[1] 12w Oligodendrocyte progenitor cells can act as cell of origin for experimental glioma

[1] 173w Gliomas are primary brain tumors mainly affecting adults. The cellular origin is unknown. The recent identification of tumor-initiating cells in glioma, which share many similarities with normal neural stem cells, has suggested the cell of origin to be a transformed neural stem cell. In previous studies, using the RCAS/tv-a mouse model, platelet-derived growth factor B (PDGF-B)induced gliomas have been generated from nestin or glial fibrillary acidic protein-expressing cells, markers of neural stem cells. To investigate if committed glial progenitor cells could be the cell of origin for glioma, we generated the Ctv-a mouse where tumor induction would be restricted to myelinating oligodendrocyte progenitor cells (OPCs) expressing 2 0 ,3 0 -cyclic nucleotide 3 0 -phosphodiesterase. We showed that PDGF-B transfer to OPCs could induce gliomas with an incidence of 33%. The majority of tumors resembled human WHO grade II oligodendroglioma based on close similarities in histopathology and expression of cellular markers. Thus, with the Ctv-a mouse we have showed that the cell of origin for glioma may be a committed glial progenitor cell.

[1] 86w Glioma is the most common primary brain tumor in adults and consists mainly of astrocytic and oligodendrocytic tumors of different malignancy grades. The most malignant type, glioblastoma (WHO grade IV), is almost always fatal with an average survival of 5-12 months after diagnosis (Ohgaki et al., 2004;Stupp et al., 2005). The heterogeneous and invasive properties of glioma cells make efficient therapy very difficult. Despite combined treatment with surgical resection, radiation therapy and chemotherapy tumors often recur and in addition have a high tendency for malignant progression.

[2] 270w Because tumor development is composed of a complex series of events involving not only tumor cells but also the surrounding tissue, in vivo methods must be used to recapitulate the oncogenic process fully. For this purpose mouse models have been developed where the resulting tumors resemble their human counterparts histologically and genetically. One such model is the well-characterized RCAS/tv-a mouse model where gene transfer is mediated by oncogene carrying RCAS (replication-competent avian sarcoma-leukosis virus long terminal repeat with splice acceptor) retroviruses to specific somatic cells in Tv-a transgenic mice (Uhrbom and Holland, 2001). RCAS is based on the subgroup-A avian leukosis virus (ALV) and expression of the receptor for ALV-A, TV-A, is restricted to avian cells. By generating transgenic mice expressing TV-A under the control of tissue or cell-type specific promoters, exclusive infection by RCAS has been obtained to TV-A-expressing cells (Holland et al., 1998;Holland and Varmus, 1998). In Ntv-a mice the Nestin promoter controls expression of TV-A directing it to neural/glial stem cells (Lendahl et al., 1990;Reynolds and Weiss, 1992). In Gtv-a mice TV-A expression is driven by the glial fibrillary acidic protein (Gfap) promoter, which is active in astrocytes (Bignami et al., 1972) and neural stem cells of the SVZ (Doetsch et al., 1999). The effect by various oncogenes cloned into RCAS on gliomagenesis has been extensively studied in the Ntv-a and Gtv-a mice (Doetsch et al., 1999;Holland et al., 2000a, b;Dai et al., 2001;Fults et al., 2002;Rao et al., 2003;Rao et al., 2004;Shih et al., 2004;Uhrbom et al., 2005;Jackson et al., 2006;Dunlap et al., 2007;Ferletta et al., 2007;Liu et al., 2007;Tchougounova et al., 2007;Binning et al., 2008).

[3] 205w The cell of origin for gliomas remains unidentified but has been suggested to be a glial stem or progenitor cell based on histopathology and expression of glial markers in the tumors. Recent findings of a glioma subpopulation of tumor-initiating cells closely resembling normal neural stem cells have implied that the cell of origin may be a transformed neural stem cell (Singh et al., 2004). Studies on experimental glioma using the RCAS/tv-a model have shown that neural/glial stem cells in Ntv-a mice and neural stem cells/astrocytes in Gtv-a mice have the capacity to generate gliomas (Holland et al., 1998;Holland and Vermus, 1998). The majority of induced tumors in both Ntv-a and Gtv-a mice were positive for NG2 chondroitin sulfate proteoglycan (NG2), a marker of early oligodendrocyte progenitor cells (OPCs) (Stallcup and Beasley, 1987;Pringle and Richardson, 1993;Nishiyama et al., 1996;Polito and Reynolds, 2005). This implied that the OPCs could represent the optimal level of cellular differentiation for oncogenic transformation to occur. To test if progenitor cells of the oligodendrocytic lineage could be the cell of origin for glioma we decided to make the Ctv-a transgenic mouse where tumorigenesis would be restricted to myelinating OPCs using the 2 0 ,3 0 -cyclic nucleotide 3 0 -phosphodiesterase (Cnp) promoter.

[1] 241w Generation and functional testing of the Ctv-a transgenic construct CNP is one of the earliest known myelin-specific proteins to be synthesized by developing oligodendro-cytes (Vogel and Thompson, 1988). Several studies have shown a specific expression of CNP during differentiation of oligodendrocytes (Scherer et al., 1994;Yuan et al., 2002) and it has been proven to faithfully direct expression of a transgene to OPCs and differentiated oligodendrocytes (Gravel et al., 1998;Chandross et al., 1999;Yuan et al., 2002). The Cnp promoter consists of two separate promoter regions resulting in two isoforms, Cnp1 (46 kDa) and Cnp2 (48 kDa), encoded by the same gene (O'Neill et al., 1997). The proximal promoter, which regulates transcription of Cnp1 mRNA, contains one open-reading frame for CNP1 protein and is mainly active after birth. The distal promoter regulates transcription of Cnp2 mRNA, which produces both CNP1 and CNP2 polypeptides and is active both in embryonic brain and postnatally (O'Neill et al., 1997). The transgenic construct contained the entire mouse Cnp promoter region coupled to the Tv-a gene followed by an ires (internal ribosomal entry site) sequence and the lacZ gene (Figure 1a). The transgenic construct was functionally tested in the CNP-expressing OLI-neu cell line (Jung et al., 1995). OLI-neu cells were transiently transfected with the transgenic construct and transfected cells were subsequently infected with supernatant containing RCAS-eGFP. Successful infection resulting in enhanced green fluorescent protein (eGFP)-positive cells was visualized with fluourescence microscopy confirming functional expression of TV-A (Figure 1b).

[2] 177w Ctv-a directs functional tv-a expression to OPCs Transgenic mice were generated, chimeras were identified and after initial analysis one founder line was selected for further investigation. To evaluate the in vitro expression of the transgene, we established primary tissue cultures from neonatal Ctv-a brains and infected them with supernatant from DF-1 cells producing RCAS-eGFP (Figure 1c). Approximately 5-10% of the cells were infected and these cells were bipolar with small, round nuclei ('progenitor-like') in early passages but soon became more differentiated containing multipolar, arbourated processes. Infected cells were analysed with immunocytochemistry (Figure 1c) and were shown to express CNP, and CNP-expressing cells were shown to express the b-Galactosidase reporter gene. There were other cell types present in the culture dish, the predominant type being larger cells reminiscent of astrocytes that were never found to express eGFP. These results confirmed that the RCAS-eGFP-infected cells were OPCs or oligodendrocytes and that the transgene was expressed exclusively in cells of the oligodendrocytic lineage. Furthermore, an X-gal staining of primary tissue cultures displayed positive cells with characteristic oligodendrocytic morphology (Figure 1d).

[3] 272w Transgene expression was further characterized with in situ immunohistochemical stainings of P6 Ctv-a mouse brains. Because neonatal and early postnatal CNP protein (most likely derived from the Cnp2 mRNA) cannot be detected with the existing antibodies although the mRNA is present (Gravel et al., 1998) we detected expression of TV-A by analysing the expression of its reporter gene b-Galactosidase. We found expression of b-Galactosidase throughout the entire brain. Most of the positive cells displayed the morphology of maturing oligodendrocytes with extended processes whereas some had a more bipolar appearance. Costainings for b-Galactosidase and CNP showed that the transgene expression was restricted to CNP-positive cells (Figure 2a). However, not all CNP-positive cells were positive for b-Galactosidase. NG2 is a marker of OPCs, which has been shown to be co-expressed with eGFP in the brains of CNP-eGFP transgenic mice at P2 (Yuan et al., 2002). More importantly, early postnatal CNP-eGFP-positive cells have been isolated from these transgenic mice and were shown to express NG2 and represent multipotent OPCs (Belachew et al., 2003). Co-staining of P6 Ctv-a mouse brains for NG2 and b-Galactosidase showed that most b-Galactosidasepositive cells were negative for NG2 (Figure 2b). The same was found for the OPC marker PDGFRa (Figure 2c). Both NG2 and PDGFRa are expressed earlier than CNP during oligodendrocyte differentiation, which could explain the result. To analyse if the Ctv-a mice would have overlapping TV-A expression with the Ntv-a mice (in Nestin-expressing cells) or the Gtv-a mice (in GFAP-expressing cells) we carried out co-stainings for b-Galactosidase and Nestin or GFAP, respectively. We could never find any Nestin or GFAPpositive cells among the b-Galactosidase-positive cells (Figures 2d and e).

[4] 111w To show that the transgene was functionally expressed in vivo, we injected newborn mice with DF-1 cells producing RCAS-eGFP retrovirus and analysed P5 brains. Most of the GFP-positive cells found were injected DF-1 cells and only a few were infected mouse cells. The GFP-positive infected mouse cells simultaneously expressed CNP (Figure 3a) and b-Galactosidase (Figure 3b) but were negative for Nestin (Figure 3c) and GFAP (Figure 3d), supporting the specificity of the TV-A expression. The DF-1 cells did not express any of the glial markers (not shown). An antibody specific for chicken Vimentin was used to show that the majority of the GFP-positive cells truly were injected DF-1 cells (Figure 3e).

[5] 31w In conclusion, we have shown that the transgene was exclusively expressed in OPCs and oligodendrocytes and that both in vitro and in vivo retroviral infections were restricted to the oligodendrocytic lineage.

[6] 152w The potential of platelet-derived growth factor B (PDGF-B), K-RAS and AKT oncogenes to induce tumors in neural/glial stem cells (Ntv-a mice) and astrocytes (Gtv-a mice) has already been thoroughly investigated using the RCAS/tv-a model (Holland et al., 2000a, b;Dai et al., 2001;Uhrbom et al., 2002Uhrbom et al., , 2005;;Tchougounova et al., 2007). We therefore selected these oncogenes in our initial tumor induction studies in the Ctv-a mice to compare the tumor-initiating capacity of the different cell types targeted. Neonatal mice were injected intracerebrally with retrovirus-producing cells and monitored for 12 weeks. By limiting the in vivo experiments to 12 weeks we minimized the effect of insertional mutagenesis and spontaneous mutations as contributing factors. RCAS-PDGF-B induced tumors in 33% of the mice (Table 1). All tumors were found in seemingly healthy mice at 12 weeks of age. AKT and K-RAS, alone or in combination, were not able to induce tumors in Ctv-a mice.

[7] 248w The PDGF-B-induced tumors were all located intraand/or periventricularly. The histopathology showed some variation but all tumors were similar to human glioma (Figures 4a and b). Most tumors resembled oligodendroglioma except three tumors that had mixed oligodendrocytic and astrocytic components. Some of the oligodendrogliomas displayed perinuclear halos, a tissue processing artifact commonly seen in human oligodendroglioma. All but one tumor were diagnosed as low-grade gliomas (WHO grade II). The increased cellular density and presence of mitotic figures in one tumor resulted in the classification as a WHO grade III glioma. Immunohistochemical analysis of the tumors displayed robust expression of the neural stem cell and OPC marker SOX2, and of the OPC markers SOX10, OLIG2, NG2 and PDGFRa in all tumors (Figure 4a). None of the tumors expressed the glial stem cell marker Nestin (not shown). The astrocytic components of the mixed oligoastrocytomas were intensely positive for GFAP and Vimentin, but negative for SOX10 (Figure 4b). In the oligodendroglial tumors only a small fraction of reactive astrocytes were positive for GFAP and Vimentin (not shown). Expression of CNP and b-Galactosidase was absent in all tumors, but CNP could be detected in surrounding normal brain, implying that a slight dedifferentiation of tumor cells had occurred (not shown). The proliferation index was 2-5%, determined by counting Ki67-positive cells (Figure 4b). Thus, the PDGF-B-induced tumors showed expression of many markers of early OPCs in line with previous result from PDGF-B-induced tumors in Ntv-a and Gtv-a mice (Dai et al., 2001;Tchougounova et al., 2007).

[8] 77w To validate that induced tumors were caused by retroviral PDGF-B, we induced additional tumors using RCAS-PDGF-B-HA. Out of 17 injected mice 6 developed tumors (35%) with the same histopathology as seen previously. All tumors were positive for hemagglutinin (HA) with 60-90% of cells within the tumor area being labeled (Figure 4b). Because PDGF-B is a secreted protein, HA-positive cells may be either infected with RCAS-PDGF-B-HA causing autocrine activation of PDGF receptors or stimulated in a paracrine manner.

[9] 184w To validate that the experimental gliomas were representative for human low-grade gliomas, we analysed nine human grade II gliomas (Figures 4c and d), consisting of four astrocytomas, four oligodendrogliomas and one oligoastrocytoma for expression of the same cellular markers. All human tumors expressed SOX2, SOX10 and OLIG2 (Figure 4c). SOX2 was expressed in almost all tumor cells of both astrocytic and oligodendrocytic origin, whereas SOX10 and OLIG2 were expressed in only some of the tumor cells. About half of the tumors were focally positive for expression of PDGFRa (Figure 4c). A majority of the tumors showed scattered areas containing CNP-positive cells with the morphological characteristics of normal oligodendrocytes and myelinated axons (Figure 4c). In the gemistocytic astrocytomas the gemistocytes were positive for both SOX2 and OLIG2 (Figure 4d). All astrocytic components of the tumors were positive for GFAP and Vimentin (Figure 4d). All tumors were negative for the glial stem cell marker Nestin (not shown). The amount of proliferating cells, as shown by Ki67 staining (Figure 4d), was 1-3%. Overall, the expression of cellular markers was similar between the mouse and human low-grade gliomas.

[10] 162w Interestingly, many of the PDGF-B-injected mice developed hydrocephalus before 3 weeks of age. These mice were removed from the study. When analysing sections of these mouse brains small hyperplastic lesions in close proximity to the ventricles were found. Immunostainings of the hyperplasias using antibodies against CNP to detect possible target cells for retroviral infection, PDGFRa to detect infected glial cells responding to PDGF-B signaling by receptor upregulation and chicken Vimentin to detect the RCAS-PDGF-B producing DF-1 cells (Figure 5). We found that the hyperplasias contained scattered DF-1 cells and cells positive for CNP and PDGFRa. This could imply that the periventricular hyperplasias were a site of tumor initiation, further supported by the fact that all gliomas showed intra-or periventricular location. No cells were positive for both CNP and PDGFRa, a plausible explanation being that infected cells responding to oncogenic PDGF-B stimulation undergo rapid dedifferentiation. This is further supported by the fact that all gliomas found in 12-week-old Ctv-a mice lacked CNP expression.

[11] 264w The O2-A glial progenitor cell that can give rise to oligodendrocytes and type-2 astrocytes in vitro has been shown to proliferate in response to PDGF and differentiate on withdrawal of PDGF (Noble et al., 1988;Raff et al., 1988). Also in vivo O2-A cells have been shown to respond to PDGF signaling by proliferation (Calver et al., 1998). In previous studies using the RCAS/tv-a model PDGF-B has been shown stimulate proliferation in vitro of both Nestin-and GFAPexpressing glial cells (Dai et al., 2001). To analyse how oncogenic activation would affect Ctv-a-derived OPC proliferation, we cultured primary cells from neonatal Ctv-a brains under conditions to enrich for OPCs (same as O2-A cells). The chronic in vivo oncogenic stimulation was mimicked by infecting OPC cultures for 17 days before the experiment. Viral infection was verified with immunofluorescence staining of the infected cells for GFP (PDGF-B), HA (AKT) and p-ERK (K-RAS) (data not shown). The number of cells on days 4 and 7 was determined and presented as fold change in relation to the number of cells on day 1 (Figure 6). Relative numbers were used because the absolute number of cells differed greatly between experiments due to variations in OPC density in the primary cultures and infection efficiency between experiments. PDGF activation was clearly more potent in stimulating proliferation than eGFP (control), K-RAS, AKT or combined K-RAS and AKT infection. The result was in line with published results from O2-A cells, supported the in vivo effect in Ctv-a mice and may explain why OPCs were susceptible to PDGF-B-induced tumorigenesis but insensitive to oncogenic K-RAS and AKT activation.

[12] 70w In addition, PDGF-B may contribute to gliomagenesis by other mechanisms such as sustaining survival or interfering with differentiation. PDGF has been shown to increase survival of O2-A cells in culture (Barres et al., 1992). In GFAP-expressing astrocytes autocrine PDGF stimulation was shown to induce dedifferentiation in vitro (Dai et al., 2001). Furthermore, the complete absence of CNP expression in Ctv-a tumor cells suggests that PDGF-B may induce dedifferentiation of OPCs.

[13] 65w The molecular pathways through which PDGF could induce gliomas in OPCs remain to be elucidated but our data together with what is previously known about PDGF in OPCs suggest that one essential property is to stimulate proliferation. The tumors occurred in a relatively short time (within 12 weeks), and if they were caused by spontaneous mutations, they would likely have taken much longer to develop.

[1] 68w The initial genetic change is crucial for tumorigenesis to occur. The effect of such an event is to a large extent dependent on the cellular context, that is, the differentiation state of the target cell as well as the surrounding tissue. Hence, to study the initial event in vivo the model used should have the capacity to specifically target the presumed cell of origin in its natural microenvironment.

[2] 218w The recent isolation of brain tumor stem cells in glioma (Singh et al., 2004) has promoted the belief that gliomas arise from transformed neural stem cells. Knowledge acquired before this study using the Ntv-a and Gtv-a mice has shown that gliomas may occur by transformation of stem cells and possibly even astrocytes, because tumors were induced in neonatal mice, which would likely have excluded the GFAP-positive SVZ neural stem cells from being infected as these first have been identified at P6 (Merkle et al., 2004). In the present investigation we have explored the ability of myelinating OPCs to be the cell of origin for glioma by developing the Ctv-a mouse. This is to the best of our knowledge the first mouse model where gliomas have been induced in a committed glial progenitor cell. Gliomas could be induced and the tumors showed histopathology and marker expression similar to human glioma. Interestingly, expression of CNP in the mouse brain tumor cells was absent and could only be found in scattered normal cells within the tumor. The fact that the tumor cells clearly expressed SOX2, OLIG2 and NG2 implied that the oncogenic stimulation had caused a slight dedifferentiation of the targeted OPCs. In conclusion, we clearly demonstrate that committed OPCs have the capacity to be the cell of origin for glioma.

[3] 136w Brain tumor stem cells have yet to be studied in experimental glioma and future investigations will aim at exploring their presence and qualities in mouse tumors of different cellular origin. We believe that the Ctv-a mouse may also be a useful model for studies of the peripheral nervous system because the Cnp promoter also targets Schwann cells (Gravel et al., 1998). In a recently described transgenic mouse model of neurofibroma, expression of the human epidermal growth factor receptor (hEGFR) was directed to CNP-expressing Schwann cells (Ling et al., 2005). These mice developed Schwann cell hyperplasias but showed no signs of glioma development, which was explained by the lack of EGFR ligands in the brain. This further supports the notion that OPCs are depending on FGF-2 and PDGF signaling rather than EGF signaling (Bogler et al., 1990).

[4] 148w If we put the result from the Ctv-a mice in the perspective of previous data using the RCAS/tv-a model (Dai et al., 2001;Tchougounova et al., 2007; Table 2) we can conclude that there are several cell types of the glial cell lineage that are sensitive to PDGF-B transformation and may serve as cell of origin for glioma. The fact that AKT and K-RAS could induce tumors only in Nestin-expressing stem cells and not in OPCs or astrocytes could imply that there are tumor suppressor pathways activated in these cells that can counteract the effects of AKT and K-RAS signaling. In astrocytes, the expression of both p16INK4a and p19ARF is possible candidates as there is a prominent increase in PDGF-Binduced tumor incidence on loss of either of these oncogenes in Gtv-a mice (Tchougounova et al., 2007). The effect of Ink4a-Arf loss in OPCs in vivo remains to be elucidated.

[1] 54w Human tissue was obtained from biopsy material collected from patients undergoing surgical treatment by informed consent and after approval by the regional ethical committee. All tumors had been clinically diagnosed as low-grade glioma and were for the purpose of the study, together with the analysis of the mouse tumors, reevaluated by a neuropathologist (TO).

[2] 138w Primary cultures were prepared and infected as described above. Infection was repeated every second day for 17 days with addition of 10 ng/ml PDGF-AA (PeproTech) and 10 ng/ml bFGF (PeproTech). Following infection, 3 Â 10 4 cells were seeded for proliferation assay with continued addition of bFGF every second day. Cells were counted 1, 4 and 7 days after plating using a Coulter Counter (Coulter Electronics, Bromma, Sweden). Within each group and each experiment the relative number of cells was calculated in relation to the number of cells by day 1 because unequal amount of cells was attached after seeding and there were great variations between each experiment. The result is an average of three independent experiments, except for PDGF in which case two experiments were performed. Statistical analysis was carried out using the GraphPad Prism 4.0 software.

[1] 93w The Ctv-a plasmid was transfected using FuGENE (Roche Diagnostics, Bromma, Sweden) into the CNP-positive immortalized oligodendroglial cell line OLI-neu (Jung et al., 1995) in passage 42 cultured in modified oligodendrocyte and oligodendrocyte precursor selective Sato media (Supplementary Materials and methods; Bottenstein and Sato, 1979) on plates coated with poly-L-lysine (Sigma-Aldrich, St Louis, MO, USA). Transfected cells were infected for 4 days with sterile filtered media from RCAS-eGFP-producing chicken fibroblasts (production of RCAS-derived retroviruses using DF-1 cells was described by Holland et al, 1998). Successful infection was visualized using a Zeiss inverted fluorescence microscope.

[2] 64w Cells were fixed in 4% paraformaldehyde (PFA), permeabilized in 0.2% Triton and blocked in 0.1% saponin, 4 mg/ml bovine serum albumin (BSA) and 2% normal goat serum (NGS) followed by incubation with a primary antibody and secondary antibody TRITC rabbit anti-mouse (Dako, Stockholm, Sweden). Slides were mounted in Immu-Mount (Shandon, Pittsburgh, PA, USA) containing 4,6-diamidino-2phenyl indole (DAPI) and evaluated using a Leica fluorescence microscope.

[3] 91w Generation and genotyping of Ctv-a transgenic mice All animal experiments were performed in accordance with the local animal ethics committee. Generation of Ctv-a mice was performed by pronuclear injection in FVB/N background at Uppsala University Transgene Facility. Pronuclear injection generated three chimera founders with germ-line transmission (A7, B8 and C16). All were phenotypically normal, viable and fertile. After initial characterization for expression of TV-A the B8 line was selected for further breeding. The founder Ctv-a mouse was crossed with FVB/N mice and the F1 generation was intercrossed to establish homozygous mice.

[4] 48w Genotyping was carried out on DNA purified from toe biopsies taken between days 7 and 12 after PCR reactions were performed using PCR beads (Amersham Biosciences, Uppsala, Sweden) and primers specific for Ctv-a, 5 0 Cnp (5 0 -CAGCT GTTGAAGATACTGAAG) and 3 0 tv-a (5 0 -GCCCTGGGG AAGGTCCTGCCC).

[5] 129w In vitro infection and immunohistochemical staining of primary brain cultures Primary cell cultures were established from neonatal mice by mechanically dissociating them using syringes with 18-and 22-gauge needles. The cell suspension was peletted, resuspended and seeded on plates. Neural/glial progenitor cells were plated on dishes coated with polyornithine and fibronectin in N2 media (Bottenstein and Sato, 1979) with addition of 10 ng/ml basic fibroblast growth factor (bFGF; PeproTech, Rocky Hill, NJ, USA). OPCs were plated in dishes coated with poly-L-lysine (Sigma-Aldrich) or polyornithine and fibronectin in Sato media with addition of 10 ng/ml bFGF and 10 ng/ml PDGF-AA (PeproTech). Astrocytes were cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco/Invitrogen, Carlsbad, CA, USA) containing 12% serum. Cells were infected for 7 days for characterization and seeded on coverslips for immunohistochemistry.

[6] 91w Primary brain cultures were infected for 5 days. Infected cells were fixed in 4% PFA for 10 min and incubated with primary antibody diluted in 0.25% Triton X-100 and 0.25% BSA in phosphate-buffered saline followed by secondary antibody. Cells were mounted in ImmunoMount (Shandon) containing DAPI. Stainings were visualized using a Zeiss 510 META confocal microscope. (2007) for PDGF-B. Pair-wise comparisons of incidence rates were calculated using Fischer's exact test for all possible combinations and a significant difference was found between Ntv-a and Gtv-a mice injected with PDGF-B (P ¼ 0.0110).

[7] 71w In situ staining of postnatal brain Brains from P6 mice or RCAS-eGFP-injected P5 mice were fixed in 4% PFA for 1 h in 4 1C, immersed in 20% sucrose at 4 1C overnight and cryosectioned. Sections were blocked using 5% milk, incubated with primary antibody diluted in blocking solution followed by secondary antibody. Slides were mounted in ImmunoMount (Shandon) containing DAPI. Stainings were visualized using a Zeiss 510 META confocal microscope.

[8] 212w In vivo tumor induction Neonatal (P1) mice were injected intracerebrally with 2 ml RCAS-producing DF-1 chicken fibroblasts (RCAS-PDGF-Bires-eGFP, RCAS-Akt or RCAS-K-Ras), which equals approximately 2 Â 10 5 cells using a 10 ml gas-tight Hamilton syringe. RCAS-PDGF-B-ires-eGFP contains a 927 bp fragment including the coding sequence of the PDGF-B chain and a fragment with the coding sequence of eGFP separated by an ires (Dai et al., 2001). However, the eGFP gene has been compromised and its expression cannot be detected in infected cells. RCAS-Akt contains the activated AKT (Akt-Myr D11-60) followed by an HA tag added to the 3' end and RCAS-K-Ras encodes the G12D point-mutation-activated K-RAS (Holland et al., 2000a, b). RCAS-PDGF-B-HA contains a 3nt 'ACC' Kozak consensus sequence in front of PDGF-B and a C-terminal HA epitope tag (Shih et al., 2004). DF-1 cells were cultured in DMEM with addition of 12% fetal calf serum (Gibco), 4 mM L-glutamine (Gibco) and 1% penicillin/streptomycin (Gibco). Mice were monitored three times per week and killed on any sign of illness or at 12 weeks. Brains were fixed in 4% formalin at 4 1C for at least 48 h, embedded in paraffin, sectioned and analysed for tumors using hematoxylin and eosin (H&E) staining. Statistical analysis was carried out using GraphPad Prism 4.0 software.

[9] 137w Immunohistochemical staining was carried out on 6 mm sections. Antigen retrieval was performed by pressure boiling in antigen unmasking solution (Vector Labs, Burlingame, CA, USA) or 1 mM EDTA (pH 8.0), sections were incubated in 1% H 2 O 2 in methanol for 30 min and blocked in 1.5% NGS followed by incubation with primary antibody. Biotinylated secondary goat anti-mouse, goat anti-rabbit, rabbit anti-goat or rabbit anti-rat antibodies were used at 1:200. Stainings were developed using Vectastain ABC system and DAB substrate kit (Vector Labs). For HA secondary antibody Alexa Flour 555 donkey anti-rabbit was used at 1:400 dilution and mounted in Immu-Mount (Shandon) containing DAPI. Pictures were taken using a Leica brightfield microscope and Zeiss 510 META confocal microscope. The proliferation index (fraction of Ki67-positive cells) was determined by counting at least 200 cells per tumor.