In vitro modeling of glioblastoma initiation using PDGF-AA and p53-null neural progenitors

Abstract

Background

Imagining ways to prevent or treat glioblastoma (GBM) has been hindered by a lack of understanding of its pathogenesis. Although overexpression of platelet derived growth factor with two A-chains (PDGF-AA) may be an early event, critical details of the core biology of GBM are lacking. For example, existing PDGF-driven models replicate its microscopic appearance, but not its genomic architecture. Here we report a model that overcomes this barrier to authenticity.

Methods

Using a method developed to establish neural stem cell cultures, we investigated the effects of PDGF-AA on subventricular zone (SVZ) cells, one of the putative cells of origin of GBM. We microdissected SVZ tissue from p53-null and wild-type adult mice, cultured cells in media supplemented with PDGF-AA, and assessed cell viability, proliferation, genome stability, and tumorigenicity.

Results

Counterintuitive to its canonical role as a growth factor, we observed abrupt and massive cell death in PDGF-AA: wild-type cells did not survive, whereas a small fraction of null cells evaded apoptosis. Surviving null cells displayed attenuated proliferation accompanied by whole chromosome gains and losses. After approximately 100 days in PDGF-AA, cells suddenly proliferated rapidly, acquired growth factor independence, and became tumorigenic in immune-competent mice. Transformed cells had an oligodendrocyte precursor-like lineage marker profile, were resistant to platelet derived growth factor receptor alpha inhibition, and harbored highly abnormal karyotypes similar to human GBM.

Conclusion

This model associates genome instability in neural progenitor cells with chronic exposure to PDGF-AA and is the first to approximate the genomic landscape of human GBM and the first in which the earliest phases of the disease can be studied directly.

Key Points.

1. PDGF-AA induces genomic instability, growth factor independence, and tumor initiating capacity (transformation) in p53 null neural progenitor cells.

2. The PDGF-AA receptor becomes redundant in transformed cells; regardless of receptor inactivation or knockout, cells retain proliferative capacity.

3. PDGFA transformed cells possess recurrent chromosomal losses and gains similar to those seen in human GBM.

Importance of the Study.

We have developed a mouse model in which the initiation, evolution, and genomic landscape of GBM can be thoroughly studied, thus paving the way for ideas about how this deadly brain cancer might be prevented, interrupted at an occult stage, or treated with very different therapies.

The genomic architecture of the isocitrate dehydrogenase (IDH) wild-type form of glioblastoma (GBM), common in older adults, consists of signature chromosomal gains and losses.¹ Despite an increasingly detailed annotation of these and other molecular alterations, as well an expanding knowledge of the processes affected by these changes, the prognosis for patients with this type of GBM has changed little. In an effort to explore and understand its origins, analyses of The Cancer Genome Atlas and other genomic datasets by multiple groups have identified gain of chromosome 7 as a putative early event in the initiation of IDH wild-type GBM.^2,3 Based on these analyses Ozawa et al further hypothesized that overexpression of platelet derived growth factor with two A-chains (PDGF-AA) at 7p22.3 initiates wild-type GBM when accompanied by loss of p53, cyclin-dependent kinase inhibitor 2A, or chromosome 10.²

To study the interaction of PDGF-AA with p53 loss, Ozawa et al overexpressed PDGF-AA in p53 null adult mice, generating tumors that were histologically indistinguishable from human GBM. More recently Koga et al engineered pluripotent human stem cells to produce human-like GBMs in immune-compromised mice by combining PDGF-AA receptor (PDGFR-α) activation with other alterations seen in human GBM, including loss of p53, phosphatase and tensin homolog (PTEN), and neurofibromatosis type 1,⁴ and Jun et al generated GBMs by engineering mice to overexpress Pdgfr-α in a Cre-recombinase dependent manner when crossed with p53-null mice.⁵ Such models suggest that activation of the PDGFR-α pathway plays a key role in GBM pathogenesis, but despite yielding tumors that resemble human GBM under the microscope, none appear to capture the chromosomal instability and progressive nature of the human disease. Furthermore, none yet add to our understanding of disease initiation, which may be critical to early intervention strategies or risk reduction.

Here, we report a model of GBM in which p53-null and heterozygous subventricular zone (SVZ) cells transform in vitro in PDGF-AA, acquiring genomic features that bear a remarkable similarity to those seen in the human disease. Aware that GBMs likely arise from progenitor cells,⁶ many of which reside in the SVZ, we prepared SVZ cultures from adult mice using a technique developed to study neural stem cells.⁷ To simulate oncogenic signaling, cultures were supplemented with epidermal growth factor plus fibroblast growth factor (EGF/FGF), or with PDGF-AA, ligands for aberrantly activated GBM pathways.⁸ Because the p53 pathway is frequently compromised in human GBM,⁸ we established cultures of null and heterozygous cells in different growth factors and monitored their phenotypes compared with wild-type cells. We found that cells with p53 compromise behaved differently than wild-type cells, and both behaved differently in PDGF-AA than in EGF/FGF: only null and heterozygous cells transformed, and only in PDGF-AA. Transformed cells displayed early genomic instability and subsequently generated GBMs in immune-competent syngeneic mice. Using this approach we observed the earliest stages in the transformation of cells that evolved to GBMs.

Materials and Methods

Mice

Eight week-old (p53-/- B6.129S2-Trp53^tm1Tyj/J, p53+/+ C57BL/6J) mice were purchased from The Jackson Laboratory. All experiments involving mice were conducted in accordance with animal care procedures at the University of Calgary, protocol #M08029.

SVZ Neurosphere Culture

Neurosphere (sphere) cultures were established in serum-free media as described⁷: NeuroCult Media (Stem Cell Technologies, #05700) supplemented with human PDGF-AA (20 ng/mL; Peprotech 100-13A) or EGF (20 ng/mL; Peprotech AF-100-15) plus FGF-2 (20 ng/mL; Peprotech AF-100-18B) and heparan sulfate (2μg/mL; Stem Cell Technologies 07980). Spheres were passaged at a diameter of 100–200 μm or every 1 to 3 weeks.

Cell Death Assay

Cell death was documented using annexinV staining (V13241, Life Technologies) as per manufacturer’s instructions. Cells were analyzed on a BD Biosciences LSRII flow cytometer using FACSDiva.

Proliferation Assay

Proliferation was assessed using Click-iT EdU Proliferation Assay (C10634, Life Technologies). For 2.5 hours, 5 × 10⁵ cells were incubated in 5-ethynyl-2′-deoxyuridine (EdU) and assessed as instructed by the manufacturer on a BD Biosciences LSRII flow cytometer using FACSDiva (software).

Syngeneic Intracranial Implantation

Cells were implanted as described.⁹ After implantation mice were monitored and euthanized at the first sign of illness (weight loss >15%). Brains were fixed by cardiac cannulation and perfusion with 4% paraformaldehyde (PFA) or submersion in 4% PFA overnight.

Staining and Immunohistochemistry

Immunohistochemistry was performed as described.¹⁰ Antibodies: anti-Nestin (1:200; Millipore MAB353), anti–glial fibrillary acidic protein (GFAP) (1:500; Millipore MAB360), and anti–oligodendrocyte transcription factor 2 (Olig2) (1:400; Millipore MABN50) with goat anti-mouse immunoglobulin G–horseradish peroxidase (IgG-HRP) secondary antibody (1:2000; Santa Cruz sc-2005). Peroxidase signal was detected using the ABC Elite kit (Vector Laboratories) and DAB (3,3′-diaminobenzidine) substrate kit (Sigma D4168) using manufacturer’s protocol.

Phospho–Receptor Tyrosine Kinase Arrays

Mouse phosphorylated receptor tyrosine kinase (RTK) arrays (ARY014, R&D Systems) were used to assess receptor phosphorylation, as per manufacturer’s instructions. Arrays were visualized with electrochemiluminescence reagent and Hyperfilm (GE Healthcare).

Viability Assays

Cell viability was assessed in triplicate with the alamarBlue viability assay (Medicorp). Cells were seeded in 96-well plates (5000 cells per well). AlamarBlue reagent was added and absorbance measured after a 6-hour incubation.

PDGFR-α Inhibition Assays

Spheres were dissociated and seeded at 2000 cells/100μL in a 96-well plate in triplicate. Each well was supplemented with PDGFR-α blocking antibody (1:100; R&D Systems), or Imatinib (Seleckchem). Cultures were incubated at 37°C for 6 days and viability assessed.

Pdgfr-α Knockdown

RNA interference (RNAi) for Pdgfr-α was obtained from Invitrogen (#4390771, ID#s71418). Control scrambled RNAi (Invitrogen, #4390843), Egfr RNAi (#4390771, lot #ASO216GD, ID: s65372), and transfection reagent Lipofectamine RNAi Max (Qiagen) were used. Transfections were carried out as per manufacturer’s protocol and expression of Pdgfr-a protein assessed (Supplementary Figure 2C).

Pdgfr-α CRISPR Gene Knockout

Short guide sequences targeting Pdgr-α exon 3 are 5′-AAGGAATCGGTCATCCCGAG-3′ and 5′-TAACCTTGCA CAATAACGGG-3′, targeting Pdgfr-α exon 6 were 5′-ACCCGA CGCCCAGGATATCG-3′. All short guide sequences were cloned into the pX458 vector. Pdgfr-α short guide RNA constructs 1, 2, and 3 were transfected with Lipofectamine 2000 (Invitrogen) into a transformed p53-null SVZ cell line as recommended by the manufacturer. See Supplementary methods for additional details.

Western Blotting

Protein was extracted as described.⁹ Antibodies included anti–PDGFR-α (1:1000; Cell Signaling 3174); anti–EGF receptor (EGFR) (1:2000; Cell Signaling 4267); anti-Olig2 (1:2000; Millipore MABN50); anti-Nestin (1:2000; Millipore MAB353); anti–neural glial antigen 2 (NG2) (1:1000; Cell Signaling 4235); anti-GFAP (1:1000; Millipore MAB360); anti–β3-tubulin (1:1000; Cell Signaling 4466); and anti–β-actin (1:5000; Cell Signaling 3700). Secondary antibodies were goat anti-mouse and anti-rabbit IgG-HRP (1:2000; Santa Cruz sc-2005 and sc-2004).

Surface Marker Assessment

Surface markers were assessed on the BD LSR II flow cytometer. IgG/REA controls were used for gating. Antibodies: anti–prominin-1 (CD133)-phycoerythrin (PE) (Miltenyi Biotec, 130-102-834); rat IgG1-PE (Miltenyi Biotec, 130-103-042); anti-CD34–fluorescein isothiocyanate (FITC) (Miltenyi Biotec, 130-105-890); REA Control-FITC (Miltenyi Biotec, 130-104-626); anti-CD44-FITC (Miltenyi Biotec, 130-102-933); rat IgG2b-FITC (Miltenyi Biotec, 130-103-088); anti–stage-specific embryonic antigen 1 (CD15)-PE (Miltenyi Biotec, 130-104-989); and REA Control-PE (Miltenyi Biotec, 130-104-612).]

Transferring EGF/FGF Cultures to PDGF-AA

SVZ cells maintained in EGF/FGF were dissociated and replated in media supplemented with PDGF-AA. At each subsequent passage, total and live cell counts and percent viability in trypan blue were assessed.

Gamma-H2AX Immunofluorescence

Staining was done as described.¹¹ Antibodies: anti–gamma H2AX (1:800; Abcam ab26350) and anti-phospho H3 (1:500; Abcam ab47297). Secondary antibodies: anti-rabbit IgG-Alexa Fluor 594 (1:800; Life Technologies A11037) and anti-mouse IgG-Alexa Fluor 488 (1:800; Life Technologies A11001). Nuclei were stained with 4′,6′-diamidino-2-phenylindole (1:1000; 1 mg/mL stock diluted to 1 μg/mL in 1x phosphate buffered saline; Sigma Aldrich D9542).

Cell Type Identification

Transformed or freshly dissected SVZ cells were plated on poly-D-lysine/laminin coated coverslips in PDGF-AA and stained as previously described.¹² Antibodies: anti-NG2 (1:400; Millipore-Sigma AB5320), anti-PDGFR-α (1:200; Cell Signaling Technology 3174), anti-GFAP (1:1000; Millipore-Sigma AB5541), anti–sex determining region Y–box 2 (Sox2) (1:500; Cell Signaling Technology 3728), anti-Nestin (1:500; Millipore-Sigma MAB353), anti-CNPase (1:1000; Millipore-Sigma MAB326), anti–β-III-tubulin (1:200; R&D Systems MAB1195), and anti-Olig2 (1:500; Millipore-Sigma MABN50). Secondary antibodies: anti-rabbit IgG-Alexa Fluor 594 (1:800; Life Technologies A11037), anti-mouse IgG-Alexa Fluor 488 (1:800; Life Technologies A11001) and anti-chicken IgY-Alexa Fluor 488 (ThermoFisher Scientific A-11039). Images were taken on a ZEISS LSM880 confocal microscope with Airyscan with a 20x/0.8 PlanApo objective. Between 300 and 700 cells were counted in each 25 mm² field of view per experiment.

Cell Cycle

Cells were harvested at indicated times and treated with Accumax to obtain a single cell suspension. Propidium iodide (PI) staining was completed as described.¹² Cells were analyzed on a FACscan Flow Cytometer (Becton Dickinson) and cell cycle distribution determined using ModFit LT 2.0 software (Verity Software House).

G-Band Karyotyping

Karyomax Colcemid (#15212-012, Gibco) was added to cells at 80–90% confluence and kept in a CO₂ incubator at 37^oC (1.5 h). Cells were collected, treated with 100 µL Accutase (#7920, Stemcell Technologies) at room temperature (5 min), suspended in 8 mL 0.075M KCl, and incubated in 37^oC (15 minutes), and Carnoy’s Fixative added. Cells were then centrifuged at 1000 rpm for 10 minutes at room temperature and pellets collected. After 3 rounds of fixation, cells were resuspended in fixative, dispersed on glass slides, and baked at 90^oC (1.5 h). Routine G-banding analysis was carried out. Twenty metaphases per line were examined.

Array Comparative Genomic Hybridization

DNA was isolated using the DNEasy extraction kit (Qiagen) and processed on the Agilent Mouse 1 × 1M array. Data preprocessing was performed using snap array comparative genomic hybridization (CGH) (background correction [method = minimum] and within-array normalization [method = median]). Within-array replicates were combined into a single value by determining mean value, and data segmented using the DNAcopy implementation of the circular binary segmentation algorithm with a default-parameter smoothing and outlier correction (smooth.CNA function). The Integrative Genomics Viewer was used to identify and visualize regions of aberrant copy number.

Statistical Analysis

GraphPad Prism v8 was used for all analyses. One-way ANOVA was used to assess statistical significance for multiple comparisons within a sample set, while the unpaired t-test was used to assess significance in sample sets of two groups.

Results

P53-Null SVZ Cells Proliferate in EGF/FGF but Only Transform in PDGF-AA

SVZ cultures were prepared as described by Reynolds and Weiss.⁷ In EGF/FGF, null cells formed large round spheres (diameter >100 µ m) within one week (Fig. 1A) and thereafter proliferated rapidly requiring weekly passaging (denoted “P”; methods). The continuous rapid expansion of these cultures led us to ask whether cells had acquired the capacity to proliferate in the absence of EGF/FGF. To test this possibility, null cells from early (P ≤ 4) and late (P ≥ 8) passage cultures were transferred to media without EGF/FGF, where they immediately stopped proliferating (Fig. 1A). Even after a year of continuous growth in EGF/FGF, null cells remained growth factor dependent. Early and late passage p53 wild-type cells behaved similarly (Supplementary Figure 1A). These findings reveal that EGF/FGF supports SVZ cell proliferation, but continuous signaling does not lead to growth factor independence regardless of the presence or absence of p53.

Fig. 1.

P53-null SVZ cell cultures acquire growth factor independence and become tumorigenic when cultured in PDGF-AA. (A) P53-null cells in media supplemented with EGF/FGF proliferate rapidly and form large spheres in early and late passage cultures but cease proliferating when EGF/FGF is removed (scale bar = 200 µM). (B) Early passage null cells display attenuated proliferation and sphere formation in PDGF-AA and remain PDGF-AA dependent. At late passages, cells rapidly form spheres in the presence or absence of PDGF-AA (scale bar = 200 µM). (C) Annexin V staining reveals that a high percentage of null SVZ cells (>70%) in early passage PDGF-AA cultures are dead or apoptotic; this proportion decreases over time. (D) EdU-labeling of null cultures supplemented with PDGF-AA shows an increase in the percentage of cells synthesizing DNA over time, increasing further when PDGF-AA independent proliferation is achieved. (E) Annexin-V/PI staining reveals that the proportion of dead, apoptotic and live null cells remains constant in EGF/FGF over time. (F) EdU-labeling shows that a high percentage EGF/FGF cultured null SVZ cells are replicating DNA and remain dependent on EGF/FGF supplementation. (G‒I) PDGF-AA independent transformed p53-null cells form infiltrating, hemorrhagic GBMs in wild-type mice. Tumor bearing mice (n = 32) had a median survival of 73 days (range, 50–420). Early passage p53-null SVZ cells cultured in PDGF-AA, or long-term EGF/FGF cultures did not form tumors. H&E staining reveals typical GBM features, including high cellularity, mitoses, and necrosis (scale bar = 50 μm). (J) Immunohistochemistry reveals presence of GBM markers, including GFAP, Nestin, and Olig2 positive cells (scale bar = 50 μm). (NS = P> 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001)

In contrast, cultures of p53-null cells in PDGF-AA immediately underwent extensive cell death and displayed attenuated proliferation. After 3 weeks, however, small irregular spheres and single cells were observed, and cultures could be passaged (Fig. 1B). Early passage PDGF-AA cultures (ie, P ≤ 4) contained spheres and extensive debris, suggesting that proliferation and cell death were occurring simultaneously. After approximately 100 days (ie, P-8), the phenotype of null cells changed dramatically; they began to proliferate rapidly, formed large asymmetric spheres, and required frequent passaging. These changes led us to ask whether they had become growth factor independent. Cells from early (P ≤ 4) and late (P ≥ 8) passages were transferred to media that did not contain PDGF-AA. Early passage cultures ceased to proliferate in the absence of PDGF-AA, whereas late passage cultures continued to expand rapidly (Fig. 1B). Unlike null cells, wild-type cells underwent extensive cell death and were not viable in PDGF-AA (Supplementary Figure 1B).

To further document the behavior of null cells, we used annexin V and PI staining to measure the fraction of dead, apoptotic, and live cells over time. We observed that <10% of cells were viable at passage 0 (P-0) versus 60% at P-12 (Fig. 1C), suggesting a gradual selection of cells capable of proliferating in PDGF-AA. We also used EdU labeling to assess rates of DNA synthesis in PDGF-AA and found that proliferation increased from P-0 to P-15 (Fig. 1D), including after P-8 when cultures displayed PDGF-AA independent growth, again implying the selection of cells capable of proliferating in PDGF-AA. In contrast, the proportions of dead, apoptotic, and live cells and their rates of proliferation remained constant in EGF/FGF (Fig. 1E, F). Attenuated proliferation followed by intense cell selection leading to the emergence of rapidly dividing growth factor independent cells was a phenotype seen only in PDGF-AA. Moreover, this unique phenotype was not seen in null cells grown in PDGF-AA plus EGF/FGF.

These findings reveal that PDGF-AA has a counterintuitive anti-proliferative effect on SVZ cells. Rather than rapid division, we observed cell death, attenuated proliferation, and gradual progression to a PDGF-AA independent proliferative state. These events were seen in multiple independent experiments (n > 40) and in 2 of 12 heterozygous p53 cultures. Heterozygous cells took approximately 10 months to become growth factor independent, and PDGF-AA independent proliferation was associated with loss of p53 protein expression (Supplementary Figure 1C, D).

The capacity of null SVZ cells to acquire growth factor independence in PDGF-AA prompted us to inquire whether they had become tumorigenic. Cells from multiple PDGF-AA independent null cell cultures were implanted into the striatum of 8- to 12-week old immune-competent mice. All mice became ill after developing infiltrating high-grade astrocytomas (ie, GBMs). Time to illness ranged from 50 to 420 days (median 73 days), suggesting that these high-grade cancers were heterogeneously aggressive (Fig. 1G, H). Hematoxylin and eosin (H&E) staining revealed cellular neoplasms characterized by frequent mitoses, areas of hemorrhage, and necrosis (Fig. 1I). Tumor cells expressed GFAP, Nestin, and Olig2 (Fig. 1J). Additionally, these high-grade astrocytic gliomas contained small numbers of CD3-positive T cells and ionized calcium binding adaptor molecule 1 (Iba1)–positive macrophages and microglia (Supplementary Figure 1E, F). Early-passage (P < 4) null cells in PDGF-AA and cells in EGF/FGF did not form tumors when implanted in mice despite long follow-up (Fig. 1GSupplementary Figure 2A).

Transformed Cells Display Phosphorylation of Multiple Receptor Tyrosine Kinases and PDGFR-α Independent Proliferation

The capacity of null SVZ cells to evolve to a growth factor independent and tumorigenic state in PDGF-AA led us to examine whether PDGFR-α was phosphorylated in the absence of exogenous PDGF-AA. Using phospho-RTK arrays, we observed 3 patterns of phosphorylation in 15 independently derived transformed lines (Fig. 2ASupplementary Figure 2B): (i) phosphorylation of PDGFR-α alone (Fig. 2A, panel I), (ii) phosphorylation of PDGFR-α plus the EGFR (panel II), and (iii) phosphorylation of PDGFR-α plus various other receptors with or without EGFR (panels III, IV). To ensure these results were not an in vitro artifact, transformed lines were grown intracerebrally in immune-competent mice and the analysis repeated on tumor tissue; PDGFR-α phosphorylation and similar secondary patterns were seen (Fig. 2A). Future work will clarify which RTKs drive the proliferation of transformed cells and which sites are phosphorylated, but these results are intriguing because activation and upregulation of EGFR, Axl receptor (AXLR), and hepatocyte growth factor receptor (HGFR) are seen in GBM and implicated in its pathogenesis.^13,14

Fig. 2.

Transformed cells have variable RTK phosphorylation and are unaffected by PDGFR-α inactivation. (A) In PDGF-AA independent transformed lines and intracerebral xenografts, multiple patterns of RTK phosphorylation were observed including: PDGFR-α alone (I); PDGFR-α with EGFR (II); PDGFR-α with EGFR and HGFR (III); PDGFR-α with HGFR and AXLR [1 = PDGFR-α, 2 = EGFR, 3 = HFGR, 4 = AXLR] (B, C, D). Antibody-mediated inhibition of PDGFR-α (B), treatment with a PDGFR-α Imatinib (C), and knockdown of PDGFR-α (D) decrease cell viability in early but not late passage transformed cultures. (E, F) Blockade or knockdown of PDGFR-α leads to a decrease in cell proliferation and sphere formation in early but not late passage transformed p53-null SVZ cultures. (NS = P > 0.05; *P < 0.05; **P < 0.01; ***P< 0.001; ****P < 0.0001)

Persistent receptor phosphorylation in the absence of PDGF-AA led us to ask whether PDGFR-α was required for growth factor independence. Accordingly, we inhibited PDGFR-α before and after transformation in cell lines with PDGFR-α or PDGFR-α plus EGFR phosphorylation using antibody-mediated blockade, Imatinib, RNAi knockdown, and clustered regularly interspaced short palindromic repeat (CRISPR) knockout. Irrespective of the phosphorylation pattern, receptor blockade, Imatinib, and RNAi decreased SVZ cell viability, proliferation, and sphere formation in early passage PDGF-AA dependent cultures, but had no effect on transformed cells (Fig. 2B–F). Likewise, Pdgfr-α CRISPR knockout did not alter proliferation; transformed cells lacking Pdgfr-α retained the ability to proliferate in the absence of PDGF-AA (Supplementary Figure 2E). These results show that PDGFR-α is required for progression to a growth factor independent state, but becomes redundant after transformation occurs.

Transformed cells have an oligodendrocyte progenitor cell (OPC)–like profile, proliferate despite DNA damage, and evade apoptosis. The capacity of null SVZ cells to transform in PDGF-AA prompted us to ask what cell types were present in the initial (ie, primary) SVZ cultures and the transformed cultures. To assess lineage profiles, we examined the stem and progenitor markers Sox2, Nestin, and GFAP; the OPC markers Olig2, NG2, and PDGFR-α; and markers of more mature cells, CNPase (oligodendrocyte lineage) and β-III-tubulin (neuronal lineage) in primary and transformed cultures. Compared with primary cultures, the percentage of transformed cells expressing stem/progenitor (Sox2, Nestin, and GFAP) or OPC (Olig2, PDGFR-α, and NG2) markers increased significantly, whereas the fracture with mature markers (CNPase and β-III-tubulin) remained low and constant (Fig. 3A, B and Supplementary Figure 3A). These data are consistent with the interpretation that stem and OPC-like null SVZ cells survived preferentially in PDGF-AA and eventually transformed.^15,16 To further characterize the features of null cells exposed to PDGF-AA, we analyzed transformed lines for the expression of stem-cell surface markers associated with human GBM. Four lines showed variable expression of CD15, CD34, CD44, and CD133 (Fig. 3C). Similar marker heterogeneity is also seen in human GBM.^17–19

Fig. 3.

Transformed p53-null SVZ cells have an OPC lineage profile, proliferate in the presence of DNA damage, and evade apoptosis. (A, B) Assessment of stem and progenitor markers in primary versus transformed null cell cultures reveals that the percentage of markers consistent with a neural stem/progenitor identity increases significantly in transformed cultures, while markers of mature neural cells remain relatively constant and low. (C) Stem and cancer related surface marker analysis reveal heterogeneous expression of CD15, CD33, CD44, and CD133 in 4 independent transformed lines (I-IV). (D) Null cultures established in EGF/FGF retain the capacity to transform when transferred to PDGF-AA, but are not viable in the absence of growth factors. (E) γH2AX staining reveals an increase in DNA damage foci in PDGF-AA cultured cells compared with EGF/FGF. (F) γH2AX foci in null SVZ cells cultured in EGF/FGF or PDGF-AA increase with time in culture and are more abundant in PDGF-AA. (G) P53 wild-type and null cultures stop synthesizing DNA 6 days after being transferred from EGF/FGF to PDGF-AA. (H) P53 Wild-type and null cells undergo increased apoptosis when transferred from EGF/FGF to PDGF-AA. (I) Trypan blue exclusion demonstrates that wild-type and null cells undergo decreased proliferation in PDGF-AA; however, after 10 days null cells resume proliferating. (NS = P> 0.05; *P < 0.05; **P< 0.01; ***P< 0.001; ****P< 0.0001)

To determine if SVZ cultures established in EGF/FGF retained the capacity to become growth factor independent and tumorigenic null cells maintained in EGF/FGF for a year were transferred to PDGF-AA. While control cultures ceased proliferating when EFG/FGF was removed, those transferred to PDGF-AA immediately displayed the “PDGF-AA phenotype”; namely, cell death and attenuated proliferation initially (Fig. 3D) followed by abrupt rapid sphere formation approximately 100 days later, and thereafter, PDGF-AA independent proliferation and GBM formation in immune-competent mice. These findings reveal that null SVZ cells retain the potential to evolve to a growth factor independent state despite prolonged culturing in EGF/FGF, suggesting that exposure to PDGF-AA is the key determinant of progression to growth factor independence in this model system.

After observing extensive SVZ cell death in PDGF-AA, we asked whether DNA damage had occurred, reasoning that damage might lead to the transformation of null cells. To assess strand breakage, we quantified γH2AX foci in null and wild-type cells in PDGF-AA or EGF/FGF and saw greater numbers of foci in PDGF-AA, irrespective of genotype (Fig. 3E and Supplementary Figure 3C, D). Then, to assess the effect of time in culture on the abundance of strand breaks in null cells in EGF/FGF versus PDGF-AA, γH2AX foci were counted in early and late passage cells and in transformed cells (Fig. 3F). In EGF/FGF, the mean number of foci per cell in early passage cultures was 4.6 versus 8.2 at later times (P < 0.0001), while in PDGF-AA, the number of foci increased from 6.30 early on to 16.4 (P< 0.0001) post-transformation. These experiments show that DNA breaks increase over time in both conditions, but are more numerous in PDGF-AA.

The ability of p53-null SVZ cells to proliferate despite DNA damage suggested that null cells in PDGF-AA are unable to activate cell cycle arrest or apoptotic functions after strand breakage due to loss of p53-dependent DNA damage checkpoint(s). To test this hypothesis, cell death and DNA replication were assessed in wild-type and null cells 6 days after PDGF-AA exposure. We observed that extensive cell death occurred and DNA replication ceased in both wild-type and null cells (Fig. 3G, H). Indeed at day 10, there were equally few viable wild-type and null cells in PDGF-AA (Fig. 3I). Thereafter, however, the culture phenotypes diverged: the number of wild-type cells continued to decline, whereas the number of null cells increased. After 20 days in PDGF-AA, there were no viable cells in the wild-type cultures, whereas the null cultures had expanded. These results demonstrate that absence of p53 allows a small population of null cells with strand breaks to evade apoptosis, proliferate in the presence of PDGF-AA, and transform.

Chromosome Instability Occurs in SVZ Cells in PDGF-AA

The capacity of null SVZ cells to tolerate DNA damage, become growth factor independent, and tumorigenic led us to search for evidence of genomic instability, a known cancer causing mechanism.²⁰ G-band karyotyping was used to assess ploidy, chromosomal integrity, and clonality in multiple SVZ cultures over time. In early passage null cell cultures in PDGF-AA, a small percentage of cells (15%) were observed to be polyploid (Fig. 4A). Clonal chromosomal alterations were infrequent in early passage cultures but losses of specific chromosomes were seen, notably 12 and 14. Late passage cultures had increasingly complex karyotypes with mosaic and near-tetraploid forms (Fig. 4B). By P-7, 90% of metaphase spreads were tetraploid, and by P-18, 100% were polyploid with clonal and subclonal populations (Fig. 4B). Furthermore, by P-18, 90% of cells had lost chromosome 2, 85% had lost chromosome 12, and 100% had lost the Y chromosome, while 60% had gained at least one copy of chromosome 11. In contrast, greater than 80% of null SVZ cells in EGF/FGF had near-diploid karyotypes at all passages tested. Although occasional gains of chromosomes were seen, there were no chromosomal losses in EGF/FGF supplemented cultures (Fig. 4C, D).

Fig. 4.

P53-null SVZ cells transformed in PDGF-AA have chromosomal instability. (A) G-band karyotyping of early passage p53-null cells cultured in PDGF-AA reveal heterozygous loss and gains of chromosomes (arrows). (B) G-band karyotypes of early (P4) to late (P18) passage p53-null cultures are shown. Transformed cells display chromosomal alterations that appear early and become increasing complex. (C) G-band karyotyping of p53-null EGF/FGF cultures reveals relatively normal chromosomal number and structure. (D) P53-null EGF/FGF cultures have relatively stable karyotypes over time. (E) aCGH of an early passage p53-null cells in EGF/FGF versus 3 pairs of early passage PDGF-AA dependent null cells and matched cell lines transformed in PDGF-AA. Cells in EGF/FGF have few alterations, whereas null cells in PDGF-AA have many large chromosomal alterations that increase over time in culture.

To further document the nature of these chromosomal alterations, high-resolution array comparative genomic hybridization (aCGH) was performed on 3 matched early and late passage null cell cultures in PDGF-AA and a single null culture in EGF/FGF. As predicted by their near normal karyotypes, null cells in EGF/FGF had no gross chromosomal alterations (Fig. 4E). However, major chromosomal changes, including whole and partial gains and losses, were seen in PDGF-AA (Fig. 4E). Although some were line specific, recurrent changes were seen, including losses of portions of chromosomes 12, 14, and 19; these regions of recurrent loss are syntenic to areas of the human genome important in GBM, or contained genes lost in GBM: chromosome 12 (syntenic to human 14 is often lost in GBM); chromosome 14 is the site of retinoblastoma (RB)²¹; and 19, the site of PTEN. There were also recurrent gains of chromosomes 3, 8, 11, and 15. These data show that null SVZ cells that transform in PDGF-AA have unstable genomes, a defining feature of GBM.

Discussion

The earliest stages in the initiation of IDH wild-type GBM are inaccessible to study, hidden within a molecularly heterogeneous disease that often develops explosively with no obvious in situ or low-grade stage. Here we address this obstacle to a deeper understanding of human GBM by developing an in vitro murine system in which the initiation of a GBM-like cancer can be studied. To our knowledge, this is the first model of GBM that closely replicates the genomic architecture of the human disease, especially recurring chromosomal gains and losses.

Here, we uncover a dichotomous role for PDGF-AA in the initiation versus maintenance of GBM. We show that PDGFR-α is required to transform null SVZ cells cultured in PDGF-AA, but is not required to maintain their oncogenic phenotype. Inactivation or knockout of PDGFR-α does not alter the proliferative capacity of SVZ cells that have transformed in PDGF-AA, nor does it affect their potential to form GBMs in the brains of immune-competent mice. This important observation further underscores that some initiators of malignant transformation will be poor therapeutic targets, a finding that is consistent with certain clinical experiences. For example, IDH wild-type GBMs are resistant to drugs that inhibit PDGFR-α signaling,²² despite the fact that gains of chromosome 7 and associated PDGF-AA overexpression are strongly implicated in GBM pathogenesis.² We also find that rapid proliferation of null SVZ cells, a phenomenon that could be readily assumed to lead to the accumulation of transformative genetic alterations, is insufficient to initiate GBMs in this model system; null cells divide rapidly in EGF/FGF but do not transform. This counterintuitive result is consistent with the observation that EGF overexpression and receptor activation are insufficient to produce GBMs in mice,^23,24 suggesting that amplification of EGFR might not be an initiating event in this disease, as is sometimes assumed. Our finding that null SVZ cells that have transformed in PDGF-AA show secondary RTK activation may be a clue to the origins of EGFR amplification in human GBM.

The most remarkable finding emerging from our work is the demonstration that chronic exposure of SVZ cells with p53 compromise (ie, null or heterozygous) to a ubiquitous growth factor leads to apoptosis, attenuated proliferation, and eventually transformation. Although the mechanisms underlying these responses remain to be elucidated, we found evidence of extensive chromosome instability only in the PDGF-AA growth factor condition. Indeed, one of the earliest alterations in pre-transformed cells was recurrent losses of specific chromosomes or chromosomal regions. High levels of DNA damage were also found in null cells, especially after prolonged exposure to PDGF-AA, but strand breaks were not unique to the PDGF-AA condition. These observations introduce the possibility that chromosome segregation errors associated with (over)exposure to PDGF-AA underlie the transformation of neural stemlike cells in which the p53 gene or its pathway has been compromised. Indeed, such segregation errors have been reported to induce mitotic arrest,²⁵ DNA breaks,²⁶ and apoptosis,²⁵ leading to gains and losses of whole chromosomes,²⁷ prominent features of this system. Intriguingly, the areas of recurrent losses in our model harbor GBM-associated tumor suppressor genes, including Rb and Pten.

One final point merits comment. The progression from PDGF-AA dependent to independent cell proliferation and to tumorigenesis over 100 days, signaled by abrupt rapid sphere formation, is reminiscent of the behavior of IDH wild-type GBM, a cancer that typically appears suddenly. The monophasic presentation of this type of GBM belies its vast molecular heterogeneity at diagnosis, a paradox suggesting that occult genome instability without visible tumor growth is a feature of the earliest stages of this cancer. In our model, a period of attenuated proliferation in PDGF-AA precedes the rapid expansion and transformation of p53-null SVZ cells. Perhaps our system can be a tool for understanding and intercepting the prodromal phase of some GBMs.

Conflict of interest statement. The authors declare no conflicts of interest.

Authorship statement. AKB, JAD, and CEB established and characterized the evolution of SVZ cultures; NC, EC, LM, HP, CC, HO, SB, and YS performed important analyses; AJG, KNT, and MAW conducted the receptor phosphorylation and PDGFR- α experiments; DB and US undertook the lineage profile analysis; SL, RM, and JAC characterized the tumors and their infiltrates; YY did the CRISPR knockdown; RW conducted the karyotype analyses; CJG, VWY, PJC, JJK, SJMJ, SW, and ECH provided insights and critical commentary; JJK and MDB made the observation that inspired this work; and MDB and JGC designed the experiments and wrote the manuscript.

Funding

This work was supported by the Terry Fox Research Institute and Foundation, the Alberta Cancer Foundation, Genome Canada, Alberta Innovates, and the family of Clark Smith.