Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evolutionarily conserved synthetic lethal interaction networks reveal targets for anticancer therapeutic… van Pel, Derek Michael 2012

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2012_fall_vanpel_derek.pdf [ 8.36MB ]
Metadata
JSON: 24-1.0073341.json
JSON-LD: 24-1.0073341-ld.json
RDF/XML (Pretty): 24-1.0073341-rdf.xml
RDF/JSON: 24-1.0073341-rdf.json
Turtle: 24-1.0073341-turtle.txt
N-Triples: 24-1.0073341-rdf-ntriples.txt
Original Record: 24-1.0073341-source.json
Full Text
24-1.0073341-fulltext.txt
Citation
24-1.0073341.ris

Full Text

  i EVOLUTIONARILY CONSERVED SYNTHETIC LETHAL INTERACTION NETWORKS REVEAL TARGETS FOR ANTICANCER THERAPEUTIC DEVELOPMENT  by Derek Michael van Pel  B.Sc., Vancouver Island University, 2006  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Biochemistry and Molecular Biology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  October, 2012  © Derek Michael van Pel, 2012   ii Abstract Cancer is a multigenic disease. The genetic distinctness of cancer cells offers a weakness that can be exploited: for example, nearly all cancers carry mutations in processes relating to the maintenance of genomic stability. As this is an essential process, this presents a weakness that can be leveraged towards inviability – a concept known as synthetic lethality. The ideal cancer therapeutic would have a broad spectrum, but genetic techniques in human cells are not sufficiently developed to identify the spectrum of synthetic lethal interactions of sets of genome stability genes easily. The use of model organisms can facilitate the identification of second-site targets for the development of anticancer therapeutics, and allows the construction of synthetic lethal interaction networks. This has the potential to identify “hub” genes having synthetic lethal interactions with many cancer- mutated orthologs. If these synthetic lethal interactions are found to be conserved in human cells, these highly connected hub genes are potential targets for therapeutic development. Assembly of a synthetic lethal interaction network of yeast orthologs of 10 genes mutated in colorectal cancer, based on data in Saccharomyces cerevisiae, previously identified five such synthetic lethal hub genes in yeast. In this thesis, the evolutionary conservation of this network is interrogated in mammalian cells. The interactions between orthologs of colorectal cancer CIN genes in yeast were found to be highly conserved in human cells. A high- throughput assay to screen for small-molecule inhibitors of the protein encoded by one such gene, FEN1, was developed and used to identify 13 compounds that inhibited FEN1 in vitro with IC50 values in the low-micromolar range or less. These compounds were applied to cells bearing mutation in the tumor suppressor CDC4, and two compounds were found to yield selective killing of CDC4-deficient cells. Finally, yeast genetic techniques were used to  iii characterize CTF4, a second highly connected hub gene within the colon cancer CIN gene network, and to expand the therapeutic range of cancers that could be selectively killed by inhibitors of Ctf4/WDHD1 or Rad27/FEN1. Taken together, these data demonstrate the considerable power of applying model organisms genetics to the discovery of new anticancer therapeutic targets.   iv Preface A version of chapters 2, 3, and 4 has been submitted for publication. Derek M. van Pel, Irene J. Barrett,Yoko Shimizu, Babu V. Sajesh, Brent J. Guppy, Tom Pfeifer, Kirk J. McManus, and Phil Hieter. (2012) An Evolutionarily Conserved Synthetic Lethal Interaction Network Identifies FEN1 as a Broad-spectrum Target for Anticancer Therapeutic Development. I performed all experiments described in these chapters with the exception of Figure 2.1 (curated/publicly available data), and Figure 4.5.  A version of chapter 5 is in preparation for publication. Derek M. van Pel, Peter C. Stirling, Sean W. Minaker, Payal Sipahimalani, and Philip Hieter. (2012) Model organism genetics predict candidate therapeutic genetic interactions at the mammalian replication fork. I performed all experiments described in chapter 5, with the exception of Figure 5.9.   v Table of Contents Abstract.................................................................................................................................... ii Preface..................................................................................................................................... iv Table of Contents .................................................................................................................... v List of Tables ........................................................................................................................... x List of Figures......................................................................................................................... xi List of Abbreviations ........................................................................................................... xiii Acknowledgements ............................................................................................................... xv Dedication ............................................................................................................................ xvii Chapter  1: Introduction ........................................................................................................ 1 1.1 Cancer is a genetic disease........................................................................................ 1 1.1.1 Cancer is a multigenic disease .......................................................................... 1 1.1.2 Types of cancer mutations ................................................................................ 2 1.2 Targeting the cancer genotype .................................................................................. 5 1.3 The use of model organism genetics to study chromosome instability .................... 7 1.3.1 Model organisms and CIN................................................................................ 7 1.3.2 High-throughput genetics in Saccharomyces cerevisiae ................................ 11 1.3.3 The conservation of genetic interactions ........................................................ 13 1.4 Chemical-genetic screening and cancer.................................................................. 14 1.4.1 Chemical-genetic interactions......................................................................... 14 1.4.2 Small-molecule inhibitor screening ................................................................ 15 1.5 Research aims ......................................................................................................... 16  vi Chapter  2: Conserved synthetic lethal interaction networks reveal potential targets for anticancer therapeutic development ................................................................................... 18 2.1 Introduction............................................................................................................. 18 2.2 Methods................................................................................................................... 20 2.2.1 Cell culture...................................................................................................... 20 2.2.2 Western blotting.............................................................................................. 20 2.2.3 RNA interference ............................................................................................ 21 2.2.4 Synthetic lethal assays, cell imaging, and compound incubation................... 21 2.3 Results..................................................................................................................... 22 2.3.1 Investigating the conservation of synthetic lethal interactions ....................... 22 2.3.2 Observed interactions are not due to off-target effects................................... 27 2.3.3 Observed interactions are specific .................................................................. 27 2.4 Discussion............................................................................................................... 28 Chapter  3: A novel fluorescence-based assay for FEN1 activity reveals potent in vitro inhibitors................................................................................................................................ 30 3.1 Introduction............................................................................................................. 30 3.2 Methods................................................................................................................... 32 3.2.1 FEN1 purification ........................................................................................... 32 3.2.2 In vitro FEN1 inhibition assay........................................................................ 33 3.3 Results..................................................................................................................... 34 3.3.1 Design and characterization of the FEN1 assay ............................................. 34 3.3.2 Pilot screen...................................................................................................... 38 3.3.3 CCBN screen .................................................................................................. 41  vii 3.4 Discussion............................................................................................................... 44 Chapter  4: FEN1 inhibitors recapitulate synthetic lethal interactions in cancer cells.. 46 4.1 Introduction............................................................................................................. 46 4.2 Methods................................................................................................................... 47 4.2.1 Cell culture...................................................................................................... 47 4.2.2 Immunofluorescent labeling ........................................................................... 48 4.2.3 siRNA ............................................................................................................. 48 4.3 Results..................................................................................................................... 48 4.3.1 Screening for compounds active in vivo ......................................................... 48 4.3.2 The effect of in vitro FEN1 inhibitors on cell growth .................................... 49 4.3.3 NSC645851 and RF00974 recapitulate the synthetic lethal interaction between FEN1 and CDC4 ............................................................................................................. 50 4.3.4 RF00974 recapitulates the interaction between FEN1 and MRE11A ............. 55 4.3.5 RF00974 recapitulates DNA repair phenotypes of FEN1 .............................. 55 4.4 Discussion............................................................................................................... 59 Chapter  5: Expanding the therapeutic potential of CTF4 and RAD27.......................... 64 5.1 Introduction............................................................................................................. 64 5.2 Methods................................................................................................................... 65 5.2.1 Strains ............................................................................................................. 65 5.2.2 Two-hybrid assay............................................................................................ 65 5.2.3 CTF assay........................................................................................................ 66 5.2.4 Sister chromatid cohesion assay ..................................................................... 66 5.2.5 Random spore analysis ................................................................................... 66  viii 5.2.6 SGA and SDL screening................................................................................. 67 5.3 Results..................................................................................................................... 67 5.3.1 Characterizing Ctf4 function using a series of point mutations...................... 67 5.3.2 Putative Ctf4 phosphorylation does not contribute to DNA damage sensitivity or cohesion defects.......................................................................................................... 73 5.3.3 Exploiting protein interaction networks to yield alternative targets to Ctf4/WDHD1.................................................................................................................. 75 5.3.4 Expanding the range of mutations targeted inhibition of Ctf4/WDHD1 or Rad27/FEN1 ................................................................................................................... 79 5.3.5 A genome-wide synthetic dosage lethality screen suggests Rad27/FEN1 as a candidate target in novel cancer genotypes .................................................................... 86 5.4 Discussion............................................................................................................... 88 5.4.1 Associations between Ctf4 and Sld5 or Mms22 ............................................. 88 5.4.2 Using Ctf4 physical interactors to identify new drug targets ......................... 89 5.4.3 Expanding the therapeutic value of Ctf4/WDHD1 and Rad27/FEN1 as targets for small-molecule inhibition.......................................................................................... 90 Chapter  6: Conclusions and future directions .................................................................. 92 6.1 The conservation of genetic interactions ................................................................ 92 6.2 Optimization of FEN1 biochemical screening........................................................ 93 6.3 Additional studies with FEN1 inhibitors ................................................................ 94 6.4 Other targets for therapeutic development.............................................................. 96 6.5 Therapeutic implications......................................................................................... 97 References.............................................................................................................................. 99  ix Appendices........................................................................................................................... 113 Appendix A siRNA pool silencing in HCT116 cells........................................................ 113 Appendix B Random siRNA in HCT116 cells. ................................................................ 116 Appendix C Individual siRNAs in HCT116 cells............................................................. 117 Appendix D List of strains used in chapter 5.................................................................... 118 Appendix E Colony counts from random spore analysis of genetic interactions between Ctf4 physical interactors and mre11Δ, bub1Δ, and sgs1Δ. ............................................... 123 Appendix F Results of SGA screen with ctf4Δ................................................................. 126 Appendix G Results of SGA screen with rad27Δ. ........................................................... 130 Appendix H Results of SDL screen with rad27Δ. ............................................................ 132   x List of Tables Table 2.1. Antibodies employed in Western blots in this study. ............................................ 20 Table 2.2. Yeast and human gene orthologs. .......................................................................... 23 Table 5.1. Quantitaion of colony sectoring and chromatid separation assay for CTF4 mutants. ................................................................................................................................................. 72 Table 5.2. Quantitation of chromatid separation assay for Ctf4 phosphorylation site mutants. ................................................................................................................................................. 74 Table 5.3. Results of validation of SGA hits with ctf4Δ......................................................... 83 Table 5.4. Results of validation of SGA hits with rad27D. .................................................... 84   xi List of Figures Figure  1.1. The concept of synthetic lethality in cancer treatment. ......................................... 8 Figure  1.2. Synthetic lethal interactions can increase the therapeutic range of second-site targets...................................................................................................................................... 11 Figure  2.1. Investigating the conservation of CIN synthetic lethal interactions between S. cerevisiae and human cells. .................................................................................................... 24 Figure  2.2. Western blots indicating knockdown efficiency of pooled siRNAs. .................. 26 Figure  2.3. Western blots indicating knockdown efficiency of individual siRNAs. ............. 27 Figure  3.1. Principle of the FEN1 assay. ............................................................................... 35 Figure  3.2. Characterization of the rate and background fluorescence of FEN1 substrates. . 37 Figure  3.3. Determination of false positive rate. ................................................................... 39 Figure  3.4. Results from pilot screen. .................................................................................... 40 Figure  3.5. Dose-response curves of selected hits from pilot screen..................................... 41 Figure  3.6. Summary of CCBN screen. ................................................................................. 42 Figure  3.7. IC50 values of compounds selected for further investigation following the CCBN screen. ..................................................................................................................................... 43 Figure  4.1. Compounds from in vitro FEN1 inhibitor screens are active on cells. ............... 49 Figure  4.2. siRNA-mediated knockdown of FEN1 in HCT116 CDC4-/- cells recapitulates a synthetic lethal interaction. ..................................................................................................... 51 Figure  4.3. RF00974 and NSC645851 show differential killing of CDC4-deficient cells.... 54 Figure  4.4. NSC645851 and RF00974 selectively kill both CDC4+/- and CDC4-/- cells. ... 54 Figure  4.5. RF00974 recapitulates the synthetic lethal interaction between FEN1 and MRE11A. ................................................................................................................................. 56  xii Figure  4.6. siRNA-mediated knockdown increases 53BP1 focus formation in HCT116 cells. ................................................................................................................................................. 57 Figure  4.7. Treatment with RF00974 leads to 53BP1 focus formation, whereas treatment with NSC645851 does not. ..................................................................................................... 58 Figure  5.1. Schematic of mutations in CTF4 alleles.............................................................. 68 Figure  5.2. Yeast two-hybrid interactions between CTF4 alleles and Sld5 and Mms22....... 71 Figure  5.3. The response of CTF4 alleles to DNA-damaging drugs. .................................... 71 Figure  5.4. CTF4 phosphoserine mutants are not sensitive to DNA-damaging agents. ........ 74 Figure  5.5. Overexpression of FEN1 but not WDHD1 is toxic to yeast. .............................. 76 Figure  5.6. Results of random spore analysis crosses between CTF4 genetic and physical interactors................................................................................................................................ 78 Figure  5.7. Gene ontology (GO) term enrichment among high-throughput genetic interaction studies. .................................................................................................................................... 81 Figure  5.8. Independent validation of SGA synthetic lethal interactions. ............................. 85 Figure  5.9. Independent reconfirmation of synthetic dosage lethal interactions with rad27Δ. ................................................................................................................................................. 87   xiii List of Abbreviations 6-FAM 6-carboxyfluorescein ANOVA Analysis of variance BHQ-1 Black hole quencher-1 BSA Bovine serum albumin CCBN Canadian Chemical Biology Network CDRD Centre for Drug Research and Development CIN Chromosome instability DAmP Decreased abundance by mRNA perturbation DAPI 4’,6-diamidino-2-phenylindole DMSO Dimethyl sulfoxide DTT Dithiothreitol FPLC Fast protein liquid chromatography GFP Green fluorescent protein GO Gene ontology HEPES 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid HC-DIM High content digital imaging microscopy HA Hydroxyapatite HU Hydroxyurea IPTG Isopropyl β-D-1-thiogalactopyranoside KD2 Known Drug, a library of compounds at CDRD MIN Microsatellite instability MMS Methyl methanesulfonate PBS Phosphate-buffered saline PCNA Proliferating cell nuclear antigen PFA Paraformaldehyde psi Pounds per square inch RFC Replication factor C RPM Rotations per minute SDL Synthetic dosage lethality or Synthetic dosage lethal  xiv SEM Standard error of the mean SGA Synthetic genetic array shRNA Small hairpin RNA siRNA Short interfering RNA SC Synthetic complete (medium) SL Synthetic lethality SS Synthetic sickness SS/L Synthetic sickness/lethality YDV Yeast Derek van Pel; yeast strains in Derek van Pel’s personal collection YEP Yeast extract peptone (medium) YPD Yeast peptone dextrose (medium)   xv Acknowledgements I would like to thank the Natural Sciences and Engineering Research Council of Canada, the Michael Smith Foundation for Health Research, the University of British Columbia, and the Faculty of Graduate Studies for providing financial support to me directly during my studies. Additionally, this work, via operating grants to the Hieter laboratory, was funded by the Canadian Institutes of Health Research, the National Institutes of Health, and the Howard Hughes Medical Institute. I am also indebted to Richard Dean, Tom Pfeifer, and Yoko Shimizu at the Centre for Drug Research and Development, for their invaluable assistance in developing and deploying the FEN1 biochemical assay described in Chapter 3. I thank Tom Lakowski and Dylan Thomas in the laboratory of Adam Frankel (UBC) for assistance in purifying FEN1 protein. I thank Lata Balakrishnan in the laboratory of Robert Bambara (University of Rochester) for guidance in FEN1 assays and purification. Finally, I appreciate the ongoing assistance of Aruna Balgi in the laboratory of Michel Roberge (UBC) with regards to HC-DIM using the Cellomics VTI. I would like to thank Irene Barrett for her indispensable advice and assistance with the genetic interaction studies described in Chapter 2. As well, her steady hand at the helm of the lab during my years here were invaluable in making sure all of our projects ran smoothly. I am indebted to all of my colleagues for their support during my Ph.D. studies, but especially Nigel O’Neil and Peter Stirling for their insightful advice with regards to my experiments, and for helping in revisions to this thesis. Finally, I thank my friends and family for their ongoing support. More specifically, I thank my parents, for always giving me their ear; I thank Megan Filiatrault, my  xvi indefatiguable baymate, for indulging me in my many foibles over the years (and contributing a few of her own); and I thank Sean Minaker, my partner in crime (or at least, in terrible life choices) for always providing both comic relief and someone with whom to ruminate and commiserate.  xvii Dedication         To my parents, for giving me life; To my colleagues, for giving me wisdom; To my students, for giving me joy. 1  Chapter  1: Introduction 1.1 Cancer is a genetic disease 1.1.1 Cancer is a multigenic disease Cancer is a disease of uncontrolled cellular growth. Specifically, cancerous cells are characterized by their ability to evade apoptosis, to replicate in the absence of extrinsic signals to do so, to ignore replication inhibitory signals, to self-renew during cell division, to initiate and maintain angiogenesis in order to bring oxygen and nutrients to the growing tumor, and ultimately to leave their tissue of origin and invade distant tissues (reviewed in Hanahan and Weinberg, 2000).   Cancer has long been suspected of being multigenic, with mutations accumulating over the lifetime of the individual before finally giving rise to cancer (Nordling, 1953). This was dramatically demonstrated by Vogelstein and colleagues (1988) in a landmark paper that examined the progression of genetic lesions in colorectal cancer, wherein it was observed that certain genetic lesions, such as loss of the TP53 locus, were always preceeded by others, such as mutation of RAS (both discussed below). With the advent of various sequencing technologies, it has been established that a series of mutations are required to predispose cells towards the cancer phenotype (for reviews, see Stratton, 2011; Hanahan and Weinberg, 2011). The mutations observed in cancer fall into two categories: “driver” mutations are those that contribute to the cancer phenotypes previously discussed, thereby driving neoplasia, angiogenesis, and invasion; on the other hand, “passenger” mutations are those occurring in genes that do not contribute to carcinogenesis, and are instead a measure of how many mitotic cell divisions the cell has undergone during its clonal expansion (reviewed in Stratton, 2011). 2 1.1.2 Types of cancer mutations  There are three classes of genes that, when mutated, can drive tumorigenesis: inactivating mutations of tumor suppressor genes; activating mutations to turn proto- oncogenes into oncogenes; and mutation of genome stability genes (reviewed in Hanahan and Weinberg, 2011). Tumor suppressor genes are responsible for inhibiting cell growth – either by preventing advancement of the cell cycle, or inducing apoptosis in response to cellular stress. These genes are defined by the fact that loss-of-function is required to predispose to tumorigenesis. One example of a tumor suppressor gene is the transcription factor p53, encoded by the TP53 gene. Originally identified as an oncogene, due to the fact that the first isolated clones possessed dominant-negative mutations (reviewed in Lane and Benchimol, 1990), p53 was later reclassified as tumor suppressor following analysis of deletions of chromosome 17p recurrently found in malignant, but not benign, colorectal tumors (Vogelstein et al., 1988; Baker et al., 1989). p53 was subsequently shown to induce transcription of p21, a potent inhibitor of cyclin-dependent kinases, which has the effect of inhibiting cell cycle progression (el-Deiry et al., 1993; Harper et al., 1993; Gu et al., 1993; Xiong et al., 1993). p53-mutant mice are extremely susceptible to the development of tumors, even in a germline-heterozygous state (Donehower et al., 1992; Jacks et al., 1994). p53 was also found to conform to Knudson’s “two-hit” hypothesis for tumor suppressor genes (Knudson, Jr., 1971), which suggests that individuals born with a heterozygous inactivating mutation in a tumor suppressor gene are strongly predisposed to developing cancer, as somatic cells must undergo only one inactivation event, or “hit,” to predispose the cell to cancer. p53 germline-heterozygous mice often underwent loss-of-heterozygosity of the remaining functional allele of TP53 3 (Jacks et al., 1994). Characterization of a gene as a tumor suppressor or oncogene can also be dependent on the genetic context; for example, CDC4, generally classified as a tumor suppressor, has been shown to degrade pro-apoptotic factors, thus exhibiting oncogenic properties (Busino et al., 2012).  Proto-oncogenes encode proteins that play a normal role in cell and tissue development, but that can acquire gain-of-function mutations such that they become constitutively active, and drive the cell through the cell cycle. One example of a family of oncogenes is the RAS family, which are mutated in one-third of colorectal cancers (Bos et al., 1987) and widely mutated in a variety of tumors (reviewed in Pylayeva-Gupta et al., 2011). RAS proteins are responsible for transducing signals from cell surface receptors to intracellular effector pathways. RAS proteins are small GTPases, and will bind GTP in their so-called “on” conformation, and GDP in the “off” conformation. Other proteins, such as guanine nucleotide exchange factors and GTPase activating proteins, facilitate the transition of RAS proteins from off to on, and from on to off, respectively. The most common oncogenic mutations found in RAS proteins prevent cleavage of GTP to GDP, or prevent binding of GTPase activating proteins. Both of these types of mutations have the effect of locking RAS in the “on” conformation, and contribute to its ability to drive cell division (Trahey and McCormick, 1987; Vogel et al., 1988). RAS has been shown to provide cell growth signals, inhibit apoptosis signals, and promote angiogenesis (reviewed in Pylayeva-Gupta et al., 2011). Another example of an oncogene is the BRC-ABL fusion, occurring as a result of a gross chromosomal rearrangement between chromosomes 9 and 22. This results in the so-called Philadelphia 4 chromosome, where the genes BCR and ABL are joined to create the oncogenic BCR- ABL fusion protein (de Klein A. et al., 1982; Ben-Neriah et al., 1986).  The third class of cancer genes are genome stability genes, which exert their effect by reducing the fidelity of DNA replication, chromosome stability, or chromosome segregation. Thus, these genes are further classified as predisposing to either a mutator phenotype or a chromosome instability (CIN) phenotype (for reviews, see Loeb, 1994; Vogelstein and Kinzler, 2004). Tumors with an increased mutation rate often carry mutations in genes involved in mismatch repair, base excision repair, or nucleotide excision repair. An increased mutation rate increases the chance of gain-of-function mutations in proto-oncogenes, or loss-of-function mutations in tumor suppressor genes. For example, mutation of the mismatch repair gene MSH2 has been associated with colorectal cancer (Fishel et al., 1993; Peltomaki et al., 1993). On the other hand, CIN tumors display an increased rate of loss or gain of whole or parts of chromosomes, or of structural rearrangements or amplifications. CIN can have the effect of increasing the dosage of proto-oncogenes through amplification or multisomy, decreasing the dosage of tumor suppressor genes through gene deletion or monosomy, or promoting the creation of oncogenic fusion proteins by gross chromosomal rearrangment. For example, deletion of the yeast gene RAD27 results in an increase in the frequency of loss of a telomeric chromosomal segment via gross chromosomal rearrangement (Chen and Kolodner, 1999). Similarly, expression of a mutant form of the human spindle assembly checkpoint gene BUB1, an ortholog of the yeast CIN gene BUB1, found in CIN cancers could abrogate the spindle assembly checkpoint, suggesting a mechanism for the generation of aneuploidy in the CIN lines from which the alleles were derived (Cahill et al., 1998). 5 1.2 Targeting the cancer genotype The genetic distinctness of cancers relative to surrounding tissue can be leveraged towards selective killing of tumor cells. The ideal anticancer therapeutic would offer completely selective killing of cancerous cells, and would be completely nontoxic to normal cells; however, this is not the case for most standard anticancer therapies used today. The changes noted in the preceeding section, such as increased mutation rate or the propensity to ignore cell cycle checkpoints, may make cancer cells more sensitive to DNA damaging agents used in chemotherapy, such as bleomycin or temozolomide, or to radiation therapy. However, DNA damaging agents suffer many drawbacks: they are non-specific, and will negatively affect all dividing cells; because of this, they have side effects, such as causing nausea, diarrhea, and blood toxicity. In addition, DNA damaging agents are themselves mutagenic (for reviews, see Arnon et al., 2001; Gill et al., 2003). Because of these properties, traditional chemotherapeutics target healthy proliferative cells, in addition to cancer cells, and thus have a narrow therapeutic window. Selective killing of cancer cells can instead be achieved by targeting the specific genetic differences that distinguish a tumor cell from normal cells. These genetic differences often make the cancer cell dependent on the activity of a particular gene or pathway for continued viability; if the gene or pathway on which the cell was dependent for survival is mutated, the cell dies. This is known as synthetic lethality: when mutations in either of two genes are independently viable, but when present together in the same cell their combination results in inviability. Two common models presented to explain synthetic lethality are that (i) the two genes function in buffering parallel pathways, with each contributing to some process essential to viability, or (ii) the two genes encode 6 proteins that form part of the same essential complex, which is partially functional in the absence or partial-function mutation of either one of the genetic interacting partners, but whose function is obliterated by absence or partial-function mutation of both (reviewed in Boone et al., 2007). With respect specifically to the DNA damage response, synthetic lethality can arise if the absence of a gene leads to the accumulation of DNA damage, and the absence of an unlinked gene makes the cell deficient in correcting this damage; in the absence of damage, the response pathway is dispensable, and in the presence of damage, the response pathway can repair it. Knowing the mutations present in a tumor can thus be a starting point to identify synthetic lethal partner genes that can lead to rational drug or therapy design. A striking example of utilizing a synthetic lethal interaction to treat cancer comes from the synthetic lethal interaction that occurs between PARP and BRCA1/2. Mutations in BRCA1 and BRCA2 are predisposing to familial breast cancer (Hall et al., 1990; Miki et al., 1994; Wooster et al., 1994; Collins et al., 1995). The proteins BRCA1 and BRCA2 are involved in homologous recombinational repair of DNA damage (reviewed in Roy et al., 2011). PARP is responsible for covalent modification of a number of substrates in the cell, some of which contribute to DNA repair (reviewed in Helleday, 2011). When BRCA1 or BRCA2 are mutated, the cell becomes deficient in homologous recombinational repair, and consequently dependent on other, PARP-mediated break repair pathways for DNA repair. siRNA-mediated knockdown of PARP in the context of mutation of BRCA1 or BRCA2 was found to strongly inhibit cell proliferation. As expected, chemical inhibition of PARP was subsequently shown to be extremely toxic to cells with mutations in BRCA1 or BRCA2, but not normal cells. Using a chemical to 7 mimic a synthetic lethal genetic interaction in this way is known as a chemical-genetic interaction, and these studies provided a powerful example of exploiting a chemical- genetic interaction specific to cancer to cause selective killing (Farmer et al., 2005; Bryant et al., 2005). Since the PARP/BRCA1/2 interaction was first reported, chemical inhibition of PARP has been shown to be effective in killing cancer cells with a small number of other genetic changes (Mendes-Pereira et al., 2009; McLellan et al., 2012). 1.3 The use of model organism genetics to study chromosome instability 1.3.1 Model organisms and CIN More than 90% of solid tumors exhibit aneuploidy, likely due to an underlying CIN phenotype (reviewed in Weaver and Cleveland, 2006). Because faithful segregation of the genetic material at mitosis is crucial for cell viability, these sub-lethal hits in an essential process could in theory be leveraged to provide selective killing (Figure 1.1). Furthermore, due to its essential nature, genes responsible for ensuring proper chromosome segregation are highly conserved across evolution (for some examples of CIN genes shown to be conserved across evolution, see: Kohler et al., 1997; Williams and McIntosh, 2002; Merkle et al., 2003; Hou and Zou, 2005; Watrin et al., 2006; Gotter et al., 2007). For example, it has been shown recently that aneuploidy leads to a general cellular stress response (Torres et al., 2007), and aneuploidy itself may lead to further CIN (Sheltzer et al., 2011). Thus, identification of the human orthologs of CIN genes discovered in model organisms will yield candidate cancer CIN genes in human cells. 8  Figure  1.1. The concept of synthetic lethality in cancer treatment. If a second-site target can be discovered that is synthetic lethal with a cancer CIN gene, it is a target for anticancer therapeutic development. A second-site target synthetic lethal with many cancer CIN genes would potentially be a broad spectrum therapeutic.  The budding yeast Saccharomyces cerevisiae has been used extensively to study CIN, given its genetic tractability and the ease with which its genome can be manipulated. Many genes responsible for maintaining genome stability were discovered by random mutagenesis screens in yeast. For example, the MCM helicase complex was originally discovered in a screen for genes responsible for maintenance of a yeast plasmid (Maine et al., 1984), and a yeast colony-color assay based on missegregation of a nonessential chromosome fragment has been used to discover many proteins that were later found to be part of the kinetochore, replication machinery, or the sister chromatid cohesion complex (Spencer et al., 1990; Strunnikov et al., 1993; Michaelis et al., 1997; 9 Guacci et al., 1997; reviewed in Stirling et al., 2012). As well, yeast has been used to identify genes that, when mutated, increase the frequency of chromosomal rearrangements (Chen and Kolodner, 1999). The completion of whole-genome arrays of yeast mutants, beginning with the yeast nonessential gene knockout collection (Winzeler et al., 1999), greatly facilitated the development of comprehensive, high-throughput CIN screens in yeast. A screen of the ~4500 nonessential yeast genes revealed 130 genes that, when deleted, yielded a CIN phenotype (Yuen et al., 2007). Expansion of this screen in subsequent years, owing to the development of collections of temperature-sensitive and hypomorphic alleles of essential genes and the application of additional chromosome stability assays (Ben-Aroya et al., 2008; Breslow et al., 2008; Stirling et al., 2011; Li et al., 2011), combined with curated literature data, found that 692 genes – more than 10% of the ~6000 genes in the yeast genome – are mutable to CIN (Stirling et al., 2011; reviewed in Stirling et al., 2012). As stated previously, the human orthologs of CIN genes discovered in yeast can be identified and sequenced in human cancers as putative human cancer CIN genes. In previous studies, the human orthologs of 200 yeast CIN genes were identified and these genes were subjected to exome sequencing in a panel of colorectal tumor cells. These studies found that SMC1A, NIPBL, and BUB1 were each mutated in >2% of tumors, MRE11A and DING were each mutated in 4% of tumors, and CDC4 (also known as FBXW7 or FBW7) was mutated in >10% of tumors. These genes are the human orthologs of, in order, SMC1, SCC2, BUB1, MRE11, PDS1, and CDC4, and together these mutations account for nearly 25% of the mutations found in the tumors analyzed. These studies demonstrated the utility of taking a cross-species approach to candidate cancer 10 gene discovery (Cahill et al., 1998; Rajagopalan et al., 2004; Wang et al., 2004; Barber et al., 2008) and suggested that if a second-site therapeutic target could be discovered that had a synthetic lethal interaction with all of the genes listed above, discovery of a small- molecule inhibitor of this second-site target could result in a therapeutic to yield selective killing of one-quarter of colorectal tumors (Figure 1.2). 11   Figure  1.2. Synthetic lethal interactions can increase the therapeutic range of second-site targets. In this example, blue circles represent query mutations, yellow circles represent synthetic lethal partners, and the red circle is a common synthetic lethal partner of both query genes. If a common synthetic lethal interactor of CDC4 and MRE11A could be found, a small-molecule inhibitor of this target could be used to selectively kill 14% of colorectal cancers.  1.3.2 High-throughput genetics in Saccharomyces cerevisiae The ideal anticancer therapeutic would be active in a broad spectrum of tumors; however, identifying second-site targets via synthetic lethal interactions in human cells on a large scale is challenging. Recent advances in human whole-genome shRNA knockdown screens offer promise as a means to determine the entire synthetic lethal interaction network of human cells (Moffat et al., 2006), though these technologies are labor-intensive and require highly specialized infrastructure and expertise. For these reasons, use of this technology is currently limited to a handful of laboratories (for recent examples, see: Marcotte et al., 2012; Adamson et al., 2012). 12  In yeast, whole-genome mutant arrays allow for the interrogation of synthetic lethal interactions in a genome-wide manner with relative ease. Most often, the phenotype assayed in yeast is cell proliferation relative to wild type, with digitally- determined colony size used as a proxy. Shortly after the release of the nonessential yeast gene deletion collection, a number of technologies were developed to use the collection to screen for genetic interactions in high-throughput (Tong et al., 2001; Pan et al., 2004; Measday et al., 2005; Collins et al., 2007). The most popular such technology is synthetic genetic array (SGA). In SGA, a strain bearing a “query” mutation is mated to the entire haploid nonessential yeast deletion collection. Following meiosis, spore germination, and selection for single and double mutants, colony size is determined. If the proliferation rate of a double mutant, relative to that of wild type, is significantly less than the multiplied effects of the two independent mutations (following a so-called multiplicative model), then this represents a “negative” genetic interaction. If the colony size of the double mutant is greater than the neutral value predicted by the model, this represents a “positive” (sometimes known as an “alleviating”) genetic interaction (reviewed in Boone et al., 2007). The query strain used for SGA also contains genetic constructs allowing for selection of MATa haploids, and counterselection against diploids following meiosis (Tong et al., 2001). Another technology is synthetic dosage lethal (SDL) screening. Here, the query strain overexpresses a gene of interest, and is screened against the deletion array. The SDL effect can be due to the overexpressed protein further perturbing a pathway already disrupted in a deletion mutant, such as sequestering a buffering protein and unmasking a synthetic lethal interaction; alternatively, overexpression could alter the 13 subcellular localization of either the overexpressed protein or another protein, which, if not counteracted by the deleted protein, results in lethality (Kroll et al., 1996).  Currently available techniques in mammalian cells do not allow for the interrogation of the genetic interaction profile of large lists of genes; as well, the possibility of uncovering false-positives following a whole-genome screen against a single gene mutation in mammalian cells necessitates time-consuming validation to eliminate. On the other hand, the most significant advantage of targets identified in model organisms is that they form part of a genetic interaction network: rather than acting as a second-site target for a single mutation, their connectedness reinforces the validity of the target in multiple genetic contexts, thereby reducing the chance of a hit being a false- positive. 1.3.3 The conservation of genetic interactions The ideal cancer drug would be active against many cancer genotypes. Given that high-throughput genetic interaction screening in human cells is in its infancy, there are as yet no genes that have been identified to have widespread synthetic lethal interactions with multiple genes mutated in cancer. In yeast, using publicly available data to screen for synthetic lethal partners of eight yeast orthologs of cancer mutated genes yields a small number of genes, on the order of five, having synthetic lethal interactions with at least five yeast cancer orthologs (Yuen et al., 2007). If these synthetic lethal interactions are conserved from yeast to human cells, then these highly connected genes are potential targets for anticancer therapeutic development. The conservation of genetic interaction networks has been studied between S. cerevisiae, the fission yeast Schizosaccharomyces pombe, and the nematode worm 14 Caenorhabditis elegans, and is somewhat controversial. Studies have typically reported a high degree of conservation of interactions, on the order of 23%, between S. cerevisiae and S. pombe (Roguev et al., 2007; Dixon et al., 2009; Frost et al., 2012); however, conservation between S. cerevisiae and the metazoan C. elegans has been reported to be as low at 0.7% and as high as 43% (Fraser, 2004; Tarailo et al., 2007). The smaller number was observed in a genome-wide screen for genetic interactions, whereas the larger one was a result of comparisons of synthetic lethal interactions in one pathway. In another study, interactions between the C. elegans orthologs each of CTF4, RAD27, CTF8, CTF18, and DCC1 with the C. elegans orthologs of each of SMC1, SMC3, SCC1, and SGS1 were found to be highly conserved (>80%), suggesting a high degree of conservation in a CIN gene network (McLellan et al., 2009). Taken together, these data would seem to support the suggestion that interactions are highly conserved within complexes, less so in pathways and processes, and less so again between whole genomes – a so-called hierarchical organization of genetic interactions, representing the conservation of functions across evolution (Dixon et al., 2009; Ryan et al., 2012). 1.4 Chemical-genetic screening and cancer 1.4.1 Chemical-genetic interactions  A chemical-genetic interaction occurs when a small molecule mimics a mutation to induce synthetic lethality, as in the PARP inhibitor/BRCA1/2 mutation example above. Model organism genetics can be used to identify targets for anticancer therapeutic development. The idea of using yeast to identify second-site targets was first advanced by Hartwell and colleagues (1997), who advocated synthetic lethal screening using yeast 15 orthologs of cancer-mutated genes as queries to identify potential second-site targets for therapeutic development. 1.4.2 Small-molecule inhibitor screening  When a potential target for anticancer therapeutic development is identified, a screening strategy must be developed to identify inhibitors. Typically, inhibitor screening is either biochemical or cell-based. In biochemical screening, the protein of interest is purified, and its biochemical activity is assayed in the presence of chemicals from a compound library. In cell-based screening, compound is applied directly to cells, and a response is measured. Each method of screening has its advantages and drawbacks. Biochemical screening is useful in that the assay system is defined, and any change in the final readout must by definition be due to the presence of compound. However, biochemical screens require downstream secondary assays to determine whether biochemical inhibitors discovered in vitro are able to reach and inhibit their target in cells or in vivo, in addition to having a greater cost of material and labour. Cell-based assays inherently demonstrate when a test compound has an effect on cells; however, the concentrations of compound required to achieve a response tend to be higher. As well, a cell-based assay must be carefully designed with counterscreens to eliminate “hits” that yield the desired response by mechanisms unrelated to the system being studied, and to identify off-target effects (for reviews, see Inglese et al., 2007; Macarron and Hertzberg, 2009). High-throughput compound screening is used extensively to find novel therapeutics, and has resulted in numerous FDA-approved compounds for the treatment of HIV, diabetes, and cancer (reviewed in Macarron et al., 2011). 16 1.5 Research aims  Given that CIN is highly prevalent in cancer, that chromosome segregation is an essential process in all cells, and that a yeast CIN gene synthetic lethal interaction network reveals a small number of genes that have synthetic lethal interactions with many yeast cancer CIN gene orthologs, the hypotheses of this research are that: (i) Synthetic lethal interactions in a cancer-relevant CIN gene network are conserved between yeast and human cells; (ii) This synthetic lethal interaction network will reveal candidate targets for anticancer therapeutic development; and (iii) Small-molecule inhibitors developed in a biochemical assay can be used to recapitulate these synthetic lethal interactions in human cells. Given the above, the research aims for this thesis are as follows: 1. To determine, using siRNA-mediated knockdown of target human orthologs, whether synthetic lethal interactions in a CIN synthetic lethal interaction network are conserved between yeast and human cells. 2. To identify small-molecule inhibitors of a protein encoded by one of the highly- connected genes via in vitro biochemical screening. 3. To use small molecule inhibitors discovered by in vitro screening to recapitulate synthetic lethal interactions observed using siRNA-mediated knockdown. 17 Finally, CTF4 and RAD27 are two of the most highly genetically connected genes in the yeast genome. This led to the final research aim of this thesis: 4. To use yeast genetic assays to further probe the interactions and functions of these two genes.  18  A version of chapters 2, 3, and 4 has been submitted for publication. Derek M. van Pel, Irene J. Barrett,Yoko Shimizu, Babu V. Sajesh, Brent J. Guppy, Tom Pfeifer, Kirk J. McManus, and Phil Hieter. (2012) An Evolutionarily Conserved Synthetic Lethal Interaction Network Identifies FEN1 as a Broad-spectrum Target for Anticancer Therapeutic Development. Chapter  2: Conserved synthetic lethal interaction networks reveal potential targets for anticancer therapeutic development 2.1 Introduction As discussed previously, classical cancer treatments have traditionally sought to exploit differences between cancer cells and somatic cells, such as targeting the increased proliferation rate of cancer cells; however, these effects will also harm normal proliferating cells. More recently, as the genotypic differences between cancerous and noncancerous cells have been discovered, the possibility of exploiting specific somatic gene mutations to yield selective tumor killing has received widespread attention. For example, whole-genome shRNA screens have been used to find synthetic lethal interactions with known oncogenes (Luo et al., 2009; Molenaar et al., 2009); screening tumors with chemical libraries has yielded compounds that cause selective killing in the context of specific mutations (Dolma et al., 2003; Ji et al., 2009; Mendes-Pereira et al., 2009; Steckel et al., 2012); and specific knowledge of biological processes has been used successfully to predict individual synthetic lethal interactions, such as between PARP and BRCA1 and 2, as described previously (Farmer et al., 2005; Bryant et al., 2005). An alternative means to discover targets for anticancer therapeutic development is to take a cross-species candidate approach in a genetically tractable model organism. The yeast Saccharomyces cerevisiae lends itself to the identification of genetic interactions in a high-throughput manner, and a wealth of these data are publicly available (Tong et al., 2001; Pan et al., 2004; Tong et al., 2004; Collins et al., 2007; Costanzo et al., 2010). 19 S. cerevisiae has been used extensively by the Hieter laboratory to identify and characterize yeast genes mutable to chromosome instability (CIN) (Spencer et al., 1990; Kouprina et al., 1992; Mayer et al., 2001; Yuen et al., 2007; Ben-Aroya et al., 2008; McLellan et al., 2009; Stirling et al., 2011). As mentioned previously, CIN genes in yeast were used to identify cognate human candidate genes for sequencing in a panel of colorectal cancers, and a small number of genes was found to account for >25% of the spectrum of mutations in a panel of these tumors (Cahill et al., 1998; Rajagopalan et al., 2004; Wang et al., 2004; Barber et al., 2008). Publicly available high-throughput synthetic lethal data for nonessential genes was combined with direct tests using tetrad analysis of temperature-sensitive mutants of essential genes, for the yeast orthologs of these colon cancer CIN genes and other colorectal cancer-mutated genes. These data were used to generate a yeast synthetic lethal interaction network (Yuen et al., 2007; McLellan et al., 2009). Screening for synthetic lethal partner genes that had ≥5 synthetic lethal interactions with cancer CIN gene orthologs yielded a small number of highly genetically connected genes, such as the replication factor CTF4, the flap endonuclease RAD27, and members of the alternative RFCCtf18 complex. This suggests that, if these interactions are conserved in human cells, targeting any one of these genes for small- molecule inhibition could prove to be a broad-spectrum anticancer therapy. Before it can be determined whether the human orthologs of the yeast highly connected genes can become candidate drug targets, the individual synthetic lethal interactions within this CIN genetic interaction network must first be directly recapitulated in human cells. Here, I use siRNA-mediated knockdown of candidate synthetic lethal interacting partner genes in human cells to investigate the conservation of 20 synthetic lethal interactions. Quantitative determination of the effect of single and double knockdown of these genes on cell proliferation demonstrates that these interactions are conserved, and suggest that WDHD1, FEN1, and the alternative RFCCHTF8 represent attractive targets for anticancer therapeutic development. 2.2 Methods 2.2.1 Cell culture HCT116 cells were purchased from ATCC and were grown in McCoy’s 5A medium supplemented 10% FBS, at 37°C within a humidified, 5% CO2 atmosphere. 2.2.2 Western blotting Western blots were performed as detailed elsewhere (McManus et al., 2006; McManus et al., 2009). Antibodies used for Western blots are described in Table 2.1. Table 2.1. Antibodies employed in Western blots in this study. Protein target Supplier Catalog number Dilution CDC4 Abcam ab12292 1:1000 FEN1 Abcam ab462 1:1000 FEN1 Abcam ab17993 1:3000 MRE11A Abcam ab397 1:10 000 RAD54B Abcam ab83311 1:1000 RNF20 Abcam ab32629 1:3000 SMC1 Abcam ab9262 1:1000 SMC3 Abcam ab9263 1:1000 WDHD1 Sigma HPA001122 1:1000 Tubulin, alpha Abcam ab7291 1:10 000 Tubulin, alpha Abcam ab18251 1:10 000 Mouse IgG (HRP- conjugated) Jackson ImmunoResearch 111-035-144 1:10 000 Rabbit IgG (HRP- conjugated) Jackson ImmunoResearch 111-035-146 1:15 000 21 2.2.3 RNA interference  Subconfluent and asynchronous HCT116 cells were transiently transfected with ON-TARGETplus siRNA pools at a total siRNA concentration of 50 nM using DharmaFECT I (Dharmacon). Cultures were replenished with fresh medium 8-12 hours after transfection. 2.2.4 Synthetic lethal assays, cell imaging, and compound incubation HCT116 cells were harvested 24 hours after siRNA transfection and replated in 96-well optical bottom plates. HCT116 cells were fixed four days after transfection in 4% paraformaldehyde/PBS for 10 minutes. Nuclei were labelled with Hoechst 33342 at 500 ng/mL. Stained nuclei were counted using a Cellomics Arrayscan VTI fluorescence imager as described previously (McManus et al., 2009). Data were normalized to GAPDH-silenced controls. Two biological replicates were performed, each consisting of at least six technical replicates. To determine the presence of a synthetic lethal interaction, the proliferative defect was calculated, and is defined as [1 – ionproliferat Observed model tivemultiplica aby  predictedion Proliferat ] x 100% (Equation 2.1)  where the predicted proliferation was the product of the proliferation of the two individual gene knockdowns, following a multiplicative model of genetic interactions (Baryshnikova et al., 2010). Synthetic lethal interactions were scored as a proliferative defect of greater than three times the average SEM of an experiment: 5% x 3 = 15%. 22 2.3 Results 2.3.1 Investigating the conservation of synthetic lethal interactions The human genes SMC1A, SMC3, NIPBL, STAG3, RNF20, FBXW7/CDC4, MRE11A, RAD54B and BLM have been found to be mutated in colorectal cancer, and together account for nearly 30% of the CIN mutational spectrum of this disease (Hiramoto et al., 1999; Rajagopalan et al., 2004; Wang et al., 2004; Kemp et al., 2005; Barber et al., 2008). SMC1A, SMC3, NIPBL, and STAG3 are all members of the sister chromatid cohesion complex, responsible for holding newly-replicated DNA together from S phase to the onset of anaphase. RNF20 plays a role in the regulation of transcription and genome stability maintenance. CDC4 is an E3 ubiquitin ligase that controls cell cycle progression via regulation of cyclin E and oncoproteins such as c-myc and KLF5 (reviwed in Wang et al., 2012). Finally, MRE11A, RAD54B and BLM are DNA metabolism proteins that play a role in DNA repair via homologous recombination. Protein BLAST was used to identify the budding yeast orthologs of these human genes (Table 2.2) and, using literature and publicly available genetic interaction data (BioGrid and the Saccharomyces Genome Database;  Stark et al., 2006; Cherry et al., 2012), I constructed a yeast synthetic lethal interaction network (Figure 2.1A) (Tong et al., 2001; Pan et al., 2004; McLellan et al., 2009; Costanzo et al., 2010). The yeast orthologs of these cancer-mutated genes share numerous synthetic lethal interacting partners: the replication and cohesion-establishment gene CTF4; the 5’ flap endonuclease RAD27, and the members of the PCNA-unloading alternative replication factor C (RFC) complex, RFCCtf18, CTF8, CTF18, and DCC1. 23 Table 2.2. Yeast and human gene orthologs. Names indicated are the names used in this work. Names in parentheses indicate common alternative gene names. Members of the cohesin complex (and SCC2/NIPBL, a cohesin loader) are indicated by shading. Yeast gene Human ortholog(s) BRE1 RNF20 CDC4 CDC4 (FBXW7, FBW7) CTF18 CHTF18 CTF4 WDHD1 (AND1) CTF8 CHTF8 DCC1 DSCC1 MRE11 MRE11A RAD27 FEN1 RDH54 RAD54B SCC1 (MCD1) RAD21 SCC2 NIPBL SCC3 (IRR1) STAG1, STAG2, STAG3 SGS1 BLM, WRN SMC1 SMC1A SMC3 SMC3 24   Figure  2.1. Investigating the conservation of CIN synthetic lethal interactions between S. cerevisiae and human cells. 25 (A) CIN synthetic lethal interaction network in yeast. Red spots are yeast orthologs of genes mutated in colorectal cancer and blue spots are genes shown to have >4 interactions with cancer-mutated orthologs. Orthologs were identified using standard protein BLAST, and interactions were curated from the Saccharomyces Genome Database and BioGrid. (B-D) Representative genetic interaction data for combinations with FEN1. HCT116 cells were transfected with siRNAs targeting the indicated genes and treated as in Methods. All treatments were normalized to GAPDH-treated controls. Blue circles, siRNA targeting the gene indicated above the graph. Red circles, siRNA targeting the gene indicated along the x- axis. Green circles, siRNA targeting the gene indicated above the graph and the gene indicated along the x- axis. Yellow triangle, the predicted growth of the double siRNA-knockdown cells, assuming a multiplicative interaction. (E) Summary of genetic interaction experiments in HCT116. Grey solid line, interaction conserved between yeast and HCT116 cells. Green dashed line, interaction not conserved between yeast and HCT116 cells. Orange dotted line, interaction not predicted based on yeast data, but observed in HCT116 cells. For a table indicating the yeast and human orthologs, refer to Table 2.2.  To investigate the conservation of this network between yeast and a human cell line, I used siRNA-mediated knockdown of potential synthetic lethal gene pairs in the chromosomally stable colorectal cancer cell line HCT116. Knockdown efficiencies were evaluated when possible by Western blots (Figure 2.2). CHTF8 was selected as a representative member of the alternative RFCCtf18, as members of the same complex typically share synthetic lethal interactions (Collins et al., 2007). All pairwise combinations between the three “central” synthetic lethal partner genes (FEN1, WDHD1, and CHTF8) and the ten outer cancer-mutated CIN genes were evaluated for synthetic lethality (Figure 2.1B-D; Appendix A) by counting, via digital imaging microscopy, the number of cells remaining following transfection. GAPDH was used as a control to elicit a silencing response, and single knockdown experiments were supplemented with GAPDH siRNA to ensure that all treatments received the same total concentration of siRNA. 26  Figure  2.2. Western blots indicating knockdown efficiency of pooled siRNAs. GAPDH was used as a control siRNA to elicit a silencing response. Protein extracts were collected from HCT116 cells three days after transfection. Antibodies used are described in Table 2.1.  In the network tested, there exist 30 possible synthetic lethal interactions, and all pairwise combinations were interrogated. Among these 30 possible combinations, 22 synthetic lethal interactions were predicted based on yeast data (Tong et al., 2001; Pan et al., 2004; Tong et al., 2004; McLellan et al., 2009; Costanzo et al., 2010). 16/22 interactions (73%) were conserved between yeast and human cells, and 6 interactions (27%) did not appear conserved in this assay. One interaction, between FEN1 and STAG1 (one of three human paralogs of yeast SCC1, along with STAG2 and STAG3), that was not predicted based on yeast data, appeared in this mammalian cell assay (Figure 2.1E and Appendix A). No interactions were observed with STAG3, which functions primarily in human meiosis (Prieto et al., 2001). As in yeast, all three central genes – WDHD1, FEN1, and CHTF8 – were highly connected to sister chromatid cohesion genes (e.g., cohesin and/or cohesin loaders) (Figure 2.1E, Table 2.2), and were highly connected in general, suggesting that WDHD1, FEN1, and the alternative RFCCHTF18 are potential targets for small-molecule inhibition as an anticancer therapy. 27 2.3.2 Observed interactions are not due to off-target effects It is possible that the observed interactions can be attributed to off-target effects resulting from one of the four siRNA duplexes used in the pooled siRNAs. To add weight to the idea that this is not the case, interactions between FEN1 and MRE11A, and FEN1 and CDC4, were further investigated by deconvolving the pooled siRNAs. The individual siRNAs yielding the greatest knockdown were identified by Western blot (Figure 2.3), and were applied to HCT116 cells in combination in the same manner as above. In this experiment, the individual siRNAs were able to recapitulate the observed interactions (Appendix B), supporting the idea that off-target effects are not responsible for the observed interactions.  Figure  2.3. Western blots indicating knockdown efficiency of individual siRNAs. GAPDH was used as a control siRNA to elicit a silencing response. Protein extracts were collected from HCT116 cells three days after transfection. Antibodies used are described in Table 2.1. From these data, siFEN1-2, siMRE11A-4, and siCDC4-2 were selected for further studies (Appendix B).  2.3.3 Observed interactions are specific  It is also possible that knockdown of the three central genes sensitizes cells nonspecifically to further siRNA-mediated target knockdown. To determine whether this is the case, I transfected cells with siRNA targeting the central genes FEN1 and WDHD1 in combination with four genes randomly selected from the genome, in the same manner as described above. Here, no genetic interactions were observed (Appendix C), suggesting that knockdown of FEN1 or WDHD1 does not sensitize cells to further siRNA-mediated knockdown in a nonspecific manner. 28 2.4 Discussion In this chapter, I have demonstrated that interactions within a defined CIN synthetic lethal interaction network are conserved between yeast and cultured human cells. Furthermore, I have provided evidence that WDHD1, FEN1, and the alternative RFCCHTF18 are attractive targets for anticancer therapeutic development, owing to the broad spectrum of cancer genotypes such inhibitors could be used to target. S. cerevisiae is a useful starting point for cross-species genetic studies due to its genetic tractability, and the ease with which defined single and double mutant genotypes can be constructed and subsequently screened in a high-throughput manner (Tong et al., 2001; Pan et al., 2004). The work presented here expands upon past proof-of-concept experiments carried out by the Hieter laboratory to examine evolutionary conservation of synthetic lethal interactions as a means to develop anti-cancer therapy (McManus et al., 2009). In the current work I found nearly 73% (16/22) of the potential interactions, derived from yeast genetic interaction data and tested in human cells, were conserved. This high conservation of CIN synthetic lethal interactions is striking when compared to previously reported interactions on a genome-wide scale. Comparison of genome-wide screens between S. cerevisiae and C. elegans found 0.7% (6/387) interactions conserved (Tischler et al., 2008), whereas comparison of interactions between S. cerevisiae and S. pombe has previously shown approximately 23% conservation (54/240 interactions conserved) (Dixon et al., 2008). Tarailo (2007) found 43% (9/21) of synthetic lethal interactions with the spindle assembly checkpoint gene MAD1 were conserved between S. cerevisiae and C. elegans. In another study, the conservation of a CIN gene synthetic lethal network, containing many of the genes investigated here, was examined between S. 29 cerevisiae and C. elegans; there, 85% of interactions tested (17/20) were conserved (McLellan et al., 2009). A further study found 82% (23/28) of interactions conserved between a cohesin gene and replication fork accessory proteins (McLellan et al., 2012). It has been suggested that genetic interactions within processes fundamental to cellular viability may be highly conserved, whereas interactions on a genome-wide scale may not (Costanzo et al., 2010; Frost et al., 2012; Ryan et al., 2012). One possibility is that perturbing a fundamental process – such as the maintenance of chromosome stability – is tolerated by the cell only up to a certain threshold. Beyond this threshold, genome rearrangements and aneuploidy would likely occur at such a high frequency as to be incompatible with viability. This agrees with the observation that moderate aneuploidy and CIN correlate with more aggressive tumor growth, but hyperaneuploidy correlates with poor tumor growth (Birkbak et al., 2011; Lee et al., 2011). The concept of using the unique genetic profile of tumor cells relative to somatic cells to selectively kill cancer has received widespread attention in recent years. Hartwell and colleagues advocated the use of both genetic and chemical genetic studies in model organisms such as S. cerevisiae to investigate possible therapies in a well-defined system (1997). Because CIN is highly prevalent in cancer (Weaver and Cleveland, 2006), and ensuring faithful chromosome segregation is an essential process, this fact can be leveraged to target tumors specifically. The work presented here demonstrates that the colon cancer CIN gene synthetic lethal interaction network analyzed in this study is highly conserved and can predict targets for therapeutic development, such as WDHD1, FEN1, and the alternative RFCCHTF18.  30   A version of chapters 2, 3, and 4 has been submitted for publication. Derek M. van Pel, Irene J. Barrett,Yoko Shimizu, Babu V. Sajesh, Brent J. Guppy, Tom Pfeifer, Kirk J. McManus, and Phil Hieter. (2012) An Evolutionarily Conserved Synthetic Lethal Interaction Network Identifies FEN1 as a Broad-spectrum Target for Anticancer Therapeutic Development. Chapter  3: A novel fluorescence-based assay for FEN1 activity reveals potent in vitro inhibitors 3.1 Introduction  As shown in the previous chapter, FEN1, WDHD1, and the alternative RFCCHTF18 are all attractive targets for anticancer therapeutic development, as they have many synthetic lethal interactions with CIN genes mutated in cancer. As such, it is of interest to attempt to develop potent inhibitors of these proteins.  In yeast, the alternative RFCCtf18 is composed of the proteins Ctf8, Ctf18, and Dcc1, in addition to the four protein subunits, Rfc2-5, that are shared among all of the RFC (Replication Factor C) complexes (Hanna et al., 2001; Mayer et al., 2001; Ben-Aroya et al., 2003). Orthologs of these proteins have been identified in human cells (Merkle et al., 2003), though most functional studies have been carried out in yeast. Loss of RFCCtf18 function has been shown to lead to a loss of sister chromatid cohesion (Hanna et al., 2001; Mayer et al., 2001; Petronczki et al., 2004). In the fission yeast Schizosaccharomyces pombe, the alternative RFCCtf18 has been implicated in stablizing replication forks during times of replication stress (Ansbach et al., 2008). It has been posited that RFCCtf18 plays a role in stabilizing translesion synthesis (Gellon et al., 2011); this is possibly achieved by removing the replication processivity factor PCNA from RPA-coated single-stranded DNA (Bylund and Burgers, 2005).  31  WDHD1 (WD repeat and HMG box DNA binding protein 1) is the human ortholog of yeast Ctf4 (Chromosome Transmission Fidelity 4). Ctf4 was first isolated as a protein required for faithful chromosome segregation (Spencer et al., 1990), and has since been shown to be a component of the replisome whose deletion leads to deficiencies in the efficient establishment of sister chromatid cohesion (Lengronne et al., 2006; Gambus et al., 2006). WDHD1 has also been implicated in cohesion establishment in mammalian cells (Bermudez et al., 2010). The roles of Ctf4 and WDHD1 in cohesion establishment have been attributed to their roles in mediating the assembly of various subcomplexes involved in DNA replication, such as GINS, MCM, and primase (Zhu et al., 2007; Gambus et al., 2009; Im et al., 2009; Tanaka et al., 2009a; Tanaka et al., 2009b; Wang et al., 2010). WD repeats are proposed to form β-propeller structures, providing a large surface area to mediate multiple protein-protein interactions (reviewed in Jawad and Paoli, 2002). This is consistent with a role for Ctf4/WDHD1 as a replication-mediating scaffolding protein.  FEN1 (Flap Endonuclease 1) is the human ortholog of Rad27. FEN1/Rad27 cleaves 5’ DNA flaps that occur during DNA replication and repair (Bambara et al., 1997; Greene et al., 1999). FEN1 is involved in preventing trinucleotide repeat expansion (Liu and Bambara, 2003), and plays a critical role in long-patch base excision repair (Liu et al., 2005) and replication fork restart (Zheng et al., 2005). Rad27, like Ctf4 and Ctf8/Ctf18/Dcc1, is a CIN gene (Yuen et al., 2007), and heterozygous knockout of FEN1 in mice leads to inflammation and increased tumor formation (Zheng et al., 2007).  While all of these proteins/complexes are valid therapeutic targets for small-molecule inhibitor screening, WDHD1 appears to be a large scaffolding protein, and it is difficult to envision a screening strategy to measure inhibition of PCNA loading onto DNA by the  32 RFCCHTF18. Thus, I chose to focus on screening for inhibition of FEN1 flap endonuclease in vitro, as its enzymatic function is well-characterized, and there exists a low-throughput, quantitative in vitro assay that can measure FEN1 activity. In this chapter, I develop a fluorescence-based FEN1 activity assay and demonstrate that it is amenable to high- throughput screening. I carry out two screens that assess approximately 30 000 compounds, and discover 13 potent small molecule inhibitors of FEN1 activity. 3.2 Methods 3.2.1 FEN1 purification His-tagged FEN1 was expressed in BL21 E. coli from pET28b(+) (a generous gift from R. Bambara, University of Rochester). An overnight culture of bacteria was diluted to OD600 of 0.1, and grown to OD600 of 0.6 before IPTG was added to a final concentration of 1 mM. Bacterial expression was induced for three hours. Bacteria were lysed in lysis buffer (50 mM NaH2PO4, 300 mM NaCl, 10 mM imidazole, pH 8.0) containing 2X protease inhibitor cocktail via a French press at 10 000 psi. The lysate was clarified by centrifugation at 30 000 rpm and 4°C for 10 min and passed through a 0.22 μm filter before being loaded onto a HisTrap FF column (1 mL, GE Healthcare) in an ÄKTAFPLC P-920 system (GE Healthcare). The column was washed in 10 volumes of wash buffer (lysis buffer + 20 mM imidazole), and FEN1 was eluted with 5 volumes of elution buffer (lysis buffer + 125 mM imidazole). The lysate was diluted with 9 volumes HI buffer (30 mM HEPES-KOH, 0.5% myo-inositol, pH 7.8) with 30 mM NaH2PO4 and concentrated in a protein concentrator (Amicon). It was then loaded onto a hydroxyapatite resin (HA Ultrogel, Pall Life Sciences). The hydroxyapatite resin was washed with 10 volumes of HI-30 mM NaH2PO4, and FEN1 was eluted with 5 volumes of HI-200 mM NaH2PO4. The eluate was diluted with 5 volumes  33 HI-30 mM KCl prior to concentration, and then loaded onto a strong cation exchange column (1 mL HiTRAP SP FF FPLC, GE Healthcare Life Sciences). The column was washed with 10 volumes of HI-30 mM KCl, then 10 volumes of HI-200 mM KCl, and FEN1 was eluted with a gradient from HI-200 mM KCl to HI-500 mM KCl over 10 column volumes. Purified FEN1 was concentrated in FEN1 dilution buffer (30 mM HEPES-KOH, 5% glycerol, 0.1 mg/mL BSA, 0.01% NP-40), and aliquots of known concentration were frozen at -80°C. A 2 L culture produced approximately 12 μmol of protein, sufficient for 2 x 109 assays at 6 fmol/assay. 3.2.2 In vitro FEN1 inhibition assay Oligonucleotides used were as follows: “template”, 5’- GGTGGACGGGTGGATTGAAATTTAGGCTGGCACGGTCG-3’, “upstream”, 5’- CGACCGTGCCAGCCTAAATTTCAATC-3’, “downstream”, 5’-6-FAM- CCAAGGCCACCCGTCCAC-BHQ-1-3’. The three oligonucleotides were annealed at equimolar amounts in annealing buffer (50 mM Tris, 50 mM NaCl, 1 mM DTT, pH 8.0). FEN1 assays were carried out with 6 fmol FEN1 and 20 nM annealed substrate in FEN1 buffer (50 mM Tris pH 8.0, 30 mM NaCl, 8 mM MgCl2, 0.1 mg/mL BSA, 2 mM DTT) in 384-well black-bottom microtitre plates. Assays were set up by pinning compound or DMSO (negative control) in 25 μL of FEN1 buffer containing enzyme and incubating at room temperature for at least 30 minutes. Reactions were initiated by addition of 25 μL of FEN1 buffer containing substrate. Assays were carried out at room temperature and kinetic reads were taken over approximately ten minutes in a Varioskan fluorescence plate reader (Thermo Fisher Scientific), using excitation and emission wavelengths of 492 nm and 517 nm, respectively.  34 The Z’ for the screens was calculated using the formula |ˆˆ| )ˆˆ(3 1 np np μμ σσ − −−  (Formula 3.1) where pσ̂  is the standard deviation of the positive controls, nσ̂  is the standard deviation of the negative controls, pμ̂  is the mean of the positive controls, and nμ̂  is the mean of the negative controls (Inglese et al., 2007). 3.3 Results 3.3.1 Design and characterization of the FEN1 assay  FEN1 cleaves overhanging 5’ DNA flaps. Other groups have devised in vitro FEN1 assays based on radiolabeling of the 5’ flap on a synthetic, three-oligonucleotide substrate (Bornarth et al., 1999; Liu and Bambara, 2003). The products are run on high-percentage polyacrylamide gels, and detected by autoradiography. However, radiodetection is not amenable to high-throughput screening, so I designed a FEN1 assay based on fluorescence detection in a microtitre plate. Based on a previous synthetic oligonucleotide substrate design (Liu and Bambara, 2003), I conceived of the substrate shown in Figure 3.1. This assay consists of an “upstream” nucleotide, bearing a one-nucleotide 3’ flap, a “downstream” oligonucleotide, bearing the 5’ flap to be cleaved by FEN1, and a “template” oligonucleotide, to which the other two oligonucleotides are annealed. In this assay, the quencher molecule is covalently linked to the downstream oligonucleotide, and quenches fluorescence of the fluorophore, linked to the 5’ flap, until FEN1 cleaves the flap. After cleavage, the fluorophore can diffuse away from the quencher and fluoresce (Figure 3.1).  35  Figure  3.1. Principle of the FEN1 assay. FEN1 cleaves the 5’ flap to which the fluorophore is attached, allowing it to diffuse away and fluoresce. 6- FAM, 6-carboxyfluorescein. BHQ-1, black hole quencher 1.  I first sought to determine whether purified FEN1 would show concentration- dependent activity on the annealed substrate. A kinetic assay was assembled with varying amounts of FEN1 acting on 100 nM substrate. Given the assay design (Figure 3.1), fluorescence would be expected to increase with time. As I sought ultimately to find compounds that decrease the rate of this fluorescence increase, I opted to carry out kinetic assays, measuring the change in fluorescence over time. I found that FEN1 showed concentration-dependent activity on 100 nM substrate  (Figure 3.2A). At the high concentrations of FEN1 used in initial optimization, the assay was nearing completion before scanning could commence; thus, in subsequent experiments I used lower amounts of FEN1 in each assay, settling on 6 fmol/assay as the optimal amount. I also noted a high background fluorescence reading that contributed to a low signal-to-noise ratio. I reasoned that the distance between the fluorophore and the quencher could be such that fluorescence quenching was not optimal. I subsequently redesigned the substrate to reposition the quencher closer to the fluorophore by shortening the distance, in nucleotides, of the annealed region of the fluorophore-bearing oligonucleotide. I found that this modification was  36 successful at reducing the background fluorescence, as the fluorescence reading at the first time point was lower (Figure 3.2B). Although it is theoretically possible to shorten the oligonucleotide further, it is likely that doing so would have compromised the annealing of the substrate, due to a smaller length of complementary base-pairing. As an alternative, I also designed a substrate with the quencher covalently linked to the 3’ one-nucleotide flap of the upstream oligonucleotide, but found that FEN1 was not able to cleave this substrate. It has subsequently been shown that the 3’ one-nucleotide flap is crucial for FEN1 substrate recognition and binding (Finger et al., 2009; Tsutakawa et al., 2011; Gloor et al., 2012).  37  Figure  3.2. Characterization of the rate and background fluorescence of FEN1 substrates. FEN1 assays were carried out as described in Methods, using 100 nM substrate and the indicated amount of purified FEN1. Substrate is shown on the left, and the change in fluorescence over time is shown on the right. (A) Substrate modeled on (Liu and Bambara, 2003). (B) Substrate modified from A, with fluorophore and substrate brought closer spatially. (C) Substrate with quencher on 1 nucleotide 3’ flap.  38 3.3.2 Pilot screen  Initial optimization of the screen was carried out by conducting FEN1 assays in a 384-well microtitre plate in the absence of library compounds. I found that the false positive rate (i.e., the rate of observing wells in which the rate of reaction was near-zero) was approximately 2% (Figure 3.3). Following optimization, a pilot screen for FEN1 inhibitors was carried out using the Centre for Drug Research and Development (CDRD) Known Drug (KD2) and Investigator libraries, which contained approximately 5500 compounds. The KD2 library was composed of previously characterized, FDA-approved drugs and compounds. The investigator library was composed of largely unrelated compounds, ordered by CDRD for other projects. Assays were carried out using a compound concentration of approximately 7 μM, following pinning from library plates at 5 mM.   After carrying out the pilot screen, I set a “hit” cutoff of 70% inhibition. 112 compounds initially met this criterion. (Figure 3.4A). The Z’ factor for the screen was found to be 0.53 (Figure 3.4B). (Z’ is a measure of the difference in the standard deviations between the positive and negative controls and the ability of a high-throughput screen to differentiate a hit from background noise, and Z’ values of >0.5 indicate an excellent assay (Iversen et al., 2006). The screen was carried out with a substrate concentration below the Km of the system, to enrich for competitive inhibitors (Figure 3.4C; Macarron and Hertzberg, 2009). It is possible for a test compound to quench fluorescence, and thus appear as a false positive due to a decrease in fluorescence signal. As a counterscreen for such false positives, I used a substrate bearing a fluorophore, but not a quencher. Although no change in fluorescence with time would be observed using this substrate, quenching compounds should  39 lead to raw fluorescence values significantly lower than those of control wells. Counterscreening, combined with susequent retesting of hits, left 27 compounds that inhibited the FEN1 reaction >70%. These compounds were tested for concentration- dependent FEN1 inhibition, of which 11 showed irreversible inhibition (which were eliminated due to the possibility of reacting nonspecifically in a cellular context), 14 showed concentration-dependent FEN1 inhibition, and 2 were eliminated due to fluorescence not detected in previous steps or non-concentration-dependent behavior. Of these remaining 14 compounds, 5 were selected for further testing and validation based on structural diversity (Figure 3.5). I used a previously described, highly potent inhibitor of FEN1 flap endonuclease activity (Tumey et al., 2005) as an internal control (Figure 3.5, upper-left panel). Interetingly, the hits from the screen were all polycyclic aromatic compounds, with hydrogen bond donors and acceptors reminiscent of DNA bases. This suggests a possible means of binding to FEN1, via interaction with the DNA-binding or active sites.  Figure  3.3. Determination of false positive rate. Change in relative fluorescence units with time is shown on the y-axis. Pos. Ctrl, wells with no FEN1 added (to simulate total inhibition). Neg. Ctrl, wells with no compound added (to simulate no inhibition).  40  Figure  3.4. Results from pilot screen. (A) Cutoff for hit designation was set at 70% inhibition. (B) Z’ factor calculated for all plates in the pilot screen. The average across all plates was 0.53. (C) A kinetic assay determined Km of this system to be approximately 25 nM. Error bars, SEM.  41   Figure  3.5. Concentration-response curves of selected hits from pilot screen. Structures and in vitro IC50 values are shown. Tumey 16 is compound 16 from (Tumey et al., 2005). Error bars, SEM.  3.3.3 CCBN screen  Next, the Canadian Chemical Biology Network (CCBN) library of approximately 25 000 compounds was screened as above. The average Z’ factor for this screen was 0.64 (Figure 3.7A), and percent inhibition data for all compounds screened (excluding autofluorescent compounds) is shown in Figure 3.7B. 279 compounds initially appeared as hits. Following counterscreening and retesting as above, 93 compounds had >90% inhibition. Screening out compounds that appeared in a previous screen by CDRD for inhibitors of another DNA metabolism enzyme (and thus, that are most likely to be non-specific), and focusing on compounds with drug-like properties, such as a small molecular weight and low number of hydrogen bond donors and acceptors to facilitate passage of the compound through the lipid bilayer (Lipinski et al., 2001), yielded 32 compounds that were selected for concentration-dependence assays. All 32 of these compounds displayed concentration-  42 dependent FEN1 inhibition. Of these, 8 compounds were selected for further characterization based on structural diversity and drug-like properties (Lipinski et al., 2001), and their IC50 values were determined (Figure 3.8). As with the pilot screen, all of the hits from the CCBN screen are polycyclic aromatic compounds. Some compounds, as above, bear electronegative groups that could be envisaged to bind FEN1 in a similar manner to DNA nucleotides (e.g., CD00862, CD08758, RF00974). However, other compounds (e.g., 5405900, RDR00966) bear structural resemblance to known DNA intercalators such as ethidium bromide and berberine. This suggests that these compounds may be interfering with FEN1 activity by binding to the substrate, rather than the enzyme; further testing would be required to differentiate between these two possibilities.  Figure  3.6. Summary of CCBN screen. (A) Z’ score for each plate in the CCBN screen. Average Z’ for the screen was 0.64. (B) Percent inhibition of all compounds tested, excluding autofluorescent compounds.   43  Figure  3.7. IC50 values of compounds selected for further investigation following the CCBN screen. FEN1 assays were carried out as in Methods.  44 3.4 Discussion  In this chapter, I devised and utilized a fluorescence-based in vitro assay for FEN1 activity. I used this assay to screen 30 000 compounds and found 13 compounds that inhibited FEN1 activity in a dose-dependent manner, with IC50 values in the low micromolar to mid-nanomolar range.  Others have screened for in vitro FEN1 inhibitors. Tumey and colleagues devised a fluorescence-based assay for FEN1 activity and, following combinatorial chemistry of primary hits, reported 26 compounds selective for FEN1 inhibition versus the closely-related endonuclease XPG (Tumey et al., 2004; Tumey et al., 2005). In another study, compounds were docked in silico to FEN1’s Mg2+ binding pocket, and compounds modeled to have high affinity were subsequently tested in a conventional radiolabel assay (Panda et al., 2009). While I was developing the assay, another fluorescence quenching-based FEN1 assay was reported, though this study did not carry out a screen for FEN1 inhibitors (Dorjsuren et al., 2010).  FEN1, in addition to its flap endonuclease activity, also possesses 5’ exonuclease and 5’ gap endonuclease activity. The gap endonuclease activity cleaves at the 5’ end of ssDNA- dsDNA interfaces and has been shown to be the primary activity of FEN1 responsible for the maintenance of trinucleotide repeats (Liu and Bambara, 2003; Singh et al., 2007). Loss of gap endonuclease activity, but retention of flap endonuclease activity, is also commonly found in FEN1 mutants in tumors, and mice engineered to contain these mutations are predisposed to autoimmune disorders, inflammation, and cancer (Zheng et al., 2007). The fact that mutants having lost all three of the enzymatic activities of FEN1 have not been observed in cancers, along with the observation that FEN1-/- mice or MEFs cannot be isolated  45 (Larsen et al., 2003) suggests that flap endonuclease activity below a certain threshold becomes incompatible with viability in mammals. This study did not investigate whether the FEN1 inhibitors isolated here inhibit any of the activities of FEN1 other than its flap endonuclease activity. As well, it has not yet been determined why abrogation of gap endonuclease activity is found in cancer. If it can be shown that specific inhibition of FEN1 gap endonuclease activity (either via small-molecule inhibition or mutation) leads to chromosome instability (CIN), then an inhibitor of gap endonuclease activity could itself selectively kill CIN tumors, as it is reasonable to predict that CIN beyond a certain threshold will lead to cell death.  High-throughput screening of diverse compound libraries has become a mainstay of pharmaceutical drug discovery, and has led to the development of highly successful therapeutics to treat a variety of conditions (reviewed in Macarron et al., 2011). The development of a simple and inexpensive fluorescence-based assay for FEN1 activity, and the demonstration that this assay is amenable to screening tens of thousands of compounds in a library suggest that the screening of millions of compounds in a large pharmaceutical company-curated library is within reach. Given the existence of related members of the FEN1 superfamily, such as XPG, GEN1, and EXO1 (Tsutakawa et al., 2011), it would be of interest to develop complementary assays for these proteins so as to demonstrate specificity before moving on to in vivo validation of compound efficacy. Regardless, the highly connected nature of FEN1 in cancer CIN gene SL networks, as demonstrated in Chapter 2, suggests that FEN1 inhibition has promise as anti-cancer therapeutic strategies. 46   A version of chapters 2, 3, and 4 has been submitted for publication. Derek M. van Pel, Irene J. Barrett,Yoko Shimizu, Babu V. Sajesh, Brent J. Guppy, Tom Pfeifer, Kirk J. McManus, and Phil Hieter. (2012) An Evolutionarily Conserved Synthetic Lethal Interaction Network Identifies FEN1 as a Broad-spectrum Target for Anticancer Therapeutic Development. Chapter  4: FEN1 inhibitors recapitulate synthetic lethal interactions in cancer cells 4.1 Introduction  Having isolated a number of potent inhibitors of FEN1 activity in vitro, my next goal was to determine which, if any, of these compounds are active in human cells. A compound that is active in vitro may be inactive in a cell-based assay for a variety of reasons: the compound may not be able to enter the cell; the compound may enter the cell but be actively pumped out; the compound may enter the cell but be inactivated by intracellular enzymes; finally, the compound may enter the cell but somehow be unable to reach its target. Furthermore, cell-based screening is essential to validate in vitro biochemical hits, and a secondary assay must be designed such that the downstream readout will be as specific to the target as possible (reviewed in Inglese et al., 2007).  Determination of whether or not a lead compound has activity in cells is typically straightforward; however, determination of specificity of response is more challenging (reviewed in Inglese et al., 2007). In yeast, technologies exist where chemical probes can be screened against genome-wide deletion or overexpression collections, which can determine with accuracy the specificity of a probe for a target (Hoon et al., 2008). However, as yet, no such technologies exist in human tissue culture. Therefore, it is important to determine whether or not putative target inhibition phenocopies loss of the target in cells and in vivo. The more phenotypes that can be shown to be recapitulated by target inhibition, the stronger  47 the body of evidence that the target of interest is indeed being inhibited, and the less likely it is that the effect is due to off-target inhibition or a widespread perturbation of the cell.  Interestingly, it is also possible for a chemical inhibitor to be specific for its target in vivo, but cause additional phenotypes not predicted by depletion of the target. For example, chemical inhibition of PARP, as compared to siRNA-mediated knockdown, has been shown to be more toxic to cells following DNA damage due to irreversible, covalent inhibition (Strom et al., 2011), suggesting that in some cases, an inhibitor would have a greater effect than a loss-of-function mutation. In this chapter, I demonstrate that a number of FEN1 inhibitors discovered in the previous chapter are effective at inhibiting cell growth. I then show that two compounds, NSC645851 and RF00974, are able to recapitulate the synthetic lethal interaction between FEN1 and CDC4, using matched CDC4+/+ and CDC4-/- cell pairs. As well, the more potent of the two compounds, RF00974, was able to recapitulate the synthetic lethal interaction with MRE11A. Finally, I show that RF00974, but not NSC645851, causes an increase in DNA damage, as evidenced by 53BP1 focus formation, phenocopying siRNA-mediated knockdown of FEN1. 4.2 Methods 4.2.1 Cell culture HCT116 cells were purchased from ATCC. HCT116 derivatives, DLD-1 and DLD-1 derivatives were gifts of Bert Vogelstein (Johns Hopkins University). 53BP1-mCherry HCT116 cells were a gift of Sam Aparicio (UBC). All cells were grown in McCoy’s 5A medium with 10% FBS. During compound incubation experiments, cells were incubated in compound of interest in 96-well optical bottom plates for approximately three days prior to  48 fixation in 4% paraformaldehyde/PBS for 10 minutes and counterstaining with 500 ng/μL Hoechst 33342/PBS. Plates were imaged using a Cellomics ArrayScan VTI. 4.2.2 Immunofluorescent labeling  Cells on cover slips were fixed in 4% paraformaldehyde for 10 minutes (5 minutes if the cell expressed a fluorescent protein) and permeabilized in 0.5% Triton-X-100/PBS for 10 minutes. Cells were mounted in Vectashield mounting medium containing 0.5 μg/mL DAPI and imaged on a Zeiss Axioplan microscope with appropriate filters. 4.2.3 siRNA siRNA transfections were carried out as described in Chapter 2. 4.3 Results 4.3.1 Screening for compounds active in vivo  Given the potential for in vitro hits to be inactive in cells, I established an iterative work flow to isolate compounds exhibiting a potent effect on cells from the hit list of 13, which progressed from the most crude, most broad, and least time-consuming assays of FEN1 activity, to more sophisticated and time-consuming assays. Given the essential nature of FEN1 in mammalian cells (Larsen et al., 2003) and my own observations that siRNA- mediated FEN1 knockdown inhibits cell growth (Chapter 2), I first elected to determine whether compounds had a dose-dependent effect on cell growth. If this was found to be the case, then the compound was applied to an independent cell line, to rule out cell-line specific genotype effects. Next, compounds were assessed for their ability to recapitulate the synthetic lethal interaction between FEN1 and CDC4 observed with siRNA-mediated knockdown, using matched cell pairs in which CDC4 had been targeted for inactivation.  49 Finally, the remaining FEN1 inhibitors were assayed for their ability to cause an increase in 53BP1 foci, indicative of increased DNA damage. 4.3.2 The effect of in vitro FEN1 inhibitors on cell growth I first asked whether compounds had a dose-dependent effect on the viability of HCT116 cells (Figure 4.1A, 4.1C), and investigated this by fixing cells after three days’ incubation in compound and using digital imaging microscopy to count the number of cells (represented by nuclei) remaining. A cutoff of EC50 < 15 μM was set, and compounds meeting this cutoff were tested in the same assay for activity in the DLD-1 cell line, an independent colorectal tumor cell line, to eliminate effects that might be specific to the HCT116 cell lines (Figure 4.1B, 4.1D). NSC645851, RF00974, RDR00966, and DP00966 were all shown to have EC50 values of < 15 μM on both cell lines.  Figure  4.1. Compounds from in vitro FEN1 inhibitor screens are active on cells. (A) Effect of compounds from pilot screen on HCT116 cells. (B) Effect of compounds from pilot screen having EC50 < 15 μM on HCT116 cells, on DLD-1 cells. (C) Effect of compounds from CCBN screen on HCT116 cells. (B) Effect of compounds from CCBN screen having EC50 < 15 μM on HCT116 cells, on DLD-1 cells. 50  4.3.3 NSC645851 and RF00974 recapitulate the synthetic lethal interaction between FEN1 and CDC4  A specific FEN1 inhibitor should recapitulate some of the phenotypes of FEN1 knockdown. Given the previously-demonstrated synthetic lethal interaction between FEN1 and CDC4, and the fact that CDC4 has been shown to be a CIN gene mutated in many tumor types (Rajagopalan et al., 2004; Koh et al., 2006; Miyaki et al., 2009; Milne et al., 2010), I was interested in determining whether any of the FEN1 inhibitors shown to be active in cells could recapitulate this interaction. Because of the high mutational load frequently observed in cancers, and the fact that I am screening for a genotype-specific effect resulting from FEN1 inhibition, I elected to take advantage of two matched pairs of cell lines, derived from either HCT116 and DLD-1 parental lines, in which both copies of CDC4 had been inactivated. The use of matched cell lines for specific analysis of a predicted phenotype in response to a chemical probe is favored (Rajagopalan et al., 2004; Ji et al., 2009; Solomon et al., 2011) over screening a chemical agent on panels of tumor cells and attempting to correlate compound response with mutational information post hoc. First, to ensure that the synthetic lethal interaction I observed following simultaneous siRNA-mediated depletion of FEN1 and CDC4 could be recapitulated with CDC4-knockout cells, I treated these cells with FEN1 siRNA and observed cell growth four days following transfection. I found that there was a statistically significant negative effect on the growth of CDC4-knockout HCT116 cells treated with FEN1 siRNA relative to GAPDH-treated CDC4- knockout cells, as compared to the same effect on CDC4-wild type cells (Figure 4.2).  51  Figure  4.2. siRNA-mediated knockdown of FEN1 in HCT116 CDC4-/- cells recapitulates a synthetic lethal interaction. Solid square, HCT116 CDC4+/+; open square, HCT116 CDC4-/-. Shown is mean ± SEM. Data were analyzed by student’s t test. *** p < 0.001.  Having demonstrated that FEN1 depletion leads to selective killing of CDC4- knockout HCT116 cells, I applied the four compounds that had the greatest effect on both HCT116 and DLD-1 cells, to the HCT116-derived matched cell line pairs. I assayed whether or not these compounds could yield selective killing of CDC4-knockout cells relative to CDC4-wild type cells by digital imaging microscopy, and found that, of the four compounds assayed, NSC645851, RF00974, and RDR00966 had a differential effect on HCT116 CDC4-/- cells relative to HCT116 CDC4+/+ cells (Figure 4.3, upper panels). Next, I asked wether the three compounds that could selectively kill CDC4-/- cells in the HCT116 background could recapitulate this effect in the DLD-1 background. These compounds were again applied to DLD-1 CDC4+/+ and DLD-1 CDC4-/- cells and assayed as before. I found that NSC645851 and RF00974 yielded selective killing of DLD-1 CDC4-/- cells relative to DLD-1 CDC4+/+ cells (Figure 4.3, lower panels). Finally, to add provide additional evidence that the observed selective killing of CDC4-/- cells over CDC4+/+ cells is a genotype-specific effect (i.e., due entirely to the CDC4  52 status of each cell line), I assayed CDC4+/- HCT116 and DLD-1 cells using NSC645851 and RF00974. I found that these two compounds yielded selective killing of CDC4+/- cells relative to CDC4+/+ cells, but that the differential tended not to be as strong as between CDC4+/+ and CDC4-/- cells, strongly implying that the observed differential killing is due to the dosage of CDC4 in these cell lines (Figure 4.4).  53   54 Figure  4.3. RF00974 and NSC645851 show differential killing of CDC4-deficient cells. HCT116-based cells (upper panels) or DLD-1-based cells (lower panels) were incubated in the indicated compounds at the indicated concentrations for 72 hours prior to fixation, staining and nuclei counting. Shown is the mean of at least 6 wells ± SEM. Data were analyzed by one-way ANOVA followed by a Tukey test. * p < 0.05; ** p < 0.01; *** p < 0.001   Figure  4.4. NSC645851 and RF00974 selectively kill both CDC4+/- and CDC4-/- cells. Experiment was performed as in Figure 4.3. Shown is the mean of 6 wells ± SEM. Data were analyzed by one- way ANOVA followed by a Tukey test. * p < 0.05; ** p < 0.01; *** p < 0.001  55 4.3.4 RF00974 recapitulates the interaction between FEN1 and MRE11A FEN1 and MRE11A were found to have a strong synthetic lethal interaction (Chapter 2), and MRE11A is mutated in 4% of colorectal tumors (Wang et al., 2004). Therefore, to extend the potential therapeutic range of FEN1 inhibitors, I decided to attempt to recapitulate this synthetic lethal interaction using the more potent of the two inhibitors RF00974. In combination with siRNA-mediated knockdown of MRE11A in HCT116 cells, RF00974 yielded selective killing (Figure 4.5A). As well, this result was confirmed by combining RF00974 treatment with a recently describe MRE11A inhibitor, mirin, which inhibits the activity of MRE11A via an unknown mechanism (Figure 4.5B; Dupre et al., 2008). 4.3.5 RF00974 recapitulates DNA repair phenotypes of FEN1  Given that FEN1 plays a fundamental role in DNA replication and repair (reviewed in Zheng et al., 2011b), I asked whether there is evidence of increased DNA damage as a result of siRNA-mediated FEN1 knockdown, or following treatment with RF00974 or NSC645851. I used HCT116 cells bearing 53BP1 stably tagged with mCherry to ask whether or not FEN1 knockdown leads to the formation of 53BP1 foci, indicative of DNA repair centers (reviewed in Noon and Goodarzi, 2011). Knockdown of FEN1 with siRNA led to an increase in 53BP1 foci (Figure 4.6). Treatment with RF00974 led to an appreciable increase in 53BP1 foci; however, treatment with NSC645851, even at high concentrations, did not (Figure 4.7). This suggests that the mechanism of NSC645851-mediated cell death may not be specific to FEN1 inhibition (see Discussion).  56  Figure  4.5. RF00974 recapitulates the synthetic lethal interaction between FEN1 and MRE11A. (A) Cells were treated with siRNA as indicated. 24 hours after transfection, cells were plated in 96-well plates. Upon adhering to the plate, RF00974 was added to the indicated concentration, and cells were incubated for a further five days. Cells were fixed, stained with Hoechst dye, and imaged as in Chapter 2. (B) Cells were plated in 96-well plates. Upon adhering to the plate, the indicated compounds were added to the indicated concentrations. Cells were incubated in compound for 72 hours prior to fixation an staining. In both panels, data shown are mean of 6 wells ± SEM. Data were analyzed by one-way ANOVA followed by a Tukey test. ***, p < 0.001. 57   Figure  4.6. siRNA-mediated knockdown increases 53BP1 focus formation in HCT116 cells. (A) The average number of 53BP1 foci per cell increases following FEN1 knockdown. (B) The fraction of cells with >15 foci/cell increases following FEN1 knockdown. Data were analyzed by student’s t test. *** p < 0.001. 58   Figure  4.7. Treatment with RF00974 leads to 53BP1 focus formation, whereas treatment with NSC645851 does not. Cells were treated with the indicated compound at the indicated concentration for 24 hours (except bleomycin; 2 hours) prior to fixation an immunofluorescent labeling. Representative images are shown. 59  4.4 Discussion  In this chapter, I have demonstrated that in vitro screens for FEN1 inhibitors can uncover compounds that are active in cells. Further, I have shown that some of these compounds elicit phenotypes that are consistent with the inhibition of FEN1 and can selectively kill CDC4-deficient cells in two independent cell line backgrounds, and cells in which MRE11A has been depleted or chemically inhibited. These data suggest that FEN1 inhibitors can recapitulate previously observed synthetic lethal interactions, and thus are viable leads as potential anticancer therapeutics.  The validity of using FEN1 as a target for anticancer therapeutics may be called into question by a number of recent observations: FEN1 mutations are associated with human lung cancer (Zheng et al., 2007; Yang et al., 2009); mutation of FEN1 to disrupt its association with PCNA in primary MEF cells leads to aneuploidy and cancer (Zheng et al., 2011a); and mice bearing mutation of FEN1 similar to those seen in cancer are highly sensitive to chemical mutagenesis, and develop autoimmune disorders, chronic inflammation, and cancers (Zheng et al., 2007; Wu et al., 2012). However, it typically takes the lifetime of an individual for somatic mutations to accumulate and lead to cancer, whereas inhibition of FEN1 in a clinical setting would last only on the order of weeks. Furthermore, as discussed previously, current chemotherapeutic agents are often highly mutagenic as well. The possibility of FEN1 inhibition offering a larger therapeutic window suggests that, even with the observations noted above, specific targeting of FEN1 may be preferable to current techniques. It is essential to determine whether compounds uncovered in a biochemical screen are active in cells. In this study, we took advantage of the fact that FEN1 depletion confers a  60 growth defect to quickly screen putative FEN1 inhibitors for activity in cells, thereby limiting the number of compounds selected for further, more detailed study. The overall goal of this thesis is to demonstrate that synthetic lethal interactions, first observed in yeast, can be recapitulated in human cells, and then recapitulated again by using inhibitors discovered in biochemical screening. CDC4 had a strong synthetic lethal interaction with FEN1, so it was selected for further study. As discussed previously, tumors are genetically complex and heterogeneous; thus, attempting to recapitulate the FEN1-CDC4 interaction using FEN1 inhibitors in different tumors with varying levels of CDC4 could be problematic. Instead, I used matched pairs of cell lines in which CDC4 had been inactivated. I showed that siRNA-mediated depletion of FEN1 did lead to selective killing of CDC4-deficient cells (Figure 4.2), thereby recapitulating the dual siRNA-mediated knockdown result observed in Chapter 2. CDC4 is frequently mutated in a variety of tumor types (Rajagopalan et al., 2004; Koh et al., 2006; Akhoondi et al., 2007; Miyaki et al., 2009; Milne et al., 2010); thus, FEN1 inhibitors could potentially be used to treat CDC4-mutant tumors. Both NSC645851 and RF00974 were able to yield selective killing of CDC4-/- cells, in two different colon cancer cell lines (Figure 4.3). Crucially, both of these compounds were also able to selectively kill heterozygous CDC4+/- cells, suggesting that FEN1 inhibition would also be a successful intervention in tumors wherein CDC4 has undergone some loss of activity, but has not been completely inactivated. Interestingly, >40% of CDC4 mutations have been reported at two arginine residues in one of CDC4’s beta-propellor domains (Akhoondi et al., 2007). Beta-propellors are important structures for mediating protein- protein interactions (reviewed in Jawad and Paoli, 2002), suggesting that these mutations are  61 critical for disrupting a particular interaction. Targeted inactivation of CDC4 has been shown to lead to an increase in CIN (Rajagopalan et al., 2004). However, CDC4 does not often undergo complete inactivation in tumors (reviewed in Davis and Tomlinson, 2012). Furthermore, a proportion of the CDC4 mutations reported in cancer do not appear to cause CIN (Kemp et al., 2005). It has been suggested recently that less than 100% CDC4 activity, but greater than 0%, is optimal for tumorigenesis (reviewed in Davis and Tomlinson, 2012). However, given that approximately one-quarter of CDC4 mutations are truncating and, likely, inactivating (Akhoondi et al., 2007), it is likely that FEN1 inhibitors would yield selective killing of these tumors. However, future studies should address, by targeted knock- in, whether FEN1 inhibition is effective at selectively killing tumors with CDC4 bearing these common arginine mutations. In addition to recapitulating the interaction between FEN1 and CDC4, RF00974 was found to recapitulate the interaction between FEN1 and MRE11A. RAD27 and MRE11 have a strong synthetic lethal phenotype in yeast (Tong et al., 2001), and this interaction was found to be conserved in Chapter 2. Given that MRE11A is mutated in 4% of colorectal tumors, recapitulating this interaction further expands the potential genotypes that could be targeted with FEN1 inhibition.  In the current work, specificity of NSC645851 and RF00974 was examined by asking whether or not they phenocopy FEN1 depletion. RF00974 phenocopied FEN1 depletion in the formation of 53BP1 foci; however, NSC645851 did not cause 53BP1 focus formation. An independent screen by the Centre for Drug Research and Development revealed that NSC645851 also strongly inhibited TDP1 (Y. Shimizu, personal communication), a phosphodiesterase responsible for removing DNA adducts (Vance and Wilson, 2002). Given  62 that both HCT116 and DLD-1 cells are deficient in mismatch repair (Bhattacharyya et al., 1995) and both yeast TDP1 and RAD27 have been shown to interact genetically with components of the mismatch repair pathway (Vance and Wilson, 2002; Costanzo et al., 2010), it is possible that inhibition of TDP1 is also partially responsible for the observed differential killing. Though this would seem to make NSC645851 a less desirable inhibitor of FEN1, it has been shown recently that even promiscuous inhibitors can be highly effective (Wahlberg et al., 2012).  Although phenocopy experiments support the suggestion that a compound targets a protein of interest, a more direct assay for specificity would be to determine whether the compound inhibits related proteins. FEN1 is closely related to a number of endonucleases, such as GEN1, XPG, and EXO1, and radiolabel-based assays for these activities exist (Liu and Bambara, 2003; Singh et al., 2007; Gloor et al., 2012). The use of cell-free assays (similar to Takawa et al., 2012) would also avoid the necessity of purifying these proteins in large quantities. Screening for FEN1 inhibitors would benefit greatly from orthologous assays for the activity of these proteins, as it would obviate the need for more time- consuming cell-based assays. Given our success in developing a high-throughput-amenable fluorescence/quenching-based FEN1 assay, deploying counterscreens for the activities of FEN1-related proteins should not prove difficult.  Finally, mouse xenograft experiments, using CDC4-mutant paired cell lines, can be conducted to determine whether or not FEN1 inhibitors are able to reach their target in an organismal context. Supplementing these assays with cell lines bearing knock-in of common CDC4 mutations would also greatly increase the therapeutic relevance of these experiments.  63  From a screen of 30 000 compounds, one potent inhibitor of FEN1 that is active in cells and recapitulates FEN1-depletion phenotypes and synthetic lethal interactions was found. Given that large pharmaceutical companies possess libraries with compounds numbering in the millions (Macarron et al., 2011), a screen for FEN1 inhibitors orders of magnitude larger would be expected to yield dozens, or even hundreds, of unique, biologically active FEN1 inhibitors. The pipeline for compound discovery presented up to this point, combined with the improvements I have suggested, should yield numerous potential anticancer FEN1 inhibitors. 64  A version of chapter 5 is in preparation for publication. Derek M. van Pel, Peter C. Stirling, Sean W. Minaker, Payal Sipahimalani, and Philip Hieter. (2012) Model organism genetics predict candidate therapeutic genetic interactions at the mammalian replication fork. Chapter  5: Expanding the therapeutic potential of CTF4 and RAD27 5.1 Introduction  The yeast gene CTF4 was originally identified in a screen for mutants that cause an increased frequency of artificial chromosome loss (Spencer et al., 1990) and was subsequently shown to be involved in DNA metabolism and progression from S phase to mitosis (Kouprina et al., 1992) and to bind to DNA polymerase α (Miles and Formosa, 1992). Orthologs of Ctf4 have been described in Apergillus nidulans (Harris and Hamer, 1995), Schizosaccharomyces pombe (Williams and McIntosh, 2002), and vertebrates (Kohler et al., 1997), and CTF4 has been shown to be a strong CIN gene in various other assays of genome integrity (Yuen et al., 2007). Ctf4 has been shown to be required for the establishment of sister chromatid cohesion (Hanna et al., 2001; Petronczki et al., 2004; Lengronne et al., 2006), and appears to mediate cohesion establishment while travelling with the replication fork (Lengronne et al., 2006).  As has been discussed extensively elsewhere in this thesis, Rad27 is a flap endonuclease responsible for cleaving 5’ DNA flap structures that occur during DNA replication and repair (reviewed in Bambara et al., 1997). Its human ortholog, FEN1, is highly conserved (Greene et al., 1999). Mutation of RAD27 causes a strong CIN phenotype (Yuen et al., 2007).  As well as playing fundamental roles in cellular DNA metabolism, CTF4 and RAD27 have many synthetic lethal interacting partners among non-essential genes in the as-yet- incomplete yeast genetic interaction map (Tong et al., 2001; Pan et al., 2004; Tong et al.,  65 2004; Collins et al., 2007; Yuen et al., 2007; Costanzo et al., 2010). As has been discussed previously, many such genetic interactions are with the yeast orthologs of genes mutated in cancers, thus recommending WDHD1, the human ortholog of Ctf4, and FEN1 as targets for the development of anticancer therapeutics based on the concept of synthetic lethality (Yuen et al., 2007). Given its potential therapeutic relevance, and the fact that little was known about the precise function of Ctf4, I undertook a series of studies to characterize Ctf4’s molecular function. Additionally, given the fact that synthetic lethal interaction data was available for CTF4 and RAD27 with only the non-essential S. cerevisiae genes, I carried out screens for genetic interactions with essential genes using temperature-sensitive and DAmP (Decreased Abundance by mRNA Perturbation) alleles (Breslow et al., 2008; Li et al., 2011). 5.2 Methods 5.2.1 Strains  For a list of strains used in this chapter, please refer to Appendix D. 5.2.2 Two-hybrid assay  Yeast two-hybrid assay was performed by mating strains carrying plasmids expressing proteins of interest fused to the GAL4 activating domain or DNA binding domain. Cells were grown to log phase in SC –Leu –Trp, then cultures were subjected to serial ten- fold dilutions and plated on SC –Leu –Trp and SC –Leu –Trp –His. Growth on medium lacking histidine was indicative of a physical interaction (Chien et al., 1991). pOAD encodes the GAL4 activating domain; pOBD2 encodes the GAL4 DNA-binding domain; pBDC/N.DEST is a Gateway-compatible vector I designed that encodes the GAL4 DNA- binding domain, and is a derivative of pBDC, itself a derivative of pOBD2. The interaction between Bir1 and Sli15 was used as a positive control.  66 5.2.3 CTF assay  Strains bearing an artificial chromosome fragment were streaked for isolation on SC – Trp to select for the chromosome fragment. This fragment encodes the SUP11 gene, which suppresses the premature STOP codon in the otherwise red pigment-producing ade2-101 strain. Colonies were picked and plated onto medium containing limiting adenine such that 200-300 colonies appeared. The number of first-division missegregation events (half- sectored colonies) leading to one white and one red sector was scored (Hieter et al., 1985). 5.2.4 Sister chromatid cohesion assay  Strains bearing genomically integrated tandem repeats of the E. coli lac operon and expressing GFP-LacI, which strongly binds the lac operon (Michaelis et al., 1997), were grown to log phase in liquid yeast peptone dextrose medium, then α-factor was added to a concentration of 0.4 μg/mL to arrest cells in G1 phase. After 3-4 hours, cells were washed four times in water, then released into YPD containing 5 μg/mL nocodazole to arrest cells at G2/M. After 3-4 hours, cells were washed in water and mounted onto slides for fluorescent imaging using a Zeiss Axioplan at 63x magnification using the GFP filter. All strains were bar1Δ to facilitate α-factor arrest. 5.2.5 Random spore analysis  Sporulated cultures of the indicated double heterozygous mutant diploids were collected in water and plated at a density such that 200-300 colonies appeared on a haploid- selective plate (SC –Leu –His –Arg + 50 mg/mL). The same number of cells was plated on single- and double-selection plates (the same as above, but lacking uracil, and/or containing 200 mg/mL G418, and/or containing 200 mg/mL cloNAT, depending on the strain). The  67 number and size of colonies on double mutant selection was compared to the other plates to assess the fitness of the double mutants (Warren et al., 2004). 5.2.6 SGA and SDL screening  Query strains were mated to the appropriate array (FLEx overexpression array; Hu et al., 2007; DAmP array; Breslow et al., 2008; essential gene temperature-sensitive allele array; Li et al., 2011), and both SGA and SDL screens were carried out and analyzed as described previously (Measday et al., 2005; Stirling et al., 2011; McLellan et al., 2012) 5.3 Results 5.3.1 Characterizing Ctf4 function using a series of point mutations  Mutants in CTF4 have been reported to be sensitive to DNA damaging drugs (Ogiwara et al., 2007), have increased chromosome missegregation (Kouprina et al., 1992), and are deficient in the establishment of sister chromatid cohesion (Hanna et al., 2001). As well, physical interactions have been reported between Ctf4 and Sld5, an essential DNA replication accessory protein (Gambus et al., 2006) and Mms22, a protein of unknown function required for double strand break repair (Ben-Aroya et al., 2010). In an attempt to separate these seemingly pleiotropic phenotypes into different regions of the Ctf4 protein, I used a series of point mutations in CTF4 that were isolated in the original CTF screen in which CTF4 was discovered (Spencer et al., 1990; Kouprina et al., 1992). I sequenced these 9 mutant alleles and found that they consist of four missense mutations and five truncating nonsense mutations, spread along the length of the protein (Figure 5.1). 68   Figure  5.1. Schematic of mutations in CTF4 alleles. WD repeats were determined from sequence analysis by Swiss-Prot, and phosphoserines were identified in (Smolka et al., 2007)  I performed a yeast two-hybrid assay with these mutants, assaying the effect of mutation on the interaction with Sld5 and Mms22. After ensuring that Ctf4 did not self-activate in the yeast two-hybrid system (Figure 5.2A), I determined that the interaction between Ctf4 and Sld5 is highly sensitive to perturbation by mutation at all points tested except the extreme N- terminus (Figure 5.2B). The interaction between Ctf4 and Mms22 appears to be most sensitive to mutation or deletion in the middle third of the primary sequence of Ctf4 (Figure 5.2B).  I next examined the sensitivity of CTF4 mutants to DNA damaging drugs using the radiomimetic compound bleomycin and the ribonucleotide reductase inhibitor hydroxyurea (Figure 5.3). The alleles fell into three groups: those with sensitivity roughly equal to the knockout; those with sensitivity roughly equal to the wild type; and those with intermediate sensitivity. Three of the five truncation mutants appeared as sensitive as ctf4Δ, and three of  69 the four missense mutants had sensitivities roughly equal to wild type. ctf4-66, the most severe truncation mutant, ctf4-43, the least severe truncation mutant, and ctf4-46, one of the two point mutants (along with ctf4-154) that could bind neither Sld5 nor Mms22, all displayed intermediate sensitivity.  70   71 Figure  5.2. Yeast two-hybrid interactions between CTF4 alleles and Sld5 and Mms22. All two-hybrid assays were carried out as in Materials and Methods. (A) CTF4 alleles in fused to the GAL4 DNA-binding domain do not self-activate. (B) Mutations in Ctf4, except those in the extreme N-terminus, severely reduce the interaction with Sld5. (C) Certain mutations of Ctf4 abrogate binding to Mms22.   Figure  5.3. The response of CTF4 alleles to DNA-damaging drugs. Cells were grown to log phase, then cultures were subjected to serial ten-fold dilution on medium containing the indicated drug at the indicated concentration.  Finally, I performed two quantitative assays to measure genome stability in these strains. I performed a quantitative chromosome loss assay, wherein colonies carrying a nonessential chromosome fragment that suppresses a red colony color are scored for first- division nondisjunction events (Hieter et al., 1985). All alleles tested had a greater frequency of chromosome missegregation than wild type (~20-60-fold greater). However, all mutants had a similar magnitude of effect compared to ctf4Δ (Table 5.1) These small differences both  72 between and within the allele set make these data uninformative for determining the function of specific domains of Ctf4. The second assay I performed was a sister chromatid cohesion assay. In this assay, M phase-arrested cells are scored for premature loss of sister chromatid cohesion, indicated by separation of chromatids marked by a GFP marker (Michaelis et al., 1997). After arresting cells in alpha factor to demonstrate that the strains I used were not aneuploid (Table 5.1), I determined that all of the alleles had an elevated frequency of cells with separated chromatids; however, whereas most mutants were approximately equal to or greater than ctf4Δ, one allele – ctf4-41 – was noticeably lower than all of the other mutants (though not statistically significantly so; Table 5.1). Interestingly, ctf4-41 was the only mutant that was able to physically interact with both Sld5 and Mms22 (Figure 5.2). Table 5.1. Quantitaion of colony sectoring and chromatid separation assay for CTF4 mutants. Allele Half-sectored colonies (%, n) Fold increase in half-sectored colonies over wild type α factor- arrested cells with separated chromatids (%) Nocodazole- arrested cells with separated chromatids (% ± SEM) WT 0.031, 3185 N/A 2.0 4.1 ± 1.2 ctf4Δ 1.72, 1981 55 2.5 21.9 ± 3.5 ctf4-66 1.40, 1002 45 2.0 35.8 ± 0.8 ctf4-65 1.94, 2111 23 8.0 24.4 ± 3.4 ctf4-50 0.78, 385 25 7.0 33.2 ± 0.4 ctf4-25 1.94, 2111 62 3.2 27.3 ± 4.5 ctf4-43 1.79, 4964 57 3.8 26.7 ± 2.5 ctf4-41 ND N/A 1.0 13.4 ± 1.6 ctf4-154 0.92, 1518 29 2.0 27.0 ± 4.0 ctf4-46 1.26, 1661 40 3.0 21.6 ± 0.3 ctf4-107 1.23, 1873 39 0.0 29.2 ± 3.6   73  Taken together, these data suggest that the ability to bind efficiently to Sld5 is critical to the cohesion establishment function of Ctf4. Furthermore, the ability of a ctf4 mutant protein to interact with Mms22 is predictive of neither its cohesion establishing ability, nor of its ability to function in the response to DNA damage. 5.3.2 Putative Ctf4 phosphorylation does not contribute to DNA damage sensitivity or cohesion defects  It has been reported that Ctf4 is phosphorylated at serines 377 and 379 constitutively, and at serine 463 in a Mec1/Tel1/Rad53-dependent manner in response to DNA damage (Smolka et al., 2007). Given that substrate phosphorylation is a major step in the DNA damage response (reviewed in Lazzaro et al., 2009), and data showing that ctf4 mutants are sensitive to DNA damage (Figure 5.3), I decided to investigate whether or not mutation of these residues would lead to an increase in CIN phenotypes. I mutated these residues in various combinations, reasoning that since serines 377 and 379 are constitutively phosphorylated, whereas serine 463 is phosphorylated in response to DNA damage, their functions may be separable. Thus, either serine 463, or serines 377 and 379 together, or all three were mutated to alanines, to make the residue non-phosphorylatable, or to aspartate, to mimic constitutive phosphorylation. The mutants were then subjected to a sister chromatid cohesion assay, or to a DNA damage sensitivity assay employing bleomycin, hydroxyurea, and the DNA alkylator MMS (Figure 5.4; Table 5.2). None of these mutations resulted in an increase in DNA damage sensitivity or a noticeable cohesion defect, suggesting that phosphorylation, if present, does not play a role in these functions of Ctf4. 74   Figure  5.4. CTF4 phosphoserine mutants are not sensitive to DNA-damaging agents. Cells were grown to log phase, then cultures were subjected to serial ten-fold dilutions before being plated on the indicated medium.  Table 5.2. Quantitation of chromatid separation assay for Ctf4 phosphorylation site mutants. Allele α factor-arrested cells with two dots (%) Nocodazole-arrested cells with two dots (%) CTF4 4.0 3.5 ctf4-S463A 3.0 0.0 ctf4-S377A/S379A 2.5 4.3 ctf4-S377A/S379A/S463A 2.0 2.5 ctf4-S463D 0.0 7.5 ctf4-S377D/S379D 0.5 1.5 ctf4-S377D/S379D/S463D 0.0 4.3 75  5.3.3 Exploiting protein interaction networks to yield alternative targets to Ctf4/WDHD1 As stated previously, CTF4/WDHD1 is a highly connected genetic hub in a CIN synthetic lethal interaction network (Yuen et al., 2007). Thus, chemical inhibitors of the activity of WDHD1 would be potential anticancer therapeutics with a broad spectrum of target genotypes. However, because the function and activity of Ctf4/WDHD1 is poorly characterized, and because it is not clear how one would screen for loss of its activity biochemically, an alternative assay for Ctf4/WDHD1 function is needed. Heterologous expression of human proteins in yeast can be toxic, and this can form the basis of high- throughput, cell-based screens for protein inhibitors (reviewed in Balgi and Roberge, 2009). The principle of the screen is that, if overexpression of a target is toxic to yeast, then an inhibitor of the activity of that protein should restore yeast growth to (ideally) wild type levels. Thus, I overexpressed WDHD1 in yeast and asked if a growth defect occurred. I found that overexpression of WDHD1 did not inhibit the growth of yeast, suggesting that still another alternative assay would be needed to inhibit this genetic hub (Figure 5.5). Interestingly, overexpression of FEN1 was toxic to yeast, suggesting this could be an alternative to the biochemical screening for FEN1 inhibitors already described.  76  Figure  5.5. Overexpression of FEN1 but not WDHD1 is toxic to yeast. (A) Strains carrying the indicated, Leu-selected plasmid were grown to log phase, then cultures were subjected to serial ten-fold dilutions and plated on the indicated medium. (B) Western blot showing expression of WDHD1 and FEN1 in strains. Strains were grown overnight in YEP + 2% raffinose, then grown for four hours in YEP + 2% indicated carbon source.  High-throughput genetic screens by other groups have shown that proteins that are part of the same complex often share genetic interactions (Krogan et al., 2006; Collins et al., 2007). Given the difficulty presented thus far in developing an assay for chemical inhibitors of Ctf4/WDHD1, I reasoned that physical interactors of Ctf4 would share its genetic interactions. Ctf4/WDHD1 is part of the replisome, and has been linked physically to the replication accessory factor Mcm10/MCM10, the catalytic subunit of DNA polymerase α Pol1/POLA1, and the GINS complex (Psf1-3 and Sld5 in yeast, GINS1-4 in humans). Also  77 linked to these proteins are the MCM helicase complex (Mcm2-7/MCM2-7) and the replication elongation factor Cdc45/CDC45L. In yeast, all of these genes are essential; thus, no high-throughput genetic interaction data was available for them at the outset of this study. I decided to perform a pilot experiment to determine whether or not these various complexes could recapitulate the reported synthetic lethal interactions of CTF4, using temperature- sensitive alleles of these essential genes from recently-described collections (Ben-Aroya et al., 2008; Li et al., 2011). I performed random spore analysis on the meiotic products of crosses of all available temperature-sensitive alleles of the above genes with three genes previously shown to exhibit synthetic sickness/lethality with ctf4Δ, and whose human orthologs have been reported mutated in cancer: the RecQ helicase SGS1, the DNA double strand break repair gene MRE11, and the spindle assembly checkpoint gene BUB1. I found that, in nearly every case, each three of the CTF4-interacting genes (SGS1, MRE11, or BUB1) was either synthetic lethal or displayed synthetic slow growth in combination with at least one member of each complex tested (Figure 5.6; Appendix E). As Pol1 and the MCM complex have a described enzymatic function, it may be possible to screen biochemically for inhibitors of the activities of these proteins. Furthermore, given the volume of genetic interactions between CTF4 and the rest of the yeast genome, these data suggest the possibility that the replication fork is generally a hub of genetic interactions. 78   Figure  5.6. Results of random spore analysis crosses between CTF4 genetic and physical interactors. These data suggest that CTF4 physical interacting proteins are also viable targets for small-molecule inhibitor development. Genes are grouped by subcomplex, with the exception of Mcm10 and Cdc45, which have not been assigned to a replication subcomplex. Note that a ts allele of PSF2 was not present in our collection, so it was not tested. 79  5.3.4 Expanding the range of mutations targeted inhibition of Ctf4/WDHD1 or Rad27/FEN1  As stated previously, CTF4 and RAD27 are among the most highly genetically connected genes in the yeast genome; however, publicly available high-throughput data is available only for the nonessential genes (Tong et al., 2001; Tong et al., 2004; Collins et al., 2007; Costanzo et al., 2010). Thus, I sought to expand the therapeutic relevance of Ctf4/WDHD1 and Rad27/FEN1 by screening for genetic interactions in collections of essential genes.  I screened ctf4Δ and rad27Δ mutants in biological and technical triplicate against a collection of temperature-sensitive alleles of essential genes (Li et al., 2011), as well as a collection of hypomorphic alleles of essential genes (DAmP alleles; Breslow et al., 2008), as essential genes have been shown previously to have an interaction density five times greater than nonessential genes (Davierwala et al., 2005). Data were analyzed as previously described (Stirling et al., 2011; McLellan et al., 2012). Negative genetic interactions among the essential genes having a p-value of <0.05 and an E – C (experimental growth minus control growth) value of < -0.3 for CTF4 are presented in Appendix F; negative genetic interactions for RAD27 are shown in Appendix G. I observed 93 genetic interactions with ctf4Δ, and 29 with rad27Δ. Both data sets were analyzed for gene ontology (GO) term enrichment (Figure 5.7A- D). Using a cutoff of 2 for fold enrichment of hit genes in the query set (i.e., the essential genes represented in the temperature-sensitive and DAmP collections; Stirling et al., 2011), hits for CTF4 are enriched for genes involved in DNA replication, response to DNA damage, cell cycle progression, and mitotic spindle dynamics, all of which are consistent with  80 previously described roles for Ctf4 (Hanna et al., 2001; Petronczki et al., 2004; Lengronne et al., 2006; Gambus et al., 2006). Likewise, RAD27 hits are mostly confined to previously- described cellular roles of Rad27, such as the DNA damage response, DNA replication, and sister chromatid cohesion (Bambara et al., 1997; Yuen et al., 2007). Analysis suggested an enrichment in chromatin silencing at the mating type cassette; however, of the five genes in the query set assigned to this GO term, four were associated directly with replication origin firing or replisome progression (Orc2, Orc3, Mcm7, Pol30), and only one, Rap1, was annotated as playing a role in silencing (specifically, at the telomere). Rap1’s role in telomere silencing, mediated by telomere binding, combined with Rad27/FEN1’s previously-described role in telomere length maintenance, may explain the observed synthetic lethality. 81   Figure  5.7. Gene ontology (GO) term enrichment among high-throughput genetic interaction studies. Shown are terms having a fold enrichment of >2. GO term cellular component (A), and biological process (B) for negative genetic interactions with ctf4Δ, and cellular component (C) and biological process (D) with rad27Δ. GO term function (E) and process (F) enrichment for SDL with rad27Δ.  Of the essential genes found to interact with CTF4, 62 genes out of 93 (67%) have human orthologs that have documented mutations in cancer. Furthermore, three of the genes in the list (CDC28/CDK4, CRM1/XPO1, and VTI1/VTI1A) appear in the cancer gene census, suggesting a causative role in cancer progression. For RAD27, 18 genes out of 29 (62%) have orthologs mutated in cancer, and one (SFH1/SMARCB1) is a cancer census gene (Futreal et al., 2004; Bamford et al., 2004).  82 A subset of these genes was selected for validation, by mating the SGA query strain to the array hit strain and, when possible, obtaining four isolates of a haploid double mutant following sporulation. Synthetic lethality was validated with ctf4Δ psf1-1, ctf4Δ rpn11-14, ctf4Δ RPL32-DAmP, rad27Δ nse4-ts2, and rad27Δ RPL32-DAmP. In total, at least one allele of 17 genes was selected for validation of interactions with CTF4, and at least one allele of 15 genes’ interactions recapitulated, suggesting a high-quality data set (Table 5.3 and Figure 5.8). For RAD27, at least one allele of 10 genes was selected for validation, and at least one allele of 4 genes’ interactions recapitulated, suggesting a higher false positive rate for the RAD27 screen (Table 5.4 and Figure 5.9). The tetrad confirmation rate was lower because tetrads were dissected only at 25°C. (Dissecting tetrads at all possible temperatures would have proven prohibitive, but would be the ideal assay to confirm a small number of interactions.) Importantly, interactions between CTF4 and CDC28, and between CTF4 and CRM1, both orthologs of cancer census genes, recapitulated; however, the interaction between RAD27 and SFH1, also a cancer census gene ortholog, did not. Taken together, these data add weight to the idea that Ctf4/WDHD1 is a strong candidate for anticancer therapeutic development, owing to its previously reported place as a hub among nonessential yeast cancer orthologs, and now its place as a hub among essential yeast cancer orthologs. The data suggest that Rad27/FEN1 also remains a candidate for anticancer therapeutic development; however, the interaction list yielded by the SGA in this study requires further validation of its quality.  83 Table 5.3. Results of validation of SGA hits with ctf4Δ. SS, synthetic sick. SS/L, synthetic sick with some synthetic lethals. SL, synthetic lethal. Allele Tetrad result Spots four isolates result, temperature (°C) cdc28-1 No interaction Weak interaction 30 cdc28-13 No interaction Strong interaction 30 cdc45-27 SS Strong interaction 30 crm1-1 No interaction Weak interaction 30, 34 dam1-1 SS/L 3/4 weak interaction 30, 1/4 no worse than dam1-1 dam1-5 No interaction Weak interaction 30 dbf4-1 SS Strong interaction 30 dbf4-3 No interaction Weak interaction 30 mob2-26 No interaction Strong interaction 34, 37 psf1-1 SL rpn11-14 SL rpn11-8 No interaction No interaction sen1-1 SS/L Weak interaction 30 spt15- I143N SS Weak interaction 25 spt15-P65S No interaction Moderate interaction 30 stu2-10 Inconclusive Moderate interaction 30 stu2-11 No interaction Strong interaction 34 taf10-ts34 SS/L Weak interaction 25 tub4-ΔDSY SS? Possibly aneuploid No interaction tub4- Y445D SS No interaction POL30- DAmP SS/L Strong interaction SMT3- DAmP SS/L Moderate interaction RPL32- DAmP SL RPS20- DAmP Inconclusive No interaction   84 Table 5.4. Results of validation of SGA hits with rad27Δ. Allele Tetrad result Spots four isolates result, temperature (°C) nse4-ts3 SS Strong interaction 30, isolate 3 throws off suppressors nse4-ts1 SS/L Moderate interaction nse4-ts2 SL nse4-ts4 No interaction 3/4 moderate interaction, 1 no worse than nse4-ts4 spt15-P65S No interaction taf1-21 No interaction No interaction POL30- DAmP SS/L Moderate interaction 25 SMT3- DAmP SS/L Moderate interaction 25 NOC4- DAmP SS (weak) RPL32- DAmP SL SFH1 1- DAmP SS No interaction SFH1 2- DAmP SS No interaction SPC34- DAmP SS (weak) No interaction 85   Figure  5.8. Independent validation of SGA synthetic lethal interactions. The key for the strain layout on the plates is shown at top-left. Cells were grown to log phase, their OD600 was normalized, and cultures were subjected to serial ten-fold dilution and plated on YPD. Strains were grown for two days at the indicated temperature. 86  5.3.5 A genome-wide synthetic dosage lethality screen suggests Rad27/FEN1 as a candidate target in novel cancer genotypes Given the etiological consequences of gene amplification and overexpression in cancers (for reviews, see Albertson, 2006; Santarius et al., 2010), the high number of genetic interacting partners of RAD27 (Tong et al., 2001; Pan et al., 2004; Tong et al., 2004; Costanzo et al., 2010; this study), the critical nature of Rad27/FEN1 in all DNA transactions (reviewed in Zheng et al., 2011b), and my development of inhibitors of FEN1 activity (Chapter 3), I was interested in expanding the therapeutic significance of FEN1 as a target for inhibition in cancer by performing a whole-genome synthetic dosage-lethal (SDL) screen with rad27Δ and a recently-described, arrayed set of yeast strains systematically overexpressing all ORFs (Hu et al., 2007 and Andrews, B. et al., unpublished). Data were analyzed as in conventional SGA, and I observed that very few genes met an E – C cutoff of -0.3. Thus, I increased the cutoff to -0.2, ending up with the list presented in Appendix H. Of the 24 genes observed to be SDL with rad27Δ, 11 (46%) have been shown to be mutated in cancer, and 8 (33%) have had amplifications reported in cancer according to the CONAN database (Forer et al., 2010). GO term analysis (Figure 5.7 E, F) was consistent with previously reported Rad27 functions, including DNA repair, DNA replication, and DNA damage response. The list was most strongly enriched for genes having DNA topoisomerase activity, picking up two such genes in a list of 24 (TOP1 and RDH54), out of 4 such genes in the entire yeast genome. As well, I detected the previously-reported SDL interaction between rad27Δ and overexpression of Pif1 (Chang et al., 2009). When these interactions were validated by independent spot assay, 21/24 negative interactions (88%) recapitulated (Figure 5.9). The one positive interaction, between rad27Δ and  87 overexpression of TPK2 (having an E – C of >0.2) was found to be spurious, and is likely due to the fact that overexpression of TPK2 alone prevents growth on galactose medium.  Figure  5.9. Independent reconfirmation of synthetic dosage lethal interactions with rad27Δ. Wild type or rad27Δ strains carrying plasmids expressing the indicated gene (or empty vector, “vector”) driven from a galactose-inducible promoter were grown to log phase in glucose medium, normalized, subjected to serial five-fold dilution, and plated on the indicated medium. Cells were grown for two days prior to imaging. 88  5.4 Discussion  In this chapter, I have used a series of point mutations of Ctf4 to further characterize its function, and determined that the physical interaction between Ctf4 and Sld5 is critical for Ctf4 to function in the establishment of sister chromatid cohesion. I showed that Ctf4 phosphorylation, if present, does not contribute to its function in cohesion establishment or the response to DNA damage, and that physical interactors of Ctf4 share at least some synthetic lethal interactions of Ctf4 itself. Finally, by performing high-throughput genetic screens, I expanded the range of genetic lesions that could be targeted by small-molecule inhibition of Ctf4/WDHD1 or Rad27/FEN1. 5.4.1 Associations between Ctf4 and Sld5 or Mms22 The specific mechanism by which Ctf4 contributes to the establishment of sister chromatid cohesion is unknown. Ctf4 appears to mediate the interaction between the MCM helicase complex and the Pol α primase complex (Zhu et al., 2007; Gambus et al., 2009; Tanaka et al., 2009a), and also between the GINS complex and Pol α (Tanaka et al., 2009a). One model suggests that the cohesin complex forms a ring whose geometry is altered by a replisome lacking or mutated in Ctf4, such that cohesion establishment is defective (Lengronne et al., 2006). Many members of the MCM, GINS, and Pol α complex have been implicated in chromosome segregation (Maine et al., 1984; Stirling et al., 2011); presumably, this is due to their role, with Ctf4, in mediating cohesion establishment. The observation that a physical association between Ctf4 and Sld5 is associated with efficient establishment of sister chromatid cohesion suggests that these two proteins coordinate in this function. Given the fact that many replisome components are destabilized in the absence of Ctf4 (Tanaka et al., 2009a; Wang et al., 2010), Ctf4 likely acts as a scaffold to hold cohesion-mediating  89 proteins together with the replisome (Tanaka et al., 2009b), a model the data presented here support.  The interaction profile of the Ctf4 alleles with Mms22 is not predictive of each mutant’s response to DNA damage, nor of its cohesion-establishing ability. This is surprising, given that Mms22, and the complex of which it is a part, plays an important role in the DNA damage response and homologous recombination. Mms22 is recruited to double strand breaks, where it must be ubiquitinated and destroyed for DNA repair to proceed. It is targeted for ubiquitination by its physical interactor Rtt101 (Ben-Aroya et al., 2010), which itself binds Orc5, a member of the origin recognition complex (Mimura et al., 2010). Rtt101 also ubiquitinates Spt16 (Han et al., 2010), a component of the FACT complex that removes histones ahead of, and replaces histones behind, elongating transcription and replication machinery. This ubiquitination is required for the association of FACT with replication origins via binding to the Mcm2-7 helicase complex. Ctf4 has also been shown to bind Spt16 competetively with Pob3 (Wittmeyer and Formosa, 1997), though strangely, loss of Pob3 slows DNA replication (Schlesinger and Formosa, 2000), suggesting Spt16-Ctf4 binding does not facilitate replication. There is a large body of literature concerning these intriguing physical interactions that warrant further study to determine the function of the physical interaction between Ctf4 and Mms22. 5.4.2 Using Ctf4 physical interactors to identify new drug targets  Given the difficulty posed in identifying biochemical inhibitors of Ctf4/WDHD1, I attempted to use genetic methods to identify potential new targets, based on the fact that physical interactors tend to share genetic interactions (Krogan et al., 2006; Collins et al., 2007). I showed that physical interactors share synthetic lethal interactions with Ctf4 in a  90 small test set, suggesting that Ctf4 physical interactors (i.e., the replication fork) form a therapeutically relevant hub of synthetic lethal interactions that could be targeted for small- molecule inhibitor development.  Ctf4 is one of the only components of the replisome that is nonessential; thus, it is one of the only replication-associated proteins for which high-throughput genetic interaction data existed at the outset of this study. These data showed that Ctf4 has a great number of synthetic lethal interactions, including many with yeast cancer orthologs. This study suggests that, if genome-wide synthetic lethal interaction profiles were obtained with the essential members of the replisome, they would share these interactions. Given that these proteins have enzymatic functions (for example, the MCM complex has ATP-dependent helicase activity; Pol α is a DNA-dependent RNA polymerase), screening for inhibitors of these proteins biochemically would be more feasible than in the case of Ctf4 (Inglese et al., 2007; Macarron and Hertzberg, 2009), if an in vitro assay for their activity amenable to high- throughput screening could be devised. 5.4.3 Expanding the therapeutic value of Ctf4/WDHD1 and Rad27/FEN1 as targets for small-molecule inhibition  Given that ctf4Δ and rad27Δ had been screened only against the nonessential deletion collections for genetic interactions, I screened them against hypomorphic and temperature- sensitive essential gene collections, reasoning that more yeast cancer orthologs that genetically interact with these genes could be uncovered in this way. 65% of the alleles that reached the cutoff in each screen had human orthologs that were mutated in cancer. Given the fact that synthetic lethal interactions between ctf4Δ and rad27Δ and nonessential yeast cancer gene orthologs appear to be highly conserved (Chapter 2), it is reasonable to expect  91 that interactions with essential yeast cancer gene orthologs would also be conserved. Thus, inhibitors developed against FEN1 or WDHD1 (or other replisome components; see previous section) would be expected to be active on cancers bearing mutations in these genes as well, thereby expanding the therapeutic range of these compounds.  In addition, a synthetic dosage-lethality screen found numerous genes that are toxic when overexpressed in a rad27Δ background; one-third of the human orthologs of these genes have been reported to be amplified in cancer, suggesting cancers bearing these genotypes may be susceptible to biochemical FEN1 inhibition.  Taken together, the data presented in this chapter suggest that the replication fork is a highly-connected genetic hub, thus making it an ideal target for the development of small- molecule anticancer therapeutics. As well, the discovery of new synthetic lethal interactions between the highly connected genes CTF4 and RAD27, and a large number of essential or amplified yeast-cancer orthologs, emphasizes the importance of developing small-molecule inhibitors of these proteins or the complexes of which they are a part. 92  Chapter  6:  Conclusions and future directions  Cancer is a multigenic disease. Chromosome instability (CIN) mutations play a significant role in cancer progression, as they allow genetic lesions in proto-oncogenes and tumor suppressor genes to accumulate (reviewed in Vogelstein and Kinzler, 2004). Because chromosome segregation is essential to cellular viability, the genes and proteins involved in this process are conserved across evolution (Yuen et al., 2007; Barber et al., 2008; McLellan et al., 2009; McLellan et al., 2012). A cross-species approach to the study of CIN that utilizes model organisms yields significantly more insight into CIN than examining CIN in cancer cells alone, due to the power of high-throughput genetics and the wide array of assay systems they offer. Because a CIN mutation is a sublethal defect in an essential process, it can be leveraged by genetic interactions to lethality. By studying the synthetic lethal interaction spectrum of yeast CIN cancer orthologs, then determining whether these synthetic lethal interactions are recapitulated in human cells, novel targets for anticancer therapeutics can be discovered. 6.1 The conservation of genetic interactions  The degree to which genetic interactions are conserved across evolution remains controversial; however, the picture that is emerging, based on the data presented in this thesis and elsewhere, is that interactions within a core process are much more highly conserved than interactions across a whole genome (Lehner et al., 2006; Roguev et al., 2007; Tarailo et al., 2007; Dixon et al., 2008; McLellan et al., 2009; Frost et al., 2012; McLellan et al., 2012). Here, I found that 70% of interactions within a CIN gene synthetic lethal interaction network were conserved (Chapter 2). This is similar to the proportion of conserved interactions (>80%) observed in two studies examining the  93 conservation of synthetic lethal interactions in related CIN cancer networks between S. cerevisiae and C. elegans (McLellan et al., 2009; McLellan et al., 2012). Considering the high degree of connectivity of the yeast orthologs of the genes examined in this work, the essential nature of ensuring proper chromosome segregation, and the fact that all of the genes interrogated are part of the same or related pathways (i.e., DNA repair, DNA replication, and chromosome segregation), this degree of conservation of synthetic lethal interactions found here is not entirely unexpected. However, the number of gene pairs probed in this thesis, 30, was small, and did not investigate all CIN or DNA repair genes. Thus, determining the conservation of synthetic lethal interactions among the entire conserved CIN gene set warrants further study. 6.2 Optimization of FEN1 biochemical screening  I carried out a successful screen for small-molecule inhibitors of FEN1 flap endonuclease activity (Chapter 3). However, FEN1 is part of a family of related 5’ nucleases, which includes three other members: XPG is active in nucleotide excision repair, and cleaves DNA bubble substrates; EXO1 plays a role in mismatch repair and cleaves DNA ends; and GEN1 is part of the homologous recombination pathway, and acts by cleaving Holliday junctions. These proteins mediate their different effects through combinatorial substrate recognition, via the presence or absence of 5’ or 3’ flaps, or dsDNA-ssDNA junctions (Tsutakawa et al., 2011). This screen was carried out under conditions that enrich for competitive inhibitors (Macarron and Hertzberg, 2009). In competitive inhibition, the compound competes with the substrate for binding to the target. Since it is unclear where on FEN1 each hit binds, it is conceivable that a compound inhibiting FEN1-substrate binding would thus also inhibit binding of another  94 member of the FEN1 family to its cognate substrate. Previous screens for FEN1 inhibitors used XPG inhibition as a counterscreen, to increase selectivity (Tumey et al., 2004; Tumey et al., 2005). Future, larger screens for FEN1 inhibitors would benefit from the addition of such counterscreens, as they would require minimal extra effort once protein had been purified. 6.3 Additional studies with FEN1 inhibitors  Two FEN1 inhibitors identified in this study were found to recapitulate the synthetic lethal interaction between FEN1 and CDC4, by combining FEN1 inhibitor with cells in which CDC4 had been inactivated (Chapter 4). As well, RF00974 recapitulated the interaction between FEN1 and MRE11A. Similar studies could be carried out for all of the genes interrogated in Chapter 2. On the other hand, genome-wide chemical-genetic profiling using whole-genome targeting shRNAs combined with FEN1 inhibitor or FEN1 siRNA could also be used to determine whether or not the FEN1 inhibitors described here recapitulate the entire spectrum of FEN1 interactions. Such a technique was recently applied to the study of sensitization to the estrogen receptor antagonist tamoxifen (Mendes-Pereira et al., 2012).  Additionally, experiments in yeast could help determine the specificity of FEN1 inhibitors, owing to the high degree of conservation between Rad27 and FEN1. FEN1 inhibitor could be applied to yeast expressing different levels of Rad27 protein – since knockout of RAD27 confers a fitness defect to yeast, application of FEN1 inhibitor should mimic this effect unless a strain is made resistant by overexpression of RAD27. As well, FEN1 inhibitors could be subjected to homozygous deletion profiling (reviewed in Ericson et al., 2010). In this experiment, the compound is applied to the nonessential  95 gene homozygous deletion collection, and the resulting interaction profile can be clustered with previously described interaction sets to determine what proteins, complexes, or pathways have similar interaction profiles. This can help to identify the target of an unknown compound, or add weight to the idea that Rad27 is the target of a FEN1 inhibitor.  In chapter 5, a number of synthetic dosage lethal (SDL) interactions were discovered between rad27Δ and genes previously shown to be overexpressed in cancer. FEN1 inhibitors can be applied to tumor lines overexpressing these genes, to determine whether or not these tumor lines are more sensitive to FEN1 inhibition than other cell lines. If such lines aren’t available, lines overexpressing these genes can be created through lentiviral-mediated transduction to create a stable expression line (for an example, see Campeau et al., 2009). The latter experiment would be the ideal proof-of- principle, as matched pairs of cell lines could be generated, eliminating many of the genetic differences that confound experiments performed on cancer-derived cell lines.  Finally, to demonstrate the viability of a FEN1 inhibitor as an anticancer drug, the experiments listed above could be performed in a mouse xenograft model. For example, CDC4 knockout cells, or cells that inducibly express shRNA to deplete MRE11A, or cells overexpressing human orthologs of hits from the SDL screen, could be xenografted on to mice fed FEN1 inhibitor. If the tumors regress as the yeast or cell culture experiments predict, and there is little or no adverse effect on the murine host, this would demonstrate that the FEN1 inhibitor used is metabolized in an organismal context, and that it can reach the tumor and induce selective killing without toxicity.  96  A less artificial means to determine a FEN1 inhibitor’s ability to selectively kill cancer cells would be to apply it to a tumor that has spontaneously arisen, with a genotype expected to be sensitive to FEN1 inhibition. MSH2 is often inactivated in hereditary nonpolyposis colon cancer, and this is often due to a loss of heterozygosity. MSH2 genetically interacts with RAD27 in yeast, and MSH2+/- mouse cells have a similar mutation rate and response to DNA damage when compared to MSH2+/+ cells. Thus, FEN1 inhibitor could be fed to germline MSH2+/- mice, and tumor-free survival could be assayed. If FEN1 inhibitor selectively kills tumors that have undergone loss of MSH2 heterozygosity, then tumor-free survival would be expected to increase. 6.4 Other targets for therapeutic development  The work presented here has focused heavily on Rad27/FEN1, due to the fact that it encodes an enzyme with measurable activity and a previously-characterized in vitro assay, thereby facilitating biochemical inhibitor screening. However, Ctf4/WDHD1 remains a potential target for anticancer therapeutic development, as WDHD1 was as highly connected in the human CIN gene interaction network as FEN1. The central challenge to the development of Ctf4/WDHD1 as a target for therapeutics remains its lack of functional characterization. The data presented here add weight to the hypothesis that Ctf4 is a scaffolding protein that mediates and stabilizes interactions at the replisome (Jawad and Paoli, 2002; Zhu et al., 2007; Tanaka et al., 2009a). Protein-protein interactions were long considered “undruggable,” but there has been much work recently to develop chemical libraries for the express purpose of disrupting these interactions, such as through the use of natural products (Fuller et al., 2009; Bauer et al., 2010). In the case of WDHD1, it has been shown by bimolecular fluorescence complementation that  97 siRNA-mediated depletion of WDHD1 reduces binding of CDC45L to MCM2, MCM2 to GINS3, and CDC45L to GINS4 (Im et al., 2009). A cell-based screen could be conducted for compounds that can reduce the binding of these proteins, measured for example by bimolecular fluorescence complementation signal. The use of all three binding pairs in the same cell (utilizing, for example, GFP, RFP, and CFP complementation) would greatly reduce the chance of detecting a false lead that reduces binding by some other mechanism (Inglese et al., 2007).  The common thread that runs through the three highly connected CIN genes presented in this work, FEN1, WDHD1, and CHTF8, is that they all play important roles during DNA replication and repair. When this work was first undertaken, yeast genetic interaction data were available for only the nonessential genes; however, most DNA replication proteins (for example, all of the subunits of DNA polymerase α, the catalytic subunits of DNA polymerases δ and ε, all of the components of the MCM helicase, and all of the components of the GINS complex) are essential, and no genetic interaction data existed for them. With the advent of essential gene mutant collections (Ben-Aroya et al., 2008; Breslow et al., 2008; Li et al., 2011), these data will no doubt be forthcoming. The data presented here lead to the prediction that these essential DNA replication and repair proteins will be as highly connected as RAD27, CTF4, and CTF8, and thus will also be attractive targets for anticancer therapeutic development. 6.5 Therapeutic implications  Taken together, the data presented here and the available yeast genetic interaction data suggest that the replication fork is a major hub of synthetic lethal interactions. Chemotherapeutic drugs have been targeting DNA for as long as they have existed – the  98 first chemotherapeutic was the DNA alkylating agent mustine (reviewed in Hirsch, 2006). The mechanism of cancer killing has historically been ascribed to targeting the cancer proliferating phenotype – damaging highly proliferative tissue more than non proliferative tissue. From the continued advancement of whole-cancer genome sequencing, combined with the types of mutation data emerging and model organism genetics, it is becoming increasingly clear that targeting DNA replication also targets the cancer genotype. DNA polymerases, for example, have been the targets of small- molecule inhibitor screens in recent years (Martin et al., 2010; Jaiswal et al., 2011; Mizushina et al., 2011).  The work presented in this thesis has demonstrated that a cross-species candidate approach identifies viable targets for anticancer therapeutic development. Whole-cancer genome sequencing technologies are advancing at such a pace that sequencing patient tumor genomes will soon become routine (Stratton, 2011). Using so-called personalized medicine, biomarkers found in tumor genotypes will indicate the most potent therapeutics; these therapeutics will potentially be developed, as in this work, by combining model organism genetic data and human cell-based validation of biochemical screening. The application of model organism genetics to the development of cancer therapy thus will be of broad and substantial clinical significance.  99 References Adamson, B., A.Smogorzewska, F.D.Sigoillot, R.W.King, and S.J.Elledge. 2012. A genome-wide homologous recombination screen identifies the RNA-binding protein RBMX as a component of the DNA-damage response. Nat. Cell Biol. %19;14:318- 328. Akhoondi, S., D.Sun, L.N.von der, S.Apostolidou, K.Klotz, A.Maljukova, D.Cepeda, H.Fiegl, D.Dafou, C.Marth, E.Mueller-Holzner, M.Corcoran, M.Dagnell, S.Z.Nejad, B.N.Nayer, M.R.Zali, J.Hansson, S.Egyhazi, F.Petersson, P.Sangfelt, H.Nordgren, D.Grander, S.I.Reed, M.Widschwendter, O.Sangfelt, and C.Spruck. 2007. FBXW7/hCDC4 is a general tumor suppressor in human cancer. Cancer Res. 67:9006- 9012. Albertson, D.G. 2006. Gene amplification in cancer. Trends Genet. 22:447-455. Ansbach, A.B., C.Noguchi, I.W.Klansek, M.Heidlebaugh, T.M.Nakamura, and E.Noguchi. 2008. RFCCtf18 and the Swi1-Swi3 complex function in separate and redundant pathways required for the stabilization of replication forks to facilitate sister chromatid cohesion in Schizosaccharomyces pombe. Mol. Biol. Cell. 19:595-607. Arnon, J., D.Meirow, H.Lewis-Roness, and A.Ornoy. 2001. Genetic and teratogenic effects of cancer treatments on gametes and embryos. Hum. Reprod. Update. 7:394- 403. Baker, S.J., E.R.Fearon, J.M.Nigro, S.R.Hamilton, A.C.Preisinger, J.M.Jessup, P.vanTuinen, D.H.Ledbetter, D.F.Barker, Y.Nakamura, R.White, and B.Vogelstein. 1989. Chromosome 17 deletions and p53 gene mutations in colorectal carcinomas. Science. 244:217-221. Balgi, A.D. and M.Roberge. 2009. Screening for chemical inhibitors of heterologous proteins expressed in yeast using a simple growth-restoration assay. Methods Mol. Biol. 486:125-37.:125-137. Bambara, R.A., R.S.Murante, and L.A.Henricksen. 1997. Enzymes and Reactions at the Eukaryotic DNA Replication Fork. J. Biol. Chem. 272:4647-4650. Bamford, S., E.Dawson, S.Forbes, J.Clements, R.Pettett, A.Dogan, A.Flanagan, J.Teague, P.A.Futreal, M.R.Stratton, and R.Wooster. 2004. The COSMIC (Catalogue of Somatic Mutations in Cancer) database and website. Br. J. Cancer. 91:355-358. Barber, T.D., K.McManus, K.W.Yuen, M.Reis, G.Parmigiani, D.Shen, I.Barrett, Y.Nouhi, F.Spencer, S.Markowitz, V.E.Velculescu, K.W.Kinzler, B.Vogelstein, C.Lengauer, and P.Hieter. 2008. Chromatid cohesion defects may underlie chromosome instability in human colorectal cancers. Proc. Natl. Acad. Sci. U. S. A. 105:3443-3448. Baryshnikova, A., M.Costanzo, S.Dixon, F.J.Vizeacoumar, C.L.Myers, B.Andrews, and C.Boone. 2010. Synthetic genetic array (SGA) analysis in Saccharomyces cerevisiae and Schizosaccharomyces pombe. Methods Enzymol. 470:145-79. Epub;%2010 Mar 1.:145-179. Bauer, R.A., J.M.Wurst, and D.S.Tan. 2010. Expanding the range of 'druggable' targets with natural product-based libraries: an academic perspective. Curr. Opin. Chem. Biol. 14:308-314.  100 Ben-Aroya, S., N.Agmon, K.Yuen, T.Kwok, K.McManus, M.Kupiec, and P.Hieter. 2010. Proteasome nuclear activity affects chromosome stability by controlling the turnover of Mms22, a protein important for DNA repair. PLoS. Genet. %19;6:e1000852. Ben-Aroya, S., C.Coombes, T.Kwok, K.A.O'Donnell, J.D.Boeke, and P.Hieter. 2008. Toward a comprehensive temperature-sensitive mutant repository of the essential genes of Saccharomyces cerevisiae. Mol. Cell. 30:248-258. Ben-Aroya, S., A.Koren, B.Liefshitz, R.Steinlauf, and M.Kupiec. 2003. ELG1, a yeast gene required for genome stability, forms a complex related to replication factor C. Proc. Natl. Acad. Sci. U. S. A. 100:9906-9911. Ben-Neriah, Y., G.Q.Daley, A.M.Mes-Masson, O.N.Witte, and D.Baltimore. 1986. The chronic myelogenous leukemia-specific P210 protein is the product of the bcr/abl hybrid gene. Science. 233:212-214. Bermudez, V.P., A.Farina, I.Tappin, and J.Hurwitz. 2010. Influence of the human cohesion establishment factor Ctf4/AND-1 on DNA replication. J. Biol. Chem. %19.. Bhattacharyya, N.P., A.Ganesh, G.Phear, B.Richards, A.Skandalis, and M.Meuth. 1995. Molecular analysis of mutations in mutator colorectal carcinoma cell lines. Hum. Mol. Genet. 4:2057-2064. Birkbak, N.J., A.C.Eklund, Q.Li, S.E.McClelland, D.Endesfelder, P.Tan, I.B.Tan, A.L.Richardson, Z.Szallasi, and C.Swanton. 2011. Paradoxical relationship between chromosomal instability and survival outcome in cancer. Cancer Res. 71:3447-3452. Boone, C., H.Bussey, and B.J.Andrews. 2007. Exploring genetic interactions and networks with yeast. Nat. Rev. Genet. 8:437-449. Bornarth, C.J., T.A.Ranalli, L.A.Henricksen, A.F.Wahl, and R.A.Bambara. 1999. Effect of flap modifications on human FEN1 cleavage. Biochemistry. 38:13347-13354. Bos, J.L., E.R.Fearon, S.R.Hamilton, V.M.Verlaan-de, J.H.van Boom, A.J.van der Eb, and B.Vogelstein. 1987. Prevalence of ras gene mutations in human colorectal cancers. Nature. 327:293-297. Breslow, D.K., D.M.Cameron, S.R.Collins, M.Schuldiner, J.Stewart-Ornstein, H.W.Newman, S.Braun, H.D.Madhani, N.J.Krogan, and J.S.Weissman. 2008. A comprehensive strategy enabling high-resolution functional analysis of the yeast genome. Nat. Methods. 5:711-718. Bryant, H.E., N.Schultz, H.D.Thomas, K.M.Parker, D.Flower, E.Lopez, S.Kyle, M.Meuth, N.J.Curtin, and T.Helleday. 2005. Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP-ribose) polymerase. Nature. 434:913-917. Busino, L., S.E.Millman, L.Scotto, C.A.Kyratsous, V.Basrur, O.O'Connor, A.Hoffmann, K.S.Elenitoba-Johnson, and M.Pagano. 2012. Fbxw7alpha- and GSK3-mediated degradation of p100 is a pro-survival mechanism in multiple myeloma. Nat. Cell Biol. 14:375-385. Bylund, G.O. and P.M.Burgers. 2005. Replication protein A-directed unloading of PCNA by the Ctf18 cohesion establishment complex. Mol. Cell Biol. 25:5445-5455. Cahill, D.P., C.Lengauer, J.Yu, G.J.Riggins, J.K.V.Willson, S.D.Markowitz, K.W.Kinzler, and B.Vogelstein. 1998. Mutations of mitotic checkpoint genes in human cancers. Nature. 392:300-303. Campeau, E., V.E.Ruhl, F.Rodier, C.L.Smith, B.L.Rahmberg, J.O.Fuss, J.Campisi, P.Yaswen, P.K.Cooper, and P.D.Kaufman. 2009. A versatile viral system for expression and depletion of proteins in mammalian cells. PLoS. One. 4:e6529.  101 Chang, M., B.Luke, C.Kraft, Z.Li, M.Peter, J.Lingner, and R.Rothstein. 2009. Telomerase is essential to alleviate pif1-induced replication stress at telomeres. Genetics. 183:779-791. Chen, C. and R.D.Kolodner. 1999. Gross chromosomal rearrangements in Saccharomyces cerevisiae replication and recombination defective mutants. Nat. Genet. 23:81-85. Cherry, J.M., E.L.Hong, C.Amundsen, R.Balakrishnan, G.Binkley, E.T.Chan, K.R.Christie, M.C.Costanzo, S.S.Dwight, S.R.Engel, D.G.Fisk, J.E.Hirschman, B.C.Hitz, K.Karra, C.J.Krieger, S.R.Miyasato, R.S.Nash, J.Park, M.S.Skrzypek, M.Simison, S.Weng, and E.D.Wong. 2012. Saccharomyces Genome Database: the genomics resource of budding yeast. Nucleic Acids Res. 40:D700-D705. Chien, C.T., P.L.Bartel, R.Sternglanz, and S.Fields. 1991. The two-hybrid system: a method to identify and clone genes for proteins that interact with a protein of interest. Proc. Natl. Acad. Sci. U. S. A. 88:9578-9582. Collins, N., R.McManus, R.Wooster, J.Mangion, S.Seal, S.R.Lakhani, W.Ormiston, P.A.Daly, D.Ford, D.F.Easton, and . 1995. Consistent loss of the wild type allele in breast cancers from a family linked to the BRCA2 gene on chromosome 13q12-13. Oncogene. %20;10:1673-1675. Collins, S.R., K.M.Miller, N.L.Maas, A.Roguev, J.Fillingham, C.S.Chu, M.Schuldiner, M.Gebbia, J.Recht, M.Shales, H.Ding, H.Xu, J.Han, K.Ingvarsdottir, B.Cheng, B.Andrews, C.Boone, S.L.Berger, P.Hieter, Z.Zhang, G.W.Brown, C.J.Ingles, A.Emili, C.D.Allis, D.P.Toczyski, J.S.Weissman, J.F.Greenblatt, and N.J.Krogan. 2007. Functional dissection of protein complexes involved in yeast chromosome biology using a genetic interaction map. Nature. 446:806-810. Costanzo, M., A.Baryshnikova, J.Bellay, Y.Kim, E.D.Spear, C.S.Sevier, H.Ding, J.L.Koh, K.Toufighi, S.Mostafavi, J.Prinz, R.P.St Onge, B.VanderSluis, T.Makhnevych, F.J.Vizeacoumar, S.Alizadeh, S.Bahr, R.L.Brost, Y.Chen, M.Cokol, R.Deshpande, Z.Li, Z.Y.Lin, W.Liang, M.Marback, J.Paw, B.J.San Luis, E.Shuteriqi, A.H.Tong, D.N.van, I.M.Wallace, J.A.Whitney, M.T.Weirauch, G.Zhong, H.Zhu, W.A.Houry, M.Brudno, S.Ragibizadeh, B.Papp, C.Pal, F.P.Roth, G.Giaever, C.Nislow, O.G.Troyanskaya, H.Bussey, G.D.Bader, A.C.Gingras, Q.D.Morris, P.M.Kim, C.A.Kaiser, C.L.Myers, B.J.Andrews, and C.Boone. 2010. The genetic landscape of a cell. Science. 327:425-431. Davierwala, A.P., J.Haynes, Z.Li, R.L.Brost, M.D.Robinson, L.Yu, S.Mnaimneh, H.Ding, H.Zhu, Y.Chen, X.Cheng, G.W.Brown, C.Boone, B.J.Andrews, and T.R.Hughes. 2005. The synthetic genetic interaction spectrum of essential genes. Nat. Genet. 37:1147-1152. Davis, H. and I.Tomlinson. 2012. CDC4/FBXW7 and the 'just enough' model of tumourigenesis. J. Pathol. 227:131-135. de Klein A., A.G.van Kessel, G.Grosveld, C.R.Bartram, A.Hagemeijer, D.Bootsma, N.K.Spurr, N.Heisterkamp, J.Groffen, and J.R.Stephenson. 1982. A cellular oncogene is translocated to the Philadelphia chromosome in chronic myelocytic leukaemia. Nature. 300:765-767. Dixon, S.J., B.J.Andrews, and C.Boone. 2009. Exploring the conservation of synthetic lethal genetic interaction networks. Commun. Integr. Biol. 2:78-81.  102 Dixon, S.J., Y.Fedyshyn, J.L.Koh, T.S.Prasad, C.Chahwan, G.Chua, K.Toufighi, A.Baryshnikova, J.Hayles, K.L.Hoe, D.U.Kim, H.O.Park, C.L.Myers, A.Pandey, D.Durocher, B.J.Andrews, and C.Boone. 2008. Significant conservation of synthetic lethal genetic interaction networks between distantly related eukaryotes. Proc. Natl. Acad. Sci. U. S. A. 105:16653-16658. Dolma, S., S.L.Lessnick, W.C.Hahn, and B.R.Stockwell. 2003. Identification of genotype-selective antitumor agents using synthetic lethal chemical screening in engineered human tumor cells. Cancer Cell. 3:285-296. Donehower, L.A., M.Harvey, B.L.Slagle, M.J.McArthur, C.A.Montgomery, Jr., J.S.Butel, and A.Bradley. 1992. Mice deficient for p53 are developmentally normal but susceptible to spontaneous tumours. Nature. %19;356:215-221. Dorjsuren, D., D.Kim, D.J.Maloney, D.M.Wilson, III, and A.Simeonov. 2010. Complementary non-radioactive assays for investigation of human flap endonuclease 1 activity. Nucleic Acids Res. Dupre, A., L.Boyer-Chatenet, R.M.Sattler, A.P.Modi, J.H.Lee, M.L.Nicolette, L.Kopelovich, M.Jasin, R.Baer, T.T.Paull, and J.Gautier. 2008. A forward chemical genetic screen reveals an inhibitor of the Mre11-Rad50-Nbs1 complex. Nat. Chem. Biol. 4:119-125. el-Deiry, W.S., T.Tokino, V.E.Velculescu, D.B.Levy, R.Parsons, J.M.Trent, D.Lin, W.E.Mercer, K.W.Kinzler, and B.Vogelstein. 1993. WAF1, a potential mediator of p53 tumor suppression. Cell. %19;75:817-825. Ericson, E., S.Hoon, R.P.St Onge, G.Giaever, and C.Nislow. 2010. Exploring gene function and drug action using chemogenomic dosage assays. Methods Enzymol. 470:233-255. Farmer, H., N.McCabe, C.J.Lord, A.N.Tutt, D.A.Johnson, T.B.Richardson, M.Santarosa, K.J.Dillon, I.Hickson, C.Knights, N.M.Martin, S.P.Jackson, G.C.Smith, and A.Ashworth. 2005. Targeting the DNA repair defect in BRCA mutant cells as a therapeutic strategy. Nature. 434:917-921. Finger, L.D., M.S.Blanchard, C.A.Theimer, B.Sengerova, P.Singh, V.Chavez, F.Liu, J.A.Grasby, and B.Shen. 2009. The 3'-flap pocket of human flap endonuclease 1 is critical for substrate binding and catalysis. J. Biol. Chem. 284:22184-22194. Fishel, R., M.K.Lescoe, M.R.Rao, N.G.Copeland, N.A.Jenkins, J.Garber, M.Kane, and R.Kolodner. 1993. The human mutator gene homolog MSH2 and its association with hereditary nonpolyposis colon cancer. Cell. 75:1027-1038. Forer, L., S.Schonherr, H.Weissensteiner, F.Haider, T.Kluckner, C.Gieger, H.E.Wichmann, G.Specht, F.Kronenberg, and A.Kloss-Brandstatter. 2010. CONAN: copy number variation analysis software for genome-wide association studies. BMC. Bioinformatics. 11:318.:318. Fraser, A. 2004. Towards full employment: using RNAi to find roles for the redundant. Oncogene. 23:8346-8352. Frost, A., M.G.Elgort, O.Brandman, C.Ives, S.R.Collins, L.Miller-Vedam, J.Weibezahn, M.Y.Hein, I.Poser, M.Mann, A.A.Hyman, and J.S.Weissman. 2012. Functional Repurposing Revealed by Comparing S. pombe and S. cerevisiae Genetic Interactions. Cell. 149:1339-1352. Fuller, J.C., N.J.Burgoyne, and R.M.Jackson. 2009. Predicting druggable binding sites at the protein-protein interface. Drug Discov. Today. 14:155-161.  103 Futreal, P.A., L.Coin, M.Marshall, T.Down, T.Hubbard, R.Wooster, N.Rahman, and M.R.Stratton. 2004. A census of human cancer genes. Nat. Rev. Cancer. 4:177-183. Gambus, A., R.C.Jones, A.Sanchez-Diaz, M.Kanemaki, D.F.van, R.D.Edmondson, and K.Labib. 2006. GINS maintains association of Cdc45 with MCM in replisome progression complexes at eukaryotic DNA replication forks. Nat. Cell Biol. 8:358-366. Gambus, A., D.F.van, D.Polychronopoulos, M.Foltman, R.C.Jones, R.D.Edmondson, A.Calzada, and K.Labib. 2009. A key role for Ctf4 in coupling the MCM2-7 helicase to DNA polymerase alpha within the eukaryotic replisome. EMBO J. 28:2992-3004. Gellon, L., D.F.Razidlo, O.Gleeson, L.Verra, D.Schulz, R.S.Lahue, and C.H.Freudenreich. 2011. New functions of Ctf18-RFC in preserving genome stability outside its role in sister chromatid cohesion. PLoS. Genet. 7:e1001298. Gill, S., R.R.Thomas, and R.M.Goldberg. 2003. Review article: colorectal cancer chemotherapy. Aliment. Pharmacol. Ther. 18:683-692. Gloor, J.W., L.Balakrishnan, J.L.Campbell, and R.A.Bambara. 2012. Biochemical analyses indicate that binding and cleavage specificities define the ordered processing of human Okazaki fragments by Dna2 and FEN1. Nucleic Acids Res. Gotter, A.L., C.Suppa, and B.S.Emanuel. 2007. Mammalian TIMELESS and Tipin are evolutionarily conserved replication fork-associated factors. J. Mol. Biol. 366:36-52. Greene, A.L., J.R.Snipe, D.A.Gordenin, and M.A.Resnick. 1999. Functional analysis of human FEN1 in Saccharomyces cerevisiae and its role in genome instability. Hum Mol Genet 8:2263-2273. Gu, Y., C.W.Turck, and D.O.Morgan. 1993. Inhibition of CDK2 activity in vivo by an associated 20K regulatory subunit. Nature. 366:707-710. Guacci, V., D.Koshland, and A.Strunnikov. 1997. A direct link between sister chromatid cohesion and chromosome condensation revealed through the analysis of MCD1 in S. cerevisiae. Cell. 91:47-57. Hall, J.M., M.K.Lee, B.Newman, J.E.Morrow, L.A.Anderson, B.Huey, and M.C.King. 1990. Linkage of early-onset familial breast cancer to chromosome 17q21. Science. 250:1684-1689. Han, J., Q.Li, L.McCullough, C.Kettelkamp, T.Formosa, and Z.Zhang. 2010. Ubiquitylation of FACT by the cullin-E3 ligase Rtt101 connects FACT to DNA replication. Genes Dev. 24:1485-1490. Hanahan, D. and R.A.Weinberg. 2000. The hallmarks of cancer. Cell. 100:57-70. Hanahan, D. and R.A.Weinberg. 2011. Hallmarks of cancer: the next generation. Cell. 144:646-674. Hanna, J.S., E.S.Kroll, V.Lundblad, and F.A.Spencer. 2001. Saccharomyces cerevisiae CTF18 and CTF4 Are Required for Sister Chromatid Cohesion. Mol. Cell. Biol. 21:3144-3158. Harper, J.W., G.R.Adami, N.Wei, K.Keyomarsi, and S.J.Elledge. 1993. The p21 Cdk- interacting protein Cip1 is a potent inhibitor of G1 cyclin-dependent kinases. Cell. %19;75:805-816. Harris, S.D. and J.E.Hamer. 1995. sepB: an Aspergillus nidulans gene involved in chromosome segregation and the initiation of cytokinesis. EMBO J. 14:5244-5257. Hartwell, L.H., P.Szankasi, C.J.Roberts, A.W.Murray, and S.H.Friend. 1997. Integrating genetic approaches into the discovery of anticancer drugs. Science. 278:1064-1068.  104 Helleday, T. 2011. The underlying mechanism for the PARP and BRCA synthetic lethality: clearing up the misunderstandings. Mol. Oncol. 5:387-393. Hieter, P., C.Mann, M.Snyder, and R.W.Davis. 1985. Mitotic stability of yeast chromosomes: a colony color assay that measures nondisjunction and chromosome loss. Cell. 40:381-392. Hiramoto, T., T.Nakanishi, T.Sumiyoshi, T.Fukuda, S.Matsuura, H.Tauchi, K.Komatsu, Y.Shibasaki, H.Inui, M.Watatani, M.Yasutomi, K.Sumii, G.Kajiyama, N.Kamada, K.Miyagawa, and K.Kamiya. 1999. Mutations of a novel human RAD54 homologue, RAD54B, in primary cancer. Oncogene. 18:3422-3426. Hirsch, J. 2006. An anniversary for cancer chemotherapy. JAMA. 296:1518-1520. Hoon, S., A.M.Smith, I.M.Wallace, S.Suresh, M.Miranda, E.Fung, M.Proctor, K.M.Shokat, C.Zhang, R.W.Davis, G.Giaever, R.P.St Onge, and C.Nislow. 2008. An integrated platform of genomic assays reveals small-molecule bioactivities. Nat. Chem. Biol. 4:498-506. Hou, F. and H.Zou. 2005. Two human orthologues of Eco1/Ctf7 acetyltransferases are both required for proper sister-chromatid cohesion. Mol. Biol. Cell. 16:3908-3918. Hu, Y., A.Rolfs, B.Bhullar, T.V.Murthy, C.Zhu, M.F.Berger, A.A.Camargo, F.Kelley, S.McCarron, D.Jepson, A.Richardson, J.Raphael, D.Moreira, E.Taycher, D.Zuo, S.Mohr, M.F.Kane, J.Williamson, A.Simpson, M.L.Bulyk, E.Harlow, G.Marsischky, R.D.Kolodner, and J.LaBaer. 2007. Approaching a complete repository of sequence- verified protein-encoding clones for Saccharomyces cerevisiae. Genome Res. 17:536- 543. Im, J.S., S.H.Ki, A.Farina, D.S.Jung, J.Hurwitz, and J.K.Lee. 2009. Assembly of the Cdc45-Mcm2-7-GINS complex in human cells requires the Ctf4/And-1, RecQL4, and Mcm10 proteins. Proc. Natl. Acad. Sci. U. S. A. 106:15628-15632. Inglese, J., R.L.Johnson, A.Simeonov, M.Xia, W.Zheng, C.P.Austin, and D.S.Auld. 2007. High-throughput screening assays for the identification of chemical probes. Nat. Chem. Biol. 3:466-479. Iversen, P.W., B.J.Eastwood, G.S.Sittampalam, and K.L.Cox. 2006. A comparison of assay performance measures in screening assays: signal window, Z' factor, and assay variability ratio. J. Biomol. Screen. 11:247-252. Jacks, T., L.Remington, B.O.Williams, E.M.Schmitt, S.Halachmi, R.T.Bronson, and R.A.Weinberg. 1994. Tumor spectrum analysis in p53-mutant mice. Curr. Biol. 4:1-7. Jaiswal, A.S., S.Banerjee, R.Aneja, F.H.Sarkar, D.A.Ostrov, and S.Narayan. 2011. DNA Polymerase beta as a Novel Target for Chemotherapeutic Intervention of Colorectal Cancer. PLoS. One. 6:e16691. Jawad, Z. and M.Paoli. 2002. Novel sequences propel familiar folds. Structure. 10:447- 454. Ji, Z., F.C.Mei, P.L.Lory, S.R.Gilbertson, Y.Chen, and X.Cheng. 2009. Chemical genetic screening of KRAS-based synthetic lethal inhibitors for pancreatic cancer. Front Biosci. 14:2904-2910. Kemp, Z., A.Rowan, W.Chambers, N.Wortham, S.Halford, O.Sieber, N.Mortensen, H.A.von, T.Gunther, M.Ilyas, and I.Tomlinson. 2005. CDC4 mutations occur in a subset of colorectal cancers but are not predicted to cause loss of function and are not associated with chromosomal instability. Cancer Res. 65:11361-11366.  105 Knudson, A.G., Jr. 1971. Mutation and cancer: statistical study of retinoblastoma. Proc. Natl. Acad. Sci. U. S. A. 68:820-823. Koh, M.S., M.Ittmann, D.Kadmon, T.C.Thompson, and F.S.Leach. 2006. CDC4 gene expression as potential biomarker for targeted therapy in prostate cancer. Cancer Biol. Ther. 5:78-83. Kohler, A., M.S.Schmidt-Zachmann, and W.W.Franke. 1997. AND-1, a natural chimeric DNA-binding protein, combines an HMG-box with regulatory WD-repeats. J. Cell Sci. 110:1051-1062. Kouprina, N., E.Kroll, V.Bannikov, V.Bliskovsky, R.Gizatullin, A.Kirillov, B.Shestopalov, V.Zakharyev, P.Hieter, and F.Spencer. 1992. CTF4 (CHL15) mutants exhibit defective DNA metabolism in the yeast Saccharomyces cerevisiae. Mol. Cell. Biol. 12:5736-5747. Krogan, N.J., G.Cagney, H.Yu, G.Zhong, X.Guo, A.Ignatchenko, J.Li, S.Pu, N.Datta, A.P.Tikuisis, T.Punna, J.M.Peregr+¡n-Alvarez, M.Shales, X.Zhang, M.Davey, M.D.Robinson, A.Paccanaro, J.E.Bray, A.Sheung, B.Beattie, D.P.Richards, V.Canadien, A.Lalev, F.Mena, P.Wong, A.Starostine, M.M.Canete, J.Vlasblom, S.Wu, C.Orsi, S.R.Collins, S.Chandran, R.Haw, J.J.Rilstone, K.Gandi, N.J.Thompson, G.Musso, P.St Onge, S.Ghanny, M.H.Y.Lam, G.Butland, A.M.taf-Ul, S.Kanaya, A.Shilatifard, E.O'Shea, J.S.Weissman, C.J.Ingles, T.R.Hughes, J.Parkinson, M.Gerstein, S.J.Wodak, A.Emili, and J.F.Greenblatt. 2006. Global landscape of protein complexes in the yeast Saccharomyces cerevisiae. Nature. 440:637-643. Kroll, E.S., K.M.Hyland, P.Hieter, and J.J.Li. 1996. Establishing genetic interactions by a synthetic dosage lethality phenotype. Genetics. 143:95-102. Lane, D.P. and S.Benchimol. 1990. p53: oncogene or anti-oncogene? Genes Dev. 4:1-8. Larsen, E., C.Gran, B.E.Saether, E.Seeberg, and A.Klungland. 2003. Proliferation failure and gamma radiation sensitivity of Fen1 null mutant mice at the blastocyst stage. Mol. Cell Biol. 23:5346-5353. Lazzaro, F., M.Giannattasio, F.Puddu, M.Granata, A.Pellicioli, P.Plevani, and M.Muzi- Falconi. 2009. Checkpoint mechanisms at the intersection between DNA damage and repair. DNA Repair (Amst). 8:1055-1067. Lee, A.J., D.Endesfelder, A.J.Rowan, A.Walther, N.J.Birkbak, P.A.Futreal, J.Downward, Z.Szallasi, I.P.Tomlinson, M.Howell, M.Kschischo, and C.Swanton. 2011. Chromosomal instability confers intrinsic multidrug resistance. Cancer Res. 71:1858- 1870. Lehner, B., C.Crombie, J.Tischler, A.Fortunato, and A.G.Fraser. 2006. Systematic mapping of genetic interactions in Caenorhabditis elegans identifies common modifiers of diverse signaling pathways. Nat. Genet. 38:896-903. Lengronne, A., J.McIntyre, Y.Katou, Y.Kanoh, K.P.Hopfner, K.Shirahige, and F.Uhlmann. 2006. Establishment of sister chromatid cohesion at the S. cerevisiae replication fork. Mol. Cell. 23:787-799. Li, Z., F.J.Vizeacoumar, S.Bahr, J.Li, J.Warringer, F.S.Vizeacoumar, R.Min, B.VanderSluis, J.Bellay, M.Devit, J.A.Fleming, A.Stephens, J.Haase, Z.Y.Lin, A.Baryshnikova, H.Lu, Z.Yan, K.Jin, S.Barker, A.Datti, G.Giaever, C.Nislow, C.Bulawa, C.L.Myers, M.Costanzo, A.C.Gingras, Z.Zhang, A.Blomberg, K.Bloom, B.Andrews, and C.Boone. 2011. Systematic exploration of essential yeast gene function with temperature-sensitive mutants. Nat. Biotechnol. 29:361-367.  106 Lipinski, C.A., F.Lombardo, B.W.Dominy, and P.J.Feeney. 2001. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 46:3-26. Liu, Y. and R.A.Bambara. 2003. Analysis of human flap endonuclease 1 mutants reveals a mechanism to prevent triplet repeat expansion. J. Biol. Chem. 278:13728-13739. Liu, Y., W.A.Beard, D.D.Shock, R.Prasad, E.W.Hou, and S.H.Wilson. 2005. DNA polymerase beta and flap endonuclease 1 enzymatic specificities sustain DNA synthesis for long patch base excision repair. J. Biol. Chem. 280:3665-3674. Loeb, L.A. 1994. Microsatellite instability: marker of a mutator phenotype in cancer. Cancer Res. 54:5059-5063. Luo, J., M.J.Emanuele, D.Li, C.J.Creighton, M.R.Schlabach, T.F.Westbrook, K.K.Wong, and S.J.Elledge. 2009. A genome-wide RNAi screen identifies multiple synthetic lethal interactions with the Ras oncogene. Cell. 137:835-848. Macarron, R., M.N.Banks, D.Bojanic, D.J.Burns, D.A.Cirovic, T.Garyantes, D.V.Green, R.P.Hertzberg, W.P.Janzen, J.W.Paslay, U.Schopfer, and G.S.Sittampalam. 2011. Impact of high-throughput screening in biomedical research. Nat. Rev. Drug Discov. 10:188-195. Macarron, R. and R.P.Hertzberg. 2009. Design and implementation of high-throughput screening assays. Methods Mol. Biol. 565:1-32.:1-32. Maine, G.T., P.Sinha, and B.K.Tye. 1984. Mutants of S. cerevisiae defective in the maintenance of minichromosomes. Genetics. 106:365-385. Marcotte, R., K.R.Brown, F.Suarez, A.Sayad, K.Karamboulas, P.M.Krzyzanowski, F.Sircoulomb, M.Medrano, Y.Fedyshyn, J.L.Koh, D.D.van, B.Fedyshyn, M.Luhova, G.C.Brito, F.J.Vizeacoumar, F.S.Vizeacoumar, A.Datti, D.Kasimer, A.Buzina, P.Mero, C.Misquitta, J.Normand, M.Haider, T.Ketela, J.L.Wrana, R.Rottapel, B.G.Neel, and J.Moffat. 2012. Essential gene profiles in breast, pancreatic, and ovarian cancer cells. Cancer Discov. 2:172-189. Martin, S.A., N.McCabe, M.Mullarkey, R.Cummins, D.J.Burgess, Y.Nakabeppu, S.Oka, E.Kay, C.J.Lord, and A.Ashworth. 2010. DNA polymerases as potential therapeutic targets for cancers deficient in the DNA mismatch repair proteins MSH2 or MLH1. Cancer Cell. 17:235-248. Mayer, M.L., S.P.Gygi, R.Aebersold, and P.Hieter. 2001. Identification of RFC(Ctf18p, Ctf8p, Dcc1p): an alternative RFC complex required for sister chromatid cohesion in S. cerevisiae. Mol. Cell. 7:959-970. McLellan, J., N.O'Neil, S.Tarailo, J.Stoepel, J.Bryan, A.Rose, and P.Hieter. 2009. Synthetic lethal genetic interactions that decrease somatic cell proliferation in Caenorhabditis elegans identify the alternative RFC CTF18 as a candidate cancer drug target. Mol. Biol. Cell. 20:5306-5313. McLellan, J.L., N.J.O'Neil, I.Barrett, E.Ferree, D.M.van Pel, K.Ushey, P.Sipahimalani, J.Bryan, A.M.Rose, and P.Hieter. 2012. Synthetic lethality of cohesins with PARPs and replication fork mediators. PLoS. Genet. 8:e1002574. McManus, K.J., I.J.Barrett, Y.Nouhi, and P.Hieter. 2009. Specific synthetic lethal killing of RAD54B-deficient human colorectal cancer cells by FEN1 silencing. Proc. Natl. Acad. Sci. U. S. A. 106:3276-3281. McManus, K.J., V.L.Biron, R.Heit, D.A.Underhill, and M.J.Hendzel. 2006. Dynamic changes in histone H3 lysine 9 methylations: identification of a mitosis-specific  107 function for dynamic methylation in chromosome congression and segregation. J. Biol. Chem. 281:8888-8897. Measday, V., K.Baetz, J.Guzzo, K.Yuen, T.Kwok, B.Sheikh, H.Ding, R.Ueta, T.Hoac, B.Cheng, I.Pot, A.Tong, Y.Yamaguchi-Iwai, C.Boone, P.Hieter, and B.Andrews. 2005. Systematic yeast synthetic lethal and synthetic dosage lethal screens identify genes required for chromosome segregation. Proc. Natl. Acad. Sci. U. S. A. 102:13956-13961. Mendes-Pereira, A.M., S.A.Martin, R.Brough, A.McCarthy, J.R.Taylor, J.S.Kim, T.Waldman, C.J.Lord, and A.Ashworth. 2009. Synthetic lethal targeting of PTEN mutant cells with PARP inhibitors. EMBO Mol. Med. 1:315-322. Mendes-Pereira, A.M., D.Sims, T.Dexter, K.Fenwick, I.Assiotis, I.Kozarewa, C.Mitsopoulos, J.Hakas, M.Zvelebil, C.J.Lord, and A.Ashworth. 2012. Genome-wide functional screen identifies a compendium of genes affecting sensitivity to tamoxifen. Proc. Natl. Acad. Sci. U. S. A. 109:2730-2735. Merkle, C.J., L.M.Karnitz, J.T.Henry-Sanchez, and J.Chen. 2003. Cloning and characterization of hCTF18, hCTF8, and hDCC1. Human homologs of a Saccharomyces cerevisiae complex involved in sister chromatid cohesion establishment. J. Biol. Chem. 278:30051-30056. Michaelis, C., R.Ciosk, and K.Nasmyth. 1997. Cohesins: chromosomal proteins that prevent premature separation of sister chromatids. Cell. 91:35-45. Miki, Y., J.Swensen, D.Shattuck-Eidens, P.A.Futreal, K.Harshman, S.Tavtigian, Q.Liu, C.Cochran, L.M.Bennett, W.Ding, and . 1994. A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1. Science. 266:66-71. Miles, J. and T.Formosa. 1992. Protein affinity chromatography with purified yeast DNA polymerase alpha detects proteins that bind to DNA polymerase. Proc. Natl. Acad. Sci. U. S. A. 89:1276-1280. Milne, A.N., R.Leguit, W.E.Corver, F.H.Morsink, M.Polak, W.W.de Leng, R.Carvalho, and G.J.Offerhaus. 2010. Loss of CDC4/FBXW7 in gastric carcinoma. Cell Oncol. 32:347-359. Mimura, S., T.Yamaguchi, S.Ishii, E.Noro, T.Katsura, C.Obuse, and T.Kamura. 2010. Cul8/Rtt101 forms a variety of protein complexes that regulate DNA damage response and transcriptional silencing. J. Biol. Chem. 285:9858-9867. Miyaki, M., T.Yamaguchi, T.Iijima, K.Takahashi, H.Matsumoto, and T.Mori. 2009. Somatic mutations of the CDC4 (FBXW7) gene in hereditary colorectal tumors. Oncology. 76:430-434. Mizushina, Y., N.Maeda, I.Kuriyama, and H.Yoshida. 2011. Dehydroaltenusin is a specific inhibitor of mammalian DNA polymerase alpha. Expert. Opin. Investig. Drugs. 20:1523-1534. Moffat, J., D.A.Grueneberg, X.Yang, S.Y.Kim, A.M.Kloepfer, G.Hinkle, B.Piqani, T.M.Eisenhaure, B.Luo, J.K.Grenier, A.E.Carpenter, S.Y.Foo, S.A.Stewart, B.R.Stockwell, N.Hacohen, W.C.Hahn, E.S.Lander, D.M.Sabatini, and D.E.Root. 2006. A lentiviral RNAi library for human and mouse genes applied to an arrayed viral high-content screen. Cell. 124:1283-1298. Molenaar, J.J., M.E.Ebus, D.Geerts, J.Koster, F.Lamers, L.J.Valentijn, E.M.Westerhout, R.Versteeg, and H.N.Caron. 2009. Inactivation of CDK2 is synthetically lethal to MYCN over-expressing cancer cells. Proc. Natl. Acad. Sci. U. S. A. 106:12968-12973.  108 Noon, A.T. and A.A.Goodarzi. 2011. 53BP1-mediated DNA double strand break repair: insert bad pun here. DNA Repair (Amst). 10:1071-1076. Nordling, C. 1953. A new theory on cancer-inducing mechanism. Br. J. Cancer. 7:68-72. Ogiwara, H., A.Ui, M.S.Lai, T.Enomoto, and M.Seki. 2007. Chl1 and Ctf4 are required for damage-induced recombinations. Biochem. Biophys. Res. Commun. 354:222-226. Pan, X., D.S.Yuan, D.Xiang, X.Wang, S.Sookhai-Mahadeo, J.S.Bader, P.Hieter, F.Spencer, and J.D.Boeke. 2004. A robust toolkit for functional profiling of the yeast genome. Mol. Cell. 16:487-496. Panda, H., A.S.Jaiswal, P.E.Corsino, M.L.Armas, B.K.Law, and S.Narayan. 2009. Amino acid Asp181 of 5'-flap endonuclease 1 is a useful target for chemotherapeutic development. Biochemistry. 48:9952-9958. Peltomaki, P., L.A.Aaltonen, P.Sistonen, L.Pylkkanen, J.P.Mecklin, H.Jarvinen, J.S.Green, J.R.Jass, J.L.Weber, F.S.Leach, and . 1993. Genetic mapping of a locus predisposing to human colorectal cancer. Science. 260:810-812. Petronczki, M., B.Chwalla, M.F.Siomos, S.Yokobayashi, W.Helmhart, A.M.Deutschbauer, R.W.Davis, Y.Watanabe, and K.Nasmyth. 2004. Sister-chromatid cohesion mediated by the alternative RF-CCtf18/Dcc1/Ctf8, the helicase Chl1 and the polymerase-alpha-associated protein Ctf4 is essential for chromatid disjunction during meiosis II. J. Cell Sci. 117:3547-3559. Prieto, I., J.A.Suja, N.Pezzi, L.Kremer, A.Martinez, J.S.Rufas, and J.L.Barbero. 2001. Mammalian STAG3 is a cohesin specific to sister chromatid arms in meiosis I. Nat. Cell Biol. 3:761-766. Pylayeva-Gupta, Y., E.Grabocka, and D.Bar-Sagi. 2011. RAS oncogenes: weaving a tumorigenic web. Nat. Rev. Cancer. 11:761-774. Rajagopalan, H., P.V.Jallepalli, C.Rago, V.E.Velculescu, K.W.Kinzler, B.Vogelstein, and C.Lengauer. 2004. Inactivation of hCDC4 can cause chromosomal instability. Nature. 428:77-81. Roguev, A., M.Wiren, J.S.Weissman, and N.J.Krogan. 2007. High-throughput genetic interaction mapping in the fission yeast Schizosaccharomyces pombe. Nat. Methods. 4:861-866. Roy, R., J.Chun, and S.N.Powell. 2011. BRCA1 and BRCA2: different roles in a common pathway of genome protection. Nat. Rev. Cancer. 12:68-78. Ryan, C.J., A.Roguev, K.Patrick, J.Xu, H.Jahari, Z.Tong, P.Beltrao, M.Shales, H.Qu, S.R.Collins, J.I.Kliegman, L.Jiang, D.Kuo, E.Tosti, H.S.Kim, W.Edelmann, M.C.Keogh, D.Greene, C.Tang, P.Cunningham, K.M.Shokat, G.Cagney, J.P.Svensson, C.Guthrie, P.J.Espenshade, T.Ideker, and N.J.Krogan. 2012. Hierarchical Modularity and the Evolution of Genetic Interactomes across Species. Mol. Cell. 46:691-704. Santarius, T., J.Shipley, D.Brewer, M.R.Stratton, and C.S.Cooper. 2010. A census of amplified and overexpressed human cancer genes. Nat. Rev. Cancer. 10:59-64. Schlesinger, M.B. and T.Formosa. 2000. POB3 is required for both transcription and replication in the yeast Saccharomyces cerevisiae. Genetics. 155:1593-1606. Sheltzer, J.M., H.M.Blank, S.J.Pfau, Y.Tange, B.M.George, T.J.Humpton, I.L.Brito, Y.Hiraoka, O.Niwa, and A.Amon. 2011. Aneuploidy drives genomic instability in yeast. Science. %19;333:1026-1030.  109 Singh, P., L.Zheng, V.Chavez, J.Qiu, and B.Shen. 2007. Concerted action of exonuclease and Gap-dependent endonuclease activities of FEN-1 contributes to the resolution of triplet repeat sequences (CTG)n- and (GAA)n-derived secondary structures formed during maturation of Okazaki fragments. J. Biol. Chem. 282:3465-3477. Smolka, M.B., C.P.Albuquerque, S.H.Chen, and H.Zhou. 2007. Proteome-wide identification of in vivo targets of DNA damage checkpoint kinases. Proc. Natl. Acad. Sci. U. S. A. 104:10364-10369. Solomon, D.A., T.Kim, L.A.az-Martinez, J.Fair, A.G.Elkahloun, B.T.Harris, J.A.Toretsky, S.A.Rosenberg, N.Shukla, M.Ladanyi, Y.Samuels, C.D.James, H.Yu, J.S.Kim, and T.Waldman. 2011. Mutational inactivation of STAG2 causes aneuploidy in human cancer. Science. 333:1039-1043. Spencer, F., S.L.Gerring, C.Connelly, and P.Hieter. 1990. Mitotic chromosome transmission fidelity mutants in Saccharomyces cerevisiae. Genetics. 124:237-249. Stark, C., B.J.Breitkreutz, T.Reguly, L.Boucher, A.Breitkreutz, and M.Tyers. 2006. BioGRID: a general repository for interaction datasets. Nucleic Acids Res. 34:D535- D539. Steckel, M., M.Molina-Arcas, B.Weigelt, M.Marani, P.H.Warne, H.Kuznetsov, G.Kelly, B.Saunders, M.Howell, J.Downward, and D.C.Hancock. 2012. Determination of synthetic lethal interactions in KRAS oncogene-dependent cancer cells reveals novel therapeutic targeting strategies. Cell Res.10. Stirling, P.C., M.S.Bloom, T.Solanki-Patil, S.Smith, P.Sipahimalani, Z.Li, M.Kofoed, S.Ben-Aroya, K.Myung, and P.Hieter. 2011. The complete spectrum of yeast chromosome instability genes identifies candidate CIN cancer genes and functional roles for ASTRA complex components. PLoS. Genet. 7:e1002057. Stirling, P.C., M.J.Crisp, M.A.Basrai, C.M.Tucker, M.J.Dunham, F.A.Spencer, and P.Hieter. 2012. Mutability and mutational spectrum of chromosome transmission fidelity genes. Chromosoma. 121:263-275. Stratton, M.R. 2011. Exploring the genomes of cancer cells: progress and promise. Science. 331:1553-1558. Strom, C.E., F.Johansson, M.Uhlen, C.A.Szigyarto, K.Erixon, and T.Helleday. 2011. Poly (ADP-ribose) polymerase (PARP) is not involved in base excision repair but PARP inhibition traps a single-strand intermediate. Nucleic Acids Res. 39:3166-3175. Strunnikov, A.V., V.L.Larionov, and D.Koshland. 1993. SMC1: an essential yeast gene encoding a putative head-rod-tail protein is required for nuclear division and defines a new ubiquitous protein family. J. Cell Biol. 123:1635-1648. Takawa, M., H.S.Cho, S.Hayami, G.Toyokawa, M.Kogure, Y.Yamane, Y.Iwai, K.Maejima, K.Ueda, A.Masuda, N.Dohmae, H.I.Field, T.Tsunoda, T.Kobayashi, T.Akasu, M.Sugiyama, S.I.Ohnuma, Y.Atomi, B.A.Ponder, Y.Nakamura, and R.Hamamoto. 2012. Histone Lysine Methyltransferase SETD8 Promotes Carcinogenesis by Deregulating PCNA Expression. Cancer Res. 72:3217-3227. Tanaka, H., Y.Katou, M.Yagura, K.Saitoh, T.Itoh, H.Araki, M.Bando, and K.Shirahige. 2009a. Ctf4 coordinates the progression of helicase and DNA polymerase alpha. Genes Cells. 14:807-820. Tanaka, H., Y.Kubota, T.Tsujimura, M.Kumano, H.Masai, and H.Takisawa. 2009b. Replisome progression complex links DNA replication to sister chromatid cohesion in Xenopus egg extracts. Genes Cells. 14:949-963.  110 Tarailo, M., S.Tarailo, and A.M.Rose. 2007. Synthetic lethal interactions identify phenotypic "interologs" of the spindle assembly checkpoint components. Genetics. 177:2525-2530. Tischler, J., B.Lehner, and A.G.Fraser. 2008. Evolutionary plasticity of genetic interaction networks. Nat. Genet. 40:390-391. Tong, A.H., M.Evangelista, A.B.Parsons, H.Xu, G.D.Bader, N.Page, M.Robinson, S.Raghibizadeh, C.W.Hogue, H.Bussey, B.Andrews, M.Tyers, and C.Boone. 2001. Systematic genetic analysis with ordered arrays of yeast deletion mutants. Science. 294:2364-2368. Tong, A.H., G.Lesage, G.D.Bader, H.Ding, H.Xu, X.Xin, J.Young, G.F.Berriz, R.L.Brost, M.Chang, Y.Chen, X.Cheng, G.Chua, H.Friesen, D.S.Goldberg, J.Haynes, C.Humphries, G.He, S.Hussein, L.Ke, N.Krogan, Z.Li, J.N.Levinson, H.Lu, P.Menard, C.Munyana, A.B.Parsons, O.Ryan, R.Tonikian, T.Roberts, A.M.Sdicu, J.Shapiro, B.Sheikh, B.Suter, S.L.Wong, L.V.Zhang, H.Zhu, C.G.Burd, S.Munro, C.Sander, J.Rine, J.Greenblatt, M.Peter, A.Bretscher, G.Bell, F.P.Roth, G.W.Brown, B.Andrews, H.Bussey, and C.Boone. 2004. Global mapping of the yeast genetic interaction network. Science. 303:808-813. Torres, E.M., T.Sokolsky, C.M.Tucker, L.Y.Chan, M.Boselli, M.J.Dunham, and A.Amon. 2007. Effects of aneuploidy on cellular physiology and cell division in haploid yeast. Science. 317:916-924. Trahey, M. and F.McCormick. 1987. A cytoplasmic protein stimulates normal N-ras p21 GTPase, but does not affect oncogenic mutants. Science. 238:542-545. Tsutakawa, S.E., S.Classen, B.R.Chapados, A.S.Arvai, L.D.Finger, G.Guenther, C.G.Tomlinson, P.Thompson, A.H.Sarker, B.Shen, P.K.Cooper, J.A.Grasby, and J.A.Tainer. 2011. Human Flap Endonuclease Structures, DNA Double-Base Flipping, and a Unified Understanding of the FEN1 Superfamily. Cell. 145:198-211. Tumey, L.N., D.Bom, B.Huck, E.Gleason, J.Wang, D.Silver, K.Brunden, S.Boozer, S.Rundlett, B.Sherf, S.Murphy, T.Dent, C.Leventhal, A.Bailey, J.Harrington, and Y.L.Bennani. 2005. The identification and optimization of a N-hydroxy urea series of flap endonuclease 1 inhibitors. Bioorg. Med. Chem. Lett. 15:277-281. Tumey, L.N., B.Huck, E.Gleason, J.Wang, D.Silver, K.Brunden, S.Boozer, S.Rundlett, B.Sherf, S.Murphy, A.Bailey, T.Dent, C.Leventhal, J.Harrington, and Y.L.Bennani. 2004. The identification and optimization of 2,4-diketobutyric acids as flap endonuclease 1 inhibitors. Bioorg. Med. Chem. Lett. 14:4915-4918. Vance, J.R. and T.E.Wilson. 2002. Yeast Tdp1 and Rad1-Rad10 function as redundant pathways for repairing Top1 replicative damage. Proc. Natl. Acad. Sci. U. S. A. 99:13669-13674. Vogel, U.S., R.A.Dixon, M.D.Schaber, R.E.Diehl, M.S.Marshall, E.M.Scolnick, I.S.Sigal, and J.B.Gibbs. 1988. Cloning of bovine GAP and its interaction with oncogenic ras p21. Nature. 335:90-93. Vogelstein, B. and K.W.Kinzler. 2004. Cancer genes and the pathways they control. Nat Med. 10:789-799. Vogelstein, B., E.R.Fearon, S.R.Hamilton, S.E.Kern, A.C.Preisinger, M.Leppert, Y.Nakamura, R.White, A.M.Smits, and J.L.Bos. 1988. Genetic alterations during colorectal-tumor development. N. Engl. J. Med. 319:525-532.  111 Wahlberg, E., T.Karlberg, E.Kouznetsova, N.Markova, A.Macchiarulo, A.G.Thorsell, E.Pol, A.Frostell, T.Ekblad, D.Oncu, B.Kull, G.M.Robertson, R.Pellicciari, H.Schuler, and J.Weigelt. 2012. Family-wide chemical profiling and structural analysis of PARP and tankyrase inhibitors. Nat. Biotechnol. %19;30:283-288. Wang, J., R.Wu, Y.Lu, and C.Liang. 2010. Ctf4p facilitates Mcm10p to promote DNA replication in budding yeast. Biochem. Biophys. Res. Commun. Wang, Z., J.M.Cummins, D.Shen, D.P.Cahill, P.V.Jallepalli, T.L.Wang, D.W.Parsons, G.Traverso, M.Awad, N.Silliman, J.Ptak, S.Szabo, J.K.Willson, S.D.Markowitz, M.L.Goldberg, R.Karess, K.W.Kinzler, B.Vogelstein, V.E.Velculescu, and C.Lengauer. 2004. Three classes of genes mutated in colorectal cancers with chromosomal instability. Cancer Res. 64:2998-3001. Wang, Z., H.Inuzuka, J.Zhong, L.Wan, H.Fukushima, F.H.Sarkar, and W.Wei. 2012. Tumor suppressor functions of FBW7 in cancer development and progression. FEBS Lett. 586:1409-1418. Warren, C.D., D.M.Eckley, M.S.Lee, J.S.Hanna, A.Hughes, B.Peyser, C.Jie, R.Irizarry, and F.A.Spencer. 2004. S-phase checkpoint genes safeguard high-fidelity sister chromatid cohesion. Mol. Biol. Cell. 15:1724-1735. Watrin, E., A.Schleiffer, K.Tanaka, F.Eisenhaber, K.Nasmyth, and J.M.Peters. 2006. Human Scc4 is required for cohesin binding to chromatin, sister-chromatid cohesion, and mitotic progression. Curr. Biol. 16:863-874. Weaver, B.A. and D.W.Cleveland. 2006. Does aneuploidy cause cancer? Curr. Opin. Cell Biol. 18:658-667. Williams, D.R. and J.R.McIntosh. 2002. mcl1+, the Schizosaccharomyces pombe homologue of CTF4, is important for chromosome replication, cohesion, and segregation. Eukaryot. Cell. 1:758-773. Winzeler, E.A., D.D.Shoemaker, A.Astromoff, H.Liang, K.Anderson, B.Andre, R.Bangham, R.Benito, J.D.Boeke, H.Bussey, A.M.Chu, C.Connelly, K.Davis, F.Dietrich, S.W.Dow, B.M.El, F.Foury, S.H.Friend, E.Gentalen, G.Giaever, J.H.Hegemann, T.Jones, M.Laub, H.Liao, N.Liebundguth, D.J.Lockhart, A.Lucau- Danila, M.Lussier, N.M'Rabet, P.Menard, M.Mittmann, C.Pai, C.Rebischung, J.L.Revuelta, L.Riles, C.J.Roberts, P.Ross-Macdonald, B.Scherens, M.Snyder, S.Sookhai-Mahadeo, R.K.Storms, S.Veronneau, M.Voet, G.Volckaert, T.R.Ward, R.Wysocki, G.S.Yen, K.Yu, K.Zimmermann, P.Philippsen, M.Johnston, and R.W.Davis. 1999. Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science. 285:901-906. Wittmeyer, J. and T.Formosa. 1997. The Saccharomyces cerevisiae DNA polymerase alpha catalytic subunit interacts with Cdc68/Spt16 and with Pob3, a protein similar to an HMG1-like protein. Mol. Cell. Biol. 17:4178-4190. Wooster, R., S.L.Neuhausen, J.Mangion, Y.Quirk, D.Ford, N.Collins, K.Nguyen, S.Seal, T.Tran, D.Averill, and . 1994. Localization of a breast cancer susceptibility gene, BRCA2, to chromosome 13q12-13. Science. 265:2088-2090. Wu, Z., Y.Lin, H.Xu, H.Dai, M.Zhou, S.Tsao, L.Zheng, and B.Shen. 2012. High risk of benzo[alpha]pyrene-induced lung cancer in E160D FEN1 mutant mice. Mutat. Res. 731:85-91. Xiong, Y., G.J.Hannon, H.Zhang, D.Casso, R.Kobayashi, and D.Beach. 1993. p21 is a universal inhibitor of cyclin kinases. Nature. 366:701-704.  112 Yang, M., H.Guo, C.Wu, Y.He, D.Yu, L.Zhou, F.Wang, J.Xu, W.Tan, G.Wang, B.Shen, J.Yuan, T.Wu, and D.Lin. 2009. Functional FEN1 polymorphisms are associated with DNA damage levels and lung cancer risk. Hum. Mutat. 30:1320-1328. Yuen, K.W., C.D.Warren, O.Chen, T.Kwok, P.Hieter, and F.A.Spencer. 2007. Systematic genome instability screens in yeast and their potential relevance to cancer. Proc. Natl. Acad. Sci. U. S. A. 104:3925-3930. Zheng, L., H.Dai, M.L.Hegde, M.Zhou, Z.Guo, X.Wu, J.Wu, L.Su, X.Zhong, S.Mitra, Q.Huang, K.H.Kernstine, G.P.Pfeifer, and B.Shen. 2011a. Fen1 mutations that specifically disrupt its interaction with PCNA cause aneuploidy-associated cancer. Cell Res. Zheng, L., H.Dai, M.Zhou, M.Li, P.Singh, J.Qiu, W.Tsark, Q.Huang, K.Kernstine, X.Zhang, D.Lin, and B.Shen. 2007. Fen1 mutations result in autoimmunity, chronic inflammation and cancers. Nat. Med. 13:812-819. Zheng, L., J.Jia, L.D.Finger, Z.Guo, C.Zer, and B.Shen. 2011b. Functional regulation of FEN1 nuclease and its link to cancer. Nucleic Acids Res. 39:781-794. Zheng, L., M.Zhou, Q.Chai, J.Parrish, D.Xue, S.M.Patrick, J.J.Turchi, S.M.Yannone, D.Chen, and B.Shen. 2005. Novel function of the flap endonuclease 1 complex in processing stalled DNA replication forks. EMBO Rep. 6:83-89. Zhu, W., C.Ukomadu, S.Jha, T.Senga, S.K.Dhar, J.A.Wohlschlegel, L.K.Nutt, S.Kornbluth, and A.Dutta. 2007. Mcm10 and And-1/CTF4 recruit DNA polymerase {alpha} to chromatin for initiation of DNA replication. Genes Dev. 21:2289.   113  Appendices Appendix A  siRNA pool silencing in HCT116 cells. Horizontal lines indicate experiments carried out on different days. siRNA NA Mean ± SEMB Normalizedpercent (%)C Expected percent (%)D Difference (%)E siGAPDH + siGAPDH 32 3173.7 ± 69 100 NA NA siGAPDH + siWDHD1 16 2952.9 ± 150.6 93 NA NA siGAPDH + siFEN1 16 2536.3 ± 89 79.9 NA NA siGAPDH + siSMC1A 16 2950.1 ± 102.5 93 NA NA siGAPDH + siSMC3 16 3034.1 ± 96.4 95.6 NA NA siGAPDH + siMRE11A 16 2901.7 ± 99.8 91.4 NA NA siGAPDH + siCDC4 16 2873.1 ± 128.4 90.5 NA NA siGAPDH + siBLM 16 2941.3 ± 67.7 92.7 NA NA siGAPDH + siNIPBL 16 2214.1 ± 98.5 69.8 NA NA siGAPDH + siSTAG1 16 2320.4 ± 133.8 73.1 NA NA siWDHD1 + siSMC1A 8 1930.1 ± 103.9 60.8 86.5 29.7 siWDHD1 + siSMC3 8 2131 ± 101.6 67.1 89 24.5 siWDHD1 + siMRE11A 8 2195 ± 120.8 69.2 85.1 18.7 siWDHD1 + siCDC4 8 1971.6 ± 60.2 62.1 84.2 26.2 siFEN1 + siSMC1A 8 1663.8 ± 101.3 52.4 74.3 29.4 siFEN1 + siSMC3 8 1491.4 ± 117.6 47 76.4 38.5 siFEN1 + siMRE11A 8 1288.8 ± 82.6 40.6 73.1 44.4 siFEN1 + siCDC4 8 1255.5 ± 95.5 39.6 72.3 45.3 siWDHD1 + siBLM 8 1484.9 ± 81.9 46.8 86.2 45.7 siWDHD1 + siNIPBL 8 1144.6 ± 52.6 36.1 64.9 44.4 siWDHD1 + siSTAG1 8 1024.3 ± 83.8 32.3 68 52.6 siFEN1 + siBLM 8 1501.8 ± 133 47.3 74.1 36.1 siFEN1 + siNIPBL 8 1271.5 ± 108.7 40.1 55.8 28.1 siFEN1 + siSTAG1 8 1335.3 ± 62.3 42.1 58.4 28 siGAPDH + siGAPDH 32 3302.9 ± 85.4 100 NA NA siGAPDH + siWDHD1 16 3146.9 ± 135.2 95.3 NA NA siGAPDH + siFEN1 16 2601.2 ± 79.6 78.8 NA NA siGAPDH + siSMC1A 16 3213.8 ± 111.6 97.3 NA NA siGAPDH + siSMC3 16 4098 ± 142.6 124.1 NA NA siGAPDH + siMRE11A 16 3850.6 ± 167.4 116.6 NA NA siGAPDH + siCDC4 16 3315.3 ± 127.4 100.4 NA NA siGAPDH + siBLM 16 3191.4 ± 105.1 96.6 NA NA siGAPDH + siNIPBL 16 2538.4 ± 94.7 76.9 NA NA siGAPDH + siSTAG1 16 2702.6 ± 82.8 81.8 NA NA siWDHD1 + siSMC1A 8 2599.1 ± 172.9 78.7 92.7 15.1 siWDHD1 + siSMC3 8 3054.4 ± 157 92.5 118.2 21.8 siWDHD1 + siMRE11A 8 2823 ± 179.4 85.5 111.1 23.1 siWDHD1 + siCDC4 8 2137 ± 135.1 64.7 95.6 32.3 siFEN1 + siSMC1A 8 1852.1 ± 83.4 56.1 76.6 26.8  114 siRNA NA Mean ± SEMB Normalizedpercent (%)C Expected percent (%)D Difference (%)E siFEN1 + siSMC3 8 2356.1 ± 108.4 71.3 97.7 27 siFEN1 + siMRE11A 8 2302.6 ± 113.5 69.7 91.8 24.1 siFEN1 + siCDC4 8 1512.9 ± 95.9 45.8 79 42.1 siWDHD1 + siBLM 8 1891.6 ± 83 57.3 92.1 37.8 siWDHD1 + siNIPBL 8 1494.8 ± 57.1 45.3 73.2 38.2 siWDHD1 + siSTAG1 8 1764.5 ± 73.6 53.4 78 31.5 siFEN1 + siBLM 8 2175.3 ± 141.7 65.9 76.1 13.5 siFEN1 + siNIPBL 8 1939.4 ± 75.6 58.7 60.5 3 siFEN1 + siSTAG1 8 1483.6 ± 69.7 44.9 64.4 30.3 siGAPDH + siGAPDH 16 2496.9 ± 80.3 100 NA NA siGAPDH + siWDHD1 8 1488.6 ± 73.1 59.6 NA NA siGAPDH + siFEN1 8 1959.5 ± 90.2 78.5 NA NA siGAPDH + siSTAG3 16 1954.3 ± 64.5 78.3 NA NA siGAPDH + siRAD54B 16 1918.6 ± 95.7 76.8 NA NA siGAPDH + siRNF20 16 1533.6 ± 45.9 61.4 NA NA siWDHD1 + siSTAG3 8 1062.4 ± 80.3 42.5 46.7 8.8 siWDHD1 + siRAD54B 8 1449.9 ± 144.1 58.1 45.8 -26.8 siWDHD1 + siRNF20 8 663.4 ± 49.6 26.6 36.6 27.4 siFEN1 + siSTAG3 8 1695.3 ± 62.8 67.9 61.4 -10.5 siFEN1 + siRAD54B 8 1225.6 ± 107.1 49.1 60.3 18.6 siFEN1 + siRNF20 8 494.4 ± 37.4 19.8 48.2 58.9 siGAPDH + siGAPDH 16 2978.7 ± 110.1 100 NA NA siGAPDH + siWDHD1 8 2542.1 ± 144.2 85.3 NA NA siGAPDH + siFEN1 8 1770.8 ± 103.4 59.4 NA NA siGAPDH + siSTAG3 16 3310.6 ± 102.6 111.1 NA NA siGAPDH + siRAD54B 16 3442.3 ± 103.8 115.6 NA NA siGAPDH + siRNF20 16 2096.3 ± 87.5 70.4 NA NA siWDHD1 + siSTAG3 8 2170.1 ± 52.7 72.9 94.9 23.2 siWDHD1 + siRAD54B 8 1881.6 ± 140.4 63.2 98.6 36 siWDHD1 + siRNF20 8 1753.8 ± 93.6 58.9 60.1 2 siFEN1 + siSTAG3 8 2164.5 ± 97.5 72.7 66.1 -10 siFEN1 + siRAD54B 8 1684.4 ± 104.1 56.5 68.7 17.7 siFEN1 + siRNF20 8 847 ± 54.6 28.4 41.8 32 siGAPDH + siGAPDH 24 3474 ± 149.3 100 NA NA siGAPDH + siCHTF8 24 2941.2 ± 126.5 84.7 NA NA siGAPDH + siSMC1A 8 2941.4 ± 184.3 84.7 NA NA siGAPDH + siSMC3 8 2947 ± 140.8 84.8 NA NA siGAPDH + siMRE11A 8 2354.4 ± 103 67.8 NA NA siGAPDH + siCDC4 8 1990.1 ± 92.6 57.3 NA NA siGAPDH + siBLM 8 3473 ± 224.8 100 NA NA siGAPDH + siNIPBL 8 2460.7 ± 48.2 70.8 NA NA siGAPDH + siSTAG1 8 2355.8 ± 68.5 67.8 NA NA siGAPDH + siSTAG3 8 2284.1 ± 302.7 65.7 NA NA siGAPDH + siRAD54B 8 3565.7 ± 463.9 102.6 NA NA siGAPDH + siRNF20 8 1662 ± 219 47.8 NA NA  115 siRNA NA Mean ± SEMB Normalizedpercent (%)C Expected percent (%)D Difference (%)E siCHTF8 + siSMC1A 8 2282.1 ± 69.3 65.7 71.7 8.4 siCHTF8 + siSMC3 8 2692.8 ± 81.5 77.5 71.8 -7.9 siCHTF8 + siMRE11A 8 2869.1 ± 167 82.6 57.4 -43.9 siCHTF8 + siCDC4 8 2115.3 ± 85.4 60.9 48.5 -25.5 siCHTF8 + siBLM 8 1926.1 ± 71.5 55.4 84.6 34.5 siCHTF8 + siNIPBL 8 1301.9 ± 88.4 37.5 60 37.5 siCHTF8 + siSTAG1 8 1428.5 ± 68.4 41.1 57.4 28.4 siCHTF8 + siSTAG3 8 2045.5 ± 271.3 58.9 55.7 -5.8 siCHTF8 + siRAD54B 8 2556 ± 330.6 73.6 86.9 15.3 siCHTF8 + siRNF20 8 1759.9 ± 236.6 50.7 40.5 -25.1 siGAPDH + siGAPDH 24 2160.6 ± 104.6 100 NA NA siGAPDH + siCHTF8 24 2120.1 ± 93 98.1 NA NA siGAPDH + siSMC1A 8 2347 ± 113.8 108.6 NA NA siGAPDH + siSMC3 8 2031.6 ± 84.3 94 NA NA siGAPDH + siMRE11A 8 2451.8 ± 88.3 113.5 NA NA siGAPDH + siCDC4 8 1620.8 ± 82.3 75 NA NA siGAPDH + siBLM 8 2566.2 ± 118.5 118.8 NA NA siGAPDH + siNIPBL 8 1952.8 ± 68.4 90.4 NA NA siGAPDH + siSTAG1 8 2059.3 ± 51.2 95.3 NA NA siGAPDH + siSTAG3 8 2665.9 ± 357.5 123.4 NA NA siGAPDH + siRAD54B 8 2421.9 ± 321.8 112.1 NA NA siGAPDH + siRNF20 8 2440.1 ± 306.5 112.9 NA NA siCHTF8 + siSMC1A 8 1984.3 ± 81.5 91.8 106.6 13.8 siCHTF8 + siSMC3 8 1867.9 ± 56.1 86.5 92.3 6.3 siCHTF8 + siMRE11A 8 2348.3 ± 74 108.7 111.4 2.4 siCHTF8 + siCDC4 8 1380.9 ± 72.1 63.9 73.6 13.2 siCHTF8 + siBLM 8 1051.3 ± 74.4 48.7 116.6 58.3 siCHTF8 + siNIPBL 8 1174.9 ± 39.2 54.4 88.7 38.7 siCHTF8 + siSTAG1 8 1889.5 ± 40.1 87.5 93.5 6.5 siCHTF8 + siSTAG3 8 1145.2 ± 146.1 53 121.1 56.2 siCHTF8 + siRAD54B 8 1578.7 ± 207.8 73.1 110 33.6 siCHTF8 + siRNF20 8 991 ± 128.5 45.9 110.8 58.6 AN; number of wells imaged BSEM; standard error about the mean CAll values are normalized relative to siGAPDH-silenced controls and shown ± SEM DCalculated by multiplying the normalized relative percentages for the two individual siRNAs ECalculated as: 1 - (Normalized Relative Percent/Expected Percent) × 100. (NA; not applicable)  116 Appendix B  Random siRNA in HCT116 cells. Horizontal lines indicate experiments carried out on different days. siRNA NA Mean ± SEMB Normalized percent (%)C Expected percent (%)D Difference (%) E siGAPDH + siGAPDH 16 3742.1 ± 146 100 NA NA siGAPDH + siITSN1 16 2930.9 ± 133 78.32 NA NA siGAPDH + siWDHD1 8 2316.6 ± 65 61.91 NA NA siGAPDH + siFEN1 8 1941.3 ± 72 51.88 NA NA siWDHD1 + siITSN1 8 2710.4 ± 166 72.43 48.49 -49.38 siFEN1 + siITSN1 8 2075.5 ± 87 55.46 40.63 -36.5 siGAPDH + siGAPDH 32 2541.2 ± 72 100 NA NA siGAPDH + siWDHD1 32 2139.1 ± 53 84.18 NA NA siGAPDH + siFEN1 32 2000.4 ± 42 78.72 NA NA siGAPDH + siITSN1 8 2495.1 ± 196 98.19 NA NA siGAPDH + siARF1 8 1087.9 ± 81 42.81 NA NA siGAPDH + siCAV1 8 2081.1 ± 125 81.9 NA NA siGAPDH + siDOT1L 8 1848.6 ± 112 72.75 NA NA siWDHD1 + siITSN1 8 2317.3 ± 91 91.19 82.65 -10.33 siFEN1 + siITSN1 8 1809.9 ± 57 71.22 77.29 7.86 siWDHD1 + siARF1 8 1046.4 ± 50 41.18 36.04 -14.27 siFEN1 + siARF1 8 1001.9 ± 90 39.43 33.7 -16.99 siWDHD1 + siCAV1 8 2274.8 ± 117 89.52 68.94 -29.85 siFEN1 + siCAV1 8 2473.9 ± 111 97.35 64.47 -51 siWDHD1 + siDOT1L 8 1739.9 ± 78 68.47 61.23 -11.81 siFEN1 + siDOT1L 8 1928.8 ± 47 75.9 57.27 -32.54 siGAPDH + siGAPDH 16 2948.1 ± 122 100 NA NA siGAPDH + siWDHD1 16 2168.4 ± 94 73.55 NA NA siGAPDH + siFEN1 16 2017.5 ± 65 68.43 NA NA siGAPDH + siCAV1 8 2605.3 ± 73 88.37 NA NA siGAPDH + siDOT1L 8 2340.6 ± 143 79.39 NA NA siWDHD1 + siCAV1 8 2764.6 ± 121 93.78 65 -44.27 siFEN1 + siCAV1 8 2009.8 ± 59 68.17 60.47 -12.73 siWDHD1 + siDOT1L 8 1955.8 ± 73 66.34 58.4 -13.6 siFEN1 + siDOT1L 8 2097.9 ± 95 71.16 54.33 -30.97 siGAPDH + siGAPDH 8 2674.3 ± 95 100 NA NA siGAPDH + siWDHD1 8 2263.9 ± 154 84.65 NA NA siGAPDH + siFEN1 8 2058.6 ± 94 76.98 NA NA siGAPDH + siARF1 8 1237.8 ± 86 46.28 NA NA siWDHD1 + siARF1 8 1100.6 ± 105 41.16 39.18 -5.04 siFEN1 + siARF1 8 1684.9 ± 97 63 35.63 -76.83 AN; number of wells imaged BSEM; standard error about the mean CAll values are normalized relative to siGAPDH-silenced controls and shown ± SEM DCalculated by multiplying the normalized relative percentages for the two individual siRNAs ECalculated as: 1 - (Normalized Relative Percent/Expected Percent) × 100. (NA; not applicable)  117 Appendix C  Individual siRNAs in HCT116 cells. Horizontal lines indicate experiments carried out on different days. siRNA NA Mean ± SEMB Normalized percent (%)C Expected percent (%)D Difference (%)E siGAPDH + siGAPDH 8 1805.3 ± 87 100 NA NA siGAPDH + siFEN1-2 8 1456.3 ± 74 80.67 NA NA siGAPDH + siMRE11A-4 8 1851.9 ± 87 102.58 NA NA siGAPDH + siCDC4-2 8 1699.5 ± 94 94.14 NA NA siFEN1-2 + siMRE11A-4 8 1086.6 ± 49 60.19 82.75 27.26 siFEN1-2 + siCDC4-2 8 574.6 ± 30 31.83 75.94 58.09 siGAPDH + siGAPDH 8 1319.8 ± 69 100 NA NA siGAPDH + siFEN1-2 8 808.5 ± 25 61.26 NA NA siGAPDH + siMRE11A-4 8 1314.1 ± 44 99.57 NA NA siGAPDH + siCDC4-2 8 1041.9 ± 63 78.94 NA NA siFEN1-2 + siMRE11A-4 8 669.4 ± 30 50.72 60.99 16.85 siFEN1-2 + siCDC4-2 8 459.3 ± 20 34.8 48.36 28.03 AN; number of wells imaged BSEM; standard error about the mean CAll values are normalized relative to siGAPDH-silenced controls and shown ± SEM DCalculated by multiplying the normalized relative percentages for the two individual siRNAs ECalculated as: 1 - (Normalized Relative Percent/Expected Percent) × 100. (NA; not applicable) 118   Appendix D  List of strains used in chapter 5. YDV number Back-ground Relevant genotype MAT 340 Y2HA pOAD pBDC a/α 342 Y2H CTF4-pOBD2 a/α 636 Y2H CTF4-pBDC/N.DEST a/α 638 Y2H ctf4-25-pBDC/N.DEST a/α 640 Y2H ctf4-41-pBDC/N.DEST a/α 642 Y2H ctf4-43-pBDC/N.DEST a/α 644 Y2H ctf4-46-pBDC/N.DEST a/α 646 Y2H ctf4-50-pBDC/N.DEST a/α 648 Y2H ctf4-65-pBDC/N.DEST a/α 650 Y2H ctf4-66-pBDC/N.DEST a/α 652 Y2H ctf4-107-pBDC/N.DEST a/α 654 Y2H ctf4-154-pBDC/N.DEST a/α 702 Y2H SLD5-pOAD pBDC a/α 678 Y2H SLD5-pOAD CTF4-pOBD2 a/α 680 Y2H SLD5-pOAD CTF4-pBDC/N.DEST a/α 682 Y2H SLD5-pOAD ctf4-25-pBDC/N.DEST a/α 684 Y2H SLD5-pOAD ctf4-41-pBDC/N.DEST a/α 686 Y2H SLD5-pOAD ctf4-43-pBDC/N.DEST a/α 688 Y2H SLD5-pOAD ctf4-46-pBDC/N.DEST a/α 690 Y2H SLD5-pOAD ctf4-50-pBDC/N.DEST a/α 692 Y2H SLD5-pOAD ctf4-65-pBDC/N.DEST a/α 694 Y2H SLD5-pOAD ctf4-66-pBDC/N.DEST a/α 696 Y2H SLD5-pOAD ctf4-107-pBDC/N.DEST a/α  119 YDV number Back-ground Relevant genotype MAT 698 Y2H SLD5-pOAD ctf4-154-pBDC/N.DEST a/α 1624 Y2H MMS22-pOAD a/α YBM319 Y2H BIR1-pOAD SLI15-pOBD2 a/α 1564 Y2H MMS22-pOAD CTF4-pOBD2 a/α 1566 Y2H MMS22-pOAD CTF4-pBDC/N.DEST a/α 1580 Y2H MMS22-pOAD ctf4-66-pBDC/N.DEST a/α 1578 Y2H MMS22-pOAD ctf4-65-pBDC/N.DEST a/α 1576 Y2H MMS22-pOAD ctf4-50-pBDC/N.DEST a/α 1568 Y2H MMS22-pOAD ctf4-25-pBDC/N.DEST a/α 1572 Y2H MMS22-pOAD ctf4-43-pBDC/N.DEST a/α 1570 Y2H MMS22-pOAD ctf4-41-pBDC/N.DEST a/α 1584 Y2H MMS22-pOAD ctf4-154-pBDC/N.DEST a/α 1574 Y2H MMS22-pOAD ctf4-46-pBDC/N.DEST a/α 1582 Y2H MMS22-pOAD ctf4-107-pBDC/N.DEST a/α 499 YPHB (Wild type) a 275 YPH ctf4-25 a 277 YPH ctf4-41 a 279 YPH ctf4-43 a 281 YPH ctf4-46 a 283 YPH ctf4-50 a 285 YPH ctf4-65 a 287 YPH ctf4-66 a 289 YPH ctf4-107 a 291 YPH ctf4-154 a 885 CohesionC ctf4-25::URA3 bar1Δ::HphMX a 608 Cohesion ctf4-41::URA3 bar1Δ::HphMX a 887 Cohesion ctf4-43::URA3 bar1Δ::HphMX a 625 Cohesion ctf4-46::URA3 bar1Δ::HphMX a  120 YDV number Back-ground Relevant genotype MAT 611 Cohesion ctf4-50::URA3 bar1Δ::HphMX a 612 Cohesion ctf4-65::URA3 bar1Δ::HphMX a 615 Cohesion ctf4-66::URA3 bar1Δ::HphMX a 619 Cohesion ctf4-107::URA3 bar1Δ::HphMX a 621 Cohesion ctf4-154::URA3 bar1Δ::HphMX a 604 Cohesion ctf4Δ::KanMX bar1Δ::HphMX a 603 Cohesion CTF4 bar1Δ::HphMX a 724 YPH ctf4-25::URA3/ctf4-25::URA3 bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 728 YPH CTF4/CTF4 bar1Δ::HphMX/BAR1 a/α 726 YPH ctf4-66::URA3/ctf4-66::URA3 bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 714 YPH ctf4-154::URA3/ctf4-154::URA3 bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 708 YPH ctf4-50::URA3/ctf4-50::URA3 bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 706 YPH ctf4-46::URA3/ctf4-46::URA3 bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 720 YPH ctf4Δ::KanMX/ctf4Δ:KanMX bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 734 YPH ctf4-43::URA3/ctf4-43::URA3 bar1Δ::HphMX/BAR1 a/α  121 YDV number Back-ground Relevant genotype MAT CFVII(RAD2.d.877)::TRP1,SUP11 710 YPH ctf4-65::URA3/ctf4-65::URA3 bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 712 YPH ctf4-107::URA3/ctf4-107::URA3 bar1Δ::HphMX/BAR1 CFVII(RAD2.d.877)::TRP1,SUP11 a/α 94 BYD ctf4Δ::HphMX a 1160 BY CTF4::URA3 can1Δ::MFA1pr-HIS3 a 1164 BY ctf4-S463A::URA3 can1Δ::MFA1pr-HIS3 a 1227 BY ctf4-S377A,S379A::URA3 can1Δ::MFA1pr-HIS3 a 1172 BY ctf4-S377A,S379A,S463A::URA3 can1Δ::MFA1pr-HIS3 a 1176 BY ctf4-S463D::URA3 can1Δ::MFA1pr-HIS3 a 1179 BY ctf4-S377D,S379D::URA3 can1Δ::MFA1pr-HIS3 a 1183 BY ctf4-S377D,S379D,S463D::URA3 can1Δ::MFA1pr-HIS3 a 1208 SGAE pARC25B α 1210 SGA WDHD1-pARC25B α 1214 SGA FEN1-pARC25B α 1420 Cohesion CTF4::URA3 bar1Δ::HphMX a 1422 Cohesion ctf4-S463A::URA3 bar1Δ::HphMX a 1431 Cohesion ctf4-S463D bar1Δ::HphMX a 1426 Cohesion ctf4-S377A,S379A::URA3 bar1Δ::HphMX a 1427 Cohesion ctf4-S377A,S379A::URA3 bar1Δ::HphMX a 1433 Cohesion ctf4-S377D,S379D::URA3 bar1Δ::HphMX a 1434 Cohesion ctf4-S377D,S379D::URA3 bar1Δ::HphMX a 1430 Cohesion ctf4-S377A,S379A,S463A::URA3 bar1Δ::HphMX a 1435 Cohesion ctf4-S377D,S379D,S463D::URA3 bar1Δ::HphMX a  122 YDV number Back-ground Relevant genotype MAT 1436 Cohesion ctf4-S377D,S379D,S463D::URA3 bar1Δ::HphMX a Y7092 SGA  α 1153 BY WDHD1-pARC25B a/α 1156 BY FEN1-pARC25B a/α 1652-1773F BY (mre11Δ::NatMX or bub1Δ::NatMX or sgs1Δ::NatMX) [(Hieter lab ts::URA3) or (Charlie Boone ts::KanMX)] a/α 888 SGA ctf4Δ::HphMX α 2534 SGA rad27Δ::HphMX α Atrp1-901 leu2-3,112 ura3-52 his3-200 gal4Δ gal80Δ LYS2::GAL1-HIS3 GAL2-ADE2 met2::GAL7-lacZ Bhis3Δ200 lys2-801 ade2-101 leu2Δ1 ura3-52 trp1Δ63 Cura3 (lys2 or LYS2) (ade2 or ADE2) (can1-100 or CAN1) (trp1 or TRP1) leu2-3,112::lacO+256(pFA559)-LEU2 his3-11,15::GFP- LacI-I12::HIS3(pFA5144) Dhis3Δ1 leu2Δ0 ura3Δ0 (lys2Δ0 or LYS2) (met15Δ0 or MET15) E his3Δ1 leu2Δ0 ura3Δ0 LYS2 lyp1Δ can1Δ::STE2pr-HIS3 FSee Appendix B. 123  Appendix E Colony counts from random spore analysis of genetic interactions between Ctf4 physical interactors and mre11Δ, bub1Δ, and sgs1Δ. KO, knockout. Alleles in bold are from the Hieter lab temperature-sensitive collection (Ben-Aroya et al., 2008); all others are from (Li et al., 2011). YDV# Temperature (oC) KO gene ts gene Non- selective G418 -URA or +NAT G418 + (-URA or NAT) Conclusions 1703 25 mre11 mcm3-1 332 164 384 99 INTERACTION 1707 25 mre11 pri2-1 380 328 276 118 INTERACTION 1710 25 mre11 mcm10-1 452 276 332 2 INTERACTION 1733 25 bub1 mcm3-1 856 512 664 124 INTERACTION 1736 25 bub1 psf1-1 196 160 112 21 INTERACTION 1743 25 bub1 pri1-M4 232 264 220 140 INTERACTION 1767 25 sgs1 pri2-1 504 380 416 173 INTERACTION 1770 25 sgs1 mcm10-1 684 584 524 140 INTERACTION 1770 25 sgs1 mcm10-1 114 126 100 7 INTERACTION 1767 25 sgs1 pri2-1 264 204 228 19 INTERACTION 1674 25 bub1 psf3-2 188 100 108 32 INTERACTION 1672 25 bub1 mcm6-1 252 128 256 67 INTERACTION 1698 30 mre11 pol1-ts 384 312 324 29 INTERACTION 1706 30 mre11 psf1-1 564 260 488 54 INTERACTION 1713 30 mre11 pri1-M4 552 456 568 14 INTERACTION 1728 30 bub1 pol1-ts 304 308 256 168 INTERACTION 1734 30 bub1 cdc46-1[mcm5] 552 600 488 112 INTERACTION 1738 30 bub1 cdc45-27 620 372 388 17 INTERACTION 1773 30 sgs1 pri1-M4 95 46 79 2 INTERACTION 1758 30 sgs1 pol1-ts 288 196 196 9 INTERACTION 1773 30 sgs1 pri1-M4 288 142 320 3 INTERACTION 1670 30 bub1 mcm4-1 172 87 128 1 INTERACTION 1652 30 mre11 sld5-2 208 156 45 1 Death is mostly due to ts 1660 30 mre11 mcm6-1 156 196 3 1 Death is due to ts  124 YDV# Temperature (oC) KO gene ts gene Non- selective G418 -URA or +NAT G418 + (-URA or NAT) Conclusions 1664 30 bub1 sld5-2 192 80 19 1 Death is mostly due to ts 1676 30 sgs1 sld5-2 208 184 45 21 Death is due to ts 1684 30 sgs1 mcm6-1 260 212 5 6 Death is due to ts 1672 30 bub1 mcm6-1 82 60 17 1 INTERACTION 1666 30 bub1 sld5-3 126 88 0 0 Death due to ts 1654 30 mre11 sld5-3 312 336 0 1 Death due to ts 1678 30 sgs1 sld5-3 320 332 1 1 Death due to ts 1705 30 mre11 cdc47-1[mcm7] 416 0 440 0 Death due to mre11 1735 30 bub1 cdc47-1[mcm7] 304 10 268 3 Death due to bub1 1763 30 sgs1 mcm3-1 132 2 156 2 Death due to sgs1 1765 30 sgs1 cdc47-1[mcm7] 122 0 128 0 Death due to sgs1 1764 30 sgs1 cdc46-1[mcm5] 288 160 240 113 Death due to sgs1 1766 30 sgs1 psf1-1 460 83 452 5 Doubles smaller 1708 32 mre11 cdc45-27 184 128 140 2 INTERACTION 1702 32 mre11 mcm2-1 376 308 304 160 INTERACTION 1732 32 bub1 mcm2-1 300 184 128 17 INTERACTION 1737 32 bub1 pri2-1 248 208 164 13 INTERACTION 1740 32 bub1 mcm10-1 232 156 75 13 INTERACTION 1704 32 mre11 cdc46-1[mcm5] 66 16 68 5 Death due to mre11; doubles smaller 1704 32 mre11 cdc46-1[mcm5] 284 11 284 1 Death due to mre11; doubles smaller 1768 32 sgs1 cdc45-27 272 11 308 6 Death due to ts; doubles smaller 1744 34 bub1 pol12-ts 29 76 0 3 Death due to bub1 1702 34 mre11 mcm2-1 324 0 372 1 Death due to ts 1689 34 sgs1 psf3-2 332 416 1 1 Death due to ts 1682 34 sgs1 mcm4-1 440 428 3 4 Death due to ts  125 YDV# Temperature (oC) KO gene ts gene Non- selective G418 -URA or +NAT G418 + (-URA or NAT) Conclusions 1658 34 mre11 mcm4-1 324 372 0 0 Death due to ts 1662 34 mre11 psf3-2 404 384 0 0 Death due to ts  126 Appendix F Results of SGA screen with ctf4Δ. Strains in bold are DAmP alleles (Breslow et al., 2008); all others are temperature- sensitive alleles (Li et al., 2011). COSMIC info ORF name Gene name E-C Human ortholog COSMICmutated? Cancer census? Mutations Missense Nonsense Indels Amplifications YDR167W taf10 -1.53 TAF10 n n YPR133C spn1 -1.50 HRC y n 1 1 0 0 1 YBR087W rfc5 -1.48 RFC3 n n YLR274W mcm5 -1.46 MCM5 y n 4 3 0 1 0 YER012W pre1 -1.30 PSMB2 n n YBL034C stu1 -1.28 none n n YDL097C rpn6 -1.27 COPS2 y n 1 0 0 1 1 YER148W spt15 -1.23 TBPL1 y n 1 1 0 0 2 YDR013W psf1 -1.18 GINS1 y n 2 1 0 0 1 YNL006W lst8 -1.17 MLST8 n n YLR045C stu2 -1.12 CKAP5 y n 3 1 0 0 0 YDR356W spc110 -1.08 CCDC18 y n 6 5 1 0 1 YLR430W sen1 -1.06 SETX y n 5 5 0 0 0 YGL093W spc105 -1.05 none n n YNR043W mvd1 -1.04 MVD y n 1 0 1 0 0 YJR006W pol31 -1.04 POLD2 y n 2 2 0 0 2 YLR103C cdc45 -1.03 CDC45 y n 2 2 0 0 2 YDR113C pds1 -1.03 none n n YGL093W spc105 -1.02 none n n YPR110C rpc40 -1.00 POLR1C n n YJL173C rfa3 -0.99 none n n YBL034C stu1 -0.99 none n n YOL094C rfc4 -0.96 RFC2 y n 2 1 1 0 1 YBL034C stu1 -0.95 none n n YDL126C cdc48 -0.93 SPATA5L1 y n 5 2 1 2 1 YBR088C POL30 -0.92 PCNA y n 1 1 0 0 1 YBR160W cdc28 -0.92 CDK4 y y 3 2 0 0 11 YDR168W cdc37 -0.90 CDC37 y n 2 2 0 0 0 YPR103W pre2 -0.89 PSMB11 y n  127 COSMIC info ORF name Gene name E-C Human ortholog COSMICmutated? Cancer census? Mutations Missense Nonsense Indels Amplifications YPL020C ulp1 -0.87 SENP5 y n 4 2 0 0 3 YPR108W rpn7 -0.87 GPS1 y n 2 2 0 0 1 YHR166C cdc23 -0.84 CDC23 y n 1 1 0 0 0 YOR249C apc5 -0.84 none n n YGR218W crm1 -0.83 XPO1 y y 8 6 0 0 1 YFR004W rpn11 -0.82 EIF3H y n 3 1 1 0 15 YFL034C- B mob2 -0.79 MOB1A y n 0 YLL004W orc3 -0.78 none n n YOR218C  -0.77 YFR027W eco1 -0.75 ESCO1 y n 1 1 0 0 0 YDL102W pol3 -0.75 POLD1 y n 5 3 0 0 1 YJR017C ess1 -0.74 PIN1 y n 3 2 1 0 0 YFR028C cdc14 -0.74 DUSP23 y n 1 1 0 0 2 YKL172W ebp2 -0.73 EBNA1BP2 n n YGR113W dam1 -0.72 none n n YMR028W tap42 -0.72 IGBP1 y n 2 2 0 0 0 YDL147W rpn5 -0.72 COPS4 y n 1 1 0 0 0 YHR164C dna2 -0.71 DNA2 y n 3 3 0 0 1 YDL003W mcd1 -0.69 RAD21L1 y n 1 0 0 0 1 YLR212C tub4 -0.69 TUBD1 y n 1 1 0 0 6 YNL006W lst8 -0.68 MLST8 n n YBR060C orc2 -0.68 ORC2 y n 5 4 0 0 1 YBR087W RFC5 -0.67 RFC3 n n YOR341W rpa190 -0.66 POLR1A y n 11 7 0 0 1 YBR060C orc2 -0.66 ORC2 y n 5 4 0 0 1 YDL102W pol3 -0.66 POLD1 y n 5 3 0 0 1 YER133W glc7 -0.65 PPP1CC y n 3 3 0 0 0 YKL045W pri2 -0.63 PRIM2 y n 2 2 0 0 4 YJR057W cdc8 -0.63 DTYMK n n YDR510W SMT3 -0.62 SUMO4 n n YFL037W TUB2 -0.61 TUBE1 y n 2 1 0 0 0  128 COSMIC info ORF name Gene name E-C Human ortholog COSMICmutated? Cancer census? Mutations Missense Nonsense Indels Amplifications YJR068W rfc2 -0.60 RFC4 y n 5 2 0 3 2 YHR191C ctf8 -0.59 CHTF8 n n YLR229C CDC42 -0.59 RHOJ y n 1 1 0 0 2 YOR259C rpt4 -0.58 PSMC6 y n 1 1 0 0 2 YJL074C smc3 -0.56 SMC3 y n 4 4 1 0 0 YNL216W rap1 -0.56 none n n YMR314W PRE5 -0.56 PSMA1 y n 2 1 0 0 0 YFL039C act1 -0.55 ACTBL2 y n 3 2 0 0 YIR008C PRI1 -0.55 PRIM1 y n YEL055C pol5 -0.54 MYBBP1A y n 5 3 0 0 0 YOL094C RFC4 -0.54 RFC2 y n 2 1 1 0 1 YOR174W med4 -0.53 none n n YER012W PRE1 -0.52 PSMB2 n n YJR076C cdc11 -0.52 SEPT5 n n 2 0 0 0 3 YLR078C BOS1 -0.49 GOSR2 y n 2 1 0 0 1 YPL124W spc29 -0.49 none n n YJL074C smc3 -0.48 SMC3 y n 4 4 1 0 0 YLR317W  -0.48 YNR003C rpc34 -0.47 POLR3F n n YDR021W fal1 -0.47 EIF4A3 y n 2 2 0 0 1 YDL017W cdc7 -0.47 CDC7 y n 6 3 0 3 1 YBL084C cdc27 -0.44 CDC27 y n 10 4 0 1 1 YGL112C taf6 -0.44 TAF6L y n 3 3 0 0 0 YNL272C SEC2 -0.43 none n n YDR168W cdc37 -0.43 CDC37 y n 2 2 0 0 0 YDR062W lcb2 -0.43 SPTLC3 n n YDL008W apc11 -0.42 ANAPC11 n n YOR117W RPT5 -0.41 ATAD3C y n 2 2 0 0 1 YGL011C SCL1 -0.41 PSMA6 n n YBR155W cns1 -0.41 TTC4 n n YDL028C mps1 -0.40 TLK1 y n 5 4 0 0 0  129 COSMIC info ORF name Gene name E-C Human ortholog COSMICmutated? Cancer census? Mutations Missense Nonsense Indels Amplifications YGR009C sec9 -0.40 SNAP23 n n YER157W cog3 -0.39 COG3 y n 2 1 0 0 2 YBR156C sli15 -0.39 none n n YBR202W mcm7 -0.38 MCM7 n n YNL114C  -0.38 YER133W GLC7 -0.37 PPP1CC y n 3 3 0 0 0 YMR197C vti1 -0.37 VTI1A y y 1 1 0 0 0 YBR109C cmd1 -0.35 CALM1 y n 1 0 0 0 2 YLR298C yhc1 -0.35 SNRPC y n 1 0 0 0 1 YFL038C YPT1 -0.33 RAB40A n n YHL015W RPS20 -0.33 RPS20 n n YLR249W yef3 -0.33 ABCF3 y n 5 2 0 0 2 YNR053C nog2 -0.33 GNL2 y n 3 2 0 0 3 YBR198C taf5 -0.32 RFWD2 y n 1 1 0 0 4 YBL092W RPL32 -0.32 RPL32 y n 1 0 1 0 1 YDR052C dbf4 -0.31 DBF4 y n 4 1 0 0 8 YNL263C YIF1 -0.31 YIF1A y n 1 1 0 0 2 YJR068W RFC2 -0.30 RFC4 y n 5 2 0 2 2   130 Appendix G Results of SGA screen with rad27Δ. Strains in bold are DAmP alleles (Breslow et al., 2008); all others are temperature- sensitive alleles (Li et al., 2011). COSMIC info ORF name Gene name E-C Human ortholog COSMIC mutated? Cancer census? Mutations Missense Nonsense Indels Amplifications YER148W spt15 -1.49 TBPL1 y n 1 1 0 0 2 YJR006W pol31 -0.79 POLD2 y n 2 2 0 0 2 YLL004W orc3 -0.69 none n n YOL094C RFC4 -0.63 RFC2 y n 2 1 0 0 1 YFL008W smc1 -0.62 SMC1B y n 6 6 0 0 0 YHL015W RPS20 -0.60 RPS20 n n YDR180W scc2 -0.57 NIPBL y n 18 9 1 1 11 YDL105W nse4 -0.56 EID3 y n 1 1 0 0 YAR007C rfa1 -0.56 RPA1 y n 2 0 0 2 1 YOR218C  -0.55 YPR110C rpc40 -0.53 POLR1C n n YER133W glc7 -0.52 PPP1CC y n 3 1 0 0 0 YNL216W rap1 -0.49 none n n YPR144C NOC4 -0.49 NOC4L y n 1 1 0 0 0 YPR176C BET2 -0.49 PGGT1B n n YBR060C orc2 -0.47 ORC2 y n 5 4 0 0 1 YHR166C cdc23 -0.45 CDC23 y n 1 1 0 0 YBR060C orc2 -0.45 ORC2 y n 5 4 0 0 1 YBR087W RFC5 -0.43 RFC3 n n YNL260C  -0.41 YBR088C POL30 -0.41 PCNA y n 1 1 0 0 0 YLR317W  -0.40 YJL074C smc3 -0.39 SMC3 y n 5 4 1 0 0 YDR510W SMT3 -0.39 SUMO4 n n YJL074C smc3 -0.38 SMC3 y n 5 4 1 0 0 YLR321C SFH1 -0.37 SMARCB1 y y 410    3 YBL092W RPL32 -0.34 RPL32 y n 1 0 1 0 1 YBR202W mcm7 -0.33 MCM7 n n YLR459W gab1 -0.32 PIGU y n 1 1 0 0 5  131 COSMIC info ORF name Gene name E-C Human ortholog COSMIC mutated? Cancer census? Mutations Missense Nonsense Indels Amplifications YLR229C CDC42 -0.32 RHOJ y n 1 1 0 0 2 YNL251C NRD1 -0.32 SCAF4 y n 2 0 2 0 YGR179C okp1 -0.32 none n n YOL139C cdc33 -0.31 EIF4E3 n n YDL165W cdc36 -0.30 CNOT2 n n   132 Appendix H Results of SDL screen with rad27Δ. “Negative interactor” denotes whether or not the gene has previously been reported to have a negative genetic interaction with rad27Δ (Tong et al., 2001; Pan et al., 2004; Tong et al., 2004; Costanzo et al., 2010). COSMIC info ORF name Gene name E-C Negative interactor? Human ortholog COSMIC mutated? Cancer census? Mutations Missense Nonsense Indels Amplifications YBR073W RDH54 -0.56 YES RAD54B y n 6 3 0 1 3 YML061C PIF1 -0.56 YES PIF1 n n YBR223C TDP1 -0.49 NO TDP1 y n 2 0 0 2 0 YBL019W APN2 -0.46 NO APEX1 y n 4 1 2 0 0 YIR002C MPH1 -0.46 NO FANCM y n 14 7 2 0 1 YOL006C TOP1 -0.41 YES TOP1MT y n 3 0 0 1 1 YHR164C DNA2 -0.40 YES DNA2 y n 3 3 0 0 1 YHR115C DMA1 -0.34 NO none n n YPL032C SVL3 -0.26 NO none n n YLR453C RIF2 -0.24 NO none n n YGL175C SAE2 -0.24 YES none n n YCR052W RSC6 -0.24 NO SMARCD2 n n YMR001C CDC5 -0.24 NO PLK4 y n 1 1 0 0 0 YNL042W BOP3 -0.24 NO none n n  a YLR032W RAD5 -0.24 YES TTF2 y n 4 4 0 0 2 YLR347C KAP95 -0.23 NO KPNB1 y n 3 1 0 0 2 YNL076W MKS1 -0.23 NO none n n YDR310C SUM1 -0.22 YES none n n YGR186W TFG1 -0.22 NO none n n YDR335W MSN5 -0.22 NO XPO5 y n 5 3 0 0 1 YGR140W CBF2 -0.21 NO none n n YGL162W SUT1 -0.21 NO none n n YBR049C REB1 -0.21 NO none n n YOR291W YPK9 -0.20 NO ATP13A5 y n 12 7 1 0 2  YPL203W TPK2 0.23 NO PRKG2 y n 8 4 0 0 0 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0073341/manifest

Comment

Related Items