Ets Factors Regulate Neural Stem Cell Depletion and Gliogenesis in Ras Pathway Glioma

Article

Ets Factors Regulate Neural Stem Cell Depletion and Gliogenesis in Ras Pathway Glioma Graphical Abstract

Authors Joshua J. Breunig, Rachelle Levy, C. Danielle Antonuk, ..., Serguei I. Bannykh, Roel G.W. Verhaak, Moise Danielpour

Correspondence [email protected]

In Brief Breunig et al. report that increased Ras signaling functions to deplete neural stem cells and expand glial progenitors in gliomagenesis. Inhibition of the upregulated Ets signaling downstream of Ras is sufficient to inhibit glioma formation by attenuating the gliogenesis necessary for tumor propagation.

Highlights d

Rapid brain tumor modeling with postnatal electroporation and transposon methodology

d

Modeling methodology allows for extensive interrogation of tumor growth mechanisms

d

d

Ras pathway mutations deplete neural stem cells and upregulate Ets factors Ets signaling block rescues Ras-mediated stem cell loss and prevents tumor formation

Breunig et al., 2015, Cell Reports 12, 258–271 July 14, 2015 ª2015 The Authors http://dx.doi.org/10.1016/j.celrep.2015.06.012

Accession Numbers GSE69254

Cell Reports

Article Ets Factors Regulate Neural Stem Cell Depletion and Gliogenesis in Ras Pathway Glioma Joshua J. Breunig,1,2,3,* Rachelle Levy,1 C. Danielle Antonuk,1 Jessica Molina,1 Marina Dutra-Clarke,1 Hannah Park,1 Aslam Abbasi Akhtar,1,2 Gi Bum Kim,1 Xin Hu,4 Serguei I. Bannykh,5 Roel G.W. Verhaak,4 and Moise Danielpour1,3,6 1Board

of Governors Regenerative Medicine Institute, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA of Biomedical Sciences, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA 3Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA 4Department of Genomic Medicine, Department of Bioinformatics and Computational Biology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA 5Department of Pathology and Laboratory Medicine, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA 6Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.celrep.2015.06.012 This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 2Department

SUMMARY

As the list of putative driver mutations in glioma grows, we are just beginning to elucidate the effects of dysregulated developmental signaling pathways on the transformation of neural cells. We have employed a postnatal, mosaic, autochthonous glioma model that captures the first hours and days of gliomagenesis in more resolution than conventional genetically engineered mouse models of cancer. We provide evidence that disruption of the Nf1-Ras pathway in the ventricular zone at multiple signaling nodes uniformly results in rapid neural stem cell depletion, progenitor hyperproliferation, and gliogenic lineage restriction. Abolishing Ets subfamily activity, which is upregulated downstream of Ras, rescues these phenotypes and blocks glioma initiation. Thus, the Nf1-Ras-Ets axis might be one of the select molecular pathways that are perturbed for initiation and maintenance in glioma. INTRODUCTION High-grade gliomas are among the most fatal primary brain tumors (Ostrom et al., 2014) and remain difficult to treat, although recent advances in radiation and chemotherapy have added a few months of survival, albeit with side effects (Stupp et al., 2009). Most treated primary gliomas inevitably evolve into secondary gliomas, which are almost always fatal (Quant et al., 2010). Understanding the molecular alteration(s) underlying the initial formation of tumors may determine critical steps required in the oncogenic process and may thereby lead to targeted therapies. Genetically engineered mouse models have been used to study initiating events in gliomagenesis by utilizing conditional knockout of tumor suppressors in promoter-defined cell populations or by viral delivery of oncogenes. Commonly investigated 258 Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors

themes include the cell(s) of mutation (i.e., cell in which the driver gene is mutated), the cell(s) of origin (the cancer initiating cell), and the nature of tumor propagating cells (proliferative cells contributing to tumor mass, which may not be able to generate the full repertoire of tumor cells) in glioma (Visvader, 2011). Initial findings suggested that brain tumor initiating cells display the properties of neural stem cells (NSCs), including the ability to self-renew and give rise to multiple daughter cell types (Alcantara Llaguno et al., 2009). Additional studies provided evidence that oligodendrocyte progenitor cells (OPCs) can directly transform into symmetrically expanding tumor-propagating cells upon acquisition of driver mutations (Assanah et al., 2006; Canoll and Goldman, 2008; Lei et al., 2011; Liu et al., 2011; Persson et al., 2010; Sugiarto et al., 2011). It remains unclear if these disparate findings reflect methodological differences or intrinsic differences in tumor types (Chen et al., 2012b; Lei and Canoll, 2011). Nevertheless, the resulting tumors in all of these models (and in patient tissue) are made up of cells histologically resembling glia (Chen et al., 2012b), suggesting that signaling pathways necessary for gliogenesis might be required for glioma regardless of the cell of origin. Interestingly, the perinatal switch from neurogenesis to gliogenesis during development has recently been found to critically involve the Ras pathway, a pathway intimately linked to glioma (Cancer Genome Atlas Research Network, 2008). Direct Ras activating mutations (e.g., Ras G12V) display increased gliogenesis (Li et al., 2014). Similarly, inactivating mutations to Nf1—a Ras inhibitor—and activating Pdgfra mutations are associated with glial proliferation and glioma (Liu et al., 2011; Paugh et al., 2013; Wang et al., 2012). Moreover, conditional knockout of the downstream effectors of Ras—Mek1 and Mek2--led to abrogated developmental gliogenesis (Li et al., 2012). Furthermore, this study identified Etv5, a member of the Ets transcription factor family, as a likely mediator of perinatal gliogenesis (Li et al., 2012). To investigate the specification of glial cells during glioma initiation, we focally targeted the postnatal day 2 ventricular zone (VZ) by combining electroporation (EP) with piggyBac transposition. This allows for mosaic genetic modification of

the proliferating NSC and progenitor populations. Specifically, we hyperactivated the Ras pathway either by EP of Kras or Hras G12V mutants or by EP of Pdgfra or Erbb2 receptor mutants. Additionally, Ras was disinhibited through Nf1 knockdown or knockout by directly targeting Nf1 with state-of-the-art miR-E based short hairpin (sh)RNAs (Fellmann et al., 2013) or using Nf1 floxed mice (Zhu et al., 2001). The ability of this model to track the transformative events resulting from disparate mutations in VZ cells with fluorescent genetic reporters led to the discovery of common events occurring during tumorigenesis. Disrupting the Ras pathway in these cells drives a depletion of NSCs and an emergence of rapidly proliferating glia that subsequently yield tumors with differing glioma subtype profiles and pathological grades. We demonstrate that this Ras-mediated glioblast specification requires increased Ets activity. Disrupting Ets signaling effectively reduces Rasmediated NSC depletion and tumor formation. Taken together, the Ets family may represent a critical component in stem and progenitor cell transformation into glia and ultimately the formation of glioma. RESULTS Stable Transgenesis of VZ Cells To evaluate the ability of transposition to mediate stable transgene expression, plasmids harboring CMV/chicken beta actin (CAG)-driven EGFP flanked by terminal repeats (PB-TR) that enable the genomic integration by PBASE (Chen et al., 2012a) were expressed in the mouse left lateral ventricle using EP at postnatal day 2 along with an electrode orientation to target the striatal wall of the left ventricle (Figures 1A and 1B). At 6 months post-EP, a large number of cells showed stable EGFP expression, while the EP of an episomal plasmid lacking PB-TR resulted in a 10-fold reduction in the number of stably fluorescing cells, indicating that PBASE transposition facilitates stable transgenesis (Figure 1C). Mouse brains were analyzed 4 hr post-EP to acutely identify the electroporated (EP-ed) cells, most of which were radial glia (RG), a bonafide NSC population (Rakic, 2003), and a small fraction expressed the progenitor markers ASCL1 and/or OLIG2, but no cell was both ASCL1+ and OLIG2+ (Figures 1D and 1E). At two days after EP, many cells remained VIMENTIN+ RG, possessing prototypical apico-basal polarity and a VZ-anchored cell body with a basal process >100 mm (Figures 1F–1G4 and S1A). Over six months, these EP-ed cells collectively gave rise to olfactory bulb neurons (Figures 1H and 1J), OPCs (Figure 1K, SOX10+/ OLIG2+ cells morphologically identifiable as multipolar glia exhibiting fewer processes when compared with astrocytes as seen in Figures S1B–S1B3), oligodendrocytes (Figures 1I [arrows], 1K, and 1L), astrocytes (Figure 1M; ALDH1L1+/GFAP+ glia with a dense cloud of processes as displayed in Figures S1C–S1C3), and ependymal cells (Figure 1N; i.e., VZ-located, apically multiciliated, cuboidal cells). Some RGs remained in the VZ, demonstrating their long-term self-renewal or quiescence (Figure 1I). We did not observe any hyperplasia or tumor formation in transposed EP-ed cells (e.g., containing only the stably inserted TagBFP2 nuclear reporter or EGFP; Figure 1I; J.J.B., R.L., data not shown).

Tumorigenesis Initiated by EP of Single Oncogenes To assess the ability of this methodology to generate glioma, we EP-ed the following constitutively active oncogenes, which have been identified as driver mutations in previous models or in patient tumors: Erbb2-V664E, Hras-G12V, Kras-G12V, and Pdgfra-D842V (Catalogue of Somatic Mutations in Cancer, 2008; Forshew et al., 2009; Janzarik et al., 2007; Ozawa et al., 2010; Schwartzentruber et al., 2012; Sharma et al., 2005; Yamaoka et al., 2000). Several strategies were employed to unambiguously visualize targeted cells. For Erbb2 plasmids, we co-expressed the mCLOVER GFP using a self-cleavable 2A peptide sequence (Figures S2A and S2B). Hras-G12V and Kras-G12V genes were fused with EGFP (Figure S2A) and Pdgfra-D842V was constructed without a reporter or tag (Figure S2A). The control gene was a fusion of EGFP with the CAAX box found in Hras C terminus, which results in membrane labeling (Figure S2C). This membrane-targeted EGFP was also used for visualization of Pdgfra D842V by co-EP of both plasmids. Stable expression of Erbb2-V664E or Hras-G12V oncogenes led to rapid development of hyperplasia between 6 days and 2 weeks (Figures 2A, 2B, S2D, and S2E). Though starting as small populations of VZ cells, the EGFP+ cells became highly infiltrative and tumors occupied a significant proportion of the forebrain by 6 weeks (Figures 2C and S2F). EP of Kras and Pdgfra mutations also resulted in tumors (Figures S2G–S2J). Co-expression of oncogenes with a ferritin-expressing plasmid permitted cell growth to be tracked in live animals using MRI. Results showed tumor growth 3 weeks post-EP specifically in the left hemisphere at the EP site (Figures 2D–2F2). Importantly, wild-type Hras or Erbb2 overexpression did not exhibit noticeable increases in proliferation or loss of apicobasal polarity (Figures S2K and S2L). Furthermore, no mice EP-ed with control plasmids developed tumors, indicating that pBase expression or random genomic insertions of plasmids are not sufficient to initiate significant tumorigenesis under these conditions. These findings show that PBASE-mediated stable integration of a single constitutively active Ras pathway mutation in the VZ niche can initiate tumors. Hras G12V and Erbb2-CA driven tumors were 100% penetrant (Figure 2G). The majority of Hras G12V animals showed hydrocephalus and doming of the skull and reached terminal endpoint within 200 days, whereas control animals (i.e., EP-ed with membrane EGFP and transposed with PBASE) did not show any abnormal symptoms or shortened lifespan (Figures 2G–2I). We observed overgrowth of the left hemisphere, abnormal vascularity, and cortical thinning in tumorbearing animals (Figure 2J). The Hras-G12V tumors were graded by a clinical neuropathologist according to the World Health Organization criteria and found to possess the hallmarks of high grade anaplastic glioma by pathology, notably an oligodendroglioma component as demonstrated by the artifactual ‘‘fried egg’’ appearance of tumor cells, necrosis, endothelial proliferation, calcification, and high cellularity (Figures 2K–2P). Erbb2-V664E animals exhibited similar pathological features (Figures S2M–S2R). All Ras and Erbb2 mutations were pathologically classified as high grade glioma. However, even littermates—treated Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors 259

Figure 1. Postnatal EP Combined with PiggyBac Transposition Targets NSCs and Progenitor Cells and Provides Sustained Transgene Expression (A) Postnatal EP of the VZ cells with EGFP-expressing plasmid. (B) PBASE catalyzes the genomic integration of gene of interest (GOI). (C) Rate of stable expression by PBASE-mediated integration, gene trapping, or episomal plasmid expression. The error bars show ± SEM; n = 3 (*p < 0.05, **p < 0.01, and paired t test). (D and E) P2 coronal section taken 4 hr after EP shows mostly RGs and few ASCL1+ and/or OLIG2+ progenitors. (F–G4) P4 section EP-ed at P2 yields largely VIMENTIN+ RGs and few non-VZ progenitors (arrowheads). (H and J–N) The multipotent VZ cells EP-ed at P2 give rise to (H and J) neurons, (K) immature, and (L) mature oligodendrocytes, (M) astrocytes, and (N) multiciliated ependymal cells. (I) 6-month section EP-ed at P2 confirms long-term stable expression. (The scale bars for A, D, F, H, and I represent 100 mm and the scale bars for G and J–N represent 10 mm).

260 Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors

Figure 2. Tumor Progression and Pathological Findings in EP-ed Mice (A–C) Stitched images showing time-series development of Hras tumors. (D and E) MRI images of 3-week brains co-EP-ed with ferretin plus Hras or EGFP. (F–F2) Micro-MRI showing ferritin+ tumor in the left hemisphere. (G) Survival analysis of control, Erbb2 CA, and Hras-G12V EP-ed mice. (H and I) Hras G12V animal presents a domed skull and hydrocephalus in the left hemisphere (J) at moribundity. (K–P) Pathological hallmarks of high-grade glioma at moribundity found in Hras tumor. (The magnification for K, M, N, and O, is 1003 and L and P is 2003).

identically—could exhibit divergent pathologies within the grade III and IV subtypes. Specifically, tumors were diagnosed as anaplastic astrocytoma, anaplastic oligodendroglioma, anaplastic oligoastrocytoma, or glioblastoma multiforme. Premature RG Depletion and Progenitor Hyperplasia through Ras Hyperactivation Employing a counting frame-based system for quantification (Figures S3A and S3B), we observed that EP-ed VZ cells rapidly lost their RG morphology and increased proliferation by 6 days (Figures 3A–3H). This conversion of morphology happened in all Hras, Pdgfra, and Kras EPs, suggesting that this effect is the result of the consistent hyperactivation of Ras that is commonly associated with gliomagenesis (Figures 3G and S2G–S2J). This

morphological change was significantly correlated with increases in OLIG2 expression, indicating potential lineage restriction (Figures 3I–3O). We employed SOX10 immunolabeling to investigate oligodendrocyte lineage restriction, as it is specific for this class of cells in the normal brain. From 6 days to 2 weeks, the number of SOX10+/EGFP+ tumor cells rapidly increased as the tumor progressed, indicating an increase in OPCs within the growing tumor (Figures 3P–3V). The acute changes in expression of oligodendrocyte cell markers were replicated using cerebral cortex-derived human neural progenitor cells (hNPCs). hNPCs EP-ed with Hras or Kras mutants became compact in nature and upregulated OLIG2 and downregulated the astrocyte lineage marker GFAP, Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors 261

Figure 3. Somatic Mutation Rapidly Depletes RG and Expands Tumor Cells (A–F) P4 and P8 coronal sections of control, Erbb, and Hras brains stained with V5, EGFP, and KI67. (F2–F5) Co-localization of KI67, EGFP, and V5 in the boxed area from (F). (G and H) Line chart comparing the quantification of EGFP-positive KI67+ cells and the RGs at 2 days and 6 days after EPs with control (A and D), Hras (C and F), Erbb (B and E), and WT-Hras (data not shown). (I–N) P4 and P8 coronal sections of control, Erbb, and Hras brains stained with V5, EGFP, PDGFRA, and OLIG2. (M2–M6) Co-localization of PDGFRA, EGFP, V5, and OLIG2 in the boxed area from (M). (O) Line chart comparing the quantification of EGFPpositive OLIG2+ at 2 days and 6 days after EPs with control (I and L), Erbb (J and M), Hras (K and N), and WTHras (data not shown). (P–U) P8 and P16 coronal sections of control, Erbb, and Hras brains stained with V5, EGFP, and SOX10. (U2–U5) Co-localization of SOX10, EGFP, and V5 in the boxed area from (U). (V) Line chart comparing the quantification of EGFPpositive Sox10+ at 6 days and 2 weeks after EPs with control (P and S), Hras (R and U), Erbb (Q and T), and WT-HRas (data not shown). The error bars show ± SEM; n = 3 mice (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, and paired t test). (The scale bar represents 25 mm for A–F, I–N, and P–U).

262 Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors

Figure 4. Nf1 Loss of Function through Knockdown or Conditional Knockout Shows Significant Reduction in RG and Striatal Gliogenesis (A) Schematic of Nf1 miR-E knockdown construct. (B) Strategy for floxed Nf1 mouse targeting by EP of Cre along with a FlEx Cre reporter. (C–E) WT CD1 brains EP-ed with shLuc, shNf1, and shNf1 plus NF1 cDNA. (F–H) Nf1fl/fl brains EP-ed with EGFP, Cre, and Cre plus NF1 cDNA. The white arrows show RG fibers; the red arrowhead shows truncated RG fiber; and the orange arrowhead shows glial clusters. (I) Histograms showing the relative quantifications of the RGs by morphology in the brains shown by (C)–(E). (J) Histograms showing the relative quantifications of the RGs by morphology in the brains shown by (C)–(H). The error bars represent ± SEM; n = 3 mice (*p < 0.05, **p < 0.01, ***p < 0.001, and paired t test). (The scale bar represents 100 mm for C–H).

while control hNPCs retained GFAP expression and a spindle shaped morphology reminiscent of RG in vitro (Figures S3C– S3F). Ras hyperactivation often results from the loss of Nf1 activity that frequently occurs in glioma patients (Dasgupta et al., 2005; Schwartzentruber et al., 2012). To test whether Nf1 deficiency mirrors Hras phenotypes, we adapted two methodologies for use with our EP technique. First, we knocked down Nf1 in EP-ed cells using the recently reported miR-E shRNA technology (Figure 4A) (Fellmann et al., 2013). In addition, we created a strategy to EP Nf1 floxed mice (Zhu et al., 2001) with a custom-made FlEx reporter plasmid and Cre (Figure 4B). In validating miR-E targeting plasmids, western blot analysis confirmed that all five candidate Nf1 shRNA sequences efficiently knocked down their cognate sensor EGFP in transfected N2a cells more strongly than a previously characterized shEGFP (Figures S4A and S4B; Matsuda and Cepko, 2007). Nf1.789 (shNf1) was used in all further experiments because its effect could be rescued by a codon-optimized human NF1 cDNA (Stowe et al., 2012; Figures S4C and S4D). When we EP-ed shNf1 and sensor EGFP, only a few ‘‘escaping’’ EGFP+ RGs remained, whereas a large increase in EGFP striatal cells could be seen in these groups (Figures S4E–S4E4). Control shRNA against firefly luciferase (luc.1309 heretofore

‘‘shLuc’’; Premsrirut et al., 2011) did not decrease sensor EGFP expression or change the overall distribution of EGFP cells (Figures S4F–S4F4). We then EP-ed littermates with shLuc, shNf1, or shNf1 + human NF1. After 6 days, shLuc control mice retained the stereotypical RGs, while shNf1 mice displayed markedly more striatal glia and a loss of RGs. These shNf1 phenotypes were partially rescued by co-expressing the human NF1 cDNA (Figures 4C– 4E). Immunostaining for NF1 protein confirmed the resistance of the human NF1 cDNA (Stowe et al., 2012) to shNf1 after coEP of both (Figures S4G–S4G3). Notably, the human NF1 signal was enriched in VZ cells, but this episomal cDNA appeared to be rapidly diluted in the escaping, proliferative glial populations (Figures S4G–S4G3). The RG depletion phenotypes were replicated in the floxed Nf1 mice EP-ed with Cre recombinase (Figures 4F and 4G) and the loss of RGs was again attenuated by co-expressing human NF1 (Figure 4H). Both shNf1-EP-ed and Cre-recombined Nf1 mice showed a significant reduction in the number of RGs (Figures 4I and 4J). These data suggest that both Ras hyperactivation and Nf1 reduction in RG leads to the cell autonomous depletion of RGs. In addition to creating focal, genetic mosaic models for cell tracing, we used this technology to test if the depletion of RGs is evidence of a change in neural stem cell potential. Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors 263

(legend on next page)

264 Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors

EP-ed plasmids are preferentially taken up by the RGs undergoing M-phase of the cell cycle (Stancik et al., 2010). Because the VZ cell cycle is roughly 12–18 hr, we hypothesized that we could electroporate the same animal twice with different plasmids, 8 hr apart, and expect to generate two genetically different cohorts of RGs (Figure S5A). To mitigate the possibility of double positive cells, we added Cre recombinase in the first round of EP to inactivate any floxed-Hras plasmid that might enter the same cell during the second EP (Figure S5A). After performing double EP with hemagglutinin (HA)-tagged, nuclear beta-glucosidase (Bgla; McCutcheon et al., 2010) + Cre in the first EP (control group) and nuclear TagBFP2 (TagBFP2nlsV5) + Ollas/FLAG-epitope-tagged Hras G12V in the second EP (HRas group), we dissociated EP-ed cells from four animals, grew them in culture with growth factors for 2 days to permit expansion and clear debris, and then re-plated cells on coverslips without growth factors to induce differentiation (Figure S5A). As with our in vivo experiments, a significant increase in OLIG2+ cells was observed in the Hras group (Figures S5B and S5C). After 7 days, the absolute number of OLIG2+ cells in the Hras group was almost 17-fold higher (290 ± 141 OLIG2+ cells in Hras-G12V versus 17 ± 7 OLIG2+ cells in controls; n = 4). Neuronal differentiation was almost completely abolished in Hras cells, as we found a total of four doublecortin-positive neuronal cells from 4,505 TagBFP2+/Hras-G12V cells sampled across four cell lines (Figures S5B and S5D). Astrocytic differentiation was decreased in Hras-G12V cells (64.8% ± 12.7% compared to 48.6% ± 16.4% GFAP cells; Figures S5B and S5E). In agreement with the disappearance of olfactory bulb (OB) neurons in vitro, OB neurogenesis appeared to decrease in vivo in Hras-G12V and Erbb2 groups (Figures S5F–S5H). These data show that Hras hyperactivation in NSCs prevents neuronal lineage specification and favors glial lineage specification. Collectively, these results suggest that the NSC character is lost acutely during a RTK/Nf1/Ras-mediated expansion of glioblastic tumor cells. Characterization of PiggyBac Tumor Propagating Cells Three Hras+ and three Erbb2+ tumors from different animals were dissociated and fluorescently sorted to give pure EGFP+ populations (Figure 5A). The Hras tumor cell lines all highly expressed PDGFRA and SOX10 (at least 80%), though one Hras cell line had significantly less PDGFRA+, SOX10+ cells (Figures 5B–5D). This PDGFRA/SOX10 co-expression was not seen in ‘‘sibling’’, control cells that were isolated from EGFP-expressing, PBASE-transduced animals (Figure 5B). RNA isolated from the six Hras-G12V and Erbb2V664E cell lines and control NSCs was analyzed by microarray. Single-sample gene set analysis (GSEA) of cell lines using the brain tran-

scriptome database (Cahoy et al., 2008) and the glioma subtype signatures (Verhaak et al., 2010) revealed that the tumor cell lines were highly heterogeneous when viewed as individual lines (Figure 5E). Erbb2-1, Erbb2-2, and Hras-1 were enriched for oligodendroglial signature, while Erbb2-3, Hras-2, and Hras-3 had a more astroglial profile (Figure 5E). These three oligodendroglial lines were found to contain more proneural/neural signatures, whereas the three astroglial lines were more mesenchymal (Figure 5F). Given that the Hras lines and Erbb2 lines were derived from same birthdate animals (i.e., EP-ed during the same surgery) from the same respective DNA mixes (i.e., containing either Hras or Erbb2 for the respective groups), we suspected that the divergence of expression profiles within tumor evolution in each group might be partially attributed to the stochastic acquisition of secondary mutations, which is suggested by several recent studies reporting that gliomas exhibit high intratumoral heterogeneity (Patel et al., 2014; Sottoriva et al., 2013). In screening several tumor suppressors, we observed that all tumors harbored at least one secondary mutation in Trp53, p16, or p19, indicating a potential mechanism for their heterogeneity (Figure S6A). Several of these Trp53 mutations (A135V, V170M, and V213M) have also been observed in other mouse models of glioma or in human glioma (Schwartzentruber et al., 2012; Song et al., 2013). To confirm that EGFP+ cells are truly tumorigenic, we performed transplantation of 100,000 tumors cells into naive postnatal day (P)2 mice and saw invasive tumors similar to the initial primary tumors marked by high SOX10 and OLIG2 expressions (Figures S6B–S6D). Etv5 Is Required for Ras-Mediated Gliomagenesis Recently, Etv5 has been reported to be critical for perinatal gliogenesis (Li et al., 2012). A preliminary microarray screen showed that many members of the Ets transcription factor family, and Pea3 subfamilies of the Ets family, were upregulated in the tumor cells when compared with NSCs (except Etv2, which does not appear to be expressed in the brain; Figures 6A, 6B, and S7A). The upregulation of the Ets family was confirmed by real-time quantitative (q)RT-PCR across Kras, Hras, and Erbb tumors (Figure 6B). Using a wellcharacterized ETV1 (aka ER81) antibody that efficiently labels deep-layer cortical neurons (A.A.A., data not shown), we observed the upregulation of ETV1 protein in tumor cells (Figures 6C–6C2). Because of the pleiotropic function of Ets factors in the brain, and due to the fact that so many members of the Ets family were simultaneously upregulated, it would be difficult to use shRNA or floxed mice to assess Ets function in tumorigenesis. Thus, we engineered a dominant-negative

Figure 5. Isolation and Microarray Analysis of Highly Pure Populations of Tumor Cells (A–C) Dissociated cells from the SVZ EP-ed with Hras, but not control EGFP, express PDGFRA and SOX10. The error bars show ± SEM (The scale bar represents 25 mm for B and C). (D) Quantification of SOX10+ and PDGFRA+ cells in control NSC line and three different Hras G12V cell lines. (E) Transcriptome comparison of cultured tumor progenitors and NSC populations. (F) Classification of cell lines according to the four subtypes of human glioblastoma multiforme (GBMs) defined by The Cancer Genome Atlas (TCGA) with the single sample Gene Set Enrichment Analysis (ssGSEA) method.

Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors 265

Figure 6. Ets Family Transcription Factors Are Upregulated in Glioma and Regulate Tumor Cell Phenotypes (A) Increased Ets family mRNA expression in tumor cells by microarray. (B) Real-time qRT-PCR confirmation of microarray data. (C–C2) Nuclear ETV1 (aka ER81) expression in cortical EGFP+ tumor cells. (D and E) Plasmid schematics for bicistronic expression of DN-Etv5 and EGFP-Hras along with control containing TagBFP2-3XFLAG-nls. Note that the P2A element ensures that all EGFP-Hras G12V cells will express upstream protein so that tumor cells cannot escape DN-Etv5. (F and G) Coronal sections of Hras+3XFLAG (F) and Hras+DN-Etv5 (G) brains at 6 days post-EP. (The scale bars represent 50 mm for A–C). (H–J) Comparison of KI67+, OLIG2, and RGs in Hras and Hras+DN-Etv5 brains. The error bars show ± SEM; n = 3 mice (*p < 0.05, **p < 0.01, and paired t test).

Etv5 (DN-Etv5), which consists of TagBFP2 fused to the DNA binding domain of Etv5. DN-Etv5 cross-reacts with other members of the Ets family to suppress their transcriptional activities by binding the GGA(A/T) consensus site (Hasegawa 266 Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors

et al., 2004). Initial attempts at co-EP of Hras G12V and DN-Etv5 indicated that cells expressing DN-Etv5 did not transform at a high rate, but escaping Hras-G12V cells quickly out-competed DN-Etv5+ populations (R.L., data not shown). We then constructed a bicistronic plasmid containing EGFP-Hras G12V and TagBFP2-DN-Etv5 separated by a P2A element (Figure 6D). We substituted the Etv5 DNA binding domain with three FLAG epitopes and a nuclear localization sequence (nls) to construct the control plasmid (Figure 6E). A constitutively active Etv5DBD-VP64 and a full-length TagBFP5Etv5 were also generated (Figure S7B). Proper function of the P2A element was confirmed by the appropriate membrane localization of Hras and nuclear localization of TagBFP2-DN-Etv5 (Figure S7C). Furthermore, a SRE-driven dual luciferase assay validated that Hras G12V activity in the P2A-containing plasmid was equivalent to co-expressed Hras G12V (Figure S7D). Moreover, coexpression of these plasmids with Pea3-driven luciferase plasmid revealed that Hras upregulated Pea3 luciferase activity and that co-expressed DN-Etv5 (i.e., Hras and DN-Etv5 in separate plasmids) or bicistronic Hras/DN-Etv5 efficiently reduces this increase (Figure S7E). Further, Pea3driven luciferase activity was increased by Etv5 overexpression or more markedly by Etv5-VP64 expression (Figure S7E).

Figure 7. Ets Inhibition Blocks Tumor Formation and Prevents Morbidity in Hras G12V Mice (A and B) Hemisphere images of Hras and Hras+DN-Etv5 brains. (C) 8-month long survival analysis indicates DN-Etv5 addition prevents morbidity (also compare with survival of animals in Figure 2G). (D) Eight months post-EP, DN-Etv5+EGFP-Hras G12V animals did not indicate any tumor presence. Cells did not express Sox10 or Olig2 (E–E2; the arrowheads denote EGFP-/SOX10+/OLIG2+ cells), but did express the astrocytic markers ALDOC and ALDH1L1 (F–F2; the red arrow denotes weakly Aldoc+/Aldh1l1+/EGFP+ cell, whereas the white arrows mark triple positive cells).

Although bicistronic TagBFP-3XFLAG-nls/Hras G12V rapidly generated proliferative hyperplasias, EP of bicistronic DN-Etv5/ Hras G12V resulted in an EGFP+ population that resembled the normal, control EGFP EP-ed brains and failed to generate tumors (Figures 6F and 6G). There was no evidence of gross hyperproliferation, RG depletion, or the increase in KI67+ proliferating cells or OLIG2+ cells in the DN-Etv5-containing group (Figures 6H–6J). At 3 weeks post-EP, DN-Etv5 animals had no hyperplasia and no tumor, which contrasted with the 3FLAG-bearing littermates (Figures 7A and 7B). The overall number of cells in DN-Etv5 mice was still comparable to those seen in six month controls (e.g., Figure 1I). DN-Etv5 animals survived well beyond 8 months, whereas the control group (TagBFP2-3FLAG-nls-P2A-Hras) died (Figure 7C). The presence of large cohorts of EGFP+ cells in these long-lived DN-Etv5 mice argues against non-specific cytotoxicity (Figure 7D). Histology of these animals did not yield evidence of tumor, but we did observe numerous hypertrophic astroglia akin to those seen in episomal Hras G12V animals (J.J.B., data not shown). These hypertrophic glia did not express SOX10 or OLIG2, but instead expressed astrocyte markers ALDOC and ALDH1L1 and had an astrocyte-like morphology (Figures 7E–7F2).

DISCUSSION We combined postnatal EP with piggyBac transposition to model glioma by generating somatic transgenic mutants. Through the ability to rapidly model several glioma driver mutations, we identified that Ras pathway glioma requires Ets family transcription factor activity in order to transform mutationharboring VZ populations into de facto tumors. Using our modeling technology with defined temporal and spatial transgene expression, we demonstrate that somatic Ras hyperactivating mutations (i.e., Ras mutations, RTK mutations, or Nf1 loss-of-function [LOF] manipulations) in VZ populations cause a rapid (<6 days) depletion of NSCs and a massive expansion of glial progenitors. Such phenomena would be difficult to observe in traditional mouse models. For example, tamoxifen-induced recombination in genetically engineered mouse models happens over days in an organ-wide population of cells and often requires multiple, staggered injections, complicating the ability to identify the very first cells that might rapidly undergo transformations. Nevertheless, additional studies in our model will be necessary to rigorously lineage trace these diverse VZ populations to see which of the initial populations (i.e., RG, ASCL1+, and/or OLIG2+) will eventually contribute Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors 267

to the tumor bulk and, therefore, whether the depletion of RG is directly tied to progenitor expansion. The molecular mechanisms of the perinatal switch from neurogenesis to gliogenesis have been an active area of investigation over the past decade (Rowitch and Kriegstein, 2010). This field of inquiry has great importance due to the underlying glial nature of tumor propagating cells. Nf1 ablation was shown to promote proliferation of immature astroglial progenitors (Dasgupta and Gutmann, 2005), while diminishing neuronal differentiation (Hegedus et al., 2007). Furthermore, alterations in neurosphere behavior have been observed in conditional Nf1 knockouts in a region-dependent manner (Lee et al., 2010, 2012). Interestingly, extracellular signal-regulated kinases (ERK) inhibition is able to rescue cell fate specification in Nf1 inactivated mice by preventing transit-amplifying progenitor cell expression of OLIG2 (Wang et al., 2012). Of particular note, it was recently reported that Mek activity is critical in regulating the neuron/glia fate switch in cortical progenitors, as these knockouts demonstrated markedly diminished gliogenesis (Li et al., 2012). Our findings further extend these previous findings by demonstrating Nf1/ Ras function in NSCs to regulate their maintenance. Specifically, Ras hyperactivity due to direct mutation, Pdgfra or Erbb2 mutation, or through Nf1 knockdown or knockout, all result in RG depletion. However, unlike Nf1 mutations, direct Ras mutations are very rarely observed in glioma (Brennan et al., 2013), so it will be important to employ other mutations and/or combinations of mutations to see if these results can be generalized. During the developmental gliogenic switch, Etv5 is the only member of the entire Ets family to be notably (>2-fold) downregulated in the Mek1/2-ablated brains (Li et al., 2012). However, we found that seven members of the Ets family are upregulated in our RTK or Ras-driven tumors. By employing a DN-Etv5, we note that the premature depletion of NSCs and glial hyperproliferation can be blocked downstream of Ras hyperactivity. Nevertheless, it is clear that Nf1 additionally functions in the transit-amplifying population (Wang et al., 2012). Given that RG and progenitor populations are simultaneously EP-ed, it would follow that DN-Etv5 is sufficient to block the downstream effects of Ras hyperactivity independently in both of these contexts (i.e., rescuing RG and preventing progenitor hyperproliferation). Interestingly, persisting glia resemble the hypertrophic astrocytes and not the OPC-like tumor cells, suggesting that heightened Ets transcriptional activity is necessary for the OPC tumor propagating cell character. Given the heterogeneity of driver mutations in glioma, it will be interesting to determine whether Ets family transcription factors are a common underlying transcriptional element in glioma or whether other signaling mechanisms are utilized by upstream drivers. Notably, we have found that there is tumor divergence even with the identical treatment of littermates (Figures 5D–5F). Specifically, Hras EP-ed (or Erbb2 EP-ed) littermates of the same age injected on the same day with the same DNA can yield progenitors falling into disparate glioma subtypes (proneural versus mesenchymal; Figure 5F) with different high grade pathological classifications (e.g., anaplastic oligodendroglioma versus 268 Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors

glioblastoma multiforme). Such findings concur with recent findings on tumor diversity in in utero EP models (Chen et al., 2014) and also with data from viral models suggesting that non-GCIMP tumor types evolve from a common proneural subtype (Ozawa et al., 2014). These findings in model systems are reinforced by the emerging findings regarding patient tumor heterogeneity (Patel et al., 2014; Sottoriva et al., 2010, 2013). However, despite the heterogeneity we saw in small groups of animals, dominantnegative inhibition of Ets signaling was able to completely rescue survival in much larger groups of animals, suggesting that Ets signaling may represent a potential choke point for tumor formation. Furthermore, it will be important to determine whether particular Ets family members are necessary for gliomagenesis or if these factors are functionally redundant. Moreover, we cannot assume that cell type properties are maintained during tumorigenesis, as certain cell characteristics can be re-acquired during tumorigenesis and subsequent de-differentiation (Visvader, 2011). Specifically, there may be tumor subpopulations that evolve Ets-independent tumor propagating capacities. Our results indicate that hyperactivation of the Ras pathway drives depletion of NSCs and the expansion of progenitors into glial tumor propagating cells. Further, we have identified Ets family signaling as a necessary downstream factor of Nf1/Ras that mediates tumor propagating cell production. These results may yield insight into previous findings regarding the tumor cell of origin in gliomas being NSCs (Alcantara Llaguno et al., 2009; Kwon et al., 2008; Wang et al., 2009; Zhu et al., 2005) or OPCs (Assanah et al., 2006; Lei et al., 2011; Liu et al., 2011; Persson et al., 2010; Sugiarto et al., 2011) given the intrinsic hierarchy of stem and progenitor lineages. Rigorous lineage tracing is still necessary to determine if the putative cell(s) of origin represent true differences in tumor subtypes or if these disparities are due to methodological differences that can be reconciled. Nevertheless, these Ets transcription factors and the resulting downstream signaling may present new clinical targets for combating glioma. EXPERIMENTAL PROCEDURES Mice and EP CD1 mice were used for most experiments, the exception being Nf1fl/flrelated experiments, which were performed according to the Cedars-Sinai Institutional Animal Care and Use Committee. Postnatal lateral ventricle EPs were performed as previously described (Breunig et al., 2012). In brief, P1–2 pups were placed on ice for 8 min until unresponsive to tail pressure. There was 1.2 ml of a plasmid DNA mix (typically 1.0 mg/ml) in Tris-EDTA buffer that was injected into the left lateral ventricle. Plasmid details are available in Table S1. Employing Signagel, platinum Tweezertrodes were used to EP with three to five pulses of 115–135 V (50 ms; separated by 950 ms) generated with the ECM 830 BTX Electroporator (Harvard Apparatus). Tissue Preparation After anesthesia, mouse brains were isolated and immersion fixed in 4% icecold paraformaldehyde (PFA) overnight. Brains were then embedded in low melting point 4% agarose and sectioned at 70 mm on a vibratome. Details can be found in the Supplemental Information. Immunohistochemistry Immunohistochemistry was performed using standard methodology (Breunig et al., 2012). Details about antibodies can be found in the Supplemental Information and Table S2.

Imaging Confocal images were collected on a Nikon A1R or C2 laser confocal microscope. For hemi or whole brain section images, the automated stitching function of Nikon Elements was used to create a seamless merged image from multiple image fields. Additional details can be found in the Supplemental Information. Quantification Sampling for quantification was based on the scheme described in the Supplemental Information. Purification of EGFP+ Tumor Cells EGFP+ wild-type (WT) NSCs (pooled from >4 animals) or EGFP+ tumor cells from a single tumor-bearing animal were first microdissected from the left hemisphere, digested in accutase, and grown as monolayers according to our previously described methods (Breunig et al., 2007). After growing to confluence, cells underwent fluorescence-activated cell sorting (FACS) for EGFP autofluorescence and were cultured as self-renewing monolayers. Further details can be found in the Supplemental Information.

Medicine Institute of Cedars-Sinai (to M.D. and J.J.B.); the Margaret E. Early foundation and the California Institute of Regenerative Medicine (to T. Town); and the Schmidt Family Foundation and the Paul and Vera Guerin Family Foundation (to M.D.). Received: February 19, 2015 Revised: April 27, 2015 Accepted: June 2, 2015 Published: July 2, 2015 REFERENCES Alcantara Llaguno, S.R., Chen, J., and Parada, L.F. (2009). Signaling in malignant astrocytomas: role of neural stem cells and its therapeutic implications. Clin. Cancer Res. 15, 7124–7129. Assanah, M., Lochhead, R., Ogden, A., Bruce, J., Goldman, J., and Canoll, P. (2006). Glial progenitors in adult white matter are driven to form malignant gliomas by platelet-derived growth factor-expressing retroviruses. J. Neurosci. 26, 6781–6790.

RNA Isolation, Microarray, and Gene Expression Analysis RNA from EGFP+ NSCs and EGFP+ tumor cells was isolated using RNeasy+ Kits according to manufacturer’s protocols and hybridized to Affymetrix Mouse Gene 1.0 ST microarrays. Neural cell type classification and ssGSEA were performed as previously described (Galvao et al., 2014; Liu et al., 2011; Verhaak et al., 2010).

Brennan, C.W., Verhaak, R.G., McKenna, A., Campos, B., Noushmehr, H., Salama, S.R., Zheng, S., Chakravarty, D., Sanborn, J.Z., Berman, S.H., et al.; TCGA Research Network (2013). The somatic genomic landscape of glioblastoma. Cell 155, 462–477.

ACCESSION NUMBERS

Breunig, J.J., Gate, D., Levy, R., Rodriguez, J., Jr., Kim, G.B., Danielpour, M., Svendsen, C.N., and Town, T. (2012). Rapid genetic targeting of pial surface neural progenitors and immature neurons by neonatal electroporation. Neural Dev. 7, 26.

Array data have been deposited in NCBI GEO and are available under accession number GEO: GSE69254. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, seven figures, and two tables and can be found with this article online at http://dx.doi.org/10.1016/j.celrep.2015.06.012. AUTHOR CONTRIBUTIONS J.J.B. conceived the models, made initial NSC depletion observations, and performed in silico plasmid design for newly created plasmids. J.J.B., R.L., C.D.A., G.K., and M.D. conceived and designed the experiments. J.J.B., R.L., C.D.A., M.D.C., and J.M. performed the in vitro experiments, histological analysis, and molecular cloning. J.J.B. and M.D. performed the in vivo EP. M.D., R.L., C.D.A., J.M., H.P., A.A.A., G.K., and J.J.B. analyzed the data and contributed to techniques. M.D., R.L., C.D.A., M.D.C., G.K., and J.J.B. wrote and edited the manuscript. G.K. helped condense the manuscript during revision. J.J.B. and G.K. revised the manuscript. X.H. and R.G.W.V. performed GSEA analysis. S.I.B. performed neuropathological examination. ACKNOWLEDGMENTS We thank J. Loturco and F. Siddiqi for the kind gift of the initial piggyBac plasmid system prior to publication and for sharing unpublished data. We thank H. Song and G. Sun for discussing unpublished data. Furthermore, we thank K. Bernau, M. Dominguez, J. Hassell, F. Matsuzaki, F. McCormick, M. Lin, N. Sestan, C. Svendsen, Y. Tanabe, and R. Tsien for generously providing additional plasmids. We would like to thank J. Rodriguez, Jr., D. Gate, and T. Town for helpful comments, experimental assistance, and financial support (from T. Town). We thank H. Zong for helpful comments on the model and for advice on glioma genomics, C. Fellman for miR-E assistance, and D. Guttmann for NF1 cDNA and antibody advice. We are thankful to S. Svendsen and D. Magoffin for critical review and editing of the manuscript. The authors would like to acknowledge support from the Samuel Oschin Comprehensive Cancer Institute Cancer Research Forum Award (to M.D. and J.J.B.); the Regenerative

Breunig, J.J., Silbereis, J., Vaccarino, F.M., Sestan, N., and Rakic, P. (2007). Notch regulates cell fate and dendrite morphology of newborn neurons in the postnatal dentate gyrus. Proc. Natl. Acad. Sci. USA 104, 20558–20563.

Cahoy, J.D., Emery, B., Kaushal, A., Foo, L.C., Zamanian, J.L., Christopherson, K.S., Xing, Y., Lubischer, J.L., Krieg, P.A., Krupenko, S.A., et al. (2008). A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J. Neurosci. 28, 264–278. Cancer Genome Atlas Research Network (2008). Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature 455, 1061–1068. Canoll, P., and Goldman, J.E. (2008). The interface between glial progenitors and gliomas. Acta Neuropathol. 116, 465–477. Catalogue of somatic mutations in cancer (COSMIC). (2008). Online database. http://cancer.sanger.ac.uk/cosmic. Chen, J., Li, Y., Yu, T.S., McKay, R.M., Burns, D.K., Kernie, S.G., and Parada, L.F. (2012a). A restricted cell population propagates glioblastoma growth after chemotherapy. Nature 488, 522–526. Chen, J., McKay, R.M., and Parada, L.F. (2012b). Malignant glioma: lessons from genomics, mouse models, and stem cells. Cell 149, 36–47. Chen, F., Becker, A.J., and LoTurco, J.J. (2014). Contribution of tumor heterogeneity in a new animal model of CNS tumors. Mol. Cancer Res. 12, 742–753. Dasgupta, B., and Gutmann, D.H. (2005). Neurofibromin regulates neural stem cell proliferation, survival, and astroglial differentiation in vitro and in vivo. J. Neurosci. 25, 5584–5594. Dasgupta, B., Li, W., Perry, A., and Gutmann, D.H. (2005). Glioma formation in neurofibromatosis 1 reflects preferential activation of K-RAS in astrocytes. Cancer Res. 65, 236–245. Fellmann, C., Hoffmann, T., Sridhar, V., Hopfgartner, B., Muhar, M., Roth, M., Lai, D.Y., Barbosa, I.A., Kwon, J.S., Guan, Y., et al. (2013). An optimized microRNA backbone for effective single-copy RNAi. Cell Rep. 5, 1704–1713. Forshew, T., Tatevossian, R.G., Lawson, A.R., Ma, J., Neale, G., Ogunkolade, B.W., Jones, T.A., Aarum, J., Dalton, J., Bailey, S., et al. (2009). Activation of the ERK/MAPK pathway: a signature genetic defect in posterior fossa pilocytic astrocytomas. J. Pathol. 218, 172–181.

Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors 269

Galvao, R.P., Kasina, A., McNeill, R.S., Harbin, J.E., Foreman, O., Verhaak, R.G., Nishiyama, A., Miller, C.R., and Zong, H. (2014). Transformation of quiescent adult oligodendrocyte precursor cells into malignant glioma through a multistep reactivation process. Proc. Natl. Acad. Sci. USA 111, E4214–E4223. Hasegawa, H., Ashigaki, S., Takamatsu, M., Suzuki-Migishima, R., Ohbayashi, N., Itoh, N., Takada, S., and Tanabe, Y. (2004). Laminar patterning in the developing neocortex by temporally coordinated fibroblast growth factor signaling. J. Neurosci. 24, 8711–8719. Hegedus, B., Dasgupta, B., Shin, J.E., Emnett, R.J., Hart-Mahon, E.K., Elghazi, L., Bernal-Mizrachi, E., and Gutmann, D.H. (2007). Neurofibromatosis-1 regulates neuronal and glial cell differentiation from neuroglial progenitors in vivo by both cAMP- and Ras-dependent mechanisms. Cell Stem Cell 1, 443–457. Janzarik, W.G., Kratz, C.P., Loges, N.T., Olbrich, H., Klein, C., Scha¨fer, T., Scheurlen, W., Roggendorf, W., Weiller, C., Niemeyer, C., et al. (2007). Further evidence for a somatic KRAS mutation in a pilocytic astrocytoma. Neuropediatrics 38, 61–63. Kwon, C.-H., Zhao, D., Chen, J., Alcantara, S., Li, Y., Burns, D.K., Mason, R.P., Lee, E.Y.-H.P., Wu, H., and Parada, L.F. (2008). Pten haploinsufficiency accelerates formation of high-grade astrocytomas. Cancer Res. 68, 3286–3294. Lee, Y., Yeh, T.H., Emnett, R.J., White, C.R., and Gutmann, D.H. (2010). Neurofibromatosis-1 regulates neuroglial progenitor proliferation and glial differentiation in a brain region-specific manner. Genes Dev. 24, 2317–2329. Lee, Y., Gianino, S.M., and Gutmann, D.H. (2012). Innate neural stem cell heterogeneity determines the patterning of glioma formation in children. Cancer Cell 22, 131–138. Lei, L., and Canoll, P. (2011). MADM gives new insights into gliomagenesis. J. Mol. Cell Biol. 3, 273–275. Lei, L., Sonabend, A.M., Guarnieri, P., Soderquist, C., Ludwig, T., Rosenfeld, S., Bruce, J.N., and Canoll, P. (2011). Glioblastoma models reveal the connection between adult glial progenitors and the proneural phenotype. PLoS ONE 6, e20041.

cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science 344, 1396–1401. Paugh, B.S., Zhu, X., Qu, C., Endersby, R., Diaz, A.K., Zhang, J., Bax, D.A., Carvalho, D., Reis, R.M., Onar-Thomas, A., et al. (2013). Novel oncogenic PDGFRA mutations in pediatric high-grade gliomas. Cancer Res. 73, 6219– 6229. Persson, A.I., Petritsch, C., Swartling, F.J., Itsara, M., Sim, F.J., Auvergne, R., Goldenberg, D.D., Vandenberg, S.R., Nguyen, K.N., Yakovenko, S., et al. (2010). Non-stem cell origin for oligodendroglioma. Cancer Cell 18, 669–682. Premsrirut, P.K., Dow, L.E., Kim, S.Y., Camiolo, M., Malone, C.D., Miething, C., Scuoppo, C., Zuber, J., Dickins, R.A., Kogan, S.C., et al. (2011). A rapid and scalable system for studying gene function in mice using conditional RNA interference. Cell 145, 145–158. Quant, E.C., Drappatz, J., Wen, P.Y., and Norden, A.D. (2010). Recurrent highgrade glioma. Curr. Treat. Options Neurol. 12, 321–333. Rakic, P. (2003). Elusive radial glial cells: historical and evolutionary perspective. Glia 43, 19–32. Rowitch, D.H., and Kriegstein, A.R. (2010). Developmental genetics of vertebrate glial-cell specification. Nature 468, 214–222. Schwartzentruber, J., Korshunov, A., Liu, X.Y., Jones, D.T., Pfaff, E., Jacob, K., Sturm, D., Fontebasso, A.M., Quang, D.A., To¨njes, M., et al. (2012). Driver mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature 482, 226–231. Sharma, M.K., Zehnbauer, B.A., Watson, M.A., and Gutmann, D.H. (2005). RAS pathway activation and an oncogenic RAS mutation in sporadic pilocytic astrocytoma. Neurology 65, 1335–1336. Song, Y., Zhang, Q., Kutlu, B., Difilippantonio, S., Bash, R., Gilbert, D., Yin, C., O’Sullivan, T.N., Yang, C., Kozlov, S., et al. (2013). Evolutionary etiology of high-grade astrocytomas. Proc. Natl. Acad. Sci. USA 110, 17933–17938.

Li, X., Newbern, J.M., Wu, Y., Morgan-Smith, M., Zhong, J., Charron, J., and Snider, W.D. (2012). MEK is a key regulator of gliogenesis in the developing brain. Neuron 75, 1035–1050.

Sottoriva, A., Verhoeff, J.J., Borovski, T., McWeeney, S.K., Naumov, L., Medema, J.P., Sloot, P.M., and Vermeulen, L. (2010). Cancer stem cell tumor model reveals invasive morphology and increased phenotypical heterogeneity. Cancer Res. 70, 46–56.

Li, S., Mattar, P., Dixit, R., Lawn, S.O., Wilkinson, G., Kinch, C., Eisenstat, D., Kurrasch, D.M., Chan, J.A., and Schuurmans, C. (2014). RAS/ERK signaling controls proneural genetic programs in cortical development and gliomagenesis. J. Neurosci. 34, 2169–2190.

Sottoriva, A., Spiteri, I., Piccirillo, S.G., Touloumis, A., Collins, V.P., Marioni, J.C., Curtis, C., Watts, C., and Tavare´, S. (2013). Intratumor heterogeneity in human glioblastoma reflects cancer evolutionary dynamics. Proc. Natl. Acad. Sci. USA 110, 4009–4014.

Liu, C., Sage, J.C., Miller, M.R., Verhaak, R.G., Hippenmeyer, S., Vogel, H., Foreman, O., Bronson, R.T., Nishiyama, A., Luo, L., and Zong, H. (2011). Mosaic analysis with double markers reveals tumor cell of origin in glioma. Cell 146, 209–221.

Stancik, E.K., Navarro-Quiroga, I., Sellke, R., and Haydar, T.F. (2010). Heterogeneity in ventricular zone neural precursors contributes to neuronal fate diversity in the postnatal neocortex. J. Neurosci. 30, 7028–7036.

Matsuda, T., and Cepko, C.L. (2007). Controlled expression of transgenes introduced by in vivo electroporation. Proc. Natl. Acad. Sci. USA 104, 1027– 1032. McCutcheon, S.C., Jones, K., Cumming, S.A., Kemp, R., Ireland-Zecchini, H., Saunders, J.C., Houghton, C.A., Howard, L.A., and Winton, D.J. (2010). Characterization of a heat resistant beta-glucosidase as a new reporter in cells and mice. BMC Biol. 8, 89.

Stowe, I.B., Mercado, E.L., Stowe, T.R., Bell, E.L., Oses-Prieto, J.A., Herna´ndez, H., Burlingame, A.L., and McCormick, F. (2012). A shared molecular mechanism underlies the human rasopathies Legius syndrome and Neurofibromatosis-1. Genes Dev. 26, 1421–1426.

Ostrom, Q.T., Gittleman, H., Liao, P., Rouse, C., Chen, Y., Dowling, J., Wolinsky, Y., Kruchko, C., and Barnholtz-Sloan, J. (2014). CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2007-2011. Neuro-oncol. 16 (4), iv1–iv63.

Stupp, R., Hegi, M.E., Mason, W.P., van den Bent, M.J., Taphoorn, M.J., Janzer, R.C., Ludwin, S.K., Allgeier, A., Fisher, B., Belanger, K., et al.; European Organisation for Research and Treatment of Cancer Brain Tumour and Radiation Oncology Groups; National Cancer Institute of Canada Clinical Trials Group (2009). Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol. 10, 459–466.

Ozawa, T., Brennan, C.W., Wang, L., Squatrito, M., Sasayama, T., Nakada, M., Huse, J.T., Pedraza, A., Utsuki, S., Yasui, Y., et al. (2010). PDGFRA gene rearrangements are frequent genetic events in PDGFRA-amplified glioblastomas. Genes Dev. 24, 2205–2218.

Sugiarto, S., Persson, A.I., Munoz, E.G., Waldhuber, M., Lamagna, C., Andor, N., Hanecker, P., Ayers-Ringler, J., Phillips, J., Siu, J., et al. (2011). Asymmetry-defective oligodendrocyte progenitors are glioma precursors. Cancer Cell 20, 328–340.

Ozawa, T., Riester, M., Cheng, Y.K., Huse, J.T., Squatrito, M., Helmy, K., Charles, N., Michor, F., and Holland, E.C. (2014). Most human non-GCIMP glioblastoma subtypes evolve from a common proneural-like precursor glioma. Cancer Cell 26, 288–300.

Verhaak, R.G., Hoadley, K.A., Purdom, E., Wang, V., Qi, Y., Wilkerson, M.D., Miller, C.R., Ding, L., Golub, T., Mesirov, J.P., et al.; Cancer Genome Atlas Research Network (2010). Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 17, 98–110.

Patel, A.P., Tirosh, I., Trombetta, J.J., Shalek, A.K., Gillespie, S.M., Wakimoto, H., Cahill, D.P., Nahed, B.V., Curry, W.T., Martuza, R.L., et al. (2014). Single-

270 Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors

Visvader, J.E. (2011). Cells of origin in cancer. Nature 469, 314–322.

Wang, Y., Yang, J., Zheng, H., Tomasek, G.J., Zhang, P., McKeever, P.E., Lee, E.Y.H.P., and Zhu, Y. (2009). Expression of mutant p53 proteins implicates a lineage relationship between neural stem cells and malignant astrocytic glioma in a murine model. Cancer Cell 15, 514–526. Wang, Y., Kim, E., Wang, X., Novitch, B.G., Yoshikawa, K., Chang, L.S., and Zhu, Y. (2012). ERK inhibition rescues defects in fate specification of Nf1-deficient neural progenitors and brain abnormalities. Cell 150, 816–830. Yamaoka, K., Mishima, K., Nagashima, Y., Asai, A., Sanai, Y., and Kirino, T. (2000). Expression of galectin-1 mRNA correlates with the malignant potential

of human gliomas and expression of antisense galectin-1 inhibits the growth of 9 glioma cells. J. Neurosci. Res. 59, 722–730. Zhu, Y., Romero, M.I., Ghosh, P., Ye, Z., Charnay, P., Rushing, E.J., Marth, J.D., and Parada, L.F. (2001). Ablation of NF1 function in neurons induces abnormal development of cerebral cortex and reactive gliosis in the brain. Genes Dev. 15, 859–876. Zhu, Y., Guignard, F., Zhao, D., Liu, L., Burns, D.K., Mason, R.P., Messing, A., and Parada, L.F. (2005). Early inactivation of p53 tumor suppressor gene cooperating with NF1 loss induces malignant astrocytoma. Cancer Cell 8, 119–130.

Cell Reports 12, 258–271, July 14, 2015 ª2015 The Authors 271