High-grade glioma (HGG) is a group of primary malignant brain tumors with dismal prognosis. Whereas adult HGG has been studied extensively, childhood HGG, a relatively rare disease, is less well-characterized. Here, we present two novel platelet-derived growth factor (PDGF)-driven mouse models of pediatric supratentorial HGG. Tumors developed from two different cells of origin reminiscent of neural stem cells (NSC) or oligodendrocyte precursor cells (OPC). Cross-species transcriptomics showed that both models are closely related to human pediatric HGG as compared with adult HGG. Furthermore, an NSC-like cell-of-origin enhanced tumor incidence, malignancy, and the ability of mouse glioma cells (GC) to be cultured under stem cell conditions as compared with an OPC-like cell. Functional analyses of cultured GC from these tumors showed that cells of NSC-like origin were more tumorigenic, had a higher rate of self-renewal and proliferation, and were more sensitive to a panel of cancer drugs compared with GC of a more differentiated origin. These two mouse models relevant to human pediatric supratentorial HGG propose an important role of the cell-of-origin for clinicopathologic features of this disease. Cancer Res; 77(3); 802–12. ©2016 AACR.

Pediatric high-grade glioma (HGG) is a group of aggressive primary central nervous system (CNS) tumors diagnosed in children (age 0–19 years) that comprises astrocytomas of WHO grade III (anaplastic astrocytoma) and IV (glioblastoma; ref. 1). Astrocytomas are diffusely growing tumors of presumable glial origin and are the most common primary malignant brain tumors of adults. Pediatric HGGs are, however, relatively rare and represent around 8%–16% of all CNS tumors in children (2–4). A majority of pediatric HGGs arise in the brainstem and are called diffuse intrinsic pontine glioma (DIPG), but a large fraction of these tumors are supratentorial and located in the cerebral hemispheres. In this article, we refer to these tumors as pediatric HGGs. Similar to adult HGGs there is no efficient treatment available. Standard treatment of care for pediatric supratentorial HGG includes maximal safe surgical resection followed by focal radiation and chemotherapy, which extends survival but rarely is curative. This makes pediatric HGGs among the most lethal cancers in children with a 5-year survival of 20%–30% (2).

Pediatric HGGs are like adult HGGs highly heterogeneous. Substantial progress has been made during the last decade in understanding the biology and genetics of pediatric HGG. They share genetic similarities with adult HGGs like alterations in receptor tyrosine kinases (RTK), activation of RTK signaling pathways, and loss of function of p53 and RB suppressor pathways (5, 6), but perhaps more importantly there are distinct differences, both in the frequency and type of mutations and copy number alterations (CNA). Mutations in PDGFRA, H3F3A, ATRX, and TP53 are significantly more common in pediatric tumors while alterations in EGFR and PTEN are more common in adult HGGs (6–8). Pediatric HGGs also have fewer and distinct CNAs (9) where amplifications of PDGFRA are the most common genomic event (5, 10).

Pediatric HGGs can be divided into molecular subgroups. Unsupervised hierarchical clustering of gene expression profiles from 53 human pediatric HGGs produced three subgroups of tumors (HC1–3) where the HC1 group was significantly enriched for PDGFRA and/or PDGFB amplifications and the HC3 group had a significant underrepresentation of gain of 1q (5). Furthermore, the pediatric HGG gene expression data was compared with previously published adult glioblastoma expression profiles (11, 12) that showed a clear separation but with a certain overlap between adult and pediatric HGGs (5). Recently, a classification based on DNA methylation that included both adult and pediatric glioblastoma was presented (8). Six subgroups of glioblastoma were identified (IDH1, K27, G34, RTK1 “PDGFRA,” Mesenchymal, RTKII “Classical”), of which only RTKII was devoid of pediatric tumors. The majority of pediatric tumors were found in the K27, G34, and RTKI groups where the two first are defined by different mutations in H3F3A. The epigenetic subgroups also showed a difference in tumor location where K27 tumors were mainly found in the midline (thalamus, brainstem, spinal cord) while all other subgroups were located in the cerebral hemispheres. The molecular characterization shows that pediatric HGG is a highly heterogeneous disease and that PDGFRA is an important driver of a subset of supratentorial pediatric HGGs.

The fact that HGGs can be found in many different locations of the brain suggests that several different cell types in various parts of the brain may develop HGG. If so, a reasonable hypothesis would be that the cell-of-origin is contributing to the heterogeneous HGG phenotype. The identity of the cell-of-origin for HGG remains unidentified but suggested by histopathology, marker expression, and gene expression of HGG to be a neural or glial stem cell. Mouse modeling has shown that many different CNS cell types have the capacity to develop glioma (13–17). Most mouse models of HGG have, however, only been related to adult HGG (17–19), and there are few characterized mouse models of pediatric supratentorial HGGs. Here we have investigated the influence of two distinct and plausible glial cells of origin for pediatric HGG development and HGG cell phenotype. We have modeled supratentorial PDGFRA-driven HGG and have compared the effect of developmental age and differentiation state of the cell-of-origin for tumor development, analyzed the relation of our models to human pediatric HGGs, and investigated the role of the cell-of-origin for tumor development and the glioma cell (GC) phenotype.

Additional descriptions of immunostainings, mouse GC and neural stem cell (NSC) culturing, proliferation and differentiation analyses, and drug response testing are available in Supplementary information.

Infection of t-va transgenic mice and secondary tumor induction in syngeneic mice

All animal experiments were performed according to Uppsala University rules and after approval by the local animal ethics committee. G/tv-a wt (20), G/tv-a;Arf−/− (21), C/tv-a wt (15), and C/tv-a;Arf−/− (22) transgenic mouse lines were used. Mice were infected at postnatal day (P) 0–2 (neonatal) or at 6–8 weeks of age (adult) through injections of 105 DF-1 cells (in 2 μL PBS) producing RCAS-eGFP or RCAS-PDGFB-HA. Coordinates in neonatal mice were 1.5 mm anterior of lambda, 0.7 mm lateral, and 1.5 mm ventral for intraventricular injection close to the subventricular zone (SVZ; ref. 23). Coordinates in adult mice were 0.5 mm anterior of bregma, 1.1 mm lateral, and 2.5 mm ventral to hit the SVZ. Mice injected with RCAS-eGFP were sacrificed 7 days later. Mice injected with RCAS-PDGFB-HA were monitored and killed upon symptoms of illness or at the endpoint of 15 weeks postinjection. PDGFB-induced tumor incidence is summarized in Table 1. 

Table 1.

Tumor incidence in response to RCAS-PDGFB-HA injected into the SVZ of neonatal or adult G/tv-a and C/tv-a wt and Arf−/− mouse lines

Mouse lineIncidence, neonatalIncidence, adult
G/tv-a wt 100% (19/19) 30.8% (4/13) 
G/tv-a;Arf−/− 100% (24/24) 100% (16/16) 
C/tv-a wt 36% (8/22) 0% (0/10) 
C/tv-a;Arf−/− 75% (36/48) 83% (15/18) 
Mouse lineIncidence, neonatalIncidence, adult
G/tv-a wt 100% (19/19) 30.8% (4/13) 
G/tv-a;Arf−/− 100% (24/24) 100% (16/16) 
C/tv-a wt 36% (8/22) 0% (0/10) 
C/tv-a;Arf−/− 75% (36/48) 83% (15/18) 

GCGFAP or GCCNP cells (105 in 2 μL PBS) were injected in P0-2 syngeneic mice as described above. Mice were monitored and killed upon symptoms of illness or at the endpoint of 20 weeks.

Histopathology analysis of mouse brain tumors

Mouse brain tissue was fixed in 4% formalin for at least 48 hours and embedded in paraffin. Six-micron sections were cut using a Microtome (HM 400 machine) and hematoxylin and eosin stained for tumor grading. Histopathologic analyses were performed in blind under the supervision of neuropathologists (I. Alafuzoff and S. Libard) and were based on the 2007 WHO classification. Features evaluated were cell density, percentage of proliferative cells (Ki67), presence of apoptotic cells, necrosis, mitotic cells, and endothelial cell proliferation in the tumor region. Ki67 immunostainings were performed on at least one section from each mouse tumor.

Mouse GC and CC cell cultures, and differentiation and proliferation analyses

All cell cultures used were newly produced for this study in the Uhrbom laboratory. Tumor and control tissues from SVZ of uninjected G/tv-a;Arf−/− and C/tv-a;Arf−/− mice were explanted and cultured under stem cell conditions (24). GCs were cultured without addition of EGF and FGF2 and CCs (NSC) were cultured in the presence of EGF and FGF2. All experiments using GCs were performed on cells in passages 10–20. The cell cultures have not been authenticated but all GC and CC cultures were tested negative for mycoplasma at two different time points (at passages 4–15 and passages 14–23) using two different methods, MycoAlert (Lonza, catalog # LT07-418) and a PCR-based method using primers (Invitrogen) Myco1 (5′-GGCGAATGGGTGAGTAACACG-3′) and Myco2 (5′-CGGATAACGCTTGCGACTATG-3′) using the KAPA kit (Techtum, catalog # 25-KK7352).

Differentiation was induced by adding 5% FBS (Sigma) to the medium for 7 days. Bromodeoxyuridine BrdU (Sigma) was added 16 hours prior to fixation. Cells were stained for BrdU and neural/glial markers. Pictures were taken using ZEN software in LSM700 confocal microscope (Carl Zeiss). Quantifications were done by counting around 100 cells from each coverslip.

Limiting dilution assay

GCs and CCs were detached using Acutase (Invitrogen) and seeded in 96-well ultra-low attachment plates (Corning) in 0.2 mL NSC medium. Seeding density ranged from 100 cells to 1 cell per well and 10 replicates were used for each cell line. Cells were left undisturbed for 7–10 days. The number of wells without spheres for each cell density was plotted against the number of cells in each well using GraphPad Prism 6.

Microarray analysis and preprocessing of the gene expression dataset

Total mRNA was extracted from HGG tissue samples and GCs along with control tissues (SVZ) and CCs using miRNeasy Mini Kit (Qiagen). RNA was labeled and hybridized on Affymetrix Mouse Gene ST 1.0 arrays at the SciLifeLab Array and Analysis Facility. Data was processed using Robust Multichip Average (RMA; ref. 25) for background correction, quantile normalization, and expression values from multiple probes corresponding to a single gene were merged. The average expression value was used to represent the gene.

Principal component analysis of gene expression data

The dataset was loaded into RStudio version 0.98.1103 (26) running R v.3.2.2 (27), and the R function prcomp with default values was used on the samples to perform the principal component analysis (PCA). The R package plot3D was used for visualization (28).

Comparison of our mouse array data to human pediatric and adult HGG samples

To compare the mouse gene expression dataset with published human HGG/GBM data, ortholog genes were matched using HUGO Gene Nomenclature Committee's (HGNC; ref. 29) Comparison of Orthology Predictions (HCOP; ref. 30) and gene names were translated according to the HGNC. Gene expression data from HGG tumor samples [Paugh and colleagues (5), Lee and colleagues (12), and Beroukhim and colleagues (11)] were downloaded from Gene Expression Omnibus (31; Series GSE19578, GSE13041) and from www.broad.mit.edu/cancer/pub/GISTIC. Gene names were harmonized according to HGNC. The datasets were compared together with our mouse dataset and an expression matrix of metagenes was generated using the metagene code for cross-platform, cross-species characterization of transcriptional profiles (32).

Analysis of drug response in mouse GCs

Mouse GCs (three GCGFAP and one GCCNP) and CCs (one CCG/tv-a and one CCC/tv-a) were tested for their response to a 72-hour treatment of 28 compounds (Table 2) using duplicate samples and repeated twice. Cell viability was measured, dose–response curves plotted, and the area under the curve (AUC) value calculated for each drug and cell sample. The fold change between GCGFAP samples and GCCNP samples were calculated on the basis of the average AUC value for each drug. The Welch t test with Benjamini–Hochberg correction was used to measure the statistical significance of the difference between the two groups. To see a trend, the base-10 logarithm of the fold change was plotted along with the corresponding negative base-10 logarithm of the adjusted P value. The scatter plot was drawn in Mathworks Matlab version R2015A (33).

Table 2.

Compounds used to test and compare the drug response of GCGFAP and GCCNP cells as measured by cell viability

Drug typeDrug nameResponse difference, BH adjusted P valueTarget(s)a
Tyrosine kinase inhibitor Imatinib 0.021 (*) BCR/ABL, KIT, RET, NTRK1, CSF1R, PDGFRA, DDR1, ABL1, PDGFRB 
 Sunitinib 0.040 (*) PDGFRB, FLT1, KIT, KDR, FLT4, FLT3, CSF1R, PDGFRA 
 Sorafenib 0.032 (*) BRAF, RAF1, FLT4, KDR, FLT3, PDGFRB, KIT, FGFR1, RET, FLT1 
 Dasatinib 0.016 (*) ABL1, SRC, EPHA2, LCK, YES1, KIT, PDGFRB, STAT5B, ABL2, FYN 
 Gefitinib 0.038 (*) EGFR 
 Lapatinib 0.500 (ns) EGFR, ERBB2 
 Crizotinib 0.589 (ns) ALK, MET 
 Erlotinib 0.589 (ns) EGFR, NR1I2 
mTOR inhibitor BEZ-235 0.026 (*) MTOR, PIK3CA, PIK3CG, PIK3CD, PIK3CBb 
 Rapamycin 0.020 (*) MTOR, FKBP1A, FGF2 
Antimetabolite 5-azacytidine 0.076 (ns) DNMT1, DNA, RNA 
 5-fluorouracil 0.016 (*) TYMS, DNA, RNA 
 6-thioguanine 0.071 (ns) DNA 
 Gemcitabine 0.039 (*) DNA, RRM1, TYMS, CMPK1 
Topoisomerase inhibitor Doxorubicin 0.020 (*) DNA, TOP2A 
 Etoposide 0.289 (ns) TOP2A, TOP2B 
 Irinotecan 0.020 (*) TOP1MT, TOP1 
Tubulin inhibitor Mebendazole 0.052 (ns) TUBA1A, TUBB4B 
 Vincristine 0.589 (ns) TUBB, TUBA4A 
Rock inhibitor GSK269962A 0.450 (ns) ROCK1, ROCK2b 
Platinum agent Oxaliplatin 0.021 (*) DNA 
Alkylating agent Melphalan 0.023 (*) DNA 
 Procarbazine 0.289 (ns) DNA 
 Temozolomide 0.153 (ns) DNA 
 Lomustine 0.721 (ns) DNA, STMN4 
 Melflufen 0.020 (*) DNAc 
Proteasome inhibitor Bortezomib 0.714 (ns) PSMB5, PSMB1 
 b-AP15 0.076 (ns) USP14, UCHL5b 
Drug typeDrug nameResponse difference, BH adjusted P valueTarget(s)a
Tyrosine kinase inhibitor Imatinib 0.021 (*) BCR/ABL, KIT, RET, NTRK1, CSF1R, PDGFRA, DDR1, ABL1, PDGFRB 
 Sunitinib 0.040 (*) PDGFRB, FLT1, KIT, KDR, FLT4, FLT3, CSF1R, PDGFRA 
 Sorafenib 0.032 (*) BRAF, RAF1, FLT4, KDR, FLT3, PDGFRB, KIT, FGFR1, RET, FLT1 
 Dasatinib 0.016 (*) ABL1, SRC, EPHA2, LCK, YES1, KIT, PDGFRB, STAT5B, ABL2, FYN 
 Gefitinib 0.038 (*) EGFR 
 Lapatinib 0.500 (ns) EGFR, ERBB2 
 Crizotinib 0.589 (ns) ALK, MET 
 Erlotinib 0.589 (ns) EGFR, NR1I2 
mTOR inhibitor BEZ-235 0.026 (*) MTOR, PIK3CA, PIK3CG, PIK3CD, PIK3CBb 
 Rapamycin 0.020 (*) MTOR, FKBP1A, FGF2 
Antimetabolite 5-azacytidine 0.076 (ns) DNMT1, DNA, RNA 
 5-fluorouracil 0.016 (*) TYMS, DNA, RNA 
 6-thioguanine 0.071 (ns) DNA 
 Gemcitabine 0.039 (*) DNA, RRM1, TYMS, CMPK1 
Topoisomerase inhibitor Doxorubicin 0.020 (*) DNA, TOP2A 
 Etoposide 0.289 (ns) TOP2A, TOP2B 
 Irinotecan 0.020 (*) TOP1MT, TOP1 
Tubulin inhibitor Mebendazole 0.052 (ns) TUBA1A, TUBB4B 
 Vincristine 0.589 (ns) TUBB, TUBA4A 
Rock inhibitor GSK269962A 0.450 (ns) ROCK1, ROCK2b 
Platinum agent Oxaliplatin 0.021 (*) DNA 
Alkylating agent Melphalan 0.023 (*) DNA 
 Procarbazine 0.289 (ns) DNA 
 Temozolomide 0.153 (ns) DNA 
 Lomustine 0.721 (ns) DNA, STMN4 
 Melflufen 0.020 (*) DNAc 
Proteasome inhibitor Bortezomib 0.714 (ns) PSMB5, PSMB1 
 b-AP15 0.076 (ns) USP14, UCHL5b 

Abbreviation: BH, Benjamini-Hochberg.

aInformation from DrugBank.

bInformation from Selleckchem.

cInformation from Oncopeptides.

Statistical analyses

Statistical analyses were performed in GraphPad Prism 6 (GraphPad Software, Inc.). For comparisons of tumor incidence and tumor grade between groups, Contingency tables and Fisher exact test was used. All other group comparisons were done using Student t test. Figures plotted from multiple experiments are presented as mean ± SD. Survival significance was evaluated using log-rank tests where the values of P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), and P < 0.0001 (****) indicate statistical significance.

Different cell types can be targeted by RCAS in the subventricular zone of neonatal G/tv-a and C/tv-a mice

We have used the RCAS/tv-a mouse model to generate HGG, where oncogene-carrying RCAS viruses are used to exclusively infect tv-a–expressing cells (20). We used the G/tv-a and the C/tv-a mouse lines in which the tv-a gene is under the control of the glial fibrillary acidic protein (GFAP; ref. 20) or the 2′, 3′-cyclic nucleotide 3′-phosphodieasterase (CNP; ref. 15) promoters, respectively. GFAP is expressed in NSCs (13) and astrocytes (14, 34) and CNP is a highly specific marker for late oligodendrocyte precursor cells (OPC) and oligodendrocytes (15, 16, 35). The mice were either wild-type (G/tv-a wt and C/tv-a wt) or carrying a targeted deletion of the Cdkn2a locus (G/tv-a;Arf−/− and C/tv-a;Arf−/−; refs. 21, 36, 37).

The NSC compartment, in this case only referring to the SVZ, has been suggested to be involved in HGG development of both children and adults (38). To direct RCAS infection to cells of the NSC compartment, neonatal mice were injected with DF-1 cells producing RCAS-eGFP in close proximity to the SVZ. Injected mice were killed 7 days later and their brains were analyzed with the purpose to characterize the infected cells. To separate between infected GFP-expressing glial cells and the injected GFP-expressing DF-1 chicken cells, we double-stained for chicken vimentin and GFP (Supplementary Fig. S1). DF-1 cells were mostly found in clumps and were intensely bright with small, round morphology, while infected mouse cells were bigger and displayed varied morphologies and cytoplasmic extensions. Only those larger, heterogeneous glial cells were included in the subsequent quantification of GFP-infected cells. Next, we performed double immunofluorescence stainings for different glial markers (GFAP, CNP, nestin, SOX2, and OLIG2) and GFP on several brain sections throughout the brains of infected mice (Fig. 1). The absolute majority of infected cells were found around the SVZ. For GFAP (Fig. 1A and B) and CNP (Fig. 1C and D) we found, as expected, that the majority of infected cells in G/tv-a mice were positive for GFAP and negative for CNP, and vice versa in C/tv-a mice. Analysis of the neural stem and progenitor cell marker nestin (NES) showed, similar to GFAP, that a high number of infected cells were nestin positive in G/tv-a mice, while essentially no infected cells were nestin positive in C/tv-a mice (Fig. 1E and F). For the glial stem and progenitor cell marker SOX2, we found positive cells in both mouse lines although significantly more in G/tv-a mice (Fig. 1G and H). The opposite was found for the glial progenitor cell marker OLIG2 that was significantly more abundant in C/tv-a mice (Fig. 1I and J).

Figure 1.

Characterization of RCAS-eGFP–infected cells in G/tv-a wt and C/tv-a;Arf−/− neonatal mice injected in close proximity to the SVZ. At least one section from at least two different mouse brains were analyzed for each group. The ratio of cells double positive for GFP and each of the neural/glial cell lineage markers in relation to the total number of infected mouse cells (i.e., GFP-positive cells with neural or glial morphology) was calculated. Values represent the mean ± SD. Scale bars, 25 μm. White arrows, cells double positive for GFP and the cell marker. Immunofluorescence staining for GFAP and GFP (A), and the result from the quantification (Student t test, P = 0.0001; B). Immunofluorescence staining for CNP and GFP (C), and the result from the quantification (Student t test, P = 0.0001; D). Immunofluorescence staining for NES and GFP (E), and the result from the quantification (Student t-test, P = 0.0001; F). Immunofluorescence staining for SOX2 and GFP (G), and the result from the quantification (Student t test, P = 0.0002). Immunofluorescence staining for OLIG2 and GFP (I), and the result from the quantification (Student t-test, P = 0.0001; J).

Figure 1.

Characterization of RCAS-eGFP–infected cells in G/tv-a wt and C/tv-a;Arf−/− neonatal mice injected in close proximity to the SVZ. At least one section from at least two different mouse brains were analyzed for each group. The ratio of cells double positive for GFP and each of the neural/glial cell lineage markers in relation to the total number of infected mouse cells (i.e., GFP-positive cells with neural or glial morphology) was calculated. Values represent the mean ± SD. Scale bars, 25 μm. White arrows, cells double positive for GFP and the cell marker. Immunofluorescence staining for GFAP and GFP (A), and the result from the quantification (Student t test, P = 0.0001; B). Immunofluorescence staining for CNP and GFP (C), and the result from the quantification (Student t test, P = 0.0001; D). Immunofluorescence staining for NES and GFP (E), and the result from the quantification (Student t-test, P = 0.0001; F). Immunofluorescence staining for SOX2 and GFP (G), and the result from the quantification (Student t test, P = 0.0002). Immunofluorescence staining for OLIG2 and GFP (I), and the result from the quantification (Student t-test, P = 0.0001; J).

Close modal

Taken together, this analysis showed that different cell types in the SVZ of neonatal G/tv-a and C/tv-a mice were targeted by RCAS. In G/tv-a mice, infected cells showed an apparent immature, neural, or glial stem cell phenotype (GFAP+/NES+/SOX2+/CNP−/OLIG2+), where the few OLIG2+ cells that were present likely were a result of that differentiation along the oligodendrocyte axis had occurred in the week after infection (39, 40). The marker expression of RCAS-infected cells in the C/tv-a mouse brains (GFAP−/NES−/SOX2+/CNP+/OLIG2+) showed a clearly differentiated phenotype toward the OPC lineage.

Developmental age and cell-of-origin affect glioma development from GFAP- and CNP-expressing cells in mouse SVZ

To analyze the role of cell-of-origin for glioma development in the context of developmental age, we injected DF-1 cells producing RCAS-PDGFB-HA in neonatal and adult G/tv-a wt, G/tv-a;Arf−/−, C/tv-a wt, and C/tv-a;Arf−/− mice. Neonatal mice were injected in the same location as above and adult mice were injected in the SVZ. Upon symptoms of illness or at the experimental endpoint of 15 weeks, the animals were killed and brain tumors scored by viewing at least five hematoxylin and eosin–stained sections from each mouse brain (Table 1). All tumors were supratentorial and no tumors were detected in the brainstem.

We found that developmental age had a significant impact on tumor initiation where neonatal mice were more susceptible to tumor development compared with their adult counterparts in both G/tv-a wt (19/19 vs. 4/13, Fisher exact test, ****, P < 0.0001) and C/tv-a wt (8/22 vs. 0/10, Fisher exact test, *, P = 0.0353) mice. There was also a significant difference in survival between tumors induced in neonatal and adult mice for both G/tv-a wt and C/tv-a wt lines (Supplementary Fig. S2A and S2B). Furthermore, in G/tv-a wt mice, there was a significant difference in malignancy between tumors induced in neonatal and adult mice (Supplementary Fig. S2C), where all tumors were high-grade in neonatal mice and low-grade in adult mice. In neonatal C/tv-a wt mice, the majority of tumors were low-grade (grade II; Fig. 2D), and no tumors were found in adult C/tv-a wt mice (Table 1).

Figure 2.

Glioma development induced by RCAS-PDGFB-HA in neonatal G/tv-a and C/tv-a wt and Arf−/− mouse lines. A, Kaplan–Meier graph showing glioma-free survival of G/tv-a wt and C/tv-a wt mice. Log-rank test, P < 0.0001. B, Kaplan–Meier graph showing glioma-free survival of G/tv-a;Arf−/− and C/tv-a;Arf−/− mice. Log-rank test, P < 0.0001. C, Representative photographs of tumor histopathology showing the most common tumor grade of each group (grade IV for G/tva wt, G/tv-a;Arf−/− and C/tv-a;Arf−/−; grade II for C/tv-a wt). Scale bar, 50 μm. D, Distribution of tumor grades. Fisher exact test, G/tv-a wt versus C/tv-a wt, P = 0.0001; G/tv-a;Arf−/− versus C/tv-a;Arf−/−, P = 0.0062; C/tv-a wt versus C/tv-a;Arf−/−, P = 0.0031.

Figure 2.

Glioma development induced by RCAS-PDGFB-HA in neonatal G/tv-a and C/tv-a wt and Arf−/− mouse lines. A, Kaplan–Meier graph showing glioma-free survival of G/tv-a wt and C/tv-a wt mice. Log-rank test, P < 0.0001. B, Kaplan–Meier graph showing glioma-free survival of G/tv-a;Arf−/− and C/tv-a;Arf−/− mice. Log-rank test, P < 0.0001. C, Representative photographs of tumor histopathology showing the most common tumor grade of each group (grade IV for G/tva wt, G/tv-a;Arf−/− and C/tv-a;Arf−/−; grade II for C/tv-a wt). Scale bar, 50 μm. D, Distribution of tumor grades. Fisher exact test, G/tv-a wt versus C/tv-a wt, P = 0.0001; G/tv-a;Arf−/− versus C/tv-a;Arf−/−, P = 0.0062; C/tv-a wt versus C/tv-a;Arf−/−, P = 0.0031.

Close modal

Comparing tumor incidence between different cells of origin in neonatal mice (Table 1), we found a significantly higher incidence in G/tv-a mice compared with C/tv-a mice irrespective of p19Arf status (wt mice, 19/19 vs. 8/22, Fisher exact test, **, P = 0.0062; Arf−/− mice, 24/24 vs. 36/48, Fisher exact test, ****, P < 0.0001). Also, tumor development was accelerated and malignancy increased in G/tv-a mice compared with C/tv-a mice (Fig. 2A–D).

Taken together, our data show that neonatal glial cells were more susceptible to tumor development than adult glial cells and that glial stem cells were more prone to develop HGG than OPC-like cells.

Gliomas induced in neonatal mice are related to subtypes of human pediatric HGGs

The result from in vivo tumor development suggested that the cell type and its age had an effect on inherent glioma cell properties dictating tumor initiation and progression. To further characterize the neonatally induced tumors, we sampled HGG and control tissue for molecular analyses (Fig. 3A). We also established GC cultures from neonatally induced tumors using stem cell conditions to enrich for glioma stem cells (Fig. 3A). In parallel, we set up CC cultures of explants from the SVZ of neonatal mice (Fig. 3A). Despite long experience of human and mouse glioma cell culturing (24, 41), HGGs induced in neonatal mice turned out to be a challenge. From G/tv-a;Arf−/− mice, we finally produced 5 cultures out of 15 explanted tumors and from Ctv-a;Arf−/− mice 1 culture out of 22 explanted tumors. The significant difference in the ability of tumor cells to be maintained in culture [5/15 vs. 1/22, Fisher exact test, P = 0.0306 (*)] strongly indicated that the originating cell type could be important for the phenotype of GCs.

Figure 3.

Overview of experimental groups and the relation of neonatal induced mouse HGG to human pediatric and adult HGG. A, The experimental groups of tumor and control tissues and cell cultures and how they were generated. B, PCA of gene expression data from HGG tissue samples and control tissue (CT) samples. C, PCA of gene expression data from GC and CC cultures. D, Hierarchical clustering of 53 pediatric HGG samples labeled with their previously described subgroup (HC1-3), 32 adult GBM samples, and our mouse HGG and GC samples. Clustering was performed using the 3,896 genes most variably expressed between the subgroups of the pediatric and adult samples.

Figure 3.

Overview of experimental groups and the relation of neonatal induced mouse HGG to human pediatric and adult HGG. A, The experimental groups of tumor and control tissues and cell cultures and how they were generated. B, PCA of gene expression data from HGG tissue samples and control tissue (CT) samples. C, PCA of gene expression data from GC and CC cultures. D, Hierarchical clustering of 53 pediatric HGG samples labeled with their previously described subgroup (HC1-3), 32 adult GBM samples, and our mouse HGG and GC samples. Clustering was performed using the 3,896 genes most variably expressed between the subgroups of the pediatric and adult samples.

Close modal

To investigate the relation and relevance of our neonatal induced tumor models to human pediatric and adult HGG, we performed transcriptome analysis of tissue samples (Fig. 3B) and cell cultures (Fig. 3C). PCA of the mouse transcriptome data showed a clear separation between the different tissue sample groups with tumor and control samples well separated, as well as tumor samples of different origin (Fig. 3B). Also, cell cultures displayed a separation between tumor and control cells but showed a less clear distinction between GCs of different origin (Fig. 3C).

To relate our neonatal mouse HGG models to human pediatric HGG, we compared our mouse transcriptome data to published human pediatric (5) and adult HGG (11, 12) gene expression data using the metagene code for cross-platform and cross-species comparisons. As mentioned above, the 53 pediatric HGG samples had previously been classified into three subgroups, HC1–3, where HC1 was associated with upregulation of cell-cycle–regulating genes and amplifications in PDGFRA and/or PDGFB, HC2 linked to neuronal development, and HC3 related to extracellular matrix receptors and cell adhesion. Hierarchical clustering of mouse and human data using the most variably expressed genes across the samples from the Paugh and colleagues dataset (5) showed a partition, with a certain overlap, of adult and pediatric human HGG samples and a relation of the mouse HGG samples to the HC2 subgroup, while the GC samples were more related to the HC1 subgroup (Fig. 3D). The separation of the mouse HGG samples and the mouse GC cultures likely reflects the lack of tumor-associated cells and the enrichment of PDGFRA-driven tumor cells in the GC samples.

In summary, these cross-species analyses provided evidence that our neonatal-induced supratentorial mouse HGG models were most related to subtypes of human pediatric HGG and GBM.

Glioma cells from GFAP-expressing cells are more malignant and have higher self-renewal and proliferation

As mouse modeling and GC culturing had showed a role of the cell-of-origin for tumor development and tumor cell phenotype, we investigated this further by performing functional analyses on four GCGFAP cultures and one GCCNP culture. The cells were first tested for their ability to generate orthotopic tumors in syngeneic mice (Fig. 4A). All cell cultures could produce secondary tumors and the difference observed in primary tumor development between different cells of origin (Fig. 2B) remained significant also in the secondary tumors (Fig. 4A). There was no significant difference in tumor grade proportions between orthotopic tumors induced by GCGFAP and GCCNP cells (data not shown).

Figure 4.

Functional analyses of mouse GCs derived from neonatal induced HGGs of different origin. All experiments were performed at least three times in duplicates except for measuring sphere sizes that were repeated twice; n equals the number of independent cell cultures (derived from different tumors) that were used. Values represent the mean ± SD. A, Kaplan–Meier graph showing glioma-free survival of syngeneic mice transplanted with GCGFAP and GCCNP. At least one litter of mice was used for each cell culture. Log-rank test, P < 0.05. B, Limiting dilution analysis of GCGFAP and GCCNP cells. Student t test, P = 0.0121. C, the diameter of spheres produced in the LDA. Approximately 30 spheres were measured from each culture. Student t test, P = 0.0001. D, Quantification of BrdU-positive cells after a 16-hour pulse of BrdU. Cells were cultured in presence (+FBS) or absence (−FBS, regular stem cell conditions) for 7 days, fixed and stained for BrdU. Approximately 100 cells were counted in each sample. Student t test was applied for comparisons, GCGFAP(−FBS) vs. GCCNP(−FBS), P = 0.0005; GCGFAP(−FBS) vs. GCGFAP(+FBS), P = 0001; GCCNP(−FBS) vs. GCCNP(+FBS), P = 0.0083; GCGFAP(+FBS) vs. GCCNP(+FBS), P = 0.0005. E, Marker expression in GCs cultured with or without addition of serum for 7 days. Quantification was done by relating the number of marker positive cells to the total number of counted cells. At least 50 cells from each sample were counted. Student t test was applied for comparisons. GCGFAP (+FBS) vs. GCCNP (+FBS; thick lines): CNP, P = 0.0001; OLIG2, P = 0.0001; GFAP, P = 0.0001; nestin, P = 0.0027; TUJ1, P = 0.0001. GCGFAP (−FBS) versus GCGFAP (+FBS; dotted lines): OLIG2, P = 0.0001; GFAP, P = 0.0001; TUJ1, P = 0.0001. GCCNP (−FBS) versus GCCNP (+FBS; thin lines): CNP, P = 0.0013; nestin, P = 0.0006; TUJ1, P = 0.0013.

Figure 4.

Functional analyses of mouse GCs derived from neonatal induced HGGs of different origin. All experiments were performed at least three times in duplicates except for measuring sphere sizes that were repeated twice; n equals the number of independent cell cultures (derived from different tumors) that were used. Values represent the mean ± SD. A, Kaplan–Meier graph showing glioma-free survival of syngeneic mice transplanted with GCGFAP and GCCNP. At least one litter of mice was used for each cell culture. Log-rank test, P < 0.05. B, Limiting dilution analysis of GCGFAP and GCCNP cells. Student t test, P = 0.0121. C, the diameter of spheres produced in the LDA. Approximately 30 spheres were measured from each culture. Student t test, P = 0.0001. D, Quantification of BrdU-positive cells after a 16-hour pulse of BrdU. Cells were cultured in presence (+FBS) or absence (−FBS, regular stem cell conditions) for 7 days, fixed and stained for BrdU. Approximately 100 cells were counted in each sample. Student t test was applied for comparisons, GCGFAP(−FBS) vs. GCCNP(−FBS), P = 0.0005; GCGFAP(−FBS) vs. GCGFAP(+FBS), P = 0001; GCCNP(−FBS) vs. GCCNP(+FBS), P = 0.0083; GCGFAP(+FBS) vs. GCCNP(+FBS), P = 0.0005. E, Marker expression in GCs cultured with or without addition of serum for 7 days. Quantification was done by relating the number of marker positive cells to the total number of counted cells. At least 50 cells from each sample were counted. Student t test was applied for comparisons. GCGFAP (+FBS) vs. GCCNP (+FBS; thick lines): CNP, P = 0.0001; OLIG2, P = 0.0001; GFAP, P = 0.0001; nestin, P = 0.0027; TUJ1, P = 0.0001. GCGFAP (−FBS) versus GCGFAP (+FBS; dotted lines): OLIG2, P = 0.0001; GFAP, P = 0.0001; TUJ1, P = 0.0001. GCCNP (−FBS) versus GCCNP (+FBS; thin lines): CNP, P = 0.0013; nestin, P = 0.0006; TUJ1, P = 0.0013.

Close modal

Self-renewal was analyzed by limiting dilution assay (LDA). We observed that the GCGFAP cultures required significantly fewer cells (18 cells) to form at least 1 sphere per well compared with the GCCNP cultures (28 cells; Fig. 4B). GCGFAP cultures also formed significantly larger spheres compared with the GCCNP culture (Fig. 4C), suggesting a difference also in proliferation. Proliferation was analyzed by measuring BrdU incorporation and GCGFAP cultures exhibited a significantly higher proliferative ability compared with the GCCNP culture (Fig. 4D, −FBS). We also investigated the response of the GC cultures to serum-induced differentiation. Cells were cultured for 7 days in presence of serum (+FBS) and BrdU incorporation was measured. A serum-induced growth inhibition occurred in both GCGFAP and GCCNP cultures (Fig. 4D, −FBS vs. +FBS) but the arrest was significantly stronger in the GCCNP cells (Fig. 4D, +FBS). We also analyzed the cell morphology and induction of glial cell markers upon serum-induced differentiation and found that all serum-cultured cells changed their morphology (Supplementary Fig. S3) and there was a significant difference in marker expression between GCs of different origin (Fig. 4E and Supplementary Fig. S3). Some of the striking differences were that the presence of serum GCGFAP cells significantly upregulated GFAP and TUJ1 and downregulated OLIG2 (Fig. 4E, dotted lines), while GCCNP cells significantly upregulated CNP and TUJ1, downregulated nestin and remained OLIG2 positive (Fig. 4E, thin lines). This produced cells of distinctively different phenotypes when comparing CNP, OLIG2, GFAP, nestin, and TUJ1 expression (Fig. 4E, thick lines), and strongly supported that the HGGGFAP and HGGCNP tumors had originated from different cell types.

As the G/tv-a;Arf−/− and C/tv-a;Arf−/− mouse lines have slightly different and mixed genetic backgrounds we also performed functional studies on the control NSC cultures (CCs) from neonatal G/tv-a;Arf−/− and C/tv-a;Arf−/− mice (Fig. 3A). We analyzed self-renewal (Supplementary Fig. S4A), proliferation as measured by BrdU incorporation (Supplementary Fig. S4B) and stem marker expression in the presence or absence of serum (Supplementary Fig. S4C) and found no significant differences when comparing CCG/tv-a and CCC/tv-a cells (Supplementary Fig. S4D), suggesting that the differences in tumor development and GC phenotype would not be due to genetic modifiers of the different mouse lines.

Glioma cells from CNP expressing cells are generally more resistant to cancer drugs

The inherent phenotypic differences of GCs of different origin encouraged us to investigate whether the cell-of-origin could also affect the drug response of these cells. To investigate this, we used a panel of 28 compounds (Table 2), of which most are in clinical use for treatment of cancer or in clinical trials for cancer, targeting a wide range of signaling pathways dysregulated in cancer cells. We tested the response of these drugs on the viability of three GCGFAP, one GCCNP, and two NSC control cell cultures (one CCG/tv-a and one CCC/tv-a). The log10 fold change of the average AUC value for CCG/tv-a and CCC/tv-a were calculated for each drug (Supplementary Table S1) and the difference between each group were compared using a Welch t test adjusted for multiple testing (Fig. 5). We found that GCGFAP displayed a higher sensitivity than GCCNP toward 14 compounds (Fig. 5 and Table 2), while no compound showed the opposite effect. To analyze whether cell proliferation, which was significantly higher for GCGFAP cells compared to GCCNP cells (Fig. 4D), could be an important factor influencing drug sensitivity, we compared the response of the GC cultures to that of the control cells that displayed a higher proliferation than any of the GC cultures (compare Supplementary Fig. S4B with Fig. 4D). We found no correlation between drug sensitivity and proliferation (Supplementary Fig. S5); in fact, for many of the compounds, the control cells were significantly more resistant. Thus, GCs of more immature origin seemed to have a higher general sensitivity to a wide range of cancer drugs.

Figure 5.

Cell viability of mouse GCGFAP and GCCNP cells in response to a panel of 28 anticancer compounds. GCGFAP and GCCNP cells were exposed to each of the compounds at different concentrations for 72 hours, after which, cell viability was measured, dose–response curves plotted, and AUC values calculated. The experiment was performed twice using at least duplicate samples for each group. The Volcano plot displays the result where the x-axis represents the log10 fold change of mean AUC values for GCGFAP and GCCNP, respectively, and the y-axis represents the negative log10 Benjamini–Hochberg (BH)–corrected P value. Red dots, drugs that show a significant difference in drug sensitivity. Significant drugs have a Benjamini–Hochberg-corrected P < 0.05.

Figure 5.

Cell viability of mouse GCGFAP and GCCNP cells in response to a panel of 28 anticancer compounds. GCGFAP and GCCNP cells were exposed to each of the compounds at different concentrations for 72 hours, after which, cell viability was measured, dose–response curves plotted, and AUC values calculated. The experiment was performed twice using at least duplicate samples for each group. The Volcano plot displays the result where the x-axis represents the log10 fold change of mean AUC values for GCGFAP and GCCNP, respectively, and the y-axis represents the negative log10 Benjamini–Hochberg (BH)–corrected P value. Red dots, drugs that show a significant difference in drug sensitivity. Significant drugs have a Benjamini–Hochberg-corrected P < 0.05.

Close modal

Pediatric supratentorial HGGs are relatively rare but highly aggressive and lethal primary brain tumors for which there is a lack of validated and relevant experimental models. Despite the histopathologic similarities of these tumors to adult HGG, recent large-scale integrated molecular data have showed distinct molecular differences between adult and pediatric HGG. There are a number of models for pediatric HGG available, in particular for brainstem gliomas (42–44). These all have in common that they originate from nestin-expressing cells using the N/tv-a mouse line (34). Here we present the characterization of two new PDGF-driven mouse models for pediatric supratentorial HGG, where either GFAP- or CNP-expressing cells are targeted. We show that there is a difference between neonatal- and adult-induced gliomas with regard to incidence and tumor development, suggesting that qualitative differences in the cell-of-origin due to age may dictate important features of glioma biology. In line with what has been found in human HGG, that pediatric HGGs carry fewer mutations than adult HGGs (5–7, 45, 46), we here show that HGG development in neonatal mice required fewer mutations than HGG development induced in adult mice, both from NSC- and OPC-like cells of origin.

Our data suggest that the as yet unknown cell-of-origin for pediatric supratentorial HGG could be a neural or glial stem cell or a glial precursor cell of the oligodendrocyte differentiation axis. The cross-species transcriptome analysis showed that both cell types produced tumors that were similar to the HC1 and HC2 subgroups of pediatric HGG and distinct from adult GBM. Importantly, our results point that the cell-of-origin could be affecting central characteristics of HGG such as disease progression and treatment response. Using the same oncogenic alterations, tumors from a less differentiated GFAP-positive cell were more aggressive, and tumor cells more tumorigenic, had higher self-renewal and proliferation as compared with tumors induced in a more differentiated, CNP-expressing OPC-like cell. However, HGG lines derived from GFAP-positive cells were in general more sensitive to a panel of selected cancer drugs, of which many are used in the clinic (temozolomide, procarbazine, vincristine) or are in clinical trials (melphalan, bortezomib, doxorubicin, etoposide, irinotecan, mebendazole, erlotinib, imatinib, lapatinib, sorafenib) for treating HGG. This argues for an important role of the cell-of-origin for the phenotype of HGG cells, which cannot be captured by the current molecular analyses of HGG tissue samples. Further investigations to validate this will require molecular and functional analyses of human pediatric supratentorial HGG cell cultures. Such cell models are rare and our attempts to establish cell cultures from experimental, neonatal-induced HGGs showed that it may be more difficult to maintain these cells in long-term culture (47), where the cell-of-origin could be one factor influencing this ability. In a recent study including 55 explanted pediatric brain tumor samples, a majority of samples could be maintained in short-term culture but only six of those reached beyond 6 months in culture (48).

It is clear that adult and pediatric HGGs are genetically, biologically, and molecularly different. Here we show that our neonatal-induced PDGF-driven HGG models in G/tv-a and C/tv-a mice display many features similar to a subset of human pediatric HGG and are molecularly related to a subgroup of pediatric HGGs. They will provide excellent models for further basic and preclinical studies to uncover the underpinnings of this devastating disease. Our finding that the cell-of-origin could influence the response to treatment further underlines the importance of identifying clinically relevant subgroups and predictive biomarkers of non-brainstem pediatric HGG.

M. Fryknäs has ownership interest (including patents) as a minor shareholder in Vivolux AB, which develops analogues of b-AP15 for clinical use. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Sreedharan, H. Weishaupt, F.J. Swartling, L. Uhrbom

Development of methodology: S. Sreedharan, Y. Xie

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Sreedharan, N.P. Maturi, Y. Xie, M. Jarvius, S. Libard, I. Alafuzoff, R. Larsson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Sreedharan, N.P. Maturi, Y. Xie, A. Sundström, M. Jarvius, H. Weishaupt, M. Fryknäs, R. Larsson, L. Uhrbom

Writing, review, and/or revision of the manuscript: S. Sreedharan, N.P. Maturi, Y. Xie, A. Sundström, M. Jarvius, R. Larsson, F.J. Swartling, L. Uhrbom

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Sreedharan, A. Sundström, I. Alafuzoff, F.J. Swartling

Study supervision: S. Sreedharan, M. Fryknäs, F.J. Swartling, L. Uhrbom

We thank Dr. Arijit Mukhopadhyay at CSIR-IGIB for providing lab facilities to S. Sreedharan. Microarray analyses were performed at the SciLifeLab Array and Analysis Facility and image analyses were partly done at the SciLifeLab BioVis Facility.

This work was supported by the Swedish Cancer Society, Swedish Childhood Cancer Foundation, Swedish Research Council, and the Swedish Society for Medical Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Louis
DN
,
Ohgaki
H
,
Wiestler
OD
,
Cavenee
WK
,
Ellison
DW
,
Branger
DF
, et al
WHO Classification of Tumours of the Central Nervous System (Revised 4th edition)
.
Lyon, France:
IARC
; 
2016
.
2.
Ostrom
QT
,
Gittleman
H
,
Fulop
J
,
Liu
M
,
Blanda
R
,
Kromer
C
, et al
CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2008-2012
.
Neuro Oncol
2015
;
17
Suppl 4
:
iv1
iv62
.
3.
Finlay
JL
,
Zacharoulis
S
. 
The treatment of high grade gliomas and diffuse intrinsic pontine tumors of childhood and adolescence: a historical - and futuristic - perspective
.
J Neurooncol
2005
;
75
:
253
66
.
4.
Broniscer
A
,
Chintagumpala
M
,
Fouladi
M
,
Krasin
MJ
,
Kocak
M
,
Bowers
DC
, et al
Temozolomide after radiotherapy for newly diagnosed high-grade glioma and unfavorable low-grade glioma in children
.
J Neurooncol
2006
;
76
:
313
9
.
5.
Paugh
BS
,
Qu
C
,
Jones
C
,
Liu
Z
,
Adamowicz-Brice
M
,
Zhang
J
, et al
Integrated molecular genetic profiling of pediatric high-grade gliomas reveals key differences with the adult disease
.
J Clin Oncol
2010
;
28
:
3061
8
.
6.
Schwartzentruber
J
,
Korshunov
A
,
Liu
XY
,
Jones
DT
,
Pfaff
E
,
Jacob
K
, et al
Driver mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma
.
Nature
2012
;
482
:
226
31
.
7.
Wu
G
,
Diaz
AK
,
Paugh
BS
,
Rankin
SL
,
Ju
B
,
Li
Y
, et al
The genomic landscape of diffuse intrinsic pontine glioma and pediatric non-brainstem high-grade glioma
.
Nat Genet
2014
;
46
:
444
50
.
8.
Sturm
D
,
Witt
H
,
Hovestadt
V
,
Khuong-Quang
DA
,
Jones
DT
,
Konermann
C
, et al
Hotspot mutations in H3F3A and IDH1 define distinct epigenetic and biological subgroups of glioblastoma
.
Cancer Cell
2012
;
22
:
425
37
.
9.
Jones
C
,
Perryman
L
,
Hargrave
D
. 
Paediatric and adult malignant glioma: close relatives or distant cousins?
Nat Rev Clin Oncol
2012
;
9
:
400
13
.
10.
Bax
DA
,
Mackay
A
,
Little
SE
,
Carvalho
D
,
Viana-Pereira
M
,
Tamber
N
, et al
A distinct spectrum of copy number aberrations in pediatric high-grade gliomas
.
Clin Cancer Res
2010
;
16
:
3368
77
.
11.
Beroukhim
R
,
Getz
G
,
Nghiemphu
L
,
Barretina
J
,
Hsueh
T
,
Linhart
D
, et al
Assessing the significance of chromosomal aberrations in cancer: methodology and application to glioma
.
Proc Natl Acad Sci U S A
2007
;
104
:
20007
12
.
12.
Lee
Y
,
Scheck
AC
,
Cloughesy
TF
,
Lai
A
,
Dong
J
,
Farooqi
HK
, et al
Gene expression analysis of glioblastomas identifies the major molecular basis for the prognostic benefit of younger age
.
BMC Med Genomics
2008
;
1
:
52
.
13.
Alcantara Llaguno
S
,
Chen
J
,
Kwon
CH
,
Jackson
EL
,
Li
Y
,
Burns
DK
, et al
Malignant astrocytomas originate from neural stem/progenitor cells in a somatic tumor suppressor mouse model
.
Cancer Cell
2009
;
15
:
45
56
.
14.
Friedmann-Morvinski
D
,
Bushong
EA
,
Ke
E
,
Soda
Y
,
Marumoto
T
,
Singer
O
, et al
Dedifferentiation of neurons and astrocytes by oncogenes can induce gliomas in mice
.
Science
2012
;
338
:
1080
4
.
15.
Lindberg
N
,
Kastemar
M
,
Olofsson
T
,
Smits
A
,
Uhrbom
L
. 
Oligodendrocyte progenitor cells can act as cell of origin for experimental glioma
.
Oncogene
2009
;
28
:
2266
75
.
16.
Liu
C
,
Sage
JC
,
Miller
MR
,
Verhaak
RG
,
Hippenmeyer
S
,
Vogel
H
, et al
Mosaic analysis with double markers reveals tumor cell of origin in glioma
.
Cell
2011
;
146
:
209
21
.
17.
Alcantara Llaguno
SR
,
Wang
Z
,
Sun
D
,
Chen
J
,
Xu
J
,
Kim
E
, et al
Adult Lineage-restricted CNS progenitors specify distinct glioblastoma subtypes
.
Cancer Cell
2015
;
28
:
429
40
.
18.
Hambardzumyan
D
,
Amankulor
NM
,
Helmy
KY
,
Becher
OJ
,
Holland
EC
. 
Modeling adult gliomas using RCAS/t-va technology
.
Transl Oncol
2009
;
2
:
89
IN6
.
19.
Ozawa
T
,
Riester
M
,
Cheng
YK
,
Huse
JT
,
Squatrito
M
,
Helmy
K
, et al
Most human non-GCIMP glioblastoma subtypes evolve from a common proneural-like precursor glioma
.
Cancer Cell
2014
;
26
:
288
300
.
20.
Holland
EC
,
Varmus
HE
. 
Basic fibroblast growth factor induces cell migration and proliferation after glia-specific gene transfer in mice
.
Proc Natl Acad Sci U S A
1998
;
95
:
1218
23
.
21.
Uhrbom
L
,
Kastemar
M
,
Johansson
FK
,
Westermark
B
,
Holland
EC
. 
Cell type-specific tumor suppression by Ink4a and Arf in kras-induced mouse gliomagenesis
.
Cancer Res
2005
;
65
:
2065
9
.
22.
Lindberg
N
,
Jiang
Y
,
Xie
Y
,
Bolouri
H
,
Kastemar
M
,
Olofsson
T
, et al
Oncogenic signaling is dominant to cell of origin and dictates astrocytic or oligodendroglial tumor development from oligodendrocyte precursor cells
.
J Neurosci
2014
;
34
:
14644
51
.
23.
Wiesner
SM
,
Decker
SA
,
Larson
JD
,
Ericson
K
,
Forster
C
,
Gallardo
JL
, et al
De novo induction of genetically engineered brain tumors in mice using plasmid DNA
.
Cancer Res
2009
;
69
:
431
9
.
24.
Jiang
Y
,
Boije
M
,
Westermark
B
,
Uhrbom
L
. 
PDGF-B can sustain self-renewal and tumorigenicity of experimental glioma-derived cancer-initiating cells by preventing oligodendrocyte differentiation
.
Neoplasia
2011
;
13
:
492
503
.
25.
Irizarry
RA
,
Hobbs
B
,
Collin
F
,
Beazer-Barclay
YD
,
Antonellis
KJ
,
Scherf
U
, et al
Exploration, normalization, and summaries of high density oligonucleotide array probe level data
.
Biostatistics
2003
;
4
:
249
64
.
26.
RStudio Team
.
RStudio: Integrated Development Environment for R
.
Boston, MA
:
Rstudio Inc.
; 
2016
.
27.
R Core Team
.
R: A Language and Environment for Statistical Computing
.
Vienna, Austria
:
R Foundation for Statistical Computing
; 
2016
.
28.
Soetaert
K
.
plot3D: Plotting Multi-Dimensional Data
.
R package version 1.1
.
Vienna, Austria
:
R Foundation for Statistical Computing
; 
2016
.
29.
Gray
KA
,
Yates
B
,
Seal
RL
,
Wright
MW
,
Bruford
EA
. 
Genenames.org: the HGNC resources in 2015
.
Nucleic Acids Res
2015
;
43
(
Database issue
):
D1079
85
.
30.
Eyre
TA
,
Wright
M
,
Lush
MJ
,
Bruford
EA
. 
HCOP: a searchable database of human orthology predictions
.
Brief Bioinform
2007
;
8
:
2
5
.
31.
Barrett
T
,
Wilhite
S
,
Ledoux
P
,
Evangelista
C
,
Kim
IF
,
Tomashevsky
M
, et al
NCBI GEO: archive for functional genomics data sets–update
.
Nucleic Acids Res
2013
;
41
(
Database issue
):
D991
5
.
32.
Tamayo
P
,
Scanfeld
D
,
Ebert
BL
,
Gillette
MA
,
Roberts
CWM
,
Mesirow
JP
. 
Metagene projection for cross-platform, cross-species characterization of global transcriptional states
.
Proc Natl Acad Sci U S A
2007
;
104
:
5959
64
.
33.
The MathWorks I
. 
MATLAB and Statistics Toolbox Release 2015A
.
Natick, MA
:
The MathWorks I
; 
2015
.
34.
Holland
EC
,
Hively
WP
,
DePinho
RA
,
Varmus
HE
. 
A constitutively active epidermal growth factor receptor cooperates with disruption of G1 cell-cycle arrest pathways to induce glioma-like lesions in mice
.
Genes Dev
1998
;
12
:
3675
85
.
35.
Persson
AI
,
Petritsch
C
,
Swartling
FJ
,
Itsara
M
,
Sim
FJ
,
Auvergne
R
, et al
Non-stem cell origin for oligodendroglioma
.
Cancer Cell
2010
;
18
:
669
82
.
36.
Kamijo
T
,
Bodner
S
,
van de Kamp
E
,
Randle
DH
,
Sherr
CJ
. 
Tumor spectrum in ARF-deficient mice
.
Cancer Res
1999
;
59
:
2217
22
.
37.
Kamijo
T
,
van de Kamp
E
,
Chong
MJ
.
Zindy
F
,
Diehl
JA
,
Sherr
CJ
, et al
Loss of the ARF tumor suppressor reverses premature replicative arrest but not radiation hypersensitivity arising from disabled Atm function
.
Cancer Res
1999
;
59
:
2464
9
.
38.
Marsh
JC
,
Goldman
S
,
Ziel
E
,
Bregman
C
,
Diaz
A
,
Byrne
R
, et al
Involvement of the neural stem cell compartment by pediatric and adult gliomas: a retrospective review of 377 cases
.
J Neurooncol
2015
;
122
:
105
10
.
39.
Menn
B
,
Garcia-Verdugo
JM
,
Yaschine
C
,
Gonzalez-Perez
O
,
Rowitch
D
,
Alvarez-Buylla
A
. 
Origin of oligodendrocytes in the subventricular zone of the adult brain
.
J Neurosci
2006
;
26
:
7907
18
.
40.
Zerlin
M
,
Milosevic
A
,
Goldman
JE
. 
Glial progenitors of the neonatal subventricular zone differentiate asynchronously, leading to spatial dispersion of glial clones and to the persistence of immature glia in the adult mammalian CNS
.
Dev Biol
2004
;
270
:
200
13
.
41.
Xie
Y
,
Bergstrom
T
,
Jiang
Y
,
Johansson
P
,
Marinescu
VD
,
Lindberg
N
, et al
The human glioblastoma cell culture resource: validated cell models representing all molecular subtypes
.
EBioMedicine
2015
;
2
:
1351
63
.
42.
Subashi
E
,
Cordero
FJ
,
Halvorson
KG
,
Qi
Y
,
Nouls
JC
,
Becher
OJ
, et al
Tumor location, but not H3.3K27M, significantly influences the blood-brain-barrier permeability in a genetic mouse model of pediatric high-grade glioma
.
J Neurooncol
2016
;
126
:
243
51
.
43.
Misuraca
KL
,
Hu
G
,
Barton
KL
,
Chung
A
,
Becher
OJ
. 
A novel mouse model of diffuse intrinsic pontine glioma initiated in Pax3-expressing cells
.
Neoplasia
2016
;
18
:
60
70
.
44.
Becher
OJ
,
Hambardzumyan
D
,
Walker
TR
,
Helmy
K
,
Nazarian
J
,
Albrecht
S
, et al
Preclinical evaluation of radiation and perifosine in a genetically and histologically accurate model of brainstem glioma
.
Cancer Res
2010
;
70
:
2548
57
.
45.
Radtke
I
,
Mullighan
CG
,
Ishii
M
,
Su
X
,
Cheng
J
,
Ma
J
, et al
Genomic analysis reveals few genetic alterations in pediatric acute myeloid leukemia
.
Proc Natl Acad Sci U S A
2009
;
106
:
12944
9
.
46.
Vogelstein
B
,
Papadopoulos
N
,
Velculescu
VE
,
Zhou
S
,
Diaz
LA
 Jr
,
Kinzler
KW
. 
Cancer genome landscapes
.
Science
2013
;
339
:
1546
58
.
47.
Sanden
E
,
Eberstal
S
,
Visse
E
,
Siesjo
P
,
Darabi
A
. 
A standardized and reproducible protocol for serum-free monolayer culturing of primary paediatric brain tumours to be utilized for therapeutic assays
.
Sci Rep
2015
;
5
:
12218
.
48.
Thirant
C
,
Bessette
B
,
Varlet
P
,
Puget
S
,
Cadusseau
J
,
Tavares Sdos
R
, et al
Clinical relevance of tumor cells with stem-like properties in pediatric brain tumors
.
PLoS One
2011
;
6
:
e16375
.