Downloaded 09/27/21 07:54 AM UTC H

Neurosurg Focus 8 (4):Article 1, 2000, Click here to return to Table of Contents

A review of astrocytoma models

HAO DING, M.D., PH.D., ANDRAS NAGY, PH.D., DAVID H. GUTMANN, M.D., PH.D., AND ABHIJIT GUHA, M.D., F.R.C.S.C. Samuel Lunenfeld Research Institute, Mt. Sinai Hospital; Department of Molecular and Medical Genetics; Labatts Brain Tumor Center, Hospital for Sick Childrens; Division of Neurosurgery, University Hospital Network, University of Toronto, Toronto,Ontario Canada; Department of Neurology, Washington University School of Medicine, St. Louis, Missouri

Despite tremendous technical improvements in neuroimaging and neurosurgery, the prognosis for patients with malignant astrocytoma remains devastating because of the underlying biology and growth characteristics of the tumor. However, our understanding of the molecular bases of these tumors has greatly increased due to study findings involving operative specimens, astrocytoma predisposing human syndromes, teratogen-induced animal and established human astrocytoma cell lines, and more recently transgenic mouse models. Appropriate small-animal models of spontaneously occurring astrocytomas, which replicate the growth and molecular characteristics found in human tumors, are essential to test the relevance and interactions of these molecular aberrations. In addition, it is hoped that relevant molecular targets will eventually be therapeutically exploited to improve patient outcomes. Appropriate animal models are also essential for testing these novel biological therapies, before they are brought to the clinic, requiring a large investment of time and money. In this paper, various astrocytoma models are discussed, with emphasis on transgenic mouse models that are of great interest to laboratory investigators.

KEY WORDS ¥ astrocytoma ¥ glioblastoma multiforme ¥ mouse ¥ gene therapy

Malignant astrocytoma represents one of the most dev- increasing understanding of the molecular pathogenesis astating tumors affecting children and adults. Surgery and and development of novel biological therapies of malig- adjuvant conventional radio- and chemotherapy have had nant astrocytomas, there is a need to develop appropriate minimal effect on changing the poor prognosis, which re- models in which to test these novel therapies. Indeed, mains at a median range of only 9 to 12 months.35 Malig- many adjuvant therapies, in which promise in various nant astrocytomas are composed of pleomorphic, hyper- existing in vitro and in vivo models of astrocytomas has proliferative infiltrative astrocytes, with areas of necrosis, been demonstrated, have been applied with little success increased tumor angiogenesis and regions of bloodÐbrain in patients; however, there has been a significant cost in barrier breakdown. These heterogeneous pathological terms of time, patient psychological and occasional phys- characteristics pose major obstacles to the effective man- ical morbidity, and economic issues. These failures are agement of gliomas.11,35 In addition to pathological het- partly based on the fact that we sometimes short-circuit erogeneity, there is associated molecular heterogeneity, the rigors of the scientific investigational process because the expression, interactions, and functional importance of of the terminal nature of the disease and because of our which we have just begun to decipher.16,39,50 It is hoped own enthusiasm to obtain a cure. This has contributed to that this increased understanding of the molecular patho- cynicism and nihilism expressed by the public and clini- genesis of astrocytomas will lead to novel therapeutic tar- cians regarding the management of malignant astrocy- gets, which can be exploited in combination with a variety tomas. Hence, it is important that we expend the time and of genetic, pharmacological, and other approaches to effort to develop and test our novel therapies in appropri- improve the overall clinical prognosis. In tandem with our ate preclinical models that most accurately reproduce the clinical, histological, and molecular characteristics found in human astrocytomas; it is hoped that this will translate Abbreviations used in this paper: CDK = cyclin dependent kinase; into improved success in subsequent clinical trials. EFGR = epidermal growth factor receptors; ES = embryonic stem; GBM = glioblastoma multiforme; GFAP = glial fibrillary acid Currently available models of astrocytomas include in protein; NF = neurofibromatosis; PDGFR = platelet-derived growth vitro cell lines derived from human malignant astrocyto- factor receptor; TSC = tuberous sclerosis complex; VEGF = vascu- mas as well as chemically or virally induced rat or mouse lar endothelial growth factor. astrocytomas. These cell lines have certainly served to in-

Neurosurg. Focus / Volume 8 / April, 2000 1

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC H. Ding, et al. crease our knowledge of the genetic aberrations in astro- nant Brain Tumor-1 gene46), amplification of EGFR and/ cytomas, permitting a variety of molecular biological ma- or EGFRvIII, and overexpression of VEGF. A second, nipulations to test the functional significance of these younger group progresses through increasing grades of alterations. However, it is well recognized that cell lines astrocytomas, and in these GBMs there are also earlier harbor different genetic profiles that are not representative genetic aberrations such as loss of the p53, p16, and renti- of the parent tumor, a fact that is even more important in noblastoma cell cycle regulatory genes in combination heterogeneous tumors such as malignant astrocytomas. with overexpression of PDGFR.1,5,18,20,21,23,28,33,45,48,56,57,63 Current in vivo astrocytoma models include growth of Chromosomal and loss of heterozygosity analyses of these cell lines in heterotopic or orthotopic locations in GBMs demonstrate amplification and loss of several ge- immunocompromised or syngenic hosts; however, most netic loci. Amplifications most commonly involve chro- fail to recapitulate the infiltrative nature of human malig- mosomes 7 containing EGFR and 12q containing the nant astrocytomas. Human tumor explant xenograft mod- CDK4, MDM2, and Rap1B loci.37,51,68 Loss of the entire els also suffer from these deficiencies, even when chromosome or several regions on chromosome 10 is also attempts have been made to implant them into the brain prevalent in GBMs as discussed previously, with PTEN/ and additionally they are plagued by an extremely low and MMAC1 expression lost in a majority of GBMs, although variable tumor formation rate. Although chemically or the actual rate of mutations is much lower.61,64,65 Mutations virally induced animal models exist, they unfortunately and loss of heterorzygosity at the p53 loci (17p) are ob- also suffer from their low incidence of tumor formation, served in two thirds of lower-grade (World Health Or- and possess significant molecular and pathological differ- ganization Grades I and II) astrocytomas but in only ap- ences from human astrocytomas. Another group of poten- proximately one third of GBMs.18,56 Last, loss of CDK tial in vivo astrocytoma models are found in familial can- inhibitors (p16 and p15) and the loss of retinoblastoma are cer predisposition syndromes that are associated with an also frequently detected, such that aberrations of the p53 or increased incidence of astrocytomas and their correspond- retinoblastoma-regulated cell cycle pathways at one level ing gene knockout mouse models. These include NF1 or another are prevalent in most astrocytomas.1,28,33,48,57,63 and NF2, TSC, and the LiÐFraumeni cancer syndrome. Unfortunately, the corresponding knockout mice do not develop astrocytomas, perhaps reflecting a combination of CANCER PREDISPOSITION SYNDROMES the low incidence of astrocytomas in these familial syn- ASSOCIATED WITH ASTROCYTOMAS AND dromes, the requirement of additional genetic aberrations CORRESPONDING MOUSE MODELS for developing astrocytomas, and the limitations and dif- ferences in mouse models as compared with humans. Examination of cancer predisposition syndromes, Despite the limitations, ongoing research is still re- which have at least one well-characterized genetic alter- quired to develop spontaneously occurring small-animal ation, and their corresponding mouse knockout models is models of astrocytomas that mimic the pathological and potentially informative. The genetic alterations demon- molecular profile of human astrocytomas; additionally, strated in the predisposition syndromes are usually found research in these animal models has the potential to ad- in the more common sporadic human counterparts; there- vance our ability to treat and investigate this disease. Ideal fore, what we learn from studying these well-defined pa- animal models of human astrocytomas should fulfill the tient cohorts and the corresponding mouse models may be following criteria: 1) tumors should arise from astrocytes; extrapolated to the sporadic tumors. 2) tumors should be transformed as characterized by im- mortalization in vitro and the ability to be transplanted Neurofibromatosis Type 1 into syngeneic or immunocompromised hosts in vivo; 3) Neurofibromatosis Type 1 is a common autosomal tumors should develop in small inexpensive animals in dominant cancer predisposition neurocutaneous syndrome which there are relatively short and predictable induction in which 15 to 20% of affected individuals develop astro- times; 4) tumors should share similar molecular profiles cytomas, commonly involving the optic pathway.13 with human astrocytomas; and 5) tumors should be re- Neurofibromin, the protein product of the NF1 gene, func- sponsive to known therapies (radio- and chemotherapy) tions as a tumor suppressor largely by accelerating the that have demonstrated efficacy in human tumors. hydrolysis and thereby inhibiting a key intracellular signaling pathway mediated by activation of Ras.13 Muta- MOLECULAR PATHOGENESIS OF tions of the NF1 gene resulting in loss of neurofibromin expression are not prevalent in sporadic astrocytomas, in ASTROCYTOMA which, indeed, its levels are elevated due to a positive A molecular pathogenic profile of astrocytoma is feedback of increased Ras activity that occurs in human slowly emerging: the loss of specific tumor suppressor astrocytomas.22,25 Recently we have had a chance to ana- genes as well as activation of critical oncogenes provide lyze biochemically one NF1-associated astrocytoma, and the rationale for development of the transgenic mouse we demonstrated that although it differs from sporadic models discussed later in this review (Fig. 1). At least astrocytomas in terms of loss of neurofibromin, it exhibits two molecular pathways have been proposed for GBMs.39 activation of the Ras and Phosphoinositide 3-OH Kinase First, patients can present de novo, with the tumor char- intracellular signaling pathways (unpublished data). The acterized by deletions on chromosome 10 harboring at mouse knockout models of the Nf1 gene do not demon- least two tumor suppressor genes (PTEN/MMAC1 [a dual strate development of astrocytomas.32 Whereas homozy- specific and lipid phosphatase]36,59 and Deleted in Malig- gous deletion of Nf1 (-/-) is embryonically lethal as a

2 Neurosurg. Focus / Volume 8 / April, 2000

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC A review of astrocytoma models

Fig. 1. Diagram illustrating the speculated molecular aberrations in the development of human astrocytomas, involving loss of tumor suppressor genes (such as p53, retinoblastoma, p16) and gain of tumor promoting genes (such as PDGFR, and EGFR). At least two molecular pathways have been postulated for GBMs: secondary GBM involving progression from a low- to high-grade astrocytoma and primary GBM involving de novo development of a GBM. Both of these types of GBMs, which are pathologically indistinguishable, are characterized by loss of PTEN/MMAC1 and other not-yet-identified tumor suppressor genes on chromosome 10 that negatively regulates the PI3-kinase and PKB/Akt sig- nalling pathway. In addition, a large proportion of GBMs has amplification and overexpression of normal and mutant EGFRs, in addition to other growth factor receptors that serve to increase activation of the p21ÐRas signaling pathway. How these molecular aberrations interact, and how they facilitate various aspects of the growth of astrocytomas, is best understood in animal models such as transgenic mouse models. In addition, these models provide reagents with which to test novel therapeutics, which may be developed based on our increased knowledge of the molecular pathogenesis of astrocytomas. WHO = World Health Organization. result of cardiovascular developmental abnormalities, the NF1. The prevalence and incidence of NF2 is 10-fold less heterozygous Nf1 mouse model (+/-) develops a variety of common than NF1; however the incidence of astrocy- tumors in late adult life but not astrocytomas.32 However, tomas, ependymomas, and meningiomas affecting the closer examination of the Nf1 heterozygote brains reveals brain and spinal cord is comparatively higher. Two groups the increased proliferation of astrocytes as compared with of investigators cloned the NF2 gene on chromosome 22q, wild-type littermates,26 with proliferation inhibited by Ras encoding a protein termed either "merlin" or "schwan- effector pathway blockade. Hence, although examination nomin." Merlin/schwannomin shares similarities with of the Nf1 mouse model may not yield much information members of the protein 4.1 family, including ezrin, radix- with regards to astrocytomas, the derivative astrocytes in, and moesin, which link cell glycoproteins to the actin with reduced neurofibromin expression have increased cytoskeleton.24 The normal function of merlin/schwanno- Ras pathway activation. The Nf1 heterozygote mice that min and how it relates to cellular transformation is not as have been mated with other relevant genetically engi- well known as that of neurofibromin but is being actively neered mice demonstrate genetic cooperativity between investigated. That Nf2 homozygous knockout mice (-/-) Nf1 and cell cycle regulatory genes with respect to astro- die early during embryogenesis (secondary to a failure to cyte growth regulation. For example, crossing of the Nf1 induce mesoderm formation), suggests that merlin is es- mice with the p53-deficient mice has been recently been sential for the development of extraembryonic struc- reported to lead to development of neurofibromas.8 tures.44 The Nf2 heterozygotes (+/-) are cancer prone and Neurofibromatosis Type 2 develop highly metastasizing osteosarcomas, lymphomas, lung adenocarcinomas, hepatocellular carcinomas, and fi- Neurofibromatosis Type 2 is another autosomal-domi- brosarcomas, although, not the classic schwannomas or nant cancer predisposition syndrome that is distinct from the associated gliomas found in patients with NF2.43 Re-

Neurosurg. Focus / Volume 8 / April, 2000 3

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC H. Ding, et al. cently, mice, in which there was overexpression of a dom- and rodents are extremely rare. The most common astro- inant inhibitory isoform of merlin specifically found in cytic tumor is the VM murine astrocytoma arising in an Schwann cells by using the myelin P0 promoter, devel- inbred mouse strain with an incidence of 1.5% at 600 oped schwannomas involving the trigeminal nerve, spinal days.17 In an effort to develop nontransgenic models for ganglia, uterus, stomach, intestine, and pancreas.19 Breed- human brain tumors, three approaches have been con- ing the Nf2 heterozygote mice with other relevant geneti- ducted using: 1) primary astrocytoma cells derived from cally altered mice may yet yield models of gliomas, and patient or from rat astrocytomas; 2) retrovirus injection; or this remains an area of ongoing research. 3) chemical mutagenesis. Xenografts obtained from established human astrocytoma cell lines or tumor explants into Tuberous Sclerosis Complex immunocompromised rodents have likewise been used. Tuberous sclerosis complex is another rare autosomal- These established cell lines suffer from being clonal in dominant neurocutaneous syndrome in which patients de- nature, whereas tumor explants have been shown to have velop hamartomatous lesions in the nervous system and low and variable implantation rates, unpredictable laten- other organs. Individuals with TSC develop a special type cies, and loss of invasive growth over time. Cell lines of astrocytoma, termed subependymal giant-cell astrocy- obtained from chemically induced astrocytomas have also toma.55 Two genes are responsible for TSC, TSC1 and been used, including C6, 9L, RG2, and T9 rat glioma TSC2, which code for hamartin and tuberin, respectively.55 cells. Unfortunately, these chemically induced models are Tuberin has been shown to function as an inactivating pro- histologically and genetically different from human astro- tein for Rap1 (another Ras-like protein that is transform- cytomas. For example, an allogeneic major histocompatibility complex has recently been identified in the C6 rat ing in certain cell types), which is similar to neurofi- 3 bromin’s role in regulating Ras.27,41 The role of this Rap1 astrocytoma model, which could explain the phenome- signaling pathway in glioma pathogenesis is underscored non in which some C6 tumors in the immunocompetent by our observations that approximately two thirds of all rats regress spontaneously even without therapeutic inter- sporadic astrocytomas either display a loss of tuberin (the vention. Therefore, we believe that the C6 astrocytoma 27,67 model in the immunocompetent rat should no longer be TSC2 gene product) or increased expression of Rap1B. used for therapeutic studies, and the available data As previously mentioned Rap1B is amplified as part of obtained in this model need to be critically reinterpreted. the second most common amplicon in human GBMs in- 50 Ethylnitrosourea that is intravenously administered to volving chromosome 12q. The Eker rat, carrying a spon- pregnant rats has also been used to induce chemically a taneously occurring inactivating mutation of TSC2, how- large number of brain tumors in the offspring. These ever, has not been an informative model for astrocytomas, tumors have variable histology (few astrocytomas), loca- as these rats do not develop astrocytoma or subependymal tion, latency, and malignant potential.2 The astrocytomas giant-cell astrocytomas. demonstrated in the litter of these chemically treated rats overexpress a variety of growth factors including PDGF- LiÐFraumeni Cancer Syndrome B, which is similar to what is found in human astro- LiÐFraumeni cancer syndrome is a prevalent familial cytomas. syndrome resulting from germline mutations in the p53 Retroviruses that induce brain tumors include members tumor suppressor locus, a common tumor suppressor gene of the Rous sarcoma virus family and simian sarcoma lost in a variety of sporadic human cancers including viruses in marmosets, whose oncogenic properties are gliomas.42 Astrocytomas, the third most common tumor caused by the overexpression of PDGF-B (v-sis).7 In a type in patients with LiÐFraumeni cancer syndrome, occur recently published model, injected recombinant retrovirus early in age and quickly become malignant in nature.42 was used to achieve cell typeÐspecific brain expression. Although p53 knockout mice (hetero- or homozygous) de- In this model, EGFRvIII has been expressed in both velop a variety of malignancies, they do not develop astro- wild-type mice and those aberrant for relevant cell cycle cytomas;12 however, cultured p53-deficient astrocytes regulatory genes such as p53, p16, retinoblastoma, and grown in culture from these mice do become transformed CDK4.29 A variety of proliferating, GFAP immunoreac- when injected either subcutaneously into immunocompro- tive astrocyte populations were induced that expressed the mised mice or intracranially into syngeneic recipients.4,6,69 inoculated transgene. In these experiments, astrocyte-like cells expressing EGFRvIII did not form tumors unless In these experiments, loss of the wild-type p53 confers a 29 growth advantage and facilitates their transformation in they also overexpressed CDK4. No cooperativity was vitro. With continued passage, p53 -/- astrocytes grow as demonstrated between EGFRvIII and p53. In addition, colonies in soft agar (anchorage-independent growth) and either loss of INK4a (p16) or overexpression of CDK4 permitted the infected astrocyte-like cells to grow without lose contact inhibition. Increased expression of EGFR, 30 PDGFR, and VEGF has been observed in the p53 (-/-)Ð senescence. These results indicate that genetic coopera- transformed astrocytes6 and also in LiÐFraumeniÐassoci- tivity between growth factor and cell cycle regulatory ated malignant astrocytomas,23 similar to sporadic human pathways is required for astrocyte transformation in this GBMs. model. It is not clear, however, whether these astrocytic collections represent true astrocytomas, because derivative cell cultures have not been tested for their ability to NONTRANSGENIC MODELS OF form tumors in another host. Although this retroviral injection strategy does not lead to germline mouse astrocy- ASTROCYTOMAS toma models, such as those potentially derived from trans- Spontaneous brain tumors found in nonhuman primates genic models for use in preclinical therapeutic testing, it

4 Neurosurg. Focus / Volume 8 / April, 2000

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC A review of astrocytoma models has the strength to test quickly the interactions among the alterations in these cells. When the genetically altered ES known genetic aberrations found in malignant human as- cells are injected into a host blastocyst, or aggregated trocytomas, as well as to shed light onto their ontogeny. within a morula-stage embryo, they have the capacity to These retroviral studies are extremely important and contribute to all tissues of the resultant chimeric mouse complement the transgenic strategies in the proceeding (Fig. 2). Most important, they can contribute to germ cells section. and transmit the genetic mutations in vivo, allowing development of established mouse lines in which the altered gene(s) are carried.47,49,52 TRANSGENIC ASTROCYTOMA MODELS Embryonic stem cellÐmediated transgenesis has several Although the knockout mouse models associated with advantages over the standard pronuclear DNA injection cancer predisposition syndromes discussed in the preced- routinely used to create transgenic models. Conventional ing sections have not yielded useful models of astrocy- pronuclear DNA injection frequently results in multiple- tomas, several other knockout mouse models with disrup- copy integration of a transgene, which can result in varia- tion of the genes that are lost in sporadic astrocytomas are tion of transgene expression, whereas ES cellÐmediated of interest. Mice lacking p16, a CDK inhibitor that is transgenesis provides a higher frequency of low-copy found on chromosome 9p that is involved in the numbers or even single copy of transgene integration.38,49 retinoblastoma-regulated cell cycle pathway, and that is In addition, transfected ES clones can be tested in vitro for commonly mutated in human astrocytoma specimens and cell typeÐspecific expression by using in vitro ES cellsÐ derived cell lines (Fig. 1), develop a variety of malignan- differentiation assays, as exemplified by astrocytic lin- cies but not astrocytomas.9,31,54 However, these mice pro- eage differentiation in which retinoic acid is used in the vide a useful background on which additional genetic mouse astrocytoma models derived in our laboratory (un- alterations associated with astrocytomas and GBMs, such published data). Finally, the ES cellÐmediated transgenic as overexpression of EGFRvIII transduced by retroviral approach also allows us to avoid the problem, found in injection, leads to development of astrocytoma-like le- a number of cases, in which expression of the transgene sions.29 Cross-breeding experiments with these p16 (-/-) is lethal, because it involves chimera production with mice with other genetically altered mice are ongoing and transgenic ES cells contributing to different levels in the may yield additional astrocytoma models. A second gene animals. of interest on chromosome 9p is p14ARF, which is also To examine specifically the effects of a genetic alter- commonly mutated in astrocytomas, and is involved indi- ation in a certain cell type, such as an astrocyte, cell-spe- rectly in the p53 pathway by its regulation of MDM2.60 A cific promoters can be used in conventional or ES cell small proportion of p14ARF knockout mice develop transgenesis. In the nervous system, tumors formed with gliomalike lesions presumably due to increasing MDM2 tissue-specific promoters in transgenic mice include: 1) levels that lead to sequestration and secondary inhibition pineoblastomas (tryptophan hydroxylase promoter); 2) of p53 function.34 The PTEN/MMAC1 on chromosome primitive neuroectodermal tumors (tyrosine hydroxylase 10q, a dual specific and lipid phosphatase that is lost in the promoter); 3) oligodendrogliomas (myelin basic promoter majority of GBMs, has also been targeted in mice. The and S-100 promoter); 4) neuroblastoma and ganglioneu- PTEN null (-/-) mice are embryonically lethal, whereas romas (dopamine--hydroxylase promoter); and 5) go- heterozygous (+/-) mice develop a plethora of tumors but nadotrophic hormoneÐreleasing tumors leutinizing hor- not astrocytomas.58 Current studies in which we use a mone/follicleÐstimulating hormone promoter).53 variety of strategies, are underway to knockout PTEN/ MMAC1 specifically in astrocytes, to avoid the embryon- Transgenic Models of Astrocytomas ic lethality. These knockout models of astrocytoma-specific genetic alteration have not resulted in useful models The GFAP promoter has been used to express oncogenes specifically in astrocytes, such as the SV40 large T of astrocytomas; however, they may augment and com- 10 66 plement strategies in which the mouse genome has been antigen and v-src. The GFAPÐSV40 large T transgenic mice were shown to develop aggressive nonastrocytic altered for a gain of function in genetic aberration associ- 10 ated with astrocytomas, such as those used in our labora- brain tumors with hyperplasia of the choroid plexus. In tory with ES cells. GFAPÐv-src transgenic mice abnormal nests of proliferating astrocytes are formed by 2 weeks of age, and these Embryonic Stem CellÐMediated Transgenesis astrocytes later evolve into overt malignant astrocytomas in the brain and spinal cord in 14% of mice by 65 weeks In the last two decades, the advent of novel and very of age.66 Hemizygosity for p53 or retinoblastoma, efficient techniques to manipulate the mouse genome, achieved by crossing these mice with the respective p53 such as in mouse ES cells, has given us the capability of and retinoblastoma knockout mice, was not shown to in- tailoring the mouse genes and genomes at will. We have crease the incidence or shorten the latency of astrocytic reached the stage at which geneticists studying mice tumors in these GFAPÐv-src mice.40 These transgenic believe there is no limitation to creating phenocopies (or astrocytomas have histological features similar to human genocopies) of any mutations or chromosomal aberrations GBMs, including expression of VEGF and angiogenic identified in human diseases. Embryonic stem cells are endothelial-specific receptors such as flt-1, flk-1, tie-1 derived from an early (blastocyst-stage) embryo and can and tie-2.62,66 be maintained in culture as undifferentiated, pluripotent We have used the GFAP promoter to express relevant cells under the proper growth conditions. A broad spec- transgenes in astrocytes, by using the ES cell strategy out- trum of strategies has been designed to create genomic lined previously (Fig. 2). Based on our prior work in

Neurosurg. Focus / Volume 8 / April, 2000 5

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC H. Ding, et al.

astrocytomas. Derivative astrocytoma cells obtained in these mice are tumorgenic when inoculated into naïve syngeneic or immunocompromised mice. Cytogenetic analysis revealed consistent clonal aneuploidies of several chromosomal regions syntenic with comparable loci altered in human astrocytomas. For example, the transgenic mouse astrocytoma cells harbored trisomy of mouse chromosome 10, an area that contains human chromosome 12q, which is the second most commonly amplified region in human GBMs.50 Direct assay conducted to deter- mine protein expression revealed decreased expression of p16, p19/p14ARF, p53, and PTEN and overexpression of proteins such as EGFR, MDM2, and CDK4, which is similar to findings in human malignant astrocytomas. In addition, these GFAP-V12Ras transgenic mice have been shown to possess genetic cooperativity, as they develop malignant astrocytomas within 2 to 3 weeks when crossed with transgenic mice in which expression of EGFRvIII in astrocytes occurs. These mice were developed using a similar strategy with the GFAP promoter in ES cells but, by themselves, did not result in astrocytoma formation (unpublished results). Additional crosses with mice harboring knockout of relevant astrocytoma genes such as p53 and retinoblastoma are ongoing and may modulate the occurrence of the GBMs in these mice. Furthermore, the mice and derived astrocytoma lines are responsive to biological therapies targeting the Ras pathway, such as with farnesyl transferase inhibitors, which have shown promise in a variety of human tumors including, as found in our own studies, astrocytomas.16 The mol- ecularÐpathological similarities with human GBMs, in addition to some of these early, promising studies, leads us to believe that this transgenic mouse GBM model may serve to increase further our knowledge of the molecular pathogenesis, as well as serve as an appropriate preclini- Fig. 2. Schematic drawing illustrating ES cell transgenesis that we have used to create the transgenic mouse GBM model. The cal model of human malignant astrocytomas. transgene (for example, activated p21-ras) is transfected into ES cells under a GFAP promoter specifically expressed in astrocytes. The promoter is activated in ES cells (by retinoic acid) and using CONCLUSIONS selection markers positive clones expressing the transgene are selected. These transfected ES cells undergo aggregations and are The generation of spontaneously occurring mouse mod- transferred into pseudopregnant mice to create chimeric embryos els of astrocytomas, based on the molecular pathogenesis composed of cells arising from the transfected and wild-type ES of human astrocytomas, would greatly advance our abili- cells, with the transgene only being expressed in a tissue-specific ty to treat human astrocytomas by serving as informative manner, such as occurs in astrocytes in the presence of the GFAP preclinical models in which to test novel therapeutic promoter. These chimeras are then crossed with normal mice to agents. The transgenic models with multiple alterations in propagate stably the transgene in a germline fashion. This and growth factor and cell cycle regulatory pathways can be other transgenic strategies have led to derivation of mouse astrocy- used to advance our understanding of the molecular toma models, which replicate the pathological and molecular pro- pathogenesis of astrocytomas. Biological therapies, such file found in human astrocytomas. as inhibitors of receptor tyrosine kinases, inhibitors of signaling pathways such as Ras, or antiangiogenesis strategies, are already being conducted in early clinical trials. which we found that one of the main signaling pathways The transgenic models mimicking the histology and mol- activated by PDGFRs and EGFRs overexpressed in ecular pathogenesis should markedly enhance our ability human GBMs involves activation of Ras,14,15,22 we have to test these and other novel agents, with an increased like- created a mouse GBM model by overexpressing onco- lihood of therapeutic success in subsequent clinical trials. genic Ras (V12Ras) in astrocytes using the GFAP promoter (unpublished data). Ninety percent of GFAP-V12Ras References transgenic mice developed GFAP-positive mutifocal 1. Arap W, Nishikawa R, Furnari FB, et al: Replacement of the astrocytomas within 2 to 4 months. These tumors demon- p16/CDKN2 gene suppresses human glioma cell growth. Can- strated a high mitotic index, nuclear pleomorphism, infil- cer Res 55:1351Ð1354, 1995 tration, and increased vascularity with overexpression of 2. Benda P, Lightbody J, Sato G, et al: Differentiated rat glial cell VEGF, similar to characteristics of human malignant strain in tissue culture. Science 161:370Ð371, 1968

6 Neurosurg. Focus / Volume 8 / April, 2000

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC A review of astrocytoma models

3. Beutler AS, Banck MS, Wedekind D, et al: Tumor gene thera- 2: loss of merlin's protective spell. Curr Opin Genet Dev 6: py made easy: allogeneic major histocompatibility complex in 87Ð92, 1996 the C6 rat glioma model. Hum Gene Ther 10:95Ð101, 1999 25. Gutmann DH, Giordano MJ, Mahadeo DK, et al: Increased neu- 4. Bogler O, Huang HJ, Cavenee WK: Loss of wild-type p53 rofibromatosis 1 gene expression in astrocytic tumors: positive bestows a growth advantage on primary cortical astrocytes regulation by p21-ras. Oncogene 12:2121Ð2127, 1996 and facilitates their in vitro transformation. Cancer Res 55: 26. Gutmann DH, Loehr A, Zhang Y, et al: Haploinsufficiency for 2746Ð2751, 1995 the neurofibromatosis 1 (NF1) tumor suppressor results in in- 5. Bogler O, Huang HJ, Kleihues P, et al: The p53 gene and its role creased astrocyte proliferation. Oncogene 18:4450Ð4459, 1999 in human brain tumors. Glia 15:308Ð427, 1995 27. Gutmann DH, Saporito-Irwin S, DeClue JE, et al: Alterations in 6. Bogler O, Nagane M, Gillis J, et al: Malignant transformation the rap1 signaling pathway are common in human gliomas. On- of p53-deficient astrocytes is modulated by environmental cues cogene 15:1611Ð1616, 1997 in vitro. Cell Growth Differ 10:73Ð86, 1999 28. Hegi ME, zur Hausen A, Ruedi D, et al: Hemizygous or homo- 7. Brickell PM: The p60c-src family of protein-tyrosine ki- zygous deletion of the chromosomal region containing the nases: structure, regulation, and function. Crit Rev Oncog 3: p16INK4a gene is associated with amplification of the EGF re- 401Ð446, 1992 ceptor gene in glioblastomas. Int J Cancer 73:57Ð63, 1997 8. Cichowski K, Shih TS, Schmitt E, et al: Mouse models of tu- 29. Holland EC, Hively WP, DePinho RA, et al: A constitutively mor development in neurofibromatosis type 1. Science 286: active epidermal growth factor receptor cooperates with disrup- 2172Ð2176, 1999 tion of G1 cell-cycle arrest pathways to induce glioma-like le- 9. Clarke AR, Maandag ER, van Roon M, et al: Requirement for sions in mice. Genes Dev 12:3675Ð3685, 1998 a functional Rb-1 gene in murine development. Nature 359: 30. Holland EC, Hively WP, Gallo V, et al: Modeling mutations in 328Ð330, 1992 the G1 arrest pathway in human gliomas: overexpression of 10. Danks RA, Orian JM, Gonzales MF, et al: Transformation of CDK4 but not loss of INK4a-ARF induces hyperploidy in cul- astrocytes in transgenic mice expressing SV40 T antigen under tured mouse astrocytes. Genes Dev 12:3644Ð3649, 1998 the transcriptional control of the glial fibrillary acidic protein 31. Jacks T, Fazeli A, Schmitt EM, et al: Effects of an Rb mutation promoter. Cancer Res 55:4302Ð4310, 1995 in the mouse. Nature 359:295Ð300, 1992 11. Daumas-Duport C, Scheithauer B, et al: Grading of astro- 32. Jacks T, Shih TS, Schmitt EM, et al: Tumour predisposition in cytomas. A simple and reproducible method. Cancer 62: mice heterozygous for a targeted mutation in Nf1. Nat Genet 2152Ð2165, 1988 7:353Ð361, 1994 12. Donehower LA, Harvey M, Slagle BL, et al: Mice deficient for 33. Jen J, Harper JW, Bigner SH, et al: Deletion of p16 and p15 p53 are developmentally normal but susceptible to spontaneous genes in brain tumors. Cancer Res 54:6353Ð6358, 1994 tumours. Nature 356:215Ð221, 1992 34. Kamijo T, Bodner S, van de Kamp E, et al: Tumor spectrum in 13. Feldkamp MM, Gutmann DH, Guha A: Neurofibromatosis type ARF-deficient mice. Cancer Res 59:2217Ð2222, 1999 1: piecing the puzzle together. Can J Neurol Sci 25:181Ð191, 35. Kleihues P, Burger PC, Scheithauer BW: The new WHO clas- 1998 sification of brain tumours. Brain Pathol 3:255Ð268, 1993 14. Feldkamp MM, Lala P, Lau N, et al: Expression of activated 36. Li J, Yen C, Liaw D, et al: PTEN, a putative protein tyrosine epidermal growth factor receptors, Ras-guanosine triphosphate, phosphatase gene mutated in human brain, breast, and prostate and mitogen-activated protein kinase in human glioblastoma cancer. Science 275:1943Ð1947, 1997 multiforme specimens. Neurosurgery 45:1442Ð1453, 1999 37. Libermann TA, Nusbaum HR, Razon N, et al: Amplification, 15. Feldkamp MM, Lau N, Guha A: Growth inhibition of astrocy- enhanced expression and possible rearrangement of EGF recep- toma cells by farnesyl transferase inhibitors is mediated by a tor gene in primary human brain tumours of glial origin. Na- combination of anti-proliferative, pro- apoptotic and anti-angio- ture 313:144Ð147, 1985 genic effects. Oncogene 18:7514Ð7526, 1999 38. Lobe CG, Nagy A: Conditional genome alteration in mice. Bio- 16. Feldkamp MM, Lau N, Guha A: Signal transduction pathways essays 20:200Ð208, 1998 and their relevance in human astrocytomas. J Neurooncol 35: 39. Louis DN, Gusella JF: A tiger behind many doors: multiple ge- 223Ð248, 1997 netic pathways to malignant glioma. Trends Genet 11: 17. Fraser H: Astrocytomas in an inbred mouse strain. J Pathol 412Ð415, 1995 103:266Ð270, 1971 40. Maddalena AS, Hainfellner JA, Hegi ME, et al: No comple- 18. Fults D, Petronio J, Noblett BD, et al: Chromosome 11p15 dele- mentation between TP53 or RB-1 and v-src in astrocytomas of tions in human malignant astrocytomas and primitive neuroec- GFAP-v-src transgenic mice. Brain Pathol 9:627Ð637, 1999 todermal tumors. Genomics 14:799Ð801, 1992 41. Maheshwar MM, Cheadle JP, Jones AC, et al: The GAP-relat- 19. Giovannini M, Robanus-Maandag E, Niwa-Kawakita M, et al: ed domain of tuberin, the product of the TSC2 gene, is a target Schwann cell hyperplasia and tumors in transgenic mice ex- for missense mutations in tuberous sclerosis. Hum Mol Genet pressing a naturally occurring mutant NF2 protein. Genes Dev 6:1991Ð1996, 1997 13:978Ð986, 1999 42. Malkin D, Li FP, Strong LC, et al: Germ line p53 mutations in 20. Guha A: Platelet-derived growth factor: a general review with a familial syndrome of breast cancer, sarcomas, and other neo- emphasis on astrocytomas. Pediatr Neurosurg 17:14Ð20, plasms. Science 250:1233Ð1238, 1990 1991 43. McClatchey AI, Saotome I, Mercer K, et al: Mice heterozygous 21. Guha A, Dashner K, Black PM, et al: Expression of PDGF and for a mutation at the Nf2 tumor suppressor locus develop a PDGF receptors in human astrocytoma operation specimens range of highly metastatic tumors. Genes Dev 12:1121Ð1133, supports the existence of an autocrine loop. Int J Cancer 60: 1998 168Ð173, 1995 44. McClatchey AI, Saotome I, Ramesh V, et al: The Nf2 tumor 22. Guha A, Feldkamp MM, Lau N, et al: Proliferation of human suppressor gene product is essential for extraembryonic devel- malignant astrocytomas is dependent on Ras activation. Onco- opment immediately prior to gastrulation. Genes Dev 11: gene 15:2755Ð2765, 1997 1253Ð1265, 1997 23. Guha A, Glowacka D, Carroll R, et al: Expression of platelet 45. Mercer WE, Shields MT, Amin M, et al: Negative growth reg- derived growth factor and platelet derived growth factor recep- ulation in a glioblastoma tumor cell line that conditionally ex- tor mRNA in a glioblastoma from a patient with Li- Fraumeni presses human wild-type p53. Proc Natl Acad Sci USA syndrome. J Neurol Neurosurg Psychiatry 58:711Ð714, 1995 87:6166Ð6170, 1990 24. Gusella JF, Ramesh V, MacCollin M, et al: Neurofibromatosis 46. Mollenhauer J, Wiemann S, Scheurlen W, et al: DMBT1, a new

Neurosurg. Focus / Volume 8 / April, 2000 7

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC H. Ding, et al.

member of the SRCR superfamily, on chromosome 10q25.3- 60. Stott FJ, Bates S, James MC, et al: The alternative product from 26.1 is deleted in malignant brain tumours. Nat Genet 17: the human CDKN2A locus, p14(ARF), participates in a regula- 32Ð39, 1997 tory feedback loop with p53 and MDM2. EMBO J 17: 47. Nagy A, Rossant J, Nagy R, et al: Derivation of completely cell 5001Ð5014, 1998 culture-derived mice from early-passage embryonic stem cells. 61. Teng DH, Hu R, Lin H, Davis T, et al: MMAC1/PTEN muta- Proc Natl Acad Sci USA 90:8424Ð8428, 1993 tions in primary tumor specimens and tumor cell lines. Cancer 48. Nishikawa R, Furnari FB, Lin H, et al: Loss of P16INK4 ex- Res 57:5221Ð5225, 1997 pression is frequent in high grade gliomas. Cancer Res 55: 62. Theurillat JP, Hainfellner J, Maddalena A, et al: Early induction 1941Ð1945, 1995 of angiogenetic signals in gliomas of GFAP-v-src transgenic 49. Pirity M, Hadjantonakis AK, Nagy A: Embryonic stem cells, mice. Am J Pathol 154:581Ð590, 1999 creating transgenic animals. Methods Cell Biol 57:279Ð293, 63. Tsuzuki T, Tsunoda S, Sakaki T, et al: Alterations of retinoblas- 1998 toma, p53, p16(CDKN2), and p15 genes in human astrocy- 50. Rasheed BK, Wiltshire RN, Bigner SH, et al: Molecular patho- tomas. Cancer 78:287Ð293, 1996 genesis of malignant gliomas. Curr Opin Oncol 11:162Ð167, 64. Wang SI, Puc J, Li J, et al: Somatic mutations of PTEN in 1999 glioblastoma multiforme. Cancer Res 57:4183Ð4186, 1997 51. Reifenberger G, Reifenberger J, Ichimura K, et al: Ampli- 65. Watanabe K, Peraud A, Gratas C, et al: p53 and PTEN gene fication of multiple genes from chromosomal region 12q13-14 mutations in gemistocytic astrocytomas. Acta Neuropathol in human malignant gliomas: preliminary mapping of the am- 95:559Ð564, 1998 plicons shows preferential involvement of CDK4, SAS, and 66. Weissenberger J, Steinbach JP, Malin G, et al: Development and malignant progression of astrocytomas in GFAP-v-src MDM2. Cancer Res 54:4299Ð4303, 1994 transgenic mice. Oncogene 14:2005Ð2013, 1997 52. Rossant J, Nagy A: Genome engineering: the new mouse genet- 67. Wienecke R, Guha A, Maize JC Jr, et al: Reduced TSC2 RNA ics. Nat Med 1:592Ð594, 1995 and protein in sporadic astrocytomas and ependymomas. Ann 53. Rovigatti U, Afanasyeva T, Brandner S, et al: Transgenic mice Neurol 42:230Ð235, 1997 as research tools in neurocarcinogenesis. J Neurovirol 4: 68. Wong AJ, Ruppert JM, Bigner SH, et al: Structural alterations 159Ð174, 1998 of the epidermal growth factor receptor gene in human gliomas. 54. Serrano M, Lee H, Chin L, et al: Role of the INK4a locus in Proc Natl Acad Sci USA 89:2965Ð2969, 1992 tumor suppression and cell mortality. Cell 85:27Ð37, 1996 69. Yahanda AM, Bruner JM, et al: Astrocytes derived from p53- 55. Short MP, Richardson EP Jr, Haines JL, et al: Clinical, neuro- deficient mice provide a multistep in vitro model for develop- pathological and genetic aspects of the tuberous sclerosis com- ment of malignant gliomas. Mol Cell Biol 15:4249Ð4259, 1995 plex. Brain Pathol 5:173Ð179, 1995 56. Sidransky D, Mikkelsen T, Schwechheimer K, et al: Clonal Manuscript received February 16, 2000. expansion of p53 mutant cells is associated with brain tumour Accepted in final form March 21, 2000. progression. Nature 355:846Ð847, 1992 This work was supported in part by a Medical Research Council 57. Srivenugopal KS, Ali-Osman F: Deletions and rearrangements of Canada Clinician Scientist Award (grant No. 9509-CLN), an inactivate the p16INK4 gene in human glioma cells. Oncogene operating grant from National Cancer Institute of Canada No. 7178 12:2029Ð2034, 1996 (to A.G.), National Institute of Neurological Disorders and Stroke 58. Stambolic V, Mak TW, Woodgett JR: Modulation of cellular grant No. NS36996 (to D.H.G.), National Cancer Institute of Can- apoptotic potential: contributions to oncogenesis. Oncogene ada grant No. 7309 (to A.N.), and a fellowship from the Heart 18:6094Ð6103, 1999 Stroke Foundation of Canada (to H.D). 59. Steck PA, Pershouse MA, Jasser SA, et al: Identification of a Address reprint requests to: Abhijit Guha, M.D., Division of candidate tumour suppressor gene, MMAC1, at chromosome Neurosurgery, The Toronto Hospital Western Division-UHN, 399 10q23.3 that is mutated in multiple advanced cancers. Nat Ge- Bathurst Street, Toronto, M5T 2S8 Ontario, Canada. email: ab.guha net 15:356Ð362, 1997 @utoronto.ca.

8 Neurosurg. Focus / Volume 8 / April, 2000

Unauthenticated | Downloaded 09/27/21 07:54 AM UTC