The main goal of my laboratory is the use of genetically-accurate mouse models of glioblastoma to understand the molecular basis for the genesis of these tumors and their response to standard therapy. In the process, we have developed preclinical trial infrastructure and imaging that supports the development of novel therapeutic approaches. My laboratory developed the RCAS/tv-a system of post-natal somatic cell-type specific gene transfer to study cancer formation in mice, and used this system to model the formation of gliomas and medulloblastomas. We demonstrated that stem cells are more sensitive to transforming events than differentiated cells, that Akt activity is elevated in human glioblastomas, and that deletion of PTEN (as occurs in human glioblastoma) in mice was causal in glioma formation and progression. Brain tumor cells that are resistant to radiation therapy occupy the perivascular niche and have stem-cell characteristics driven by a combination of Akt and notch activities. We have shown that nitric oxide produced by endothelial cells promote stem-cell characteristics in perivascular cells through cGMP, PKG and Notch signaling. We have further characterized therapeutic DNA Damage Response pathway in gliomas and compared this data with that from human GBM. Finally, one recent breakthrough in the field is the molecular subdivision of the GBMs. Our contribution to this was demonstrating proteomic evidence that specific signaling pathway activity characterizes these subgroups and that the mouse models were specific mimics of the molecular GBM subgroups. The human and mouse glioma data has led to molecularly-stratified clinical trials for GBM patients.
We have been using our immunocompetent models of gliomas to better understand the immunologic response to the tumor and to its response to standard therapy. Different glioma subtypes have different immune characteristics, a feature mimicked by our mouse models. The Abscopal effect, where radiation or other localized therapy induces the body’s immune system to recognize and target tumors outside the field of therapy as foreign. We have created experimental paradigms that have bilateral tumors of different subtypes, or where one tumor is treated and the immunologic response measured on the other side. Ultimately, we hope to better understand when and how checkpoint inhibitors, CAR T cells and oncolytic vectors can be used in combination with standard therapy.
One topic in this area is the issue of stem-ness in tumor cells and what drives this character. The work evolves from our previous work showing that stem like cells are located in the PVN and are driven by NO signaling mentioned in the abstract among others.
My laboratory has a long-standing collaboration with Franziska Michor of the Computational Biology department at the Dana Farber. We combine mathematical modeling with mouse modeling to understand the likelihood of events in the evolution of gliomas development or in optimizing therapy based on parameters obtained from mouse models. In these projects we have:
Many laboratories are studying the biology of these tumors (and other tumor types), but few are trying to understand the biology of how these tumors respond to therapy. This is conceptually important because the disease that kills people in the western world is a treated and recurrent tumor, not an untreated tumor. Therefore, we have spent effort in developing the technologies to understand how these tumors respond to standard therapy using the same rigor that we have studied the biology of the tumor in the first place.
In order to perform preclinical trials in mice, we need to identify tumors, quantify their size, and follow them over time non-invasively. One approach that we have used is by MRI scanning with T2 weighted images or with T1 weighted images with and without contrast as is done in people. However, MRI only measures anatomic structure and not biologic processes. Therefore, we have developed bioluminescence imaging strategies for use in preclinical trials of brain tumor-bearing mice. We initially developed a reporter mouse that expressed luciferase from the E2F1 promoter that measures proliferation and a Gli responsive promoter measuring SHH signaling. We are now developing genetic backgrounds that activate luciferase expression by cre recombinase activity that will allow us to “see” the tumor cells in vivo that have been deleted for PTEN, or that have knocked down INK4a/arf. This will allow us to easily identify mice with tumors and to count live tumor cells in vivo non-invasively.
Gliomas are composed of not only tumor cells per se but also reactive astrocytes, microglia, endothelial cells and pericytes. Multiple lines of evidence indicate that many if not all of the cells that make up the stroma in these tumors contribute to the tumor biology and may be valid therapeutic targets.
We have also developed a modified version of the RCAS/tv-a system that achieves loss-of-function combined with lineage tracing using short hairpins and florescent tags. This system is able to mimic the mesenchymal GBMs by combining knockdown the combination of NF1 and p53 while lineage tracing each of these two events from specific cell types, with a penetrance of essentially 100%. We are using this model to understand the evolution of mesenchymal GBM from proneural ones and understand the complexity of these tumors. This type of lineage tracing allows us to appreciate the cellular heterogeneity in ways that germline strategies are unable to. We also have developed a new model of ependymoma by expressing a commonly occurring gene fusion (C11orf95/RELA) with this system.
PDGF signaling characterizes the proneural subgroup of GBM and is sufficient to induced similar tumors in mice. One might think that inhibition of PDGFR would be a good therapeutic strategy for at least the proneural GBM subgroup. However, several trials of PDGFR inhibitors have been done in humans with GBM and none have been successful. A simple explanation is that the patients were not stratified to PDGFR active tumors prior to enrolling in these trials. However, there are several additional more interesting possibilities as to why this might be the case, and we are investigating under what circumstances PDGFR inhibition might be effective. One contributing factor is likely to be cellular heterogeneity of these tumors where subclones of cells within the tumor express PDGFR while others express EGFR in humans, and in mice similar results can be seen. A second contributing factor in the resistance to PDGFR inhibition is the fact that most of the gene expression changes that accompany the oncogenic transformation of olig2 expressing cells by PDGFR in vivo are not reversed by PDGFR inhibitors in vivo, even when that inhibition achieves a full cycle arrest. Additionally, mutant forms of PDGFR alpha found in some GBM appear to reduce effect of PDGFR inhibition. Finally, we have found that additional alterations found in human gliomas such as loss of Ink4a/arf, p53 or PTEN enhance oncogenic character of these tumors and prevent PDGFR inhibition of achieving full cell cycle arrest.
Cancer-driving mutations are found across a wide range of tumor types, yet are often only present in a subset of tumor cells, making early detection and subsequent treatment of cancer difficult. The identification of a unique oncogenic driver that is found in nearly all tumor cells, across various cancer subtypes, would be a highly valuable biomarker as well as a promising diagnostic and profitable therapeutic target. We have identified one such target in the form of a splice variant of the TrkB neurotrophin receptor. This variant is expressed highly across nearly all human cancers when compared to normal tissues and forced expression drives multiple tumor types in mice. Furthermore, forced expression of this splice form of TrkB, when combined with loss of PTEN, is sufficient to induce cancers from many organ sites in mice. We are working on understanding the mechanism and implications of these findings.
YAP1 is a transcriptional co-activator and a proto-oncogene. Several different YAP1 gene fusions have been identified in various human cancers. Here, we show that overexpression of several of these gene fusions in mice is sufficient to cause local tumor formation. Each of these YAP1 fusion proteins exert YAP activity, and also exert activity of the C’-terminal fusion partners. These fusion proteins evade the negative Hippo pathway regulation due to constitutive nuclear localization and resistance to degradation. Combined point mutations in YAP1 (S127/397A-YAP1) that achieve these functions also induces tumor formation in vivo. Genetic disruption of the TEAD binding domain of these oncogenic YAP1 fusions is sufficient to inhibit tumor formation in vivo, while pharmacological inhibition of the YAP1-TEAD interaction also reduces the YAP activity of the fusion proteins in vitro.