microarray Analysis for Determination of Gene Expression Profile Total RNA (500 ng) from three replicates of GSCs (NNI-8, passage 34) transduced with antimiR-138 or scrambled control was converted to biotinylated cRNA using TargetAmp Nano-g Biotin-aRNA labeling kit (Epicenter) and isolated using QIAGEN columns. cRNA was hybridized on HumanWG-6 v3.0 array (Illumina). Normalized data were analyzed with Illumina BeadStudio, and analysis was performed on PARTEK platform. Quantitative RT-PCR RNA was isolated from GSC lines (NNI-1, NNI-4, NNI-8, NNI-11, and NNI-12 [passage #35–40]) and from NSC lines (fetal whole-brain-derived Lonza, SVZ derived, and FL derived [passage #3–6]). cDNA was synthesized using miRNA RT assay (TaqMan), and expression levels of miRNA were analyzed using the 7900 fast RT-PCR system (Applied biosystems) using miR-138-specific primers. U6 probe was used as an endogenous control. DDCt was used to determine the transcript abundance. miRNA Northern Blot Analysis Northern blot analysis was carried out using 15 mg of small RNA-enriched total RNA from GSCs (NNI-8, passage #36). RNA was separated on 15% denaturating Urea-PAGE, transferred to nylon membrane (Ambion), and UV crosslinked (Stratagene). LNA probes for miR-138 and U6 (Exiqon) were end labeled with T4 polynucleotide kinase. Following hybridization and washing the blot was processed. Preparation of Lentiviral Stocks and Transduction Stable expression of antimiRs was carried out using miRZip, a lentiviral expression vector (System Biosciences). Mature functional antimiR-138 sequence is CGGCCTGATTCACAACACCAGCT. The H1 expression cassette provides constitutive RNA polymerase III-dependent transcription of antimiR transcripts. CMV promoter supports expression of copGFP (fluorescent reporter) and puromycin-N-acetyl transferase (drug-selectable marker) for detection and selection of transduced cells, respectively. Lentivirus expressing luciferase under human PGK promoter (Campeau et al., 2009) was obtained from Addgene. Third-generation lentiviruses were produced in Lenti-X 293T (Clontech) with packaging mix consisting of three constructs, pMDLg/pRRE, pRSV-Rev, and pMD2.G, from Addgene. Supercoiled DNA constructs were prepared using QIAGEN maxi preps. Soft Agar Assay Single-cell suspension of 2 3 103 cells (NNI-1, NNI-4, and NNI-8; passage #35–42) was plated in medium containing 0.3% noble agar (Difco) seeded in 6-well plates containing 1% noble agar. Cells were cultured for 2– 3 weeks with growth factor supplementation. Lentiviral-transduced cells were subjected to puromycin selection. Cells were incubated with 1 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma-Aldrich), and colonies were counted using MATLAB software. Cell-Viability Assay Lentivirus-transduced GSCs (NNI-1, NNI-4, and NNI-8; passage #35–42) were plated at a density of 5,000 cells/well for growth curve analysis. Live cells were determined at the indicated number of days after plating using the Cell Titer-Glo Luminescent Cell Viability Assay kit (Promega).

Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors 599

Caspase-3/7 Assay Lentivirus-transduced GSCs (NNI-1, NNI-4, and NNI-8; passage #35–42) were plated at a density of 5,000 cells/well, and Caspase-3/7 activity was measured by Caspase-Glo 3/7 Assay (Promega). Relative units were normalized to the number of viable cells. Cell-Cycle Analysis Lentivirus-transduced GSCs (NNI-8, passage #38) were plated at a density of 100,000 cells/well, labeled with 10 mM EdU for 5 hr prior to harvesting, and processed using the Click-iT EdU Alexa Fluor 647 Flow Cytometry assay (Invitrogen), and cell nuclei were counterstained with 25 mg/ml propidium iodide (Sigma-Aldrich). Samples were subject to EdU incorporation analysis on a BD FACS caliber (Becton Dickinson). Data were analyzed using WINMDI 2.9 software. TUNEL Assay Lentiviral-transduced GSCs (NNI-8, passage #38) cultured for indicated time points were fixed, and paraffin-embedded sections were subjected to DeadEnd fluorometric TUNEL assay (Promega). Cell nuclei were counterstained with DAPI. Western Blotting Cell lysate (NNI-8; passage #38) was resolved by SDS-PAGE and transferred to PVDF membranes (Millipore). For immunoblots, primary antibodies used were anti-PARP, anti-LASP1, anti-BTG2, anti-OLIG1 from Abcam or antiGADD45a, anti-EFNA1, anti-CCND1, anti-PANX2, anti-TXNIP from Santa Cruz Biotechnology or anti-AURKA (Cell Signaling), and anti-phosphoHistone3 (Millipore). Secondary antibodies used were anti-rabbit HRP (Santa Cruz Biotechnology) and anti-mouse HRP (Jackson Laboratory). All blots were stripped and reprobed with b-actin antibodies as loading control. Luciferase Assay Wild-type 30 UTR reporter constructs of BLCAP, CASP3, and MXD1 constructs (GeneCopoeia) were cotransfected with pCDH-miR-138 or negative control vector into HEK293T cells using Effectene (QIAGEN). Firefly and Renilla luciferase activities were measured 24 hr post-transfection using Dual-Luciferase Reporter System (Promega). The firefly luminescence was normalized to Renilla luminescence as an internal control for transfection efficiency. For mutated constructs, miR-138 binding site was mutated using the QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene). miR-138 binding site accagc was substituted with cttgat, gaagta, and gcagaa in BLCAP, MXD1, and CASP3, respectively. miR-138 ISH Five micron sections were processed and boiled in pretreatment solution (Panomics), washed in PBS, followed by protease (Panomics) treatment 37 C. Sections were incubated with LNA probes (50 DIG-labeled LNA probes specific for miR-138 or scrambled probe with no homology to known vertebrate miRNAs [Exiqon]) in hybridization buffer (Roche) at 51 C for 4 hr. Sections were blocked with 10% goat serum and incubated with anti-DIG alkaline phosphatase (Roche). miRNA bound LNA probes were detected by Fast Red Substrate (Panomics). After counterstaining with DAPI, slides were mounted using FluorSave (Merck). Image acquisition was performed with Olympus FluoView FV1000 using TRITC filter.

GSC Luciferase Preparation and Bioluminescence Imaging GSC luciferase stables (NNI-8; passage #40–45) were established by selecting for puromycin (1 mg/ml) after transducing with lentivirus expressing luciferase under human PGK promoter (Addgene). For bioluminescence imaging, GSCs bearing scramble control or antimiR-138 and expressing firefly luciferase were injected into the right forebrains of NOD-SCID mice, and Xenogen system was used for imaging. After an intraperitoneal dose of 150 mg/kg of D-luciferin, mice were anesthetized, and imaging was performed using IVIS Spectrum Imaging System (Xenogen). Quantification was based on total flux (photons/ s) of emitted light as a measure of the relative number of viable cells. Bioluminescence signals were analyzed using Living Image software (IVIS Living Image v3.0). Data Preprocessing of miRNA Expression Profiling miRNA microarray expression data from 494 samples were downloaded from TCGA GBM database. Outlier removal and batch effect correction procedures were performed. Quality assessment to identify poor-quality chips was done by utilizing several visualization methods and statistical indicators on Agilent Total Gene Signal. The statistical indicators were the median of log2 intensity, log intensity ratio M, relative log expression (RLE), and correlation among samples. Samples that fail in more than two indicators were identified as outliers and subsequently removed. Invariant set normalization (ISN) was performed to normalize the data. Nonparametric Combat software (Johnson et al., 2007) that applies an empirical Bayes approach was utilized to correct batch effect. A total of 57 miRNA data sets as outliers were removed from analysis. Survival analysis was performed using Cox proportional hazard regression model and DDg method. Clinical information for 596 patients, each containing data such as time to last follow-up/progression/recurrence, vital status, and types of treatments received, is downloaded. Data-driven survival analysis is a published method of assigning patients to distinct groups according to their risk for death (Motakis et al., 2009). For each gene of interest, the method estimates the expression cutoff of the gene expression by maximizing the separation of the survival curves associated with the low- and high-risk patients. Statistical Analysis Values are reported as the mean ± SE. Statistical significance between two samples was determined with two-tailed Student’s t test using GraphPad. Human Brain Tissue Human brain tumor and control tissue specimens were obtained from the Medical University of Graz. Biopsy specimens were registered in the respective biobank and kept anonymous. The research project was authorized by the ethical commission of the Medical University Graz (EK-number 21-520 ex 09/10). The study protocol was in accordance with the ethical guidelines of the Helsinki declaration and the Austrian law. ACCESSION NUMBERS Microarray data have been deposited in the GEO database under the accession number GSE37366. SUPPLEMENTAL INFORMATION Supplemental Information includes three figures and can be found with this article online at http://dx.doi.org/10.1016/j.celrep.2012.07.012. LICENSING INFORMATION

Intracranial Implantation of GSCs Intracranial transplantation of GSCs (NNI-8, passage # 40–45) into 8-week-old NOD/SCID/IL2rg mice (Jackson Laboratory) was performed in accordance with the Institutional Animal Care and Use Committee-approved protocol. Luciferase-expressing GSCs bearing either antimiR-138 or scramble control were orthotopically transplanted following washing and resuspension in PBS. A total of 15 3 104 cells in 3 ml was injected stereotactically into the forebrain of immunodeficient mice and maintained till the development of neurological symptoms. Brain collected from the euthanized mice was fixed in 4% PFA, embedded and sectioned, and subjected to H&E.

This is an open-access article distributed under the terms of the Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 Unported License (CC-BY-NC-ND; http://creativecommons.org/licenses/by-nc-nd/3.0/ legalcode). ACKNOWLEDGMENTS We thank Drs. Carol Tang and Beng Ti Ang from National Neuroscience Institute, Singapore, for the generous gift of primary GSCs and Dr. Jeremy Crook

600 Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors

and Nao Kobayashi for the gift of normal human NSCs. We thank Professor Davor Solter and Mr. Sathish Balu for assistance with in vivo experiments. We thank Dr. Benedict Yan from National University Hospital, Singapore, for histology assistance. We acknowledge the TCGA database for access to GBM data. We thank Dr. Kwoh Chee Keong for useful discussion and Dr. Tang Zhiqun for assistance in data preprocessing and batch effect correction for the TCGA data and Ian Jun Jie Tay and Qian Yi Lee for help with microarray analysis. This work was supported by IMI OncoTrack to J.H. and Agency for Science Technology and Research Investigatorship award to P.S. Received: July 26, 2011 Revised: February 2, 2012 Accepted: July 27, 2012 Published online: August 23, 2012

Foong, C.S., Ng, F.S., Phong, M., Toh, T.B., Chong, Y.K., Tucker-Kellogg, G., Campbell, R.M., Ang, B.T., and Tang, C. (2011). Cryopreservation of cancerinitiating cells derived from glioblastoma. Front Biosci (Schol Ed) 3, 698–708. Gal, H., Pandi, G., Kanner, A.A., Ram, Z., Lithwick-Yanai, G., Amariglio, N., Rechavi, G., and Givol, D. (2008). MIR-451 and Imatinib mesylate inhibit tumor growth of Glioblastoma stem cells. Biochem. Biophys. Res. Commun. 376, 86–90. Galli, R., Binda, E., Orfanelli, U., Cipelletti, B., Gritti, A., De Vitis, S., Fiocco, R., Foroni, C., Dimeco, F., and Vescovi, A. (2004). Isolation and characterization of tumorigenic, stem-like neural precursors from human glioblastoma. Cancer Res. 64, 7011–7021. Garden, A.S., Maor, M.H., Yung, W.K., Bruner, J.M., Woo, S.Y., Moser, R.P., and Lee, Y.Y. (1991). Outcome and patterns of failure following limited-volume irradiation for malignant astrocytomas. Radiother. Oncol. 20, 99–110.

REFERENCES Amati, B., Brooks, M.W., Levy, N., Littlewood, T.D., Evan, G.I., and Land, H. (1993). Oncogenic activity of the c-Myc protein requires dimerization with Max. Cell 72, 233–245. Ayer, D.E., Kretzner, L., and Eisenman, R.N. (1993). Mad: a heterodimeric partner for Max that antagonizes Myc transcriptional activity. Cell 72, 211–222.

Godlewski, J., Bronisz, A., Nowicki, M.O., Chiocca, E.A., and Lawler, S. (2010). microRNA-451: a conditional switch controlling glioma cell proliferation and migration. Cell Cycle 9, 2742–2748. Guessous, F., Zhang, Y., Kofman, A., Catania, A., Li, Y., Schiff, D., Purow, B., and Abounader, R. (2010). microRNA-34a is tumor suppressive in brain tumors and glioma stem cells. Cell Cycle 9, 1031–1036.

Ball, K.L., Lain, S., Faˆhraeus, R., Smythe, C., and Lane, D.P. (1997). Cell-cycle arrest and inhibition of Cdk4 activity by small peptides based on the carboxyterminal domain of p21WAF1. Curr. Biol. 7, 71–80.

Han, S.H., Jeon, J.H., Ju, H.R., Jung, U., Kim, K.Y., Yoo, H.S., Lee, Y.H., Song, K.S., Hwang, H.M., Na, Y.S., et al. (2003). VDUP1 upregulated by TGF-beta1 and 1,25-dihydorxyvitamin D3 inhibits tumor cell growth by blocking cell-cycle progression. Oncogene 22, 4035–4046.

Bao, S., Wu, Q., McLendon, R.E., Hao, Y., Shi, Q., Hjelmeland, A.B., Dewhirst, M.W., Bigner, D.D., and Rich, J.N. (2006a). Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature 444, 756–760.

Hemmati, H.D., Nakano, I., Lazareff, J.A., Masterman-Smith, M., Geschwind, D.H., Bronner-Fraser, M., and Kornblum, H.I. (2003). Cancerous stem cells can arise from pediatric brain tumors. Proc. Natl. Acad. Sci. USA 100, 15178– 15183.

Bao, S., Wu, Q., Sathornsumetee, S., Hao, Y., Li, Z., Hjelmeland, A.B., Shi, Q., McLendon, R.E., Bigner, D.D., and Rich, J.N. (2006b). Stem cell-like glioma cells promote tumor angiogenesis through vascular endothelial growth factor. Cancer Res. 66, 7843–7848.

Johnson, W.E., Li, C., and Rabinovic, A. (2007). Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8, 118–127.

Bottoni, A., Piccin, D., Tagliati, F., Luchin, A., Zatelli, M.C., and degli Uberti, E.C. (2005). miR-15a and miR-16-1 down-regulation in pituitary adenomas. J. Cell. Physiol. 204, 280–285. Calin, G.A., and Croce, C.M. (2006). MicroRNA signatures in human cancers. Nat. Rev. Cancer 6, 857–866. Campeau, E., Ruhl, V.E., Rodier, F., Smith, C.L., Rahmberg, B.L., Fuss, J.O., Campisi, J., Yaswen, P., Cooper, P.K., and Kaufman, P.D. (2009). A versatile viral system for expression and depletion of proteins in mammalian cells. PLoS One 4, e6529.

Kim, T.M., Huang, W., Park, R., Park, P.J., and Johnson, M.D. (2011). A developmental taxonomy of glioblastoma defined and maintained by MicroRNAs. Cancer Res. 71, 3387–3399. Kobayashi, N.R., Sui, L., Tan, P.S., Lim, E.K., Chan, J., Choolani, M., and Crook, J.M. (2010). Modelling disrupted-in-schizophrenia 1 loss of function in human neural progenitor cells: tools for molecular studies of human neurodevelopment and neuropsychiatric disorders. Mol. Psychiatry 15, 672–675.

Chan, J.A., Krichevsky, A.M., and Kosik, K.S. (2005). MicroRNA-21 is an antiapoptotic factor in human glioblastoma cells. Cancer Res. 65, 6029–6033.

Kondo, M., Ji, L., Kamibayashi, C., Tomizawa, Y., Randle, D., Sekido, Y., Yokota, J., Kashuba, V., Zabarovsky, E., Kuzmin, I., et al. (2001). Overexpression of candidate tumor suppressor gene FUS1 isolated from the 3p21.3 homozygous deletion region leads to G1 arrest and growth inhibition of lung cancer cells. Oncogene 20, 6258–6262.

Chong, Y.K., Toh, T.B., Zaiden, N., Poonepalli, A., Leong, S.H., Ong, C.E., Yu, Y., Tan, P.B., See, S.J., Ng, W.H., et al. (2009). Cryopreservation of neurospheres derived from human glioblastoma multiforme. Stem Cells 27, 29–39.

Kru¨tzfeldt, J., Rajewsky, N., Braich, R., Rajeev, K.G., Tuschl, T., Manoharan, M., and Stoffel, M. (2005). Silencing of microRNAs in vivo with ‘antagomirs’. Nature 438, 685–689.

Colorado, P.C., Torre, A., Kamphaus, G., Maeshima, Y., Hopfer, H., Takahashi, K., Volk, R., Zamborsky, E.D., Herman, S., Sarkar, P.K., et al. (2000). Antiangiogenic cues from vascular basement membrane collagen. Cancer Res. 60, 2520–2526.

Lai, C.P., Bechberger, J.F., and Naus, C.C. (2009). Pannexin2 as a novel growth regulator in C6 glioma cells. Oncogene 28, 4402–4408.

Corsten, M.F., Miranda, R., Kasmieh, R., Krichevsky, A.M., Weissleder, R., and Shah, K. (2007). MicroRNA-21 knockdown disrupts glioma growth in vivo and displays synergistic cytotoxicity with neural precursor cell delivered S-TRAIL in human gliomas. Cancer Res. 67, 8994–9000. D’Urso, P.I., D’Urso, O.F., Storelli, C., Mallardo, M., Gianfreda, C.D., Montinaro, A., Cimmino, A., Pietro, C., and Marsigliante, S. (2012). miR-155 is up-regulated in primary and secondary glioblastoma and promotes tumour growth by inhibiting GABA receptors. Int. J. Oncol. 41, 228–234.

Lavon, I., Zrihan, D., Granit, A., Einstein, O., Fainstein, N., Cohen, M.A., Cohen, M.A., Zelikovitch, B., Shoshan, Y., Spektor, S., et al. (2010). Gliomas display a microRNA expression profile reminiscent of neural precursor cells. Neurooncol. 12, 422–433. Legler, J.M., Ries, L.A., Smith, M.A., Warren, J.L., Heineman, E.F., Kaplan, R.S., and Linet, M.S. (1999). Cancer surveillance series [corrected]: brain and other central nervous system cancers: recent trends in incidence and mortality. J. Natl. Cancer Inst. 91, 1382–1390.

Esquela-Kerscher, A., and Slack, F.J. (2006). Oncomirs—microRNAs with a role in cancer. Nat. Rev. Cancer 6, 259–269.

Li, Y., Guessous, F., Zhang, Y., Dipierro, C., Kefas, B., Johnson, E., Marcinkiewicz, L., Jiang, J., Yang, Y., Schmittgen, T.D., et al. (2009). MicroRNA-34a inhibits glioblastoma growth by targeting multiple oncogenes. Cancer Res. 69, 7569–7576.

Fan, D.G., Zhao, F., Ding, Y., Wu, M.M., Fan, Q.Y., Shimizu, K., Dohjima, T., Nozawa, S., Wakahara, K., Ohno, T., et al. (2011). BLCAP induces apoptosis in human Ewing’s sarcoma cells. Exp. Biol. Med. (Maywood) 236, 1030–1035.

Liu, D.P., Wang, Y., Koeffler, H.P., and Xie, D. (2007). Ephrin-A1 is a negative regulator in glioma through down-regulation of EphA2 and FAK. Int. J. Oncol. 30, 865–871.

Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors 601

Lu, J., Getz, G., Miska, E.A., Alvarez-Saavedra, E., Lamb, J., Peck, D., SweetCordero, A., Ebert, B.L., Mak, R.H., Ferrando, A.A., et al. (2005). MicroRNA expression profiles classify human cancers. Nature 435, 834–838. Ma, L., Teruya-Feldstein, J., and Weinberg, R.A. (2007). Tumour invasion and metastasis initiated by microRNA-10b in breast cancer. Nature 449, 682–688. Malzkorn, B., Wolter, M., Liesenberg, F., Grzendowski, M., Stu¨hler, K., Meyer, H.E., and Reifenberger, G. (2010). Identification and functional characterization of microRNAs involved in the malignant progression of gliomas. Brain Pathol. 20, 539–550. Motakis, E., Ivshina, A.V., and Kuznetsov, V.A. (2009). Data-driven approach to predict survival of cancer patients: estimation of microarray genes’ prediction significance by Cox proportional hazard regression model. IEEE Eng. Med. Biol. Mag. 28, 58–66. Ng, F.S., Toh, T.B., Ting, E.H., Koh, G.R., Sandanaraj, E., Phong, M., Wong, S.S., Leong, S.H., Kon, O.L., Tucker-Kellogg, G., et al. (2012). Progenitorlike traits contribute to patient survival and prognosis in oligodendroglial tumors. Clin. Cancer Res. 18, 4122–4135. Nyberg, P., Xie, L., Sugimoto, H., Colorado, P., Sund, M., Holthaus, K., Sudhakar, A., Salo, T., and Kalluri, R. (2008). Characterization of the antiangiogenic properties of arresten, an alpha1beta1 integrin-dependent collagen-derived tumor suppressor. Exp. Cell Res. 314, 3292–3305. Obernosterer, G., Leuschner, P.J., Alenius, M., and Martinez, J. (2006). Posttranscriptional regulation of microRNA expression. RNA 12, 1161–1167. Papagiannakopoulos, T., Shapiro, A., and Kosik, K.S. (2008). MicroRNA-21 targets a network of key tumor-suppressive pathways in glioblastoma cells. Cancer Res. 68, 8164–8172. Reynolds, B.A., and Rietze, R.L. (2005). Neural stem cells and neurospheres— re-evaluating the relationship. Nat. Methods 2, 333–336. Rouault, J.P., Falette, N., Gue´henneux, F., Guillot, C., Rimokh, R., Wang, Q., Berthet, C., Moyret-Lalle, C., Savatier, P., Pain, B., et al. (1996). Identification of BTG2, an antiproliferative p53-dependent component of the DNA damage cellular response pathway. Nat. Genet. 14, 482–486. Scherr, M., Venturini, L., Battmer, K., Schaller-Schoenitz, M., Schaefer, D., Dallmann, I., Ganser, A., and Eder, M. (2007). Lentivirus-mediated antagomir

expression for specific inhibition of miRNA function. Nucleic Acids Res. 35, e149. Schraivogel, D., Weinmann, L., Beier, D., Tabatabai, G., Eichner, A., Zhu, J.Y., Anton, M., Sixt, M., Weller, M., Beier, C.P., and Meister, G. (2011). CAMTA1 is a novel tumour suppressor regulated by miR-9/9* in glioblastoma stem cells. EMBO J. 30, 4309–4322. Shao, S., Wang, Y., Jin, S., Song, Y., Wang, X., Fan, W., Zhao, Z., Fu, M., Tong, T., Dong, L., et al. (2006). Gadd45a interacts with aurora-A and inhibits its kinase activity. J. Biol. Chem. 281, 28943–28950. Silber, J., Lim, D.A., Petritsch, C., Persson, A.I., Maunakea, A.K., Yu, M., Vandenberg, S.R., Ginzinger, D.G., James, C.D., Costello, J.F., et al. (2008). miR-124 and miR-137 inhibit proliferation of glioblastoma multiforme cells and induce differentiation of brain tumor stem cells. BMC Med. 6, 14. Singh, S.K., Hawkins, C., Clarke, I.D., Squire, J.A., Bayani, J., Hide, T., Henkelman, R.M., Cusimano, M.D., and Dirks, P.B. (2004). Identification of human brain tumour initiating cells. Nature 432, 396–401. Slack, F.J., and Weidhaas, J.B. (2006). MicroRNAs as a potential magic bullet in cancer. Future Oncol. 2, 73–82. Wang, A., Alimova, I.N., Luo, P., Jong, A., Triche, T.J., and Wu, L. (2010). Loss of CAK phosphorylation of RARalpha mediates transcriptional control of retinoid-induced cancer cell differentiation. FASEB J. 24, 833–843. Wang, J., Wang, H., Li, Z., Wu, Q., Lathia, J.D., McLendon, R.E., Hjelmeland, A.B., and Rich, J.N. (2008). c-Myc is required for maintenance of glioma cancer stem cells. PLoS One 3, e3769. Yao, J., Duan, L., Fan, M., Yuan, J., and Wu, X. (2007). Overexpression of BLCAP induces S phase arrest and apoptosis independent of p53 and NF-kappaB in human tongue carcinoma : BLCAP overexpression induces S phase arrest and apoptosis. Mol. Cell. Biochem. 297, 81–92. Zheng, H., Ying, H., Yan, H., Kimmelman, A.C., Hiller, D.J., Chen, A.J., Perry, S.R., Tonon, G., Chu, G.C., Ding, Z., et al. (2008). Pten and p53 converge on c-Myc to control differentiation, self-renewal, and transformation of normal and neoplastic stem cells in glioblastoma. Cold Spring Harb. Symp. Quant. Biol. 73, 427–437.

602 Cell Reports 2, 591–602, September 27, 2012 ª2012 The Authors