Identification of Genetic Vulnerabilities in Glioblastoma and Other Cancers Using CRISPR-Cas9-based Functional Genomic Approaches

Abstract

Glioblastoma (GBM) is the most common and aggressive form of brain cancer. It remains one of the deadliest cancers, with 90% of adult patients dying within two years of diagnosis with standard-of-care treatments. Although directly targeting oncogenic driver mutations is a popular strategy in developing new therapeutics, it has been met with limited success. Thus, we take the alternative, unbiased approach of using functional genomic screens to identify molecular vulnerabilities that arise in cancer cells due to the oncogenic state. In order to identify genes required for the proliferation and survival of GBM cells, the Paddison Lab previously performed lentiviral genome-scale pooled RNAi and CRISPR-Cas9 outgrowth screens in patient-derived GBM stem-like cells (GSCs) and human nontransformed neural stem cells (NSCs). Retesting screen hits using single CRISPR sgRNAs can prove challenging with lentivirus due to protracted windows of indel formation (editing over the course of ~5-12 days), resulting in phenotypically mixed populations. Here, to address this challenge, I optimized a method for direct nucleofection of ribonucleoprotein complexes (RNPs) composed of chemically synthesized 2'-O-methyl 3’phosphorothioate-modified sgRNA and purified Cas9 protein. With this technique, we can routinely achieve >90% indel formation as well as targeted genomic deletions in only 3 days, even in hyperdiploid cells (such as GSCs), with no need to create clonal lines for simple loss-of-function experiments. Additionally in this body of work, to further characterize candidate hits specific to GSCs (and not NSCs), we performed comprehensive pooled outgrowth retests of all putative screen hits that scored preferentially in GSC isolates in our whole-genome CRISPR-Cas9 screens. We then used these results to identify both GBM gene dependency groups and context-specific vulnerabilities, which was facilitated by incorporating published functional genomic screening data (DepMap database) into the analyses. By creating co-dependency networks, we identified GSC-specific gene vulnerability groups related to mitochondrial protein processing and turnover and membrane trafficking as well as metabolic enzymes and regulators. Furthermore, we also identified predicted context-specific vulnerabilities and further investigated the dsRNA-editing enzyme ADAR, which is required in cancer cells that have an interferon-stimulated gene expression signature, and the adapter protein EFR3A, which is required in cancer cells that have low expression of its paralog EFR3B. In addition, we investigated the particularly strongly GSC-specific screen hit FBXO42. We found this F-box protein to be essential in a subset of GSCs in vitro and in vivo but nonessential in NSCs in vitro, in addition to being required in subsets of cells of various other cancer types, suggesting the potential for both a large therapeutic window and broad applicability. Furthermore, we demonstrated that the ubiquitin ligase role of FBXO42 is responsible for the viability phenotype, but likely not through p53 as a previous study might suggest. In searching for possible interactors, we determined that FBXO42 and the gene CCDC6 (Coiled-Coil Domain Containing 6, a common translocation partner for receptor kinase oncogenes) are both necessary to promote viability in FBXO42 loss-sensitive cells, likely working together rather than redundantly. Lastly, we demonstrated that the GSC-specific viability loss is due to an extended metaphase arrest upon FBXO42 knockout due to prolonged spindle assembly checkpoint activation. Altogether, this body of work both improves upon existing CRISPR-Cas9 technologies and uses this technology as part of an effort to identify potential novel therapeutic opportunities in GBM and other cancers.