Astroglial c-Myc overexpression predisposes mice to primary malignant gliomas.

Malignant astrocytomas are common human primary brain tumors that result from neoplastic transformation of astroglia or their progenitors. Here we show that deregulation of the c-Myc pathway in developing astroglia predisposes mice to malignant astrocytomas within 2-3 weeks of age. The genetically engineered murine (GEM) gliomas harbor a molecular signature resembling that of human primary glioblastoma multiforme, including up-regulation of epidermal growth factor receptor and Mdm2. The GEM gliomas seem to originate in an abnormal population of glial fibrillary acidic protein-expressing cells in the ventricular zone and, analogous to human glioblastomas, exhibit molecular and morphological heterogeneity. Levels of connexin 43 in the majority of the tumors are unaltered from normal tissue, indicating that GEM tumors have retained the capacity to establish syncytial networks. In line with this, individual glioma foci are composed of a mixture of actively proliferating cells expressing c-Myc and proliferating cell nuclear antigen and less dividing bystander cells that express glial fibrillary acidic protein and the broad complex tramtrack bric-a-brac/poxvirus and zinc finger domain protein HOF. A subset of the transgenic mice harbored, in addition to brain tumors, vestigial cerebellums in which granule cell migration and radial Bergman glial cell differentiation were disturbed. These observations argue for a window of vulnerability during astrocyte development where c-Myc overexpression is sufficient to trigger the neoplastic process, presumably by inducing the sustained growth of early astroglial cells. This is in contrast to most other transgenic studies in which c-Myc overexpression requires co-operating transgenes for rapid tumor induction.

Of the tumors that originate in the human brain, astrocytederived gliomas are the most common. They range from slow growing pilocytic astrocytomas to high grade variants such as anaplastic astrocytomas and glioblastoma multiforme (1,2). Astrocytomas are clinically troubling because they diffusely infiltrate the brain, impeding complete surgical resection of neoplastic cells, and because lower grade tumors often progress to higher grade variants. Because the latter respond poorly to radiotherapies and chemotherapies, malignant astrocytomas are currently untreatable. Glioblastoma multiforme rank among the most malignant brain lesions, with a median survival of less than 12 months.
High grade gliomas can be classified into primary and secondary glioblastomas (3). Primary glioblastomas appear mainly de novo and are often associated with deregulation of epidermal growth factor receptor (EGFR) 1 expression. Secondary glioblastomas, on the other hand, develop from the progression of lower grade tumors that often carry mutations in the p53 gene. In addition to EGFR and p53 mutations, a number of genetic alterations affecting interphasing biochemical pathways have been correlated with glioma pathogenesis. This involves gain-of-function and loss-of-function mutations in genes regulating p53-mediated apoptosis (p16 ARF and Mdm2), cell cycle arrest (pRb, cdk4, and p16 Ink4a ), cell proliferation, and survival (Ras, mitogen-activated protein kinase, phosphatidylinositol 3-kinase, and Akt) (4). c-Myc is a basic helix-loop-helix leucine zipper transcription factor that triggers cell proliferation and apoptosis, functions that seem to be effectuated mainly through activation of the cdk4/pRb/E2F pathway (5). c-Myc functions as a sequence-specific DNA-binding protein in association with its partner Max to facilitate expression of E-box-containing target genes (6). Moreover, c-Myc has recently been assigned roles in chromatin remodeling of target promoters, a process that involves recruitment via the N-terminal transactivation domain of co-activator complexes including histone acetylases, ATPase/helicase activities, as well as an ATP-dependent chromatin remodeling complex (7,8). In line with its involvement in a wide range of neoplasms, there is circumstantial evidence that suggests c-Myc also plays a role in gliomagenesis, including deregulation of the c-Myc pathway in human glioblastoma multiforme (9,10). Although the molecular mechanism(s) underlying this expression is poorly understood, it may be mediated by various genetic changes associated with glioma progression including: increased platelet-derived growth factor receptor signaling, which is commonly observed during early stages of glioma pathogenesis (2) and which can trigger deregulated c-Myc expression through Src tyrosine kinase signaling (11)(12)(13), amplification and rearrangement of the c-Myc locus (14), and inactivation of the c-Myc pathway antagonist Mxi1 (15,16). In addition, genetic lesions affecting the Ras/phosphatidylinositol 3-kinase/Akt pathway, which are often present in glioblastoma multiforme tumors, can prolong the half-life of c-Myc and hence contribute to c-Myc accumulation in high grade gliomas (17).
In this study, we test the hypothesis that overexpression of human c-Myc 2 in astroglia triggers gliomagenesis. The c-Myc locus consists of three exons that encode various alternatively translated c-Myc isoforms, designated c-Myc 1, c-Myc 2, and c-Myc S (18,19). Translation of c-Myc 2 and S are initiated from AUG codons in exon II, whereas c-Myc 1 is initiated from a non-AUG (CUG) codon in exon I. Although both c-Myc 2 and S promote cell proliferation and apoptosis and have been associated with cellular transformation (19,20), c-Myc 1 seems to play a separate role in growth arrest control (18). Here we prepared transgenic mice with a human c-Myc 2-and S-encoding minigene under transcriptional control of an astrocytespecific promoter. We report that transgenic mice overexpressing c-Myc 2 and S proteins develop early onset genetically engineered murine (GEM) malignant gliomas that show features of human primary malignant gliomas.

EXPERIMENTAL PROCEDURES
Transgene Construction, Generation, and Screening of Mice-To direct transgenic expression of c-Myc 2 and c-Myc S in astroglia, the myelin basic protein promoter of the myelin basic protein/c-Myc plasmid (21) was removed by digestion with HindIII and BamHI and replaced with a PCR product encompassing 2.2 kb of the human GFAP gene transcriptional regulatory sequences (22). This places the genomic sequence containing exons II and III of the human c-Myc gene under control of the GFAP promoter. The transgene fragment was excised from the plasmid backbone by EcoRI digestion, gel-purified, and microinjected at a concentration of 6 ng/l into fertilized mouse eggs as previously described (23). Transgenic mice were identified by PCR analysis of DNA samples prepared from tail biopsies using the following PCR program: one cycle at 96°C for 5 min; 35 cycles at 96°C for 30 s, 64°C for 30 s, and 75°C for 30 s; and one cycle at 75°C for 8 min. The primer sets used were: 5Ј-TACAAGCATGAGCCACCCCAC-3Ј (GFAP promoter primer, sense) and 5Ј-GGAGAATCGGACACATCCTCG-3Ј (human c-Myc primer, antisense) or the human c-Myc specific primers 5Ј-CGATTCCTTCTAACAGAAATGTCCTGAG-3Ј and 5Ј-AGATTTGG-CTCAATGATATATTTGCCAG-3Ј.
Immunohistochemistry-In some cases, the mice were given an intraperitoneal injection of 0.1 mg of BrdUrd/g of mouse weight (diluted in physiological saline) 3 h before the mice were sacrificed. The brains and optic nerves were removed from transgenic and control mice and immediately frozen in liquid N 2 . Tissue sections (20 m) were cut with a cryostat microtome, collected on gelatinized glass slides, fixed with methanol for 5 min, and either stained with hematoxylin or subjected to immunohistochemistry. The fixed sections were preincubated 15 min in phosphate-buffered saline containing 2% bovine serum albumin to reduce background, and immunohistochemistry was carried out as previously described using phosphate-buffered saline with 2% bovine serum albumin as the antibody diluent (24). The primary antibodies used at 1:100 dilution are mouse CNPase (Neomarkers), mouse NeuN (Chemicon), mouse GFAP (Chemicon), mouse VEGF (Santa Cruz), mouse PCNA (Sigma), rabbit GFAP (Sigma), rabbit c-Myc (Santa Cruz), rabbit cleaved caspase-3 (Cell Signaling Technology), and rabbit HOF antisera (25). For staining with sheep bromodeoxyuridine antisera (1:200 dilution; Research Diagnostics), the fixed sections were first treated for 30 s each with 2 M HCl and 0.1 M borate. Primary antibodies were detected with tetramethylrhodamine B isothiocyanate-conjugated swine antirabbit and fluorescein isothiocyanate-conjugated goat anti-mouse (DAKO) and donkey anti-sheep (Research Diagnostics Inc.) at 1:100 dilution. TUNEL labeling was carried out using the in situ cell death detection kit (Roche Applied Science).

RESULTS
Approximately half of a total of 14 transgenic mice, established by pronuclear microinjection of a c-Myc 2/S recombinant minigene, developed neurological disabilities (Fig. 1, A and B, and Table I). The animals were killed at the time they pre- The transcriptional start site is indicated by an arrow. C-K and M, gross and histopathological appearance of representative GEM gliomas from affected F 0 transgenic mice, demonstrating the range of abnormalities observed: enlarged cerebrum (C-E); bulging out of the temporal cortex (D and E, asterisks); enlargement of the lateral ventricle (E and F); diffuse and widespread areas of increased cell proliferation (F and G, asterisks); center of cell death with pseudo-palisading (H, arrow); a well circumscribed tumor localized to the diencephalon (I, asterisk); areas of necrosis observed in a large tumor (J, arrow); scattered GEM glioma cell infiltration in the cerebrum without cell death centers (K); and optic nerve glioma (M, arrow). L, optic nerve from a normal mouse (arrow). N-Q, brain histopathology in the Tg-7 F 1 transgenic mice, showing ataxia and balance difficulties (N); enlargement of the cerebrum and atrophy of the cerebellum compared with a normal brain (O); infiltrating areas of increased cellularity in the cerebrum together with pseudo-palisading (P); and the presence of a vestigial cerebellum (Q). LV, lateral ventricle; Cer, cerebellum.

c-Myc-induced Malignant Gliomas in Mice
sented with terminal disease, which in most cases occurred before 3 weeks of age (Table I). Affected animals harbored markedly enlarged cerebrums, enlargement of the lateral ventricles, as well as increased cellularity in the cerebral hemispheres ( Fig. 1, C-H, K, O, and P). The increment in cellularity involved the neocortex, striatum, and subventricular zone of all of the affected animals ( Table I). The high cellularity of the tumors as well as the presence of small and large centers of cell death with or without pseudo-palisading are reminiscent of glioblastoma multiforme ( Fig. 1, H, J, and P). A subset of animals harbored a highly scattered and widespread tumor cell infiltration in the cerebral cortex without cell death centers (Fig. 1K). This pattern resembles that of the extensively infiltrating human malignant glial cell neoplasm gliomatosis cerebri, which lacks the necrotic centers characteristic of glioblastoma multiforme. A large circumscribed tumor was disclosed in one mouse that occupied the dorso-ventral axis of the diencephalon, i.e. from the thalamus to the optic nerves ( Fig. 1, I and M). This tumor exhibited a high degree of cellularity, necrotic centers, and microvessels but showed low to moderate spreading of tumor cells outside the diencephalon. A subset of animals harbored, in addition to cerebral glioma, a large intracranial hemorrhage that compressed and atrophied the surrounding structure (data not shown). Two founder mice, designated Tg-7 and Tg-8, gave rise to transgenic progeny that became disoriented and ataxic during the second and third postnatal weeks ( Fig.  1N and Table I). The mice harbored markedly enlarged cerebrums that showed profiles of diffuse infiltrating malignant glioma lesions with pseudo-palisading and central cell death centers (Fig. 1, O and P). The cerebellums of the mice were rudimentary, lacking deep fissures and folia (Fig. 1Q). In line with the specificity of the GFAP promoter, other organs appeared normal in affected GFAP/c-Myc transgenic mice.
Transgenic animals harbored a spectrum of abnormal brain cells ranging from an increased cellular density around the lateral ventricles to diffuse infiltration of neoplastic cells in the cerebral cortex. To reveal the cellular origin of these cell populations, we investigated whether or not they expressed the astrocytic intermediate filament protein GFAP. As shown in Fig. 2 (A-F), the abnormal cell populations in both the neocortex and periventricular zone express GFAP, indicating that they originate in cells of the astroglial lineage. In line with a glial cell origin, there was no expression of the neuronal differ-entiation marker NeuN in these cells (Fig. 2G). There were numerous hypertrophic GFAP-positive astrocytes located in the vicinity of tumor foci, and processes from some of these cells appeared to be in direct physical contact with tumor cells (Fig.  2H). Whether these cells represent normal reactive astroglia that develop in response to brain pathology or they constitute a subpopulation of diffusely proliferating neoplastic cells is not clear at the present. Moreover, GFAP-expressing cells also infiltrated the white matter of affected brains in association with a severe hypomyelination (Fig. 2, I-L).
A characteristic feature of human malignant gliomas is that they are often composed of a mixture of proliferating, dying, and metabolically stressed cells that express angiogenic factors. To determine whether or not the GEM gliomas show some of these characteristics, we first investigated the pattern of BrdUrd labeling in brains of affected transgenic mice following administration of BrdUrd 3 h before killing. This short time frame ensures that only actively dividing cell populations are labeled. Using BrdUrd labeling and immunohistochemical localization, numerous BrdUrd-positive cells were revealed in both diffuse and circumscribed glioma lesions (Fig. 3, A-D). On the contrary, only a few positive nuclei were disclosed in agematched nontransgenic control animals, and they appeared mainly in regions of postnatal neurogenesis such as the subgranular zone of the dentate gyrus (data not shown). In line with a high tumor grade, there was also morphological evidence of regressive processes in the GEM gliomas, including circumscribed and small centers of cell death (Fig. 1). Thus, to reveal whether apoptosis plays a role in the continued loss of cells in the tumors, we used molecular markers of programmed cell death including TUNEL labeling and staining with an anti-cleaved caspase 3 antibody. Numerous TUNELpositive nuclei were scattered around large necrotic centers in the GEM gliomas (Fig. 3, E and F). As mentioned above, some GEM gliomas showed a widespread diffuse infiltration of the cerebral cortex in the absence of cell death centers. These gliomas, however, were TUNEL-positive and harbored cleaved caspase 3-expressing apoptotic cells scattered within tumor foci (Fig. 3, G-J). Expression of the vascular endothelial cell growth factor (VEGF) is up-regulated in human malignant glioma cells located at the periphery of necrotic centers (26,27). In line with this, VEGF expression was also up-regulated in GEM glioma cells surrounding cell death  (Figs. 3, K and L, and 4). We next investigated the pattern of c-Myc expression in various GEM gliomas. As shown in Fig. 4, expression of the c-Myc transgene was revealed in tumor biopsies from c-Myc 2/S transgenic mice but not in brain biopsies from control brains. We consistently detected two c-Myc specific bands of ϳ64 and 52 kDa in Western blots. The 64-kDa band is consistent with the molecular mass of c-Myc 2, and the 52-kDa band is consistent with that of c-Myc S. The c-Myc S isoform is 100 amino acids shorter than c-Myc 2, corresponding to a difference of about 11 kDa. Notably, the c-Myc S isoform seems to predominate in a well circumscribed glioma, whereas a more equal distribution of the two c-Myc isoforms is seen in diffuse infiltrating tumors (Fig. 4). PCNA is a component of DNA polymerase Ѩ and is widely used as a nuclear marker of continuously dividing tumor cell populations (28). As expected for actively dividing malignant cells, there was a strong correlation between the pattern of PCNA expression and that of c-Myc in the tumors (Fig. 4). Because expression of a number of signaling molecules is enforced in human malignant gliomas, we next determined whether this was also the case in the early onset GEM gliomas. Overexpression and/or amplification of EGFR occur in many (ϳ40% of cases) human primary glioblastomas but less often in lower grade gliomas and secondary glioblastomas (3). Hence, EGFR overexpression is characteristic of de novo malignant lesions and presumably responsible for activation of the Ras pathway in these tumors, because activating p21 Ras mutations are uncommon in gliomas (29). As shown in  Fig. 4, we found EGFR overexpression in tumor biopsies from various GEM gliomas. Moreover, the p53-negative regulator Mdm2 is also overexpressed in many human primary glioblastomas (3). Mdm2 has been noted to bind to p53 to both inhibit transcription and to target p53 for degradation resulting in an increased resistance of tumor cells to programmed cell death (30). In line with a primary malignancy, a subset of GEM gliomas showed a marked increase in Mdm2 expression (Fig.  4). Thus, the c-Myc-expressing GEM gliomas seem to harbor a molecular signature resembling that of human primary glioblastomas. Cerebral edema occurs in association with most malignant brain tumors including gliomas (31). Affected GFAP/c-Myc transgenic mice showed features of brain edema, including swelling of the brain, demyelination of white matter tracts, as well as reactive glioses (Figs. 1, D and O, and 2, H-J). Hence, we investigated the pattern of expression of the astroglial water channel protein aquaporin-4 in GEM gliomas because this protein has been claimed to be up-regulated in human edematous malignant gliomas (32). We found a moderate increase in aquaporin-4 expression in diffuse infiltrating GEM tumors (Fig. 4). In contrast, there was no discernable change in expression of the cell proliferation-associated transcription factor E2F1 in the tumors (Fig. 4).

c-Myc-induced Malignant Gliomas in Mice
We next investigated the pattern of connexin 43 (Cx 43) expression in different GEM gliomas, because this astroglial gap junction protein has been noted to play important roles in glioma cell invasion as well as in the establishment of astroglial tumor syncytia (33,34). Moreover, a variable expression of Cx 43 has been disclosed in human glioblastoma multiforme tumors (33,(35)(36)(37). In line with a role of the protein in astroglioma cell spreading and invasion, there was a marked lack of Cx 43 expression in a well circumscribed GEM glioma as opposed to diffuse infiltrating lesions (Fig. 4). The fact that the latter tumors express the astroglial gap junction protein Cx 43 (Fig. 4) indicates that they have the capacity to establish syncytial tumor networks between GEM glioma cells and astroglia (33,34). Consistent with the presence of tumor syncytia, only a subpopulation of cells within individual GEM tumor foci seems to express the c-Myc transgene (Fig. 5,  A-D). Moreover, these c-Myc-expressing cells appear to constitute an actively dividing subpopulation because they coexpress the PCNA antigen (Fig. 5, E-H). Thus, diffuse infiltrating GEM tumors seem to comprise both actively dividing true neoplastic cells as well as a less proliferating bystander cell population.
We next investigated the possibility that the less dividing cell population of the GEM tumors express the novel astroglial marker HOF. The HOF gene encodes a nuclear broad complex tramtrack bric-a-brac/poxvirus and zinc finger domain factor expressed in both developing and mature astrocytes as well as in reactive astroglia during brain pathology (25). As shown in Fig. 6 (A-C), levels of HOF expression are up-regulated in some but not all GEM gliomas. Using BrdUrd labeling and immunohistochemical localization, we were unable to detect either BrdUrd and HOF (Fig. 6, D-G) or PCNA and HOF (Fig. 6, H and I) double-labeled cells in diffuse GEM lesions. Thus, HOF FIG. 4. Molecular signatures of GEM gliomas. Western blot analysis of protein extracts from control cerebral cortex (Cont) and GEM gliomas 1-4. GEM glioma 3 is highly circumscribed. Equivalent protein amounts were loaded on separate protein gels, blotted, and probed with the indicated antibodies. In some cases, the probed blots were stripped and reprobed with a different antibody. The results demonstrate that all GEM gliomas exhibit elevated levels of c-Myc 2, c-Myc S, and the proliferation marker PCNA. In the circumscribed GEM glioma 3, a substantial increase in c-Myc S is observed together with a reduced expression of Cx 43 and aquaporin-4 (AQP4) and a marked increase in VEGF, relative to the diffuse GEM gliomas 1, 2, and 4. Whereas EGFR is elevated in all the GEM gliomas, expression of Mdm2 is variable, and there is no observable change in E2F1 protein levels in GEM gliomas compared with the control extract.

c-Myc-induced Malignant Gliomas in Mice
is unlikely to be expressed in actively dividing GEM tumor cells. Moreover, HOF is expressed in cells with both large and small sized nuclei (Fig. 6I). Co-immunolocalization experiments show that HOF expression in GEM gliomas is often confined to cells that show robust expression of the astroglial differentiation marker GFAP (Fig. 6, J and K). Taken together, these data suggest that diffuse infiltrating GEM tumors, like human malignant gliomas, are a mixed cell population comprising tumorous and less malignant reactive elements.
As shown above, affected F 1 or F 2 transgenic mice of the Tg-7 and Tg-8 strains harbor sausage-shaped cerebellar rudiments (Fig. 1, O and Q). The deficiency in these animals seems to affect the developmental transformation of vestigial cerebellums into a mature structure with distinct internal granule cell and molecular layers (Fig. 7, A, E, I, M, and Q). During normal cerebellar development, granule neuron precursors in the external granule cell layer migrate inward through the molecular layer along the processes of radial Bergman glia to settle in the internal granule cell layer (38). Because radial Bergmann glia are thought to be essential for granule cell migration as well as morphological transformation of the cerebellum during development (39), disruption of the differentiated functions of these cells may in part explain the cerebellar phenotype in the animals. Moreover, the GFAP promoter has been noted to be active in radial Bergman glia in some transgenic mouse models (40,41). In normal cerebellums, the GFAP-positive radial processes of Bergman glia extend from a subpial position across the molecular layer to terminate in the Purkinje cell layer at the border of the internal granule cell layer (Fig. 7, B-D). In line with a proposed defect in Bergman glia development, the rudimentary cerebellums show a pronounced lack of radial glia in the subpial region (Fig. 7, F, N, and R). Because mature cerebellar granule cells, as opposed to immature cells, express the neuronal differentiation marker NeuN (25), we performed NeuN immunohistochemistry to investigate whether differentiated granule neurons were present in the vestigial cerebellums in the absence of a distinct internal granule cell layer. Consistent with a Bergman glia-mediated migration failure of granule cells, numerous NeuN-positive granule cells were scattered randomly in subpial locations (Fig. 7, G, H, O, P, S, and T). On the contrary, only few NeuN-positive cells were located in cell-dense regions at the center of vestigial cerebellums (Fig.  7, I and K) where GFAP-expressing glial cells prevail (Fig. 7, J and L).

DISCUSSION
Our results show that deregulation of the c-Myc pathway in astroglia predisposes mice to early onset gliomas. Consistent with an early astroglial origin of tumor cells is the observation that an abnormal population of GFAP-expressing cells were manifest following the onset of gliogenesis in the postnatal subventricular zone of the lateral ventricles (42). The early astroglial onset of the GEM gliomas may be explained by a window of vulnerability during astrocyte development where deregulation of the c-Myc pathway effectively sustains cell proliferation. Perhaps the continued presence of active c-Myc/ Max complexes in developing astroglia prevents induction of terminal differentiation by opposing (e.g. Mad/Max) complexes of the Myc/Mad/Max network (6,43). In the absence of counterbalanced apoptosis, the increase in cellularity that results from sustained Myc/Max-mediated proliferation of astroglial precursors may be instrumental in the short latency of GEM gliomas in this model. Deregulation of the c-Myc pathway has also been noted to trigger genomic instability (44 -47). This instability coupled with an increased cellularity may accelerate the process of neoplastic progression of c-Myc-overexpressing astroglia, resulting in induction of co-operating oncogenes, including those involved in cell survival and proliferation. Furthermore, it has been suggested that human primary glioblastomas originate in primitive glial precursor cells as opposed to secondary glioblastomas that may arise from dedifferentiation of low grade gliomas (4). In addition to early astroglial origin, the molecular signature of the GEM tumors resembles that of human primary glioblastoma multiforme, including EGFR and Mdm2 up-regulation.
In contrast to this study our previous efforts to establish a transgenic mouse model for oligodendrocytic gliomas using the

c-Myc-induced Malignant Gliomas in Mice
same human c-Myc transgene but under control of a myelin gene promoter (myelin basic protein/c-Myc) were less successful (21,24). Although some of the transgenic mice harbored abnormal cellular profiles in the striatum and neocortex in association with BrdUrd positive cells, there is no evidence of an increased cellularity and/or GEM gliomas in these mice (21). 2 On the contrary, affected myelin basic protein/c-Myc mice developed a lethal programmed cell death disorder characterized by a widespread and scattered oligodendrocyte loss and pronounced hypomyelination of white matter tracts. Deregulation of the c-Myc pathway in developing oligodendroglia seems to be compliant with the ability of these cells to differentiate but not with the cell cycle arrest required for myelin formation. Although the foci of diffuse GEM gliomas in GFAP/c-Myc mice harbor numerous apoptotic cells, it remains to be determined whether this is primarily a result of aberrant cell cycle activity in differentiating astroglia and/or a consequence of metabolic stress.
The 150-amino acid N-terminal transactivation domain of human c-Myc 2 comprises two conserved Myc homology boxes, designated MBI (amino acids 45-65) and MBII (amino acids 128 -144). The MBI box plays a role in stabilization of c-Myc 2 through phosphorylations of Thr 58 and Ser 62 (5). The MBII subregion interacts with TRRAP, which in turn functions as a co-activator that recruits histone acetylase activities and hence links c-Myc to nucleosome remodeling processes (8,43,48). c-Myc S proteins arise from a leaky ribosome scanning of the c-Myc 2 transcript, resulting in translation initiation at two closely spaced internal methionines (Met 101 and Met 110 ) located upstream of the MBII subregion (19). We found that the majority of GEM tumors expressed comparable levels of c-Myc 2 and c-Myc S, with the exception of a highly circumscribed malignant tumor that harbored a severalfold increase in c-Myc S. The synthesis of comparable levels of the two c-Myc isoforms has been noted in some hematopoietic tumor cell lines (19). Although c-Myc S has retained many of the biological activities of c-Myc 2, including E-box binding via heterodimerization with Max, triggering of cell proliferation, and apoptosis (19,20), it remains to be determined whether the N-terminal transactivation domain-truncated c-Myc S isoforms have oncogenic potential in vivo.
Astroglia function as syncytial networks in which numerous cells are connected by gap junctional pores and where small secondary messengers such as inositol 1,4,5-trisphosphate, cAMP, ATP, and ions are shared by the cytoplasm of groups of cells (49). Moreover, astroglial syncytia also establish gap junctional communication with neurons and oligodendroglia and play pivotal roles in supplying the former with essential energy 2 F. Lihme and N. A. Jensen, unpublished data. ). An area of increased cellularity in the transgenic cerebellum (I, asterisk) coincides with increased GFAP staining (J, asterisk and L, indicated by wЈ) and decreased NeuN staining (K). In some animals, disorganized Bergman-like glia are observed, but they lack clear radial processes (R, indicated by #). m, molecular layer; g, granular layer; w, white matter.

c-Myc-induced Malignant Gliomas in Mice
metabolites as well as in glutamate recycling at synaptic terminals (50). The main gap junctional protein of astroglia, Cx 43, is expressed in many human malignant gliomas (33,(35)(36)(37). The establishment of Cx 43-mediated syncytial tumor networks with astroglia has been noted to be important for glioma cell invasion in vivo (34). Consistent with this finding, we observed a pronounced lack of Cx 43 protein in a well circumscribed GEM glioma that exhibited little tumor cell spreading. On the contrary, Cx 43 protein was present in biopsies from diffuse infiltrating GEM lesions. The survival of metabolically active glioma cells depends on an efficient supply of energy metabolites. Assuming that GEM gliomas establish functional gap junctions with bystander astroglia, this could serve as an important route for energy metabolite supply in diffuse infiltrating tumor cells in the absence of adequate vascularization. Furthermore, the foci of diffuse infiltrating GEM tumor cells appear morphologically similar, although they comprise cells with distinct molecular signatures. Our results show that the diffuse GEM lesions are a mixed cell population comprising actively proliferating elements that co-express c-Myc and PCNA and a subpopulation of less dividing bystander cells that co-express HOF and GFAP. A complicated problem in neurooncology is the difficulty of distinguishing true astrocytic tumor cells from reactive glia. Furthermore, although reactive astroglia rarely divide, GFAP is not a reliable marker for nondividing cells because actively dividing astroglia have been noted to occasionally express the intermediate filament protein (51). The finding that HOF is a marker of cells that are not actively dividing in diffuse infiltrating GEM lesions suggests that the protein in association with GFAP could serve as a marker of nontumorous bystander cells in human malignant gliomas. However, whether HOF is a definitive marker for nondividing reactive elements in diffuse infiltrating gliomas including those in humans is under investigation. Our preliminary results indicate that the pattern of HOF expression in human grade IV malignant gliomas resembles that disclosed in GEM gliomas. 3 In addition to GEM gliomas, a subset of the transgenic mice also developed a deficiency affecting morphogenesis of the cerebellum. The deficiency correlates with a pronounced lack of radial Bergman glia in the molecular layer, resulting in impaired granule cell migration. Hence, we surmise that sustained c-Myc expression in Bergmann glia blocks the differentiation of these cells into an ordered array of radial rails traversing the molecular layer. In the absence of this glial rail, migrating granule neurons born in the external granular cell layer settle and differentiate prematurely into NeuN-expressing granule neurons in the presumptive molecular layer. In addition, some affected transgenic mice seem to harbor a migration deficiency of pyramidal neurons in the archicortex in which the morphology of Ammon's horn of the hippocampus is abnormal, consisting of a wide band of noncompacted pyramidal cells (data not shown). However, the exact mechanisms of these phenotypes as well as the role played by the Myc pathway remain to be determined.
In summary, we have shown that developing astroglia are highly vulnerable to deregulation of the c-Myc pathway leading to formation of early onset GEM tumors that show features of human primary malignant glioblastoma multiforme. Based on these findings, we surmise that sustained c-Myc activity in primitive glial cell precursors is instrumental in the development of at least a subset of human primary malignant gliomas, although this remains to be determined. An interesting feature of this model is the molecular heterogeneity of the GEM tumors comprising both an actively dividing cell population and a less dividing bystander population that expresses the novel astroglial marker HOF in association with GFAP. Furthermore, the ability to obtain affected transgenic offspring through heterozygous crosses of the Tg-8 line greatly facilitates ongoing studies on the biology of these tumors.