Project DRIVE: A Compendium of Cancer Dependencies and Synthetic Lethal Relationships Uncovered by Large-Scale, Deep RNAi Screening.

Elucidation of the mutational landscape of human cancer has progressed rapidly and been accompanied by the development of therapeutics targeting mutant oncogenes. However, a comprehensive mapping of cancer dependencies has lagged behind and the discovery of therapeutic targets for counteracting tumor suppressor gene loss is needed. To identify vulnerabilities relevant to specific cancer subtypes, we conducted a large-scale RNAi screen in which viability effects of mRNA knockdown were assessed for 7,837 genes using an average of 20 shRNAs per gene in 398 cancer cell lines. We describe findings of this screen, outlining the classes of cancer dependency genes and their relationships to genetic, expression, and lineage features. In addition, we describe robust gene-interaction networks recapitulating both protein complexes and functional cooperation among complexes and pathways. This dataset along with a web portal is provided to the community to assist in the discovery and translation of new therapeutic approaches for cancer.

Identification of regulators of polyploidization presents therapeutic targets for treatment of AMKL.

The mechanism by which cells decide to skip mitosis to become polyploid is largely undefined. Here we used a high-content image-based screen to identify small-molecule probes that induce polyploidization of megakaryocytic leukemia cells and serve as perturbagens to help understand this process. Our study implicates five networks of kinases that regulate the switch to polyploidy. Moreover, we find that dimethylfasudil (diMF, H-1152P) selectively increased polyploidization, mature cell-surface marker expression, and apoptosis of malignant megakaryocytes. An integrated target identification approach employing proteomic and shRNA screening revealed that a major target of diMF is Aurora kinase A (AURKA). We further find that MLN8237 (Alisertib), a selective inhibitor of AURKA, induced polyploidization and expression of mature megakaryocyte markers in acute megakaryocytic leukemia (AMKL) blasts and displayed potent anti-AMKL activity in vivo. Our findings provide a rationale to support clinical trials of MLN8237 and other inducers of polyploidization and differentiation in AMKL.

A physical and regulatory map of host-influenza interactions reveals pathways in H1N1 infection.

During the course of a viral infection, viral proteins interact with an array of host proteins and pathways. Here, we present a systematic strategy to elucidate the dynamic interactions between H1N1 influenza and its human host. A combination of yeast two-hybrid analysis and genome-wide expression profiling implicated hundreds of human factors in mediating viral-host interactions. These factors were then examined functionally through depletion analyses in primary lung cells. The resulting data point to potential roles for some unanticipated host and viral proteins in viral infection and the host response, including a network of RNA-binding proteins, components of WNT signaling, and viral polymerase subunits. This multilayered approach provides a comprehensive and unbiased physical and regulatory model of influenza-host interactions and demonstrates a general strategy for uncovering complex host-pathogen relationships.

Synthetic lethal interaction between oncogenic KRAS dependency and STK33 suppression in human cancer cells.

An alternative to therapeutic targeting of oncogenes is to perform "synthetic lethality" screens for genes that are essential only in the context of specific cancer-causing mutations. We used high-throughput RNA interference (RNAi) to identify synthetic lethal interactions in cancer cells harboring mutant KRAS, the most commonly mutated human oncogene. We find that cells that are dependent on mutant KRAS exhibit sensitivity to suppression of the serine/threonine kinase STK33 irrespective of tissue origin, whereas STK33 is not required by KRAS-independent cells. STK33 promotes cancer cell viability in a kinase activity-dependent manner by regulating the suppression of mitochondrial apoptosis mediated through S6K1-induced inactivation of the death agonist BAD selectively in mutant KRAS-dependent cells. These observations identify STK33 as a target for treatment of mutant KRAS-driven cancers and demonstrate the potential of RNAi screens for discovering functional dependencies created by oncogenic mutations that may enable therapeutic intervention for cancers with "undruggable" genetic alterations.

Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1.

The proto-oncogene KRAS is mutated in a wide array of human cancers, most of which are aggressive and respond poorly to standard therapies. Although the identification of specific oncogenes has led to the development of clinically effective, molecularly targeted therapies in some cases, KRAS has remained refractory to this approach. A complementary strategy for targeting KRAS is to identify gene products that, when inhibited, result in cell death only in the presence of an oncogenic allele. Here we have used systematic RNA interference to detect synthetic lethal partners of oncogenic KRAS and found that the non-canonical IkappaB kinase TBK1 was selectively essential in cells that contain mutant KRAS. Suppression of TBK1 induced apoptosis specifically in human cancer cell lines that depend on oncogenic KRAS expression. In these cells, TBK1 activated NF-kappaB anti-apoptotic signals involving c-Rel and BCL-XL (also known as BCL2L1) that were essential for survival, providing mechanistic insights into this synthetic lethal interaction. These observations indicate that TBK1 and NF-kappaB signalling are essential in KRAS mutant tumours, and establish a general approach for the rational identification of co-dependent pathways in cancer.

A public genome-scale lentiviral expression library of human ORFs.

Functional characterization of the human genome requires tools for systematically modulating gene expression in both loss-of-function and gain-of-function experiments. We describe the production of a sequence-confirmed, clonal collection of over 16,100 human open-reading frames (ORFs) encoded in a versatile Gateway vector system. Using this ORFeome resource, we created a genome-scale expression collection in a lentiviral vector, thereby enabling both targeted experiments and high-throughput screens in diverse cell types.

CDK8 is a colorectal cancer oncogene that regulates beta-catenin activity.

Aberrant activation of the canonical WNT/beta-catenin pathway occurs in almost all colorectal cancers and contributes to their growth, invasion and survival. Although dysregulated beta-catenin activity drives colon tumorigenesis, further genetic perturbations are required to elaborate full malignant transformation. To identify genes that both modulate beta-catenin activity and are essential for colon cancer cell proliferation, we conducted two loss-of-function screens in human colon cancer cells and compared genes identified in these screens with an analysis of copy number alterations in colon cancer specimens. One of these genes, CDK8, which encodes a member of the mediator complex, is located at 13q12.13, a region of recurrent copy number gain in a substantial fraction of colon cancers. Here we show that the suppression of CDK8 expression inhibits proliferation in colon cancer cells characterized by high levels of CDK8 and beta-catenin hyperactivity. CDK8 kinase activity was necessary for beta-catenin-driven transformation and for expression of several beta-catenin transcriptional targets. Together these observations suggest that therapeutic interventions targeting CDK8 may confer a clinical benefit in beta-catenin-driven malignancies.

Scoring diverse cellular morphologies in image-based screens with iterative feedback and machine learning.

Many biological pathways were first uncovered by identifying mutants with visible phenotypes and by scoring every sample in a screen via tedious and subjective visual inspection. Now, automated image analysis can effectively score many phenotypes. In practical application, customizing an image-analysis algorithm or finding a sufficient number of example cells to train a machine learning algorithm can be infeasible, particularly when positive control samples are not available and the phenotype of interest is rare. Here we present a supervised machine learning approach that uses iterative feedback to readily score multiple subtle and complex morphological phenotypes in high-throughput, image-based screens. First, automated cytological profiling extracts hundreds of numerical descriptors for every cell in every image. Next, the researcher generates a rule (i.e., classifier) to recognize cells with a phenotype of interest during a short, interactive training session using iterative feedback. Finally, all of the cells in the experiment are automatically classified and each sample is scored based on the presence of cells displaying the phenotype. By using this approach, we successfully scored images in RNA interference screens in 2 organisms for the prevalence of 15 diverse cellular morphologies, some of which were previously intractable.

The transcription factor Eyes absent is a protein tyrosine phosphatase.

Post-translational modifications provide sensitive and flexible mechanisms to dynamically modulate protein function in response to specific signalling inputs. In the case of transcription factors, changes in phosphorylation state can influence protein stability, conformation, subcellular localization, cofactor interactions, transactivation potential and transcriptional output. Here we show that the evolutionarily conserved transcription factor Eyes absent (Eya) belongs to the phosphatase subgroup of the haloacid dehalogenase (HAD) superfamily, and propose a function for it as a non-thiol-based protein tyrosine phosphatase. Experiments performed in cultured Drosophila cells and in vitro indicate that Eyes absent has intrinsic protein tyrosine phosphatase activity and can autocatalytically dephosphorylate itself. Confirming the biological significance of this function, mutations that disrupt the phosphatase active site severely compromise the ability of Eyes absent to promote eye specification and development in Drosophila. Given the functional importance of phosphorylation-dependent modulation of transcription factor activity, this evidence for a nuclear transcriptional coactivator with intrinsic phosphatase activity suggests an unanticipated method of fine-tuning transcriptional regulation.

Functional screening identifies miR-315 as a potent activator of Wingless signaling.

The existence of vast regulatory networks mediated by microRNAs (miRNAs) suggests broad potential for miRNA dysfunction to contribute to disease. However, relatively few miRNA-target interactions are likely to make detectable contributions to phenotype, and effective strategies to identify these few interactions are currently wanting. We hypothesized that signaling cascades represent critical points of susceptibility to miRNA dysfunction, and we developed a strategy to test this theory by using quantitative cell-based screens. Here we report a screen for miRNAs that affect the Wingless (Wg) pathway, a conserved pathway that regulates growth and tissue specification. This process identified ectopic miR-315 as a potent and specific activator of Wg signaling, an activity that we corroborated in transgenic animals. This miR-315 activity was mediated by direct inhibition of Axin and Notum, which encode essential, negatively acting components of the Wg pathway. Genetic interaction tests substantiated both of these genes as key functional targets of miR-315. The ability of ectopic miR-315 to activate Wg signaling was not a trivial consequence of predicted miRNA-target relationships because other miRNAs with conserved sites in the Axin 3' UTR neither activated Wg outputs nor inhibited an Axin sensor. In summary, activity-based screening can selectively identify miRNAs whose deregulation can lead to interpretable phenotypic consequences.

Tumor Intrinsic Efficacy by SHP2 and RTK Inhibitors in KRAS-Mutant Cancers.

KRAS, an oncogene mutated in nearly one third of human cancers, remains a pharmacologic challenge for direct inhibition except for recent advances in selective inhibitors targeting the G12C variant. Here, we report that selective inhibition of the protein tyrosine phosphatase, SHP2, can impair the proliferation of KRAS-mutant cancer cells in vitro and in vivo using cell line xenografts and primary human tumors. In vitro, sensitivity of KRAS-mutant cells toward the allosteric SHP2 inhibitor, SHP099, is not apparent when cells are grown on plastic in 2D monolayer, but is revealed when cells are grown as 3D multicellular spheroids. This antitumor activity is also observed in vivo in mouse models. Interrogation of the MAPK pathway in SHP099-treated KRAS-mutant cancer models demonstrated similar modulation of p-ERK and DUSP6 transcripts in 2D, 3D, and in vivo, suggesting a MAPK pathway-dependent mechanism and possible non-MAPK pathway-dependent mechanisms in tumor cells or tumor microenvironment for the in vivo efficacy. For the KRASG12C MIAPaCa-2 model, we demonstrate that the efficacy is cancer cell intrinsic as there is minimal antiangiogenic activity by SHP099, and the effects of SHP099 is recapitulated by genetic depletion of SHP2 in cancer cells. Furthermore, we demonstrate that SHP099 efficacy in KRAS-mutant models can be recapitulated with RTK inhibitors, suggesting RTK activity is responsible for the SHP2 activation. Taken together, these data reveal that many KRAS-mutant cancers depend on upstream signaling from RTK and SHP2, and provide a new therapeutic framework for treating KRAS-mutant cancers with SHP2 inhibitors.

New vision from Eyes absent: transcription factors as enzymes.

Post-translational modifications such as phosphorylation provide versatile context-specific strategies for modulating transcription factor activity. In the prevailing view of this process, modifying enzymes indirectly influence gene expression by shuttling to and from the nucleus where they alter the activity of their target transcription factors. However, a new paradigm has recently been suggested from studies of Eyes absent (EYA), a member of a conserved network of transcriptional regulators implicated in the development of numerous tissues and organs including the eye, ear, muscle and kidney. These findings indicate that EYA operates both as a transcriptional coactivator and as the prototype of a novel class of protein tyrosine phosphatases. The regulatory potential of such a bifunctional transcription factor is enormous and suggests a new layer of dynamic regulation in which transcription factors themselves might provide intrinsic enzymatic activities to fine-tune nuclear output.

Generation of High-Throughput Three-Dimensional Tumor Spheroids for Drug Screening.

Cancer cells have routinely been cultured in two dimensions (2D) on a plastic surface. This technique, however, lacks the true environment a tumor mass is exposed to in vivo. Solid tumors grow not as a sheet attached to plastic, but instead as a collection of clonal cells in a three-dimensional (3D) space interacting with their neighbors, and with distinct spatial properties such as the disruption of normal cellular polarity. These interactions cause 3D-cultured cells to acquire morphological and cellular characteristics which are more relevant to in vivo tumors. Additionally, a tumor mass is in direct contact with other cell types such as stromal and immune cells, as well as the extracellular matrix from all other cell types. The matrix deposited is comprised of macromolecules such as collagen and fibronectin. In an attempt to increase the translation of research findings in oncology from bench to bedside, many groups have started to investigate the use of 3D model systems in their drug development strategies. These systems are thought to be more physiologically relevant because they attempt to recapitulate the complex and heterogeneous environment of a tumor. These systems, however, can be quite complex, and, although amenable to growth in 96-well formats, and some now even in 384, they offer few choices for large-scale growth and screening. This observed gap has led to the development of the methods described here in detail to culture tumor spheroids in a high-throughput capacity in 1536-well plates. These methods represent a compromise to the highly complex matrix-based systems, which are difficult to screen, and conventional 2D assays. A variety of cancer cell lines harboring different genetic mutations are successfully screened, examining compound efficacy by using a curated library of compounds targeting the Mitogen-Activated Protein Kinase or MAPK pathway. The spheroid culture responses are then compared to the response of cells grown in 2D, and differential activities are reported. These methods provide a unique protocol for testing compound activity in a high-throughput 3D setting.

Functional dissection of eyes absent reveals new modes of regulation within the retinal determination gene network.

The retinal determination (RD) gene network encodes a group of transcription factors and cofactors necessary for eye development. Transcriptional and posttranslational regulation of RD family members is achieved through interactions within the network and with extracellular signaling pathways, including epidermal growth factor receptor/RAS/mitogen-activated protein kinase (MAPK), transforming growth factor beta/DPP, Wingless, Hedgehog, and Notch. Here we present the results of structure-function analyses that reveal novel aspects of Eyes absent (EYA) function and regulation. We find that the conserved C-terminal EYA domain negatively regulates EYA transactivation potential, and that GROUCHO-SINE OCULIS (SO) interactions provide another mechanism for negative regulation of EYA-SO target genes. We have mapped the transactivation potential of EYA to an internal proline-, serine-, and threonine-rich region that includes the EYA domain 2 (ED2) and two MAPK phosphorylation consensus sites and demonstrate that activation of the RAS/MAPK pathway potentiates transcriptional output of EYA and the EYA-SO complex in certain contexts. Drosophila S2 cell two-hybrid assays were used to describe a novel homotypic interaction that is mediated by EYA's N terminus. Our data suggest that EYA requires homo- and heterotypic interactions and RAS/MAPK signaling responsiveness to ensure context-appropriate RD gene network activity.

Using Drosophila to decipher how mutations associated with human branchio-oto-renal syndrome and optical defects compromise the protein tyrosine phosphatase and transcriptional functions of eyes absent.

Eyes absent (EYA) proteins are defined by a conserved C-terminal EYA domain (ED) that both contributes to its function as a transcriptional coactivator by mediating protein-protein interactions and possesses intrinsic protein tyrosine phosphatase activity. Mutations in human EYA1 result in an autosomal dominant disorder called branchio-oto-renal (BOR) syndrome as well as congenital cataracts and ocular defects (OD). Both BOR- and OD-associated missense mutations alter residues in the conserved ED as do three missense mutations identified from Drosophila eya alleles. To investigate the molecular mechanisms whereby these mutations disrupt EYA function, we tested their activity in a series of assays that measured in vivo function, phosphatase activity, transcriptional capability, and protein-protein interactions. We find that the OD-associated mutations retain significant in vivo activity whereas those derived from BOR patients show a striking decrease or loss of in vivo functionality. Protein-protein interactions, either with its partner transcription factor Sine oculis or with EYA itself, were not significantly compromised. Finally, the results of the biochemical assays suggest that both loss of protein tyrosine phosphatase activity and reduced transcriptional capability contribute to the impaired EYA function associated with BOR/OD syndrome, thus shedding new light into the molecular mechanisms underlying this disease.

An epigenetic mechanism of resistance to targeted therapy in T cell acute lymphoblastic leukemia.

The identification of activating NOTCH1 mutations in T cell acute lymphoblastic leukemia (T-ALL) led to clinical testing of γ-secretase inhibitors (GSIs) that prevent NOTCH1 activation. However, responses to these inhibitors have been transient, suggesting that resistance limits their clinical efficacy. Here we modeled T-ALL resistance, identifying GSI-tolerant 'persister' cells that expand in the absence of NOTCH1 signaling. Rare persisters are already present in naive T-ALL populations, and the reversibility of their phenotype suggests an epigenetic mechanism. Relative to GSI-sensitive cells, persister cells activate distinct signaling and transcriptional programs and exhibit chromatin compaction. A knockdown screen identified chromatin regulators essential for persister viability, including BRD4. BRD4 binds enhancers near critical T-ALL genes, including MYC and BCL2. The BRD4 inhibitor JQ1 downregulates expression of these targets and induces growth arrest and apoptosis in persister cells, at doses well tolerated by GSI-sensitive cells. Consistently, the GSI-JQ1 combination was found to be effective against primary human leukemias in vivo. Our findings establish a role for epigenetic heterogeneity in leukemia resistance that may be addressed by incorporating epigenetic modulators in combination therapy.

ZFHX4 interacts with the NuRD core member CHD4 and regulates the glioblastoma tumor-initiating cell state.

Glioblastoma (GBM) harbors subpopulations of therapy-resistant tumor-initiating cells (TICs) that are self-renewing and multipotent. To understand the regulation of the TIC state, we performed an image-based screen for genes regulating GBM TIC maintenance and identified ZFHX4, a 397 kDa transcription factor. ZFHX4 is required to maintain TIC-associated and normal human neural precursor cell phenotypes in vitro, suggesting that ZFHX4 regulates differentiation, and its suppression increases glioma-free survival in intracranial xenografts. ZFHX4 interacts with CHD4, a core member of the nucleosome remodeling and deacetylase (NuRD) complex. ZFHX4 and CHD4 bind to overlapping sets of genomic loci and control similar gene expression programs. Using expression data derived from GBM patients, we found that ZFHX4 significantly affects CHD4-mediated gene expression perturbations, which defines ZFHX4 as a master regulator of CHD4. These observations define ZFHX4 as a regulatory factor that links the chromatin-remodeling NuRD complex and the GBM TIC state.

The JAK2/STAT3 signaling pathway is required for growth of CD44⁺CD24⁻ stem cell-like breast cancer cells in human tumors.

Intratumor heterogeneity is a major clinical problem because tumor cell subtypes display variable sensitivity to therapeutics and may play different roles in progression. We previously characterized 2 cell populations in human breast tumors with distinct properties: CD44+CD24- cells that have stem cell-like characteristics, and CD44-CD24+ cells that resemble more differentiated breast cancer cells. Here we identified 15 genes required for cell growth or proliferation in CD44+CD24- human breast cancer cells in a large-scale loss-of-function screen and found that inhibition of several of these (IL6, PTGIS, HAS1, CXCL3, and PFKFB3) reduced Stat3 activation. We found that the IL-6/JAK2/Stat3 pathway was preferentially active in CD44+CD24- breast cancer cells compared with other tumor cell types, and inhibition of JAK2 decreased their number and blocked growth of xenografts. Our results highlight the differences between distinct breast cancer cell types and identify targets such as JAK2 and Stat3 that may lead to more specific and effective breast cancer therapies.

CK1epsilon is required for breast cancers dependent on beta-catenin activity.

BACKGROUND: Aberrant beta-catenin signaling plays a key role in several cancer types, notably colon, liver and breast cancer. However approaches to modulate beta-catenin activity for therapeutic purposes have proven elusive to date. METHODOLOGY: To uncover genetic dependencies in breast cancer cells that harbor active beta-catenin signaling, we performed RNAi-based loss-of-function screens in breast cancer cell lines in which we had characterized beta-catenin activity. Here we identify CSNK1E, the gene encoding casein kinase 1 epsilon (CK1epsilon) as required specifically for the proliferation of breast cancer cells with activated beta-catenin and confirm its role as a positive regulator of beta-catenin-driven transcription. Furthermore, we demonstrate that breast cancer cells that harbor activated beta-catenin activity exhibit enhanced sensitivity to pharmacological blockade of Wnt/beta-catenin signaling. We also find that expression of CK1epsilon is able to promote oncogenic transformation of human cells in a beta-catenin-dependent manner. CONCLUSIONS/SIGNIFICANCE: These studies identify CK1epsilon as a critical contributor to activated beta-catenin signaling in cancer and suggest it may provide a potential therapeutic target for cancers that harbor active beta-catenin. More generally, these observations delineate an approach that can be used to identify druggable synthetic lethal interactions with signaling pathways that are frequently activated in cancer but are difficult to target with the currently available small molecule inhibitors.

Subtype-specific genomic alterations define new targets for soft-tissue sarcoma therapy.

Soft-tissue sarcomas, which result in approximately 10,700 diagnoses and 3,800 deaths per year in the United States, show remarkable histologic diversity, with more than 50 recognized subtypes. However, knowledge of their genomic alterations is limited. We describe an integrative analysis of DNA sequence, copy number and mRNA expression in 207 samples encompassing seven major subtypes. Frequently mutated genes included TP53 (17% of pleomorphic liposarcomas), NF1 (10.5% of myxofibrosarcomas and 8% of pleomorphic liposarcomas) and PIK3CA (18% of myxoid/round-cell liposarcomas, or MRCs). PIK3CA mutations in MRCs were associated with Akt activation and poor clinical outcomes. In myxofibrosarcomas and pleomorphic liposarcomas, we found both point mutations and genomic deletions affecting the tumor suppressor NF1. Finally, we found that short hairpin RNA (shRNA)-based knockdown of several genes amplified in dedifferentiated liposarcoma, including CDK4 and YEATS4, decreased cell proliferation. Our study yields a detailed map of molecular alterations across diverse sarcoma subtypes and suggests potential subtype-specific targets for therapy.