Oncogene (2007) 26, 6289–6296 & 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc ORIGINAL ARTICLE Loss of Arf causes tumor progression of PDGFB-induced oligodendroglioma

E Tchougounova1, M Kastemar1, D Bra˚ sa¨ ter1,3, EC Holland2, B Westermark1 and LUhrbom 1

1Rudbeck Laboratory, Department of Genetics and Pathology, Uppsala University, Uppsala, Sweden and 2Departments of Surgery (Neurosurgery), Neurology and Cancer Biology and Genetics, Memorial Sloan-Kettering Cancer Center, New York, NY, USA

In a subset of gliomas, the platelet-derived The tumors are believed to originate from a glial (PDGF) signaling pathway is perturbed. This is usually an progenitor cell or an astrocyte based on the expression early event occurring in low-grade tumors. In high-grade of various glial-specific mRNAs and in the gliomas, the subsequent loss of the INK4a-ARF locus tumors. Gliomas are diagnosed based on malignancy is one of the most common . Here, we dissected into four grades (I–IV). Owing to infiltrative growth and the separate roles of Ink4a and Arf in PDGFB-induced a strong tendency of malignant progression of tumor oligodendroglioma development in mice. We found that cells, the prognosis is poor even when diagnosed with a there were differential functions of the two tumor low-grade tumor. The oligodendrogliomas are divided suppressor . In tumors induced from astrocytes, both into grade II oligodendroglioma and grade III anaplas- Ink4a-loss and Arf-loss caused a significantly increased tic oligodendroglioma. They commonly display over- incidence compared to wild-type mice. In tumors induced expression of the epidermal from glial progenitor cells there was a slight increase in (EGFR) (Reifenberger et al., 1996) and of ligands and tumor incidence in Ink4aÀ/À mice and Ink4a-ArfÀ/À mice receptors of the platelet-derived growth factor (PDGF) compared to wild-type mice. In both progenitor cells and family (Di Rocco et al., 1998). In high-grade oligoden- astrocytes, Arf-loss caused a pronounced increase in tumor drogliomas, the malignant progression is often asso- malignancy compared to Ink4a-loss. Hence, Ink4a-loss ciated with homozygous of the INK4a-ARF contributed to tumor initiation from astrocytes and Arf-loss locus (Cairncross et al., 1998; Bigner et al., 1999), which caused tumor progression from both glial progenitor cells is one of the most frequent mutations found in human and astrocytes. Results from in vitro studies on primary glioma (Jen et al., 1994; Schmidt et al., 1994). INK4a- brain cell cultures suggested that the PDGFB-induced ARF encodes the tumor suppressor proteins p16INK4A activation of the mitogen-activated kinase pathway and p14ARF (p19Arf in mouse), which regulates the RB via extracellular signal-regulated kinase was involved in the and P53 pathways, respectively. Their separate roles in initiation of low-grade oligodendrogliomas and that the suppression of spontaneous tumor development in vivo additional loss of Arf may contribute to tumor progression have been studied in mice with targeted deletions of the through increased levels of cyclin D1 and a phosphoinositide corresponding genes (Serrano et al., 1996; Kamijo et al., 3-kinase-dependent activation of p70 ribosomal S6 kinase 1997a; Sharpless et al., 2001, 2004). Arf-null mice causing a strong proliferative response of tumor cells. showed a high frequency of spontaneous tumors in Oncogene (2007) 26, 6289–6296; doi:10.1038/sj.onc.1210455; diverse tissue types, whereas Ink4a-null mice demon- published online 16 April 2007 strated a low frequency of spontaneous tumors with a more restricted tissue distribution. The combined loss Keywords: oligodendroglioma; PDGF; mouse model; of Ink4a-Arf displayed a phenotype similar to that of Ink4a; Arf; p70S6K Arf-null mice. The somatic cell transfer model replication- competent avian leukemia virus splice acceptor/avian Introduction leukemia virus receptor (RCAS/tv-a) has been used to study oncogenic mechanisms of tumor genes during Gliomas, comprising astrocytomas and oligodendro- gliomagenesis. Infection with RCAS retroviruses carry- gliomas, are the most common primary tumors of the ing specific genetic mutations can be directed to specific central nervous system (Kleihues and Cavenee, 1997). brain cell populations using transgenic mice expressing tv-a from cell type-specific promoters. In Ntv-a mice, the Correspondence: Dr LUhrbom, Rudbeck Laboratory,Department of nestin promoter that supports transfer to neural/glial Genetics and Pathology, Uppsala University, SE-751 85, Uppsala, progenitor cells controls expression of tv-a.InGtv-a, Sweden. mice tv-a is expressed by the glial fibrillary acidic protein E-mail: [email protected] (GFAP) promoter that is active primarily in astrocytes 3Current address: Ludwig Institute for Cancer Research, Karolinska Institutet, Box 240, SE-171 77 Stockholm, Sweden. in the newborn mouse brain. In the Ntv-a mice, we have Received 29 June 2006; revised 6 February 2007; accepted 2 March 2007; previously shown that various types of gliomas (astro- published online 16 April 2007 cytomas and oligodendrogliomas) can be induced in Role of Ink4a and Arf in PDGF-induced gliomagenesis E Tchougounova et al 6290 wild-type mice by infecting with KRas þ Akt or PDGFB more of the following features had to be present: high retrovirus (Dai et al., 2001; Uhrbom et al., 2002). When tumor cell density with mitotic figures (Figure 1c), these oncogenic stimuli were combined with the loss of cellular and nuclear pleomorphism, microvascular pro- Ink4a-Arf tumor incidence, latency and malignancy liferation and pseudopalisading necrosis (Figure 1d). were increased (Dai et al., 2001; Uhrbom et al., 2002). Grade III tumors normally displayed grade II tumor Furthermore, for KRas-induced astrocytomas, we have histology in some parts of the tumor. Grade II tumors shown that there were functionally independent roles of were mostly found in asymptomous 12-week-old ani- loss of either Ink4a or Arf (Uhrbom et al., 2005). mals whereas grade III tumors usually presented earlier In the present study, we have investigated the than 12 weeks. Some tumors in Gtv-a mice had the individual contributions of loss of p16Ink4a and p19Arf histopathology of mixed oligo-astrocytomas. All tumors in PDGFB-induced oligodendroglioma development. were positive for the oligodendroglial cell marker NG2 We found that the main consequence of Ink4a-loss (Figure 1e) and the neural stem cell marker Sox2 was to render astrocytes susceptible to PDGFB-induced (Figure 1f). In addition, most tumors showed areas oncogenesis and that the primary result of Arf-loss was positive for Gfap (Figure 1g) and nestin (Figure 1h). increased tumor malignancy. Results from in vitro In Ntv-a mice, significant increase in tumor incidence studies on primary brain cell cultures supported these was found upon loss of either Ink4a (Fischer’s exact findings and suggested a mechanism by which test, P ¼ 0.0496 (*)) or Ink4a-Arf (Fischer’s exact test, tumor initiation was caused by activation of the P ¼ 0.0159 (*)) compared to wild-type mice (Table 1). In mitogen-activated protein kinase (MAPK) pathway via Gtv-a mice, the separate loss of Ink4a or Arf had a more extracellular signal-regulated kinase (Erk), and tumor striking effect (Table 1). There was a significant and progression occurred upon loss of p19Arf resulting in similar increase in tumor incidence in Ink4aÀ/À and increased levels of cyclin D1 and activation of the ArfÀ/À mice compared to wild-type mice (Fischer’s phosphoinositide 3-kinase (PI3K) pathway via p70 exact test, Po0.0001 (***) for both comparisons), and ribosomal S6 kinase (p70S6K). the combined loss of Ink4a-Arf did not further increase the tumor incidence. In line with what has previously been shown for Kras-induced gliomagenesis (Holland Results et al., 2000; Uhrbom et al., 2002), the incidence of PDGFB-induced transformation was significantly high- To dissect the effect of loss of each of the Ink4a-Arf gene er in Ntv-a wild-type mice than in Gtv-a wild-type mice products in vivo, Ntv-a and Gtv-a wild-type mice and (Fischer’s exact test, P ¼ 0.0159 (*)). However, upon mice lacking p16Ink4a, p19Arf or both were injected with loss of either Ink4a or Arf, or combined loss of both RCAS–PDGFB–EGFP (Table 1). As previously des- Ink4a-Arf the difference in tumor incidence between cribed (Dai et al., 2001), the PDGFB-induced tumors Ntv-a and Gtv-a mice disappeared. Thus, the major displayed a histopathology similar to human oligoden- effect of Ink4a-loss in PDGFB-induced oligodendro- droglioma (grade II) (Figure 1a and b) or anaplastic glioma formation was to increase the susceptibility of oligodendroglioma (grade III) (Figure 1c and d). The the target cells in Gtv-a mice to PDGFB-induced tumors consisted of small tumor cells with regular, transformation. round nuclei and some tumors displayed the ‘peri- The main consequence of Arf-loss was increased nuclear halo’ feature where the cytoplasm was cleared malignancy and shorter survival of tumor-bearing mice. from the tumor cell nuclei (Figure 1a and c). They grew This was most obvious in Gtv-a mice but the trend could diffusely into the normal brain parenchyma and also be seen in Ntv-a mice. In Gtv-a mice, there was a displayed occurrence of the ‘secondary structures of significant difference in survival between Ink4aÀ/À mice Scherer’ including tumor cells lining up in the white and ArfÀ/À mice, which was not further increased by matter tract of the corpus callosum (Figure 1b), the combined loss of Ink4a-Arf (Figure 2a). This perivascular satellitosis, perineuronal satellitosis and difference was paralleled by a significant increase in subpial infiltration (Scherer, 1938). To be classified as malignancy of ArfÀ/À tumors (Fischer’s exact test, an anaplastic oligodendroglioma, the addition of two or P ¼ 0.0001 (***)) and Ink4a-ArfÀ/À tumors (Fischer’s

Table 1 PDGFB-induced glioma incidence and malignancy grade Line Genetic background RCAS Number of mice Number of tumors Glioma incidence Grade II/III

Ntv-a Wild-type PDGFB 28 14 50 11/3 Ntv-a Ink4aÀ/À PDGFB 28 22 79 21/1 Ntv-a ArfÀ/À PDGFB 35 22 63 10/12 Ntv-a Ink4a-ArfÀ/À PDGFB 20 17 85 6/11 Gtv-a Wild-type PDGFB 31 5 16 5/0 Gtv-a Ink4aÀ/À PDGFB 28 23 82 18/5 Gtv-a ArfÀ/À PDGFB 26 20 77 3/17 Gtv-a Ink4a-ArfÀ/À PDGFB 28 19 68 5/14

Abbreviations: PDGFB, platelet-derived growth factor B; RCAS, replication-competent avian leukemia virus splice acceptor.

Oncogene Role of Ink4a and Arf in PDGF-induced gliomagenesis E Tchougounova et al 6291

Figure 1 H&E-stained sections showing the histopathology of oligodendroglioma grade II (a and b) and anaplastic oligodendroglioma grade III (c and d) and immunohistochemical stainings for glial cell markers in a grade III tumor from Ntv-a ArfÀ/À mouse (e–h). (a) A tumor displaying the ‘perinuclear halo’ feature where the cytoplasm is cleared from the tumor cell nuclei in an oligodendroglioma grade II. (b) Intrafascicular queing of tumor cells in the white matter tract of corpus callosum in a grade II tumor. (c) A grade III tumor showing the ‘perinuclear halo’ feature. (d) Pseudopalisading necrosis in a grade III tumor. Immunohistochemical analysis of the expression of NG2 (e), Sox2 (f), Gfap (g) and Nestin (h). Scale bar. 50 mm. exact test, P ¼ 0.0015 (**)), when compared with exact test, Po0.0001 (***)). Thus, loss of Arf caused Ink4aÀ/À tumors. In Ntv-a mice significant differences tumor progression in both Ntv-a and Gtv-a mice. in survival were found between wild-type mice and To investigate the molecular mechanisms by which Ink4aÀ/À mice or Ink4a-ArfÀ/À mice (Figure 2b), the separate loss of Ink4a and Arf contributed to which could be explained by increased tumor incidence PDGFB-induced oligodendroglioma development and malignancy, respectively. The proportion of grade in vitro experiments on primary brain cells infected with III tumors in Ntv-a mice showed a rising trend with loss either RCAS-PDGFB-HA or RCAS-X were performed. of Arf and Ink4a-Arf. There was a significant difference Expression of PDGFB was confirmed with Western blot in tumor malignancy when comparing Ntv-a wild-type for HA on cell lysates derived from the infected cells of mice with Ntv-a Ink4a-ArfÀ/À mice (Fischer’s exact different genetic backgrounds (Figure 3a). Activation of test, P ¼ 0.0292 (*)), and even more convincing so when the MAPK and the PI3K pathways were studied since comparing Ntv-a Ink4aÀ/À mice and Ntv-a ArfÀ/À they have been shown to be activated in human glioma. mice (Fischers’s exact test, P ¼ 0.0006 (***)) and Ntv-a There was a clear activation of Erk in all cells upon Ink4aÀ/À mice and Ntv-a Ink4a-ArfÀ/À mice (Fischer’s chronic PDGFB stimulation (Figure 3b). Activation of

Oncogene Role of Ink4a and Arf in PDGF-induced gliomagenesis E Tchougounova et al 6292

Figure 2 Kaplan–Meier graphs of glioma-free survival in PDGFB-injected Gtv-a mice (a) and Ntv-a mice (b). Statistical analyses using the log-rank test was performed on all possible comparisons and showed significant differences in survival between Ntv-a wild-type and Ntv-a Ink4aÀ/À, P ¼ 0.0236 (*); Ntv-a wild- type and Ntv-a Ink4a-ArfÀ/À, P ¼ 0.0003 (***); Ntv-a Ink4aÀ/À and Ntv-a Ink4a-ArfÀ/À, P ¼ 0.0467 (*); Ntv-a ArfÀ/À and Ntv-a Ink4a-ArfÀ/À, P ¼ 0.0348 (*); Gtv-a wild-type and Gtv-a Ink4aÀ/À, Po0.0001 (***); Gtv-a wild-type and Gtv-a ArfÀ/À, Po0.0001 (***); Gtv-a wild-type and Gtv-a Ink4a-ArfÀ/À, Po0.0001 (***); Gtv-a Ink4aÀ/À and Gtv-a ArfÀ/À, P ¼ 0.0017 (**); Ntv-a wild- type and Gtv-a wild-type, P ¼ 0.0041 (**); Ntv-a ArfÀ/À and Gtv-a ArfÀ/À, P ¼ 0.0116 (*).

Akt could not be detected in any cells even after long exposures (Figure 3b). In addition, elevated total levels of Erk and Akt expression were found in ArfÀ/À and Ink4a-ArfÀ/À cells, particularly upon PDGFB stimulation. These results suggested a central role for Figure 3 Western blot analysis of possible downstream activa- the MAPK pathway during PDGFB-induced tumor tors of chronic PDGF signaling in primary Ntv-a cell cultures initiation. from wild-type, Ink4aÀ/À, ArfÀ/À and Ink4a-ArfÀ/À mice. The To analyse possible downstream targets of the PI3K primary cells were continuously infected for 7 days with either RCAS-PDGFB-HA (B) or control RCAS-X (X). Actin was used as pathway, the activation of the serine/threonine kinase a loading control. (a) Expression of PDGFB protein by RCAS- p70S6K was analysed. It has previously been shown that PDGFB-HA infection (HA), cyclin D1 (cycD1) and p16Ink4a.(b) PDGFB stimulation of human glioma cells can induce Detection of total (tErk, tAkt) and activated (pErk, pAkt) levels of activation of p70S6K (Van Brocklyn et al., 1997). We Erk and Akt proteins. As a positive control for pAkt, protein found that chronic PDGFB stimulation in combination extract from DF-1 chicken fibroblasts producing RCAS-AKT was used (not shown). (c) Activated (pp70S6K) and total (p70S6K) with Arf-loss (in ArfÀ/À and Ink4a-ArfÀ/À cells) levels of p70S6K. Protein extract from NIH3T3 cells was used as a showed a high activation of p70S6K as compared to positive control for p70S6K (not shown). (d) Detection of cyclin D1 wild-type and Ink4aÀ/À cells (Figure 3c). Also, the total protein and activated and total forms of p70S6K protein in level of p70S6K was pronouncedly higher in these cells. PDGFB-infected cells treated with the PI3K inhibitor LY294002 or the MEK inhibitor U0126. The quantification of fold levels of To investigate if the PI3K pathway and/or the MAPK protein was analysed using Image Gauge v3.45. pathway were involved in the activation of p70S6K, we

Oncogene Role of Ink4a and Arf in PDGF-induced gliomagenesis E Tchougounova et al 6293 treated the cells with the PI3K inhibitor LY294002 or Discussion the mitogen-induced extracellular kinase (MEK) inhi- bitor U0126. The PI3K inhibitor was shown to inhibit There were significant and cell type-specific conse- specifically the activation of p70S6K whereas the MEK quences of loss of Ink4a and Arf in PDGFB-induced inhibitor had no effect on its activation (Figure 2d). oligodendroglioma formation in Ntv-a and Gtv-a mice. Hence, the PDGFB-induced activation of p70S6K was The main findings were that (i) in Gtv-a mice (transfor- conveyed via PI3K in an Akt-independent manner. mation of astrocytes) loss of Ink4a had a major impact Since activation of p70S6K previously has been on tumor initiation and caused an increased tumor shown to lead to an upregulation of cyclin D1 (Grewe incidence similar to that of loss of Arf or Ink4a-Arf,and et al., 1999), we investigated the status of cyclin D1 (ii) loss of Arf in both Ntv-a and Gtv-a mice caused in our cells. We found that the expression of cyclin increased malignancy of tumors which was not further D1 was induced in PDGFB-stimulated cells and that enhanced by the combined loss of Ink4a-ArfÀ/À. The the induction was higher in ArfÀ/À and Ink4a-ArfÀ/À clear effect of Ink4a-loss on the tumor incidence in Gtv-a cells (Figure 3a). To investigate if there was a direct mice but not in Ntv-a mice implies a differential role for functional relationship between the activation of Ink4a in astrocytes versus glial progenitor cells, keeping p70S6K and the level of cyclin D1, chronically astrocytes in a non-permissive differentiation state for PDGFB-infected cells were treated with LY294002 or PDGFB-induced transformation. The fact that all U0126 (Figure 3d). The result showed no such direct tumors induced from astrocytes were positive for the regulation of cyclin D1 levels by p70S6K. The PDGFB- neural stem cell marker Sox2 strongly indicated that a induced expression of cyclin D1 was clearly independent dedifferentiation process had occurred in these tumors. of the activation of the PI3K pathway. This implies The main consequence of Arf-loss was to contribute to that one mechanism by which loss of Arf contributes tumor progression in both Ntv-a and Gtv-a mice and in to tumor progression in our in vivo mouse model is addition shorter survival of tumor-bearing Gtv-a mice. via PDGFB-dependent expression and activation of The in vitro experiments suggested a mechanism where p70S6K and expression of cyclin D1. the PDGFB-induced activation of the MAPK pathway The observation that ArfÀ/À cells but not Ink4a- leading to a slight increase in proliferation was sufficient ArfÀ/À cells showed higher basal levels of cyclin D1 and for the initiation of low-grade oligodendrogliomas. The p70S6K led us to investigate the expression of p16Ink4a in additional loss of Arf contributed to tumor progression these cells. We found that the level of p16Ink4a was due to increased levels of cyclin D1 and a PI3K-dependent increased in the Arf-null cells compared to wild-type but apparently Akt-independent activation of p70S6K cells (Figure 3a). This was in accordance with a previous causing a strong proliferative response of the cells. report showing that Arf-null cells retained expression The intriguing finding that the basal levels of both of p16Ink4a, which was increased during serial passage p70S6K and cyclin D1 were considerably higher in the of the cells (Serrano et al., 1996; Kamijo et al., 1997a). A ArfÀ/À cells compared to wild-type, Ink4aÀ/À and plausible explanation of the increased levels of cyclin D1 Ink4a-ArfÀ/À cells directed us investigate the status of and p70S6K found in our ArfÀ/À cells could be that p16Ink4a. The rationale was that in normal mouse cells they are the consequence of a compensatory mechanism the levels of p16Ink4a and p19Arf are increased during required to override the growth inhibitory signal caused serial passage eventually leading to replicative senes- by the high expression of p16Ink4a. We also found cence (Zindy et al., 1997). Cells deficient for Arf or that p16Ink4a was induced upon PDGF stimulation, an Ink4a-Arf may escape senescence (Serrano et al., 1996; observation in line with a previous report coupling Kamijo et al., 1997a) albeit maintaining a high expres- PDGF stimulation with nuclear localization of b- sion of p16Ink4a in the Arf-null cells. We reasoned that the catenin and increased expression of p16Ink4a (Fischer increased levels of cyclin D1 and p70S6K found in our et al., 2006). ArfÀ/À cells could be the consequence of a compensa- Primary cells from Ntv-a wild type, Ntv-a Ink4aÀ/À, tory mechanism required to counteract the effect of the Ntv-a ArfÀ/À and Ntv-a Ink4a-ArfÀ/À mice were p16Ink4a expression. However, the influence of modifier investigated for their proliferative response to PDGFB genes cannot be entirely excluded. The ArfÀ/À line was stimulation (Figure 4a and b). All cells showed increased made in WW6 ES cells which are 75% 129/Sv, 20% proliferation in response to PDGFB stimulation but the C57Bl6 and 5%SJL(Ioffe et al., 1995) and the Ink4a- combinations of PDGFB stimulation and loss of ArfÀ/À line was made in RW4 ES cells which are 100% Arf or Ink4a-Arf resulted in the highest proliferative 129/Sv (Schumacher et al., 2003). Yet, we find it unlikely response. Ntv-a wild type and Ntv-a Ink4aÀ/À cells were that the minor difference in genetic background would much less responsive to PDGFB stimulation and also be the sole cause of such a prominent cellular effect. had significantly lower basal proliferation rate. This For the development of successful targeted therapeu- was in agreement with the in vivo situation where tic treatment of human glioma, it is essential to know immunohistochemical staining of tumors for Ki-67 how different combinations of mutations collaborate to clearly showed more positive cells in Ntv-a ArfÀ/À activate downstream oncogenic signaling pathways. (Figure 4c and d) and Ink4a-ArfÀ/À mice (data not High-grade astrocytomas, both from patients and shown and Figure 4d) than in wild-type (Figure 4c from several different mouse models, have displayed and d) and Ink4aÀ/À mice (data not shown and activation of the PI3K pathway via Akt as a seemingly Figure 4d). obligatory event (Holland et al., 2000; Bachoo

Oncogene Role of Ink4a and Arf in PDGF-induced gliomagenesis E Tchougounova et al 6294 ab221 Ntv-a wild type; RCAS-X 40 Ntv-a wild type; RCAS-PDGFB-HA RCAS-X Ntv-a Ink4a–/–; RCAS-X RCAS-PDGFB-HA 20 Ntv-a Ink4a–/–; RCAS-PDGFB-HA 2 Ntv-a Arf–/–; RCAS-X Ntv-a Arf–/–; RCAS-PDGFB-HA 19 Ntv-a Ink4a-Arf–/–; RCAS-X 30 2 Ntv-a Ink4a-Arf–/–; RCAS-PDGFB-HA

218

217 20

16

2 Fold increase Number of cells

215 10

214

213 0 012345678 WTInk4a –/– Arf–/– Ink4a- Day Arf–/– c

d 40 35 30 25 20 15 10 positive tumor cells Percentage of Ki-67 5 0 WT Ink4a –/– Arf–/– Ink4a- Arf–/–

Figure 4 Proliferation assay on primary Ntv-a cell cultures from wild-type, Ink4aÀ/À, ArfÀ/À and Ink4a-ArfÀ/À mice. The primary cells were continuously infected for 7 days with either RCAS-PDGFB-HA or control RCAS-X. (a) Each time point is calculated from three independent experiments using duplicate samples. Bars show s.e. (b) Fold increase in cell number at day 7 normalized to Ntv-a wild-type cells infected with RCAS-X. (c) Immunohistochemical detection of Ki-67 in two representative tumors from Ntv-a wild-type and Ntv-a ArfÀ/À mice. Scale bar, 50 mm. (d) Percentage of Ki-67-positive tumor cells from Ntv-a wild-type, Ntv-a Ink4aÀ/À, Ntv-a ArfÀ/À and Ntv-a Ink4a-ArfÀ/À mice. Each result shows the mean from two different tumors and error bars show s.d.

et al., 2002; Charest et al., 2006). Here, we have shown et al., 1996) were crossed with the Ntv-a and Gtv-a transgenic that also for the development of high-grade oligoden- mouse lines. Heterozygous mice were subsequently inter- drogliomas the PI3K pathway is activated, although in an crossed and all injections were made in the F4 or later apparently Akt-independent manner. Future efforts will be generation of mice. The Ink4aÀ/À line was made in WW6 ES focused on further dissecting the signaling events in cells, injected into C57Bl6 blastocysts and chimeras were mated with C57Bl6 mice. The ArfÀ/À line was made in WW6 oligodendroglioma development in order to try to identify ES cells, injected into C57Bl6 blastocysts and chimeras were unique targets for this tumor type. mated with C57Bl6 mice. The Ink4a-ArfÀ/À line was made in RW4 ES cells, injected into C57Bl6 blastocysts and chimeras Materials and methods were mated with C57Bl6 mice. Neonatal mice were injected in the right cerebral hemisphere Generation of mice, in vivo virus infection and tumor with DF-1 chicken fibroblasts producing RCAS-PDGF-EGFP surveillance and monitored as described (Uhrbom et al., 2002). Experi- The ArfÀ/À mice (Kamijo et al., 1997b), the Ink4aÀ/À mice ments were performed in accordance with the local animal (Sharpless et al., 2001) and the Ink4a-ArfÀ/À mice (Serrano ethics committee.

Oncogene Role of Ink4a and Arf in PDGF-induced gliomagenesis E Tchougounova et al 6295 Primary tissue culture and proliferation assay (DakoCytomation, Glostrup, Denmark) and rabbit polyclonal Primary brain cell cultures were prepared from neonatal Ntv-a anti-Sox2 (Chemicon). wild-type, Ntv-a Ink4aÀ/À, Ntv-a ArfÀ/À and Ntv-a Ink4a- The quantification of Ki-67-positive tumor cells was deter- Arf À/À mice as described (Uhrbom et al., 2005). Cells were mined from two representative tumors each from Ntv-a wild- expanded for two passages and stored as frozen aliquots. To type, Ntv-a Ink4aÀ/À, Ntv-a ArfÀ/À and Ntv-a Ink4a-ArfÀ/À infect the primary Ntv-a cells, supernatants from DF-1 chicken mice. The percentage of Ki-67-positive nuclei out of 500 fibroblast cells producing either RCAS-PDGFB-HA or RCAS- randomly selected cells was calculated. X (control virus) were used. Conditioned media from the respective retroviral producing cells was collected after 24 h, Protein extraction and Western blot analyses sterile-filtered through 0.45 mm filters and added to the primary Infected cells were washed twice with and incubated in serum- Ntv-a cell cultures. The infection was repeated every day for free Dulbecco’s modified Eagle’s medium for 19 h. The small 1week. molecule inhibitors LY 294002 (10 mM in dimethyl sulfoxide For the proliferation assay, 2 Â 104 infected cells per 35-mm (DMSO); Promega, Madison, WI, USA) and U0126 (10 mM in dish were plated, duplicate samples for each time point. The DMSO; Promega) or only DMSO were added 4 h before cell number per dish was determined at days 1, 4 and 7 using a collection of cells. Protein extracts and Western blots were Coulter counter (Coulter Electronics, Bromma, Sweden). The performed as described previously (Uhrbom et al., 2005). proliferation assay was repeated three times. Membranes were immunoblotted at 41C overnight with the primary antibodies p44/42 MAPK, phospho-p44/42 MAPK (Cell Signaling, Beverly, MA, USA), phospho-Akt (Thr308) Histopathology and statistical analyses of tumor numbers (Cell Signaling), Akt (Cell Signaling), cyclin D1 (H-295; Santa Mouse brains were fixed in formalin, cut into five pieces, Cruz Biotechnology, Santa Cruz, CA, USA), HA-probe (Y-11) embedded in paraffin, sectioned and analysed for presence or (Santa Cruz), p16Ink4a (M-156; Santa Cruz), phospho-p70S6 absence of tumor tissue by viewing hematoxylin and eosin kinase (Thr 421/Ser 424)-R (Santa Cruz), p70S6 kinase (C-18; (H&E)-stained sections. Statistical analyses were performed Santa Cruz), actin (Chemicon, Temecula, CA, USA). Horse- in the GraphPad Software Prism 4.0a using the log-rank radish peroxidase-coupled secondary antibodies (Amersham test applied to Kaplan–Meier graphs, and the Fischer’s exact Biosciences, Little Chalfont, Buckinghamshire, UK) were test for the comparison of incidence rates and malignancy used and the reaction was visualized with SuperSignal West Pico grades. Chemiluminescent Substrate (Pierce, Cramlington, Northum- berland, UK) on Hyperfilm-ECL(Amersham, Amersham Immunohistochemical analyses and quantification of Biosciences). Ki-67-positive cells Immunostainings were performed as described previously Acknowledgements (Uhrbom et al., 2005). Antibodies used were monoclonal anti-GFAP (Chemicon, Hampshire, UK), monoclonal anti- This work has been supported by grants from the Swedish nestin (Pharmingen, BD Biosciences, Stockholm, Sweden), Cancer Society, The Swedish Research Council, A˚ ke Wiberg’s rabbit polyclonal anti-NG2 chondroitin sulfate proteoglycan Foundation, Magn. Bergvall’s Foundation and Jeansson’s (Chemicon, Temecula, CA, USA), rat monoclonal anti-Ki-67 Foundation.

References

Bachoo RM, Maher EA, Ligon KL, Sharpless NE, Chan SS, expression in human oligodendrogliomas. Neurosurgery 42: You MJ et al. (2002). receptor and 341–346. Ink4a/Arf: convergent mechanisms governing terminal Fischer AN, Fuchs E, Mikula M, Huber H, Beug H, Mikulits differentiation and transformation along the neural stem W. (2006). PDGF essentially links TGF-beta signaling to cell to astrocyte axis. Cancer Cell 1: 269–277. nuclear beta-catenin accumulation in hepatocellular carci- Bigner SH, Rasheed BK, Wiltshire R, McLendon RE. (1999). noma progression. Oncogene. Morphologic and molecular genetic aspects of oligodendro- Grewe M, Gansauge F, Schmid RM, Adler G, Seufferlein T. glial neoplasms. Neuro-oncol 1: 52–60. (1999). Regulation of cell growth and cyclin D1 expres- Cairncross JG, Ueki K, Zlatescu MC, Lisle DK, Finkelstein sion by the constitutively active FRAP-p70s6K pathway in DM, Hammond RR et al. (1998). Specific genetic predictors human pancreatic cancer cells. Cancer Res 59: 3581–3587. of chemotherapeutic response and survival in patients Holland EC, Celestino J, Dai C, Schaefer L, Sawaya RE, with anaplastic oligodendrogliomas. J Natl Cancer Inst 90: Fuller GN. (2000). Combined activation of Ras and Akt in 1473–1479. neural progenitors induces glioblastoma formation in mice. Charest A, Wilker EW, McLaughlin ME, Lane K, Gowda R, Nat Genet 25: 55–57. Coven S et al. (2006). ROS fusion activates a Ioffe E, Liu Y, Bhaumik M, Poirier F, Factor SM, Stanley P. SH2 domain-containing phosphatase-2/phosphatidylinositol (1995). WW6: an embryonic stem cell line with an inert 3-kinase/mammalian target of rapamycin signaling axis to genetic marker that can be traced in chimeras. Proc Natl form glioblastoma in mice. Cancer Res 66: 7473–7481. Acad Sci USA 92: 7357–7361. Dai C, Celestino JC, Okada Y, Louis DN, Fuller GN, Holland Jen J, Harper JW, Bigner SH, Bigner DD, Papadopoulos N, EC. (2001). PDGF autocrine stimulation dedifferentiates Markowitz S et al. (1994). Deletion of p16 and p15 genes in cultured astrocytes and induces oligodendrogliomas and brain tumors. Cancer Res 54: 6353–6358. oligoastrocytomas from neural progenitors and astrocytes Kamijo T, Zindy F, Roussel MF, Quelle DE, Downing JR, in vivo. Genes Dev 15: 1913–1925. Ashmun RA et al. (1997a). Tumor suppression at the mouse Di Rocco F, Carroll RS, Zhang J, Black PM. (1998). INK4a locus mediated by the alternative reading frame Platelet-derived growth factor and its receptor product p19ARF. Cell 91: 649–659.

Oncogene Role of Ink4a and Arf in PDGF-induced gliomagenesis E Tchougounova et al 6296 Kamijo T, Zindy F, Roussel MF, Quelle DE, Downing JR, Sharpless NE, Bardeesy N, Lee KH, Carrasco D, Castrillon Ashmun RA et al. (1997b). Tumor suppression at the mouse DH, Aguirre AJ et al. (2001). Loss of p16Ink4a with INK4a locus mediated by the alternative reading frame retention of p19Arf predisposes mice to tumorigenesis. product p19ARF. Cell 91: 649–659. Nature 413: 86–91. Kleihues P, Cavenee WC (eds). Pathology and Genetics. Sharpless NE, Ramsey MR, Balasubramanian P, Castrillon Tumours of the nervous system, International Agency for DH, DePinho RA. (2004). The differential impact of Research on Cancer: Lyon, 1997. p16(INK4a) or p19(ARF) deficiency on cell growth and Reifenberger J, Reifenberger G, Ichimura K, Schmidt EE, tumorigenesis. Oncogene 23: 379–385. Wechsler W, Collins VP. (1996). Epidermal growth factor Uhrbom L, Dai C, Celestino JC, Rosenblum MK, Fuller GN, receptor expression in oligodendroglial tumors. Am J Pathol Holland EC. (2002). Ink4a-Arf loss cooperates with KRas 149: 29–35. activation in astrocytes and neural progenitors to generate Scherer H. (1938). Structural development in gliomas. Am J glioblastomas of various morphologies depending on acti- Cancer 34: 333–351. vated Akt. Cancer Res 62: 5551–5558. Schmidt EE, Ichimura K, Reifenberger G, Collins VP. (1994). Uhrbom L, Kastemar M, Johansson FK, Westermark B, CDKN2 (p16/MTS1) gene deletion or CDK4 amplification Holland EC. (2005). Cell type-specific tumor suppression by occurs in the majority of glioblastomas. Cancer Res 54: Ink4a and Arf in Kras-induced mouse gliomagenesis. Cancer 6321–6324. Res 65: 2065–2069. Schumacher A, Arnhold S, Addicks K, Doerfler W. (2003). Van Brocklyn JR, Vandenheede JR, Fertel R, Yates AJ, Staurosporine is a potent activator of neuronal, glial, Rampersaud AA. (1997). Ganglioside GM1 activates the and ‘CNS stem cell-like’ neurosphere differentiation in mitogen-activated protein kinase Erk2 and p70 S6 kinase in murine embryonic stem cells. Mol Cell Neurosci 23: U-1242 MG human glioma cells. J Neurochem 69: 116–125. 669–680. Zindy F, Quelle DE, Roussel MF, Sherr CJ. (1997). Expres- Serrano M, Lee H, Chin L, Cordon-Cardo C, Beach D, sion of the p16INK4a tumor suppressor versus other INK4 DePinho RA. (1996). Role of the INK4a locus in tumor family members during mouse development and aging. suppression and cell mortality. Cell 85: 27–37. Oncogene 15: 203–211.

Oncogene