The promise of “precision oncology” relies on decoding the molecular signatures of tumors to make predictions about effective therapies. The prevailing wisdom is that precision therapies will arise from identifying and targeting "drivers" of oncogenic transformation (e.g., mutated oncogenes). However, this approach has met with limited clinical success, particularly for some of the most devastating and difficult to treat cancers. Glioblastoma multiforme (GBM) is the most aggressive and common form of brain cancer in adults: approximately 90% of GBM patients die within two years of diagnosis with current standard of care therapy. To identify novel GBM therapeutic targets, we have developed a paradigm that focuses not on targeting oncogenic drivers, but instead on normal cellular pathways that are sensitized uniquely in GBM cells and not their healthy counterparts. We have isolated patient-derived GBM “stem cells” (GSCs) and compared them with non-GBM neuronal progenitor cells (NPCs). Every gene is then systematically inactivated in each set of samples to identify genes essential for the self-renewal and growth of only the GSCs. As a result, we have now identified 100s of candidate targets, which are essential to GSCs but which have reduced requirement in NPCs and other normal cells.
BubR1 is a multiple functional protein that plays key roles implicated in mitotic checkpoint control, mitotic timing, and regulating kinetochore-microtubule attachment. We found that certain GBM isolates are sensitive to inhibition of BubR1's GLE2p-binding sequence (GLEBS) domain. We have shown, for example, that in BubR1 knockout cells that BubR1's GLEBs domain becomes essential for kinetochore-microtubule attachment and viability only after oncogenic activation of the RTK/Ras-pathway (Ding et al., Cancer Discovery, 2013). Importantly, we were able to find a key molecular indicator on the kinetochore that predicts whether GBM or other cell types will be sensitive to loss of BubR1's GLEBs domain function. We are currently working with Dr. Jim Olson's group (Clinical Research Division) to identify molecules that will interfere with BubR1's GLEBs domain activity.
BuGZ was isolated from an RNAi screen targeting putative human transcription factors to identify key regulators of GSC expansion. Its official name is ZNF207, a previously uncharacterized C2–H2 zinc-finger domain gene and putative transcription factor. We renamed the gene BuGZ (Bub3 interacting GLEBS and Zinc finger domain containing protein) and demonstrated that BuGZ is a novel kinetochore component that binds to and stabilizes Bub3 during interphase and mitosis.Just like BubR1, BuGZ binds to Bub3 through a highly conserved GLEBS domain. As with BubR1, we found the cancer-specific requirement for BuGZ was limited to its GLEBS domain, which mediates BuGZ kinetochore localization (Toledo et al., Dev. Cell, 2014). One key implication from our work is that oncogenic transformation leads to added requirement for BuGZ function. Further, cancer requirement for BuGZ may be more widespread than BubR1 based on our assessment in brain tumor isolates. For example, isolates that are resistant to BubR1-GLEBs domain inhibition are still sensitive to loss of BuGZ. As with BubR1, we are working with Jim Olson's group to identify small molecules or peptides that will interfere with BuGZ's GLEBs domain activity.
PHF5A is a highly conserved plant homeodomain (PHD)-zinc finger domain protein that facilitates interactions between the U2 small nuclear ribonucleoprotein (snRNP) complex and DNA/RNA helicases. We found that in GSCs, but not in untransformed controls, PHF5A facilitates recognition of exons with unusual C-rich 3' splice sites (ss) in thousands of essential genes. PHF5A knockdown in GSCs, but not in untransformed NSCs, astrocytes, or fibroblasts, inhibited splicing of these genes, leading to cell cycle arrest and loss of viability. Further, induction of knockdown of PHF5A in patient-derived xenograft brain tumors in mice led to highly significant survival benefits (Hubert et al., Genes and Dev., 2013). We are currently working with Jim Olson's and Marc Ferrer's groups (Fred Hutch and NIH/NCATs) to identify small molecules that trigger splicing defects similar to PHF5A inhibition in GBM cells.
HumanPKMYT1/Myt1 encodes a dual specificity (threonine and tyrosine) protein kinase homologous to WEE1 that localizes to the endoplasmic reticulum-Golgi complex and, at least in vitro, can inhibit CyclinB/CDK1 activity, by phosphorylating CDK1's ATP binding domain at Thr14. PKMYT1 arose as a key GSC-lethal hit in recent genome-wide CRISPR-Cas9 screens. Mechanistic studies showed that in non-transformed cells PKMYT1 acts redundantly with WEE1 to facilitate proper mitotic entry in primary neural progenitors via phosphorylating CDK1-Thr14 and -Tyr15, which inhibits CyclinB/CDK1 activity. However, in GSCs and NPCs with activated EGFR and AKT1 the redundancy is broken causing PKMYT1 to become essential (Toledo et al., Cell Reports, 2015)
We are currently comprehensively retesting all possible CRISPR-Cas9 and RNAi hits from each of our focus set and genome-wide screens using custom made lentiviral libraries. This will identify top scoring cancer-lethal genes. These targets and each of our published candidate therapeutic targets are in the process of being evaluated by our collaborators Drs. Jim Olson (Clinical Research Division) and Eric Holland (Human Biology Division) for clinical translation. Part of this evaluation process involves generating mouse models of target inhibition (e.g., inducible shRNA transgenics or Floxed allele) to examine "therapeutic window" in the context of mouse models of cancer (e.g., glioma) and also to examine specific on-target liabilities (i.e., are any normal tissues affected by inhibition?). Another arm of the effort includes generation of gene activity reporter assays for high throughput screens for molecular inhibitors of our cancer lethal targets (led by Dr. Drew Mhyre in the Olson Lab).
Our lab studies the regulatory features of the genome that can be passed from one generation to another — either from mother to daughter cells or from parent to child — without altering the actual DNA coding sequences of genes. The epigenentic regulatory features most commonly studied are chemical modifications of DNA (such as methylation) or histone proteins (the proteins that help pack down chromosomal DNA in the nucleus). From our published studies, we have examined roles for histone function during ESC exist from the pluripotent state (Schaniel et al., Stem Cells 2009) and roles for particular histone marks (e.g., H3K9me2) during hematopoietic stem/progenitor cells lineage commitment (Chen et al., 2012 Genes and Dev.). From our hematopoietic stem cell (HSC) studies, HSCs are developmentally "reprogrammed" to have little or no histone H3 lysine 9 methylation in the primitive state. Upon lineage commitment, H3K9me2 marks are nucleated at specific sites in the genome and then spread across the entire genome (Chen et al., 2012 Genes and Dev.). Increase in this mark coincides with global changes in chromatin structure during differentiation (Schones et al., Epigenetics and Chromatin 2014). One possibility is that the absence of this histone mark promotes developmental plasticity in uncommitted stem and progenitor populations.
Chemical modifications of mRNA occur during or after gene transcription. We are currently performing broad genomic surveys of the impact of mRNA methylation (i.e., N6-methyladenosine) on regulation of key gene mRNAs required for progenitor cell lineage commitment and stem cell self-renewal (Kuppers et al., in preparation).
There are over 200 different cell types in the human body, each with a specialized function, which arise during development and adult tissue homeostasis from transient or established progenitor cells resident in tissues and organs. Two emerging themes in disease research emphasize why it is crucial that we understand how cell identities are formed and maintained in mammals. First is the notion that cancer cells may arise from maligned development programs. In addition to co-opting growth and survival promoting pathways, tumor cells hijack molecular pathways that are normally involved in developmental processes such as cell fate determination. The existence of cancer stem cells, which may play vital roles in tumor progression, maintenance, and recurrence, underscores this notion. Second is the notion that, with the successful isolation of human embryonic stem cells (hESCs) and induced pluripotent stem cells (iPSCs), we can develop techniques to harness their developmental potential in the laboratory for clinical applications, such as cell replacement therapies for neurodegeneration, spinal cord injury, liver dysfunction, severe burns, blood disorders, etc. Through the use of defined, in vitro embryonic and somatic stem cell systems, we will find and characterize gene products affecting stem cell self-renewal, differentiation, proliferation, and survival.
To date we have examined several aspects of regulation of cell identify and cell growth:
A critical knowledge gap for the human genome has arisen from our inability to resolve genes by their key functional domains. For example, there are currently 5494 conserved protein family (Pfam) domains in the human genome with putative activities (e.g., methyltransferase-like domain). However, only a fraction of these predictions has been validated; and homology based-inference is no guarantee for functional similarity. Moreover, ~45% of the proteome is devoid of Pfam domains and is even less well characterized. This Pfam void, nonetheless, can have critical functions, such as harboring disordered protein segments with key roles in post-translational modifications, recruitment of binding partners, or conformational variability (e.g., phase transitions). Thus, although the human genome is decorated with predicted protein domains, most require validation, while many others await discovery. This represents a critical impediment for both basic and disease-focused biomedical research, where years, if not decades, can be spent resolving gene functions.
To overcome this barrier, we have developed a CRISPR-Cas9-based mutagenesis strategy that employs tiling sgRNAs across genes as a means to resolve their domain structure in living cells (aka "CTiL"). CTiL succeeds, we contend, because certain genic regions are more phenotypically constrained and less mutable than others, resulting in unique mutational signatures for each gene, with constrained regions scoring as phenotypic "peaks". Due to the composition and penetrance of Cas9-driven indel formation, CTiL works effectively in diploid and even severely aneuploid cells, underscoring its applicability to numerous cell-based models.
We used this approach for 48 well characterized kinetochore-associated genes in four different human cell lines. We computationally defined 224 functional regions, of which 59 are novel. Preliminary evidence suggests that these novel domains can contribute to protein-protein interactions responsible for sub-cellular localization of proteins at kinetochores and kinetochore function (Herman et al., in preparation).