Abstract NG02: A network of deeply conserved synthetic-lethal interactions for exploration of precision cancer therapy

An emerging therapeutic strategy for cancer is to induce selective lethality in a tumor by exploiting interactions between its driving mutations and specific drug targets. Here, we develop a resource of synthetic-lethal interactions between genes mutated in cancer, including many tumor suppressor genes (TSG), and selective chemical inhibitors including many FDA-approved drugs, using an integrative multi-species approach. Whereas, targeting oncogenes with either chemical inhibitors or therapeutic antibodies has proven to be highly effective for cancer therapy, it is not currently feasible to restore the function of mutated or deleted TSGs in the clinical setting. Rather than targeting a TSG directly, it is possible to exploit a “synthetic lethal” genetic interactions between the TSG and another gene, such that simultaneous disruption of both gene functions causes rapid and selective cell death. For example, cells deficient for BRCA1 have a reduced capacity for repairing double-stranded DNA breaks and are especially vulnerable to further perturbations in alternate DNA repair pathways. Oliparib, an FDA-approved drug, exploits this principle by targeting a component of the base excision repair pathway, PARP1, thus causing selective cell death in BRCA1-/- or BRCA2 -/- cells. Recent efforts to map synthetic-lethal interactions in cancer typically fall into one of several categories. First, populations of tumor genomes may be analyzed statistically to detect pairs of genes that are seldom co-mutated in the same tumor, with one interpretation being that loss-of-function of both genes is synthetically lethal. While promising, such approaches are under-powered to test many relevant interactions, due to the already low frequency of mutation for most TSGs and the quadratic number of gene pairs that must be tested for co-mutation. Second, synthetic-lethal interactions may be mapped by directed combinatorial disruptions in human cell lines, using pairwise RNAi knockdowns, RNAi or drug treatments in cell lines with TSG loss-of-function or, conceivably in the near future, the CRISPR-Cas9 system. While such directed approaches can test relevant interactions in an unbiased manner, the largest screens performed to-date (∼10,000 gene pairs) still fall quite short of the required throughput to interrogate the potential interaction space of millions of human gene pairs involving a TSG. A complementary strategy for mapping synthetic lethal interactions in cancer is to leverage conservation with genetic interactions first identified in model species. In the yeasts S. cerevisiae and S. pombe, techniques such as synthetic genetic arrays (SGA) and Pombe Epistasis Mapper (PEM) enable genetic interactions to be measured in an unbiased and high-throughput manner, with minimal off-target effects since the genes are disrupted by complete and specific knockout of the open reading frame. Such interactions are numerous and found to be significantly conserved across species, especially for the core conserved pathways in which TSGs typically operate such as the cell cycle, genome maintenance and metabolic growth. Many TSGs important for human cancer were first identified and studied in yeast, which also provides an accessible model system in which to study mechanism of action for effects first observed in humans. Nonetheless, it is unclear whether and to what extent synthetic lethal interactions observed for core conserved processes can be ultimately translated for clinical application. Multiple factors have been postulated to influence whether an interaction will be translatable, including genetic, epigenetic, and environmental context as well as the strength, redundancy, and network topology of the interaction. To study such factors, however, would require a large cross-species dataset of genetic interactions relevant to cancer genes and functions. Here, we generate a comprehensive multi-species synthetic lethal network as a resource for the study of cancer and the design of targeted therapy. Leveraging the throughput and precise gene disruption of SGA technology, we experimentally test ∼78,000 potential interactions to generate a network that includes quantitative tests for interaction among all yeast orthologs of human TSGs and genes that are currently targetable by selective inhibitors (“druggable” targets or DT). Guided by these data, we target 2,352 TSG-drug combinations in human HeLa cells, resulting in a validated network of 172 “deeply conserved” interactions, called CoCaNet (Conserved Cancer Network). Having created this resource of conserved synthetic lethal interactions we explored three possible applications. First we demonstrate that synthetic lethal relationships in the conserved network are strongly predictive of cell survival in orthogonal survival assays and in alternate cell lines. We validated synthetic lethal interactions between the TSG RAD17 and all five synthetic lethal partners in CoCaNet (CHEK1, CHEK2, TOP2, TOP3A, CSNK1G1) in clonogenic assays using HeLa cells. For the TSG XRCC3 five of seven synthetic lethal partners (HDAC1, HDAC2, HDAC6, IMPDH1, RABGGTB) were confirmed in clonogenic assay in LN428 glioblastoma cells. Second, we examined the clinical relevance of these synthetic lethal networks. Using gene expression and clinical survival data of breast cancer patients from the METABRIC database (Curtis et al., Nature 2012) we tested our hypothesis that co-under expression of genes in CoCaNet would reduce the fitness of a tumor and associate with better clinical outcome. As expected overall survival was 8.6 years for those patients in the top 90th percentile of synthetic lethal interactions vs. 7.3 years for the 10th percentile (log-rank p Third, we use the overlapping yeast and human synthetic lethal networks to learn the ‘rules’ that govern whether an interaction observed in yeast will be conserved in humans. We annotated each gene pair with multiple observations including whether we had observed interaction conservation with yeast, the degree to which the genes are co-expressed, whether the gene products are linked by a protein-protein interaction, and whether the genes are known to co-function in the same Gene Ontology biological process. Training on the overlapping yeast and human networks we integrated these multiple lines of evidence into a combined Log Likelihood Score classifier. Applying this classifier of cross-species conservation to the complete yeast network we are able to predict an expanded human network of over 11,000 prioritized synthetic sick or lethal interactions for pre-clinical and ultimately clinical exploration, each backed by data from budding yeast for investigating drug mode of action. Citation Format: John Paul Shen, Rohith Srivas, Chih Cheng Yang, Su Ming Sun, Jian Feng Li, Andrew Gross, James Jensen, Kate Licon, Ana Bojoquez-Gomez, Kristin Klepper, Haico van Attikum, Pedro Aza-Blanc, Robert Sobol, Trey Ideker. A network of deeply conserved synthetic-lethal interactions for exploration of precision cancer therapy. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr NG02.