Funded Grants

Large-scale screening for therapeutic targets in a mouse model of glioma.

Gliomas are the most frequently occurring primary malignancies in the central nervous system (CNS). They are a heterogeneous group of tumors that display histologic similarities to glia, which include astrocytes and oligodendrocytes; astrocytomas account for the majority of these tumors [1, 2]. The most malignant form, glioblastoma multiforme (GBM), is one of the most aggressive and lethal forms of cancer, with a median survival of about one year [3]. GBMs can present as one of two distinct subtypes. Primary or “de novo” GBMs arise without any prior clinical or histological evidence of a lower grade precursor lesion and more commonly affect older patients (mean age of 62 years). Secondary or “progressive” GBMs progress from a lower grade glioma and typically develop in younger patients (median age of 45 years) [4]. In general, GBMs exhibit unique properties compared to most other solid tumors. For example, gliomas are highly infiltrative from the outset, thus low-grade (secondary) gliomas disperse throughout the brain. In addition, mutations in the TP53 tumor suppressor gene are frequent in low-grade tumors. This feature departs from many other cancers where TP53 mutations tend to appear later in tumor development during the progression phase [5, 6]. Most significantly, gliomas are resistant to conventional anticancer therapies including radiation and chemotherapy. The consequence is that there has been a stark absence of improvement in prognosis over the last thirty years [7]. Over the last two decades, advances in cancer-related molecular analyses has allowed probing of human brain tumor DNA for presence of mutations in genes known to contribute to many types of cancers. Not surprisingly, mutations have been identified in genes involved in cell cycle and apoptosis regulation (INK4A, CDK4, RB, TP53), and growth factor receptor signaling (EGF, PDGF, PTEN) [8, 9]. More advanced analysis through The Cancer Genome Atlas (TCGA) project, which undertook a comprehensive genetic analysis of human GBM, has provided confirmatory evidence of such mutations and revealed a hierarchy of mutations in terms of frequency [10]. Five cancer-related genes were reported to be most frequently mutated in GBM: p53; Pten; EGFR; Rb; and NF1. These results strongly implicate but do not prove that these genes are causal in brain tumor development.

Among the failed efforts to develop anti-glioma therapies are cell lines established from human tumors. Although a partial advance, these cell lines suffer from many drawbacks. First and foremost, the actual establishment in culture is a selective and inefficient process in which only a few tumor cells survive that then give rise to the cell line. It is now apparent that this process imposes dramatic alterations in the established cells that make them poor representatives of the original tumor from which they were derived. Indeed when reintroduced into the brains of mice (xenografts), these cells form tumors that do not resemble the original tumors and furthermore, these transplanted tumors often respond to therapies that have no effect on patients. In summary, conventional approaches toward deriving therapies including the use of established glioma cell lines and transplantation of xenografts (human glioma cells transplanted into mice) appear to be of limited value. An alternative approach that has gained increasing attention is the development of genetic mouse models that develop glioma. Physiologically relevant genetic mouse models of high-grade astrocytomas provide one approach to study crucial questions in tumor biology; the rationale for using mouse models is that they allow us to probe and investigate the mechanisms of cancer initiation and progression in ways that cannot be done in humans. In the clinic, doctors typically only see the patient once a tumor is significantly advanced. The mechanism by which the tumor initiates and progresses, and in which cell type this occurs has been, for the most part, a black hole. Understanding the natural history of tumor formation could lead to important insights that will have certain impact on prevention and cure. To this end, genetic lesions are engineered in the mouse to generate animal models that in one fashion or another replicate the human malignancy and, thus, allow for in vivo investigation of tumor initiation and development. Many of these mouse models involve introduction of oncogenic mutations in the germline or specific cell subpopulations in the brain [8, 11]. Unfortunately, most of these animal models of glioma suffer from two limitations. First, they often rely on oncogenic activation of genes that are either not clinical features of human glioma, or which target multiple cellular signaling pathways thus complicating analysis. Second, the strategies employed in generating these tumors do not permit identification of the target cells that give rise to the tumors. This latter aspect therefore impedes rigorous determination of precursor/product relationships in cancer formation and thus clinical validation. That is: are the cells giving rise to the tumor in the mouse model the same ones that do so in human glioma?

To overcome these issues, our laboratory has focused on generating genetic mouse models of glioma that resemble the human condition. To do so, we genetically engineered mice to lose one of two brain copies of either two or three of the most frequently mutated tumor suppressor genes in glioma: NF1, p53 and Pten [10] (TCGA GBM Disease Working Group; http://cancergenome.nih.gov/dataportal/). These mouse models developed glioma in all cases and that by all criteria currently used by neuropathologists who routinely diagnose brain tumors, the mouse tumors were essentially indistinguishable from the human tumors [12]. In one mouse model (Mut3) where Nf1 and p53 were mutated, tumors initiated as low-grade tumors and progressed through the WHO-defined criteria of Grades II, III, and IV (GBM); thus these mice represent a progressive model of secondary glioma. A second mouse model that included mutation of the Pten gene along with Nf1 and p53, also developed astrocytomas with 100% penetrance [13] however the tumors that developed were de novo, high-grade tumors. These data provided experimental evidence that Pten tumor suppressor loss is not contributing to tumor initiation but instead is critical for tumor progression to high grade GBM. Another critical question regarding gliomas is how they originate. Patients present with glioma essentially throughout the brain but where the tumors arose is not known. Our mouse models provide a unique opportunity to study the natural history of tumor development. We can sacrifice mice well before they show illness and ask where we first find evidence of abnormality. Such studies led us to observe that in most cases the first evidence of deregulated cell division occurred in the subventricular zone (SVZ) region of the brain where adult stem cells reside. We therefore proposed that glioma might be a stem/progenitor cell disease and in a recently published study [14] we provide the first conclusive evidence that in mouse models, only SVZ derived stem/progenitor cells can give rise to glioma. Mutation of these genes outside this region failed to yield tumors [14]. We have further shown that we can culture the tumor cells in a very efficient process and these cultured cells, when reimplanted into mouse brains, give rise to tumors that are indistinguishable from the original tumors. These tumor-derived neurospheres (TD-NFCs) therefore mimic the original tumors and hold the promise of retaining tumor properties that will hopefully make them valuable substrates for discovering tumor killing agents [13, 14]. We can show that TD-NFCs exhibit abnormal growth and differentiation properties and we plan to harness these differences to seek out compounds or gene pathways that force the TD-NFCs to either cease dividing or undergo cell death. We have also discovered tumor precursor cells that do not yet have the full complement of tumor-causing mutations. This is an important finding because it uncovers an intermediate step toward tumor development and therefore a potential means to uncover processes that serve as a prelude to tumor formation. We believe that these pretumorigenic cells may be extremely important to study as an intermediate step toward the full-blown tumor phenotype.

In summary, we have generated uniquely powerful mouse models of human glioma that develop tumors with 100% penetrance. In addition to the remarkable pathological and molecular similarities of these mouse tumors to human glioma, we have identified the neural stem/progenitor cells as the requisite source of these tumors and moreover have demonstrated our capacity to both culture and transplant primary cells from the tumors [14]. Further, our discovery that the source of these tumors is the stem/progenitor lineage provides us with precise precursor/intermediate/product sources of cells for molecular comparisons, cell-based screens, and cellular validation. This strikingly physiologically relevant mouse model of idiopathic glioma puts us in a unique position to uncover important features that may lead to new therapies, in addition to providing us with a genetic system to both screen and preclinically validate any novel opportunities for therapy. This grant application describes our experimental plans towards this end; we will exploit the TD-NFCs from our mouse glioma models for high throughput screens to identify pathways and molecules that are fundamental for tumor initiation and progression and to seek strategies to impede this process in mice with a view toward clinical applications.