Identification of Cancer-Specific Essential Genes in Glioblastoma Multiforme

Thesis (Ph.D.)--University of Washington, 2012Glioblastoma multiforme (GBM) is the most common and aggressive type of brain cancer and is notoriously drug and radiation resistant. Current therapies and trials have repeatedly failed to substantially increase the survival of patients over the last two decades. Although new therapeutic approaches are urgently needed, sophisticated strategies to identify novel GBM therapies have been precluded by several biological and technical barriers. First, established GBM cell lines poorly represent human glioblastoma. Second, powerful gene discovery and manipulation tools were not yet developed or applied to GBM. To overcome these historic challenges, we have applied both focused and genome-wide functional genetic screening techniques to multiple primary glioma stem cell (GSC) cultures that accurately model essential elements of this disease, as well as normal neural stem cells (NSCs). From these unbiased screens, we show that GSCs have an added requirement for the mitotic spindle checkpoint BUB1B to suppress lethal consequences of altered kinetochore function. Specifically, the activity of the Gle2-binding-sequence (GLEBS) domain within BUB1B is required to suppress lethal kinetochore-microtubule attachment defects in GBM isolates and genetically transformed cells. We also identified the pre-mRNA splicing protein PHF5A as a key GSC-specific essential gene. PHF5A knockdown or pharmacologic inhibition of its complex triggers a G2/M arrest in GSCs but not in NSCs or other untransformed cells. We show that PHF5A facilitates recognition of exons with unusual C-rich 3' splice sites and its suppression results in the cancer-specific mis-splicing of thousands of genes. PHF5A suppression compromises GBM tumor formation and maintenance in vivo and results in the regression of established GBM tumors. Rather than being isolated to GSCs, this sensitivity to splicing inhibition is conferred by Ras-mediated transformation of multiple normal cell types, suggesting that a therapy targeting PHF5A may be applicable to a wide range of cancer types