Essential Gene Profiles in Breast, Pancreatic, and Ovarian Cancer Cells

Published OnlineFirst December 29, 2011; DOI: 10.1158/2159-8290.CD-11-0224 ReseaRch aRticLe Essential Gene Profiles in Breast, Pancreatic, and Ovarian Cancer Cells Richard Marcotte4, Kevin R. Brown1, Fernando Suarez4, Azin Sayad1, Konstantina Karamboulas1, Paul M. Krzyzanowski4, Fabrice Sircoulomb4, Mauricio Medrano3,4, Yaroslav Fedyshyn1, Judice L.Y. Koh1, Dewald van Dyk1, Bohdana Fedyshyn1, Marianna Luhova1, Glauber C. Brito1, Franco J. Vizeacoumar1, Frederick S. Vizeacoumar5, Alessandro Datti5,7, Dahlia Kasimer1, Alla Buzina1, Patricia Mero1, Christine Misquitta1, Josee Normand4, Maliha Haider4, Troy Ketela1, Jeffrey L. Wrana2,5, Robert Rottapel3,4,6, Benjamin G. Neel3,4, and Jason Moffat1,2 Downloaded from cancerdiscovery.aacrjournals.org on October 1, 2021. © 2012 American Association for Cancer Research. Published OnlineFirst December 29, 2011; DOI: 10.1158/2159-8290.CD-11-0224 The BATTLE Trial: Personalizing Therapy for Lung Cancer rEsEarCh artiClE a bstRact Genomic analyses are yielding a host of new information on the multiple genetic abnormalities associated with specific types of cancer. A comprehensive de- scription of cancer-associated genetic abnormalities can improve our ability to classify tumors into clinically relevant subgroups and, on occasion, identify mutant genes that drive the cancer phenotype (“drivers”). More often, though, the functional significance of cancer-associated mutations is difficult to discern. Genome-wide pooled short hairpin RNA (shRNA) screens enable global identification of the genes essential for cancer cell survival and proliferation, providing a “functional genomic” map of human cancer to complement genomic studies. Using a lentiviral shRNA library targeting ~16,000 genes and a newly developed, dynamic scoring approach, we identified essential gene profiles in 72 breast, pancreatic, and ovarian cancer cell lines. Integrating our results with cur- rent and future genomic data should facilitate the systematic identification of drivers, unanticipated synthetic lethal relationships, and functional vulnerabilities of these tumor types. siGNifiCaNCE: This study presents a resource of genome-scale, pooled shRNA screens for 72 breast, pancreatic, and ovarian cancer cell lines that will serve as a functional complement to ge- nomics data, facilitate construction of essential gene profiles, help uncover synthetic lethal relationships, and identify uncharacterized genetic vulnerabilities in these tumor types. Cancer Discovery; 2(2); 172–89. ©2011 AACR. iNtRODUctiON (“synthetic lethality”) can arise as a consequence of the genetic Recent technological advances have revolutionized our abnormalities in cancer cells, as illustrated by the sensitivity understanding of cancer genetics. Transcriptional profiling, of BRCA1/2-mutant breast cancer cells to PARP inhibitors copy number variation, and deep sequencing have revised (7, 8). The systematic identification of such synthetic lethal the classification of many tumors into molecular subtypes relationships might suggest new drug targets (9). Comparison that provide improved prognostic information compared of the genetic abnormalities and functional vulnerabilities of with conventional clinical and histopathologic classifica- cancer cells should help researchers identify new drivers and tion schemes (1, 2). Yet often, these subtypes provide little provide insight into the complex systems biology of cancer. functional information about the molecular events that drive RNA interference technology has enabled genome-wide cancer cell behavior. Genome-wide sequencing studies have loss-of-function screens in mammalian cells. Most screens identified hundreds to thousands of mutations in individual have used siRNAs, usually arrayed in multiwell plates. tumors (3–6), yet it frequently is difficult to know which of Arrayed screens have focused mainly on specific gene fami- these are essential for pathogenesis (i.e., “drivers”), as opposed lies, such as kinases, phosphatases, or selected candidate to “passenger” mutations. Even when a driver oncogene (e.g., genes, and have yielded new insights into mechanisms of KRAS, MYC) or tumor suppressor gene (e.g., TP53, BRCA1/2) cancer cell signal transduction, cell division, and cell death is known, these can be poor targets for drug development. (10, 11). Cell proliferation assays in multiwell plates are In addition, unanticipated gene/pathway dependencies usually constrained to a few population doublings, and gene “knockdowns” in these conditions typically last for at most a week. Therefore, siRNA screens are, by nature, 1 a uthors’ affiliations: Donnelly Centre and Banting & Best Department of transient, and might underestimate the roles of long-lived Medical Research, Departments of 2Molecular Genetics and 3Medical Biophysics, University of Toronto; 4Campbell Family Cancer Research proteins on a given phenotype. Moreover, given their cost Institute, Ontario Cancer Institute, Princess Margaret Hospital, University and the need for extensive automation to interrogate most Health Network; 5Samuel Lunenfeld Research Institute; 6Division of of the genome, siRNA screens usually are performed on only Rheumatology, Department of Medicine, St. Michael's Hospital, Toronto, limited numbers of cell lines and might fail to capture the 7 Canada ; Department of Experimental Medicine and Biochemical Sciences, genetic heterogeneity in cancer. These properties make it University of Perugia, Perugia, Italy difficult to use arrayed screening approaches to construct Note: Supplementary data for this article are available at Cancer Research extensive functional genomic maps of cancer cells. Online (http://cancerres.aacrjournals.org/). The more recent development of large retroviral- or lenti- R. Marcotte, K.R. Brown, and F. Suarez contributed equally to this article. viral-based short hairpin RNA (shRNA) libraries facilitates Corresponding authors: Jason Moffat, Donnelly Centre, 160 College genome-wide screening of cultured cancer cells in a pooled for- Street, Room 802, University of Toronto, Toronto, Canada M5S 3E1. Phone: 416-978-0336; Fax: 416-978-8287; E-mail: [email protected]. mat (12–14), providing a potential solution to the limitations Robert Rottapel. E-mail: [email protected]; Benjamin G. Neel. of arrayed screens. Cells are infected with these libraries at a E-mail: [email protected] low multiplicity of infection (MOI) and allowed to proliferate doi: 10.1158/2159-8290.CD-11-0224 for 3 to 4 weeks, after which shRNAs that have been selectively ©2011 American Association for Cancer Research. depleted (referred to as “dropouts”) or enriched are identified FEBRUARY 2012 CANCER DISCOVERY | 173 Downloaded from cancerdiscovery.aacrjournals.org on October 1, 2021. © 2012 American Association for Cancer Research. Published OnlineFirst December 29, 2011; DOI: 10.1158/2159-8290.CD-11-0224 rEsEarCh artiClE Marcotte et al. on custom microarrays or by deep sequencing. Pooled screens our confidence in the essentiality score derived for each gene. have been used to define genes necessary for cancer cell pro- We also developed a set of heuristics to classify shRNAs as liferation/survival in cell culture (12–14), genes that enhance fast, continuous, or slow dropouts, based on the rate at which or interfere with the action of specific oncogenes (15), or an shRNA disappeared from the bulk population of cells genes that enhance the effects of antineoplastic drugs, sug- during the screen (see Methods and Supplementary Table gesting potential new combination therapies (16, 17). S3). Examples of these profiles are shown at the right of Most large-scale pooled shRNA screens have surveyed Figure 1A. Using heuristics designed to identify the most po- cancer cell lines representing multiple histotypes but usu- tent shRNAs in the fast, continuous, or slow classes resulted ally with few representatives of any one tumor type, or they in the classification of ~2% of the shRNAs in the library into have focused on cell lines from different histotypes bearing one of these categories, with 40% fast, 30% continuous, and the same genetic abnormality (e.g., KRAS mutations) (15, 30% slow dropouts. These classification criteria largely re- 18). As an initial step toward a more comprehensive un- stricted hairpins to a single class. Moreover, dropout behav- derstanding of the vulnerabilities of breast cancer (BrCa), ior largely appeared to be characteristic of the gene targeted pancreatic ductal adenocarcinoma (PDAC), and high-grade by the hairpin rather than the shRNA itself: within any cell serous ovarian carcinoma (HGS-OvCa), we performed near line, a given gene almost always fell into a single dropout genome-wide pooled shRNA screens on 72 cancer cell lines class (Fig. 1B). Altering our heuristics would allow us to clas- and established a unique informatics approach to monitor sify more hairpins but would result in greater overlap be- the dynamic evolution of cancer cell populations. We chose tween dropout classes and also lead to classification of less breast cancer because the extensive genomic information potent hairpins. On average, ~0.4% of shRNAs were enriched and subtype classification schemes that exist for this tu- in any given cell line; due to our shRNA bar code detection mor type facilitate integrated genomic/functional genomic procedures (see Supplementary Table S4 and Methods), this analysis. Ongoing genomic efforts should provide similar is almost certainly a substantial underestimate of the true information for PDAC and

Load more