While medulloblastoma, a pediatric tumor of the cerebellum, is characterized by aberrations in developmental pathways, the majority of genetic determinants remain unknown. An unbiased Sleeping Beauty transposon screen revealed MyoD as a putative medulloblastoma tumor suppressor. This was unexpected, as MyoD is a muscle differentiation factor and not previously known to be expressed in cerebellum or medulloblastoma. In response to deletion of one allele of MyoD, two other Sonic hedgehog-driven mouse medulloblastoma models showed accelerated tumor formation and death, confirming MyoD as a tumor suppressor in these models. In normal cerebellum, MyoD was expressed in the proliferating granule neuron progenitors that are thought to be precursors to medulloblastoma. Similar to some other tumor suppressors that are induced in cancer, MyoD was expressed in proliferating medulloblastoma cells in three mouse models and in human medulloblastoma cases. This suggests that although expression of MyoD in a proliferating tumor is insufficient to prevent tumor progression, its expression in the cerebellum hinders medulloblastoma genesis. Cancer Res; 73(22); 6828–37. ©2013 AACR.
Brain tumors are among the leading cause of childhood cancer-related deaths, and medulloblastoma is the most common pediatric brain malignancy with largely undetermined molecular pathogenesis. On the basis of molecular signatures, medulloblastomas are broadly categorized into four main subgroups: SHH-driven, WNT-driven, and the poorly characterized Group 3 and Group 4 variants (1). However, the vast majority of genetic drivers of this highly heterogeneous cancer remain unknown.
In rare instances, medulloblastomas can show evidence of differentiation along non-neuronal lineages as evidenced by melanin production or expression of muscle markers (2, 3). The rare variant of medulloblastoma that shows some microscopic features resembling muscle are classified by the World Health Organization as “medullomyoblastoma” (4). Diagnosis of these tumors is typically made by immunohistochemical staining for myogenic markers like fast myosin, desmin, and myoglobin (5). Transcription factors that drive myogenic differentiation in medulloblastoma have not been reported previously.
MyoD, a basic-helix-loop transcription factor, is a critical lineage-restricted master regulator of skeletal muscle development (6, 7). Exogenous expression of MyoD is sufficient to drive non-muscle cells (e.g., fibroblasts, chondroblasts, and others) into the skeletal muscle lineage (8, 9). MyoD function or expression during normal cerebellar development or tumorigenesis, however, remains unknown.
Findings in a variety of cancers implicate MyoD as a possible tumor suppressor. MyoD is epigenetically silenced in solid tumors, including prostate and colon cancer (10, 11) as well as during immortalization of cell lines (12, 13), yet the functional significance of this has not been elucidated. MyoD has not previously been considered as a possible tumor suppressor in medulloblastoma.
Here, we identify MyoD as a candidate medulloblastoma tumor suppressor in an unbiased Sleeping Beauty transposon-based in vivo screen. We confirm that single allele loss of MyoD is sufficient to accelerate Sonic hedgehog (Shh)-driven medulloblastoma genesis and that the chromosomal region that expresses MyoD is deleted in some human medulloblastomas. We show that MyoD is expressed in normal cerebellar development in the cells that are thought to be precursors of Shh-driven medulloblastomas. Our study of MyoD as a novel tumor suppressor in medulloblastoma adds a new dimension to the functional versatility of this lineage-restricted muscle determinant while providing a unique insight into the critical regulation of gene expression in medulloblastoma.
Materials and Methods
Smo transgenic and MyoD+/− mouse lines
The ND2:SmoA1 (SmoA1), ND2:SmoA2 (SmoA2) transgenic mouse lines, PtchF/F Math1-Cre conditional knockout and MyoD+/− mice and genotyping protocols have been described previously (14–17).
Human tissue samples
Collection and use of human tissue samples were approved by the Institutional Review Boards of each institution.
Histopathology, immunohistochemistry, and immunofluorescence
Mice were euthanized using CO2 inhalation and tissue snap-frozen for RNA studies or fixed in 10% formalin for histologic examination. Formalin-fixed paraffin-embedded tissues were cut into 4 μm sections. For immunohistochemistry, anti-MyoD (5.8A, BD Biosciences 1:200) followed by anti-mouse Fab frag-ME kit/CSA detection kit and anti-Ki67 (Novocastra, 1:200) were used. Data were confirmed using additional MyoD antibodies (rat Active motif 1:75; rabbit Santa Cruz Biotechnology M-318). Slides were developed using DAB Plus reagent followed by DAKO Hematoxylin counterstain (DAKO). For human MYOD detection, anti-MYOD (Novocastra, 1:80) and (Epitomics, 1:500) were used. For immunofluorescence assays, the following antibodies and protocols were used: MyoD—rat anti-MyoD (Active Motif 1:75) followed by goat anti-Rat HRP (1:500) and CSAII Amplification Reagent with fluorescein isothiocynate. This was confirmed by mouse anti-MyoD (5.8A, BD Biosciences 1:200), followed by anti-mouse Fab-fragment ME-kit, secondary CSA-SA Alexa 350 polymer (Molecular probes, Invitrogen). Antigen-retrieval was performed using the Biocare Rodent Decloaker system. Ki67—rabbit anti-Ki67 (Novocastra, 1:100) followed by goat anti-rabbit Alexa 647 (Molecular Probes, Invitrogen) pseudocolored red in image. Math1—rabbit anti-Math1 (LS Bio, 1:50) followed by goat anti-rabbit Alexa 647 (Molecular Probes, Invitrogen) pseudocolored red in image. NeuN—mouse anti-NeuN (Millipore/Chemicon, 1:75) followed by unconjugated Rabbit anti-mouse with ME kit and goat anti-rabbit Alexa 647 (Molecular Probes, Invitrogen) pseudocolored red in image. 4′, 6-diamidino-2-phenylindole (DAPI) was used as nuclear counterstain.
Images were acquired using the following methods: (i) transmitted light color images of stained tissue sections were acquired on a Nikon E800 microscope fitted with a Nikon 10/0.45 or 20/0.75 Plan Apo objective and Photometrics Coolsnap cf color CCD camera; (ii) 3-D stacks of optical sections were acquired on an Applied Precision Deltavision microscope fitted with an Olympus 100/1.35 UPlan Apo oil immersion objective, and a Photometrics Coolsnap HQ CCD camera. The image stacks were deconvolved using the manufacturer's SoftWorx software; (iii) tissue sections were imaged with a Aperio ScanScope FL slide scanner or 3D Histech's Panoramic 250 Flash whole slide scanner using a 20× objective. Minimal Image adjustments in accordance with journal regulations were made using Image J or Adobe Photoshop CS5.
Image processing method
Regions of interest (ROI; 1024 × 1024) were manually extracted using ImageScope software v184.108.40.2062 from scans acquired with an Aperio ScanScope FL slide scanner using a 20× objective. Cell abundance and protein expression measurements of all DAPI-positive cells costained with anti-Ki67 and anti-MyoD were quantified at the single cell level using a classifier rule set generated with CellProfiler and CellProfiler Analyst (version r11710; refs. 18, 19). Cumulative distribution functions of the classification rule versus fractional relative cell frequency were generated in GraphPad Prism 6.02 to assess protein expression level changes.
RNA was extracted using Qiagen RNeasy mini kit from whole tumor lysates (n = 3/group) and the Agilent 2100 Bioanalyzer Expert was used for quality assessment. Samples were processed at the FHCRC Genomics Shared Resource according to Illumina standard protocols. Array analysis was carried out using the Illumina MouseWG-6 v2.0 Expression BeadChip Kit. Microarray data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO series accession number GSE51219. Data were processed using the Bioconductor package “limma” (20) and quantile normalized using the lumi bioconductor package (21–24). A probe was determined to being significantly differentially expressed if: |logFC| >0.585 and adjacent P < 0.05.
Quantitative reverse transcription PCR
RNA was isolated using miRNAeasy Kit (Invitrogen), DNase (Qiagen) treated, and converted to cDNA using High Capacity Reverse Transcription kit (Applied Biosystems). Reactions were set up using ABI SYBR green or TaqMan Master Mix and run on an ABI 7900HT Fast Real-Time PCR System. TaqMan Gene Expression Assays were used for mouse MyoD, b2m, Human MYOD, PPIA. For SYBR (Invitrogen) assays, primers (Supplementary Document S1) were designed using Primer3 software (25). Data were analyzed using SDS 2.3 software. All conditions were run in triplicates and normalized to b2m or Ppia/PPIA endogenous controls.
Western blot analysis
Protein lysates were prepared using RIPA Buffer (Millipore) with Halt Protease Inhibitor Cocktail (Pierce), and Phosphatase Inhibitor Cocktails (Calbiochem/Sigma). Twenty-five micrograms of protein from each sample were subject to SDS-PAGE using NuPAGE Novex Bis-Tris gels, transferred to nitrocellulose membranes using X-Cell SureLock Mini cell (Invitrogen), probed with primary and corresponding secondary antibodies (Supplementary Document S1). Proteins were detected using ECL chemiluminiscent substrate (Pierce).
Molecular classification of human medulloblastomas
The molecular classification of medulloblastoma tumors using a nanoString-based assay was described previously (26). Briefly, the RNA expressions of markers were measured using a nanoString assay. The expression values were log-transformed, batch-corrected, normalized to endogenous controls, and used as features for class prediction using the Prediction Analysis for Microarrays (PAM; ref. 27) algorithm, as implemented in the pamr package (v 1.51). The class predictions were then filtered using predefined confidence score thresholds for bona fide predictions.
For the analysis of quantitative reverse transcription (qRT)-PCR data, statistical significances of differences between means from two groups were tested using two-tailed Student t test. Survival curves were plotted using Kaplan–Meier method (28) and compared using two-sided log-rank test (29). Statistical analyses were performed in R statistical systems (http://www.r-project.org). Survival analyses used animal death times as events and mice that were still alive at the time of analysis were censored. A nonparametric Kolmogorov–Smirnov statistical test was performed to determine whether differences in MyoD single-cell expression measurements from each genotype (cumulative distribution functions) are statistically significant. The level of significance for all tests was 0.05 (α)
Genomic loss of MYOD is observed in medulloblastoma
The Sleeping Beauty Transposon system is an unbiased, in vivo genetic tool allowing identification of oncogenes and tumor suppressor genes through random integration and clonal expansion in a model of medulloblastoma (30). Using this system, MyoD was identified as a gene-centric common insertion site (gCIS; Fig. 1A). The targeting of MyoD by loss-of-function insertions suggested a selective pressure to reduce MyoD expression.
Further to this finding, we investigated whether a similar phenomenon occurred in human medulloblastomas. While no mutations were observed in MYOD across a cohort of previously sequenced tumors (0/310; refs. 31–34), copy number analysis revealed hemizygous deletion of the 11p arm encompassing the MYOD genomic loci (11p15.1) in 6% (47/827) of medulloblastomas (Fig. 1B). This cytogenetic event was observed in 2 of 76 WNT tumors, 3 of 266 SHH tumors, 7 of 168 Group 3 tumors, and more enriched in the highly aggressive Group 4 tumors (35/317).
Loss of MyoD accelerates tumorigenesis in mouse models of medulloblastoma
Our lab previously generated and characterized two mouse models of medulloblastoma (14, 15, 35). To directly assess whether MyoD reduction functionally contributed to medulloblastoma genesis in vivo, we crossed these two different medulloblastoma mouse models, SmoA1 or SmoA2 mice, to MyoD+/− mice (17) to obtain mice with reduced MyoD. Interestingly, SmoA1 or SmoA2 mice homozygous null for MyoD were born in sub-Mendelian ratios with compromised general health and the majority died within a few weeks of postnatal life. The cause of death is unknown but unrelated to tumorigenesis (Supplementary Fig. S1). Heterozygous reduction in MyoD expression led to significant acceleration of tumor formation in both MyoD+/−; SmoA1 and MyoD+/−; SmoA2 mice as compared with MyoD+/+; SmoA1 and MyoD+/+; SmoA2 mice, respectively (Fig. 2A and B). In stage-matched tumors, we observed a trend towards a higher proliferative index in the faster onset MyoD+/−; SmoA2 tumors compared with MyoD+/+; SmoA2 (Supplementary Fig. S2). We validated the reduction of MyoD in the SmoA2 MyoD+/− mice at the mRNA and protein level, confirming our genetic model (n = 5, P < 0.05; Fig. 2C and Supplementary Fig. S2). IHC analysis suggests that the reduction in MyoD in the MyoD+/−; SmoA2 tumors compared to the MyoD+/+;SmoA2 group, stems from possibly both an overall reduction in the number of MyoD+ cells as well as the level of expression at a cellular level (Supplementary Fig. S2). These functional mouse genetic experiments, together with the Sleeping Beauty screen, confirmed MyoD to be a tumor suppressor in medulloblastoma genesis.
MyoD is expressed in the developing mouse cerebellum and hyperproliferative cells of SmoA2 mice
The finding that the MyoD locus was a gene-centric common insertion site in the Sleeping Beauty screen and that reduced MyoD accelerated medulloblastomas in both the SmoA1 and SmoA2 models clearly demonstrated a functional role of MyoD as a medulloblastoma tumor suppressor in vivo. Because MyoD expression has not previously been reported in the developing cerebellum or in medulloblastoma, we sought to carefully characterize MyoD expression in both normal cerebella and in mouse medulloblastomas.
We first assessed MyoD expression in the brains of the SmoA2 mouse medulloblastoma model. Immunohistochemical (IHC) analysis revealed MyoD expression in the expanded external granule layer (EGL) and hyperproliferative lesions within the interior of the SmoA2 developing cerebellum (Fig. 3A). MyoD is expressed in the normal developing cerebellum as well, restricted to the Ki67+ outermost, undifferentiated EGL from postnatal day (P) 0 to P15 in mice (Fig. 1B). A comparison of WT versus the SmoA2 cerebella revealed pronounced differences in MyoD expression (Fig. 3A and B). Interestingly, MyoD expression is fully silenced in the mature normal cerebellum (P30) yet persists in the undifferentiated proliferating cells of the SmoA2 cerebellum through adulthood. While our focus was on the role of MyoD in the context of medulloblastoma genesis, its expression in normal development prompted us to evaluate MyoD−/− cerebella at P5, which did not show any apparent abnormalities.
MyoD is expressed in proliferating medulloblastoma cells in three SHH medulloblastoma mouse models
While some tumor suppressors are absent or diminished in cancer, others are upregulated in proliferating cancer cells in an unsuccessful attempt to regulate proliferation or differentiation. In our mouse medulloblastoma models, the latter seems to be the case. We found high levels of MyoD in three independent SHH-subgroup mouse models—SmoA1 (14, 15), SmoA2 (35), and PtchF/F Math1-Cre conditional knockout (Ptch cko) tumors (Fig. 4A; ref. 16). Importantly, MyoD mRNA and protein were not expressed in regions of cerebellar dysplasia, that commonly occur in the adult SmoA2 mice, but rather only in neoplastic tissue (Fig. 4B and C). The lack of expression of MyoD in SmoA2-expressing nontumor cells suggests that MyoD is not a direct target of the conditionally active Smoothened protein, but its expression is maintained as a consequence of the cellular transformation process. To confirm our hypothesis, we examined the relationship between proliferation (via Ki67) and MyoD expression using immunofluorescence. MyoD expression was observed exclusively in Ki67+ cells in tumors derived from SmoA1, SmoA2, and Ptch cko mice (Fig. 4D and Supplementary Fig. S3). The proportion of MyoD+ cells appeared to be highest in the outer region of the tumors known to develop from the hyperproliferative EGL, as compared with the inner core (Supplementary Fig. S4). No proliferating (Ki67+) or MyoD+ cells were detected in the WT adult cerebellum. Thus, like P53, TP73, pRb2/p130, ARF, and p16INK4a in selected types of cancer (36–41), MyoD appears to be a tumor suppressor that is expressed in response to oncogenic signaling, but insufficient as a single protein to prevent medulloblastoma progression.
MyoD colocalizes with Math1 in granule neuronal precursors
Granule neuron precursors (GNP) are a transiently proliferative population of cells that form the EGL and are considered to be the cell-of-origin for Shh medulloblastomas (42). To characterize the relationship of MyoD to normal cerebellar development and Shh-driven medulloblastoma genesis, we carried out dual immunofluorescence for MyoD with Math1, a GNP marker and NeuN, a marker of differentiated granule neurons. Our results demonstrate that MyoD is observed exclusively in the Math1 compartment and mutually exclusive with NeuN (Fig. 5A and B). Thus, it appears that MyoD is expressed in proliferating normal GNPs during the final stage before cell-cycle arrest and differentiation. It is possible, but unproven that Shh induction of MyoD in these normal cells and in proliferating Shh-induced medulloblastoma cells is part of an oncofetal development program that is aberrantly activated in medulloblastoma.
The tumor suppressor function of MyoD is not executed through the canonical myogenic differentiation program
To determine whether the canonical myogenic differentiation program was involved in increased tumorigenicity following loss of MyoD, we performed gene expression analysis comparing MyoD+/+; SmoA2 with MyoD+/−;SmoA2 tumors. Surprisingly, no single gene candidate passed statistical significance between the two genotypes. To identify subtle transcriptional differences possibly beyond the detection limit of an array-based approach, we carried out qRT-PCR analyses on canonical genes involved in the MyoD-mediated skeletal muscle differentiation program. Myf5, Myog, Desmin, Cdh15 showed no difference between the two groups. The only gene involved in myogenic differentiation that differed between the two genotypes was Id3, which was reduced by approximately two-fold (Supplementary Fig. S5). Notably, Id3 is involved in many differentiation programs, including hematopoeisis and neurogenesis. These tumors did not show any histologic evidence of muscle differentiation and were negative for myogenic markers, desmin and myoglobin by immunohistochemistry as well. Taken together, these data demonstrate that MyoD does not utilize the canonical myogenic differentiation program as the basis of tumor suppression.
MyoD is expressed in a subset of human medulloblastomas
Medulloblastomas with myogenic differentiation (medullomyoblastomas) have been described on the basis of histopathologic criteria. To determine whether MyoD was expressed in human medulloblastomas and learn whether expression correlated with molecular subtype, we performed qRT-PCR analysis of MYOD expression in a cohort of primary medulloblastomas (n = 22) after establishing molecular subgroup affiliations. MYOD expression was detected in 36% (8/22) of medulloblastomas across the four molecular subgroups, most prominent in Group 3 medulloblastomas. FH-MB01, the tumor with the highest level of MYOD expression (Fig. 6A), was confirmed to express nuclear MYOD protein by IHC analysis albeit sparse and heterogeneous throughout the tumor section (Fig. 6B). There was no histologic evidence of differentiated skeletal muscle cells, consistent with the patient's pathology report and our findings that MyoD did not appear to execute a canonical muscle differentiation program in brain or brain tumors. Subgroup-specific expression patterns of MYOD were confirmed across an independent and non-overlapping validation series (n = 103; Supplementary Fig. S6).
Medulloblastoma is a pediatric malignancy characterized by aberrant developmental pathways. Here, we present the first report of MyoD expression in GNPs during normal cerebellar development restricted to the outermost proliferative and undifferentiated EGL in mice. While quiescent in the normal adult cerebella (i.e., beyond P15 as the GNPs in the EGL exit the cell cycle and migrate inwards to form the differentiated internal granular layer (IGL)), MyoD expression is retained in the neoplastic expansion of the EGL in mouse models of medulloblastoma. We demonstrate through genetic studies with multiple lines of mice that loss of one allele of MyoD significantly accelerates tumorigenesis in vivo establishing its role as a tumor suppressor gene in medulloblastoma.
The absence of detectable MyoD expression in nontumor cells from SmoA1 or SmoA2 mice that differentiate into mature neurons, suggests that MyoD is not a direct target of conditionally activated Smoothened protein; rather the developmental expression of MyoD occurs as a result of neoplastic transformation.
MyoD expression in Ki67+ tumor cells is consistent with other examples of antiproliferative networks being activated by oncogenic signals such as p53 induction by oncogenes E1A/Myc/E2F or ARF activation by Ras/TSC/mTORC1 signaling as responses to counter abnormal hyperproliferative cues (36, 39). In cancer, overexpression of tumor suppressors have been described as inherent protective responses—similar examples include TP73 overexpression in medulloblastoma associated with improved survival outcomes (37); pRb2/p130 in hepatocellular carcinoma (38); p16INK4a induction in HPV-induced tumors as an attempt to arrest proliferation (40); wild-type p53 overexpression in human glioma (41). The normal developmental expression of MyoD in proliferative GNPs only during a period of intense growth suggests an important role in regulation of cell proliferation before neuronal maturation.
Ki67 is expressed in all active phases of the cell cycle (absent in G0; ref. 43), whereas MyoD has been shown to undergo a biphasic expression pattern in proliferating myoblasts, peaking in the middle of G1 and at the end of G2 (44). Our results therefore suggest that the Ki67+/MyoD+ medulloblastoma cells may be in G1 or G2. Since cell-cycle exit and induction of differentiation can occur in G1 or in late G2/mitosis, a strong expression of MyoD may define a temporal window in which differentiation can occur (44). Moreover, MyoD has also been shown to induce cell-cycle arrest independent of differentiation in normal as well as transformed cells (45, 46). Therefore, functions of MyoD in cell-cycle regulation (47–49) provide important leads for further understanding the mechanism of action of MyoD as a tumor suppressor.
Pro-proliferative molecules like Id3 are known to antagonize the function of MyoD in the myogenic program (6) as well as play key neurodevelopmental functions such as inhibition of differentiation during neurogenesis (50). Our observation of elevated Id3 following the loss of MyoD together with the known contributions of Id genes to tumor growth, metastasis, and vascularization in cancer (50, 51) suggest a possible connection of Id3 to the MyoD-regulated tumor suppressor network in medulloblastoma. A recent study by Bai and colleagues shows oncogene Otx2 to repress the MyoD enhancer in medulloblastoma cells whereby the loss of Otx2 causes activation of the myogenic differentiation program in vitro (4). These results directly support our finding of a tumor suppressor role of MyoD in medulloblastoma in vivo. As MyoD is a functionally versatile molecule that binds thousands of genes to exert its complex genetic and epigenetic regulatory functions, further investigation of the downstream targets in the MyoD-regulated tumor suppressor network remains an important future step.
The role of MyoD in tumor suppression and in mammalian brain development is novel and fascinating, with a highly complex underlying biology. On the basis of our current findings, we propose the following model for MyoD activity in cerebellar development and medulloblastoma genesis: During development, MyoD is transiently expressed in the proliferating GNPs potentially playing a key role in the maturation of the normal cerebellum through mechanisms unrelated to its myogenic targets. While MyoD is normally silenced as GNPs differentiate into mature neurons in the normal cerebellum, medulloblastoma tumor cells either retain MyoD or reinduce expression of MyoD under the influence of oncogenes such as the constitutively active Shh pathway. While aberrant expression of MyoD in proliferating medulloblastoma cells is insufficient to prevent tumor formation, reduction of endogenous MyoD expression in developing cerebellum is sufficient to accelerate tumor formation. This is reminiscent of multiple other tumor suppressors that happen to be activated by oncogenic pathways, yet fail to fully impede tumor progression (36–41).
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: J. Dey, A.M. Dubuc, M.D. Taylor, J.M. Olson
Development of methodology: J. Dey, J.M. Olson
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Dey, K.D. Pedro, M.D. Taylor, J.M. Olson
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Dey, A.M. Dubuc, D. Thirstrup, B.H. Mecham, P. Northcott, D.J. Shih, S.J. Tapscott, M.L. LeBlanc, M.D. Taylor, J.M. Olson
Writing, review, and/or revision of the manuscript: J. Dey, A.M. Dubuc, M.D. Taylor, J.M. Olson
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Dey, K.D. Pedro
Study supervision: J. Dey, J.M. Olson
Provided supporting data: X. Wu
Human tissue was obtained from the Brain Tumor Tissue Bank, London, Ontario funded by The Brain Tumor Foundation of Canada, NICHD Brain Tissue Bank for Developmental Disorders at the University of Maryland, Baltimore, MD and Seattle Children's Hospital Cancer Biorepository, Seattle, WA, with appropriate approval from Institutional Review Boards. The authors thank Dr. Julie-Randolph Habecker, Kimberly Melton and the FHCRC Experimental Histopathology Shared Resource, Ying-tzang Tien at University of Washington (UW) Histopathology Lab, Dr. Paul Swanson and Farinaz Shokri at UW Medicine Anatomic Pathology Lab, Dr. Lauren Snider, Barbara Pullar for colony management and genotyping, Dr. Martine Roussel, Dr. Daisuke Kawaichi for reagents, Dr. Julio Vazquez and David L. McDonald at FHCRC Scientific Imaging, Dr. Jeff Delrow and Ryan Basom at FHCRC Genomics Shared Resource, Betty Lu and Marc Remke for Nanostring analysis, Drs. Richard Klinghoffer, Michelle Cook Sangar, Michelle Lee, Sally Ditzler, Ilona Tretyak for guidance and technical help.
This work was supported by NIH grants 5R01CA11456705 and 5R01CA112350; NIH NIAMS R01AR045113; The Pediatric Brain Tumor Foundation (M.D. Taylor). J. Dey was supported by Predoctoral Developmental Biology Training Grant 5T32HD007183 from the National Institute of Child Health and Human Development (NICHD), A.M. Dubuc was supported by Canadian Institutes of Health Research (CIHR) Vanier Scholarship, B.H. Mecham was supported by U54CA149237 from the Integrative Cancer Biology Program of the National Cancer Institute.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.