Genetic Determinants of Malignancy in a Mouse Model for Oligodendroglioma¹

Gliomas are the leading cause of cancer death in children and the most common primary brain tumor in adults (1, 2). Oligodendroglioma constitute ∼20% of glial tumors (2); however, the cell of origin for oligodendroglioma is unknown. Although the morphology of tumor cells resembles that of normal oligodendrocytes, several markers known to be present in oligodendrocytes are absent in oligodendroglial tumors (3). Recent evidence suggests that oligodendroglioma does arise from oligodendrocytes; however, as basic helix-loop-helix transcription factors, Olig-1 and Olig-2 are expressed throughout oligodendroglial development and are present in tumors (4, 5, 6).

Oligodendrogliomas are characterized by: (a) dysregulation of the EGFR³ growth stimulatory pathway; (b) deletion of the tumor suppressors ink4a/arf and PTEN; and (c) mutation in P53 (7, 8, 9). EGFR is frequently expressed at high levels in both low- and high-grade oligodendrogliomas (7); whereas overexpression of EGFR in astrocytomas is associated with EGFR amplification and occurs only in high-grade tumors (10). Over half of oligodendrogliomas show combined loss of the short arm of chromosome 1 and long arm of chromosome 19 (11, 12). Combined 1p/19q loss appears to be inversely related to p53 mutation, consistent with distinct molecular subtypes (13, 14). The subgroup of oligodendroglioma in which chromosome 1p is deleted is sensitive to cytotoxic agents and irradiation and correlates with improved survival (15, 16). However, despite the fact that oligodendrogliomas show an improved survival rate when compared with astrocytomas, most patients with oligodendrogliomas die of their disease.

A major problem with the most common gliomas relates to their infiltrative nature. Even low-grade astrocytoma and oligodendrogliomas are highly infiltrative of surrounding brain tissue, precluding complete surgical resection. The inability to completely resect these tumors is responsible for the nearly inevitable local recurrence experienced by most patients with infiltrating gliomas. A model which recapitulates the infiltrative nature of these tumors would be very useful for biological and preclinical studies to improve therapies for this disease.

To explore the role of EGFR in the pathogenesis of oligodendroglioma, we created transgenic mice that expressed v-erbB, a transforming homologue of EGFR, under the control of the S100β promoter. S100β is expressed in both oligodendroglia and astrocytes early in brain development and is detectable in a wide variety of glial tumors (17, 18, 19). Because S100β is expressed in central nervous system stem cells (20, 21), we hypothesized that the S100β-v-erbB transgene would recapitulate high-level expression of EGFR during the pathogenesis of glioma.

A 9-kb EcoRI fragment containing the S100β gene (22) was cloned into the NotI site of pGEM13Zf+ (Promega, Madison, WI). A 1.8-kb XhoI-HindIII fragment containing v-erbB ES4 was ligated to a 2.1-kb fragment containing the intron and poly(A) site form of SV40. The resulting 3.9-kb fragment containing v-erb-intron-poly(A) was cloned into an AccI site immediately 5′ to the transcription initiation site of S100β (22). The resulting 13-kb insert was cleaved from vector DNA using NotI and used to generate transgenic mice (23). Founders were derived from C57B6/JX DBA/2 (The Jackson Laboratory, Bar Harbor, ME) and were subsequently back-crossed into FVB/N. Mice deleted for p53 were in strain FVB/N. Mice deleted for Rb1 and ink4a/arf were outbred. Tail DNA was analyzed by Southern analysis using a v-erbB probe or by PCR using primers 5′-CTCACAGCAATCTCAAAGCTCCCC-3′ and 3′ primer 5′-AGCCTCCAAAGTCAGGTTGATGAGC-3′.

Brain, tumor, and grossly abnormal organs were submitted for pathology. The entire brain was cut in 8-μm sections, and every 10th section stained was analyzed for tumor. For immunohistochemistry, 8-μm sections were deparaffinized, hydrated, and treated with the appropriate mouse monoclonal antibodies directed against GFAP (Biogene, San Ramon, CA), v-erbB (24), S100β (Ventana, Inc., Tucson, AZ), or NeuN (Chemicon, Inc., Temcula, CA). Human oligodendrogliomas were from the University of California at San Francisco Brain Tumor Research Center. Tissues were lysed, and 5 μg of protein supernatant were loaded into 10% SDS-PAGE gels. Membranes were incubated with antisera directed against v-erbB, pyruvate kinase, or 4G10 (Upstate Biotechnology, Waltham, MA).

The 1.8-kb XhoI-HindIII fragment of v-erbB was cloned in both orientations into pTopo 2.1 (Invitrogen). The two resulting v-erbB pTopo 2.1 plasmids were linearized with BamHI, which cuts out a fragment within v-erbB. The resulting sense and antisense probes were generated using a Ribo MAX kit and digoxigenin-labeling ribonucleoside triphosphate mix (Roche, Indianapolis, IN). Frozen sections (10 μm) were hybridized overnight at 65°C and visualized with an alkaline phosphatase-conjugated antidigoxigenin antibody (Roche).

Fresh primary tumors were incubated in a solution of DNAase1 (0.2 mg/ml in Ca²⁺- and Mg²⁺-free HBSS media), Pronase (0.5 mglml), and Collagenase (0.2 mg/ml). Tumors were incubated at 37° for 45 min with stirring, filtered through a 100-μm sterile screen, centrifuged, and plated in DMEM-H with 20% fetal bovine serum in a 37°C CO₂ incubator.

Genomic DNA or plasmid DNA from the construct in Fig. 1 A was digested with BamHI, NotI, or KpnI, respectively. Samples (5 μg) representing 1.5 × 10⁶ copies of mouse genome were resolved, along with dilutions of the v-erbB gene representing 0.5, 2, 8, 32, and 128 × 10⁶ copies. DNA was transferred to membranes and probed with 20 × 10⁶ cpm of a v-erbB probe in ExpressHyb (Clonetech, Palo Alto, CA).

Tumor tissue was scraped from unstained slides and incubated twice in xylene and twice in ethanol at 55°C. Samples were dissolved in PCR buffer containing 0.5% Tween and Proteinase K (0.4 mg/ml) and incubated at 55°C for 3 days.

Mice were anesthetized with Avertin (23) and imaged using a 2.0 Tesla magnet interfaced with a Bruker Omega console (Bruker Instruments, Inc., Fremont, CA). Serial transverse sections (2 mm) were acquired using a T2-weighted sequence (repetition time/echo time: 2500/80 ms), followed by a T1-weighted precontrast sequence (repetition time/echo time: 500/12 ms). Gadolinium pentate (100 μl of 50 μm stock) was then delivered by tail vein injection.

CGH was performed and analyzed as described previously (25). Each tumor was analyzed twice, inverting the red FITC-NEN and green Alexa 568 fluors for the separate analyses.

We cloned v-erbB adjacent to the transcriptional start site in the murine S100β promotor (Fig. 1,A). Three of 15 founders developed tumors. Transgene dosage had no obvious impact on penetrance, varying from two to eight copies in tumor-prone and one to eight copies in tumor-free lines of mice (Fig. 2, A and B). Each of three lines predisposed to tumors had a similar penetrance. All tumors examined in animals from two lines had easily detectable levels of V-erbB, although the level of expression did not correlate with pathologic grade (Fig. 2,C). V-erbB is transforming, whereas EGFR transforms only in the presence of ligand. To compare the phosphotyrosine pattern in v-erbB-driven tumors with that in human oligodendroglioma, we analyzed five human and four murine oligodendroglioma by Western analysis using a phosphotyrosine antibody (Fig. 2 D). Levels of phosphotyrosine in murine tumors were less variable than in human tumors and were either similar to or modestly exceeded the levels observed in human tumors. The phospho-proteins represented in this blot are unknown; however, the mobility of the highest molecular weight band present in murine tumors is consistent with that of V-erbB.

We performed in situ hybridization of histological brain sections from p60 (60 days old) transgenic mice using a v-erbB probe. High levels of expression were seen diffusely throughout the cortex and subcortical white matter (Fig. 3, A and B), within the granule and Purkinjie layers of the cerebellum (Fig. 3, C and D), and in the subventricular zone (Fig. 3, E and F). We did not observe transgene expression outside of the central nervous system (data not shown). S100β was detected in multiple brain regions at birth. Expression increased during the first few weeks of life, with regional expression patterns comparable with those shown in Fig. 3 (M. J. B. and W. A. W., data not shown).

The appearance of tumors in these animals was in each case pathognomonic for human oligodendroglioma (26): (a) monomorphic fields of cells; (b) round homogeneous nuclei without indentations; and (c) swollen clear cytoplasm forming a prominent perinuclear halo (Fig. 1). Infiltration of human gliomas is a key feature that contributes to their poor therapeutic response, and the failure of xenografted tumors to recapitulate this characteristic has limited the utility of such models (2, 26). Remarkably, murine tumors were infiltrative in a manner analogous to human oligodendroglioma, evidenced by their easily recognized nuclei invading the surrounding brain parenchyma (Fig. 1 E). Tumors stained for S100β (data not shown) but were negative for both the astrocytic marker GFAP and neuronal marker NeuN (data not shown).

Other typical features of human oligodendroglioma were also observed in murine tumors. These included preferential growth along white matter tracts, a pathological feature of both clinical and scientific significance in human gliomas (Fig. 4,A). The dense network of branching capillaries characteristic of human oligodendroglioma (26) was readily observed (Fig. 4,B). Tumors were always low grade, as evidenced by uniform appearance of nuclei, infrequent mitotic figures, modest cellularity, and absence of necrosis or endothelial proliferation. Occasional tumors showed rhythmic pallisading of nuclei typical of human oligodendroglioma (Fig. 4,C). Electron microscopic analysis was consistent with oligodendroglioma (Fig. 4,D) in that neither neuronal nor ependymal features were observed (27). Like their human counterparts, low-grade murine oligodendrogliomas demonstrated an intact blood–brain barrier by MRI (28, 29), as evidenced by their failure to show enhancement after injection of i.v. contrast (Fig. 4, E and F).

To assess the penetrance of oligodendroglioma, 38 transgenic animals from a single line were observed for 1 year. All symptomatic animals were autopsied. Half of this cohort died of glioma or was euthanized (and autopsied) for signs of tumor by 6 months (Fig. 5 A). After 1 year, 24 deaths from oligodendroglioma were documented. In 4 mice, the cause of death was uncertain. In 2 mice without oligodendroglioma, euthanasia was secondary to either an abdominal mass or a soft tissue leg tumor.

In humans, deletion of ink4a/arf is found in high-grade anaplastic oligodendroglioma (12, 30) and is associated with a poor prognosis (16). Nearly half of the gliomas arising in transgenic animals carrying mutations in ink4a/arf showed features of high-grade anaplastic oligodendroglioma (26), including increased cellularity, endothelial proliferation, necrosis, and increased nuclear atypia, although clear evidence of oligodendroglial origin remained (Fig. 6, A–C). MRI analysis from high-grade tumors revealed a disrupted blood–brain barrier, as evidenced by marked extravasation of contrast material in tumor tissue. (Fig. 6, D–F). In comparison, images of low-grade tumors showed no contrast enhancement (Fig. 4, E and F). High-grade tumors frequently showed ring enhancement (Fig. 6 E), an imaging feature illustrative of central tumor necrosis and associated with poor clinical outcome (16).

Tumors developed with decreased latency and increased penetrance in transgenic mice homozygous for loss of ink4a/arf, in comparison with transgenic littermates wild type at the ink4a/arf locus. Three-quarters of these animals died of glioma by 2 months, and >90% succumbed to oligodendroglioma by 6 months (Fig. 5,B). Of 45 transgenic, ink/arf+/− mice followed for 1 year, ∼90% died of oligodendroglioma (Fig. 5 B).

Ink4a and arf are distinct tumor suppressors that share a common exon but encode nonhomologous proteins (31). Loss of ink4a affects Rb signaling, whereas loss of arf affects p53 signaling. Oligodendrogliomas in S100β-v-erbB transgenic mice heterozygous for p53 were typically of high grade and had latency and penetrance comparable with that observed in transgenic mice heterozygous for ink4a/arf (Fig. 5 B). Of 52 transgenic, p53+/− mice followed for 6 months, 80% died of oligodendroglioma. Mutation in p53 is relatively uncommon in human oligodendroglioma, occurring in ∼10–15% of tumors (13, 14).

The pattern of tumor development and pathology of tumors in transgenic animals heterozygous for Rb1 were indistinguishable from animals wild type at Rb1 (Fig. 5 B and data not shown).

Transgenic animals heterozygous for either p53 or ink4a/arf had longer latency but similar tumor incidence compared with transgenic animals homozygous for deletion of ink4a/arf (Fig. 5,B). To examine whether animals heterozygous for p53 or ink4a/arf lost the remaining wild-type allele, we examined DNA from normal tissue, primary tumors, and tumor-derived cell lines arising in transgenic animals heterozygous for p53 or ink4a/arf. Tumors and tumor-derived cell lines gave identical results, showing loss of the functional p53 or ink4a/arf allele. Both wild-type and mutant alleles were detectable in somatic tissues as expected in these heterozygous animals (Fig. 7).

To identify spontaneous genetic aberrations that contributed to the development of oligodendroglioma, we used CGH to analyze 22 tumors and tumor-derived cell lines from transgenic animals, transgenic animals deleted for p53, and transgenic animals deleted for ink4a/arf. Nine samples showed gain for chromosome 15, and five showed gain of proximal chromosome 10 (Fig. 8). Nine samples showed loss of chromosome 4. In four cases, the loss of chromosome 4 occurred distal to band C3–C6 that contains ink4a/arf (Mouse Genome Database 9/02).⁴ The distal part of chromosome 4 is orthologous to human 1p36, which is commonly lost in oligodendroglioma in humans (11, 32, 33, 34). Changes on chromosomes 4 and 15 were found in primary tumors, as well as tumor-derived cell lines, whereas gains on chromosome 10 were restricted to cell lines.

To evaluate the significance of overexpression of EGFR in human oligodendrogliomas, we expressed v-erbB using the S100β promoter. Animals transgenic for S100β-v-erbB developed oligodendroglioma with the pathognomonic histology and pathology observed in human oligodendroglioma. Tumors were highly infiltrative, invaded along white matter tracts, and developed branching thin-walled capillaries, reflecting histopathologic features characteristic of human oligodendroglioma. Our finding that aberrant EGFR signaling can cause oligodendroglioma in transgenic mice supports a causal role for overexpression of EGFR in human tumors.

Deletion of ink4a/arf is common in high-grade oligodendroglioma in humans and is associated with shortened survival (16). In mice predisposed to low-grade oligodendroglioma, deletion of ink4a/arf led to the development of tumors with shortened latency, increased penetrance, and increased pathologic grade. Loss of p53 phenocopied the loss of ink4a/arf, whereas the loss of Rb1 had little effect. The survival curve of transgenic mice heterozygous for p53 indicated that loss of p53 could enhance tumorigenesis in mice expressing v-erbB and that defective p53 signaling through arf accounted for the increased penetrance observed in transgenic mice deleted for ink4a/arf. Our data also suggest that alteration in the p53 pathway alone is sufficient to explain the transition from low to high grade. Importantly, nuclear atypia, mitotic figures, endothelial proliferation, and necrosis observed in high-grade tumors are mediated by molecular networks in which p53 plays a prominent role. Nuclear atypia reflects genomic changes and arises as the result of genomic instability, a well-known result of p53 loss. Loss of p53 also leads to increased proliferation as a result of enhanced cell division and decreased apoptosis (31, 35). Similarly, microvascular proliferation could occur from dysregulation of angiogenic factors normally inhibited by p53 (36, 37). The lack of a precise balance between cellular and microvascular proliferation is a likely cause of necrosis. The association between p53 mutation and grade in human tumors is uncertain.

Because S100β is expressed in both astrocytes and oligodendrocytes (17, 19, 38), tumors arising in both cell types might have been expected. Overexpression of EGFR is an early event in oligodendroglioma (7) but occurs late in the pathogenesis of astrocytoma (10, 26). It is therefore possible that the expression of v-erbB under control of S100β was insufficient to initiate transformation of cells in which astrocytomas arise, whereas oligodendroglia were more readily transformed by this molecule. We did see occasional astrocytic differentiation in tumors both based on histology and on GFAP immunoreactivity (data not shown), but this occurred in <10% of cases.

Recently, Dai et al. (39) described a model for oligodendroglioma based on intracerebral inoculation of a recombinant avian retrovirus carrying an allele of PDGFB. This virus was injected into the brains of transgenic mice in which expression of the viral receptor (tva) was controlled by promoters active in either astrocytes or astrocytic precursors. Dai et al. observed oligodendroglial tumors when targeting their transgene to GFAP-positive cells and concluded that tumorigenesis in their model could arise through dedifferentiation of committed astrocytes. Importantly, however, the GFAP-tva mice used for studies by Dai et al. were also used by others to show that neural stem cells arise from GFAP-positive progenitors (40). Therefore, it remains possible that oligodendroglioma in GFAP-tva mice could arise through primary transformation of a glial progenitor cell, rather than through dedifferentiation of a committed astrocyte.

The incomplete penetrance and variable latency of murine oligodendroglioma suggest that additional genetic events contributed to gliomagenesis. We found evidence that the genetic alterations found in human oligodendroglioma may also occur spontaneously in our model. CGH analysis of 22 tumors and tumor-derived cell lines revealed common regions of gain on chromosomes 10 and 15 and loss on chromosome 4. The gain on proximal 10 lies in a region of orthologous with human chromosome 6q21–25.⁵ Candidate genes mapping to this interval include oligodendrocyte transcription factor 3, the Myb proto-oncogene, and protein tyrosine phosphatase, receptor type, K (February 2003 freeze).⁶ The gain on chromosome 15 is orthologous with regions of human chromosomes 8q and 12q, which are gained in human gliomas (41, 42). Potential candidate genes mapping to this region include the Myc proto-oncogene, focal adhesion kinase Ptk2, and platelet-derived growth factor B polypeptide. The region of loss on chromosome 4 is orthologous with human chromosome 1p, which is commonly lost in human oligodendroglioma and defines a group of tumors that is relatively responsive to therapy (16). The middle of chromosome 4 also contains ink4a/arf; however, four tumors showed loss of the distal end of chromosome 4 (Fig. 8) in a region distal to ink4a/arf (Mouse Genome Database 9/02).⁴ For these three tumors, the region of loss lies in a region orthologous to that of human 1p36. Such findings raise the possibility of discovering novel genes related to tumor development using genomic strategies.

We thank Qi-Wen Fan, Susan Hill, Martin McMahon, Randy Schatzman, and Andreas Trumpp for plasmids and reagents. We also thank Martha Simmons for NeuN studies and Gregg Magrane, Robert Flandermeyer, and Fred Waldman for CGH analyses. Finally, we thank J. Michael Bishop, in whose lab this project was initiated.