Sonic hedgehog regulates Bmi1 in human medulloblastoma brain ...

Oncogene (2012) 31, 187–199

& 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc

ORIGINAL ARTICLE

Sonic hedgehog regulates Bmi1 in human medulloblastoma brain tumor-initiating cells X Wang1, C Venugopal1, B Manoranjan1, N McFarlane1, E O’Farrell1, S Nolte1, T Gunnarsson2, R Hollenberg2, J Kwiecien3, P Northcott4, MD Taylor4, C Hawkins5 and SK Singh1,2,6,7 1 McMaster Stem Cell and Cancer Research Institute, McMaster University, Hamilton, Ontario, Canada; 2Department of Surgery, McMaster University, Hamilton, Ontario, Canada; 3Department of Pathology and Molecular Medicine, McMaster University, Hamilton, Ontario, Canada; 4Arthur and Sonia Labatt Brain Tumor Research Centre and Department of Surgery, The Hospital for Sick Children and University of Toronto, Toronto, Ontario, Canada; 5Arthur and Sonia Labatt Brain Tumor Research Centre and Department of Pediatric Laboratory Medicine, The Hospital for Sick Children and University of Toronto, Toronto, Ontario, Canada; 6 Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, Ontario, Canada and 7Department of Neuroscience, Faculty of Health Sciences, McMaster University, Hamilton, Ontario, Canada

Bmi1 is a key stem cell regulatory gene implicated in the pathogenesis of many aggressive cancers, including medulloblastoma. Overexpression of Bmi1 promotes cell proliferation and is required for hedgehog (Hh) pathwaydriven tumorigenesis. This study aimed to determine if Sonic hedgehog (Shh) modulates the key stem cell regulatory gene Bmi1 in childhood medulloblastoma brain tumor-initiating cells (BTICs). Although current literature suggests that there is a correlation between Shh pathway genes and Bmi1 expression, it is unclear whether there is indeed a direct regulatory mechanism. To address whether Shh induces expression of Bmi1, stem cell-enriched populations from medulloblastoma cell lines and primary samples were treated with Shh ligand and KAAD-cyclopamine (Shh antagonist). Our data indicate that Bmi1 expression positively correlates with increasing Shh ligand concentrations. Chromatin immunoprecipitation reveals that Gli1 preferentially binds to the Bmi1 promoter, and Bmi1 transcript levels are increased and decreased by Gli1 overexpression and downregulation, respectively. Knockdown experiments of Bmi1 in vitro and in vivo demonstrate that Hh signaling not only drives Bmi1 expression, but a feedback mechanism exists wherein downstream effectors of Bmi1 may, in turn, activate Hh pathway genes. These findings implicate Bmi1 and Hh as mutually indispensable pathways in medulloblastoma BTIC maintenance. Recent molecular characterization of medulloblastoma also reveals that Bmi1 is overexpressed across all subgroups of medulloblastoma, particularly in the most aggressive subtypes. Lastly, despite recent identification of BTIC markers, the molecular characterization of these cell populations remains unclear. In this work, we propose that the BTIC marker CD133 may segregate a cell population with a Hh-receptor phenotype, thus demonstrating a cell–cell interaction between the CD133 þ Hh receptor cells and the CD133 Hh-secreting cells. Correspondence: Dr SK Singh, Department of Surgery, McMaster University Stem Cell and Cancer Research Institute, MDCL 5027, 1200 Main Street West, Hamilton, Ontario, Canada L8N 3Z5. E-mail: [email protected] Received 22 November 2010; revised 4 May 2011; accepted 8 May 2011; published online 20 June 2011

Oncogene (2012) 31, 187–199; doi:10.1038/onc.2011.232; published online 20 June 2011 Keywords: Bmi1; sonic hedgehog; medulloblastoma; brain tumor-initiating cells; CD133; chromatin immunoprecipitation

Introduction Medulloblastoma is the most common malignant pediatric brain tumor, representing 20% of newly diagnosed childhood central nervous system malignancies (Crawford et al., 2007). Despite current multimodality treatment of surgical resection, radiation and chemotherapy, jointly yielding an 80% 5-year survival rate, there remains significant treatment-induced morbidity and long-term clinical sequelae (Gilbertson, 2004; Huse and Holland, 2010). Recent molecular classification of medulloblastoma has not only reconceptualized the heterogeneity that exists within pathological subtypes, but has given context to the role of key developmental signaling pathways in the pathogenesis of this tumor (Pomeroy et al., 2002; Thompson et al., 2006; Kool et al., 2008; Northcott et al., 2010). In normal development, hedgehog (Hh) functions as a morphogen to induce cell identities in the ventral spinal cord, and as a mitogen to drive proliferation of granule neuron precursors (Fuccillo et al., 2006). Medulloblastoma, a primitive neuroectodermal tumor, is thought to arise from immature neural progenitors in the cerebellum (Lee et al., 2005). Genomic alterations in components of the Hh signaling pathway have been identified in up to 25% of human medulloblastomas (Michael et al., 2008), consisting of inactivating mutations of Ptch1 and Sufu and/or activating mutations of Smo (Taylor et al., 2002; Hallahan et al., 2004; Oliver et al., 2005; Mao et al., 2006; Thomas et al., 2009; Huse and Holland, 2010). More recently, dysregulated Hh signaling has been shown to be sufficient to generate

Shh regulates Bmi1 in medulloblastoma BTICs X Wang et al

188

tumors in either neural stem cell (NSC) or precursor cell populations, implicating a more primitive cell of origin (Schuller et al., 2008; Yang et al., 2008). Polycomb protein Bmi1 is a key regulator of hematopoietic, NSC and brain tumor-initiating cell (BTIC) populations. This gene is implicated in the pathogenesis of brain tumors such as glioma (Bruggeman et al., 2007) and medulloblastoma (Leung et al., 2004; Wiederschain et al., 2007; Michael et al., 2008), and is an important epigenetic regulator of fate determination and proliferation in stem cell populations (Valk-Lingbeek et al., 2004; Bracken et al., 2006). Previous reports have suggested that there is a potential link between Hh signaling and Bmi1, alluding to a novel regulatory mechanism whereby an external signaling morphogen interacts with cell-intrinsic epigenetic pathways controlling cell fate programs (Valk-Lingbeek et al., 2004). Here, we propose that the Gli family transcription factors may activate the promoter of Bmi1, thus regulating its expression. In addition to mechanistically demonstrating Bmi1 regulation by Sonic hedgehog (Shh), we investigated the consequences of attenuating Bmi1 expression on Hh signaling. In the present study, we have used both in vitro and human–mouse xenograft models to demonstrate that knockdown of Bmi1 leads to significant inhibition of Hh pathway genes. Lastly, tumorigenic stem cells have been identified in a variety of brain tumors (Singh et al., 2003, 2004), lending support to the cancer stem cell hypothesis. However, the role of Hh and Bmi1 signaling in the regulation of stem cells has not yet been elucidated. We demonstrate that BTIC marker CD133 (Yin et al., 1997; Singh et al., 2003, 2004) may help to identify different populations of cells that secrete or respond to Shh. This understanding in combination with recent molecular classification of medulloblastoma may ultimately provide subgroupspecific targets for therapeutic interventions.

Results Characterization of Daoy medulloblastoma cell line in NSC media reveals a BTIC phenotype Having previously characterized BTIC populations from primary human medulloblastoma (Singh et al., 2003, 2004), we applied this experimental model system to 10 primary pediatric medulloblastomas (see Table 1). We have called these specimens ‘BTIC patient isolates’ to emphasize the fact that these cells are not renewable cell lines, but are cell isolates minimally cultured (24 h to o1 week) under stem cell conditions to select for stem cell populations. Owing to limited primary sample availability, we have supplemented our experimental work on human primary medulloblastoma BTICs with studies of the Daoy medulloblastoma cell line. In order to characterize the Daoy cell line, we cultured the cells in NSC media to enrich for a stem cell-like phenotype. After 3 days in NSC media conditions, Daoy cells were able to form primary spheres. Flow cytometric analysis revealed Oncogene

Table 1 BTIC patient isolates: clinicopathological data Sample ID BT BT BT BT BT BT BT BT BT BT

1 2 3 4 5 6 7 8 9 10

Gender

Age at diagnosis

M M M M M M F M F M

14 9 4 16 2 1 6 9 3 12

Pathological subtype Anaplastic Anaplastic Anaplastic Classic Anaplastic Desmoplastic Classic Anaplastic Classic Classic

Abbreviations: BT, brain tumor; BTIC, brain tumor-initiating cell; F, female; M, male.

that over the course of the enrichment for a stem cell-like phenotype, these cells increased expression of NSC marker CD133 from 3.28% (Figure 1a) to 14.57% (Figure 1b, representative sample shown). Further characterization reveals that these cells are able to form secondary and tertiary spheres following a limiting dilution assay, with a sphere frequency of 21 spheres and 26 spheres per 2000 cells, respectively (Figure 1c). When compared with primary medulloblastoma samples, the Daoy cells enriched for stem-like cells were comparable in terms of sphere formation. Proliferation assay revealed a similar proliferative capacity for both patient samples and the Daoy cell line (Figure 1d). Further molecular characterization of the Daoy cell line, both in our hands (data not shown) and by various other investigators, reveals a global activation of Hh pathway genes including Shh, Gli family and Ptch1, as well as overexpression of Bmi1 (Dahmane et al., 2001; Leung et al., 2004). These gene signatures recapitulate human primary medulloblastoma BTIC expression profiles and thus the Daoy cell line grown in NSC conditions act as a suitable surrogate. Two other human medulloblastoma cell lines, D341 and D283, derived from a cerebellar medulloblastoma and a peritoneal medulloblastoma metastasis, respectively, were also grown in stem cell conditions and subsequently characterized in our laboratory. These cell lines demonstrated less robust and minimally renewable stem cell populations that did not reflect the stem cell profile or properties of self-renewal and proliferation found in primary human medulloblastoma BTICs (Supplementary Figures 1a–d). As only the Daoy medulloblastoma cell line accurately recapitulated the stem cell properties of primary human medulloblastoma BTICs, we utilized the Daoy cell line to generate supportive data in all subsequent experiments. Hh signaling through Gli1 modulates Bmi1 expression Due to the role of Bmi1 in medulloblastoma proliferation, we hypothesized that the Hh pathway may directly regulate Bmi1 expression. We treated primary human medulloblastoma cultures with the Shh soluble ligand and found a corresponding increase in its downstream gene targets, notably Gli1 (Figure 2a and Supplementary


189

Figure 1 Daoy medulloblastoma cell line grown in NSC conditions expresses BTIC marker CD133 and displays similar self-renewal and proliferative capacity as primary medulloblastoma. (a) Daoy cell line grown in ATCC (American Type Culture Collection) adherent conditions exhibits low CD133 expression via flow cytometry (n ¼ 3, representative flow plot shown). (b) Daoy cell line grown in NSC media conditions forms tumor spheres and shows increased CD133 expression when assayed by flow cytometry (n ¼ 3, representative flow plot shown). (c) Self-renewal from primary medulloblastoma samples (n ¼ 8, BT 2–9, see Table 1) is similar when compared with Daoy cell line grown in NSC conditions in both secondary (n ¼ 2) and tertiary sphere formation assays (n ¼ 3). (d) Proliferative capacity of primary medulloblastoma samples (n ¼ 2, BT 6 and BT 7, see Table 1) is comparable with Daoy cell line grown in NSC conditions (n ¼ 2).

Figure 2a). Interestingly, we were able to observe Shhinduced dosage-dependent increase in Bmi1 expression (Supplementary Figure 2b), with a significant change after 5 mg/ml (Figure 2a). To interrogate the mechanism of Bmi1 regulation by Hh, both primary medulloblastoma and Daoy cells were subjected to chromatin immunoprecipitation (ChIP). We used a microChIP technique adapted for rare stem cell populations (Dahl and Collas, 2008) to determine whether the Hh target Gli1, the major effector in cancer

(Clement et al., 2007; Schuller et al., 2008; Read et al., 2009; Ayrault et al., 2010), could bind to the promoter of Bmi1. Both primary medulloblastoma and Daoy cells expressed a relative enrichment of Gli1 (Figure 2b) and Gli2 (Supplementary Figure 3) at the Bmi1 promoter. To determine the functional effects of Gli1 on Bmi1 transcription, we transiently transfected Gli1 overexpression vectors and Gli1 small interfering RNA into Daoy spheres. We observed a significant increase and decrease of Bmi1 transcript levels with Gli1 Oncogene


190

Figure 2 Hh pathway controls Bmi1 expression through promoter activation. (a) Daoy sphere cells treated with Shh ligand show increased Gli1 (n ¼ 2, P ¼ 0.0018) and Bmi1 (n ¼ 2, P ¼ 0.0229) gene expression. (b) ChIP experiment indicates that Gli1 is enriched at the Bmi1 promoter compared with IgG control in both primary medulloblastoma (n ¼ 2, BT 6 and BT 10, P ¼ 0.0348) and Daoy cell line (P ¼ 0.0048). (c) siGli1 treatment of Daoy spheres shows a significant decrease in Gli1 (n ¼ 3, P ¼ 0.0014) and Bmi1 (n ¼ 3, P ¼ 0.0165) transcript levels. Scrambled RNA was used as a control. (d) Gli1 overexpression treatment of Daoy spheres shows a significant increase in Gli1 (n ¼ 3, Po0.001) and Bmi1 (n ¼ 3, P ¼ 0.0157) transcript levels. Empty vector was used as a control. (e) KAAD-cyclopamine treatment of Daoy spheres (n ¼ 2) results in decreased Gli1 expression (P ¼ 0.0102) and unchanged Bmi1 expression at both the transcript level (P ¼ 0.2613) and (f) protein level via flow cytometry (representative sample shown). *#Po0.05; **Po0.01; ***Po0.001; ns, P40.05.

overexpression (Figure 2c) and knockdown (Figure 2d), respectively. These results lend functional relevance to the enrichment of Gli1 at the Bmi1 promoter. Given the capability of Shh to regulate Bmi1, we then asked whether attenuating the Hh pathway would in turn silence Bmi1 expression. We treated Daoy spheres with Hh blocker, KAAD-cyclopamine, and found that despite an expected decrease in downstream Gli1 expression, Bmi1 remains largely unchanged at both the transcript level (Figure 2e) and the protein level (Figure 2f, representative sample shown). This finding suggests that other mechanisms may be at play to sustain Bmi1 expression. To test the effects of Hh blockade on Bmi1 in vivo, we generated xenografts from Daoy medulloblastoma spheres as previously described (Singh et al., 2004). We performed intracranial injections of 106 single cells from Daoy medulloblastoma spheres into NOD-CB17SCID mice, and mice were subsequently treated every other day with either 25 mg/kg of cyclopamine (Toronto Research Chemicals, Toronto, ON, Canada) or vehicle (45% HBC (2-hydropropyl-b-cyclodextrin) in phosphateOncogene

buffered saline; Sigma, Oakville, ON, Canada) until end point. The resulting human tumor xenografts were fixed, embedded in paraffin and stained with Bmi1 and Gli1. We found decreased expression of both Bmi1 and Gli1 in tumors when Hh signaling is blocked by cyclopamine (Supplementary Figure 4a). This finding was supported by further experiments applying flow cytometry and cell sorting to human Daoy sphere xenografts treated with cyclopamine. The majority of GFPhigh Daoy cells expressed Bmi1, representing 78% of the cell population (Supplementary Figure 4b). Mixing experiments were performed with mouse brain cells and human Daoy sphere cells stably transduced with a green fluorescent protein (GFP)expressing lentivector (CS-H1-shRNA-EF-1 a-EGFP, see Materials and methods), to determine the lowest detection threshold for GFPhigh human cells in the mouse brain (Supplementary Figure 4c). We then performed intracranial injection of GFP þ human Daoy spheres into mouse brains, subsequently treating mice with a control vehicle or with cyclopamine. Xenografted mouse brains were harvested, and dissociated to single-cell suspension for flow cytometric quantification and


191

sorting of GFP þ and GFP populations. We observed a significant reduction in Bmi1 þ GFPhigh human tumor cells with cyclopamine treatment (Supplementary Figure 4a, right panel), suggesting that these stem-like cells are specifically targeted by cyclopamine. Bmi1 knockdown reduces cell proliferation and attenuates Hh pathway genes in vitro, and decreases infiltration in vivo We obtained lentiviral constructs CS-H1-BMI1 shRNAEF-1a-EGFP expressing short hairpin RNA (shRNA) targeting Bmi1 and CS-HI Luc shRNA-EF-1a-EGFP expressing the target sequence for Luciferase (as a negative control), as a kind gift from Professor Atsushi Iwama (Chiba University, Japan). We obtained a stable knockdown of Bmi1 transcript and protein in Daoy sphere cultures (Figure 3a). Bmi1 knockdown demonstrated decreased proliferative capacity in vitro (Figure 3b), and a 70–80% decrease in the expression of Hh pathway genes Gli1 (Figure 3c) and Smo (Figure 3d), which may suggest a potential feedback mechanism. To test whether these findings are recapitulated in vivo, we performed intracranial injections of Daoy medulloblastoma spheres stably knocked down for Bmi1 into nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice (n ¼ 19). We found that despite consistent tumor volumes between the Bmi1 knockdown and wild type (Figure 3e), a clear survival benefit was observed when Bmi1 is knocked down (Figure 3f, Po0.007). Histological examination of tumor sections (Figure 3g) revealed that although Daoy-generated medulloblastomas were invasive and infiltrative, Bmi1 knockdown tumors appeared significantly less invasive and were comparatively well circumscribed, likely accounting for the enhanced survival of these animals. These findings suggest that Bmi1 may increase the potential of medulloblastoma BTICs to invade and infiltrate the brain, resulting in poorer survival (Glinsky et al., 2005). CD133 cells express the Hh-secretor phenotype, whereas CD133 þ cells express the Hh-receptor phenotype To explore the expression of Hh pathway components and Bmi1 in different stem cell compartments within medulloblastoma, we sorted primary human medulloblastoma samples for the BTIC marker CD133, which enriches for stem cell populations in Daoy sphere cultures. As CD133 was originally described as the NSC (Uchida et al., 2000) and BTIC (Singh et al., 2004) marker, its ability to exclusively isolate tumor-initiating cells has been debated in the literature, especially in adult glioblastoma. In this heterogeneous tumor, CD133 cells in long-term culture were found to be capable of tumor initiation in mice (Chen et al., 2010), although many recent publications maintain that not only is the CD133 þ fraction enriched in BTIC capacity (Wang et al., 2010; Yan et al., 2011), but the features of ‘stemness’ marked by CD133 are associated with

increasing malignancy and poor patient outcome (Bao et al., 2006; Thon et al., 2008; McCord et al., 2009; Pallini et al., 2010). In medulloblastoma, evidence more uniformly supports the application of CD133 as a BTIC marker, as CD133 þ multipotent NSCs were described in the postnatal cerebellum (Lee et al., 2005), and further studies showed that CD133 enriched for brain tumor-initiating capacity and radioresistance in primary and Daoy medulloblastoma-derived tumors (Fan et al., 2006; Blazek et al., 2007; Annabi et al., 2008, 2010; Pistollato et al., 2010; Yu et al., 2010). In contrast to previous experiments performed on unsorted cells, we found a differential expression of Hh pathway components between the CD133 þ stem-like cells and the CD133 late progenitors or differentiated cells. It appears that CD133 cells show increased expression of Shh (Figure 4a), whereas CD133 þ cells show enhanced expression of Hh receptor genes Smo and Ptch1 (Figure 4b). This may suggest that CD133 cells form a regulatory niche for CD133 þ cells within the tumor or sphere microenvironment. To further explore this possibility in sphere cultures, we treated sorted primary samples with Hh antagonist, KAAD-cyclopamine. We found that in CD133 þ cells, KAAD-cyclopamine induces a significant decrease in expression of Hh pathway genes Gli1 (Figure 4c) and Ptch1 (Figure 4d), whereas in the negative population, there is no apparent response of Hh pathway receptors or downstream targets. Finally, although KAAD-cyclopamine treatment of Daoy spheres exerts no significant effect on Bmi1 transcript levels in unsorted cells (Figures 2c and d), Hh blockade induces a significant decrease in the Bmi1 transcript level (Figure 4e) in sorted CD133 þ cells, with no apparent effect on the majority of CD133 population. Cyclopamine treatment of Daoy spheres also significantly decreases CD133 protein expression (Figure 4f), suggesting that it targets this cellular population. These data suggest that CD133 þ stem-like cells alone possess the Hh pathway receptive components required to modulate Bmi1 expression in cultured primary human medulloblastoma. Bmi1 expression increases with more aggressive medulloblastoma molecular subtypes, and gene transcription profiles on CD133-sorted Daoy cells confirm cell–cell interactions within the tumor niche Bmi1 has been cited as a marker of progression in many cancer models, and it reliably predicts recurrence, poor treatment response, metastatic potential and death from cancer (Glinsky et al., 2005). Functional genomic platforms have reproducibly shown that medulloblastoma can be categorized into four molecular subclasses that are distinct in terms of both prognosis and predicted therapeutic responses (Pomeroy et al., 2002; Thompson et al., 2006; Kool et al., 2008; Northcott et al., 2010). Medulloblastoma subtypes A and B are characterized by exclusive aberrant activation of the Wnt and Shh signaling pathways, respectively. Activation of the Wnt signaling pathway characterizes Oncogene


192

Figure 3 Knockdown of Bmi1 reduces cell proliferation in vitro and disrupts Hh signaling. (a) Knockdown of Bmi1 in Daoy cell line via a lentiviral short hairpin vector reveals a stable knockdown of Bmi1 transcript (n ¼ 3, P ¼ 0.0028) and protein. (b) Daoy cell line knocked down for Bmi1 shows decreased proliferation compared with wild type (n ¼ 2, day 7, P ¼ 0.0156). (c) Knockdown of Bmi1 drastically reduces Gli1 transcript expression (n ¼ 3, P ¼ 0.0001) and (d) Smo transcript expression levels (n ¼ 3, P ¼ 0.0006). (e) Tumor xenografts in NOD/SCIDs between Daoy knocked down for Bmi1 (n ¼ 7) and wild type (n ¼ 12) reveal no significant difference in tumor volume, (f) despite a significant survival advantage when Bmi1 is knocked down (P ¼ 0.007). (g) Histological analysis of xenografts with Bmi1 knockdown (n ¼ 7) reveals tumors with a marked reduction in infiltration when compared with wild type (n ¼ 12). *Po0.05; **Po0.01; ***Po0.001; ns, P40.05.

a distinct molecular subgroup of medulloblastomas associated with a favorable prognosis (Ellison et al., 2005; Fattet et al., 2009). In contrast, medulloblastoma molecular subtypes C and D are associated with Oncogene

metastatic disease and poor patient outcome (Northcott et al., 2010). Northcott et al. (2010) recently demonstrated that medulloblastomas can be reliably assigned to these subgroups through immunohistochemistry,


193

Figure 4 Prospectively sorted CD133 primary medulloblastoma BTICs show CD133 cells express Shh ligand, whereas CD133 þ cells express Hh receptor and are responsive to KAAD-cyclopamine treatment. (a) Primary medulloblastoma (n ¼ 3, BT 3–5) sorted for CD133 reveals that the CD133 þ populations express the Hh-receptor phenotype with low Shh expression and high Ptch1 and Smo expression, (b) whereas CD133 populations express the Hh-secretor phenotype with high Shh expression and low Ptch1 and Smo expression (n ¼ 3, BT 3–5). (c) Sorted primary medulloblastoma treated with KAAD-cyclopamine shows a decrease in Gli1 expression only in the CD133 þ fraction (n ¼ 2, BT 6 and BT 8, P ¼ 0.0047), whereas no significant change was observed in the CD133 population (n ¼ 2, BT 6 and BT 8, P ¼ 0.3319). (d) Sorted primary medulloblastoma treated with KAAD-cyclopamine shows a decrease in Ptch1 expression only in the CD133 þ fraction (n ¼ 3, BT 6, BT 8, and BT 9, P ¼ 0.0128), whereas no significant change was observed in the CD133 population (n ¼ 3, BT 6, BT 8 and BT 9, P ¼ 0.1311). (e) Sorted primary medulloblastoma treated with KAAD-cyclopamine shows a decrease in Bmi1 expression only in the CD133 þ fraction at both the transcript level (n ¼ 4, BT 6–8, and BT 9, P ¼ 0.7310), whereas no significant change was observed in the CD133 population (n ¼ 4, BT 6–8 and BT 9, P ¼ 0.0047), and (f) protein level by flow cytometry (representative sample BT 9 shown). *Po0.05; **Po0.01; ns, P40.05.

allowing for further prospective investigation into the cellular origin, stem cell properties and specific genetic events found in each subgroup. Given that our data suggest that Bmi1 is a transcriptional target of Gli1, we hypothesized that Bmi1 expression would be highest in Shh-driven molecular subtype B. We probed for expression of Bmi1 in a medulloblastoma data set derived from exon array profiling of 14 normal brain samples and 103 primary medulloblastomas (Northcott et al., 2010). Although Bmi1 expression is elevated

across all medulloblastoma molecular subtypes compared with normal cerebellar controls, we found that its levels are highest in the most aggressive subtypes of medulloblastoma, groups C and D (Figure 5a). These subtypes, not yet characterized by activation of a distinct signaling pathway, are associated with poor survival and invasive metastatic disease, and they remain refractory to current treatment modalities. To interrogate the possibility that other stem cell signaling pathways that are even more potent activators of Bmi1 Oncogene


194

than Shh may be active in subtypes C and D, we searched the exon array data set for the expression of a recently well-characterized and potent activator of Bmi1, Twist1 (Yang et al., 2010). Interestingly, we found Twist1 expression to be highest in medulloblastoma subtypes C and D (Supplementary Figure 5). Yang et al. (2010) reported that Twist1 and Bmi1 act cooperatively to repress expression of both E-cadherin and p16Ink4a. This mechanism may be active in nonhedgehog-driven medulloblastoma, suggesting that Twist1 may play a specific role in Bmi1 regulation in subtypes C and D. To confirm the hypothetical cell–cell interactions found between CD133-sorted cell populations in primary human medulloblastoma, we sorted human Daoy cells for CD133 and subjected purified populations to quantitative real-time–polymerase chain reaction (qRT– PCR) for Hh pathway components and target genes. We found that again, Shh transcript was elevated in the CD133 cells, and Hh receptors Ptch1, Smo and downstream effector Gli1 transcripts were elevated in CD133 þ cells (Figure 5b). Furthermore, Math1, a known marker of external granule layer cells of the cerebellum (Lee et al., 2005) and the putative progenitor cell of origin for medulloblastoma (Flora et al., 2009; Ayrault et al., 2010), showed increased expression in the CD133 þ stem-like cells. Lastly, known Hh target gene N-myc (Oliver et al., 2003) shows increased transcript levels in the CD133 þ Hh-receiving cells. These data, together with experiments performed on primary human medulloblastoma BTICs (Figure 4), allowed us to propose a model of cell–cell interactions that may occur between the CD133 þ BTIC and supportive CD133 cells within the tumor niche (Figure 5c).

Discussion

Figure 5 Bmi1 expression increases with more aggressive medulloblastoma molecular subtypes, and may contribute to cell–cell interactions between the BTIC and tumor niche. (a) Data from Affymetrix exon array profiling of 14 normal cerebellar (CB) samples (9 fetal, 5 adult) and 103 primary medulloblastomas (WNT ¼ 8, SHH ¼ 33, group C ¼ 27, group D ¼ 35) show that Bmi1 expression is significantly higher in all medulloblastoma subgroups compared with normal controls (Po0.05), with the highest expression observed in group D, an aggressive subtype associated with metastases and poor outcome. Data are presented as log2-transformed signal intensity (that is, expression). (b) qRT– PCR performed on CD133-sorted Daoy medulloblastoma cells (n ¼ 3) shows that Shh transcript is increased in the CD133 population, and Ptch1, Smo, Gli1, N-myc and Math1 transcript levels are increased in CD133 þ cells. (c) Proposed model of BTIC and tumor niche cell–cell interactions, based on CD133 cells secreting Shh to their CD133 þ BTIC neighbors, effecting increased Bmi1 expression. *Po0.05; **Po0.01.

Oncogene

In this work, we propose the existence of an external signaling morphogen interacting with cell-intrinsic epigenetic pathways controlling tumorigenesis. Using both primary medulloblastoma BTICs and the Daoy medulloblastoma cell line grown in tumor sphere conditions, we demonstrate that a critical developmental pathway (Hh signaling) may influence epigenetic mechanisms through promoter activation of the key stem cell regulator, Bmi1. Through transcript and protein expression analysis, we were able to validate the use of Daoy grown in tumor sphere conditions, as it recapitulates the stem cell phenotype of primary medulloblastoma samples. The interaction of the morphogen Hh with cellintrinsic epigenetic mechanisms in both primary medulloblastoma and Daoy cells is implied by a relative enrichment of both Gli1 and Gli2 at the Bmi1 promoter. Recent evidence has identified a Gli1-dependent transcription program that drives proliferation of both granule neuron progenitors, and their neoplastic transformation to cause medulloblastoma (Kimura et al., 2005; Ayrault et al., 2010). Ayrault et al. (2010) have


195

shown that Math1 collaborates with the pro-proliferative Gli1-dependent transcriptional program to influence medulloblastoma development, and Clement et al. (2007) demonstrate that Gli1 drives cancer stem cell self-renewal in glioma models. We therefore focused on the functional effects of Gli1 on Bmi1 transcription, and Gli1 knockdown and overexpression experiments confirmed its function as a Bmi1 activator. Although these results lend functional relevance to the enrich ment of Gli1 at the Bmi1 promoter, future work should address the corresponding potential role of Gli2 in modulating Bmi1 function. We further performed comparative examination of the expression of Bmi1 through modulation of the Hh pathway using small molecule inhibitors and agonists. Despite a dosage-dependent increase in Bmi1 via Hh signaling, we were unable to observe a change in Bmi1 with Hh blockade. This may suggest that there are Hhindependent mechanisms that regulate Bmi1 (Subkhankulova et al., 2010). However, an alternate explanation may be offered by the fact that the majority of unsorted population comprises CD133 cells, which possess few Hh receptors and would show indifference to KAADcyclopamine-mediated Hh signaling blockade. Given the downstream regulation of Bmi1 by Hh, we explored whether Bmi1 knockdown would have significant functional effects. With a stable shBmi1 knockdown in Daoy spheres, we observed a marked reduction in proliferation in vitro. In vivo xenotransplantation results revealed that NOD/SCID mice injected with Daoy cells knocked down for Bmi1 had a similar tumor volume compared with wild type. However, a significant survival advantage was observed in mice with tumors knocked down for Bmi1. Histological examination revealed that knocking down Bmi1 led to reduced tumor infiltration. This corroborates recent data suggesting that elevated Bmi1 expression is associated with increasing tumor aggressiveness and predicts poor prognosis (Glinsky et al., 2005; Zakrzewska et al., 2011). Thus, we demonstrate that Bmi1 plays a critical role in the BTIC-driven tumorigenesis in medulloblastoma. In normal cerebellar development, proliferation is spatially regulated by Shh secretion by a discrete cell population (Purkinje neurons) to cells possessing the required Hh receptor components (external granule precursors) in order to elicit a controlled growth response (Dahmane and Ruiz i Altaba, 1999; Wechsler-Reya and Scott, 1999; Ruiz i Altaba et al., 2002; Lee et al., 2010). To interrogate whether this cell–cell interaction may be preserved within the BTIC population, we fractionated medulloblastoma samples into BTIC and non-BTIC populations, through sorting for the stem cell marker CD133. Interestingly, we found that CD133 þ cells react to Hh differently than their negative counterparts. Here, we hypothesize that CD133 þ fractions express the necessary receptors and downstream targets (such as Bmi1 and N-Myc) to respond to Hh signaling, whereas CD133 fractions secrete the soluble ligand to affect a response from CD133 þ populations. This putative regulatory func-

tion observed among cellular sub-fractions of a tumor invokes the important role of the tumor niche in maintaining tumor growth (Blanpain et al., 2004; Bendall et al., 2007). Given current efforts toward the molecular characterization of medulloblastomas into distinct subtypes (Pomeroy et al., 2002; Thompson et al., 2006; Kool et al., 2008; Northcott et al., 2010), we were interested to see if Bmi1 is implicated in any particular molecular subgroup of medulloblastomas. Although Bmi1 expression is elevated across all medulloblastoma molecular subtypes, its levels are highest in the most aggressive subtypes of medulloblastoma, groups C and D. These subtypes are associated with poor survival and invasive metastatic disease, and they remain refractory to current treatment modalities. Recent genomic platforms have shown that medulloblastomas of different molecular subtypes may originate from distinct precursor and stem cell populations in the cerebellum (Gilbertson and Ellison, 2008; Johnson et al., 2009; Ellison, 2010; Northcott et al., 2010; Parsons et al., 2010). Many mouse models have elegantly established that CD133 þ ventricular zone stem cells are distinct from Math1 þ external granule layer cells (Lee et al., 2005, 2010; Gibson et al., 2010), and strong evidence suggests that although non-hedgehog pathway medulloblastoma may originate from the former, the Shh subtype of medulloblastoma likely originates from the latter (Rao et al., 2004; Schuller et al., 2008; Read et al., 2009; Sutter et al., 2010). Our application of BTIC culture conditions to human primary medulloblastomas and medulloblastoma cell lines allows for the enrichment of both stem and progenitor cell populations that exist within tumor spheres. Although Hh pathway activation at the level of transcription mainly characterizes medulloblastoma molecular subtype B, it is well known that Shh, Notch and Wnt as well as other key developmental and stem cell pathway genes are likely preferentially activated in the stem and progenitor cell populations enriched by our model system. Thus, the BTIC model system applied to medulloblastoma offers a means to study stem cell signaling pathways that may not be active in every cell within the tumor, but are enriched in the various stem and progenitor cells (whether a Math1 þ external granule layer precursor, CD133 þ ventricular zone stem cell or CD15 þ progenitor cell) that have been identified as putative cells-of-origin for medulloblastoma. To interrogate the possibility that other stem cell signaling pathways that are even more potent activators of Bmi1 than Shh may be active in subtypes C and D, we observed that the expression of a recently wellcharacterized and potent activator of Bmi1, Twist1, was also highest in medulloblastoma subtypes C and D. A ‘stem cell signature’ comprising genes such as Twist1 and Bmi1 may help to further define current molecularbased classification by specifically identifying the less well-characterized non-Shh/Wnt subtypes (C and D) that are biologically aggressive and more refractory to current therapy. Coupled with our novel observation of cell–cell interactions that exist within medulloblastoma Oncogene


196

sub-populations, these data may assist in the search for alternative therapeutic targets in medulloblastoma that are more specific and produce fewer side effects.

unsorted cells, and purities for CD133 þ sorted cells and CD133 sorted cells from magnetic-activated cell sorting were determined by flow cytometry. The purity of CD133 þ cell fractions from brain tumors was 85±3.4%, and the CD133 cell fraction purity was 99.5±0.3%. The appropriate isotype control served as the negative control for every experiment.

Materials and methods Dissociation of primary medulloblastoma tissue and neurosphere culture Human brain tumor samples (see Table 1) were obtained from consenting patients, as approved by the Hamilton Health Sciences/McMaster Health Sciences Research Ethics Board. Samples were dissociated in artificial cerebrospinal fluid containing 0.2 Wunisch unit/ml Liberase Blendzyme 3 (Roche, Laval, QC, Canada), and incubated at 37 1C in a shaker for 15 min. The dissociated tissue was then filtered through 70 mm cell strainer and collected by centrifugation at 1500 r.p.m. for 3 min. Tumor cells were resuspended in tumor sphere medium consisting of a chemically defined serum-free neural stem cell medium, and plated in ultra-low attachment plate (Corning, Ottawa, ON, Canada). The components of our complete neural stem cell media per 500 ml include: Dulbecco’s modied Eagle’s medium/F12 (450 ml; Invitrogen, Burlington, ON, Canada), N2-supplement (5 ml; Invitrogen), HEPES (5 ml; Wisent), glucose (3 g; Invitrogen), N-acetylcysteine (60 mg/ml; Sigma), neural survival factor-1 (10 ml; Lonza, Mississauga, ON, Canada), epidermal growth factor (20 ng/ml; Sigma), basic fibroblast growth factor (20 ng/ml; Invitrogen), leukemic inhibitory factor (10 ng/ml; Millipore, Etobicoke, ON, Canada) and antibiotic–antimycotic (10 mg/ml; Wisent, Saint-Bruno, QC, Canada). Growth factors added to the media include human recombinant epidermal growth factor (20 ng/ml: Sigma), basic fibroblast growth factor (20 ng/ml; Invitrogen) and leukemic inhibitory factor (10 ng/ml; Chemicon, Etobicoke, ON, Canada). Red blood cells were removed using RBC lysis buffer (Stemcell Technologies, Vancouver, BC, Canada). The BTIC patient isolates shown in Table 1 are not renewable cell lines, but are cell isolates minimally cultured (24 h to o1 week) under stem cell conditions to select for stem cell populations. Daoy medulloblastoma cell line was also cultured in tumor sphere medium conditions. Treatments of neurospheres with Hh agonists and antagonists Cells from dissociated tumors were treated with biologically active, unmodified amino-terminal recombinant human Shh (R&D Systems, Minneapolis, MN, USA) and KAAD-cyclopamine (Stemcell Technologies) (Berman et al., 2002). We tested different concentrations and determined that the optimum activation or inhibition was obtained with 5 mg/ml Shh and 5 mM KAAD-cyclopamine. Tomatidine and/or dimethylsulfoxide vehicle were used as a negative control for KAAD-cyclopamine. Neurospheres were treated for 3 days and then collected for further assays. Magnetic-activated cell sorting and flow cytometric cell sorting CD133 þ cells from dissociated spheres were purified by magnetic-activated cell sorting (Miltenyi Biotec, Auburn, CA, USA) using microbeads conjugated to CD133 antibodies or cell sorting (MoFlo XDP; Beckman Coulter, Mississauga, ON, Canada) using allophycocyanin-labeled anti-CD133 antibodies (Miltenyi Biotec). Cell sorting was used when 4100 000 viable cells were obtained from tumor samples, and magnetic bead sorting was used for tumor samples o100 000 cells. Over 80% of samples processed for this work were sorted by flow cytometer. The percentage expression of CD133 for each group of Oncogene

Western blotting Denatured total protein (20 mg) was separated using 10% sodium dodecyl sulphate–polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membrane. Western blots were probed with monoclonal mouse-anti-human Bmi1 antibody (Upstate) and anti-GAPDH antibody (Abcam, Cambridge, MA, USA). The secondary antibody was horseradish peroxidaseconjugated goat anti-mouse IgG (H þ L) (Bio-Rad, Mississauga, ON, Canada) in a dilution of 1:15 000. The bands were visualized using an Immobilon Western kit (Millipore). Gli1 small interfering RNA and overexpression Daoy spheres were treated with human Gli1 small interfering RNA (Santa Cruz Biotechnology, Santa Cruz, CA, USA) as per the manufacturer’s instructions to knock down Gli1 mRNA expression. Following treatment for 48 h, Gli1 mRNA expression and Bmi1 mRNA levels were measured by RT– PCR to determine the efficiency of knockdown and effects, respectively. Cells treated with scrambled small interfering RNA (Santa Cruz Biotechnology) were used as controls. Daoy spheres were transfected with Gli1 expression vector or empty vector pCMV6-XL5 (Origene, Rockville, MD, USA) with Lipofectamine 2000 (Invitrogen) as per the manufacturer’s instructions. Following treatment for 48 h, the Gli1 and Bmi1 mRNA levels were assessed by RT–PCR. Limiting dilution assay After primary sphere formation was noted, spheres were dissociated to single cells with Liberase Blendzyme 3 and replated in 96 microwell plates in 0.2 ml tumor sphere medium as previously described (Singh et al., 2003). The secondary and tertiary sphere formation rate was calculated from the number of spheres forming from 2000 dissociated cells to quantify stem cell frequency. Cell proliferation assay Cell proliferation assay was performed on days 0, 3, 5 and 7 using CyQuant Cell Proliferation Kit (Invitrogen) as per the manufacturer’s instructions or by the Countess (Invitrogen). The plates were read with FLUOstar Omega fluorescence Microplate reader (BMG Labtech, Nepean, ON, Canada) at lEx ¼ 485, lEm ¼ 525. Proliferation was calculated as fold increase in viable cells over day 0. Viral production and transduction Lentiviral vectors (CS-H1-shRNA-EF-1 a-EGFP) expressing shRNA that targets human Bmi1 (target sequence: sh-BMI1-1, 50 -CAGATGAAGATAAGAGAAT-30 ) and luciferase were kind gifts from Dr Atushi Iwama. Replication-incompetent lentivirus was produced by co-transfection of the expression vector and ViraPower packaging mix (Invitrogen) containing an optimized mixture of three packaging plasmids: pLP1, pLP2 and pLP/VSVG into HEK 293FT cells. Viral supernatant was harvested 48 h after transfection, filtered through a 0.45-mm cellulose acetate filter and ultracentrifuged at 25 000 g for 2 h at 4 1C in a Beckman refrigerated centrifuge. The viral pellet was resuspended in 0.5 ml of phosphate-buffered saline and stored at 80 1C.


197 Quantitative real-time–polymerase chain reaction Total RNA from samples were isolated using the Qiagen RNeasy Micro kit (Qiagen, Toronto, ON, Canada) and reverse transcribed using Invitrogen’s Superscript III First Strand Synthesis kit (Invitrogen). Quantitative PCR was performed using the Chroma4 (Bio-Rad) with iQSYBR Green qPCR kit (Quanta VWR, Mississauga, ON, Canada). Data were presented as the ratio of the gene of interest to GAPDH (glyceraldehyde-3-phosphate dehydrogenase) as control. The program Primer3 (NCBI, Primer-BLAST, http://www.ncbi. nlm.nih.gov/tools/primer-blast) was used to design primer sequences, which are provided in Table 2. Chromatin immunoprecipitation ChIP assays were performed using LOW cell# ChIP kit (Diagenode, Sparta, NJ, USA) according to the manufacturer’s instructions. Briefly, chromatin that was crosslinked to transcription factors was immunoprecipitated using anti-Gli1 (Upstate). The input DNA was defined as an aliquot of sheared chromatin before immunoprecipitation, and was used to normalize the sample to the amount of chromatin added to each ChIP. DNA was purified from input, immunoprecipitated samples with corresponding antibodies. qPCR was performed using the SYBR Green system on the Chroma4 RT–PCR system (Bio-Rad) with Bmi1 promoter-forward, 50 -CCACTCTGCCTTCAGCGGTGCA-30 ; Bmi1 promoterreverse, 50 -CATACTACGATTATTTCATAGTTGC-30 . In vivo injections of BTICs, cyclopamine treatment and immunostaining of BTIC xenograft tumors BTIC xenografts from Daoy medulloblastoma cell line and Daoy knocked down for Bmi1 were generated as previously described (Singh et al., 2004). Briefly, BTIC samples were injected into the right frontal lobe of NOD-CB17-SCID mouse brains according to Research Ethics Board-approved protocols (n ¼ 19). Mice were injected with biological replicates consisting of 106 single-cell suspensions. Following 2 weeks of incubation, mice were treated every other day with either 25 mg/kg of cyclopamine (Toronto Research Chemicals) or vehicle (45% HBC; Sigma) until end point (Romer et al., 2004; Liu et al., 2006). The resulting human tumor xenografts were fixed, embedded in paraffin and stained with Bmi1 (Rabbit polyclonal antibody, 1:50; Abgent, Brockville, MD, USA), and Gli1 (Rabbit monoclonal antibody, 1:100; Cell Signaling, Pickering, ON, Canada) according to previously published protocols (Singh et al., 2003).

Table 2 qRT–PCR primers Bmi1 F Bmi1 R Gli1 F Gli1 R Gli2 F Gli2 R Smo F Smo R Ptch1 F Ptch1 R Shh F Shh R Math1 F Math1 R Sox2 F Sox2 R Oct4 F Oct4 R N-myc F N-myc R GAPDH F GAPDH R

50 -GGAGGAGGTGAATGATAAAAGAT-30 50 -AGGTTCCTCCTCATACATGACA-30 50 -CCCAAGGCTCTAGGTGGAAC-30 50 -CTGGGGGTAATGGGAAAAGA-30 50 -CAGCATCTCTTGCCACCATT-30 50 -AAGCCGGATCAAGGAGATGT-30 50 -TCTACGTCAATGCGTGCTTC-30 50 -CACAAACCAAACCACACCAG-30 50 -GGGATTAAAAGCAGCGAACC-30 50 -TGTAGGAGCGCTTCTGTGGT-30 50 -GCGGAAGGTATGAAGGGAAG-30 50 -GCCAAAGCGTTCAACTTGTC-30 50 -TGCACTTCTCGACTTTCGAGGACA-30 50 -AACTTGCCTCATCCGAGTCACTGT-30 50 -TCAGGAGTTGTCAAGGCAGAGAAG-30 50 -GCCGCCGCCGATGATTGTTATTAT-30 50 -GATGGCGTACTGTGGGCCC-30 50 -CAAAACCCGGAGGAGTCCCA-30 50 -TGGTAAAGAATGAGAAGGCCGCCA-30 50 -TCTAGCAAGTCCGAGCGTGTTCAA-30 50 -TGCACCACCAACTGCTTAGC-30 50 -GGCATGGACTGTGGTCATGAG-30

Abbreviation: qRT–PCR, quantitative real-time–polymerase chain reaction.

Statistical analysis Biological replicates from at least three tumors are compiled for each experiment for statistical power; unique samples were not pooled before analysis. Respective data represent means±s.e.m., n values are listed in the figure legends. Student’s t-test analyses were performed accordingly using the Prism 4.03 software package (GraphPad Software, La Jolla, CA, USA). The independent Student’s t-test was used to compare the continuous variables between two groups. The Kaplan– Meier estimation method was used for overall survival analysis, and a log-rank test was used to compare differences. The level of statistical significance was set at 0.05 for all tests.

Conflict of interest The authors declare no conflict of interest.

Acknowledgements Exon array profiling Expression of Bmi1 and Twist1 across human pediatric medulloblastomas is presented in boxplot format. The data are derived from Affymetrix (Santa Clara, CA, USA) Human 1.0 exon array profiling of 14 normal cerebellar samples (9 fetal, 5 adult) and 103 primary medulloblastomas (WNT ¼ 8, SHH ¼ 33, group C ¼ 27, group D ¼ 35). Data are presented as log2-transformed signal intensity (that is, expression).

We are very grateful to Atsushi Iwama for kind provision of Bmi1 shRNA lentiviral construct. We thank Merry Wang for artistic illustrations. This work was supported by funds from the Department of Surgery at McMaster University, the Ontario Institute for Cancer Research (OICR) and the JP Bickell Foundation. XW held a training award from the Stem Cell Network.

References Annabi B, Doumit J, Plouffe K, Laflamme C, Lord-Dufour S, Beliveau R. (2010). Members of the low-density lipoprotein receptor-related proteins provide a differential molecular signature between parental and CD133+ DAOY medulloblastoma cells. Mol Carcinog 49: 710–717. Annabi B, Rojas-Sutterlin S, Laflamme C, Lachambre MP, Rolland Y, Sartelet H et al. (2008). Tumor environment dictates medulloblastoma

cancer stem cell expression and invasive phenotype. Mol Cancer Res 6: 907–916. Ayrault O, Zhao H, Zindy F, Qu C, Sherr CJ, Roussel MF. (2010). Atoh1 inhibits neuronal differentiation and collaborates with Gli1 to generate medulloblastoma-initiating cells. Cancer Res 70: 5618–5627. Oncogene


198 Bao S, Wu Q, McLendon RE, Hao Y, Shi Q, Hjelmeland AB et al. (2006). Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature 444: 756–760. Bendall SC, Stewart MH, Menendez P, George D, Vijayaragavan K, Werbowetski-Ogilvie T et al. (2007). IGF and FGF cooperatively establish the regulatory stem cell niche of pluripotent human cells in vitro. Nature 448: 1015–1021. Berman DM, Karhadkar SS, Hallahan AR, Pritchard JI, Eberhart CG, Watkins DN et al. (2002). Medulloblastoma growth inhibition by hedgehog pathway blockade. Science 297: 1559–1561. Blanpain C, Lowry WE, Geoghegan A, Polak L, Fuchs E. (2004). Selfrenewal, multipotency, and the existence of two cell populations within an epithelial stem cell niche. Cell 118: 635–648. Blazek ER, Foutch JL, Maki G. (2007). Daoy medulloblastoma cells that express CD133 are radioresistant relative to CD133 cells, and the CD133+ sector is enlarged by hypoxia. Int J Radiat Oncol Biol Phys 67: 1–5. Bracken AP, Dietrich N, Pasini D, Hansen KH, Helin K. (2006). Genome-wide mapping of Polycomb target genes unravels their roles in cell fate transitions. Genes Dev 20: 1123–1136. Bruggeman SW, Hulsman D, Tanger E, Buckle T, Blom M, Zevenhoven J et al. (2007). Bmi1 controls tumor development in an Ink4a/Arf-independent manner in a mouse model for glioma. Cancer Cell 12: 328–341. Chen R, Nishimura MC, Bumbaca SM, Kharbanda S, Forrest WF, Kasman IM et al. (2010). A hierarchy of self-renewing tumorinitiating cell types in glioblastoma. Cancer Cell 17: 362–375. Clement V, Sanchez P, de Tribolet N, Radovanovic I, Ruiz i Altaba A. (2007). HEDGEHOG-GLI1 signaling regulates human glioma growth, cancer stem cell self-renewal, and tumorigenicity. Curr Biol 17: 165–172. Crawford JR, MacDonald TJ, Packer RJ. (2007). Medulloblastoma in childhood: new biological advances. Lancet Neurol 6: 1073–1085. Dahl JA, Collas P. (2008). A rapid micro chromatin immunoprecipitation assay (microChIP). Nat Protoc 3: 1032–1045. Dahmane N, Ruiz i Altaba A. (1999). Sonic hedgehog regulates the growth and patterning of the cerebellum. Development 126: 3089–3100. Dahmane N, Sanchez P, Gitton Y, Palma V, Sun T, Beyna M et al. (2001). The Sonic Hedgehog-Gli pathway regulates dorsal brain growth and tumorigenesis. Development 128: 5201–5212. Ellison DW. (2010). Childhood medulloblastoma: novel approaches to the classification of a heterogeneous disease. Acta Neuropathol 120: 305–316. Ellison DW, Onilude OE, Lindsey JC, Lusher ME, Weston CL, Taylor RE et al. (2005). beta-Catenin status predicts a favorable outcome in childhood medulloblastoma: the United Kingdom Children’s Cancer Study Group Brain Tumour Committee. J Clin Oncol 23: 7951–7957. Fan X, Matsui W, Khaki L, Stearns D, Chun J, Li YM et al. (2006). Notch pathway inhibition depletes stem-like cells and blocks engraftment in embryonal brain tumors. Cancer Res 66: 7445–7452. Fattet S, Haberler C, Legoix P, Varlet P, Lellouch-Tubiana A, Lair S et al. (2009). Beta-catenin status in paediatric medulloblastomas: correlation of immunohistochemical expression with mutational status, genetic profiles, and clinical characteristics. J Pathol 218: 86–94. Flora A, Klisch TJ, Schuster G, Zoghbi HY. (2009). Deletion of Atoh1 disrupts Sonic Hedgehog signaling in the developing cerebellum and prevents medulloblastoma. Science 326: 1424–1427. Fuccillo M, Joyner AL, Fishell G. (2006). Morphogen to mitogen: the multiple roles of hedgehog signalling in vertebrate neural development. Nat Rev Neurosci 7: 772–783. Gibson P, Tong Y, Robinson G, Thompson MC, Currle DS, Eden C et al. (2010). Subtypes of medulloblastoma have distinct developmental origins. Nature 468: 1095–1099. Gilbertson RJ. (2004). Medulloblastoma: signalling a change in treatment. Lancet Oncol 5: 209–218. Gilbertson RJ, Ellison DW. (2008). The origins of medulloblastoma subtypes. Annu Rev Pathol 3: 341–365. Oncogene

Glinsky GV, Berezovska O, Glinskii AB. (2005). Microarray analysis identifies a death-from-cancer signature predicting therapy failure in patients with multiple types of cancer. J Clin Invest 115: 1503–1521. Hallahan AR, Pritchard JI, Hansen S, Benson M, Stoeck J, Hatton BA et al. (2004). The SmoA1 mouse model reveals that notch signaling is critical for the growth and survival of sonic hedgehog-induced medulloblastomas. Cancer Res 64: 7794–7800. Huse JT, Holland EC. (2010). Targeting brain cancer: advances in the molecular pathology of malignant glioma and medulloblastoma. Nat Rev Cancer 10: 319–331. Johnson R, Wright KD, Gilbertson RJ. (2009). Molecular profiling of pediatric brain tumors: insight into biology and treatment. Curr Oncol Rep 11: 68–72. Kimura H, Stephen D, Joyner A, Curran T. (2005). Gli1 is important for medulloblastoma formation in Ptc1+/- mice. Oncogene 24: 4026–4036. Kool M, Koster J, Bunt J, Hasselt NE, Lakeman A, van Sluis P et al. (2008). Integrated genomics identifies five medulloblastoma subtypes with distinct genetic profiles, pathway signatures and clinicopathological features. PLoS One 3: e3088. Lee A, Kessler JD, Read TA, Kaiser C, Corbeil D, Huttner WB et al. (2005). Isolation of neural stem cells from the postnatal cerebellum. Nat Neurosci 8: 723–729. Lee EY, Ji H, Ouyang Z, Zhou B, Ma W, Vokes SA et al. (2010). Hedgehog pathway-regulated gene networks in cerebellum development and tumorigenesis. Proc Natl Acad Sci USA 107: 9736–9741. Leung C, Lingbeek M, Shakhova O, Liu J, Tanger E, Saremaslani P et al. (2004). Bmi1 is essential for cerebellar development and is overexpressed in human medulloblastomas. Nature 428: 337–341. Liu S, Dontu G, Mantle ID, Patel S, Ahn NS, Jackson KW et al. (2006). Hedgehog signaling and Bmi-1 regulate self-renewal of normal and malignant human mammary stem cells. Cancer Res 66: 6063–6071. Mao J, Ligon KL, Rakhlin EY, Thayer SP, Bronson RT, Rowitch D et al. (2006). A novel somatic mouse model to survey tumorigenic potential applied to the Hedgehog pathway. Cancer Res 66: 10171–10178. McCord AM, Jamal M, Williams ES, Camphausen K, Tofilon PJ. (2009). CD133+ glioblastoma stem-like cells are radiosensitive with a defective DNA damage response compared with established cell lines. Clin Cancer Res 15: 5145–5153. Michael LE, Westerman BA, Ermilov AN, Wang A, Ferris J, Liu J et al. (2008). Bmi1 is required for Hedgehog pathwaydriven medulloblastoma expansion. Neoplasia 10: 1343–1349, 5p following 1349. Northcott PA, Korshunov A, Witt H, Hielscher T, Eberhart CG, Mack S et al. (2010). Medulloblastoma comprises four distinct molecular variants. J Clin Oncol 29: 1408–1414. Oliver TG, Grasfeder LL, Carroll AL, Kaiser C, Gillingham CL, Lin SM et al. (2003). Transcriptional profiling of the Sonic hedgehog response: a critical role for N-myc in proliferation of neuronal precursors. Proc Natl Acad Sci USA 100: 7331–7336. Oliver TG, Read TA, Kessler JD, Mehmeti A, Wells JF, Huynh TT et al. (2005). Loss of patched and disruption of granule cell development in a pre-neoplastic stage of medulloblastoma. Development 132: 2425–2439. Pallini R, Ricci-Vitiani L, Montano N, Mollinari C, Biffoni M, Cenci T et al. (2010). Expression of the stem cell marker CD133 in recurrent glioblastoma and its value for prognosis. Cancer 117: 162–174. Parsons DW, Li M, Zhang X, Jones S, Leary RJ, Lin JC et al. (2010). The genetic landscape of the childhood cancer medulloblastoma. Science 331: 435–439. Pistollato F, Rampazzo E, Persano L, Abbadi S, Frasson C, Denaro L et al. (2010). Interaction of hypoxia-inducible factor-1alpha and Notch signaling regulates medulloblastoma precursor proliferation and fate. Stem Cells 28: 1918–1929. Pomeroy SL, Tamayo P, Gaasenbeek M, Sturla LM, Angelo M, McLaughlin ME et al. (2002). Prediction of central nervous system


199 embryonal tumour outcome based on gene expression. Nature 415: 436–442. Rao G, Pedone CA, Del Valle L, Reiss K, Holland EC, Fults DW. (2004). Sonic hedgehog and insulin-like growth factor signaling synergize to induce medulloblastoma formation from nestinexpressing neural progenitors in mice. Oncogene 23: 6156–6162. Read TA, Fogarty MP, Markant SL, McLendon RE, Wei Z, Ellison DW et al. (2009). Identification of CD15 as a marker for tumor-propagating cells in a mouse model of medulloblastoma. Cancer Cell 15: 135–147. Romer JT, Kimura H, Magdaleno S, Sasai K, Fuller C, Baines H et al. (2004). Suppression of the Shh pathway using a small molecule inhibitor eliminates medulloblastoma in Ptc1(+/)p53(/) mice. Cancer Cell 6: 229–240. Ruiz i Altaba A, Palma V, Dahmane N. (2002). Hedgehog-Gli signalling and the growth of the brain. Nat Rev Neurosci 3: 24–33. Schuller U, Heine VM, Mao J, Kho AT, Dillon AK, Han YG et al. (2008). Acquisition of granule neuron precursor identity is a critical determinant of progenitor cell competence to form Shh-induced medulloblastoma. Cancer Cell 14: 123–134. Singh SK, Clarke ID, Terasaki M, Bonn VE, Hawkins C, Squire J et al. (2003). Identification of a cancer stem cell in human brain tumors. Cancer Res 63: 5821–5828. Singh SK, Hawkins C, Clarke ID, Squire JA, Bayani J, Hide T et al. (2004). Identification of human brain tumour initiating cells. Nature 432: 396–401. Subkhankulova T, Zhang X, Leung C, Marino S. (2010). Bmi1 directly represses p21Waf1/Cip1 in Shh-induced proliferation of cerebellar granule cell progenitors. Mol Cell Neurosci 45: 151–162. Sutter R, Shakhova O, Bhagat H, Behesti H, Sutter C, Penkar S et al. (2010). Cerebellar stem cells act as medulloblastoma-initiating cells in a mouse model and a neural stem cell signature characterizes a subset of human medulloblastomas. Oncogene 29: 1845–1856. Taylor MD, Liu L, Raffel C, Hui CC, Mainprize TG, Zhang X et al. (2002). Mutations in SUFU predispose to medulloblastoma. Nat Genet 31: 306–310. Thomas WD, Chen J, Gao YR, Cheung B, Koach J, Sekyere E et al. (2009). Patched1 deletion increases N-Myc protein stability as a mechanism of medulloblastoma initiation and progression. Oncogene 28: 1605–1615. Thompson MC, Fuller C, Hogg TL, Dalton J, Finkelstein D, Lau CC et al. (2006). Genomics identifies medulloblastoma subgroups that

are enriched for specific genetic alterations. J Clin Oncol 24: 1924–1931. Thon N, Damianoff K, Hegermann J, Grau S, Krebs B, Schnell O et al. (2008). Presence of pluripotent CD133(+) cells correlates with malignancy of gliomas. Mol Cell Neurosci 43: 51–59. Uchida N, Buck DW, He D, Reitsma MJ, Masek M, Phan TV et al. (2000). Direct isolation of human central nervous system stem cells. Proc Natl Acad Sci USA 97: 14720–14725. Valk-Lingbeek ME, Bruggeman SW, van Lohuizen M. (2004). Stem cells and cancer; the polycomb connection. Cell 118: 409–418. Wang R, Chadalavada K, Wilshire J, Kowalik U, Hovinga KE, Geber A et al. (2010). Glioblastoma stem-like cells give rise to tumour endothelium. Nature 468: 829–833. Wechsler-Reya RJ, Scott MP. (1999). Control of neuronal precursor proliferation in the cerebellum by Sonic Hedgehog. Neuron 22: 103–114. Wiederschain D, Chen L, Johnson B, Bettano K, Jackson D, Taraszka J et al. (2007). Contribution of polycomb homologues Bmi-1 and Mel-18 to medulloblastoma pathogenesis. Mol Cell Biol 27: 4968–4979. Yan X, Ma L, Yi D, Yoon JG, Diercks A, Foltz G et al. (2011). A CD133-related gene expression signature identifies an aggressive glioblastoma subtype with excessive mutations. Proc Natl Acad Sci USA 108: 1591–1596. Yang MH, Hsu DS, Wang HW, Wang HJ, Lan HY, Yang WH et al. (2010). Bmi1 is essential in Twist1-induced epithelial-mesenchymal transition. Nat Cell Biol 12: 982–992. Yang ZJ, Ellis T, Markant SL, Read TA, Kessler JD, Bourboulas M et al. (2008). Medulloblastoma can be initiated by deletion of Patched in lineage-restricted progenitors or stem cells. Cancer Cell 14: 135–145. Yin AH, Miraglia S, Zanjani ED, Almeida-Porada G, Ogawa M, Leary AG et al. (1997). AC133, a novel marker for human hematopoietic stem and progenitor cells. Blood 90: 5002–5012. Yu CC, Chiou GY, Lee YY, Chang YL, Huang PI, Cheng YW et al. (2010). Medulloblastoma-derived tumor stem-like cells acquired resistance to TRAIL-induced apoptosis and radiosensitivity. Childs Nerv Syst 26: 897–904. Zakrzewska M, Zakrzewski K, Gresner SM, Piaskowski S, ZalewskaSzewczyk B, Liberski PP. (2011). Polycomb genes expression as a predictor of poor clinical outcome in children with medulloblastoma. Childs Nerv Syst 27: 79–86.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene