SEARCH

SEARCH BY CITATION

Keywords:

  • Brain tumor-initiating cell;
  • Medulloblastoma;
  • Stem cells;
  • FoxG1;
  • Bmi1;
  • Self-renewal

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Brain tumors represent the leading cause of childhood cancer mortality, of which medulloblastoma (MB) is the most frequent malignant tumor. Recent studies have demonstrated the presence of several MB molecular subgroups, each distinct in terms of prognosis and predicted therapeutic response. Groups 1 and 2 are characterized by relatively good clinical outcomes and activation of the Wnt and Shh pathways, respectively. In contrast, groups 3 and 4 (“non-Shh/Wnt MBs”) are distinguished by metastatic disease, poor patient outcome, and lack a molecular pathway phenotype. Current gene expression platforms have not detected brain tumor-initiating cell (BTIC) self-renewal genes in groups 3 and 4 MBs as BTICs typically comprise a minority of tumor cells and may therefore go undetected on bulk tumor analyses. Since increasing BTIC frequency has been associated with increasing tumor aggressiveness and poor patient outcome, we investigated the subgroup-specific gene expression profile of candidate stem cell genes within 251 primary human MBs from four nonoverlapping MB transcriptional databases (Amsterdam, Memphis, Toronto, Boston) and 74 NanoString-subgrouped MBs (Vancouver). We assessed the functional relevance of two genes, FoxG1 and Bmi1, which were significantly enriched in non-Shh/Wnt MBs and showed these genes to mediate MB stem cell self-renewal and tumor initiation in mice. We also identified their transcriptional regulation through reciprocal promoter occupancy in CD15+ MB stem cells. Our work demonstrates the application of stem cell data gathered from genomic platforms to guide functional BTIC assays, which may then be used to develop novel BTIC self-renewal mechanisms amenable to therapeutic targeting. STEM Cells2013;31:1266–1277


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Medulloblastoma (MB) is the most frequent malignant pediatric brain tumor, representing 20% of newly diagnosed childhood central nervous system malignancies [1]. Despite current multimodal therapy consisting of surgical resection, radiation, and chemotherapy jointly yielding an 80% 5-year survivorship, significant treatment-induced morbidity and long-term clinical sequelae are common and adversely affect quality-of-life [1–4]. Given the limitations of current clinico-pathological parameters to accurately predict treatment response [1, 2, 2], multiple genomic platforms have been used to characterize the aberrant expression of signaling pathways in MB. This has reconceptualized the heterogeneity that exists within pathological subgroups along with giving context to the role of key stem cell signaling pathways in MB pathogenesis [6–10].

The recent molecular classification of MB consists of four subgroups, each distinct in terms of prognosis and predicted therapeutic response. Groups 1 and 2 are characterized by upregulation of genes in the canonical Wnt or Sonic hedgehog (Shh) pathways, respectively [7, 8, 10-13]. These two subgroups are separated from each other and all others by principal component analysis, and both are associated with relatively good clinical outcomes [8]. Wnt-driven MBs are primarily characterized by monosomy 6, activating mutations of CTNNB1, and an excellent prognosis [7, 10, 11, 14], whereas targeted Shh inhibitors have shown promising results in patients with Shh-driven MBs [15, 16]. However, group 3 and 4 MBs are associated with metastatic disease and poor patient outcome [6–8]. These aggressive MBs, which have been collectively labeled as “non-Shh/Wnt” subgroups [4], remain refractory to current treatment modalities and lack aberrant activation of known signaling pathways. Furthermore, recent genomic meta-analyses have revealed evidence for the existence of subtypes within group 3 and 4 MBs, now labeled as 3α, 3β, 4α, and 4β, each with distinct gene mutations and transcriptome profiles [6, 13, 13]. Recently, two mouse models of c-MYC-driven MB [18, 19] have been shown to recapitulate a subclass of group 3 human MBs (group 3α), implicating c-MYC amplification as a driver of tumor formation in this subset. Other reports have shown non-Shh/Wnt MBs to express genes associated with increased photoreceptor [6] or neuronal differentiation [7] and more recently the exclusive expression of NPR3 [8], KCNA1 [8], or FSTL5 [20]. However, these reports do not address the role of stem cell regulatory genes in driving MB tumorigenesis, which is in keeping with the minimal literature on examining selectively activated developmental signaling pathways in non-Shh/Wnt MBs [21–23], and even fewer reports linking aberrant MB stem cell self-renewal signaling to patient prognosis and outcome [24].

The cancer stem cell (CSC) hypothesis suggests that tumors are organized into a hierarchy in which only a rare clonal population of cells, termed CSCs or tumor-initiating cells (TICs), have the ability to initiate, proliferate, and maintain tumor growth. This is in sharp contrast to all other cells of the bulk tumor, characterized by a limited proliferative capacity and restricted lineage potential [25, 26]. By applying culture conditions and assays used to characterize normal neural stem cells (NSCs), we were among the first researchers to identify cells from a variety of human brain tumors (including MB), termed brain tumor-initiating cells (BTICs), which exhibited the stem cell properties of self-renewal and differentiation [27, 28]. A unique property by which stem cells may induce oncogenesis is self-renewal, defined as the ability of the parental cell to generate an identical daughter cell and a second daughter cell of the same or different phenotype [29, 30]. Given that normal NSCs must maintain a relative balance between self-renewal and differentiation, brain tumorigenesis may be conceptualized as a disease of unregulated BTIC self-renewal [25, 26]. Despite the promise of novel therapies targeting the MB BTIC self-renewal machinery to improve outcomes in those who remain refractory to current treatment, the regulatory mechanisms of these pathways remain unknown.

In this study, we have characterized the role of MB stem cell-specific pathways specifically enriched in non-Shh/Wnt MBs. Since prior work has demonstrated increasing BTIC frequency to be associated with tumor aggressiveness [27, 28] and poor patient outcome [24, 31, 31], we have investigated the likelihood of BTIC self-renewal genes in specifying group 3 and 4 non-Shh/Wnt MBs. While these MBs and their further fractionated subtypes may display characteristic gene mutations, amplifications, or gene expression profiles [13], we hypothesized that all of these MB subtypes may be aberrantly driven by stem cell regulatory genes promoting enhanced BTIC self-renewal. If BTIC self-renewal genes and their signaling pathways are specifically identified in aggressive MB subgroups, mechanistic studies of their regulatory function can then be conducted through application of our in vitro and in vivo BTIC models.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Stem Cell Gene Profiling in Non-Shh/Wnt Subtype MB

MB microarray data for 103 MBs were downloaded from Gene Expression Omnibus (GEO) (GSE21140). These data consisted of processed Affymetrix CEL files (Affymetrix Human Exon 1.0 ST Array [transcript (gene) version]) that had undergone gene-level analysis (CORE content), quantile normalization (sketch), and summarization using PLIER with PM-GCBG background correction for 103 MBs. Clinical annotations for each MB tumor, including subgroup (Wnt, Shh, group 3, group 4) were also downloaded from this location. Expression values for group 3 and 4 tumors were pooled into one group titled, “non-Shh/Wnt.” The expression of each gene was plotted as the normalized fluorescence intensity of the corresponding affymetrix probes. Similarly, raw Affymetrix datasets for 62 primary human MBs, 40 primary human MBs, 15 Daoy MB cell line samples, and 46 primary human MBs were, respectively, downloaded and processed from GEO (GSE10327, GSE12992, GSE7578) or http://www.stjuderesearch.org/site/data/medulloblastoma (Thompson et al. [10]) and normalized using robust multi-array average [33]. When available, clinical annotations for each tumor were also downloaded from these locations. When multiple probe sets identified the same gene, the probe set with the highest mean expression across all patients in the dataset was used.

nCounter System (NanoString) Gene Expression Profiling

RNA was extracted from formalin-fixed paraffin-embedded (FFPE) tissue using Qiagen RNeasy FFPE kit. Exactly 250 ng of RNA was run for each patient sample. Analysis using nCounter Gene Expression system was conducted at the Centre for Translational and Applied Genomics [BC Cancer Agency, Vancouver, BC]. A custom codeset synthesized by nCounter (NanoString Technologies, Seattle, WA) was designed including 22 MB-specific subgrouping gene probes [34] in addition to other genes of interest which include: CD15 (RefSeq NM_002033.2), BMI1 (NM_005180.5), FOXG1 (NM_005249.3), SOX2 (NM_003106.2), and POU5F1 (RefSeq NM_002701.4). The recommendations outlined by NanoString Technologies were all followed regarding mRNA sample preparation, hybridization, detection and scanning, and data normalization.

Dissociation of Primary MB Tissue and Neurosphere Culture

Human brain tumor samples (supporting information Table 3) were obtained from consenting patients, as approved by the Hamilton Health Sciences/McMaster Health Sciences Research Ethics Board. Briefly, samples were dissociated in artificial cerebrospinal fluid containing 0.2 Wunisch unit/mL Liberase Blendzyme 3 (Roche, Basel, Switzerland, http://www.roche-applied-science.com) filtered through 70 mm cell strainer. Tumor cells were resuspended in tumor sphere medium consisting of a chemically defined serum-free NSC medium and plated in an ultra-low attachment plate (Corning, Acton, MA, http://www.corning.com/lifesciences). The components of our complete NSC media per 500 mL include: Dulbecco's modified Eagle's medium/F12 (450 mL; Invitrogen, Carlsbad, CA, http://www.invitrogen.com), N2-supplement (5 mL; Invitrogen), HEPES (5 mL; Wisent, Saint-Jean-Baptiste de Rouville, QC, Canada, http://www.wisent.ca), glucose (3 g; Invitrogen), N-acetylcysteine (60 mg/mL; Sigma), neural survival factor-1 (10 mL; Lonza), epidermal growth factor (20 ng/mL; Sigma), basic fibroblast growth factor (20 ng/mL; Invitrogen), leukemic inhibitory factor (10 ng/mL; Chemicon, Temecula, CA, http://www.chemicon.com). The BTIC patient isolates shown in supporting information Table 3 are not renewable cell lines but rather minimally cultured cell isolates (24 hours to <1 week) within stem cell conditions to select for stem cell populations. Daoy MB (American Type Culture Collection) and H9 human embryonic stem cell-derived NSCs (Invitrogen) were also cultured in tumor sphere medium conditions and as per manufacturer's instructions, respectively.

Quantitative Real-Time-Polymerase Chain Reaction

Total RNA from samples was isolated using the Qiagen RNeasy Micro kit (Qiagen, Hilden, Germany, http://www1.qiagen.com) and reverse transcribed using Invitrogen's Superscript III First Strand Synthesis kit (Invitrogen). Quantitative polymerase chain reaction (PCR) was performed using the Chroma4 (Bio-Rad, Hercules, CA, http://www.bio-rad.com) with iQSYBR Green qPCR kit (Quanta VWR). 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 provided in supporting information Table 4.

Viral Production and Transduction

Lentiviral vectors (CS-H1-shRNA-EF-1 a-EGFP) expressing short hairpin RNA (shRNA) that targets human Bmi1 and luciferase were kind gifts from Dr. Atushi Iwama. Lentiviral vectors expressing shRNA that targets FoxG1 and ORF expression clone for FOXG1 were from GeneCopoeia (MD, USA, http://www.genecopoeia.com). Replication-incompetent lentivirus was produced by cotransfection of the expression vector and ViraPower packaging mix (Invitrogen). Viral supernatant was harvested 48 hours after transfection, filtered through a 0.45-mm cellulose acetate filter and precipitated by PEGit (System Biosciences). The viral pellet was resuspended in 0.5 mL of phosphate-buffered saline and stored at 80°C. Daoy cells were transduced with lentiviral vectors and treated with the respective antibiotics after 48 hours of transduction to develop stable lines.

Western Blotting

Denatured total protein (20 mg) was separated using 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membrane. Western blots were probed with monoclonal mouse-anti-human Bmi1 antibody (Upstate, Charlottesville, VA, http://www.upstate.com), polyclonal rabbit-anti-human FoxG1 antibody (Abcam, Cambridge, U.K., http://www.abcam.com), and anti-GAPDH antibody (Abcam). The secondary antibody was horseradish peroxidase-conjugated goat anti-mouse IgG (Bio-Rad) or goat anti-rabbit IgG (Sigma). The bands were visualized using an Immobilon Western kit (Millipore, Billerica, MA, http://www.millipore.com).

Immunohistochemistry

The original histologic slides were reviewed on all cases to confirm the diagnosis of MB. A tissue microarray was subsequently constructed from the original blocks; triplicate 1 mm cores were extracted from every tumor resection performed on the patient. A total of 83 patients were included. There was patient outcomes data available for 74 of the 83 cases. Antibodies to NRP3, KCNA1, and SFRP1 were used to subgroup the group 3, group 4, and SHH tumor subtypes as previously described [8]. b-Catenin was used for the identification of the WNT subgroup. For these proposed subgroup-specific antibodies, the Ventana Benchmark XT autostainer was used; vendors, clones, dilutions, and protocols are listed as follows: b-catenin (BD, ab610154) antibody, dilution 1:100, protocol cc1; SFRP (Abcam, ab4193) antibody, dilution 1:200, protocol cc1; NPR3 (Sigma, HPA031065) antibody, dilution 1:30, protease protocol; KCNA1 (Abcam, ab32433) antibody, dilution 1:200, protocol cc1. CD15 (Beckman Coulter, Fullerton, CA, http://www.beckmancoulter.com, IM1954U) immunohistochemical staining was performed as previously published [35]. All immunohistochemical stains (with the exception of b-catenin) were scored semiquantitatively using a four point scale [0-3] that took into consideration the intensity and the diffusivity of the staining; subsequently, these scores were binarized into “low” [0, 1] or “high” [2, 3] categories for survival analysis. b-Catenin was scored according to the presence of nuclear staining only; cases were considered to be either nuclear positive or negative.

Neurosphere Size Assay to Determine BTIC Proliferative Potential

Proliferative potential of MB stem cells was determined according to neurosphere size. 200 single cells were plated in a 96-microwell plate in 0.2 mL tumor sphere medium. After 5 days in culture, the size (diameter) of resulting secondary neurospheres was determined using light microscopy and Metamorph 7.1 (Molecular Devices, Union City, CA, http://www.moleculardevices.com). Similarly, neurosphere size was determined 3 days after siRNA treatment.

Cell Proliferation Assay

Single cells were plated at a density of 1,000 cells per well in complete tumor sphere media in triplicate, in a 96-well plate (100 mL per well) and incubated for 3 days. Alamar Blue (Invitrogen), a Resazurin-based fluorescent indicator of cell metabolism, was added (20 mL) to each well approximately 18 hours prior to the readout time point. Fluorescence was measured using a FLUOstar Omega Fluorescence 556 Microplate reader (BMG LABTECH) at excitation and emission wavelengths of 535 nm and 600 nm, respectively, according to recommendations given by the manufacturer. Readings were analyzed using Omega analysis software. Proliferation was calculated as fold increase in viable cells over day 0.

Secondary Sphere Formation Assay

After primary sphere formation was noted, spheres were dissociated to single cells and replated in tumor sphere medium as previously described [27]. To quantify stem cell frequency, the secondary and tertiary sphere formation rate was calculated from the number of spheres forming from 2,000 dissociated single cells.

Differentiation Assay

Cells were differentiated in NSC media supplemented with 20% fetal bovine serum which was replaced every other day. Differentiation was carried out for 7 days after which, differentiated cells were trypsinized and harvested for subsequent gene expression analyses.

Chromatin Immunoprecipitation

Chromatin immunoprecipitation (ChIP) assays were performed using Low cell# ChiP kit (Diagenode) according to the manufacturer's instructions. Briefly, crosslinked chromatin was immunoprecipitated using anti-Bmi1 (Upstate) and anti-FoxG1 (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 and immunoprecipitated samples with corresponding antibodies. Quantitative PCR (qPCR) was performed using the SYBR Green system on the Choma4 RT-PCR system (Bio-Rad). ChIP Primers are displayed in supporting information Table 4.

Flow Cytometric Cell Sorting

Primary Daoy tumor spheres were dissociated to single cell suspension. CD15+ cells were enriched using flow cytometric sorting (MoFlo XDP) using PE-labeled anti-CD15 antibodies (Beckman Coulter). The percentage expression of CD15 for each group of unsorted cells and purities for CD15+/CD15− sorted cells from flow cytometric sorting was determined. The appropriate isotype control served as the negative control.

In Vivo MB Stem Cell Injections and H&E Staining of Xenograft Tumors

BTIC xenografts from Daoy MB cell line expressing control vector, OE FoxG1, and shBmi1+OE FoxG1 were generated as previously described [36]. Briefly, BTIC samples were injected into the right frontal lobe of NOD-CB17-SCID mouse brains according to Research Ethics Board-approved protocols (n = 9). Mice were injected with biological replicates consisting of 106 single-cell suspensions. The resulting human tumor xenografts were fixed, embedded in paraffin for hematoxylin and eosin (H&E) staining; images were taken using an Aperio Slide Scanner and analyzed using ImageScope v11.1.2.760 software (Aperio).

Statistical Analysis

For all in vitro studies, biological replicates from at least three tumors are compiled for each experiment in order to achieve statistical power; unique samples were not pooled before analyses. Data represent mean ± SEM, n values are listed in figure legends. Student's t test analyses were performed accordingly, using the Prism 4.03 software package (GraphPad Software). The independent Student's t test was used to compare the continuous variables between two groups. The level of statistical significance was set at 0.05 for all tests.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Identification of Stem Cell Genes Preferentially Expressed in Non-Shh/Wnt Subgroup MB

Current unsupervised hierarchical clustering analyses of all published MB functional genomics platforms have produced four molecular subgroups: Wnt, Shh, group 3, and group 4 [13]. Groups 3 and 4 share a poor clinical outcome, and as their names denote, have yet to be defined by activation of specific signaling pathways [4]. Given prior work has shown increasing BTIC frequency to be associated with features of increasing tumor aggressiveness [27, 28] and reduced patient survival [24, 31, 31], we hypothesized that a candidate gene approach for known stem cell-associated genes would specify non-Shh/Wnt poor-outcome MBs. Therefore, we took the novel approach of probing four existing MB transcriptional databases (Boston [6], Amsterdam [7], Toronto [8], Memphis [10]) along with a recent NanoString-subgrouped cohort of MBs for differential stem cell gene expression patterns in Wnt, Shh, group 3, group 4, and non-Shh/Wnt MBs.

Examination of our candidate gene list (described in supporting information Table 1 with their regulatory functions) [18, 19, 22, 27, 28, 37-62] consisting of genes implicated in the self-renewal and proliferation of a number of normal and CSC populations showed enriched expression in the non-Shh/Wnt subgroup (Fig. 1). Interestingly, we were able to identify FoxG1 and Bmi1 as being preferentially expressed in non-Shh/Wnt MBs (Fig. 2, supporting information Table 2). Although Bmi1 has been previously studied by our group [36] and others, FoxG1 warranted further investigation as it was the strongest overall predictor of groups 3 and 4, and the most independently predictive gene of these tumors (supporting information Table 2). In addition, Fasano et al. [42] have shown Bmi1 and FoxG1 to act as comediators of normal NSC self-renewal. FoxG1 is a forkhead-box family transcription factor that is developmentally active in forebrain specification, and it has been shown to be frequently dysregulated in MB [63]. The functional relevance of these two genes in MB stem cells was subsequently assessed using in vitro stem cell assays.

thumbnail image

Figure 1. Multiple candidate stem cell genes are preferentially expressed in non-Shh/Wnt subgroup medulloblastomas. (A–D): Affymetix exon array data in four independent datasets shows a differential expression pattern in which many candidate cancer stem cell genes are associated with poor outcome group 3 and 4 medulloblastoma (MBs). (E): NanoString gene expression data for select genes showing enrichment in group 3 and 4 MBs. Abbreviation: shh, sonic hedgehog.

Download figure to PowerPoint

thumbnail image

Figure 2. FoxG1 and Bmi1 expression identify non-Shh/Wnt subgroup medulloblastomas. (A–D): Affymetix exon array and (E) NanoString data show enriched FoxG1 and Bmi1 expression in poor-outcome MB subgroups (p < .05). Data are presented as log2-transformed signal intensity (that is, expression).

Download figure to PowerPoint

Bmi1 and FoxG1 Function as Regulators of MB Stem Cell Properties

Probing of MB transcriptional databases revealed a strong association of Bmi1 and FoxG1 with group 3 and 4 MBs, but their functional role in MB remains unknown. We therefore decided to apply our in vitro and in vivo BTIC model systems to further interrogate the role of these genes in MB stem cells.

Having previously characterized BTIC populations from primary human MB [27, 28], we applied this experimental model system to three primary pediatric MBs (supporting information Table 3). We have called these specimens “MB stem cell patient isolates” to emphasize the fact that these cells are not renewable cell lines but are cell isolates minimally cultured (24 hours to 1 week) under conditions that select for stem cell populations. Owing to limited primary sample availability, we have supplemented our experimental work on human primary MB stem cells with studies of the Daoy [36] and MED8a [64] cell lines.

Given that all current MB transcriptional databases do not profile cells enriched for tumor sphere populations, we assessed the differential transcript levels of the two stem cell genes most preferentially expressed in non-Shh/Wnt subgroup MBs, FoxG1 and Bmi1, within normal NSCs, Daoy and MED8a tumor spheres, and primary MB stem cells. Primer sequences are listed in supporting information Table 4. We found FoxG1 and Bmi1 transcript levels to be significantly elevated in Daoy spheres, primary MB stem cells (Fig. 3A, 3B), and MED8a spheres (supporting information Fig. 4A, 4B) when compared to NSCs. No significant difference was observed in the expression of FoxG1 and Bmi1 between Daoy spheres, MED8a spheres, and primary MB stem cells.

thumbnail image

Figure 3. FoxG1 and Bmi1 are significantly expressed in primary MB stem cells and Daoy tumor spheres. (A): FoxG1 transcript levels are significantly elevated in primary human MB stem cells (n = 3, BT1-3, p = .0075) and Daoy MB tumor spheres (n = 3, p = .0146) when compared to normal human NSCs (n = 3). (B) Similarly, Bmi1 transcript levels are significantly elevated in primary human MB stem cells (n = 3, BT1–3, p = .0191) and Daoy MB tumor spheres (n = 3, p = .0238) when compared to normal human NSCs (n = 3). *, p < .05; **, p < .01. Abbreviations: MB, medulloblastoma; hNSC, human neural stem cell.

Download figure to PowerPoint

thumbnail image

Figure 4. FoxG1 and Bmi1 function to regulate proliferation and self-renewal of medulloblastoma stem cells. Daoy tumor spheres knocked down for FoxG1 (shFoxG1), Bmi1 (shBmi1), and FoxG1+Bmi1 (shFoxG1+shBmi1) showed a decrease in (A, B) sphere size (n = 3, p = .034; n = 3, p = .027; n = 3, p = .003, respectively) when compared to control cells. Conversely, Daoy tumor spheres overexpressing FoxG1 (OE FoxG1) showed an increase in (A, B) sphere size (n = 3, p = .0003), but a reduction compared to control cells when cotransduced shBmi1 (n = 3, p = .042). (C): Daoy tumor sphere proliferative potential was significantly reduced with shFoxG1, shBmi1, shFoxG1+shBmi1 (n = 3, p = .0103; n = 3, p = .0481; n = 3, p = .0017, respectively) and significantly elevated in OE FoxG1 cells (n = 3, p = .0003). Together, OE FoxG1+shBmi1 reduced the proliferative potential of tumor spheres, suggesting Bmi1 as a possible feedback regulator of FoxG1 expression (n = 3, p = .0113). (D): Self-renewal capacity was further diminished following shFoxG1, shBmi1, and shFoxG1+shBmi1 (n = 3, p = .0170; n = 3, p = .0431; n = 3, p = .00087, respectively) when compared to control cells. In contrast, self-renewal capacity is enhanced in OE FoxG1 cells (n = 3, p = .0412) and reduced in OE FoxG1+shBmi1 cells (n = 3, p = .0285) compared to control. Scale bar = 100 mm. *, p < .05; **, p < .005; ***, p < .0005.

Download figure to PowerPoint

To interrogate the functional significance of FoxG1 and Bmi1 in regulating stem cell properties essential for promoting MB tumorigenesis, we generated FoxG1 overexpression constructs (OE FoxG1) and performed shRNA-mediated knockdown of both genes (shBmi1, shFoxG1) using multiple constructs. The efficiency of gene knockdown and overexpression was validated at protein and transcript levels for all constructs (supporting information Fig. 1). In vitro limiting dilutions assays for each construct correlated with the levels of FoxG1 and Bmi1 (supporting information Fig. 2). shFoxG1-2 and shBmi1-3, displaying the best gene and functional knockdowns, were chosen for further study. Tumor sphere size in cells treated with shFoxG1 and shBmi1 or OE FoxG1 was markedly reduced or increased, respectively, in comparison to spheres generated from cells treated with control vector (Fig. 4A, 4B). To assess whether these genes may drive MB tumor cell proliferation, we also measured the proliferative potential of tumor spheres. Cells treated with shRNA and overexpression constructs were less and more proliferative, respectively, than those cells treated with control vector (Fig. 4C). Since brain tumorigenesis may be attributed to unregulated BTIC self-renewal, we assessed the self-renewal capacity of Daoy tumor spheres. Secondary tumor sphere formation was significantly diminished following knockdown, which suggested both genes to independently function as putative BTIC self-renewal genes (Fig. 4D). Since Bmi1 has been shown to co-operate with FoxG1 in maintaining the self-renewal machinery of normal NSCs [42], we cotransduced both shBmi1 and shFoxG1 constructs to determine if there may be an additive effect in self-renewal reduction. Interestingly, we observed a significant reduction in self-renewal compared to cells transduced with control vector. Finally we performed differentiation assays on primary MB stem cells, Daoy, and Med8a tumor spheres. Although we found Daoy and Med8a tumor spheres to be refractory to standard differentiation conditions, in minimally cultured primary human MB stem cells, we observed a decrease in neural lineage marker expression with OE FoxG1 (supporting information Fig. 5a) and an increase with shFoxG1 cells (supporting information Fig. 5b). Furthermore, FoxG1 expression decreased with differentiation of primary human MB stem cells (supporting information Fig. 5c), suggesting a role for FoxG1 in maintaining a stem cell state.

thumbnail image

Figure 5. Bmi1 and FoxG1 display differential binding at promoters within distinct CD15+ enriched stem cells. Neither (A) FoxG1 nor (B) Bmi1 appear to enrich at the Bmi1 or FoxG1 promoters, respectively, in unsorted tumor sphere populations (n = 5, p = .3253; n = 5, p = .8848, respectively). Cell sorting for CD15 shows CD15+ cells to contain significantly elevated transcript levels of (C)FoxG1 (n = 3, p = .0016) and (D)Bmi1 (n = 3, p < .0001) when compared to CD15− cells. Chromatin immunoprecipitation experiments in CD15-sorted tumor sphere populations demonstrate significant enrichments for (E) FoxG1 at the Bmi1 promoter (n = 5, p = .0151) and (F) Bmi1 at the FoxG1 promoter (n = 5, p = .0044) in the CD15+ cell fraction. There were no enrichments observed in the CD15− cell fraction for either (E) FoxG1 at the Bmi1 promoter (n = 5, p = .0759) or (F) Bmi1 at the FoxG1 promoter (n = 5, p = .2323). CD15− cells display increased expression levels of downstream targets of FoxG1 and Bmi1, (G)p16 (n = 3, p = .0062) and (H)p21 (n = 3, p = .0015) when compared to CD15+ cells. *, p < .05; **, p < .005; ***, p < .0001.

Download figure to PowerPoint

Therefore, it became apparent that Bmi1 and FoxG1 function as mediators of self-renewal, proliferation, and differentiation in MB and may in fact be interacting with one another in a manner akin to normal NSCs [42]. As hypothesized, OE FoxG1 demonstrated a significant increase in secondary sphere formation compared to control cells (Fig. 4D). However, OE FoxG1+shBmi1 cells displayed a significant reduction in self-renewal capacity compared to control cells, suggesting FoxG1 overexpression to not completely rescue the reduced self-renewal capacity observed with Bmi1 knockdown (Fig. 4D).

FoxG1 and Bmi1 Show Reciprocal Promoter Occupancy Only in CD15+ MB Stem Cells

Although FoxG1 and Bmi1 have previously been shown to interact with one another through correlative overexpression and knockdown studies of either gene [42] and in our hands have been implicated in co-operating with one another to regulate self-renewal in MB tumor spheres, the mechanism for their interaction has yet to be elucidated. We hypothesized that both proteins may regulate each other's transcriptional activity through binding at one another's promoter. Consequently, we performed ChIP experiments to assess the enrichment of FoxG1 at the Bmi1 promoter (Fig. 5A) and Bmi1 at the FoxG1 promoter (Fig. 5B). We observed no difference for enrichment at either promoter and hypothesized that this may in fact resemble the current inability of genomic profiling of bulk tumors to identify stem cell genes preferentially expressed in poor outcome MBs, since our ChIP experiments were performed on unsorted tumor sphere populations that had not been sort-enriched for putative MB stem cells. Having validated CD15 protein expression as a marker of aggressive, poor-outcome subgroup 3 and 4 MBs from 74 MBs (supporting information Fig. 6), we performed flow cytometric cell sorting for CD15 [38, 39], a marker which allows for sorting of distinct and reproducible positive and negative populations across multiple MB model systems and cell lines (supporting information Fig. 3). We then assessed FoxG1 and Bmi1 transcript levels in CD15+ and CD15− cells and found that both genes were significantly expressed in the CD15+ cell fraction when compared to the CD15− cells (Fig. 5C, 5D). We then performed our ChIP experiments on CD15-sorted MB tumor cells and found a marked enrichment for FoxG1 at the Bmi1 promoter (Fig. 5E) and Bmi1 at the FoxG1 promoter (Fig. 5F) in the CD15+ cell fraction with no significant enrichment at either promoter in the CD15− population.

thumbnail image

Figure 6. Bmi1 is a novel downstream target of FoxG1, through which it exerts an increase in tumorigenicity. (A): Bmi1 levels are reduced and increased following shFoxG1 and OE FoxG1, respectively, when compared to stable control cells (n = 3, p = .0063; n = 3, p = .0001, respectively). OE FoxG1+shBmi1 resulted in a significant decrease in Bmi1 levels compared to control cells (n = 3, p = .0001), revealing that FoxG1 overexpression may not be sufficient to rescue Bmi1 knockdown at the transcript level. (B): A positive feedback with Bmi1 on FoxG1 expression was demonstrated with a reduction in FoxG1 levels compared to control cells with shBmi1 (n = 3, p = .0209) and an elevation in FoxG1 expression within OE FoxG1+shBmi1 cells (n = 3, p = .0143). (C): In vivo analysis (top panel, ×20; bottom panel, ×100) demonstrated OE FoxG1 MB stem cells to generate much larger and infiltrative tumors compared to smaller, well-circumscribed control tumors. Interestingly, OE FoxG1+shBmi1 tumors were smaller than OE FoxG1 tumors and more circumscribed than control tumors. (D): Model of FoxG1 regulation of Bmi1 expression and Bmi1 feedback on FoxG1 expression to promote in vivo tumor aggressiveness and enhanced in vitro stem cell self-renewal. Scale bar = 100 mm. *, p < .05; **, p < .0001.

Download figure to PowerPoint

The cell cycle inhibitors p16 and p21 are well-defined downstream targets for repression by FoxG1 and Bmi1 to promote self-renewal in a number of tissues and malignancies [42, 65-69]. We found both p16 (Fig. 5G) and p21 (Fig. 5H) levels to be significantly elevated in the CD15− cell fraction when compared to CD15+ MB stem cells, suggesting a novel mechanism for MB stem cell self-renewal by which FoxG1 and Bmi1 may interact with one another to co-operatively inhibit p16 and p21 in CD15+ MB stem cells.

In Vitro and In Vivo Functional Characterization of FoxG1/Bmi1 Interactions

In order to further interrogate the mechanistic significance of FoxG1 and Bmi1 sharing promoter occupancy, we investigated the expression of Bmi1 levels in our stable shFoxG1, OE FoxG1, and OE FoxG1+shBmi1 cell lines. Interestingly, Bmi1 levels were reduced and increased following FoxG1 knockdown and overexpression, respectively, when compared to stable control cells (Fig. 6A). These data suggest Bmi1 to be a novel downstream target for FoxG1 signaling. Following OE FoxG1+shBmi1, we observed a significant decrease in Bmi1 levels when compared to control cells, revealing that FoxG1 overexpression may not be sufficient to rescue Bmi1 knockdown at the transcript level. Bmi1 regulation of FoxG1 was assessed following shBmi1 (Fig. 6B), and a positive feedback was identified as FoxG1 expression was significantly reduced in shBmi1 cells compared to control. Moreover, OE FoxG1+shBmi1 cells displayed elevated FoxG1 levels compared to control. These data allude to the presence of additional novel downstream targets of FoxG1 signaling aside from Bmi1, which may have a stronger affinity for positively regulating FoxG1 expression.

In vivo analysis of FoxG1 and Bmi1 functional relevance on MB stem cell-driven tumorigenesis demonstrated OE FoxG1 MB stem cells to generate much larger and infiltrative tumors compared to smaller, well-circumscribed control tumors (Fig. 6C). In vitro self-renewal analysis showed a reduction in self-renewal capacity in OE FoxG1+shBmi1 cells compared to OE FoxG1 and control cells (supporting information Fig. 2). This finding mirrored in vivo characteristics of tumor size and invasiveness, as OE FoxG1+shBmi1-generated tumors were smaller than OE FoxG1 tumors and more circumscribed than control tumors. Our previously published in vivo findings have implicated Bmi1 in regulating tumor size in MB [36] glioblastoma (GBM) [70], and it appears that Bmi1 may also critically regulate tumor size and invasiveness downstream of FoxG1.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

The classification of MB has progressively evolved to include a combination of clinico-pathological and molecular factors that reflect the clinically relevant and potentially prognostic aspects of tumor biology [4]. Our understanding of the pathogenesis and basic biology of MB has been advanced considerably by the application of integrated genomics platforms to characterize molecular subgroups [6–10]. It has since been widely accepted that a combination of clinical and molecular factors will afford a more reliable means of assigning disease risk in patients with MB [71, 72], allowing for a de-escalation of current therapy and a progression toward risk-tailored, individualized, and targeted patient therapy.

Herein, we have characterized the functional and clinical significance of two BTIC-specific genes, FoxG1 and Bmi1 in regulating self-renewal of MB stem cells. Since increasing BTIC frequency is associated with decreased survival and poor patient outcome [70], we hypothesized that genes governing BTIC self-renewal may segregate aggressive, poor-outcome MBs from other subgroups. We applied this clinical outcomes-driven rationale to 251 primary human MBs through several published nonoverlapping MB transcriptional databases [6-8, 10] and 74 recently NanoString-subgrouped primary MBs for candidate genes previously implicated in CSC and BTIC populations. This allowed us to ascertain a differential stem cell gene expression profile within Wnt, Shh, and non-Shh/Wnt MB subgroups. A subsequent series of step-wise knockdown, overexpression, in vitro, in vivo, and ChIP analyses have allowed us to demonstrate for the first time that FoxG1 and Bmi1 are preferentially expressed in MB stem cells and co-operatively function to regulate their self-renewal and tumorigenicity (Fig. 6D). Our model illustrates that Bmi1 is a novel downstream target of FoxG1, through which it exerts an increase in stem cell self-renewal and tumorigenicity by inhibition of p16 and p21. This finding is of great clinical significance as BTIC self-renewal genes and their regulatory pathways may serve as novel factors used to characterize poor-outcome non-Shh/Wnt MBs in future studies.

Our ChIP data demonstrates that a differential transcriptional regulatory mechanism exists between unsorted and sort-enriched stem cell populations, in which Bmi1 and FoxG1 show reciprocal promoter binding only in CD15+ MB stem cells. Our data further suggests this difference to repress cell cycle inhibitors, p16 and p21, in CD15+ MB stem cells, permitting extensive self-renewal and proliferation in this cell fraction, while CD15− cells are maintained under the control of these cell cycle inhibitors. Recently, we have identified additional cell–cell differences between MB BTIC and non-BTICs based on the characterization of Shh-receiving and Shh-signaling cells, respectively [36]. In GBM BTICs, we have observed time-dependent functional differences in Bmi1 based on the stage of BTIC differentiation and cell type as Bmi1 contributes to self-renewal in BTICs but regulates proliferation and cell fate determination in non-BTICs [70]. Given these critical differences in sort-enriched BTIC and non-BTIC cell compartments of a heterogeneous tumor such as brain tumors, future characterization of MB functional genomics platforms would benefit from subgrouping based on sort-enriched MB tumor cells for elucidating novel pathways that may be aberrantly activated in MB BTICs.

The study of BTICs has reconceptualized the heterogeneity of brain tumors in terms of their biological framework and predicted therapeutic response. However, several limitations remain in our ability to characterize these rare clonal populations of cells and this is particularly true for a rare tumor such as MB. Although our stem cell profiling technique shows promise to further characterize current high-risk MBs, the small primary sample size for our in vitro and in vivo functional studies limits the true determination of its specificity as a unique and independent identifier of high-risk disease. Although several stem cell self-renewal genes (such as Nanog, Oct4, and Lgr5) did not show enriched expression in high-risk non-Shh/Wnt MBs, our findings illustrate that regulation of self-renewal may be regional and context-dependent, explaining why some self-renewal genes critical to embryonic or epithelial stem cells may not be active in a neural microenvironment. Primary human MB BTIC cultures are technically challenging, provide limited cell numbers for data acquisition, and require specific infrastructure; therefore, our platform is unlikely to be widely adapted for routine laboratory use at this point. However, continued study of larger numbers of human MB BTIC specimens will eventually elucidate key stem cell signaling pathways and molecular mechanisms of self-renewal that could provide specific targets for poor-outcome MBs. Specifically, as current transcriptional profiling of MB consists of the cellular bulk tumor population, rare stem cell fractions may be missed and therefore, future transcriptional analyses of stem cell genes may benefit from tissue banks consisting of prospectively sorted BTIC RNA. Current practice for MB molecular profiling from a single tumor biopsy or resection sample may also underestimate the genomic landscape [34], providing the BTIC model system as an ideal approach for addressing intratumor heterogeneity.

CONCLUSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Our work provides encouraging evidence to suggest that identifying and subsequently modulating stem cell populations in MB may improve clinical outcome for those patients with metastatic and high-risk disease. Future implications of our model system may take advantage of our in vitro BTIC self-renewal assay, which has previously been validated to correlate with clinical outcome in pediatric brain tumors, including MB [24]. With the recent application of the NanoString assay to MB molecular subgrouping [73], academic hospitals may discern low-level stem cell gene transcript levels within paraffin-embedded tissues in conjunction with our established in vitro self-renewal assays. Such analyses may assist in rapidly determining the stem cell phenotype of a given MB and may in turn promote more targeted and individualized treatment for a patient-specific BTIC profiles.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

B.M. is supported by a Canadian Institutes of Health Research Vanier Canada Graduate Scholarship. S.K.S. is supported by the Neurosurgical Research and Education Foundation and American Association of Neurological surgeons, Pediatric Section, the Ontario Institute for Cancer Research, and McMaster University Department of Surgery.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
sc-12-0914_sm_SupplFigure1.pdf866KSupplementary Figure 1: FoxG1 and Bmi1 knockdown lead to an efficient reduction in their transcript and protein levels. shRNA-mediated knockdown of (a) FoxG1 revealed a significant knockdown in shFoxG1-2 construct when compared to cells transduced with a control shRNA construct (n=3, P=0.0047). All other constructs, shFoxG1-1, shFoxG1-3, shFoxG1-4 had varying levels of knockdown but were not significant at transcript level (n=3, P=0.5669, n=3, P=0.2636 and n=3, P=0.1425, respectively). (b) shRNA-mediated knockdown of (b) Bmi1 revealed a significant knockdown in shBmi1-3 construct when compared to cells transduced with a control shRNA construct (n=3, P=0.001). All other constructs, shBmi1-1 and shBmi1-2, had varying levels of knockdown but were not significant at transcript level (n=3, P=0.7923, and n=3, P=0.2064, respectively). (c) Overexpression construct for FoxG1 revealed a significant increase in FoxG1 expression compared to cells transduced with a control construct (n=3, P=0.0018). **P<0.05, ***P<0.005
sc-12-0914_sm_SupplFigure2.pdf1474KSupplementary Figure 2: Functional characterization of FoxG1 knockdown/overexpression and Bmi1 knockdown constructs through in vitro limiting dilution assays. Characterization of (a) 4 shFoxG1 and (b) 3 shBmi1 constructs with (c) shFoxG1-2 and shBmi1-3 showing the greatest reduction in sphereforming capacity. (d) Cells transduced with OE FoxG1 and OE FoxG1+shBmi1 constructs displayed an increase and decrease in sphere-forming capacity, respectively.
sc-12-0914_sm_SupplFigure3.pdf1231KSupplementary Figure 3: Representative flow cytometric cell sorting plot for CD15 in Daoy and Med8a medulloblastoma stem cells. (a) Flow plot showing side scatter against forward scatter for Daoy MB stem cells. (b) Flow plot of sort sample showing 5.16% of cells being CD15+. (c) Flow plot showing side scatter against forward scatter for Med8a MB stem cells. (d) Flow plot of sort sample showing 43.29% of cells being CD15+.
sc-12-0914_sm_SupplFigure4.pdf1387KSupplementary Figure 4: FoxG1 and Bmi1 are differentially enriched in Med8a medulloblastoma cells. (a) FoxG1 (n=3, P=0.00044) and (b) Bmi1 (n=3, P=0.0031) expression are significantly elevated in Med8a MB cells grown in NSC conditions compared to normal human NSCs (n=3). (c) FoxG1 is specifically enriched in CD15+ Med8a MB stem cells compared to CD15- cells (n=5, P=0.00041), however, (d) no difference is observed in Bmi1 expression between CD15+ and CD15- cells (n=5, P=0.7332). (e) Flow plot (left) showing side scatter against forward scatter for Med8a MB stem cells along with (right) FoxG1 and Bmi1 levels showing 99.39% positivity for FoxG1 and 3.12% positivity for Bmi1. **P<0.005, ***P<0.0005
sc-12-0914_sm_SupplFigure5.pdf1005KSupplementary Figure 5: FoxG1 regulation of MB BTIC differentiation. A general trend demonstrating a (a) decrease in GFAP, MAP2, and OLIG2 expression was observed in OE FoxG1 cells compared to control cells (n=3), whereas an (b) increase in neural differentiation marker expression was observed in shFoxG1 cells when compared to control cells (n=3). (c) Additionally, FoxG1 expression was found to decrease with differentiation in primary human MB BTICs (n=3, P=0.0482). *P<0.05
sc-12-0914_sm_SupplFigure6.pdf2080KSupplementary Figure 6: Putative MB BTIC marker, CD15, is preferentially expressed in subgroup 3 and 4 MBs. Representative images of differential cytoplasmic CD15 expression in (a) Wnt, (b) Shh, (c) Group 3, and (d) Group 4 MBs. (e) Graphical representation of low/no cytoplasmic CD15 staining to high cytoplasmic positivity within MB subgroups. CD15 staining is marked in non-Shh/Wnt subgroups. (Magnification: ×200)
sc-12-0914_sm_SupplTable1.pdf13KSupplementary Table 1: Candidate gene list
sc-12-0914_sm_SupplTable2.pdf12KSupplementary Table 2: Statistical evaluation of candidate genes
sc-12-0914_sm_SupplTable3.pdf7KSupplementary Table 3: Medulloblastoma stem cell patient isolates: Clinicopathological data
sc-12-0914_sm_SupplTable4.pdf8KSupplementary Table 4: qRT-PCR & ChIP primers

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.