We develop and use in-vivo models for neural cancers to: 1) Identify genetic events that promote tumorigenesis. 2) Study cancer stem and progenitor cells. 3) Evaluate new targets, chemical genetic approaches, and mechanistic rationales for combining targeted agents.
Stem cell biology, genetics, and developmental therapeutics in glioma. Aberrant EGFR signaling features prominently in glioma, the most common primary adult brain tumor. We generated a mouse model for glioma by over-expressing EGFR under the S100 beta promoter (Weiss, 2003). Expression of oncogenes in rare cancer-stem-like cells in the subventricular zone led to differentiation block and aberrant glial differentiation, resulting in astrocytoma (Persson, In revision). In contrast, murine oligodendrogliomas arose from abundant oligodendroglial progenitors in white matter. We described a progenitor origin for this more favorable form of glioma, demonstrating that a progenitor rather than a stem-cell origin underlies the improved outcome in patients (Persson, 2010). We were among the first to describe oncogene addiction driven by activated EGFR (Fan, 2002). We described and characterized dual inhibitors of PI3K and mTOR, demonstrating that these drugs blocked mTOR inhibitor-driven activation of Akt, that EGFR signaling to Akt was dispensable for arrest, that EGFR signaling to Protein Kinase C alpha was central to the ability of PI3K to signal to mTOR, and that blockade of PI3K, mTOR and autophagy converted cytostatic PI3K/mTOR inhibitors into cytotoxic agents (Fan, 2006-2010). RapaLink-1, an mTORC1 targeted mTOR kinase inhibitor was well-tolerated and showed superior efficacy, as compared to either parental drug (Fan, 2017). Activated alleles of EGFR occur in brain and lung-cancers, yet EGFR inhibitors benefit only lung cancer. We traced this differential response to lower occupancy rates of EGFR inhibitors in brain as compared to lung cancer mutants (Barkovich 2012). EGFR is frequently co-amplified with EGFRvIII. We showed co-expression of EGFR and vIII in individual cells in human tumors, demonstrated that vIII was a substrate for EGFR, and that co-expression drove therapeutic resistance through activating STAT signaling (Fan et al, 2013). We showed that EGFR and EGFRvIII cooperate to shape the tumor microenvironment in glioblastoma, promoting recruitment of immunosuppressive tumor-associated macrophages (An et al, 2018).
Targeted expression of MYCN generates in-vivo models of neuroblastoma and medulloblastoma. Neuroblastoma is the third most common tumor of childhood. The proto-oncogene MYCN is amplified in ~50% of high-risk incurable neuroblastoma. We generated transgenic mice that mis-expressed MYCN in neural crest, that developed neuroblastoma, and that remain the standard GEM model used by the community (Weiss, 1997). Genome-wide screens revealed genetic alignment with human tumors (Weiss, 2002, Hackett, 2003). Inhibitors of PI3 and mTOR kinases blocked MYCN in-vivo, disrupted MYCN-directed signaling between tumor and vascular cells, and led to angiogenic collapse (Chantery, 2012). We identified altered neurotransmitter signaling through GABA as contributing to human and murine neuroblastoma, and described the alternative splicing landscape (Hackett, 2014; Chen, 2015). Murine neuroblastoma tumors mutant at p53 modeled relapsed, drug-resistant neuroblastoma (Chesler, 2006-8). MYCN blockade reduced VEGF signaling, promoting vascular collapse (Chanthery, 2012). We synthesized and solved the co-crystal structure of a new class of MYC/MYCN-degrading drugs that block a kinase-independent MYC/MYCN stabilizing function of Aurora Kinase, potently degrading MYCN in-vivo (Gustafson, Meyerowitz, 2014). We developed a non-germline GEM model for neuroblastoma by transducing N-myc into the mouse neural crest. This model is fully penetrant in B6 mice, the work-horse strain for immunology. We are developing this as-yet unpublished model to understand how neuroblastoma tumors so effectively suppress the local immune system, and to develop immuno-oncology approaches.
MYCN is mis-expressed in the majority of medulloblastoma tumors. We used the Tet system to regulate MYCN expression and to image tumor-associated firefly luciferase expression in-vivo. Targeted expression of MYCN to the brains of transgenic mice led to luciferase and MYCN-positive medulloblastoma, (Swartling, 2010). We also transduced MYCN into murine neural stem cells, separately cultured from prenatal or postnatal mice, with cells from hindbrain generating medulloblastoma, and from forebrain generating glioma. Orthotopic transduction of prenatal cerebellar stem cells drove SHH-dependent, while both prenatal brainstem and postnatal cerebellar stem cells drove SHH-independent disease (Swartling 2012). Thus, distinct neural stem cell populations generated disparate brain tumors in response to MYCN.
Genome-wide sequencing efforts have generally failed to identify new driver mutations for the majority of high-risk neuroblastoma and medulloblastoma. In contrast, copy number analyses have identified recurrent regions of variation. Regions of gain or loss on any human chromosome correspond to multiple different chromosomes in the mouse, which is challenging to model. Thus, we are incorporating known driver mutations into engineered human induced pluripotent stem cells, have generated non-germline humanized mouse models for neuroblastoma (in progress) and medulloblastoma (Huang et al, Cell-Stem-Cell 2019), and are developing comparable models for glioma. These human based xenograft models represent a genetic platform to test whether copy number variation can drive cancers, and to develop therapies.