A Novel Gene Overexpression Plasmid Library and Its Application in Mapping Genetic Networks by Systematic Dosage Suppression

Home , FCF1, Library (biology)

A Novel Gene Overexpression Plasmid Library and its application in Mapping Genetic Networks by Systematic Dosage Suppression

Leslie Joyce Magtanong

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Molecular Genetics University of Toronto

A Novel Gene Overexpression Plasmid Library and its application in Mapping Genetic Networks by Systematic Dosage Suppression

Leslie Joyce Magtanong

Doctor of Philosophy

Department of Molecular Genetics University of Toronto

2011

Abstract

Increasing gene dosage provides a powerful means of probing gene function, as it tends to cause a gain-of-function effect due to increased gene activity. In the budding yeast, Saccharomyces cerevisiae, systematic gene overexpression studies have shown that in wild-type cells, overexpression of a small subset of genes results in an overt phenotype. However, examining the effects of gene overexpression in sensitized cells containing mutations in known genes is a powerful means for identifying functionally relevant genetic interactions. When a query mutant phenotype is rescued by gene overexpression, the genetic interaction is termed dosage suppression. I comprehensively investigated dosage suppression genetic interactions in yeast using three approaches. First, using one of two novel plasmid libraries cloned by two colleagues and myself, I systematically performed dosage suppression screens and identified over 130 novel dosage suppression genetic interactions for more than 25 essential yeast genes. The plasmid libraries, called the molecular barcoded yeast ORF (MoBY-ORF) 1.0 and 2.0, are designed to streamline dosage analysis by being compatible with high-throughput genomics technologies that can monitor plasmid representation, including barcode microarrays and next-generation sequencing methods. Second, I describe a detailed analysis of the novel dosage suppression interactions, as well as of literature-curated interactions, and show that the gene pairs exhibiting dosage suppression are often functionally related and can overlap with physical as well as negative genetic interactions. Third, I performed a systematic categorization of dosage suppression genetic

interactions in yeast and show that the majority of the dosage suppression interactions can be assigned to one of four general mechanistic classifications. With this comprehensive analysis, I conclude that systematically identifying dosage suppression genetic interactions will allow for their integration into other genetic and physical interaction networks and should provide new insight into the global wiring diagram of the cell.

iii

Acknowledgments

I have many people to thank for all of their support and encouragement throughout the years. First and foremost, I would like to thank my parents, Vic and Anita, my sisters, Lisa and Jill, and my husband, Scott Dixon, for their unconditional support during my time in Toronto. I also want to thank the professors, postdocs, and students who have provided valuable feedback and suggestions for my various experiments, presentations, and manuscripts. In particular, I acknowledge and am grateful to my supervisory committee, Drs. Brenda Andrews, Barbara Funnell, and Howard Lipshitz, who have been incredibly supportive of my research and abilities as a doctoral student. I thank the fellow graduate students who have contributed to my research. In particular, I thank Cheuk Hei Ho, who spearheaded the development of the MoBY-ORF plasmid libraries and gave me many helpful suggestions throughout my research; a postdoc, Sarah Barker, and the various technicians and summer students, all of whom were integral to the development of the MoBY-ORF plasmid libraries; Wei Jiao and Anastasia Baryshnikova, who did invaluable computational work for this project; and Sondra Bahr, a talented technician who made a significant contribution to the dosage suppression studies. Finally, I would like to thank my supervisor, Dr. Charlie Boone, whose intelligence and support for me will always be remembered.

This work would not have been possible without assistance provided by members of the scientific community. In particular, I thank Andrew Smith, and Drs. Larry Heisler, Marinella Gibella, and Corey Nislow, who provided access to and assistance with their microarray facilities. I also thank the Natural Sciences and Engineering Research Council (NSERC), the Canadian Institutes of Health Research (CIHR), and the University of Toronto for financial support.

Table of Contents Page Abstract ii Acknowledgments iv List of Tables ix List of Figures x List of Appendices xi List of Electronic Tables xii

Chapter One: Introduction 1 1.1 General Introduction 2 1.2 Genetic Interactions 2 1.2.1 Negative Genetic Interactions 4 1.2.1.1 Complex Haploinsufficiency 4 1.2.2 Positive Genetic Interactions 5 1.2.3 Synthetic Dosage Effects: Lethality and Suppression 7 1.3 Investigating Genetic Interactions in S. cerevisiae in a Systematic 7 Manner 1.3.1 Development of Genome-wide Strain Collections 8 1.3.1.1 Loss-of-Function Strains: The Deletion Strain Collection 8 1.3.1.1.1 Barcoded strains and barcode microarrays 8 1.3.1.2 Essential Gene Strain Collections 13 1.3.1.2.1 tetO promoter collection 13 1.3.1.2.2 URA3-marked temperature-sensitive allele collection 15 1.3.1.2.3 DAmP allele collection 15 1.3.2 Gene overexpression 16 1.3.2.1 The Yeast Two-Hybrid S. cerevisiae ORF Array 17

1.3.2.2 The PCUP1-GST Library 21

1.3.2.3 The PGAL1/10-GST Library 21 1.3.2.4 The Movable ORF Library 22 1.3.2.5 The FLEXGene ORF Collection 23 1.3.2.6 The Yeast Genome Tiling Collection 23

1.3.2.7 Summary and Comparison of Various Existing Overexpression 24 Libraries 1.4 Systematic Identification of Genetic Interactions in Yeast 25 1.4.1 Synthetic Genetic Array (SGA) Analysis 25 1.4.1.1 Application of SGA to SDL analysis 27 1.4.1.2 Application of SGA to genetic mapping 27 1.4.1.3 Application of SGA to array-based high-content screening 27 1.4.2 Diploid-based Synthetic Lethal Analysis on Microarrays (dSLAM) 29 1.4.3 Genetic Interaction Mapping (GIM) 30 1.4.4 Summary of Genetic Interaction Mapping Strategies 30 1.5 Next-Generation Sequencing 31 1.6 Summary and Rationale 32

Chapter Two: The MoBY-ORF 1.0 Yeast Plasmid Library 34 2.1 Introduction 35 2.2 Results 35 2.2.1 Construction of a library of molecular barcoded yeast ORFs 35 2.2.2 Verification of constructed clones by sequencing 36 2.2.3 Assessment of clone function using temperature-sensitive mutants 39 2.2.4 Complementation cloning to identify drug-resistant mutants and 39 compound mode-of-action 2.3 Summary 39 2.4 Methods 40 2.4.1 Yeast Strains 40 2.4.2 Growth Media 40 2.4.3 Clone Construction and Analysis 40 2.4.4 Sequence Confirmation of the MoBY-ORF Collection Barcodes and 42 3’ ORF Junctions 2.4.5 Functional Complementation of Essential Genes 42

Chapter Three: Mapping Genetic Networks by Systematic Dosage 43 Suppression 3.1 Introduction 44 3.2 Results 46 3.2.1 Construction of the MoBY-ORF 2.0 plasmid library 46 3.2.2. Dosage suppression analysis of temperature-sensitive conditional 46 mutants Methods used to identify dosage suppressors 46 Description of results 53 3.2.3 An integrated dosage suppression genetic interaction network 57 Network overview 58 Identification of a genetic link between PKA signaling and the 58 kinetochore 3.2.4 Distribution of dosage suppressors across cellular processes 61 3.2.5. Overlap of dosage suppression interactions with protein-protein and 64 negative genetic network edges 3.2.6 Mechanistic categorization of dosage suppression interactions 64 Dosage suppression decision tree for categorizing dosage 64 suppression interactions Description of categories 64 3.3 Discussion 73 3.4 Methods 76 3.4.1 Growth media 76 3.4.2 Clone construction and analysis 76 3.4.3 Plasmid pool preparation 77 3.4.4 Cloning of dosage suppressors with the 2µ MoBY-ORF library 77 3.4.5 Yeast barcode microarray hybridization and data analysis 80 3.4.6 Empirical determination of raw barcode microarray intensity cutoff 80 for identification of candidate dosage suppressors 3.4.7 Assessing fitness of barcoded yeast strains by Illumina/Solexa 81 sequencing

vii

3.4.8 Confirmation of candidate dosage suppressors and test for reciprocal 82 suppression 3.4.9 Overlap of dosage suppression genetic interactions with other types of 83 interactions 3.4.10 Analysis of functional relatedness 83 3.4.11 Identifying gene clusters in the integrated dosage suppression network 83

Chapter Four: Conclusions and Future Directions 84 4.1 General Overview 85 4.2 The MoBY-ORF gene overexpression libraries: present and future 85 applications 4.3 Dosage suppression genetic interaction networks: illuminating a new 88 facet of genetics 4.4 Understanding the mechanistic basis of dosage suppression 90 4.5 Concluding thoughts 92

Chapter Five: References 94

Chapter Six: Appendices 112

viii

List of Tables Page

1.1 Overview of existing essential gene strain collections for S. cerevisiae 14 1.2 Overview of gene overexpression plasmid libraries for S. cerevisiae 18 3.2.1 Overlap of dosage suppression interactions with other types of 65 interactions 3.2.2 Distribution of dosage suppression gene pairs annotated in the 66 Saccharomyces Genome Database 3.2.3 Gene pairs tested for reciprocal suppression 71 3.2.4 Yeast strains used in this study 78

List of Figures Page

1.1 The barcoded kanMX cassette of the S. cerevisiae deletion collection 9 1.2 Barcode microarray method in yeast 11 1.3 Plasmid libraries in S. cerevisiae 19 2.2.1 Plasmid map of p5472 37 2.2.2 Construction of the MoBY-ORF library by homologous 38 recombination in yeast 3.2.1 Schematic of a plasmid in the MoBY-ORF 2.0 plasmid library 47 3.2.2 Plasmid map of p5476 48 3.2.3 MAGIC with the MoBY-ORF 1.0 plasmid library 49 3.2.4 Using the MoBY-ORF 2.0 library to identify candidate dosage 51 suppressors by barcode microarray 3.2.5 Empirical determination of raw barcode microarray intensity cutoff 54 for identification of candidate dosage suppressors 3.2.6 Dosage suppression genetic interaction network for S. cerevisiae 59 3.2.7 Properties of the yeast dosage suppression network 62 3.2.8 Decision tree used to categorize dosage suppression interactions 67 3.2.9 Mechanisms of dosage suppression in yeast 69

List of Appendices Page

6.1 All confirmed spot dilutions performed based on dosage suppression 113 screens reported in this study 6.2 Unique dosage suppression interactions identified in this study 120

List of Electronic Tables (found on the DVD accompanying this thesis)

3.2.1 Dosage suppression interactions identified in this study (.xls file) 3.2.2 Dosage suppression interactions annotated in the Saccharomyces Genome Database (.xls file)

xii

Chapter One Introduction to Genetic Interactions

1 2

1.1 GENERAL INTRODUCTION

In the post-genome sequence era, a major goal in biology is to understand the function of every gene encoded in an organism’s genome. One method to elucidate a gene’s function is to generate a loss-of-function allele, either through targeted genomic replacement/deletion (Shoemaker, Lashkari et al. 1996) or to generate the functional equivalent by depleting the target gene messenger RNA (mRNA) using RNA interference (RNAi)(Fire et al., 1998), and see if this results in a phenotype that differs from wild type. In the budding yeast Saccharomyces cerevisiae, genome-wide systematic replacement of each predicted open reading frame (ORF) with a G418 drug resistance cassette, kanMX, showed that ~19% of all genes are essential under standard laboratory conditions (Giaever et al., 2002); a similar percentage (17.5%) of essential genes was observed in a pilot systematic deletion study using the unrelated fission yeast Schizosaccharomyces pombe genome (Decottignies, Sanchez-Perez et al. 2003). In the nematode Caenorhabditis elegans, systematic RNAi experiments targeting ~86% (16,757/19,427) of all coding genes demonstrated that reduced function of ~1,000 genes (6%) resulted in a phenotype consistent with an essential function (e.g. embryonic lethal, larval lethal, larval arrest; (Kamath, Fraser et al. 2003)). Compared to yeast, the lower percentage of essential genes observed in C. elegans may be due to either a greater redundancy inherent in a multicellular organism or technical limitations associated with the RNAi technique or phenotypic analysis. Genome-wide RNAi libraries have also been developed for other sequenced multicellular organisms, including Drosophila melanogaster (Boutros, Kiger et al. 2004; Dietzl, Chen et al. 2007), Arabidopsis thaliana (Schwab, Ossowski et al. 2006), Danio rerio (Pickart, Klee et al. 2006), and cultured mouse and human cells (Kittler, Putz et al. 2004; Paddison, Silva et al. 2004; Moffat, Grueneberg et al. 2006). What is clear from all of these efforts is that reducing or eliminating the function of most genes has no overt phenotypic effect; therefore, understanding gene function on a genomic scale requires moving beyond loss of function paradigms. As discussed below in greater detail, one approach which has been pursued with great vigor for almost a decade now is the analysis of gene deletions or gene knockdowns in the context of other genetic alterations, to investigate genetic interactions systematically.

1.2 GENETIC INTERACTIONS

Epistasis is a term describing the interaction of genes that are not allelic to one another. According to the Merriam-Webster dictionary, the word epistasis is derived from the Greek words epi plus histanai, which mean “to cause to stand”. Two related but distinct definitions of epistasis are both useful in explaining genetic interactions. William Bateson’s definition of epistasis (Bateson 1909), which Patrick Phillips refers to as “compositional epistasis” (Phillips 2008), is not a quantitative measurement, but rather is directly related to the Greek derivation of the word. When two single mutants (gene a and gene b) are combined, and mutant phenotype A masks, or “stops”, mutant phenotype B, from manifesting, then gene a is defined as being epistatic to gene b; the gene being masked is defined as hypostatic (from the Greek words hypo plus histanai, which mean “to stand under”). R.A. Fisher’s definition of epistasis (Fisher 1918), which Phillips refers to as “statistical epistasis” (Phillips 2008), is based on population genetics studies, and involves the quantitative measurement of a genetic interaction. Each single mutant is given a fitness measurement relative to wild type; therefore, if no genetic interaction occurs, then the fitness of the double mutant is the expected additive effect of the two fitness values (Phillips 2008). Any deviation from this model, then, is considered a genetic interaction. Note that Bateson’s strict definition of epistasis is actually one type of genetic interaction (masking or suppression) and, thus, is encompassed by the broader Fisher definition.

A genetic interaction, therefore, occurs when an unexpected phenotype manifests from the combination of at least two mutations. The unexpected, or mutant, phenotype implies that a functional relationship of some kind exists between the gene products. The identification and analysis of genetic interactions can therefore help unravel genetic and biochemical pathways and networks and illuminate the underlying structure of biological systems (Tong, Evangelista et al. 2001; Tong, Lesage et al. 2004; Lehner, Crombie et al. 2006; Byrne, Weirauch et al. 2007; Dixon, Fedyshyn et al. 2008; Roguev, Bandyopadhyay et al. 2008; Costanzo, Baryshnikova et al. 2010) To date, however, the majority of studies have relied upon loss-of-function mutants. Genetic interactions involving gain-of-function mutations or gene overexpression are likely to be equally informative, but this direction has been pursued much less vigorously.

As discussed further in the following sections, the nature of the functional relationship uncovered by a genetic interaction depends on what types of mutants are used and what type of genetic interaction is observed. A genetic interaction can reflect a phenotype that is either worse (negative) or better (positive) than expected. Several mathematical definitions of what the expected value for a genetic interaction should be have been reported in the literature, including the Product (or multiplicative), Additive, Log, and Min models (Mani, St Onge et al. 2008). In S. cerevisiae, the Min model was initially used to define genetic interactions qualitatively (Tong, Evangelista et al. 2001), but more recently, the Product model has gained favor as more quantitative phenotypic measurements and data analysis models have been developed (Mani, St Onge et al. 2008; Costanzo, Baryshnikova et al. 2010). Therefore, in what follows below, I will introduce negative and positive genetic interactions in the context of the Product model. Furthermore, unless otherwise noted, the phenotype under consideration is typically strain fitness, as defined by colony size on solid agar or growth rate in liquid culture.

1.2.1 Negative Genetic Interactions

Negative genetic interactions, which are also referred to as enhancer or aggravating interactions, occur when a double mutant is less fit than the product of its cognate single mutants. The most extreme example of a negative genetic interaction is synthetic lethality, where the combination of two single loss of function (LOF) mutants, which alone are viable, results in an inviable double mutant phenotype (Dobzhansky 1946; Guarente 1993). When the query allele is a null allele of a non-essential gene, a synthetic lethal interaction suggests a between-pathway interaction, whereby the two pathways normally are able to “buffer”, or rescue, one another in the event that one pathway is compromised (Tong, Evangelista et al. 2001; Tong, Lesage et al. 2004). When the query allele is a conditional allele of an essential gene, synthetic lethality can also result from a within-pathway interaction (Tong, Evangelista et al. 2001; Mnaimneh, Davierwala et al. 2004; Tong, Lesage et al. 2004).

1.2.1.1 Complex Haploinsufficiency

Complex haploinsufficiency (CHI) is a diploid-specific negative genetic interaction. CHI interactions are more commonly known as dominant enhancers or second-site noncomplementation in Drosophila (Reuter and Wolff 1981; Ashburner 1982; Cutforth and Rubin 1994) and unlinked or non-allelic non-complementation in worms and yeast (Bisson and Thorner 1982; Kusch and Edgar 1986; Stearns and Botstein 1988; Yook, Proulx et al. 2001). In a binary CHI interaction, two recessive mutations in different genes fail to complement one another as heterozygous diploid double mutants. CHI interactions have been shown to occur between genes encoding proteins that physically interact within a protein complex (Bisson and Thorner 1982; Rine and Herskowitz 1987; Stearns and Botstein 1988; Vinh, Welch et al. 1993; Baetz, Krogan et al. 2004; Haarer, Viggiano et al. 2007). As an example, in a screen designed to identify new alleles of the yeast β-tubulin gene, TUB2, the first conditional allele of the yeast α-tubulin gene, TUB1, was isolated (Stearns and Botstein 1988); α- and β-tubulin heterodimerize to form tubulin, which can then polymerize to form microtubules (Luduena, Shooter et al. 1977). Two models of CHI have been described in the literature (Yook 2005). In both models, single mutations of two genes encoding proteins that function in the same pathway do not cause a mutant phenotype. In the “Poison” model, when the two mutations are present in the same cell, the mutant proteins physically interact and act as poisons that interfere with the complex’s normal function (Hays, Deuring et al. 1989). In the “Dosage” model, when the two mutations are present in the same cell, the overall reduction in pathway activity leads to a mutant phenotype (Kidd, Bland et al. 1999). A systematic genome-wide analysis of CHI interactions was reported in yeast (Haarer, Viggiano et al. 2007). This screen searched for deletion alleles that resulted in a general enhancement of the growth defect associated with a diploid strain deleted for one of two copies of the ACT1 gene, which encodes yeast actin. The authors identified over 200 genes that had a CHI interaction with ACT1. Approximately 15% of the genes were previously either poorly characterized or had no known function. Interestingly, the authors showed that some functional specialization related to actin appears to be associated with ribosome function because several individual genes of ribosomal paralog pairs had CHI interactions with actin (Haarer, Viggiano et al. 2007).

1.2.2 Positive Genetic Interactions

A positive genetic interaction occurs when the combination of two mutations results in a phenotype that is fitter than the product of its cognate single mutations. Also referred to as alleviating interactions, positive genetic interactions can be divided into different categories based on the resulting double mutant phenotype. Several classes of positive interactions have been described (Drees, Thorsson et al. 2005; St Onge, Mani et al. 2007); however, since this is a rapidly evolving field where the terminology remains unsettled, only definitions of symmetric and asymmetric positive interactions are presented below.

In a symmetric positive interaction, the two single LOF mutant phenotypes and the cognate double mutant phenotype are all indistinguishable. This type of interaction most often occurs because the two genes encode proteins found in either the same complex or biological pathway; therefore, a pathway or complex is compromised by removal of one component but is not further compromised by removal of another component (Drees, Thorsson et al. 2005; Collins, Miller et al. 2007; St Onge, Mani et al. 2007). As an example, St. Onge et al. (St Onge, Mani et al. 2007) found that the fitness of the single mutant strains rad55Δ and rad57Δ, and of the double mutant strain rad55Δ rad57Δ were indistinguishable from one another in quantitative growth assays. The two proteins Rad55p and Rad57p form a heterodimer that is involved in Rad51p nucleoprotein filament extension in yeast (Sung 1997), consistent with the fact that disruption of either gene alone will destroy the activity of the entire complex and explaining why the single and double mutants have the same fitness.

Two types of asymmetric positive interactions involving LOF mutations have been described. Masking occurs when the fitness of the double mutant is equal to or greater than the fitness of the least sick single mutant. Genetic suppression occurs when the fitness of the double mutant is equal to or greater than the fitness of the sickest single mutant (Drees, Thorsson et al. 2005; Segre, Deluna et al. 2005; St Onge, Mani et al. 2007; Breslow, Cameron et al. 2008). Several different mechanisms of genetic suppression have been identified in the literature (Prelich 1999; Hodgkin 2005), including (but not limited to) altering the levels of the mutant protein (McCusker, Yamagishi et al. 1991), altering the activity of the mutant protein (Sandrock, O'Dell et al. 1997), and altering the activity of the biological pathway in which the mutant protein operates (Stevenson, Rhodes et al. 1992).

1.2.3 Synthetic Dosage Effects: Lethality and Suppression

Whereas a loss-of-function mutation in a gene may not result in a mutant phenotype, overexpression (gain-of-function) of the same gene may provide insight into its function. In yeast, 3-15% of genes are toxic when overexpressed in wild-type cells, with the percentage dependent on the method of overexpression (see Section 2.2 for a detailed explanation of overexpression in yeast) (Sopko, Huang et al. 2006; Jones, Stalker et al. 2008). Gene overexpression can also be done in a (loss-of-function) mutant background in order to identify genetic interactions and, by extension, elucidate gene function. In this case, suppression or enhancement of the mutant phenotype is called dosage suppression or synthetic dosage lethality (SDL), respectively. In a dosage suppression interaction, overexpression improves or rescues the mutant phenotype. Conversely, in an SDL interaction, overexpression exacerbates a sick or induces a lethal phenotype. Dosage suppression studies have been used classically to identify new protein interaction partners of a particular query allele; for example, the G1 cyclins, CLN1 and CLN2, were initially discovered as dosage suppressors of cdc28-1, a temperature-sensitive allele of the essential gene encoding the cyclin-dependent kinase Cdc28p (Reed, Hadwiger et al. 1989). Subsequent studies showed that the cyclins bind to and activate the enzymatic activity of Cdc28p (Richardson, Wittenberg et al. 1989; Tyers and Futcher 1993), suggesting that overexpression of the cyclins may in some way stabilize Cdc28-1p. SDL studies have been useful in identifying function-specific relationships; for example, Kroll et al. (Kroll, Hyland et al. 1996) showed that overexpression of ORC6, which encodes a member of the origin recognition complex, has synthetic dosage lethal interactions specifically with alleles of three other genes involved in replication, but not with genes required for chromosome segregation. Conversely, overexpression of the chromosome segregation gene CTF13 is specifically toxic in chromosome segregation mutants but has no phenotype in replication mutants (Kroll, Hyland et al. 1996). These small-scale studies demonstrate that overexpression is a powerful way to investigate gene function and identify genetic interactions.

1.3 INVESTIGATING GENETIC INTERACTIONS IN S. CEREVISIAE IN A SYSTEMATIC MANNER

Classical genetic and biochemical analysis has focused on the function of one or a small number of genes and/or proteins in any one study. In the last decade, there has been a decided shift towards experimental paradigms that attempt to examine the role of most or all the genes/proteins in parallel. These systematic approaches have largely been pioneered in the budding yeast S. cerevisiae.

1.3.1 Development of Genome-wide Strain Collections

The study of genetic interactions in yeast has been greatly advanced by the availability of several genome-wide collections of genetic “reagents”. Each of these collections has unique characteristics that make it useful in both investigating specific aspects of gene function and genetic interactions.

1.3.1.1 Loss-of-Function Strains: The Deletion Strain Collection

Each strain in the S. cerevisiae deletion collection is precisely deleted for one open reading frame (ORF) from its ATG start codon to its stop codon; specifically, each ORF has been replaced with a dominant drug-resistant cassette, kanMX, which confers resistance to the antibiotic geniticin (G418). Furthermore, each cassette is barcoded with two unique 20 nucleotide tags flanked by common primer sequences (Figure 1.1). This allows for highly parallel analyses of all deletion strains, as the relative abundance of each strain can be measured from a barcode microarray readout (Shoemaker, Lashkari et al. 1996). Gene deletion strains for non-essential genes are available as haploids or homozygous diploids, while for essential genes, strains are available as heterozygous diploids (Giaever, Chu et al. 2002).

1.3.1.1.1 Barcoded strains and barcode microarrays

One innovation that makes the yeast deletion strain collection described above especially useful in genomic studies is the unique set of barcodes carried by each deletion strain (Shoemaker, Lashkari et al. 1996). Flanking the kanMX cassette used to replace each ORF are

Figure 1.1

10 two 20 nucleotide-long sequence tags, referred to as barcodes, which are unique to each ORF. The “Uptag” and “Downtag” barcodes are found at the 5’ and 3’ ends respectively of the kanMX cassette. These unique barcodes are flanked by common primer sequences, thereby allowing for PCR amplification of all Uptags and Downtags present in a heterogeneous population of barcoded strains. The relative abundance of each barcoded strain in the population is measured by hybridizing the PCR products onto a barcode microarray and then detecting the fluorescence intensity of each feature (Figure 1.2)(Shoemaker, Lashkari et al. 1996; Pierce, Davis et al. 2007). Functional profiling of the yeast genome using the deletion strain collections in barcode microarray experiments has provided valuable insight into drug target identification (Giaever, Flaherty et al. 2004) and mechanisms of haploinsufficiency (Deutschbauer, Jaramillo et al. 2005). More recently, >1,100 barcode microarray experiments using both the heterozygous and homozygous deletion strain collections were conducted for 408 unique chemical stress conditions; remarkably, 97% of the deletion strains screened displayed a phenotype in at least one of the conditions tested, helping to resolve the previous conundrum of why an organism would maintain so many seemingly non-essential genes (Hillenmeyer, Fung et al. 2008).

Molecular barcodes are being applied in novel ways. For example, in theory, any strain collection that is barcoded can be subjected to analysis by barcode microarray. To facilitate barcoding of any S. cerevisiae collection, Yan et al. (Yan, Costanzo et al. 2008) developed a set of “Barcoder” donor strains in which the HO locus of each strain has been replaced by a uniquely barcoded kanMX cassette. An S. cerevisiae collection can be barcoded by using synthetic genetic array (SGA) technology and selection methods (described below in Section 3.1). In proof-of-principle experiments, the above group barcoded 1,402 decreased abundance in mRNA production, or DAmP, allele strains (described below in Section 2.1.5). Through barcode microarray experiments, they showed that barcoding did not adversely affect fitness, as the fitness of barcoded DAmP strains was virtually indistinguishable from that of the cognate non- barcoded DAmP strains (Yan, Costanzo et al. 2008). This validates the barcodes as a useful way to molecularly tag any strain of interest. In this thesis, we constructed a genome-wide set of uniquely barcoded yeast overexpression plasmids for use in gene dosage studies, as described in Chapters 2 and 3.

Figure 1.2

Figure 1.2 Barcode microarray method in yeast. a) Each yeast strain is tagged with a unique molecular ‘barcode’ indicated by the different colors. The barcode can either be integrated into the genome or contained on a plasmid; either way, the downstream analysis is identical. The barcoded yeast strains are pooled for subsequent analyses. b) The pool is grown competitively in the desired screening condition (e.g. mutant allele, drug treatment). c) Barcoded DNA is extracted. d) Universal primers are used to PCR-amplify the Uptags and Downtags from the heterogeneous cell population. e) PCR products are hybridized to a barcode microarray. The intensity of the barcode on the array is indicative of the relative abundance of the barcoded strain in the pool.

1.3.1.2 Essential Gene Strain Collections

Deletion strains of essential genes are by definition inviable and thus impossible to propagate as either haploids or homozygous diploids. As a class, however, essential genes are of great interest for a number of reasons. Most obviously, the fact that a gene is required for viability immediately suggests that it is crucial for some vital and likely interesting aspect of cell function. Using cell division as an example, the classical cdc screens in yeast identified 32 essential genes all required for some aspect of this process, including DNA replication, transition between stages of the cell cycle, and cytokinesis (Hartwell, Culloti et al. 1970; Hartwell 1971; Hartwell, Culotti et al. 1974).

Due to the biological importance of essential genes, much effort has been expended to generate genetic reagent sets that allow the problem of lethality to be circumvented and thereby facilitate genomic studies of these genes. In addition to the heterozygous diploid collection, several groups have developed conditional or hypomorphic alleles of yeast essential genes, which can be maintained in a haploid cell. This allows for a more direct comparison to the haploid deletion collection of non-essential genes. It also facilitates looking at genetic interactions involving essential genes, as haploid cells of opposite mating types carrying selectable query alleles can be mated and selected for, in order to ultimately generate double mutant haploids (see below, Sections 1.4.1 and 1.4.3). The development and characterization of several essential gene strain collections is described below and summarized in Table 1.1.

1.3.1.2.1 tetO promoter collection

Mnaimneh et al. (Mnaimneh, Davierwala et al. 2004) produced 602 strains of essential gene conditional alleles that are regulated by the small molecule doxycycline, a tetracycline R analogue. In each strain, a kan -tetO7-TATA cassette was inserted into the promoter just upstream of the ORF; additionally, each strain expresses tandem copies of the tet repressor, which binds tetO7 and prevents transcription of the associated ORF. Therefore, in the absence of doxycycline, this repressor is not active, and transcription of the cognate essential gene proceeds normally. However, once doxycycline is added, the repressor binds tetO7 and represses

Table 1.1 Overview of existing essential gene strain collections for S. cerevisiae.

Collection Number of Nature of hypomorphic allele Reference Name essential ORFs in collection tetO 602 Tetracycline- or doxycycline- (Mnaimneh, promoter repressible promoter directly Davierwala et al. upstream of ORF;  partial 2004) repression of gene expression URA3-ts 250 Temperature-sensitive (Ben-Aroya, Coombes mutation in ORF;  partially et al. 2008) functional protein DAmP 842 Drug-resistant cassette inserted (Breslow, Cameron et between ORF stop codon and al. 2008) 3’ UTR;  destabilized RNA transcript

15 transcription of the essential gene. Using this system, the authors were able to predict and subsequently verify the function of several previously uncharacterized essential genes (Mnaimneh, Davierwala et al. 2004).

1.3.1.2.2 URA3-marked temperature-sensitive allele collection

Temperature-sensitive (ts) strains are a classical reagent for investigating essential gene function in haploid cells. Ben-Aroya et al. (Ben-Aroya, Coombes et al. 2008) developed a method called “diploid shuffling” to systematically generate ts alleles for ~250 essential genes. First, PCR mutagenesis was used to make random point mutations in a particular essential gene. Second, the mutagenized PCR product was cloned into a vector so that it ligates immediately downstream of the 5’ half of kanMX (kan), and upstream of a URA3 marker, followed by the 3’ half of kanMX (MX). Third, the vector is linearized to free the kan-mutagenized allele-URA3-MX fragment, and this fragment is transformed into the corresponding heterozygous diploid strain, and URA3 transformants are selected. Fourth, the transformant is sporulated, and MATa URA3 haploids are selected. Fifth, haploids are replica plated at both the permissive and restrictive temperatures, and any colonies that drop out specifically at the latter represent candidate ts alleles. By screening this collection in several assays, it was possible to identify roles for ten poorly characterized essential ORFs in RNA processing and six uncharacterized essential ORFs in chromosome segregation (Ben-Aroya, Coombes et al. 2008).

1.3.1.2.3 DAmP allele collection

In the decreased abundance by mRNA perturbation (DAmP) approach to creating hypomorphic alleles, a heterologous coding sequence, such as an antibiotic resistance marker, is inserted between the stop codon of the ORF and the 3’ UTR. When a transcript is produced from the DAmP ORF, it is unstable and results in a two- to ten-fold decrease in the amount of mRNA that is produced (Schuldiner, Collins et al. 2005). Taking advantage of this technique, Breslow et al. (Breslow, Cameron et al. 2008) created a collection of 842 DAmP allele strains for S. cerevisiae essential genes. By screening a subset of the strains in a highly quantitative growth assay, the authors showed that the DAmP alleles result in a range of measurable fitness defects

16 that are comparable to the range of growth defects they quantified for the haploid deletion strains; however, it should be noted that many of the DAmP strains do not show growth defects. They also demonstrated the usefulness of this collection by performing several chemical-genetic screens, in which they correctly identified the known protein targets of several compounds, including 5-fluorouracil, sulfanilamide, and clotrimazole, whose targets are CDC21, FOL1/FOL3, and ERG11 respectively (Breslow, Cameron et al. 2008).

1.3.2 Gene overexpression

Gene overexpression is a complementary approach to loss-of-function studies and can provide insight into the function of genes with no deletion phenotype. Furthermore, gene overexpression is relevant to the molecular mechanisms of numerous diseases, such as cancer, where amplification and overexpression of oncogenes encoding c-Myc and the Src family kinases (SFKs) are implicated in disease initiation and progression (Little, Nau et al. 1983; Meyer and Penn 2008; Kim, Song et al. 2009). A better understanding of how gene overexpression perturbs genetic networks of simpler organisms could conceivably help shed light on the general principles governing the effects of oncogene overexpression in human cancer cells.

Several libraries are available to examine gene overexpression phenotypes in yeast. These libraries can be distinguished on the basis of how gene overexpression is achieved. The two most common methods of gene overexpression are the heterologous expression system and the high- copy expression vector.

In a heterologous expression plasmid, an ORF is placed under the control of a strong, inducible promoter; therefore, gene expression is regulated by repressing or activating the promoter. One commonly used inducible promoter is GAL1/10 (Schneider and Guarente 1991). This promoter contains upstream activating sequence (UAS) sites to which the transcription factor Gal4p can bind. In the presence of glucose, Gal4p is inactive, and transcription from GAL1/10 is repressed. When glucose is replaced by galactose in the growth medium, this molecule binds to and activates Gal4p, which then activates transcription from GAL1/10. In the

17 presence of galactose, transcription of the associated GAL1/10 ORF can be induced 1000-fold (St John and Davis 1979; St John and Davis 1981). This massive increase in gene product present within the cell is one explanation for the observation that ~15% of yeast ORFs expressed from GAL1/10 are toxic to yeast when overexpressed (Liu, Krizek et al. 1992; Akada, Yamamoto et al. 1997; Stevenson, Kennedy et al. 2001; Sopko, Huang et al. 2006; Vavouri, Semple et al. 2009). Other examples of inducible promoters, from which various levels of transcription are initiated, include the CUP1 promoter, induced by adding copper to the growth medium (Labbe and Thiele 1999; Martzen, McCraith et al. 1999) the MET25 promoter, repressed by adding excessive methionine to the growth medium (Thomas, Cherest et al. 1989), and the tetO-CYC1 TATA promoter, repressed by adding tetracycline or its analogue doxycycline to the growth medium (Gari, Piedrafita et al. 1997; Boyer, Badis et al. 2004).

An alternative way to confer gene overexpression is to place the gene on a high-copy plasmid, namely YEp (yeast episome) plasmid expression vectors that contain an origin of replication from a naturally occurring yeast plasmid called the 2µ circle. The 2µ circle exists at ~60 copies per cell and maintains itself at such levels in part by its ability to drive site-specific recombination that ultimately leads to copy number amplification (Murray 1987). The development of YEp vectors containing various prototrophic markers has allowed for in-house generation of random high-copy genomic libraries (Ma, Kunes et al. 1987). From genomic libraries, gene expression is regulated by the endogenous promoter and terminator sequences, but gene overexpression occurs because multiple copies of the expression plasmid are present within the cell (provided the concentration of activating transcription factors is not limiting).

The development and characterization of several genome-wide overexpression libraries is described below. The features of the libraries are summarized in Table 1.2 and depicted schematically in Figure 1.3. The availability of these libraries in an arrayed format greatly facilitates genome-wide investigations of genetic interactions in S. cerevisiae.

1.3.2.1 The Yeast Two-Hybrid S. cerevisiae ORF Array

Figure 1.3

Figure 1.3 Plasmid libraries in S. cerevisiae.

Schematic representations of the vector backbone and major features for each of the plasmid libraries in yeast are shown. a) Yeast Two-Hybrid ORF Library. b) PCUP1-GST Library. c) PGAL1/10-GST Library. d) Moveable ORF Library. e) FLEXGene ORF Collection. f) Yeast Genome Tiling Collection. See Sections 1.3.2.1 to 1.3.2.6, respectively for detailed descriptions of the plasmid libraries.

The ORF plasmid library developed by Uetz et al. (Uetz, Giot et al. 2000) was the first genome-wide collection amenable to investigating full-length protein-protein interactions using the yeast two-hybrid (Y2H) assay in an arrayed format. In this library, each strain carries a plasmid that contains one ORF cloned in frame with a Gal4 transcription-activation domain. The background yeast strain of this library has integrated yeast two-hybrid reporter genes (James, Halladay et al. 1996). Using this arrayed library, the authors screened for protein-protein interactions of 192 query ORFs and successfully identified 281 interacting gene pairs for 87 query ORFs (Uetz, Giot et al. 2000). Importantly, this landmark paper was the first description of an arrayed, high-density, semi-automated screening procedure for S. cerevisiae, which showed the potential to use an arrayed format as a basis for systematic genome-wide screening.

1.3.2.2 The PCUP1-GST Library

The PCUP1-GST library was the first systematic S. cerevisiae ORF plasmid library designed for purification of biochemically active gene products. In total, 6,080 ORFs fused N- terminally in frame with the coding sequence for the GST protein tag were individually cloned downstream of the inducible CUP1 promoter (Martzen, McCraith et al. 1999). The GST tag allows for easy purification of any ORF of interest. The authors used this library in biochemical proof-of-principle experiments to rapidly identify three previously uncharacterized ORFs with distinct biochemical activities: CPD1/YGR247W had substrate-specific cyclic phosphodiesterase activity, POA1/YBR022W had substrate-specific phosphatase activity, and CTM1/YHR109W had cytochrome c methyltransferase activity (Martzen et al., 1999).

1.3.2.3 The PGAL1/10-GST Library

In the PGAL1/10-GST library, each of the 5,800 yeast strains carries a plasmid that contains a single yeast ORF under the regulation of the inducible GAL1/10 promoter; to facilitate biochemical studies, each ORF is also N-terminally tagged with GST (Zhu, Bilgin et al. 2001; Sopko, Huang et al. 2006).

This plasmid library was first used in proteome chip experiments. The authors overexpressed all 5,800 ORFs in yeast and purified the protein products; these were spotted onto a microscope slide to generate a yeast proteome microarray (Zhu, Bilgin et al. 2001). In proof-of- principle experiments, the authors demonstrated that this proteome microarray can be used to detect both protein-protein and protein-lipid interactions. To identify calmodulin-binding proteins, the authors probed the array with biotinylated calmodulin in the presence of calcium, and subsequently detected bound calmodulin with Cy3-labeled streptavidin. They identified both known and novel calmodulin-interacting proteins (Zhu, Bilgin et al. 2001). To identify phosphoinositide (PI)-binding proteins, the authors probed the array with six different types of PI-containing liposomes that all also contained biotinylated phosphotidylcholine (PC); the biotinylated lipid facilitated detection of protein-PI interactions by binding to Cy3-labeled streptavidin. They identified 150 PI-binding proteins, including 52 that corresponded to uncharacterized proteins. Of the 98 known proteins, 45 were either membrane-associated or predicted to be membrane-associated, and another 8 were known to be involved in lipid metabolism, suggesting a significant enrichment for bona fide lipid-regulated proteins (Zhu, Bilgin et al. 2001).

Sopko et al. (Sopko, Huang et al. 2006) transformed this plasmid library into a yeast strain that is compatible with synthetic genetic array (SGA) technology (described below in Section 3.1)(Tong, Evangelista et al. 2001; Tong, Lesage et al. 2004). The authors performed a comprehensive investigation to identify the set of genes in this library that is toxic when overexpressed in yeast and identified 769 ORFs that cause lethality upon overexpression. They also demonstrated how this overexpression system is useful in synthetic dosage lethality studies (discussed below in Section 3.1.1).

1.3.2.4 The Movable ORF Library

In the movable ORF (MORF) plasmid library, a single yeast ORF was cloned into a 2µ vector backbone with a galactose-inducible promoter and a C-terminal His6-HA tag (Gelperin, White et al. 2005). This collection is “movable” because each ORF is flanked by directional att recombination sequences, and therefore can be subcloned into any att-containing vector. The

23 vector-ORF junctions for every plasmid in this library were sequenced, with complete ORF sequence verification for 55% of the collection. Protein expression levels were examined for 5,573/5,854 ORFs. The authors showed that 48 ORFs previously characterized as “dubious”, meaning they have very limited experimental evidence for their existence and do not have orthologs in other Saccharomyces species, were efficiently expressed using this overexpression system. To demonstrate the utility of this library to identify non-C-terminal post-translational modifications, the authors conducted a global analysis of glycosylation, as the preservation of the native N terminus of the proteins allows native processing to occur; they confirmed 109 new glycoproteins and identified another 345 candidate glycoproteins (Gelperin, White et al. 2005).

1.3.2.5 The FLEXGene ORF Collection

The Full Length EXpression-Ready (FLEXGene) plasmid collection comprises >5,000 S. cerevisiae ORFs (~87% of the protein-coding genes), each cloned into a Gateway donor vector (Hu, Rolfs et al. 2007). The ORFs in this collection contain their native stop codons, thereby allowing for either N-terminally tagged or native proteins to be overexpressed. Each ORF in this collection was PCR-amplified and fully sequenced from the start to the stop codons. To demonstrate the usefulness of this sequence-verified collection, the authors selected a subset of ORFs for protein overexpression, and applied a select set to protein binding microarrays in order to examine DNA-binding specificities. In particular, they were able to correctly identify the consensus DNA-binding sequence for the well-characterized transcription factor Rap1p (Hu, Rolfs et al. 2007).

1.3.2.6 The Yeast Genome Tiling Collection

The Yeast Genome Tiling Collection (YGTC) is a collection of minimally overlapping, ordered genomic fragments covering >97% of the S. cerevisiae genome (Jones, Stalker et al. 2008). To generate this library, S. cerevisiae genomic DNA was partially digested and then ligated into a 2µ vector, and the resulting ligation products were transformed into bacteria. Over 13,000 transformants were picked for sequencing of the vector-insert junctions. The resultant minimal tiling collection is a set of ~1,600 plasmids equally covering all 16 yeast chromosomes,

24 including centromeres, non-protein coding genes, and dubious ORFs. The authors used this collection to rapidly identify small genomic regions that, when overexpressed, cause specific transcription defects, and ultimately identified four novel transcriptional regulators (Jones, Stalker et al. 2008).

1.3.2.7 Summary and Comparison of Various Existing Overexpression Libraries

By comparing the individual features of the overexpression libraries described in Sections 1.3.2.1 to 1.3.2.6, it is apparent that each library has its own unique strengths and weaknesses.

With the exception of the YGTC, all of the systematic plasmid libraries described above rely on an inducer for overexpression. Several studies have shown that constitutive gene overexpression results in toxic effects, with ~15% of GAL-GST ORFs causing some type of overexpression phenotype in wild type haploid cells (Sopko, Huang et al. 2006). The YGTC is 2µ-based, so gene overexpression is due to high copy number. Jones et al. (Jones, Stalker et al. 2008) observed that less than 3% of the tiling collection plasmids caused toxicity when transformed into wild type haploid cells. This observation suggests that the cell can tolerate significant dosage increases of most ORFs when ORF expression is controlled by endogenous regulatory sequences.

As implied by the library names, every ORF in the CUP1-GST, the MORF, and the

PGAL1/10-GST overexpression libraries is tagged either N- or C-terminally with a non-native protein sequence. The tag is extremely useful in biochemical studies, as it facilitates ORF protein purification, but it might interfere with protein-protein interactions, protein turnover or protein stability. As well, some post-translational modifications required for proper protein function (e.g. N-terminal myristylation) or organelle import (e.g. mitochondrial import) are found at the termini of proteins, so the presence of the tag could also interfere with these modifications. The resulting protein, therefore, may not be properly localized or functional.

In all of the heterologous expression libraries, each plasmid contains a single ORF. In the YGTC, however, the average insert size in the YGTC is 8.7 kb (Jones, Stalker et al. 2008). Since the yeast genome generally contains one protein-coding gene every 2 kb (Goffeau, Barrell et al. 1996), most plasmids in this library contain >1 ORF. When this library is used in a genetic screen, deconvolution of the ORF ultimately responsible for the observed phenotype one is investigating is likely required.

In principle, the ideal plasmid-based gene overexpression library would have the following features: 1) endogenous regulatory sequences controlling ORF expression; 2) no epitope tag; 3) a single ORF per plasmid; and, 4) unique barcodes to facilitate parallel analysis of strain pools. A plasmid library that combines the “best” features of these libraries would complement, and in some cases possibly supplant, the resources that are currently available. Indeed, the construction, validation and application of such a library is the goal of the present thesis.

1.4 SYSTEMATIC IDENTIFICATION OF GENETIC INTERACTIONS IN YEAST

The development of the S. cerevisiae genome deletion collection (Giaever, Chu et al. 2002) was a seminal event in the history of functional genomics, as it allowed, for the first time, the systematic analysis of gene function across the majority of genes in an organism in parallel. However, as noted above (Section 1.1), most single gene deletion mutants have no obvious phenotype under standard growth conditions and, alone, this library is therefore unable to provide insight into the function of many genes. As discussed above, small-scale experimentation suggests that the interactions between genes contain rich functional information. Thus, one way to exploit the single gene deletion collection is to combine individual deletions together in higher order combinations. Several methods have been developed to accomplish this, as described in the following sections.

1.4.1 Synthetic Genetic Array (SGA) Analysis

The haploid deletion collection led to the development of synthetic genetic array (SGA) technology (Tong, Evangelista et al. 2001), which has two defining features. The first is that this technology employs robotic pinning of the collection onto any type of solid (agar) media. The second is that the query strain used has a mating type-specific reporter gene that allows for efficient and selection of haploid cells of only one mating type, thereby bypassing the requirement to dissect tetrads. In SGA, the deletion strains are placed in an ordered array format. This input array is mated to a query strain that is of opposite mating type and marked with nourseothricin (NAT)(Goldstein and McCusker 1999). Diploids are selected on media containing G418 and NAT. The diploids are then pinned onto sporulation media so that they can undergo meiosis to produce recombinant haploid cells. MATa haploids are selected on media lacking histidine; the media also contains two drugs, canavanine and S-2-aminoethyl-L-cysteine (S- AEC), specifically meant to kill any residual diploid cells. Subsequent pinning onto haploid selective media containing G418 and NAT ultimately produces the systematic double mutant array. Fitness is determined by quantifying double mutant colony size, which is compared to the cognate single mutant fitnesses. Using this technology, thousands of negative, i.e. synthetic lethal (SL)/synthetic sick (SS), interactions have been identified (Tong, Evangelista et al. 2001; Tong, Lesage et al. 2004; Costanzo, Baryshnikova et al. 2010). Extrapolation of this network suggests the existence of approximately 100,000 SL/SS interactions between non-essential genes in S. cerevisiae (under the standard conditions used). This technology has also been used to identify positive, i.e. alleviating, interactions for subsets of functionally related genes (Schuldiner, Collins et al. 2005; Collins, Miller et al. 2007; Laribee, Shibata et al. 2007; Nagai, Dubrana et al. 2008; Wilmes, Bergkessel et al. 2008).

SGA analysis was also used to identify genetic interactions involving essential genes by employing various tetO7 alleles as array strains. This work found both previously identified and novel genetic interactions for a temperature-sensitive allele of CDC40 (Mnaimneh, Davierwala et al. 2004). In a subsequent study, Davierwala et al. (Davierwala, Haynes et al. 2005) screened nine ts alleles against an array of 147 tetO7 strains. The genetic network produced from these screens showed that essential genes have approximately five-fold more SS/SL interactions than non-essential genes. Together with the results from Tong et al. (Tong, Lesage et al. 2004), the

27 prediction is that ~200,000 SS/SL interactions between all genes (essential and non-essential) in the S. cerevisiae genome.

1.4.1.1 Application of SGA to SDL analysis

As demonstrated in Sopko et al. (Sopko, Huang et al. 2006), the array-based method of identifying genetic interactions is not limited to the deletion collection nor limited to identifying synthetic lethal interactions. By mating the PGAL1/10-GST arrayed yeast strain library to a query strain deleted for the cyclin-dependent kinase PHO85 and employing a slightly modified SGA selection method, Sopko et al. (Sopko, Huang et al. 2006) overexpressed the PGAL1/10-GST ORFs and looked for synthetic dosage lethality (SDL) interactions to identify genetic interactors of PHO85. Importantly, the SDL genetic interactions identified in this screen identified proteins that were subsequently shown to be novel bona fide Pho85p phosphorylation targets (Sopko, Huang et al. 2006; Sopko, Huang et al. 2007). These results illustrate the usefulness of approaches that combine loss- and gain-of-function alleles as a means to explore genetic interaction networks.

1.4.1.2 Application of SGA to genetic mapping

The SGA methodology has also been used to map second-site mutations present in particular query strains, a technique referred to as Synthetic Genetic Array Mapping (SGAM)(Costanzo and Boone 2009). Since the position of every deletion in the deletion array is known, one can look for a second linkage group that is not linked to the query allele. SGAM has been used to identify recessive suppressors of genes required for polarized morphogenesis (Jorgensen, Nelson et al. 2002) and genome integrity (Chang, Bellaoui et al. 2005). More recently, SGAM was used to map dominant suppressors of the yeast ortholog of the Shwachman- Bodian-Diamond Syndrome protein, the human disease gene defective in Shwachman-Diamond Syndrome (SDS); this analysis provided the first molecular mechanistic explanation underlying SDS (Menne, Goyenechea et al. 2007).

1.4.1.3 Application of SGA to array-based high-content screening

In all applications described above, the fitness of the single and double mutants, as inferred from colony size, has been used as the phenotype from which genetic interactions are quantified. An alternative to measuring colony size as an indicator of fitness and thus a potential genetic interaction is to quantify the readout from a reporter gene, such as GFP, in a double mutant colony and then compare it to the readout from the cognate single mutants. Any deviations from the expected fluorescence may be due to genetic interactions that regulate a specific biological process. When a selectable fluorescent reporter gene is in the background of an SGA query strain, SGA technology can be employed to generate an array of double mutant strains carrying the reporter. Subsequent quantification of fluorescence can be done by either high-throughput flow cytometry, which can detect overall intensity, or microscopy, which can detect subcellular distribution in addition to intensity.

To comprehensively identify genes required in protein folding in the endoplasmic reticulum (ER), Jonikas et al. (Jonikas, Collins et al. 2009) used high-throughput flow cytometry to quantify single-cell overall fluorescence of a reporter gene that was activated as part of the cell’s unfolded protein response (UPR). They first used SGA technology to introduce the fluorescent reporter into the deletion collection and measured single mutant fluorescence intensities. Based on these results, they selected 340 genes, constructed pairwise double mutants expressing the reporter gene, and quantified the fluorescence to identify genetic interactions that contribute to regulation of the UPR. Since the reporter gene had a basal level of expression in wild-type cells, the authors were able to identify both positive and negative genetic interactions, which had lower and higher fluorescence intensities, respectively, when compared to expected values. They identified genetic interactions between a set of six poorly characterized ORFs and, in subsequent biochemical experiments, showed that these genes encode members of a novel transmembrane protein complex that may play a role in folding ER membrane proteins (Jonikas, Collins et al. 2009).

In another example, Vizeacoumar et al. (Vizeacoumar, van Dyk et al. 2010) developed a microscopy-based approach called high-content screening (HCS)-SGA, which uses high-content microscope imaging to examine subcellular phenotypes associated with various single and

29 double mutants. As a proof-of-principle, the authors used the tubulin reporter gene, GFP-TUB1, to investigate the spindle disassembly pathway. First, they used SGA technology to introduce GFP-TUB1 into the haploid deletion array; second, they crossed this array to two query strains, bni1Δ and bim1Δ, to generate double mutants carrying the fluorescent reporter. The authors used microscopy to image wild-type, single mutant and double mutant strains. They employed machine learning to detect and define the wild type fluorescence pattern, and then to detect and rank mutant strains based on their deviation from wild type. From their extensive analysis, the authors identified 122 genes that had not been previously implicated in spindle disassembly; furthermore, they were able to elaborate on the spindle disassembly pathway, and identified a novel role for sumoylation at the kinetochore as a mode of regulating this pathway.

1.4.2 Diploid-based Synthetic Lethal Analysis on Microarrays (dSLAM)

Diploid-based synthetic lethal analysis on microarrays (dSLAM; (Pan, Yuan et al. 2004)) takes a different approach to double mutant construction and analysis. In dSLAM, the diploid heterozygous, kanMX-marked and barcoded deletion collection is pooled and transformed en masse with a linearized, URA3-marked disruption cassette targeting the query gene of interest. After transformants are selected, diploids are sporulated, and double mutant haploids are subsequently isolated on selective media. The double mutants are pooled and their barcodes amplified using universal primers, one of which is conjugated to the fluorescent dye Cy3. Concurrently, a single mutant pool is also generated using the same selection steps (with the exception that uracil is added to the medium since no URA3 cassette is transformed into this pool) and serves as a control sample for comparison. For the control sample, one of the universal primers used for PCR amplification of the barcodes is conjugate to Cy5. The Cy3- and Cy5- labelled PCR products are competitively hybridized to a barcode microarray, and a control/experiment (C/E) ratio is calculated for each barcode. High C/E ratios are indicative of synthetic sick or synthetic lethal interactions, meaning the barcode is under-represented in the experimental (double mutant) pool as compared to the control (single mutant) pool. Conversely, low C/E ratios are indicative of potential synthetic suppression interactions. Using this method, Pan et al. (Pan, Yuan et al. 2004) identified SS/SL interactions for CIN8, a kinesin motor protein involved in mitotic spindle assembly and chromosome segregation, and CHI interactions for

TUB1. Recently, Kim et al. (Kim, Zhao et al. 2009) used dSLAM to investigate the genetic requirements underlying the unfolded protein response, identifying vesicular trafficking as an important compensatory mechanism for defects in ER protein maturation.

1.4.3 Genetic Interaction Mapping (GIM)

A third synthetic genetic interaction methodology is the Genetic Interaction Mapping (GIM) approach (Decourty, Saveanu et al. 2008). The GIM method combines features of SGA technology and dSLAM in order to identify genetic interactions in yeast (Decourty, Saveanu et al. 2008). In GIM, the query strain carries a hygromycin-resistant plasmid, while the query gene ORF is replaced by a MATα2-NATR cassette, which is expressed only in MATα cells. An SGA- like approach is used to mate the query strain to the deletion array, and then to pin the mating products onto diploid selection media containing hygromycin and G418. The diploids are then pooled, and sporulation is performed en masse in liquid. Double mutant haploid cells are selected for in rich media liquid containing NAT and G418. Genomic DNA is extracted from the pool, and the barcodes are PCR-amplified and hybridized to a barcode microarray. To normalize the microarray results, a reference screen, in which the MATα2-NATR cassette replaces a neutral ORF, is done in parallel. Using the GIM method, the authors identified novel genetic interactions, both positive and negative, that participate in mRNA decapping (Decourty, Saveanu et al. 2008). They extended their method to 41 query strains, all involved in RNA metabolism, and show that their results have statistically significant overlap with previously published high- throughput genetic interaction studies (Decourty, Saveanu et al. 2008).

1.4.4 Summary of Genetic Interaction Mapping Strategies

With the exception of synthetic dosage lethality studies (Sopko, Huang et al. 2006), all of the genome-wide genetic interaction mapping studies in yeast have used loss-of-function mutants for identifying negative and positive interactions. With the development of several different genome-wide overexpression libraries, it is now feasible to systematically examine synthetic dosage lethality and suppression interactions in different genetic backgrounds.

Recent proof-of-principle experiments in other microorganisms have shown it is possible to detect genetic interactions using next-generation sequencing methods ((van Opijnen, Bodi et al. 2009), see below). A broader application of these novel methods has the potential to make the detection of genetic interactions, especially when using barcoded strains and plasmids, much more streamlined and quantitative, and these methods will therefore be introduced in greater detail in the next section.

1.5 NEXT-GENERATION SEQUENCING

For 30 years, Sanger-based methods were the primary means to generate DNA sequence data (Hunkapiller, Kaiser et al. 1991; Shendure and Ji 2008). The Sanger method is accurate, allows long read lengths (500-1000 bp / template) and the cost per base of sequence has come down exponentially since the initial development of this technique. Moreover, advances in sample handling and automation, such as micro-capillary-based automated sequencers, enabled Sanger sequencing to be applied to large genomes, including that of humans, but at tremendous cost (Lander, Linton et al. 2001; Venter, Adams et al. 2001). More recently, a number of next- generation (also known as ‘second-generation’ or ‘deep’) sequencing technologies have been commercialized, and are known by names such as 454, Solexa, Helicos and SOLiD sequencing (Shendure and Ji 2008). Each of these methods relies on a unique set of chemistry and imaging tools. Common to all approaches is the ability to generate sequence data at vastly lower cost per base and in a much shorter timeframe compared to traditional Sanger sequencing, albeit with the caveats of generally shorter read lengths per template (30-400 bp, depending on the system), higher error rates and more complex downstream bioinformatic analysis.

While next-generation sequencing methods have not yet been described to identify genetic interactions in yeast, they have been successfully used in another unicellular organism to identify novel genetic interactions. van Opijnen et al. (van Opijnen, Bodi et al. 2009) used the Solexa/Illumina platform to establish genetic interaction profiles of five query genes in the gram- positive bacterium Streptococcus pneumoniae as follows. First, the authors generated five query strains by replacing each of the five query genes with a drug-resistant marker. Second, they established transposon insertion libraries for each query strain, obtaining between 10,000-25,000

32 transposon mutants in each library. Third, they pooled the transposon mutants for each strain. Fourth, a sample of the pool was grown in liquid from lag phase to late exponential phase, at which time the DNA was harvested; to serve as a control, while DNA was also harvested from a pre-liquid growth pool sample. Fifth, DNA samples were prepared for sequencing on a Solexa/Illumina flow cell lane and sequenced using the Illumina Genome Analyzer. By introducing a unique DNA sequence “tag” for each of the five samples during preparation of the samples for sequencing, the authors were able to sequence DNA samples from all 5 query strains on the same lane (van Opijnen, Bodi et al. 2009). The sequence reads from each query strain were then separated based on the different DNA sequence tags. Based on the abundance of reads for a given transposon insertion site in the different strains, it was possible to quantify which strains ‘dropped out’ and which strains were ‘over-represented’ in a given background, similar to what is done using barcode microarrays. Using this next-generation sequencing approach, the authors identified 97 high-confidence genetic interactions for S. pneumoniae involved in transcriptional regulation and carbohydrate transport (van Opijnen, Bodi et al. 2009). These results show how next-generation sequencing can be applied in principle to the detection of genetic interactions. In particular, they highlight how a single run in one lane of the Solexa/Illumina platform can, in principle, detect genetic interactions for multiple strains in parallel. A similar application of next generation sequencing to the quantification of genetic interactions in S. cerevisiae has been suggested (Smith, Heisler et al. 2009), but yet to be reported.

1.6 SUMMARY AND RATIONALE

Here I have introduced the various means which have been used to identify genetic interactions and how this knowledge has been useful in discovering novel roles for individual genes and in understanding the overall functional organization of the cell. Most genetic interaction studies to date have used of loss-of-function alleles, in part because the methods to generate such reagents are simple and well established. It is clear, however, from both small- and large-scale studies that the analysis of gain-of-function mutations can provide information that is either complementary to that obtained using loss of function alleles, or in fact inaccessible using loss-of-function alleles only. Indeed, the analysis of essential gene function is by definition

33 especially difficult when using loss-of-function methods alone. An additional consideration is that the time and cost involved in generating the existing loss-of-function interaction maps has proven to be substantial; new methods such as next-generation sequencing may be able to help bring down these resource costs and increase the amount of data that can be generated in a given amount of time.

In this thesis, I address these issues. First, I describe the cloning of two novel gene overexpression libraries that are designed to streamline gene dosage studies. Second, I describe the application of one of the libraries in gene dosage studies, specifically in dosage suppression screens that use barcode microarrays and next-generation sequencing as readouts for identifying candidate dosage suppressors. Third, I show that dosage suppression interactions represent a new functional edge in the yeast interaction landscape that can be used to both investigate gene function and identify novel functional relationships in the cell.

Chapter Two The MoBY-ORF 1.0 Yeast Plasmid Library

The work reported in this chapter was a collaboration between me, Leslie J. Magtanong, Cheuk Hei Ho, and Sarah Barker. Cheuk Hei Ho and Bilal Sheikh, a former computer support person in the Boone lab, designed the primers used in PCR amplification of each ORF along with native upstream and downstream sequence. Cheuk Hei Ho and I did the PCR amplification reactions. Cheuk Hei Ho, Sarah Barker, and I did the yeast and bacterial transformations, plasmid extractions, restriction digests, and agarose gel analyses to evaluate the restriction digest results. Sarah Barker did the sequencing and functional tests of MoBY-ORF 1.0 clones. Cheuk Hei Ho performed all of the chemical genetics experiments using MoBY-ORF 1.0. A complete description of this work was published in Nature Biotechnology (2010), 27:369-77; doi:10.1038/nbt.1534 (URL: http://www.nature.com/nbt/journal/v27/n4/full/nbt.1534.html).

34 35

2.1 INTRODUCTION

As described in the general Introduction (Chapter 1), one goal of this project is to identify dosage suppression genetic interactions for essential genes using a new overexpression library in combination with barcode microarrays and next-generation sequencing. In this Chapter, I describe the cloning and construction of the first of two new plasmid overexpression libraries for yeast that we developed, called the molecular barcoded yeast ORF (MoBY-ORF) libraries. We designed the MoBY-ORF vector backbone to be compatible with an in vivo bacterial cloning method called mating-assisted genetic integrated cloning (MAGIC)(Li and Elledge 2005), which uses homologous recombination to transfer DNA sequences from one vector backbone to another. In the next chapter (Chapter 3), I discuss how we generated the second MoBY-ORF library by using MAGIC to transfer the barcoded ORFs from the low-copy vector to a high-copy vector, and then pursued dosage suppression genetic interaction studies of S. cerevisiae essential genes using the high-copy overexpression MoBY-ORF library.

The MoBY-ORF 1.0 plasmid library is a low-copy overexpression library due to the centromere (CEN) sequence found on each plasmid, which constrains it to 1-3 copies per cell (Tschumper and Carbon 1983). Each plasmid in the library contains a single ORF that is flanked by ~900 bp and ∼250 bp of native upstream and downstream genomic sequence respectively, and is therefore largely under the control, to an extent, of its endogenous regulatory elements. Additionally, each plasmid in the MoBY-ORF library carries unique barcodes which act as molecular tags (Shoemaker, Lashkari et al. 1996); these features make this collection first yeast ORF library amenable to barcode microarray analysis (Pierce, Fung et al. 2006). While this plasmid library has many genetic and chemical-genetic applications, we first used the library to identify the mode-of-action of several bioactive compounds. In the process, we discovered a new class of sterol-binding chemicals.

2.2 RESULTS

2.2.1 Construction of a library of molecular barcoded yeast ORFs

The MoBY-ORF 1.0 library consists of plasmids that each carry a pair of oligonucleotide barcodes and a single yeast ORF that is flanked by its native upstream and downstream genomic sequences. The plasmid vector p5472 carries a URA3 selectable marker and a yeast centromere, which maintains one to three copies of the plasmid per cell (Figure 2.2.1). The vector was designed to be compatible with an in vivo bacterial cloning method, mating-assisted genetically integrated cloning (MAGIC)(Li and Elledge 2005), which facilitates the rapid construction of recombinant DNA molecules, enabling the barcoded clones to be transferred efficiently to other vector backbones, such as a high-copy vector. The barcode cassettes were obtained from the yeast deletion mutant collection (Giaever, Chu et al. 2002) and comprise two unique 20- nucleotide DNA sequences (labeled the UPTAG and DNTAG) flanking a dominant selectable marker (kanMX) that confers resistance to the drug G418/kanamycin. The barcodes can be amplified with universal primers, enabling cells carrying a specific ORF to be quantitatively detected with a microarray having probes that hybridize to the barcode sequences (Pierce, Fung et al. 2006). Each plasmid was constructed in a three-step process. First, each yeast ORF was PCR-amplified from an average of 900 bp upstream of the start codon to an average of 250 bp downstream of the stop codon using a DNA template isolated from the sequenced S288C strain. In addition, the kanMX barcode sequences that uniquely identify the ORF were PCR-amplified from the appropriate strain in the yeast deletion collection. Second, the plasmid was assembled by homologous recombination by transforming yeast with the ORF, the barcode PCR products and linearized p5472 (Figure 2.2.2). Third, recombinant plasmids were recovered and used to transform bacteria to facilitate plasmid DNA isolation and subsequent diagnostic restriction digests to confirm the sizes of both fragments.

2.2.2 Verification of constructed clones by sequencing

Each clone in the MoBY-ORF library was sequenced to confirm the 3’ portion of the gene and the barcodes. We identified 4,396 ORFs (88.7%) with two unique barcodes, but 560 with only one barcode (344 with only an UPTAG and 216 with a DNTAG), as the other barcode was either not unique within the collection (i.e. multiple clones contained the same barcode sequence), or it had no corresponding sequence on the Affymetrix TAG4 microarray (Pierce, Fung et al. 2006). In summary, the MoBY-ORF 1.0 library contains 4,956 uniquely barcoded 37

Figure 2.2.1

!"#$%&'#()*+#,-./01 2'3$&#!*4*5*6*7$#8'9*4:%;3&%*;#(!81#<'=7';9' $%&/

-+,(%)!+#+, ')#(%)1*+, !"# *+,3(%)1*#, *+,3(%)"2#+, 2'3$&#!*4*5*6*7$#8'9*4:%;3&%*;#(!81#<'=7';9' !>#$%&'#()*+#,-./01

$%&-(((%)./*, &'()*a &'(+ !"#$% -+,(%)"1+#, !"#$%&' !"#$

!"#(%)*"+#,

')#(%)*+//, '(#0(%)+"+1,

'(#0(%)+22#,

!"#$%&'(%#)(*+(),-./0(123(4*%%*5(6347*8(9#4:9*53(+*8(723(;*<=>?@A("&98#8B( 4*57#&5$(723(+*""*C&5D(3"3%357$E(F@$(G9#$3>)#&8$(HI>JJ-(K(J,J>JH,L(354*'3(723( $3MN3543$( N$3'( +*8( B3#$7( 2*%*"*D*N$( 834*%9&5#7&*50( !"#$%&'(( GJ-.O>PJO,L( 354*'3$(723(B3#$7(43578*%383(#5'(&'()$3MN35430(*'&+)GP,PP>-PPQL(354*'3$(723( *'&+)4*'&5D($3MN3543(+*8($3"347&*5(&5(B3#$70(!,')G-.,H>,-J.L(354*'3$(723(42"*8> #%)235&4*"(4*'&5D($3MN3543(+*8($3"347&*5(&5(9#4738�(RUL.ȖG./->JI.-L(354*'3$( 723(9#4738&#"(*8&D&5(*+(83)"&4#7&*50(-./0)GJIO,>J-J/L(354*'3$(723(*8&D&5(*+(78#5$+38( $3MN3543R(#5'(FP(GP.>.JL(K(F-(GJO->//HL(354*'3(723(2*%*"*D*N$(834*%9&5#7&*5( $3MN3543$(+*8(;STUV(4"*5&5D0

Figure 2.2.2

ORFs, representing ~90% of all non-dubious ORFs annotated in the Saccharomyces Genome Database (SGD). This collection is available from Open Biosystems.

2.2.3 Assessment of clone function using temperature-sensitive mutants

To assess the functionality of our clones, we first introduced 254 different MoBY-ORF plasmids into a synthetic genetic array (SGA)(Tong, Evangelista et al. 2001) query strain. We then used the SGA method to cross the plasmids into a set of corresponding temperature- sensitive mutants covering alleles of the same 254 essential genes, and tested the transformants for functional complementation at the restrictive temperatures. In total, 17 clones failed to rescue the temperature sensitivity of the corresponding mutant strain, suggesting that ~93% of the clones in the library should be functional.

2.2.4 Complementation cloning to identify drug-resistant mutants and compound mode-of-action

Cheuk Hei Ho, a graduate student in the Boone lab, developed and used a strategy to efficiently clone drug-resistant genes by complementation with the MoBY-ORF library. In proof-of-concept experiments, he successfully identified the mutated loci in a first mutant conferring drug resistance to cycloheximide and a second mutant conferring drug resistance to rapamycin: CYH2 and FPR1, respectively. Cheuk Hei went on to use this strategy to eventually identify the ergosterol biosynthesis pathway as the target of two natural compounds, stichloroside and theopalauamide, that are structurally different but have virtually identical chemical-genetic profiles, which catalog the set of deletion mutants that are hypersensitive to a chemical compound (Parsons, Lopez et al. 2006). By identifying the target pathway of these two compounds, stichloroside was characterized as a member of the α-tomatine class of sterol- binding compounds, while theopalauamide and another compound, theonellamide A, belong to a novel class of sterol-binding compounds.

2.3 SUMMARY

In this Chapter, I described the cloning and construction of the MoBY-ORF library, the first yeast plasmid library amenable to analysis by barcode microarray. The MoBY-ORF 1.0 collection contains barcoded plasmids for ~90% of all non-dubious ORFs annotated in the Saccharomyces Genome Database (SGD, www.yeastgenome.org). While the MoBY-ORF 1.0 library has many applications, we used the library in chemical-genetic experiments to identify the mode-of-action of several bioactive compounds. Through this process, we discovered a novel class of steroid-binding chemicals.

2.4 METHODS

2.4.1 Yeast Strains

Y1239 (BY4741) MATa ura3Δ0 leu2Δ0 his3Δ met15Δ; Y1241 (BY4743) MATa/α his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 ura3Δ0/ura3Δ0 met15Δ0/+ lys2Δ0/+; Y7092 MATα can1Δ::STE2 pr-Sp_his5 lyp1Δ; his3Δ1 leu2Δ0 ura3Δ0 met15Δ0

2.4.2 Growth Media

Yeast strains were grown in YPD (1 % yeast extract, 2 % peptone, 2 % glucose) with G418 (200 µg/ml) to select for the plasmid. SD-URA+G418 (0.17 % yeast nitrogen base without amino acid or ammonium sulfate, 0.1% L-glutamic acid sodium salt hydrate, 0.2% amino acid supplement minus URA, 2% glucose) was used in homologous recombination cloning for library construction. Bacteria were grown 2YT (1 % yeast extract, 1.6 % tryptone, 0.5 % sodium chloride) with tetracycline at 5 µg/ml, kanamycin at 50 µg/ml, and chloramphenicol at 12.5 µg/ml.

2.4.3 Clone Construction and Analysis

Each ORF in the library was amplified using ORF-specific primers (Operon)(50 mM) with 2.1 U Roche HF Taq (Roche) or 3 U TaKaRa ExTaq (Takara Bio Inc.) polymerase, 0.2 mM dNTPs, and 4 ng S. cerevisiae genomic DNA (extracted from strain Y1239) in 50 µl reactions. 41

The following cycling conditions were used: 94°C for 2 min, 39 cycles of 94°C for 15 sec, 55°C for 30 sec, and 68°C for 4 min, followed by 68°C for 10 min. A sample of each ORF PCR product was visualized for the correct size on a 1% agarose gel, and upon confirmation, the remaining PCR product was column-purified (Invitrogen PureLink PCR Purification Kit). Barcoded kanMX cassettes were PCR-amplified using one of three templates. We first amplified the kanMX cassettes from whole cell lysate from the deletion strains (template 1). If we were unable to amplify from whole cell lysate, we prepared genomic DNA from the deletion strain (template 2). If we were still unable to amplify from either of the two templates, we synthesized oligonucleotide primers that contained the barcodes and PCR-amplified a kanMX cassette from a kanMX cassette-carrying plasmid (template 3). Template 1 was made from 5µl of S. cerevisiae heterozygous deletion strain cell lysate prepared by zymolyase digestion (~2x107 cells in 30 µl zymolyase at 5 mg/ml). Template 2 was 4 ng of S. cerevisiae heterozygous deletion strain genomic DNA prepared by standard phenol-chloroform extraction. The kanMX cassette using these two templates was PCR-amplified in two overlapping fragments. Template 3 was 0.2 ng of plasmid P1970, which carries the kanMX4 cassette. The kanMX cassette for template 3 was PCR- amplified as a single fragment. Each PCR reaction contained 1 U Roche HF Taq (Roche) or 3 U TaKaRa ExTaq (Takara Bio Inc.) polymerase, and 0.2 mM dNTPs. The following cycling conditions were used for all three templates: 94°C for 2 min, 39 cycles of 94°C for 15 sec, 50°C for 1 min, and 72°C for 2 min, followed by 72°C for 7 min. A wild-type diploid strain (Y1241) was co-transformed using standard procedures with the ORF, the kanMX PCR products and XhoI-linearized p5472 (Figure 2.2.2). Yeast transformants were selected for positively on

SDMSG-URA+G418. Recombinant plasmids were extracted from individual yeast transformants by zymolyase digestion followed by a miniprep plasmid preparation (Macherey Nagel). Miniprep DNA was transformed into competent P5505 [Dlac-169 rpoS(Am) robA1 creC510 hsdR514 DuidA(MluI):pir-116 endA(BT333) recA1 F’(lac+ pro+ DoriT:tet)](provided by Gwenael Baedis), followed by selection on 2YT containing tetracycline at 5 µg/ml, kanamycin at 50 µg/ml, and chloramphenicol at 12.5 µg/ml. Miniprep DNA was prepared from a single bacterial transformant, doubly digested using BamHI and EcoRI (Fermentas), and run out on a 0.8% agarose gel to confirm vector and ORF fragment sizes . Two individuals performed gel analysis independently, and the results were compared to determine clone validity. If the primary BamHI/EcoRI digest was ambiguous, a secondary digest using NotI/HindIII was performed. 42

ORF fragment sizes for both double digests, along with complete ORF sequence and barcodes, can be obtained from the MoBY-ORF database: http://moby.ccbr.utoronto.ca.

2.4.4 Sequence Confirmation of the MoBY-ORF Collection Barcodes and 3’ ORF Junctions

For each clone, two sequencing reactions were performed, the first covering the UPTAG and 3’ ORF junction and the second covering the DNTAG (with sequencing primers 5’- TATACATGGGGATGTATGGGC-3’ and 5’-GGGCAACAACAGATGGCTG-3’ respectively). The sequencing reads were analyzed using computational scripts developed to identify the barcode position, based on the adjacent universal primer sequences (5’- GACCTGCAGCGTACG-3’ for the UPTAG and 5’CGGTGTCGGTCTCGTAG-3’ for the DNTAG), and the identified barcodes were extracted. Since the reverse complement of the UPTAGs were sequenced (to accommodate reading the 3’ ORF junction within the same read), the UPTAGs were transposed into the common 5’-3’ top-strand notation. In cases where barcodes failed to be extracted or there was an ambiguous nucleotide call within the barcode, a manual review of the sequence was performed and barcodes recorded. The nucleotide BLAST program was used to map the 3’ ORF junction sequence to the full genome sequence from the Saccharomyces Genome Database (SGD, www.yeastgenome.org), and the position results were compared to the expected ORF annotation for confirmation of the clone.

2.4.5 Functional Complementation of Essential Genes

350 temperature sensitive strains (strain background Y7092) marked with natMX4 were selected for our complementation studies. Clones for the ORF of each ts allele were transformed into a complementary mating strain (Y1239). The temperature sensitive strains were mated with yeast carrying the cognate MoBY-ORF plasmid, and the diploids were taken through SGA (Tong, Evangelista et al. 2001) to obtain the double positive progeny: NATR representing the ts allele and G418R for the MoBY-ORF plasmid. These progeny were grown at both 26oC and 35oC and scored for rescue of the temperature sensitive phenotype. For strains in which the initial temperature sensitive nature was subtle, the NATR G418R cells were individually assessed at higher temperatures.

Chapter Three Mapping Genetic Networks by Systematic Dosage Suppression

Cheuk Hei Ho, Sarah Barker, and I, Leslie J. Magtanong were involved in MoBY- ORF 2.0 construction. Cheuk Hei Ho, Sondra Bahr, Elena Kuzmin and I carried out barcode microarray experiments to identify candidate dosage suppressors. Cheuk Hei Ho, Sondra Bahr, Elena Kuzmin, Kerry Andrusiak, and Anna Kobylianski did transformations to confirm suppressors. Andrew Smith carried out all the sample preparation and data analysis of the Bar-seq experiments. Wei Jiao and Anastasia Baryshnikova did the computational analysis of features of dosage suppression gene pairs, which included the overlap with other types of interactions, shared gold standard GO terms, heat maps displaying frequencies within and between biological processes, and identification of clusters in the integrated dosage suppression genetic interaction network. I developed the decision tree to categorize dosage suppression interactions. Cheuk Hei Ho and I were involved in re-categorizing unknown dosage suppression interactions. Cheuk Hei Ho performed the reciprocal suppression tests. All electronic tables can be found on the DVD accompanying this thesis document.

43 44

3.1 INTRODUCTION

Increasing gene dosage provides a powerful means of probing gene function as it tends to cause an increase in gene activity referred to as a gain-of-function effect (Sopko, Huang et al. 2006; Vavouri, Semple et al. 2009). Gene overexpression is relevant to the molecular mechanisms of diseases, such as cancer, where gene amplification and gain-of-function mutations are prominently implicated in disease initiation and progression (Santarius, Shipley et al. 2010). In yeast, systematic analysis of gene overexpression has revealed that only a subset of genes cause an overt phenotype when overexpressed in wild-type cells (Moriya, Shimizu- Yoshida et al. 2006; Sopko, Huang et al. 2006; Jones, Stalker et al. 2008; Kaizu, Moriya et al. 2010). However, examining gene overexpression in sensitized cells containing mutations in genes of known function is an effective way to probe gene activity because it can identify functionally relevant genetic interactions (Rine 1991; Prelich 1999; Sopko, Huang et al. 2006; Boone, Bussey et al. 2007; Dixon, Costanzo et al. 2009).

Genetic interaction networks map functional connections occurring both within and between cellular pathways. An extensive genetic interaction network based upon loss-of-function double mutant analysis has recently been described for budding yeast (Costanzo, Baryshnikova et al. 2010). In addition, a large-scale physical interaction network for yeast has been assembled from multiple data sources (Gavin, Bosche et al. 2002; Krogan, Cagney et al. 2006; Tarassov, Messier et al. 2008; Yu, Braun et al. 2008). Direct interactions between genes or gene products are referred to as ‘edges’ in biological networks. The edges in genetic interaction networks largely complement those found in protein-protein interaction networks, as only a small fraction of gene pairs that show a genetic interaction also physically interact (Costanzo, Baryshnikova et al. 2010). This implies that an integrated network is more informative than either type of network alone. Despite the mapping of these genome-scale networks, our knowledge of the cell remains incomplete and novel methodologies that map new types of genetic interactions should improve our global understanding of the functional wiring diagram of the cell and provide further insight into the roles of specific genes.

Suppression or enhancement of a query mutant phenotype by gene overexpression is called dosage suppression (DS) or synthetic dosage lethality (SDL) respectively. Identification of SDL interactions in yeast has largely relied on plasmid libraries containing genes whose overexpression can be induced from the heterologous GAL1 promoter (Schneider and Guarente 1991). Using this approach, SDL interactions have been identified for genes encoding kinetochore components and members of the origin recognition complex (Kroll, Hyland et al. 1996; Measday, Hailey et al. 2002; Measday, Baetz et al. 2005). More recently, an arrayed galactose-inducible overexpression library was developed and used in SDL experiments to identify novel kinase-substrate relationships (Sopko, Huang et al. 2006; Sopko, Huang et al. 2007).

Classical gene dosage suppression studies in yeast have been productively performed using random genomic high-copy libraries (Ma, Kunes et al. 1987); however, plasmids in these libraries often carry large genomic fragments and the suppressing gene must be identified through an additional round of experiments. The recent availability of an ordered tiling library (Jones, Stalker et al. 2008) has simplified the identification of genomic fragments with suppressing activity but may not precisely identify the key gene of interest. To facilitate the process of gene identification and to enable facile and systematic gene dosage analysis, we generated an overexpression plasmid library, MoBY-ORF 2.0, that is compatible with high- throughput genomics technologies, such as barcode microarrays and next-generation sequencing methods (Pierce, Davis et al. 2007; Smith, Heisler et al. 2009), which can monitor plasmid representation in pools of transformed strains. As proof-of-principle, we use the MoBY-ORF 2.0 library for dosage suppression screens of an extensive collection of temperature-sensitive conditional alleles of essential genes. Analysis of data from these screens demonstrates the utility of systematic studies of dosage suppression analysis in yeast to provide a new class of functional edge in the global yeast interaction network.

3.2 RESULTS

3.2.1 Construction of the MoBY-ORF 2.0 plasmid library

We developed a high-copy (2µ-based) plasmid library in which each plasmid contains a DNA insert composed of a single yeast ORF with its native upstream and downstream genomic sequences, along with a kanMX marker flanked by two unique 20-nucleotide molecular barcode tags (Figure 3.2.1). The DNA insert was derived from the low-copy (CEN-based) molecular barcoded yeast ORF (MoBY-ORF) 1.0 library (Ho, Magtanong et al. 2009) and transferred to the high copy vector (Figure 3.2.2) by mating-assisted genetically integrated cloning (MAGIC) (Li and Elledge 2005) (Figure 3.2.3). The final MoBY-ORF 2.0 plasmid library contains 4547 clones (representing 4499 ORFs), of which 91% have 2 usable barcodes, 5% have a unique uptag only, and 4% have a unique downtag only; therefore, all plasmids in the library are represented by at least one barcode.

3.2.2. Dosage suppression analysis of temperature-sensitive conditional mutants

The MoBY-ORF 2.0 plasmid library provides a reagent set tailored for gene dosage analysis because the barcodes enable highly parallel assessment of individual plasmid abundance within a mixed population. While there are numerous applications of the MoBY-ORF 2.0 plasmid library, we developed a multi-step protocol to explore its use in identifying dosage suppressors of conditional temperature-sensitive (ts) alleles of essential genes (Figure 3.2.4). First, a ts strain was transformed with the MoBY-ORF 2.0 plasmid library, aiming for >10-fold representative coverage of each plasmid. Accordingly, in ~50,000 transformants, each plasmid should be represented ~11 times. Second, the transformants were pooled, and 50,000 cells were plated on selective media and incubated at both permissive and semi-permissive temperatures. Third, colonies appearing after 3 days were pooled. Fourth, the barcodes from the dosage suppressor (semi-permissive temperature) pool were PCR-amplified using biotinylated universal primers; concurrently, the barcodes from the control (permissive temperature) pool were PCR- amplified using non-biotinylated universal primers. Fifth, the PCR products from the dosage suppressor and starting pools were competitively hybridized to a barcode microarray. The

Figure 3.2.1

Figure 3.2.2

Figure 3.2.3

Figure 3.2.3 MAGIC with the MoBY-ORF 1.0 plasmid library. a. The donor bacterial strain carrying the CEN-based MoBY-ORF plasmid is mated to a recipient bacterial strain carrying a 2µ-based vector. b. Expression of a restriction enzyme, I-SceI, releases the barcoded ORF and linearizes the recipient vector. Expression of a recombinase in the recipient strain induces homologous recombination between the barcoded ORF and linearized vector using the MAGIC sequences (filled yellow circles). c. The recipient strain is grown on media containing ampicillin, kanamycin, and DL- chlorophenylalanine that selects both against any non-recombined vector and for the recombinant plasmid.

Figure 3.2.4

Figure 3.2.4 Using the MoBY-ORF 2.0 plasmid library to identify candidate dosage suppressors by barcode microarray.

A ts mutant strain is transformed with the MoBY-ORF 2.0 plasmid library. In this case, the ts mutation is in a gene designated as C. Transformants (represented in different colors) are pooled and plated at permissive and semi-permissive temperatures. Barcoded plasmids are extracted from colonies present at the semi-permissive temperature after 3 days, and barcodes are amplified using biotinylated universal primers. Concurrently, barcoded plasmids are extracted from the control pool grown at the permissive temperature, and barcodes are amplified using non-biotinylated universal primers. The biotinylated and non-biotinylated PCR products are mixed and competitively hybridized to a barcode microarray. Only the microarray-bound biotinylated PCR products (indicated with blue circles) will bind to the streptavidin-conjugated fluorescent dye used to detect hybridization to the microarray. Barcodes with intensities 5-fold and higher above background represent candidate dosage suppressors of the ts allele. In this case, barcodes from plasmid C are detected. This is expected, as plasmid C carries the wild-type gene of the ts allele c. Barcodes from plasmid A are also detected; therefore, gene A represents a candidate dosage suppressor.

53 competition between biotinylated and non-biotinylated PCR products allowed for candidate dosage suppressors to be identified by their barcodes as having a significantly higher signal intensity, which was defined empirically as any signal 5-fold or higher above background. (Figure 3.2.5; see Methods 3.4.6). To confirm dosage suppression interactions, candidate dosage suppressor plasmids were individually transformed into the cognate ts strain, and spot dilutions were performed at the semi-permissive temperature.

When selecting the 40 different query genes for dosage suppression screens, we focused on genes with a variety of different functional roles and relatively few previously published dosage suppression interactions (Appendix 6.1). In total, we performed confirmation transformations and spot dilution assays to validate 214 different suppressing plasmids (Appendix 6.1). The wild-type complementing ORF was recovered for all but 3 query strains. We expected a few cases in which we would not observe complementation because the 2µ MoBY-ORF library does not contain wild-type plasmids for ~20% of yeast ORFs, including those of the remaining 3 strains, and most (~93%) but not all of the PCR-amplified genes are functional (Ho, Magtanong et al. 2009). Of the remaining 168 extragenic dosage suppressors, three plasmid clones carried genes immediately next to the wild-type complementing ORF (Appendix 6.1). Since each MoBY-ORF 2.0 plasmid carries native upstream and downstream ORF sequence, two ORFs on a single plasmid can occur if the intergenic region is relatively small; in these three cases, the plasmids each contained the respective wild-type query gene. For several query genes, we screened multiple alleles and recovered the same suppressor gene.

For 8 query genes, the wild-type complementing ORF was the only clone recovered. For the remaining 32 query genes, we identified at least one dosage suppressor, with a total of 150 extragenic dosage suppressors. The number of dosage suppressors varied widely from one to 24 (Appendix 6.2; Electronic Table 3.2.1), but on average, we recovered ~5 dosage suppressors per query gene. For RFA3, which encodes a subunit of the Replication Protein A (RPA), a highly conserved single-stranded DNA binding protein involved in DNA replication, repair, and recombination (Brill and Stillman 1991), we only identified one dosage suppressor, RFA2, whose product forms a complex with Rfa3p (Gavin, Bosche et al. 2002; Dickson, Krasikova et al. 2009). For CDC48, which encodes the yeast ortholog of the mammalian ATPase p97 (Ye,

Figure 3.2.5

Figure 3.2.5 Continued

Figure 3.2.5 Empirical determination of raw barcode microarray intensity cutoff for identification of candidate dosage suppressors.

Three different ts strains were used in the determination: cdc48-9 (a.), stu1-5 (b.), and nse3-ts5 (c.). Bar graphs were generated to display the raw barcode microarray intensities. Only the barcodes with raw intensities >200 average fluorescence units (a.f.u.) are shown. Dotted lines are drawn at raw microarray intensities of 2000, 1000, and 500.

Meyer et al. 2003), we identified 24 dosage suppressors. These dosage suppressors were derived from numerous different pathways, and included UBX4, which encodes a UBX (ubiquitin regulatory X) domain-containing protein that interacts with Cdc48p (Decottignies, Evain et al. 2004), and RPL17A, which encodes a protein subunit of the large ribosome (Mager, Planta et al. 1997). In total, we found 137 novel dosage suppression interactions for 29 query genes. We identified novel interactions for 26 query genes that had three or fewer literature-curated dosage suppressors, including 13 query genes that had no previously known dosage suppressors (Appendix 6.2; Electronic Table 3.2.1). A significant portion of the novel dosage suppressors we identified are functionally related to their respective query genes, as 40% (p-value < 2.6 x 10-10) of the gene pairs have shared gold standard Gene Ontology (GO) terms (Myers, Barrett et al. 2006) (Appendix Table 6.2.2; Electronic Table 3.2.1), indicating that the dosage suppressors identified in our screens represent functionally relevant interactions.

As a complement to the barcode microarray analysis, we also used a next-generation sequencing method called Bar-seq (Smith, Heisler et al. 2009; Smith, Heisler et al. 2010) to measure the abundance of each barcode sequence present in the dosage suppressor pools. We employed a modified version of Bar-seq (Smith, Heisler et al. 2009), multiplexing 25 independent experiments at once, such that 75 independent dosage suppressor pools of varying complexity of barcode representation were analyzed in three lanes. Candidate dosage suppressors were identified as those whose barcode sequences exceeded 5% of the total sequencing reads/experiment and having greater than 500 raw sequencing counts for each dosage suppressor pool. Using spot dilutions, we attempted to confirm dosage suppressors that represented anywhere from 5% to >99% of the sequencing reads within a unique dosage suppressor pool. Approximately 37% of the dosage suppressors we identified were confirmed by both microarray and sequencing methods (Appendix 6.2; Electronic Table 3.2.1). Despite identifying fewer interactions, the Bar-seq method was more precise than the microarray-based approach. Specifically, using spot dilutions, 64% of the interactions identified by Bar-seq were confirmed as dosage suppressors compared to a 20% confirmation rate for interactions identified by barcode microarray.

3.2.3 An integrated dosage suppression genetic interaction network

Our screens for dosage suppressors of essential gene phenotypes with the MoBY-ORF 2.0 library identified mostly novel interactions, emphasizing the need for systematic analysis to generate a comprehensive view of the dosage suppression genetic network. Nonetheless, other dosage suppression interactions have been reported previously and thus to generate a global view of the current dosage suppression network, we combined our results with a set of 1503 dosage suppression genetic interactions that were annotated in the Saccharomyces Genome Database (Electronic Table 3.2.2) to map a network containing 1077 genes (including 437 essential query genes) and 1640 interactions (Figure 3.2.6a). As mentioned above, most query genes have only a few dosage suppressors, while a small set of genes has a large number of dosage suppression interactions. The network was visualized in Cytoscape using a forced-directed layout (Shannon, Markiel et al. 2003), such that genes that share common dosage suppression interactions formed distinct clusters. Markov clustering (MCL) analysis (van Dongen 2002) identified nine clusters, each containing 30 or more genes, that correspond to specific bioprocesses. Similar to the synthetic genetic network (Costanzo, Baryshnikova et al. 2010), the relative distance between these clusters appears to reflect shared functionality (Figure 3.2.6a). For example, the functional relationships between vesicle-mediated transport, exocytosis, and cell polarity and morphogenesis are illustrated by the relatively close proximity of their corresponding gene clusters to one another in the network, along with a significant number of dosage suppression interactions that occur between genes functioning in these different bioprocesses. This suggests that dosage suppression interactions, like other forms of genetic interactions (Dixon, Costanzo et al. 2009; Costanzo, Baryshnikova et al. 2010), can be used to independently cluster genes on the basis of functional interrelatedness.

A detailed look at specific interactions can provide new mechanistic insight into particular pathways and complexes. For example, we screened two components of the essential MIND (Mtw1p including Nnf1p-Nsl1p-Dsn1p) kinetochore complex (Figure 3.2b), which participates in bridging centromeric heterochromatin and kinetochore microtubule-associated proteins (MAPs) and motors (De Wulf, McAinsh et al. 2003; Pagliuca, Draviam et al. 2009). No dosage suppressors have been mapped previously for NSL1 and DSN1, but here we identified a network of 31 interactions and 28 genes impinging on these two query genes (Appendix 6.2;

Figure 3.2.6

Figure 3.2.6 Dosage suppression genetic interaction network for S. cerevisiae. a. Integrated network diagram of dosage suppression genetic interactions annotated in the literature and identified in this study. Genes are represented as nodes and interactions are represented as edges. Colored nodes indicate sets of genes enriched for GO biological processes summarized by the indicated terms. The nodes were distributed using a force-directed layout, such that genes (nodes) that share common dosage suppression interactions form distinct clusters. b. Dosage suppression provides new biological insight into functional relationships. Novel dosage suppressors were identified for two components of the MIND kinetochore complex, NSL1 and DSN1. Green: kinetochore function; blue: PKA signaling; red: ribosome biogenesis.

Electronic Table 3.2.1). Among the dosage suppressors identified for both NSL1 and DSN1 were the S-phase transcription factor HCM1 and a ribosomal biogenesis gene, FCF1. In a previous study, HCM1 was predicted to directly upregulate transcription of both genes (Pramila, Wu et al. 2006), and nsl1(ts) hcm1Δ and dsn1(ts) hcm1Δ double mutants each display a negative genetic interaction (Costanzo, Baryshnikova et al. 2010); therefore, our results are consistent with both the in silico predictions and the double mutant phenotypes. We also identified SPC105 as a DSN1 dosage suppressor. Spc105p forms a complex with Kre28p, and this complex also acts as an essential kinetochore linker complex.

Two redundant genes that downregulate PKA signaling at various steps in the pathway, GPB1 and GPB2, were also identified as dosage suppressors of DSN1 (Figure 3.2b). A third gene, GIS2, is not well characterized but is thought to act as a negative regulator, similar to PDE2, in PKA signaling (Balciunas and Ronne 1999). While a genetic link between the PKA pathway and the Dam1p-Duo1p (or DASH) kinetochore- and microtubule-associated complex has been previously observed (Li, Li et al. 2005), our results are the first to identify a possible functional relationship between downregulation of PKA signaling and the MIND complex. Thus, our results support the hypothesis that attenuating PKA signaling contributes to proper kinetochore function.

3.2.4 Distribution of dosage suppressors across cellular processes

We examined the occurrence of dosage suppression genetic interactions within and across different cellular processes. The heat map identified functions enriched (yellow) or depleted (blue) for dosage suppression interactions relative to the expected frequency of a random gene set (Figure 3.2.7a). Consistent with connectivity of other biological networks (Gavin, Bosche et al. 2002; Krogan, Cagney et al. 2006; Tarassov, Messier et al. 2008; Yu, Braun et al. 2008; Costanzo, Baryshnikova et al. 2010), we found that genes involved in the same biological process were highly enriched for dosage suppression interactions. Importantly, we also observed a significant number of dosage suppression interactions between distinct yet related bioprocesses (Figure 3.2.7a). For example, the growth defect of mutants compromised for cell polarity and morphogenesis pathways are suppressed by overexpression of genes involved in

Figure 3.2.7

Figure 3.2.7 Properties of the yeast dosage suppression network. a. Frequency of dosage suppression genetic interactions within and across biological processes for the integrated dosage suppression network. The frequency of gene pairs exhibiting dosage suppression interactions was measured for 19 broadly defined functional gene sets (Costanzo, Baryshnikova et al. 2010); blue: below the frequency of random pairs; black: statistically indistinguishable from a random set of gene pairs; yellow: above the frequency of random pairs. Dosage suppressor gene function is on the x-axis, and query ORF gene function is on the y-axis. The diagonal represents within-process interactions. The red line in the color scale indicates the frequency of interactions expected by chance (0.0005). b. Scaled square Venn diagram showing the fraction of dosage suppression gene pairs that also exhibit negative genetic and protein-protein interactions. Only gene pairs known to be tested for both genetic and physical interactions were considered. Light blue: gene pairs showing dosage suppression interactions only; red: gene pairs showing dosage suppression and negative genetic interactions; dark blue: gene pairs showing dosage suppression and physical interactions; yellow: gene pairs showing dosage suppression, negative genetic and physical interactions.

64 several different functional categories, including those that act at various steps in intracellular vesicle-mediated transport.

3.2.5 Overlap of dosage suppression interactions with protein-protein and negative genetic network edges

We explored the overlap of dosage suppression genetic interactions with three other interaction networks: the physical (protein-protein) (Gavin, Bosche et al. 2002; Krogan, Cagney et al. 2006; Tarassov, Messier et al. 2008; Yu, Braun et al. 2008), negative genetic, and positive genetic networks (Costanzo, Baryshnikova et al. 2010). We found that both our experimentally- derived and literature-curated dosage suppression networks were enriched significantly for both physical and negative genetic interactions but not positive genetic interactions (Table 3.2.1). Despite this overlap with physical and negative genetic interactions, most dosage suppression interactions (68%) in the integrated network did not overlap with any previously mapped network edge (Figure 3.2.7b; Appendix 6.2; Electronic Tables 3.2.1 and 3.2.2). Importantly, these unique dosage suppression interactions were enriched for co-annotated gene pairs (55%; p- value << 10-16). Thus, dosage suppression identifies a new type of interaction capable of covering novel and functionally relevant network space, one that has not been interrogated previously by the currently established interaction mapping approaches.

3.2.6 Mechanistic categorization of dosage suppression interactions

General mechanistic categories of second-site genetic suppression have been described previously (Prelich 1999; Hodgkin 2005). We have extended this analysis to dosage suppression genetic interactions by developing a decision tree to systematically categorize dosage suppression genetic interactions in yeast (Figure 3.2.8). with the remaining interactions (13%) falling into an unknown category (Figure 3.2.8; Table 3.2.2; Appendix 6.2; Electronic Tables 3.2.1 and 3.2.2)

First, a gene pair was determined to be functionally related if it was co-annotated to the same GO term within a gold standard set of terms (Myers, Barrett et al. 2006). Based on this

Table 3.2.1 Overlap of dosage suppression interactions with other types of interactions.

Dosage suppression Dosage suppression

(SGD) (this study) Type of # tested # overlap. # tested # overlap. p-value a p-value a interaction pairs pairs pairs pairs Physical 1503 b 525 << 10-16 150 27 << 10-16 Negative 254 c 59 << 10-16 57 c 8 1.03x10-5 genetic Positive 254 c 4 0.52 57 c 0 N/A genetic a p-values based on hypergeometric test. b Bait-hit gene pairs annotated in the Saccharomyces Genome Database as “Dosage Rescue” in which the bait is an essential gene. c Subset of gene pairs from b that have been screened for genetic interactions (Costanzo, Baryshnikova et al. 2010).

Table 3.2.2 Distribution of dosage suppression gene pairs annotated in the Saccharomyces Genome Database.

Category Number of Dosage Suppression Gene Pairs Functionally Related (No Protein-Protein Interaction) 732 a Functionally Related (With Protein-Protein Interaction) 515 a, b Chaperone 13 b RNA Processing/Protein Synthesis 69 Unknown 174 TOTAL 1503 a Gene pairs were determined to be functionally related if they shared a GO term found in the gold standard set of terms (Myers, Barrett et al. 2006) as annotated in the Saccharomyces Genome Database. b Protein-protein interactions as annotated in the Saccharomyces Genome Database.

Figure 3.2.8

68 standard and consistent with our previous results (Figures 3.2.6 and 3.2.7), we found that ~80% of all gene pairs with reported dosage suppression interactions were co-annotated to the same GO term (Table 3.2.2; Appendices 6.1 and 6.2; Electronic Tables 3.2.1 and 3.2.2), confirming that dosage suppression interactions can identify genes that participate in the same general biological pathway or process (Figure 3.2.9a and 3.2.9b).

Second, functionally related gene pairs that exhibit a physical interaction between their gene pairs were grouped into the “Complex Component” category, because the physical interaction between the gene products affects the activity of the complex (Figure 3.5b). For example, the G1 cyclins, CLN1 and CLN2, were initially discovered as dosage suppressors of cdc28-1, a ts allele of the essential gene encoding the cyclin-dependent kinase Cdc28p (Reed, Hadwiger et al. 1989). Subsequent studies showed that the cyclins bind to and activate Cdc28p (Richardson, Wittenberg et al. 1989; Tyers and Futcher 1993). A direct physical interaction between a mutant query gene product and its dosage suppressor may also reflect the ability of the dosage suppressor to stabilize a protein complex containing the query mutant protein. Based on this hypothesis, reciprocal dosage suppression (Gene A suppresses query mutant b and Gene B suppresses query mutant a) may be expected between two essential gene products belonging to the same protein complex. We identified 28 dosage suppression interactions in our screens in which both the query mutant and the dosage suppressor were essential (Table 3.2.3; Appendix 6.2.1). Eight of the 27 gene pairs exhibited reciprocal dosage suppression (Table 3.2.3), such that growth defects associated with mutations in either gene can be suppressed by overexpressing its partner. Interestingly, all eight gene pairs also shared a physical interaction among their gene products, which is highly unlikely to occur by chance (Table 3.2.3; p-value << 10-16). Similarly, 75% of reciprocal dosage suppression interactions reported in the literature also share a physical interaction (Appendix 6.2). Thus, the strong overlap between reciprocal dosage suppression and physical interactions provides evidence to support a mechanism whereby phenotypic suppression is mediated by increased protein complex stability.

In the absence of any physical interactions, however, functionally related genes can still exhibit dosage suppression interactions (Figure 3.2.9a). Mutations in SEC3, which encodes an essential member of the exocyst complex (Finger and Novick 1997) that transports secretory

Figure 3.2.9

Figure 3.2.9 Mechanisms of dosage suppression in yeast. a. Functional Relationship: A dosage suppressor can function either upstream or downstream of its respective mutant query allele in the same biological process. In the example shown, at the semi-permissive temperature, the function of the mutant allele b gene product is impaired and unable to transmit information to a downstream effector. A dosage suppressor, encoded by gene A, can act upstream of the mutant allele to activate the pathway. b. Complex Component: A dosage suppressor can be a gene that encodes an interacting protein of the mutant gene product which is required for its normal function. At the semi-permissive temperature, the mutant protein b predominantly occurs in an unfolded state, likely because the mutation renders the gene product unstable, and is therefore unable to interact with its normal physical partner(s). Overexpression of a dosage suppressor, protein A, increases the levels of properly folded mutant protein so that the physical complex can execute its essential function. c. Chaperone: A dosage suppressor can affect the amount of the mutant gene product. In the example shown, the dosage suppressor protein does not normally interact with the mutant gene product. At the semi-permissive temperature, the mutant protein b is unfolded, but overexpression of a dosage suppressor, such as a chaperone (protein A), can re-fold and stabilize the mutant protein, enabling it to carry out its essential function. d. RNA Processing/Ribosome: A dosage suppressor can be a gene that acts during transcription or translation. In the example show, the dosage suppressor protein A is normally involved in some aspect of transcription. At the semi-permissive temperature, transcription of mutant allele b leads to a poor quality mRNA product that may be translated but more likely will be degraded. By increasing some aspect of transcription, however, it might be possible to improve the quality of the mRNA product, which can then be translated instead of degraded, leading to enough functional mutant protein for the cell to be viable.

Table 3.2.3 Gene pairs tested for reciprocal suppression.

Dosage Suppressor Reciprocal Suppression Known Protein-Protein ts allele Gene Name Observed? Interaction? ame1-4 OKP1 No Yes apc11-13 APC2 Yes Yes ccl1-ts4 KIN28 Yes b Yes cdc11-1 YEF3 No No cdc11-4 a CDC3 No Yes cdc11-5 a CDC3 No Yes cdc11-5 CDC10 No Yes cdc14-1 CDC33 No No cdc24-H CDC24 Yes c Yes cdc48-3 NOP1 No No dsn1-7 SPC105 No Yes ipl1-1 SDS22 No No nse3-ts4 NSE1 No Yes nsl1-5 DSN1 No Yes orc2-3 CDC48 No No orc2-3 a ORC3 Yes Yes orc3-70 a ORC2 Yes Yes pol12-ts HYP2 No No scc4-4 SCC2 No Yes sec14-3 YPT1 No No sec17-1 SEC18 No Yes sec26-11D26 BET1 No No sec26-11D26 OLE1 No No sec26-11D26 SEC11 No No sec26-11D26 SEC22 Yes Yes stu1-5 NOP1 No No taf12-9 a TAF4 Yes Yes taf12-W486stop a TAF4 Yes Yes taf8-ts7 TAF10 Yes Yes taf9-ts2 TAF4 Yes Yes a CDC11-CDC3, ORC2-ORC3 and TAF12-TAF4 gene pairs are only counted once; however, we screened two different alleles of CDC11 and TAF12 and recovered CDC3 and TAF4 respectively with both alleles. b reported by (Valay, Simon et al. 1993) c reported by (Ziman and Johnson 1994; Richman, Sawyer et al. 1999; Barale, McCusker et al. 2006).

72 vesicles from the trans-Golgi network to the plasma membrane, are suppressed by overexpression of either of two functionally redundant plasma membrane t-SNAREs, SSO1 and SSO2, involved in secretory vesicle fusion to the plasma membrane (Aalto, Ronne et al. 1993). Dosage suppression interactions have also been identified for gene pairs that represent gene duplications (Tanaka, Nakafuku et al. 1990; Roberg, Crotwell et al. 1999; Angus-Hill, Schlichter et al. 2001; Norgaard, Westphal et al. 2001) and gene pairs that have the same molecular function (Matsui and Toh-e 1992; Arevalo-Rodriguez, Cardenas et al. 2000; Han, Audhya et al. 2002; Chloupkova, LeBard et al. 2003); in both cases, the biochemical activity of the dosage suppressor can functionally substitute for the mutant gene product (Rine 1991).

If a dosage suppression gene pair is not co-annotated to the same gold standard GO term, then a dosage suppressor may be characterized as a chaperone suppressor or an RNA processing/ribosome suppressor, or an unknown suppressor. Chaperones, such as heat shock proteins (HSPs) or RNA stability factors, can act as dosage suppressors by stabilizing the levels of a query gene product (Figure 3.2.9c). Indeed, increased dosage of HSPs has been shown to suppress diverse sets of genes which do not share any obvious functional relationship (Shirayama, Kawakami et al. 1993; Shea, Toyn et al. 1994; Kosodo, Imai et al. 2001; Orlowski, Machula et al. 2007). Genes encoding ribosomal subunits and RNA processing factors have also been identified as dosage suppressors for a variety of query genes. While the molecular mechanism of dosage suppression in these cases is not well understood, it is possible that the suppressing genes may lead to increased transcription or translation of the ts query gene product (Figure 3.2.9d). The ts mutation may specifically lead to transcriptional repression of RNA processing and/or ribosomal subunit genes. For example, transcriptional profiling of myo1Δ cells, which are deficient for the single yeast type II myosin heavy chain (Watts, Shiels et al. 1987), showed down-regulated expression of many genes, including several ribosomal subunit genes (Rodriguez-Quinones, Irizarry et al. 2008), and overexpression of some of these genes can rescue the associated mutant phenotypes (Diaz-Blanco and Rodriguez-Medina 2007; Rodriguez- Quinones, Irizarry et al. 2008).

Gene pairs that were not co-annotated to the same gold standard GO term, or considered a chaperone suppressor or an RNA processing/ribosome suppressor, were classified as having an

73 unknown dosage suppression mechanism. Interestingly, within the “unknown” category, we identified a small number of dosage suppression gene pairs whose protein products have a known physical interaction but the two genes do not share any gold standard GO terms in common (Appendix 6.2; Electronic Table 3.2.2). For every interaction initially placed in the unknown category, including gene pairs both with and without known physical interactions between their gene products, we went back to the primary literature in an effort to find other functional information that might support re-classification of a particular interaction into a mechanistic category. By doing so, we were able to re-classify 13% of the unknown interactions into one of the other mechanisms of dosage suppression. In total we found that a relatively small fraction (7%) of dosage suppression interactions fall into the chaperone and RNA processing/ribosome category. Using our dosage suppression decision tree, we classified the remaining ~87% of the dosage suppression interactions within the integrated network into one of four general mechanistic categories.

3.3 DISCUSSION

We report a simple and efficient method to clone dosage suppressors in yeast. Central to this method was the development of the high-copy MoBY-ORF 2.0 library which was derived from the low-copy MoBY-ORF 1.0 library using the MAGIC bacterial mating and recombination system for transfer of the ORF and its unique molecular barcodes (Li and Elledge 2005; Ho, Magtanong et al. 2009). Because each gene in the MoBY-ORF 2.0 library is under the control of its own regulatory sequences, it should be particularly useful for systematic dosage suppression analysis of genetic or chemical perturbations (Hoon, Smith et al. 2008) that compromise cellular fitness. Preserving the gene under its own regulation may reduce the potential for synthetic dosage lethality, which has been the focus of most of the previous studies of gene dosage (Liu, Krizek et al. 1992; Akada, Yamamoto et al. 1997; Stevenson, Kennedy et al. 2001; Boyer, Badis et al. 2004; Sopko, Huang et al. 2006) and can confound dosage suppression screens.

We validated the use of the MoBY-ORF 2.0 plasmid library in dosage suppression experiments using a barcode microarray and a deep sequencing readout (Bar-seq). The two

74 methods identified a largely overlapping set of candidate genes; however, each technique also identified several unique hits. This is not unexpected, due to differences in the dynamic range and sensitivity of the two techniques. The barcode microarray provided a larger list of candidate and greater number of total confirmed dosage suppressors than did Bar-seq; however, Bar-seq had a significantly higher precision (confirmation) rate. Barcodes that are present in a dosage suppressor pool but have mutations (either point or deletion/insertion) may fail to hybridize to a barcode microarray and therefore would lead to a false negative. However, barcodes carrying mutations could be identified by Bar-seq, as the analysis can be customized to allow for nucleotide mismatches in a particular barcode. Nevertheless, if a barcode is present at a relatively low amount in a dosage suppressor pool, it may not be above a given cutoff used in Bar-seq; however, it may still be identified on a barcode microarray due to the signal detection method, which allows for amplification of weak but true barcode PCR product hybridization. The difference in confirmation rates we observed may in part be attributed to how we harvested the candidate suppressor colonies, which involved washing the plates completely, and, therefore, may include some general background colonies.

Gene pairs exhibiting dosage suppression are highly enriched for physical interactions between the encoded proteins (Table 3.2.1). This observation supports the hypothesis that increased expression of wild-type genes can rescue mutations in a target gene because their functional association results directly from some sort of physical interaction between the respective encoded proteins (Reed, Hadwiger et al. 1989). Thus, direct interaction between a mutant query gene product and its dosage suppressor may serve to stabilize a complex containing the query mutant protein.

Dosage suppression genetic interactions were also enriched for negative genetic interactions (Table 3.2.1). A negative genetic interaction is defined as a double mutant fitness defect that is significantly stronger than expected, given the two single mutant fitness defects (Dixon, Costanzo et al. 2009). In a dosage suppression genetic interaction, the fitness of one single mutant is improved by overexpressing the wild-type copy of a second gene. Thus, the interacting gene can be viewed as behaving in opposite ways; it has decreased activity in tests for negative genetic interactions, and increased activity in an overexpression suppression test, which

75 is logically consistent because the query mutant phenotype is enhanced and suppressed, respectively.

In contrast, we do not see a significant enrichment of positive genetic interactions for dosage suppression gene pairs (Table 3.2.1). Positive genetic interactions occur when a double mutant shows a fitness defect that is less severe than expected based on the fitness of the corresponding single mutants (Mani, St Onge et al. 2008; Dixon, Costanzo et al. 2009). This means that positive genetic interactions would only be logically consistent if the loss-of-function allele and the overexpressed gene showed the same suppression phenotype. In rare cases, overexpressed genes can lead to a dominant-negative phenotype; however, large-scale analysis has shown that most dosage phenotypes can be attributed to a gain-of-function role associated with the overexpressed gene (Sopko, Huang et al. 2006), which supports the lack of overlap between these types of interactions.

Mechanistic classification of the dosage suppression interactions revealed that the vast majority of dosage suppression interactions (80%) occur between functionally related genes, of which ~60% are between gene pairs whose products have no known physical interactions. Thus, dosage suppression represents a functionally relevant yet unique type of genetic interaction.

Conditional alleles are being developed for the majority of essential genes in yeast (Ben- Aroya, Coombes et al. 2008) (Z. Li and C. Boone, unpublished data, and P. Heiter, personal communication), and thus the potential exists to map a dosage suppression genetic interaction network for the entire spectrum of essential genes. This mapping effort could extend to the majority of nonessential genes on the dosage suppression network by creating ts alleles of each gene within the context of a synthetic lethal background (Costanzo, Baryshnikova et al. 2010). With on the order of ~5 dosage suppression interactions per query gene, the dosage suppression network offers the potential of a wealth of new functional information and connections. While we demonstrate the utility of this type of genetic interaction in yeast, an analogous mapping of genetic interactions should be possible in mammalian cells and metazoan model systems. We conclude that a global dosage suppression map adds a highly prevalent and new type of

76 functional ‘edge’ that can be integrated into the construction of a complete cellular landscape comprising all types of genetic and physical interactions.

3.4 METHODS

3.4.1 Growth Media

Yeast strains were grown in SD-LEU (0.67% yeast nitrogen base, 0.2% amino acid supplement minus LEU, 2% glucose) or SC (0.67% yeast nitrogen base, 0.2% amino acid supplement, 2% glucose) medium. Bacteria were grown in 2X YT (1% yeast extract, 1.6% tryptone, 0.5% sodium chloride) or in YE (0.5% yeast extract, 1% NaCl).

3.4.2 Clone construction and analysis

MoBY-ORF, v1.0 bacterial strains (Ho, Magtanong et al. 2009) were inoculated from frozen stocks in 96-well plates into a shallow 96-well plate in which each well had 100 µl 2X YT containing tetracycline at 5 µg/ml, kanamycin at 50 µg/ml, and chloramphenicol at 12.5 µg/ml. Cultures were grown for ~16 hours at 37oC. P5530 (genotype: lacIQrrnB3 ΔlacZ4787 hsdR514 Δ(araBAD)567 Δ(rhaBAD)568 galU95 ΔendA9:FRT ΔrecA635:FRT umuC:ParaBAD-I-SceI- FRT), the MAGIC recipient strain carrying plasmid p5476, was inoculated into 5 ml of YE+Gluc, 0.2% glucose, spectinomycin at 10 µg/ml, and carbenicillin at 200 µg/ml. Cultures were grown for ~22 hours at 30oC.

The following day, the OD600 values of the recipient strain and of 3 bacterial (donor) strains from the MoBY-ORF 96-well plate were taken; the average of the 3 wells was used as the average OD600 for the entire plate. Cultures were diluted to OD600 ~0.10 and mixed together for mating in a 1:1 ratio in a total volume of 100 µl in a fresh 96-well plate. Cells were shaken at 30oC for 2 hours, at which time L-arabinose was added to a final concentration of 0.2% to each well. Cells were incubated without shaking at 37oC for 2 hours, and then transferred to a shaking incubator at 37oC for 2 hours. 2 µl of the 100 µl mating reaction were plated onto YE+Glyc

77 containing 0.2% glycerol, 0.2% DL-chlorophenylalanine, carbenicillin at 200 µg/ml, and kanamycin at 50 µg/ml, and incubated at 41oC overnight.

Mating products were streaked out for individual colonies onto 2X YT containing 0.2% glucose, carbenicillin at 200 µg/ml, and kanamycin at 50 µg/ml and incubated at 37oC overnight. Miniprep DNA was prepared from a single bacterial colony, doubly digested using XhoI and EcoRI (Fermentas), and resolved on a 0.8% agarose gel to confirm vector and insert fragment sizes. Two individuals performed gel analysis independently, and the results were compared to determine clone validity. If the primary XhoI/EcoRI digest was ambiguous, a secondary digest using BamHI/HindIII was performed. ORF fragment sizes for both double digests, along with complete ORF sequence and barcodes, can be obtained from the MoBY-ORF database: (http://moby.ccbr.utoronto.ca).

3.4.3 Plasmid pool preparation

Individual E. coli transformants containing a barcoded high-copy plasmid were grown in 100 µl of 2X YT containing glucose (0.2%), carbenicillin (200 µg/ml) and kanamycin (50 µg/ml) at 37°C for 15 hours in a shallow 96-well plate. 55 µl of each culture was mixed to form the E. coli MoBY-ORF version 2.0 pool. Plasmid DNA was prepared from the E. coli pool.

3.4.4 Cloning of dosage suppressors with the 2µ MoBY-ORF library

Each temperature-sensitive query strain (Table 3.2.4) was transformed with MoBY-ORF v2.0; ≥ 50,000 transformants were pooled and was frozen in 15% glycerol. For identification of suppressors, a sample of the transformant pool was thawed, and 50,000 cells were plated onto SD-LEU. The incubation temperatures for each strain were dependent on the observed restrictive temperature for the untransformed temperature-sensitive mutant (Z. Li, unpublished observations). For a given strain at a particular temperature, colonies that appeared after 3 days were pooled (to form the “dosage suppressor pool”) and stored at -80oC in 15% glycerol. To isolate suppressing plasmids, a sample of the dosage suppressor pool was thawed, and plasmids

Table 3.2.4 Yeast strains used in this study. Yeast Strain Genotype Y5041 MATa cdc42-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y5300 MATa cdc24-H::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y5361 MATa sec17-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y5583 MATa sec14-3::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y5587 MATa sec18-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y5588 MATa sec18-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6270 MATa sec19-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6271 MATa scc2-4::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6323 MATa smc3-42::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6415 MATa ccl1-ts4::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6417 MATa cdc10-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6426 MATa cdc11-2::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6432 MATa cdc11-5::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6434 MATa cdc23-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6437 MATa cdc23-4::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6462 MATa kin28-ts::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6494 MATa sec22-3::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6525 MATa cdc11-4::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y6538 MATa nop1-3::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7236 MATa cdc3-3::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7354 MATa cdc14-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7425 MATa ipl1-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7488 MATa cdc28-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7731 MATa cdc48-9::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7827 MATa cdc48-3::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7829 MATa cdc48-2::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y7912 MATa okp1-5::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8076 MATa apc2-8::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8183 MATa stu1-5::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8198 MATa cep3-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8224 MATa orc2-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8270 MATa cdc33-E72G::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8309 MATa cdc36-16::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8324 MATa sec11-2::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8437 MATa orc2-3::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8846 MATa orc3-70::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y8950 MATa pol12-ts::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9020 MATa cdc11-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ

Table 3.2.4 Continued Yeast Strain Genotype Y9313 MATa dsn1-7::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9315 MATa dsn1-7::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9320 MATa dsn1-8::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9423 MATa spc105-15::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9452 MATa prp6-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9495 MATa arc35-5::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9587 MATa prp4-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9588 MATa prp4-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9606 MATa ole1-m2::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9831 MATa taf12-9::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9834 MATa taf12-W486stop::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9854 MATa smc2-8::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9918 MATa rpn11-14::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y9987 MATa stu2-11::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10087 MATa sds22-5::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10121 MATa yef3-F650S::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10137 MATa ypt1-3::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10166 MATa tel2-7::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10247 MATa arp4-G161D::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10306 MATa taf4-18::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10626 MATa rfa3-313::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y10803 MATa taf8-ts7::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11140 MATa ame1-4::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11188 MATa sec26-11D26::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11213 MATa taf9-ts2::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11232 MATa hyp2-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11237 MATa med4-6::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11357 MATa tcp1-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11359 MATa apc11-13::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11386 MATa bet1-1::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11413 MATa taf10-ts34::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11540 MATa nsl1-5::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11745 MATa scc4-4::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11860 MATa nse3-ts4::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y11912 MATa nse5-ts2::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ Y12371 MATa nse1-16::KanR ura3Δ0 leu2Δ0 his3Δ met15Δ

80 were isolated using a modified miniprep protocol of the Qiagen miniprep kit as previously described (Butcher and Schreiber 2006).

3.4.5 Yeast barcode microarray hybridization and data analysis

PCR amplification of the barcodes and TAG4 microarray hybridization were performed as previously described (Pierce, Fung et al. 2006). For each array, a competitive hybridization was performed. Biotinylated universal TAG4 primers were used to PCR-amplify the barcodes from the dosage suppressor pool, while non-biotinylated universal TAG4 primers were used to PCR-amplify the barcodes from the original transformant pool. Each hybridization mix contained 9:1 (v/v) non-biotinylated:biotinylated PCR product. A signal 5-fold or higher above background, determined empirically, was used as the cutoff for identifying candidate dosage suppressors.

3.4.6 Empirical determination of raw barcode microarray intensity cutoff for identification of candidate dosage suppressors.

50,000 cells, representing a pooled sample of the transformants of each strain were plated on selected medium and incubated at the semi-permissive temperature. Colonies appearing after 3 days at the semi-permissive temperature were individually picked and pooled. We anticipated that the plasmid pool contained in this mixture would identify a distinct set of barcodes that display raw intensity signals at a level above the standard barcode microarray background cutoff (200 average fluorescence units (a.f.u.)) and thereby identify a cutoff for candidate dosage suppressors. Barcoded plasmids were extracted from the pooled colonies, the barcodes were PCR-amplified, and a competitive microarray hybridization was performed as described. a. For cdc48-9, 24 barcodes were above the standard background. Using a cutoff of 2000, 14/24 barcodes (58%) of the barcodes in the sample are retrieved. Using a cutoff of 1000, 19/24 barcodes (80%) of the barcodes in the sample are retrieved. Using a cutoff of 500, 21/24 barcodes (88%) of the barcodes in the sample are retrieved. b. For nse3-ts5, 6 barcodes were above the standard background.

Using a cutoff of 2000, 3/6 barcodes (50%) of the barcodes in the sample are retrieved. Using a cutoff of 1000, 4/6 barcodes (66%) of the barcodes in the sample are retrieved. Using a cutoff of 500, 4/6 barcodes (66%) of the barcodes in the sample are retrieved c. For stu2-11, 10 barcodes were above the standard background. Using a cutoff of 2000, 4/10 barcodes (40%) of the barcodes in the sample are retrieved. Using a cutoff of 1000, 5/10 barcodes (50%) of the barcodes in the sample are retrieved. Using a cutoff of 500, 5/10 barcodes (50%) of the barcodes in the sample are retrieved.

Because for two of the three strains, there was no difference in the number of barcodes retrieved between the 500 and 1000 raw intensity cutoffs, and for one strain, there is a marginal (<10%) difference, we decided to use the more stringent cutoff of 1000 raw a.f.u. for identification of candidate dosage suppressors.

3.4.7 Assessing fitness of barcoded yeast strains by Illumina/Solexa sequencing

Each 20mer uptag and barcode was amplified with composite primers comprising the sequences of the common barcode primers and the sequences required for attachment to the Illumina/Solexa slide.

For the Uptags the following primers were used: 5’- AATGATACGGCGACCACCGACACTCTTTCCCTACACGACGCTCTTCCGATCTNN NNNGTCGACCTGCAGCGTACG -3’ (Forward) and 5’- CAAGCAGAAGACGGCATACGAGCTCTTCCGATCTGATGTCCACGAGGTCTCT - 3’ (Reverse).

For the Downtags the following primers were used: 5’- AATGATACGGCGACCACCGACACTCTTTCCCTACACGACGCTCTTCCGATCTNN NNNCGGTGTCGGTCTCGTAG -3’ (Forward) and

5’- CAAGCAGAAGACGGCATACGAGCTCTTCCGATCTGAAAACGAGCTCGAATTCAT CG -3’ (Reverse).

The 5’ portion (in Bold) is the sequences incorporated into the F and R primer, respectively which are required for Illumina/Solexa cluster formation. The variable sequence (italics) represents the 10mer indexing tag used for multiplexing. The 3’ portion (underlined) represents the common primer flanking the barcode and is required to amplify the MoBY-ORF barcodes. PCR amplification was conducted in 100µL volumes, using Invitrogen Platinum PCR Supermix (Cat. No. 11306-016) with the following conditions: 95°C/3 min; 25 cycles of 94°C/30 sec, 55°C/30 sec, 68°C/30 sec; followed by 68°C/10 min. PCR product was then purified with Qiagen MinEluteTM 96 UF PCR Purification Kit (Cat. No. 28051). Following PCR purification, DNA was quantified with the Invitrogen Quant-iTTM dsDNA BR Assay Kit (Cat No. Q32853) and then adjusted to a concentration of 10µg/mL. Equal volumes of normalized DNAs were then pooled. Theses pool DNA samples (25-plex) consist of 130bp PCR products that was gel purified from 12% polyacrylamide TBE gels using the crush and soak method(Sambrook, Russell et al. 2001) followed by ethanol precipitation. Samples were used directly for cluster formation. Each lane was sequenced once, using the standard single-read sequencing primer. The sequence was as follows: multiplexing tag – common primer (U1 or D1) – MoBY-ORF barcode. The first 5 bases represent the multiplexing tag allowed post-sequencing assignment of each amplicon to a particular experiment. The identity of each clusters multiplexing tag was determined allowing 0 mismatches. The last 20 bases were used to identify which ORFs were potential dosage suppressors. Clusters were binned according to their multiplexing tag, than tallied using the MoBY-ORF barcodes. These tallies were transformed into a percentage of total counts for that experimental bin. Within each 25-plex, we removed the ORF barcodes that corresponded to wild-type alleles of other temperature sensitive mutants that were screened. A potential dosage suppressor was defined by greater than 5% of the Bar-seq counts for a particular experimental bin.

3.4.8 Confirmation of candidate dosage suppressors and test for reciprocal suppression

Individual 2µ MoBY-ORF plasmids were transformed into the corresponding temperature-sensitive strain using standard methods. Spot dilutions were performed on SD-LEU media using standard methods and incubated at the same temperature at which the dosage suppressor was initially identified (for candidate dosage suppressors) or the corresponding semi- permissive temperature of the temperature-sensitive strain (for reciprocal suppression tests).

3.4.9 Overlap of dosage suppression genetic interactions with other types of interactions

The literature-curated dosage suppression dataset and protein-protein interaction dataset were downloaded from the Saccharomyces Genome Database (SGD, www.yeastgenome.org) on May 20th, 2010. Dosage suppression data were filtered to include only gene pairs containing an essential gene as the query ORF. The list of essential genes was downloaded from Saccharomyces Genome Deletion Project (www- sequence.stanford.edu/group/yeast_deletion_project) on March 20th, 2010.

3.4.10 Analysis of functional relatedness

Two genes sharing a dosage suppression interaction were considered to be functionally related if they are co-annotated to the same Gene Ontology (GO) term. Only GO terms from a published gold standard were considered (Myers, Barrett et al. 2006). Gene Ontology annotations were downloaded from the Saccharomyces Genome Database (SGD, www.yeastgenome.org) on May 11th, 2010.

3.4.11 Identifying gene clusters in the integrated dosage suppression network

The integrated network was clustered using the Markov Clustering algorithm (van Dongen 2002). Nine clusters containing more than 30 genes were tested for functional enrichment using the BiNGO plugin for Cytoscape (Maere, Heymans et al. 2005). The Gene Ontology Biological Process term showing the highest enrichment in a particular cluster was used to label the cluster on the network.

Chapter Four Conclusions and Future Directions

84 85

4.1 General Overview

The goal of this thesis was two-fold: first, to construct the first barcoded library of gene expression and overexpression plasmids covering every gene in the Saccharomyces cerevisiae genome; and, second, to use this library to identify genes that, when overexpressed from a multicopy plasmid, suppress the lethality of temperature-sensitive alleles of essential genes. This type of genetic interaction is called dosage suppression. Gene overexpression generally leads to a gain-of-function effect (Sopko, Huang et al. 2006; Vavouri, Semple et al. 2009); therefore, dosage suppression genetic interactions can provide information that is difficult or indeed impossible to obtain using loss-of-function alleles alone. One important reason for pursuing dosage suppression interaction studies is that gain-of-function-based genetic interactions are likely to provide a rich source of functional information (Dixon, Costanzo et al. 2009) because these interactions are largely unique and rich in functionally coherent links. Moreover, the global structure of dosage suppression genetic interactions, which should reveal a functional wiring diagram of the cell highlighting functional relationships between different processes, remains largely unexplored. A second important reason is that the mechanistic basis of individual dosage suppression interactions is still poorly understood; therefore, a systematic survey of dosage suppression interactions may help reveal different general mechanisms of dosage suppression. In this thesis, for the first time, I was able to explore the relationship between functional edges in the dosage suppression network and those in the existing S. cerevisiae genetic and physical interaction networks. As an extension of this analysis, I also mined the S. cerevisiae literature to identify four general mechanisms of dosage suppression. Together, this thesis described the first comprehensive analysis of dosage suppression in any organism.

4.2 The MoBY-ORF gene overexpression libraries: present and future applications

Although it is possible to envision engineering the genome to contain multiple, integrated copies of a wild-type gene, the simplest experimental approach to studying dosage suppression is to use plasmid-based gene overexpression. One necessary reagent for investigating dosage suppression genetic interactions is a gene overexpression library. Over the years, several such libraries have been developed for yeast. However, as I previously described (see Introduction,

Section 1.3.2.7), each of these libraries has certain features that are not optimal for studying dosage suppression in a systematic and genome-wide manner. We reasoned that the ideal plasmid-based gene overexpression library would have the following features: 1) endogenous regulatory sequences controlling ORF expression; 2) no epitope tag; 3) a single ORF per plasmid; and, 4) unique barcodes to facilitate parallel analysis of strain pools. In this thesis, I described my role in the cloning and application of two related gene overexpression libraries in which each plasmid has all four of these features. The MoBY-ORF 1.0 and 2.0 plasmid libraries are respectively low-copy (CEN-based) and high-copy (2µ-based) gene overexpression libraries that can be used in a variety of dosage analysis studies. In this thesis, I described the first systematic approach specifically developed to studying dosage suppression in yeast. The MoBY- ORF 2.0 plasmid library is central to this method because it streamlines dosage analysis by being compatible with high-throughput genomics technologies that can monitor plasmid representation, including barcode microarrays and next-generation sequencing methods (Pierce, Davis et al. 2007; Smith, Heisler et al. 2009).

The MoBY-ORF 2.0 plasmid library is a versatile reagent that can be applied in systematic gene dosage studies to identify novel enhancers (dosage lethality) or suppressors (dosage suppression) of a phenotype of interest. The library can be used to identify genetic interactions occurring in either a conditional (e.g. temperature-sensitive) or constitutive (e.g. deletion) mutant strain. One limitation of the library is that, alone, it may not be able to identify dosage lethal interactions in certain genetic contexts. For example, if a dosage lethal interaction occurs in a deletion or some other type of constitutive mutant, then the transformed mutant will not be identified because the cell will be dead or arrested. Such an interaction could be detected, however, if the query mutant strain is first transformed with the wild-type complementing ORF carried on a plasmid containing a counterselectable marker, such as URA3, before mass transformation with the MoBY-ORF 2.0 plasmid library. After the mass transformation, replica plating the transformants onto media containing a drug used for counterselection, which in the case of URA3 is 5-fluoroorotic acid (5’-FOA), will allow for growth of cells that have lost the URA3-containing plasmid. However, if a dosage lethal interaction does occur, then the colony will not appear on the 5’-FOA-containing media, and this could be detected using a barcode readout. This strategy has successfully been used in an analogous manner in small-scale studies

87 to identify functionally relevant bypass suppressors of essential genes (Li, Routt et al. 2000; Marcoux, Cloutier et al. 2000; Kurischko, Weiss et al. 2005).

In the short term, an obvious future goal is to elaborate upon the dosage suppression network that we have begun to construct (see Chapter 3, Results, and below). For the majority of essential genes in yeast, ts alleles in a defined genetic background have been constructed (Ben- Aroya, Coombes et al. 2008)(Z. Li and C. Boone, unpublished data). Therefore, it should now be feasible to apply the systematic screening method I developed to construct an extensive network of dosage suppression interactions. It should also be possible to systematically investigate dosage suppression of single deletion mutants of non-essential genes that have quantifiable fitness defects (Costanzo, Baryshnikova et al. 2010); upon transformation with the MoBY-ORF 2.0 plasmid library, a strain carrying such an allele may have improved fitness that is a result of a dosage suppression genetic interaction. Interestingly, one mechanism of dosage suppression of a deletion allele is by hyperactivation of a redundant, parallel pathway. By virtue of its definition, essential pathways are not suppressed by overexpression of a parallel pathway. The use of the MoBY-ORF library may therefore allow for existence of such parallel, compensatory pathways to be investigated.

Another envisioned application of the existing MoBY-ORF 2.0 plasmid library is in higher-order genetics studies. Triple loss-of-function mutant analysis has been examined for specific genes on a small-scale (Lam, Krogh et al. 2008; Nugent, Johnsson et al. 2010) and a handful of genome-wide synthetic lethal screens have been mapped using a double mutant query strain (Tong, Lesage et al. 2004), but no large-scale systematic analysis has been reported. Such mutant analysis can highlight both overlapping and unique roles for individual genes. For example, dosage suppression genetic interactions between a query gene and two paralogous genes have been reported (Drebot, Johnston et al. 1993; Baudin-Baillieu, Tollervey et al. 1997; Helliwell, Schmidt et al. 1998; Kota, Melin-Larsson et al. 2007; Demmel, Beck et al. 2008). If no double mutant phenotype is observed between the query gene and a loss-of-function allele of only one of the two paralogs, this observation suggests the other paralog may be compensating for the loss-of-function. To determine if this compensation may be occurring, one could create a triple mutant comprising the mutant query gene and loss-of-function alleles of the two paralogs.

If a triple genetic interaction is observed, as determined by significant deviation from the predicted product model phenotype, then another functional edge representing an interaction between three different genes can be added to the overall interaction network of the cell. In another example, one could look at dosage suppression of a double deletion mutant. Double deletion mutants that exhibit a negative genetic interaction usually identify functional relationships occurring between redundant pathways that effectively buffer one another in the event that the activity of one pathway is compromised (Costanzo, Baryshnikova et al. 2010). By identifying dosage suppression interactions of synthetic lethal double mutants, perhaps with a ts allele of one gene, it might be possible to identify pathway-specific activities. As well, the novel activities of other previously known genes and/or pathways contributing to the double mutant genetic interaction might also be illuminated.

Analogous to the systematic analysis reported for genes that, under standard laboratory conditions, cause lethality upon overexpression in yeast (Sopko, Huang et al. 2006), the MoBY- ORF 2.0 plasmid library could be used to identify genes that either increase or decrease tolerance of certain environmental conditions, such as osmotic stress or pH change. For example, knowledge of overexpressed genes that increase tolerance to salt stress and ethanol has industrial applications in crop production and fermentation processes respectively (Hong, Lee et al. ; Hou, Cao et al. 2009; Hong, Lee et al. 2010; Hong, Lee et al. 2010; Sun, Guo et al. 2010). By definition, an inducible gene overexpression system requires an external molecule to achieve increased gene dosage. When looking at phenotypes under a certain environmental condition, using an inducer may effectively alter that condition in an unknown way and confound the experimental results. By contrast, no external requirements for gene overexpression are necessary with MoBY-ORF plasmids, making it more likely that appropriate context-dependent dosage effects will be identified.

4.3 Dosage suppression genetic interaction networks: illuminating a new facet of genetics

In comparison to the extensive genetic and physical interaction networks that have been determined for yeast and other organisms (Tong, Evangelista et al. 2001; Gavin, Bosche et al. 2002; Krogan, Cagney et al. 2006; Tarassov, Messier et al. 2008; Yu, Braun et al. 2008;

Costanzo, Baryshnikova et al. 2010), the existing dosage suppression genetic interaction network for yeast is quite small and for other organisms is largely non-existent. The limiting factor was that, before this work, no approach existed that enabled dosage suppression genetic interactions to be studied in a systematic manner. What distinguishes our protocol from previous dosage suppression studies is the method used in identifying candidate dosage suppressors from a dosage suppression screen, which is more rapid and efficient than traditional dosage suppression screens at identifying candidate dosage suppressors. Furthermore, unlike the initial results from screens using certain types of gene overexpression libraries (specifically, those in which a given plasmid contains more than one ORF), no ambiguity exists as to which ORF is the candidate dosage suppressor. In our proof-of-principle set of dosage suppression screens, we used barcode microarrays to identify candidate dosage suppressors for a selected set of query ts strains. We confirmed 214 dosage suppression interactions for ~50 ts strains queried. After accounting for multiple alleles screened and strains for which only the wild-type clone was recovered, we ultimately identified 150 dosage suppression interactions for 32 query genes. Our dosage suppression genetic interaction network is the first of its kind for any eukaryote.

Our dosage suppression network allowed us to explore the relationship between the dosage suppression genetic interactions and other known genetic and physical interactions. Using our experimental network, we examined the overlap of negative genetic, positive genetic and protein-protein interactions with dosage suppression interactions. We found that gene pairs exhibiting dosage suppression are enriched for negative genetic interactions, as well as for physical interactions between the encoded proteins. This was expected, as (1) essential genes encoding components of protein complexes are known to have numerous negative genetic interactions with other components of the same complex (Davierwala, Haynes et al. 2005) ; and (2) previous small-scale studies have suggested that physical stabilization of protein structure is a plausible mechanism of dosage suppression (Reed, Hadwiger et al. 1989). On the other hand, no significant overlap with positive genetic interactions is observed. This is likely due to the fact that we only analyzed dosage suppression interactions in which the queries were essential genes, as previous studies have reported that positive genetic interactions are not generally observed for genes encoding proteins found in the same complex when at least one member of the complex is essential (Bandyopadhyay, Kelley et al. 2008) (Baryshnikova et al., in press). Interestingly, many

90 gene pairs that exhibit dosage suppression interactions have not been identified as having a functional interaction in either of these more expansive genetic or physical interaction networks. Therefore, dosage suppression genetic interactions represent a unique type of functional edge that can be used in developing a complete network of all interactions occurring within the cell.

Looking ahead, we can envision a complete map of dosage suppression interactions for all essential and non-essential genes in S. cerevisiae. As noted above, this information is likely to provide a unique insight into the structure of the S. cerevisiae genetic network that will be of exceptional value on its own. More speculative questions include, for example, the conservation of this network in other species. It should be possible to develop similar tools for other single- celled organisms, and it will be interesting to determine whether these interactions are conserved in other species such as S. pombe. The impacts of gene dosage on multi-cellular animals, where cellular-level redundancy can compensate for the loss of individual cells, will be interesting to explore, as will issues related to cell non-autonomous processes such as inter-cellular communication and tissue function. Indeed, a better global understanding of increased gene dosage could have implications for diseases characterized by gene copy gains, such as Down’s syndrome or various types of cancer (Santarius, Shipley et al. 2010). A complete S. cerevisiae gene dosage interaction map will be an important starting point for these studies.

4.4 Understanding the mechanistic basis of dosage suppression

What are the underlying mechanisms of dosage suppression? To empirically determine how dosage suppression occurs, we mined the S. cerevisiae literature for dosage suppression interactions involving essential gene queries to determine what the biological relationship, if any, was between a dosage suppressor gene and its respective query gene. We found that, analogous to mechanisms described for second-site suppression, dosage suppression interactions involving essential gene queries fall into a small number of functional categories. As expected, many (>80%) dosage suppressors are genes that have some sort of functional relationship with the query gene. This functional relationship may be some sort of physical interaction that is required for activation of a protein complex or subcellular localization to a particular site of action. However, no physical interaction is required, as dosage suppression interactions have been

91 reported for gene pairs that act at distinct steps within a more general biological process. For those dosage suppression gene pairs that do not have an annotated functional relationship, the types of dosage suppressors typically fall into one of two categories: 1) an RNA processing/ribosomal gene; or 2) a gene with chaperone activity which stabilizes the RNA or protein product. As more and more functional relationships between genes and their products are identified, and more types of query alleles, such as deletion and environmental condition- dependent alleles, it is very possible that other mechanisms of dosage suppression will be determined.

Transcription factors that are functionally unrelated to their respective query genes have been identified as dosage suppressors. Upon transcription factor overexpression, the assumption is that the expression of its targets, which are either normally not expressed under the tested conditions or expressed and functioning at basal levels, is increased, leading to a physiological change that results in viability of a cell that would normally be arrested (or dead) in the semi- permissive condition. Performing gene expression analysis, using either microarrays or deep sequencing methods, on a mutant strain that is overexpressing a transcription factor dosage suppressor may provide insight into the biological changes going on in such a cell as well as possibly highlight previously unknown genetic relationships.

In our experimental network, we identified several ORFs that suppress seemingly functionally unrelated query genes. It is possible that the query and dosage suppressor genes are indeed functionally related, but an unexplored aspect of dosage suppression is allele-specific, gene non-specific suppression. Such suppression is known for second-site suppression; for example, amber mutations are suppressed by the corresponding tRNA mutations (Murgola 1985). No equivalent mechanism has been described for dosage suppression. Identification of such suppressors requires knowing the nature of the query mutation, be it a point mutation, a base pair insertion or deletion, or some other type of genomic alteration. Knowledge of such dosage suppressors may allow for refinement of the dosage suppression genetic interaction network, as well as, perhaps, the assignment of new functional information to genes that behave as dosage suppressors exclusively by this mechanism.

Previous reports have shown that heat shock induces a transient, albeit significant, repression of protein biosynthesis genes (Warner 1999; Gasch, Spellman et al. 2000). Therefore, protein biosynthesis genes are expected to represent a general class of dosage suppressors in screens that rely on heat shock to induce a conditional phenotype. In our screens, we used strains harboring temperature-sensitive query alleles that cause mutant phenotypes when incubated at a higher (semi-permissive) temperature. Not surprisingly, approximately one-third of the novel interactions we identified involved a dosage suppressor that encoded a gene that is (or is predicted to be) involved protein biosynthesis. The common assumption is that dosage suppression by overexpression of a protein biosynthesis gene is likely non-specific; therefore, we would expect that a given protein biosynthesis gene would suppress a variety of temperature- sensitive alleles of unrelated genes. In fact, we did observe a few such interactions: specifically, two dosage suppressors each interacted with three different query genes. These dosage suppressors could be “frequent flyers”, or non-specific dosage suppressors. However, (1) no definition currently exists that would allow for the confident identification of a “frequent flyer” (i.e. 50%, 75%, 95% of all screens – the criteria are unclear), (2) in most interactions, dosage suppression of a given query gene was only observed with one or two protein biosynthesis genes and, (3) it is likely that some dosage suppression interactions involving protein biosynthesis genes have functional relevance, as non-canonical (e.g. extraribosomal) functions for ribosomal genes has been reported (Warner and McIntosh 2009), and functional specialization of ribosomal paralogs has been observed (Haarer, Viggiano et al. 2007; Komili, Farny et al. 2007). Expanding dosage suppression screens to the rest of the genome will allow for the more accurate identification and analysis of non-specific dosage suppressors. As well, detailed functional studies may also reveal additional novel roles for protein biosynthesis genes in other processes.

4.5 Concluding thoughts

Dosage suppression is an underappreciated type of genetic interaction. The MoBY-ORF 2.0 plasmid library enables the systematic and streamlined study of gene dosage effects, including dosage suppression, in unprecedented detail and resolution. Future developments should enable the complete mapping of dosage suppression interactions in S. cerevisiae as well as other organisms. Investigating the genetic networks underlying improved fitness upon gene

93 overexpression may shed light on mechanisms of certain human diseases. Finally, the integration of dosage suppression genetic interactions into other types of functional networks will improve our overall understanding of the functional wiring diagram of the cell and contribute to discovering the function of every gene encoded in an organism’s genome.

Chapter Five References

94 95

Aalto, M. K., H. Ronne and S. Keranen (1993). "Yeast syntaxins Sso1p and Sso2p belong to a family of related membrane proteins that function in vesicular transport." EMBO J 12(11): 4095-104. Akada, R., J. Yamamoto and I. Yamashita (1997). "Screening and identification of yeast sequences that cause growth inhibition when overexpressed." Mol Gen Genet 254(3): 267-74. Angus-Hill, M. L., A. Schlichter, D. Roberts, H. Erdjument-Bromage, P. Tempst and B. R. Cairns (2001). "A Rsc3/Rsc30 zinc cluster dimer reveals novel roles for the chromatin remodeler RSC in gene expression and cell cycle control." Mol Cell 7(4): 741-51. Arevalo-Rodriguez, M., M. E. Cardenas, X. Wu, S. D. Hanes and J. Heitman (2000). "Cyclophilin A and Ess1 interact with and regulate silencing by the Sin3-Rpd3 histone deacetylase." EMBO J 19(14): 3739-49. Ashburner, M. (1982). "The genetics of a small autosomal region of Drosophila melanogaster containing the structural gene for alcohol dehydrogenase. III. Hypomorphic and hypermorphic mutations affecting the expression of hairless." Genetics 101(3-4): 447-59. Baetz, K. K., N. J. Krogan, A. Emili, J. Greenblatt and P. Hieter (2004). "The ctf13-30/CTF13 genomic haploinsufficiency modifier screen identifies the yeast chromatin remodeling complex RSC, which is required for the establishment of sister chromatid cohesion." Mol Cell Biol 24(3): 1232-44. Balciunas, D. and H. Ronne (1999). "Yeast genes GIS1-4: multicopy suppressors of the Gal- phenotype of snf1 mig1 srb8/10/11 cells." Mol Gen Genet 262(4-5): 589-99. Bandyopadhyay, S., R. Kelley, N. J. Krogan and T. Ideker (2008). "Functional maps of protein complexes from quantitative genetic interaction data." PLoS Comput Biol 4(4): e1000065. Barale, S., D. McCusker and R. A. Arkowitz (2006). "Cdc42p GDP/GTP cycling is necessary for efficient cell fusion during yeast mating." Mol Biol Cell 17(6): 2824-38. Bateson, W. (1909). Mendel's Principles of Heredity. Cambridge, Cambridge University Press. Baudin-Baillieu, A., D. Tollervey, C. Cullin and F. Lacroute (1997). "Functional analysis of Rrp7p, an essential yeast protein involved in pre-rRNA processing and ribosome assembly." Mol Cell Biol 17(9): 5023-32.

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(2): 248-58. Bisson, L. F. and J. Thorner (1982). "Mutations in the pho80 gene confer permeability to 5'- mononucleotides in Saccharomyces cerevisiae." Genetics 102(3): 341-59. Boone, C., H. Bussey and B. J. Andrews (2007). "Exploring genetic interactions and networks with yeast." Nat Rev Genet 8(6): 437-49. Boutros, M., A. A. Kiger, S. Armknecht, K. Kerr, M. Hild, B. Koch, et al. (2004). "Genome- wide RNAi analysis of growth and viability in Drosophila cells." Science 303(5659): 832-5. Boyer, J., G. Badis, C. Fairhead, E. Talla, F. Hantraye, E. Fabre, et al. (2004). "Large-scale exploration of growth inhibition caused by overexpression of genomic fragments in Saccharomyces cerevisiae." Genome Biol 5(9): R72. Breslow, D. K., D. M. Cameron, S. R. Collins, M. Schuldiner, J. Stewart-Ornstein, H. W. Newman, et al. (2008). "A comprehensive strategy enabling high-resolution functional analysis of the yeast genome." Nat Methods 5(8): 711-8. Brill, S. J. and B. Stillman (1991). "Replication factor-A from Saccharomyces cerevisiae is encoded by three essential genes coordinately expressed at S phase." Genes Dev 5(9): 1589-600. Butcher, R. A. and S. L. Schreiber (2006). "A microarray-based protocol for monitoring the growth of yeast overexpression strains." Nat Protoc 1(2): 569-76. Byrne, A. B., M. T. Weirauch, V. Wong, M. Koeva, S. J. Dixon, J. M. Stuart, et al. (2007). "A global analysis of genetic interactions in Caenorhabditis elegans." J Biol 6(3): 8. Chang, M., M. Bellaoui, C. Zhang, R. Desai, P. Morozov, L. Delgado-Cruzata, et al. (2005). "RMI1/NCE4, a suppressor of genome instability, encodes a member of the RecQ helicase/Topo III complex." EMBO J 24(11): 2024-33. Chloupkova, M., L. S. LeBard and D. M. Koeller (2003). "MDL1 is a high copy suppressor of ATM1: evidence for a role in resistance to oxidative stress." J Mol Biol 331(1): 155-65. Collins, S. R., K. M. Miller, N. L. Maas, A. Roguev, J. Fillingham, C. S. Chu, et al. (2007). "Functional dissection of protein complexes involved in yeast chromosome biology using a genetic interaction map." Nature 446(7137): 806-10.

Costanzo, M., A. Baryshnikova, J. Bellay, Y. Kim, E. D. Spear, C. S. Sevier, et al. (2010). "The genetic landscape of a cell." Science 327(5964): 425-31. Costanzo, M. and C. Boone (2009). "SGAM: an array-based approach for high-resolution genetic mapping in Saccharomyces cerevisiae." Methods Mol Biol 548: 37-53. Cutforth, T. and G. M. Rubin (1994). "Mutations in Hsp83 and cdc37 impair signaling by the sevenless receptor tyrosine kinase in Drosophila." Cell 77(7): 1027-36. Davierwala, A. P., J. Haynes, Z. Li, R. L. Brost, M. D. Robinson, L. Yu, et al. (2005). "The synthetic genetic interaction spectrum of essential genes." Nat Genet 37(10): 1147-52. De Wulf, P., A. D. McAinsh and P. K. Sorger (2003). "Hierarchical assembly of the budding yeast kinetochore from multiple subcomplexes." Genes Dev 17(23): 2902-21. Decottignies, A., A. Evain and M. Ghislain (2004). "Binding of Cdc48p to a ubiquitin-related UBX domain from novel yeast proteins involved in intracellular proteolysis and sporulation." Yeast 21(2): 127-39. Decottignies, A., I. Sanchez-Perez and P. Nurse (2003). "Schizosaccharomyces pombe essential genes: a pilot study." Genome Res 13(3): 399-406. Decourty, L., C. Saveanu, K. Zemam, F. Hantraye, E. Frachon, J. C. Rousselle, et al. (2008). "Linking functionally related genes by sensitive and quantitative characterization of genetic interaction profiles." Proc Natl Acad Sci U S A 105(15): 5821-6. Demmel, L., M. Beck, C. Klose, A. L. Schlaitz, Y. Gloor, P. P. Hsu, et al. (2008). "Nucleocytoplasmic shuttling of the Golgi phosphatidylinositol 4-kinase Pik1 is regulated by 14-3-3 proteins and coordinates Golgi function with cell growth." Mol Biol Cell 19(3): 1046-61. Deutschbauer, A. M., D. F. Jaramillo, M. Proctor, J. Kumm, M. E. Hillenmeyer, R. W. Davis, et al. (2005). "Mechanisms of haploinsufficiency revealed by genome-wide profiling in yeast." Genetics 169(4): 1915-25. Diaz-Blanco, N. L. and J. R. Rodriguez-Medina (2007). "Dosage rescue by UBC4 restores cell wall integrity in Saccharomyces cerevisiae lacking the myosin type II gene MYO1." Yeast 24(4): 343-55. Dickson, A. M., Y. Krasikova, P. Pestryakov, O. Lavrik and M. S. Wold (2009). "Essential functions of the 32 kDa subunit of yeast replication protein A." Nucleic Acids Res 37(7): 2313-26.

Dietzl, G., D. Chen, F. Schnorrer, K. C. Su, Y. Barinova, M. Fellner, et al. (2007). "A genome- wide transgenic RNAi library for conditional gene inactivation in Drosophila." Nature 448(7150): 151-6. Dixon, S. J., M. Costanzo, A. Baryshnikova, B. Andrews and C. Boone (2009). "Systematic mapping of genetic interaction networks." Annu Rev Genet 43: 601-25. Dixon, S. J., Y. Fedyshyn, J. L. Koh, T. S. Prasad, C. Chahwan, G. Chua, et al. (2008). "Significant conservation of synthetic lethal genetic interaction networks between distantly related eukaryotes." Proc Natl Acad Sci U S A 105(43): 16653-8. Dobzhansky, T. (1946). "Genetics of Natural Populations. Xiii. Recombination and Variability in Populations of Drosophila Pseudoobscura." Genetics 31(3): 269-90. Drebot, M. A., G. C. Johnston, J. D. Friesen and R. A. Singer (1993). "An impaired RNA polymerase II activity in Saccharomyces cerevisiae causes cell-cycle inhibition at START." Mol Gen Genet 241(3-4): 327-34. Drees, B. L., V. Thorsson, G. W. Carter, A. W. Rives, M. Z. Raymond, I. Avila-Campillo, et al. (2005). "Derivation of genetic interaction networks from quantitative phenotype data." Genome Biol 6(4): R38. Finger, F. P. and P. Novick (1997). "Sec3p is involved in secretion and morphogenesis in Saccharomyces cerevisiae." Mol Biol Cell 8(4): 647-62. Fisher, R. A. (1918). "The correlations between relatives on the supposition of Mendelian inheritance." Trans R Soc Edingburgh 52: 399-433. Gari, E., L. Piedrafita, M. Aldea and E. Herrero (1997). "A set of vectors with a tetracycline- regulatable promoter system for modulated gene expression in Saccharomyces cerevisiae." Yeast 13(9): 837-48. Gasch, A. P., P. T. Spellman, C. M. Kao, O. Carmel-Harel, M. B. Eisen, G. Storz, et al. (2000). "Genomic expression programs in the response of yeast cells to environmental changes." Molecular Biology of the Cell 11(12): 4241-4257. Gavin, A. C., M. Bosche, R. Krause, P. Grandi, M. Marzioch, A. Bauer, et al. (2002). "Functional organization of the yeast proteome by sustematic analysis of protein complexes." Nature 415: 141-147.

Gelperin, D. M., M. A. White, M. L. Wilkinson, Y. Kon, L. A. Kung, K. J. Wise, et al. (2005). "Biochemical and genetic analysis of the yeast proteome with a movable ORF collection." Genes Dev 19(23): 2816-26. Giaever, G., A. M. Chu, L. Ni, C. Connelly, L. Riles, S. Veronneau, et al. (2002). "Functional profiling of the Saccharomyces cerevisiae genome." Nature 418(6896): 387-91. Giaever, G., P. Flaherty, J. Kumm, M. Proctor, C. Nislow, D. F. Jaramillo, et al. (2004). "Chemogenomic profiling: identifying the functional interactions of small molecules in yeast." Proc Natl Acad Sci U S A 101(3): 793-8. Goffeau, A., B. G. Barrell, H. Bussey, R. W. Davis, B. Dujon, H. Feldmann, et al. (1996). "Life with 6000 genes." Science 274(5287): 546, 563-7. Goldstein, A. L. and J. H. McCusker (1999). "Three new dominant drug resistance cassettes for gene disruption in Saccharomyces cerevisiae." Yeast 15(14): 1541-53. Guarente, L. (1993). "Synthetic enhancement in gene interaction: a genetic tool come of age." Trends in Genetics 9: 362-366. Haarer, B., S. Viggiano, M. A. Hibbs, O. G. Troyanskaya and D. C. Amberg (2007). "Modeling complex genetic interactions in a simple eukaryotic genome: actin displays a rich spectrum of complex haploinsufficiencies." Genes Dev 21(2): 148-59. Han, G. S., A. Audhya, D. J. Markley, S. D. Emr and G. M. Carman (2002). "The Saccharomyces cerevisiae LSB6 gene encodes phosphatidylinositol 4-kinase activity." J Biol Chem 277(49): 47709-18. Hartwell, L. H. (1971). "Genetic Control Of The Cell Division Cycle in Yeast." Experimental Cell Research 69: 265-276. Hartwell, L. H., J. Culloti and B. Reid (1970). "Genetic control of cell division in yeast, I. Detection of mutants." Proc. Natl. Acad. Sci. USA 66: 352-359. Hartwell, L. H., J. Culotti, J. R. Pringle and B. J. Reid (1974). "Genetic control of cell division cycle in yeast." Science 183: 46-51. Hays, T. S., R. Deuring, B. Robertson, M. Prout and M. T. Fuller (1989). "Interacting proteins identified by genetic interactions: a missense mutation in alpha-tubulin fails to complement alleles of the testis-specific beta-tubulin gene of Drosophila melanogaster." Mol Cell Biol 9(3): 875-84.

100

Helliwell, S. B., A. Schmidt, Y. Ohya and M. N. Hall (1998). "The Rho1 effector Pkc1, but not Bni1, mediates signalling from Tor2 to the actin cytoskeleton." Curr Biol 8(22): 1211-4. Hillenmeyer, M. E., E. Fung, J. Wildenhain, S. E. Pierce, S. Hoon, W. Lee, et al. (2008). "The chemical genomic portrait of yeast: uncovering a phenotype for all genes." Science 320(5874): 362-5. Ho, C. H., L. Magtanong, S. L. Barker, D. Gresham, S. Nishimura, P. Natarajan, et al. (2009). "A molecular barcoded yeast ORF library enables mode-of-action analysis of bioactive compounds." Nat Biotechnol 27(4): 369-77. Hodgkin, J. (2005). "Genetic suppression." WormBook: 1-13. Hong, M. E., K. S. Lee, B. J. Yu, Y. J. Sung, S. M. Park, H. M. Koo, et al. "Identification of gene targets eliciting improved alcohol tolerance in Saccharomyces cerevisiae through inverse metabolic engineering." J Biotechnol. Hong, M. E., K. S. Lee, B. J. Yu, Y. J. Sung, S. M. Park, H. M. Koo, et al. (2010). "Identification of gene targets eliciting improved alcohol tolerance in Saccharomyces cerevisiae through inverse metabolic engineering." J Biotechnol 149(1-2): 52-9. Hong, M. E., K. S. Lee, B. J. Yu, Y. J. Sung, S. M. Park, H. M. Koo, et al. (2010). "Identification of gene targets eliciting improved alcohol tolerance in Saccharomyces cerevisiae through inverse metabolic engineering." J Biotechnol 149(1-2): 52-59. Hoon, S., A. M. Smith, I. M. Wallace, S. Suresh, M. Miranda, E. Fung, et al. (2008). "An integrated platform of genomic assays reveals small-molecule bioactivities." Nat Chem Biol 4(8): 498-506. Hou, L., X. Cao, C. Wang and M. Lu (2009). "Effect of overexpression of transcription factors on the fermentation properties of Saccharomyces cerevisiae industrial strains." Lett Appl Microbiol 49(1): 14-9. Hu, Y., A. Rolfs, B. Bhullar, T. V. Murthy, C. Zhu, M. F. Berger, et al. (2007). "Approaching a complete repository of sequence-verified protein-encoding clones for Saccharomyces cerevisiae." Genome Res 17(4): 536-43. Hunkapiller, T., R. J. Kaiser, B. F. Koop and L. Hood (1991). "Large-scale and automated DNA sequence determination." Science 254(5028): 59-67. James, P., J. Halladay and E. A. Craig (1996). "Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast." Genetics 144(4): 1425-36.

101

Jones, G. M., J. Stalker, S. Humphray, A. West, T. Cox, J. Rogers, et al. (2008). "A systematic library for comprehensive overexpression screens in Saccharomyces cerevisiae." Nat Methods 5(3): 239-41. Jonikas, M. C., S. R. Collins, V. Denic, E. Oh, E. M. Quan, V. Schmid, et al. (2009). "Comprehensive characterization of genes required for protein folding in the endoplasmic reticulum." Science 323(5922): 1693-7. Jorgensen, P., B. Nelson, M. D. Robinson, Y. Chen, B. Andrews, M. Tyers, et al. (2002). "High- resolution genetic mapping with ordered arrays of Saccharomyces cerevisiae deletion mutants." Genetics 162(3): 1091-9. Kaizu, K., H. Moriya and H. Kitano (2010). "Fragilities caused by dosage imbalance in regulation of the budding yeast cell cycle." PLoS Genet 6(4): e1000919. Kamath, R. S., A. G. Fraser, Y. Dong, G. Poulin, R. Durbin, M. Gotta, et al. (2003). "Systematic functional analysis of the Caenorhabditis elegans genome using RNAi." Nature 421(6920): 231-7. Kidd, T., K. S. Bland and C. S. Goodman (1999). "Slit is the midline repellent for the robo receptor in Drosophila." Cell 96(6): 785-94. Kim, J. H., Y. Zhao, X. Pan, X. He and H. F. Gilbert (2009). "The unfolded protein response is necessary but not sufficient to compensate for defects in disulfide isomerization." J Biol Chem 284(16): 10400-8. Kim, L. C., L. Song and E. B. Haura (2009). "Src kinases as therapeutic targets for cancer." Nat Rev Clin Oncol 6(10): 587-95. Kittler, R., G. Putz, L. Pelletier, I. Poser, A. K. Heninger, D. Drechsel, et al. (2004). "An endoribonuclease-prepared siRNA screen in human cells identifies genes essential for cell division." Nature 432(7020): 1036-40. Komili, S., N. G. Farny, F. P. Roth and P. A. Silver (2007). "Functional specificity among ribosomal proteins regulates gene expression." Cell 131(3): 557-71. Kosodo, Y., K. Imai, A. Hirata, Y. Noda, A. Takatsuki, H. Adachi, et al. (2001). "Multicopy suppressors of the sly1 temperature-sensitive mutation in the ER-Golgi vesicular transport in Saccharomyces cerevisiae." Yeast 18(11): 1003-14.

102

Kota, J., M. Melin-Larsson, P. O. Ljungdahl and H. Forsberg (2007). "Ssh4, Rcr2 and Rcr1 affect plasma membrane transporter activity in Saccharomyces cerevisiae." Genetics 175(4): 1681-94. Krogan, N. J., G. Cagney, H. Yu, G. Zhong, X. Guo, A. Ignatchenko, et al. (2006). "Global landscape of protein complexes in the yeast Saccharomyces cerevisiae." Nature 440(7084): 637-43. 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. Kurischko, C., G. Weiss, M. Ottey and F. C. Luca (2005). "A role for the Saccharomyces cerevisiae regulation of Ace2 and polarized morphogenesis signaling network in cell integrity." Genetics 171(2): 443-55. Kusch, M. and R. S. Edgar (1986). "Genetic studies of unusual loci that affect body shape of the nematode Caenorhabditis elegans and may code for cuticle structural proteins." Genetics 113(3): 621-39. Labbe, S. and D. J. Thiele (1999). "Copper ion inducible and repressible promoter systems in yeast." Methods Enzymol 306: 145-53. Lam, A. F., B. O. Krogh and L. S. Symington (2008). "Unique and overlapping functions of the Exo1, Mre11 and Pso2 nucleases in DNA repair." DNA Repair (Amst) 7(4): 655-62. Lander, E. S., L. M. Linton, B. Birren, C. Nusbaum, M. C. Zody, J. Baldwin, et al. (2001). "Initial sequencing and analysis of the human genome." Nature 409(6822): 860-921. Laribee, R. N., Y. Shibata, D. P. Mersman, S. R. Collins, P. Kemmeren, A. Roguev, et al. (2007). "CCR4/NOT complex associates with the proteasome and regulates histone methylation." Proc Natl Acad Sci U S A 104(14): 5836-41. 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(8): 896-903. Li, J. M., Y. Li and S. J. Elledge (2005). "Genetic analysis of the kinetochore DASH complex reveals an antagonistic relationship with the ras/protein kinase A pathway and a novel subunit required for Ask1 association." Mol Cell Biol 25(2): 767-78. Li, M. Z. and S. J. Elledge (2005). "MAGIC, an in vivo genetic method for the rapid construction of recombinant DNA molecules." Nat Genet 37(3): 311-9.

103

Li, X., S. M. Routt, Z. Xie, X. Cui, M. Fang, M. A. Kearns, et al. (2000). "Identification of a novel family of nonclassic yeast phosphatidylinositol transfer proteins whose function modulates phospholipase D activity and Sec14p-independent cell growth." Mol Biol Cell 11(6): 1989-2005. Little, C. D., M. M. Nau, D. N. Carney, A. F. Gazdar and J. D. Minna (1983). "Amplification and expression of the c-myc oncogene in human lung cancer cell lines." Nature 306(5939): 194-6. Liu, H., J. Krizek and A. Bretscher (1992). "Construction of a GAL1-regulated yeast cDNA expression library and its application to the identification of genes whose overexpression causes lethality in yeast." Genetics 132(3): 665-73. Luduena, R. F., E. M. Shooter and L. Wilson (1977). "Structure of the tubulin dimer." J Biol Chem 252(20): 7006-14. Ma, H., S. Kunes, P. J. Schatz and D. Botstein (1987). "Plasmid construction by homologous recombination in yeast." Gene 58(2-3): 201-16. Maere, S., K. Heymans and M. Kuiper (2005). "BiNGO: a Cytoscape plugin to assess overrepresentation of gene ontology categories in biological networks." Bioinformatics 21(16): 3448-9. Mager, W. H., R. J. Planta, J. G. Ballesta, J. C. Lee, K. Mizuta, K. Suzuki, et al. (1997). "A new nomenclature for the cytoplasmic ribosomal proteins of Saccharomyces cerevisiae." Nucleic Acids Res 25(24): 4872-5. Mani, R., R. P. St Onge, J. L. t. Hartman, G. Giaever and F. P. Roth (2008). "Defining genetic interaction." Proc Natl Acad Sci U S A 105(9): 3461-6. Marcoux, N., S. Cloutier, E. Zakrzewska, P. M. Charest, Y. Bourbonnais and D. Pallotta (2000). "Suppression of the profilin-deficient phenotype by the RHO2 signaling pathway in Saccharomyces cerevisiae." Genetics 156(2): 579-92. Martzen, M. R., S. M. McCraith, S. L. Spinelli, F. M. Torres, S. Fields, E. J. Grayhack, et al. (1999). "A biochemical genomics approach for identifying genes by the activity of their products." Science 286: 1153-1155. Matsui, Y. and A. Toh-e (1992). "Isolation and characterization of two novel ras superfamily genes in Saccharomyces cerevisiae." Gene 114(1): 43-9.

104

McCusker, J. H., M. Yamagishi, J. M. Kolb and M. Nomura (1991). "Suppressor analysis of temperature-sensitive RNA polymerase I mutations in Saccharomyces cerevisiae: suppression of mutations in a zinc-binding motif by transposed mutant genes." Mol Cell Biol 11(2): 746-53. Measday, V., K. Baetz, J. Guzzo, K. Yuen, T. Kwok, B. Sheikh, et al. (2005). "Systematic yeast synthetic lethal and synthetic dosage lethal screens identify genes required for chromosome segregation." Proc Natl Acad Sci U S A 102(39): 13956-61. Measday, V., D. W. Hailey, I. Pot, S. A. Givan, K. M. Hyland, G. Cagney, et al. (2002). "Ctf3p, the Mis6 budding yeast homolog, interacts with Mcm22p and Mcm16p at the yeast outer kinetochore." Genes Dev 16(1): 101-13. Menne, T. F., B. Goyenechea, N. Sanchez-Puig, C. C. Wong, L. M. Tonkin, P. J. Ancliff, et al. (2007). "The Shwachman-Bodian-Diamond syndrome protein mediates translational activation of ribosomes in yeast." Nat Genet 39(4): 486-95. Meyer, N. and L. Z. Penn (2008). "Reflecting on 25 years with MYC." Nat Rev Cancer 8(12): 976-90. Mnaimneh, S., A. P. Davierwala, J. Haynes, J. Moffat, W. T. Peng, W. Zhang, et al. (2004). "Exploration of essential gene functions via titratable promoter alleles." Cell 118(1): 31- 44. Moffat, J., D. A. Grueneberg, X. Yang, S. Y. Kim, A. M. Kloepfer, G. Hinkle, et al. (2006). "A lentiviral RNAi library for human and mouse genes applied to an arrayed viral high- content screen." Cell 124(6): 1283-98. Moriya, H., Y. Shimizu-Yoshida and H. Kitano (2006). "In vivo robustness analysis of cell division cycle genes in Saccharomyces cerevisiae." PLoS Genet 2(7): e111. Murgola, E. J. (1985). "tRNA, suppression, and the code." Annu Rev Genet 19: 57-80. Murray, J. A. (1987). "Bending the rules: the 2-mu plasmid of yeast." Mol Microbiol 1(1): 1-4. Myers, C. L., D. R. Barrett, M. A. Hibbs, C. Huttenhower and O. G. Troyanskaya (2006). "Finding function: evaluation methods for functional genomic data." BMC Genomics 7: 187. Nagai, S., K. Dubrana, M. Tsai-Pflugfelder, M. B. Davidson, T. M. Roberts, G. W. Brown, et al. (2008). "Functional targeting of DNA damage to a nuclear pore-associated SUMO- dependent ubiquitin ligase." Science 322(5901): 597-602.

105

Norgaard, P., V. Westphal, C. Tachibana, L. Alsoe, B. Holst and J. R. Winther (2001). "Functional differences in yeast protein disulfide isomerases." J Cell Biol 152(3): 553-62. Nugent, R. L., A. Johnsson, B. Fleharty, M. Gogol, Y. Xue-Franzen, C. Seidel, et al. (2010). "Expression profiling of S. pombe acetyltransferase mutants identifies redundant pathways of gene regulation." BMC Genomics 11: 59. Orlowski, J., K. Machula, A. Janik, E. Zdebska and G. Palamarczyk (2007). "Dissecting the role of dolichol in cell wall assembly in the yeast mutants impaired in early glycosylation reactions." Yeast 24(4): 239-52. Paddison, P. J., J. M. Silva, D. S. Conklin, M. Schlabach, M. Li, S. Aruleba, et al. (2004). "A resource for large-scale RNA-interference-based screens in mammals." Nature 428(6981): 427-31. Pagliuca, C., V. M. Draviam, E. Marco, P. K. Sorger and P. De Wulf (2009). "Roles for the conserved spc105p/kre28p complex in kinetochore-microtubule binding and the spindle assembly checkpoint." PLoS ONE 4(10): e7640. Pan, X., D. S. Yuan, D. Xiang, X. Wang, S. Sookhai-Mahadeo, J. S. Bader, et al. (2004). "A robust toolkit for functional profiling of the yeast genome." Mol Cell 16(3): 487-96. Parsons, A. B., A. Lopez, I. E. Givoni, D. E. Williams, C. A. Gray, J. Porter, et al. (2006). "Exploring the mode-of-action of bioactive compounds by chemical-genetic profiling in yeast." Cell 126(3): 611-25. Phillips, P. C. (2008). "Epistasis--the essential role of gene interactions in the structure and evolution of genetic systems." Nat Rev Genet 9(11): 855-67. Pickart, M. A., E. W. Klee, A. L. Nielsen, S. Sivasubbu, E. M. Mendenhall, B. R. Bill, et al. (2006). "Genome-wide reverse genetics framework to identify novel functions of the vertebrate secretome." PLoS ONE 1: e104. Pierce, S. E., R. W. Davis, C. Nislow and G. Giaever (2007). "Genome-wide analysis of barcoded Saccharomyces cerevisiae gene-deletion mutants in pooled cultures." Nat Protoc 2(11): 2958-74. Pierce, S. E., E. L. Fung, D. F. Jaramillo, A. M. Chu, R. W. Davis, C. Nislow, et al. (2006). "A unique and universal molecular barcode array." Nat Methods 3(8): 601-3.

106

Pramila, T., W. Wu, S. Miles, W. S. Noble and L. L. Breeden (2006). "The Forkhead transcription factor Hcm1 regulates chromosome segregation genes and fills the S-phase gap in the transcriptional circuitry of the cell cycle." Genes Dev 20(16): 2266-78. Prelich, G. (1999). "Suppression mechanisms: themes from variations." Trends Genet 15(7): 261-6. Reed, S. I., J. A. Hadwiger, H. E. Richardson and C. Wittenberg (1989). "Analysis of the Cdc28 protein kinase complex by dosage suppression." J Cell Sci Suppl 12: 29-37. Reuter, G. and I. Wolff (1981). "Isolation of dominant suppressor mutations for position-effect variegation in Drosophila melanogaster." Mol Gen Genet 182(3): 516-9. Richardson, H. E., C. Wittenberg, F. Cross and S. I. Reed (1989). "An essential G1 function for cyclin-like proteins in yeast." Cell 59(6): 1127-33. Richman, T. J., M. M. Sawyer and D. I. Johnson (1999). "The Cdc42p GTPase is involved in a G2/M morphogenetic checkpoint regulating the apical-isotropic switch and nuclear division in yeast." J Biol Chem 274(24): 16861-70. Rine, J. (1991). "Gene overexpression in studies of Saccharomyces cerevisiae." Methods Enzymol 194: 239-51. Rine, J. and I. Herskowitz (1987). "Four genes responsible for a position effect on expression from HML and HMR in Saccharomyces cerevisiae." Genetics 116(1): 9-22. Roberg, K. J., M. Crotwell, P. Espenshade, R. Gimeno and C. A. Kaiser (1999). "LST1 is a SEC24 homologue used for selective export of the plasma membrane ATPase from the endoplasmic reticulum." J Cell Biol 145(4): 659-72. Rodriguez-Quinones, J. F., R. A. Irizarry, N. L. Diaz-Blanco, F. E. Rivera-Molina, D. Gomez- Garzon and J. R. Rodriguez-Medina (2008). "Global mRNA expression analysis in myosin II deficient strains of Saccharomyces cerevisiae reveals an impairment of cell integrity functions." BMC Genomics 9: 34. Roguev, A., S. Bandyopadhyay, M. Zofall, K. Zhang, T. Fischer, S. R. Collins, et al. (2008). "Conservation and rewiring of functional modules revealed by an epistasis map in fission yeast." Science 322(5900): 405-10. Sambrook, J., D. W. Russell and Cold Spring Harbor Laboratory. (2001). Molecular cloning : a laboratory manual. Cold Spring Harbor, N.Y., Cold Spring Harbor Laboratory.

107

Sandrock, T. M., J. L. O'Dell and A. E. Adams (1997). "Allele-specific suppression by formation of new protein-protein interactions in yeast." Genetics 147(4): 1635-42. 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(1): 59-64. Schneider, J. C. and L. Guarente (1991). "Vectors for expression of cloned genes in yeast: regulation, overproduction, and underproduction." Methods Enzymol 194: 373-88. Schuldiner, M., S. R. Collins, N. J. Thompson, V. Denic, A. Bhamidipati, T. Punna, et al. (2005). "Exploration of the function and organization of the yeast early secretory pathway through an epistatic miniarray profile." Cell 123(3): 507-19. Schwab, R., S. Ossowski, M. Riester, N. Warthmann and D. Weigel (2006). "Highly specific gene silencing by artificial microRNAs in Arabidopsis." Plant Cell 18(5): 1121-33. Segre, D., A. Deluna, G. M. Church and R. Kishony (2005). "Modular epistasis in yeast metabolism." Nat Genet 37(1): 77-83. Shannon, P., A. Markiel, O. Ozier, N. S. Baliga, J. T. Wang, D. Ramage, et al. (2003). "Cytoscape: a software environment for integrated models of biomolecular interaction networks." Genome Res 13(11): 2498-504. Shea, J. E., J. H. Toyn and L. H. Johnston (1994). "The budding yeast U5 snRNP Prp8 is a highly conserved protein which links RNA splicing with cell cycle progression." Nucleic Acids Res 22(25): 5555-64. Shendure, J. and H. Ji (2008). "Next-generation DNA sequencing." Nat Biotechnol 26(10): 1135-45. Shirayama, M., K. Kawakami, Y. Matsui, K. Tanaka and A. Toh-e (1993). "MSI3, a multicopy suppressor of mutants hyperactivated in the RAS-cAMP pathway, encodes a novel HSP70 protein of Saccharomyces cerevisiae." Mol Gen Genet 240(3): 323-32. Shoemaker, D. D., D. A. Lashkari, D. Morris, M. Mittmann and R. W. Davis (1996). "Quantitative phenotypic analysis of yeast deletion mutants using a highly parallel molecular bar-coding strategy." Nature Genetics 14: 450-456. Smith, A. M., L. E. Heisler, J. Mellor, F. Kaper, M. J. Thompson, M. Chee, et al. (2009). "Quantitative phenotyping via deep barcode sequencing." Genome Res 19(10): 1836-42.

108

Smith, A. M., L. E. Heisler, R. P. St Onge, E. Farias-Hesson, I. M. Wallace, J. Bodeau, et al. (2010). "Highly-multiplexed barcode sequencing: an efficient method for parallel analysis of pooled samples." Nucleic Acids Res. Sopko, R., D. Huang, N. Preston, G. Chua, B. Papp, K. Kafadar, et al. (2006). "Mapping pathways and phenotypes by systematic gene overexpression." Mol Cell 21(3): 319-30. Sopko, R., D. Huang, J. C. Smith, D. Figeys and B. J. Andrews (2007). "Activation of the Cdc42p GTPase by cyclin-dependent protein kinases in budding yeast." EMBO J 26(21): 4487-500. St John, T. P. and R. W. Davis (1979). "Isolation of galactose-inducible DNA sequences from Saccharomyces cerevisiae by differential plaque filter hybridization." Cell 16(2): 443-52. St John, T. P. and R. W. Davis (1981). "The organization and transcription of the galactose gene cluster of Saccharomyces." J Mol Biol 152(2): 285-315. St Onge, R. P., R. Mani, J. Oh, M. Proctor, E. Fung, R. W. Davis, et al. (2007). "Systematic pathway analysis using high-resolution fitness profiling of combinatorial gene deletions." Nat Genet 39(2): 199-206. Stearns, T. and D. Botstein (1988). "Unlinked noncomplementation: isolation of new conditional-lethal mutations in each of the tubulin genes of Saccharomyces cerevisiae." Genetics 119: 249-260. Stevenson, B. J., N. Rhodes, B. Errede and G. F. Sprague, Jr. (1992). "Constitutive mutants of the protein kinase STE11 activate the yeast pheromone response pathway in the absence of the G protein." Genes Dev 6(7): 1293-304. Stevenson, L. F., B. K. Kennedy and E. Harlow (2001). "A large-scale overexpression screen in Saccharomyces cerevisiae identifies previously uncharacterized cell cycle genes." Proc Natl Acad Sci U S A 98(7): 3946-51. Sun, S. J., S. Q. Guo, X. Yang, Y. M. Bao, H. J. Tang, H. Sun, et al. (2010). "Functional analysis of a novel Cys2/His2-type zinc finger protein involved in salt tolerance in rice." J Exp Bot 61(10): 2807-18. Sung, P. (1997). "Yeast Rad55 and Rad57 proteins form a heterodimer that functions with replication protein A to promote DNA strand exchange by Rad51 recombinase." Genes Dev 11(9): 1111-21.

109

Tanaka, K., M. Nakafuku, F. Tamanoi, Y. Kaziro, K. Matsumoto and A. Toh-e (1990). "IRA2, a second gene of Saccharomyces cerevisiae that encodes a protein with a domain homologous to mammalian ras GTPase-activating protein." Mol Cell Biol 10(8): 4303- 13. Tarassov, K., V. Messier, C. R. Landry, S. Radinovic, M. M. Serna Molina, I. Shames, et al. (2008). "An in vivo map of the yeast protein interactome." Science 320(5882): 1465-70. Thomas, D., H. Cherest and Y. Surdin-Kerjan (1989). "Elements involved in S- adenosylmethionine-mediated regulation of the Saccharomyces cerevisiae MET25 gene." Mol Cell Biol 9(8): 3292-8. Tong, A. H., G. Lesage, G. D. Bader, H. Ding, H. Xu, X. Xin, et al. (2004). "Global mapping of the yeast genetic interaction network." Science 303(5659): 808-13. Tong, A. H. Y., M. Evangelista, A. B. Parsons, H. Xu, G. D. Bader, N. Page, et al. (2001). "Systematic genetic analysis with ordered arrays of yeast deletion mutants." Science 294: 2364-2368. Tschumper, G. and J. Carbon (1983). "Copy number control by a yeast centromere." Gene 23(2): 221-32. Tyers, M. and B. Futcher (1993). "Far1 and Fus3 link the mating pheromone signal transduction pathway to three G1-phase Cdc28 kinase complexes." Mol Cell Biol 13(9): 5659-69. Uetz, P., L. Giot, G. Cagney, T. Mansfield, R. S. Judson, J. R. Knight, et al. (2000). "A comprehensive analysis of protein-protein interactions in Saccharomyces cerevisiae." Naure 403: 623-627. Valay, J. G., M. Simon and G. Faye (1993). "The kin28 protein kinase is associated with a cyclin in Saccharomyces cerevisiae." J Mol Biol 234(2): 307-10. van Dongen, S. (2002). "A cluster algorithm for graphs. (National Research Institute for Mathematics and Computer Science in The Netherlands, Amsterdam)." van Opijnen, T., K. L. Bodi and A. Camilli (2009). "Tn-seq: high-throughput parallel sequencing for fitness and genetic interaction studies in microorganisms." Nat Methods 6(10): 767- 72. Vavouri, T., J. I. Semple, R. Garcia-Verdugo and B. Lehner (2009). "Intrinsic protein disorder and interaction promiscuity are widely associated with dosage sensitivity." Cell 138(1): 198-208.

110

Venter, J. C., M. D. Adams, E. W. Myers, P. W. Li, R. J. Mural, G. G. Sutton, et al. (2001). "The sequence of the human genome." Science 291(5507): 1304-51. Vinh, D. B., M. D. Welch, A. K. Corsi, K. F. Wertman and D. G. Drubin (1993). "Genetic evidence for functional interactions between actin noncomplementing (Anc) gene products and actin cytoskeletal proteins in Saccharomyces cerevisiae." Genetics 135(2): 275-86. Vizeacoumar, F. J., N. van Dyk, S. V. F, V. Cheung, J. Li, Y. Sydorskyy, et al. (2010). "Integrating high-throughput genetic interaction mapping and high-content screening to explore yeast spindle morphogenesis." J Cell Biol 188(1): 69-81. Warner, J. R. (1999). "The economics of ribosome biosynthesis in yeast." Trends Biochem Sci 24(11): 437-40. Warner, J. R. and K. B. McIntosh (2009). "How common are extraribosomal functions of ribosomal proteins?" Mol Cell 34(1): 3-11. Watts, F. Z., G. Shiels and E. Orr (1987). "The yeast MYO1 gene encoding a myosin-like protein required for cell division." EMBO J 6(11): 3499-505. Wilmes, G. M., M. Bergkessel, S. Bandyopadhyay, M. Shales, H. Braberg, G. Cagney, et al. (2008). "A genetic interaction map of RNA-processing factors reveals links between Sem1/Dss1-containing complexes and mRNA export and splicing." Mol Cell 32(5): 735- 46. Yan, Z., M. Costanzo, L. E. Heisler, J. Paw, F. Kaper, B. J. Andrews, et al. (2008). "Yeast Barcoders: a chemogenomic application of a universal donor-strain collection carrying bar-code identifiers." Nat Methods 5(8): 719-25. Ye, Y., H. H. Meyer and T. A. Rapoport (2003). "Function of the p97-Ufd1-Npl4 complex in retrotranslocation from the ER to the cytosol: dual recognition of nonubiquitinated polypeptide segments and polyubiquitin chains." J Cell Biol 162(1): 71-84. Yook, K. (2005). "Complementation." WormBook: 1-17. Yook, K. J., S. R. Proulx and E. M. Jorgensen (2001). "Rules of nonallelic noncomplementation at the synapse in Caenorhabditis elegans." Genetics 158(1): 209-20. Yu, H., P. Braun, M. A. Yildirim, I. Lemmens, K. Venkatesan, J. Sahalie, et al. (2008). "High- quality binary protein interaction map of the yeast interactome network." Science 322(5898): 104-10.

111

Zhu, H., M. Bilgin, R. Bangham, D. Hall, A. Casamayor, P. Bertone, et al. (2001). "Global analysis of protein activities using proteome chips." Science 293(5537): 2101-5. Ziman, M. and D. I. Johnson (1994). "Genetic evidence for a functional interaction between Saccharomyces cerevisiae CDC24 and CDC42." Yeast 10(4): 463-74.

Chapter Six Appendices

112 113

Appendix 6.1 All confirmed spot dilutions performed based on dosage suppression screens reported in this study.