My laboratory research focuses on diffuse gliomas, a diverse group of primary brain tumors. The most malignant of these, glioblastoma (GBM), is currently diagnosed by microscopic morphology and treated empirically with concurrent fractionated external beam radiation (XRT) and the DNA alkylating agent temozolomide (TMZ) to yield a median overall survival of 12-14 months. Neither diagnosis nor therapy are based upon the underlying molecular alterations responsible for gliomagenesis, the roles of which are becoming increasingly defined using genetically-engineered mouse (GEM) models. The main goal of our work is to establish a direct link between preclinical drug development in glioma GEM and the rational design of human clinical trials involving patients with molecular subtypes of tumor via comparative molecular analyses. To facilitate this work, we have established collaborations with several UNC and non-UNC investigators, including Gary Johnson and Fernando Pardo-Manuel de Villena at UNC, as well as Mike Berens and Jason Huse at TGen and MD Anderson.
GEM are valuable resources for investigation of the genetic basis of neoplasia. However, these models are not ideal for preclinical drug development. We have developed a series of GEM orthotopic allograft model of gliomas that not only recapitulates the growth pattern of human tumors in vivo but also represents a genetically-tractable model system for drug development. Because clinical trials involve patients with recurrent tumors that have failed standard treatments, identification of molecular correlates of therapy resistance in glioma GEMM will facilitate preclinical drug development with these models and inform future molecular marker-based clinical trial design. We hypothesize that therapy-induced molecular changes in allograft GEM models are similar to recurrent human GBM. To investigate this hypothesis, we are harvesting tumors terminally and systematically during experimental therapies and analyzing them for genomic and proteomic (kinome profiling) techniques to define therapy-specific molecular effects. Genomics data from untreated allografts will be compared with those from its conditional, inducible counterpart to validate the biological fidelity of this model system. Therapy-induced changes in allografts will be compared to those in recurrent human GBM to identify potential markers of therapy resistance.
The relationship between neuroglial ontogeny and gliomagenesis remains unclear. We have shown that GFAP+ astrocytes are susceptible to genetically-induced gliomagenesis in over ten inducible GEM models using CreER and conditional oncogenic alleles to mutate the core intracellular signaling pathways altered in human gliomas: the G1/S cell cycle, receptor tyrosine kinase/mitogen-activated protein kinase/phosphatidylinositol-3-kinase (RTK/MAPK/PI3K) pathways. Three of these models develop low-grade astrocytomas that rapidly progress to large, lethal GBM, suggesting that secondary genetic events develop stochastically. Similar GEM models of GBM targeting Nestin+ neural stem cells (NSC) have recently been described. However, comparative GEM modeling and genomic studies targeting different brain cells with identical genetic lesions are lacking. We hypothesize that cell-of-origin dictates human subtype-specificity of GBM in GEM. To address this question, we are currently examining the genotype-specific effects on tumor signatures of GEM GBM. We utilize genomics techniques to identify secondary mutations that occur during tumor development and to define their effects on molecular tumor subtypes at the mRNA, DNA, epigenetic, and kinome levels. Finally, although conditional, inducible GEMM are designed based on the molecular abnormalities present in human GBM, the extent to which they recapitulate human GBM molecular biology has yet to be established. Therefore, we are performing comparative genomics between mouse and human GBM to define the effects of cell-of-origin and genetics on human GBM subtypes. Such data will permit future drug/biomarker development studies for specific subtypes of human GBM.
Precision medicine promises to revolutionize oncology by tailoring treatments to specific somatic mutations within a patient’s tumor. Yet this approach fails to account for the effect of host genetics (germline variations) on tumor evolution and treatment response. It also fails to account for possible genetic interactions between host and tumor. GBM is a genomically diverse disease with fatal outcomes and few effective treatments. Despite the fact that GBM was the first tumor to be characterized by The Cancer Genome Atlas (TCGA), and several genome-wide association studies have linked specific polymorphisms to disease susceptibility, little is known about the impact of host genetics on its biology or treatment response. We have established collaborations with Fernando Pardo-Manuel de Villena (UNC) and Mike Berens (TGen) with expertise in population genetics of inbred and outbred mouse strains, specifically the Collaborative Cross (CC), and human GBM biology and preclinical models, respectively, to complement our experience with GEM models of GBM, to address whether germline variations contribute to differences in GBM evolution or response to therapy.
We are doing so using a unique experimental approach that utilizes non-germline GEM (nGEM) models. Germline GEM tumors are driven by predefined mutation(s) in specific cell types in their native environment. GEM are essential tools for functional validation of GBM genes, such as NF1 and TP53, but their use has been limited by variable tumor penetrance/latency and lack of cell culture counterparts. To overcome these limitations, the Miller lab developed nGEM models that target specific mutations to predefined brain cells implicated in the origin of GBM, including Nf1;Trp53 deletion mutations in oligodendrocyte progenitor cells (OPC). Transplanting cultured nGEM cells into the brains of syngeneic mice mimics the human disease. Using this nGEM model in the context of the CC will permit us to experimentally define host genes that influence GBM evolution and treatment response. We will use this system to test the hypothesis that genetic polymorphisms affecting expression of ligand-receptor(s) that mediate paracrine effects within the tumor microenvironment impact GBM evolution and response to therapy.
Precision medicine also fails to account for the dynamic state of tumor kinomes – the repertoire of expressed kinases. Indeed, researchers still view kinome circuits as static and remain focused on the small subset of kinases that are mutated in cancer. “Driver mutations” in kinases such as BRAF (melanoma), ERBB2 (breast cancer), and BCR-ABL (leukemia) have justified this narrow focus, but the problem is that many potentially important kinases for drug discovery remain understudied. This continues, in part, because of a lack of understanding of the entire kinome and appropriate methods to study its dynamics. Our collaborator, Gary Johnson (UNC), developed a novel, unbiased proteomics technique - multiplex inhibitor beads-mass spectrometry (MIB-MS) - to examine dynamic, drug-induced changes in the activation state of the kinome en masse. This includes “understudied kinases” that lack selective inhibitors, antibody reagents, characterized networks & cellular functions, and defined disease relevance. Thus, a major question in cancer biology remains which understudied kinases in the kinome “dark matter” are critical signaling nodes where targeted drug modulation would elicit clinical responses.
Through our collaborations with Mike Berens (TGen) and Jason Huse (MD Anderson), we have combined our expertise in glioma GEM models to incorporate patient-derived xenograft (PDX) models into our experimental armamentarium. PDX accurately recapitulate the genomic heterogeneity and pathological features of GBM and thus represent the most biologically faithful models of human GBM to date. Through the Ivy Clinical Trial Consortium, Dr. Berens developed and genetically characterized PDX from a “basket” Feasibility Trial where recurrent GBM patients are prospectively recommended targeted therapies based on genome profiling. These PDX harbor 2 groups of mutations:
Group 1: Validated mutations in the G1/S checkpoint (CDKN2A) and RTK/MAPK/PI3K (EGFR, NF1, PTEN) core GBM pathways.
Group 2: More recently described mutations in ATRX and IDH1, in combination with TP53.
Whereas PDX are limited by requirement for immunodeficient hosts, GEM model tumors are driven by predefined mutation(s) in specific cell types in their native environment. GEM are essential for functional validation of GBM target genes, but their use in preclinical drug development has been limited by variable penetrance/latency and lack of cell culture counterparts. Moreover, the typical gene-centric modelling approach has largely ignored the impact of cellular origin on disease pathogenesis. To overcome these limitations, we developed non-germline GEM (nGEM) models that target specific mutations to predefined cells implicated in the origin of GBM subtypes, including neural stem cells (NSC), astrocytes (AC), and oligodendrocyte progenitors (OPC). Transplanting these cultured cells into syngeneic hosts mimics the human disease. We are thus developing nGEM models with the same Gr1/Gr2 driver mutations as our PDX models to:
1. credential PDX models against human GBM by MIB-MS kinome proteomics
2. develop nGEM models from distinct cells of origin that are genetically-matched to specific PDX
3. credential PDX and nGEM models by high-throughput drug screening and monitoring of the dynamic transcriptome and kinome response.
Our work will help realize the promise of precision medicine in neuro-oncology. Combining PDX and nGEM models and credentialing both with comprehensive molecular analyses will help elucidate the role of mutations and cellular origin in gliomas and the response of their mutated or aberrant signaling circuits to unique combinations of targeted kinase inhibitors. Genomically annotated, syngeneic nGEM models will be useful for future preclinical development of drugs targeting the tumor microenvironment and intact immune system.
