I o

HIGH-THROUGHPUT APPROACHES AND APPLICATIONS FOR CHEMOGENOMICS

A DISSERTATION SUBMITTED TO THE DEPARTMENT OF GENETICS AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Shawn Hoon December 2008

UMI Number: 3343956 Copyright 2009 by Hoon, Shawn

All rights reserved.

INFORMATION TO USERS

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

®

UMI UMI Microform 3343956 Copyright 2009 by ProQuest LLC. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest LLC 789 E. Eisenhower Parkway PO Box 1346 Ann Arbor, Ml 48106-1346

© Copyright by Shawn Hoon 2009 All Rights Reserved

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

(Ronald W. Davis)

N

Principal Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor_^f Philosophy.

(Andrew Fire) I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

(Corey Nislow) I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the dej£§£_af_[)octor of Philosophy.

(Timothy Stearns) Approved for the University Committee on Graduate Studies.

^fc-^. jtf~*phin

0.1

Abstract

The intersection between chemistry and biology has always been important for understanding the molecular processes of life. One of the key challenges today is to determine how chemicals regulate biological systems which has implications for how chemicals may be used for research as well as the development of therapeutics. Genomic technologies have great strides in this area and continue to do so. In this dissertation, I describe four areas of work that sought to advance this field: 1) Development of a cost-effective and extensible platform that integrates three distinct assays with the goal of improving the characterization of the bioactivity of small-molecules and making the technology accessible. This platform is built on the robust TAG4 barcode microarray and simultaneously resolves the fitness of strains derived from pools of a) homozygous deletion mutants, b) heterozygous deletion mutants, and c) genomic library transformants; 2) Development of methods to screen drug combinations systematically and to identify genetic mutants that modify the cellular response to drug combinations thereby allowing the mechanistic basis for the interaction to be inferred; 3) Glyoxal and methyglyoxal are reactive carbonyls, formed as by-products of metabolism and found in a large number of environmental sources such as cigarette smoke, automobile exhaust and thermal processing of food. This dissertation describes the results of using multiple genomewide assays to identify the genetic basis for carbonyl stress resistance; 4) Gain-of-function (GOF) genetic screens are a powerful way of generating phenotypes and have been used with great success in multiple biological systems. I describe the generation of a yeast pool expressing ~ 12,000 different human genes and demonstrate that toxic human genes can be identified via a competitive growth assay.

IV

0.2

Acknowledgments

There have been many that have made this journey possible. I would like to thank the following: - my parents, Simon and Eileen for your unconditional love and support and for never questioning why I do what I do (imagine, 'yeast research'). - my parents-in-law especially to my father-in-law who passed away during the writing of this thesis. - my sister Sherlynn for your love and support - my grandmother and my late grandfather. - my advisor Ron, who has provided vision, invaluable guidance and a wonderful environment for me to explore and develop ideas, and most importantly the room to fail. Solving difficult problems is an easy thing to say but Ron is a testament to how that can be achieved. - my committee, Tim, Andy, John for inspiring me with your intellect, ideas and guidance. - all members of SGTC past and present, in particular the supportive and fun-loving HIP-HOP group: Bob, Julia, Maureen, Sarah, Will, Ana Maria, Eula, Molly, Dan J, Kara, Michael, Sundari for being a part of such a special place that is SGTC. I would also like to thank Joella, Donna, Mary K, Wendy Christiansen all of whom have been very patient and helpful. - Special thanks to my Genetics class, Nanibaa, Jeff, Evonne (and Josh), Matt (and Amy), Sarah (and John), Heather and Dan Riordan for your invaluable friendship (and getting me through Gene 203). - The rest of the Genetics department especially to Arend Sidow. - the Cyert lab for inviting my family into theirs. - Corey Nislow and Guri Giaever for taking me under their wings. Your mentorship has been invaluable and your unfailing optimism has kept me going many times. - Chris Tan/Venki/Elia/Sam/Sumanty from days at IMCB who gave me my start in biology. - Gene, for getting me started in this whole adventure and who continues to amaze me. - My two greatest achievements, Linus and Amelia. Linus, you humble me with your energy, curiosity and optimism. Amelia, I can already see the strength you possess and I look forward to watching you grow up. - And most importantly, Hooi See, without whom, this adventure would not have been worthwhile. This is as much yours as it is mine.

v

Contents

1

2

0.1

Abstract

iv

0.2

Acknowledgments

v

Introduction

1

1.1

Yeast as a model organism

1

1.2

Discovering chemical mechanism of action

2

1.3

Chemogenomics in yeast

2

1.4

Organization

4

1.5

Bibliography

5

A n integrated platform of genomic assays reveals small molecule bioactivities

11

2.1

Abstract

11

2.2

Introduction

12

2.3

Results

12

2.3.1

A Chemogenomics Platform

12

2.3.2

Integrating gene-dose assays improves target identification

14

2.3.3

Cantharidin and Calyculin A have distinct effects in vivo

15

2.3.4

Characterization of novel bioactive compounds

16

2.3.5

Comparison of profiles to compound structure

17

2.4

Discussion

18

2.5

Materials and Methods

19

2.5.1

Reagents

19

2.5.2

Individual strain growth analysis

23

2.5.3

Deletion-pool construction and screening conditions

23

2.5.4

Multicopy Pool Construction and screening conditions

23

2.5.5

Genomic DNA preparation, TAG PCR, and microarray hybridization

....

24

2.5.6

Plasmid isolation and insert PCR amplification and microarray hybridization

24

2.5.7

Microarray Analysis

24

2.5.8

Chemical descriptor annotation

25

vi

2.5.9

3

4

Acknowledgements

25

2.6

Tables

26

2.7

Figures

29

2.8

Bibliography

40

A framework for analyzing drug-drug interactions

45

3.1

Abstract

45

3.2

Introduction

45

3.3

Results

46

3.3.1

Growth assay for generating dose response surfaces

46

3.3.2

Cantharidin and Calyculin A synergize to inhibit yeast growth

47

3.3.3

Chemical combinations exert distinct effects

47

3.3.4

Mechanistic insights into drug interactions

47

3.4

Discussion

48

3.5


49

3.5.1

Dose response surface analysis

49

3.5.2

Quantifying deletion strain drug interactions

50

3.5.3

Yeast Interaction Network

51

3.6

Tables

52

3.7

Figures

53

3.8

Bibliography

58

A global perspective of the genetic basis for carbonyl stress resistance

61

4.1

Abstract

61

4.2

Introduction

61

4.3

Results

63

4.3.1

Methyglyoxal and glyoxal inhibit yeast growth

4.3.2

Different HOG pathway components are required for resistance to MG and Glyoxal

4.3.3

63 63

Genomewide fitness profiling identifies differences between the cellular effects of MG and GLY

64

4.3.4

Carbonyl adduct detoxification

64

4.3.5

Identifying glyoxal resistant strains

65

4.3.6

A sensitized suppressor screen to understand glyoxal stress

65

4.3.7

Overexpression of aldehyde reductases suppress glyoxal toxicity

66

4.3.8

NADPH generating pathways buffer glyoxal resistance

66

4.3.9

Role of PKA Signaling

66

4.3.10 The HOG pathway suppresses glyoxal sensitivity

vn

67

4.3.11 Quantitative epistasis analysis

67

4.3.12 Loss of Fpsl and Ervl4 abolishes requirement for HOG pathway

68

4.4

Discussion

69

4.5


70

4.5.1

Strains and Reagents

70

4.5.2

Plasmids

70

4.5.3

Growth Assay

70

4.5.4

Genomewide screening procedure

70

4.5.5

Microarray Analysis

71

4.5.6

Double mutant Strain Construction

71

4.5.7

Growth assay for epistasis analysis

71

4.5.8

Fluorescence Microscopy

72

4.6

Tables

73

4.7

Figures

75

4.8

Bibliography

84

5 Identifying human proteins toxic to yeast

93

5.1

Abstract

93

5.2

Introduction

93

5.3

Results

94

5.3.1

Generating a yeast pool expressing human yeast proteins

94

5.3.2

Human Toxicity Screen

95

5.3.3

Confirmation of Toxic human genes

95

5.3.4

Function Enrichment of toxic human genes

95

5.3.5

Structural variants within the toxic ORF set

96

5.3.6

ANG nuclear targeting signal modulates toxicity

96

5.4

Discussion

97

5.5


97

5.5.1

Pool Construction

97

5.5.2

Pool Growth

98

5.5.3

Plasmid purification, clone amplification and DNA labeling

98

5.5.4

Chip hybridization

98

5.5.5

Chip Analysis

99

5.6

Tables

100

5.7

Figures

103

5.8

Bibliography

108

viii

6

Conclusion and perspectives

113

6.1

Conclusion

113

6.2

Perspectives

113

6.2.1

Small-molecule screening

113

6.2.2

Drug interactions

114

6.2.3

Characterization of toxic compounds

114

6.2.4

Yeast and Beyond

115

6.3

Bibliography

116

IX

List of Tables 2.1

Primers for cloning ORFs by gap-repair

20

2.2

Individual strains used in this study

21

2.3

Individual plasmids used in this study

22

2.4

Reference compounds used in this study

26

2.5

List of top 20 gene-compound pairs with the highest combined DSP-MSP activity scores 27

2.6

ECFP_4 Tanimoto Coefficient Similarity of 9-anilinoacridine derivatives to m-AMSA

28

3.1

Fit results for drug surface modeling

52

4.1

Multicopy suppressors identified by MSP

73

4.2

Plasmids used in this study

74

5.1

Biological Functions Over-represented in Toxic Gene set

100

5.2

Distribution of polymorphic and splice variants

102

x

List of Figures 2.1

An integrated chemogenomics screening platform

29

2.2

Multicopy Suppression Profiling

30

2.3

Data integration improves small-molecule target prediction

31

2.4

Data integration improves small-molecule target prediction (2)

32

2.5


33

2.6

Cantharidin and Calyculin A have distinct effects in vivo (2)

34

2.7

CRGl is a major suppressor of cantharidin sensitivity.

35

2.8


36

2.9

Characterization of novel bioactive compounds (2)

37

2.10 Correlation of chemogenomic profiles with compound structure similarity.

38

2.11 Chemogenomics profiles reveal structure activity relationships

39

3.1

Best fit of dose response surfaces for 10 compound pairs

53

3.2

Cantharidin and calyculin A are synergistic in inhibiting growth

54

3.3


55

3.4


56

3.5

Identifying biological processes targeted by drug combinations

57

4.1

Methylgloxal and glyoxal inhibit yeast growth

75

4.2

Identifying deletion strains resistant to glyoxal stress

76

4.3

Sensitized glyoxal suppressor screen

77

4.4

MSP Confirmation growth curves

78

4.5

cdc20+/- is sensitive to glyoxal

79

4.6

Quantitative epistasis analysis

80

4.7

Genetic interaction profiles predict pathways

81

4.8


82

4.9

Model for glyoxal resistance

83

5.1

Human Pool Construction

103

xi

5.2

Human Toxicity Screen outline

104

5.3

Identifying toxic human genes

105

5.4

Features of the Human toxic gene set

106

5.5

Missense mutation in ANG modifies toxicity to yeast

107

xn

Chapter 1

Introduction 1.1

Yeast as a model organism

Yeast has proven to be a remarkable model for the study of many fundamental questions in biology ranging from cell-cycle regulation[20, 21], and gene expression regulation[32] to genome evolution[45], population genetics[48] and ecology[31]. Genetic manipulation is cheap and easy making it one of the most attractive model organisms [2]. By sequence similarity, approximately 30% of yeast genes have a corresponding mammalian homolog[2]. For this reason, yeast has also proven to be an excellent model for many human diseases. Taken together, these features make yeast attractive as a rapidprototyping system, for testing risky ideas with the eventual goal of applying the same technology and concepts to higher organisms. In 1996, Saccharomyces cerevisiae became the first eukaryote to have a fully sequenced genome[18]. The enumeration of all open reading frames (ORF) gave researchers for the first time a complete 'parts' list. With this and the deluge of other sequenced genomes that followed, sequence analysis went from a cottage industry to a full-blown industrial revolution. Virtual pipelines were fed DNA sequences at one end and genes and their putative functions came out of the other. However, sequence analysis can only take us so far. It is extremely humbling that even with Saccharomyces cerevisiae, one of the most well studied organisms to date, 21% of genes still have no known function (even with a very loose definition of 'known') [39]. This is no surprise but at least, a gene list can be used as a report card to periodically assess our progress. And progress has not been slow. Many developments are being made in genomics with yeast being the vanguard of this revolution. One of the first applications was microarray based expression profiling which allowed the entire transcriptional response of a cell to be measured[14]. A pioneering study characterizing the transcription response of cells to a variety environmental conditions provided for the first time, a dynamic picture of cellular systems[14]. Recent highlights include the use of tiling arrays for mapping the location nucleosome[33] and full-length transcripts[9] as well as the study of complex traits[41, 19]. 1

CHAPTER

1.2

1.

INTRODUCTION

2

Discovering chemical mechanism of action

Bioactive compounds axe widely used to modulate protein function and can be effective tools to understand cell physiology, and when combined with appropriate medicinal chemistry efforts, serve as a starting point for therapeutics. However, identifying the in vivo targets of any bioactive compound remains a considerable challenge. Indeed, in light of several high-profile drug withdrawals from the market [47, 13], understanding the spectrum of off-target effects of any compound in an unbiased manner has never been more important. Another important application of chemical characterization is chemical testing. It has become increasing apparent that we need to understand the effects of everyday chemicals that we are exposed to. For a long time, the use of chemicals in consumer products have operated on the premise that all chemicals are considered safe until proven otherwise, placing the burden of chemical testing on regulatory agencies. This is now beginning to change as public concern has made chemical safety a priority. For example, the implementation of REACH in the European Union (EU) represents a significant paradigm shift in the regulation of chemical safety where chemicals have to be proven safe in order to be used thereby placing the burden of proof on the industry[5]. What does this mean for chemical testing research? The lack of regulation for chemical testing has meant the relatively slow adoption of genomics technologies compared to drug discovery. With a changing regulatory landscape, genomics technology will need to be rapidly adopted because of the cost of traditional animal-based testing[8]. There are about 100,000 different chemicals in use in the EU of which 30,000 will require registration with REACH of which less than 1 percent have been thoroughly tested [3]. An estimated 45 million laboratory animals is required to meet this legislation and the economical and political costs associated with animal testing present a formidable challenge [24]. Yeast-based approaches have a unique opportunity to fill this gap to become the Ames test of the 21st century[46].

1.3

Chemogenomics in yeast

To advance the discovery of gene function, a concerted effort by an international consortium led to the completion of the Yeast Knockout Collection (YKO), a set of deletion strains for almost all annotated yeast genes. This collection comprises 21 000 strains (haploid strains for both MATa and MATa mating types, heterozygous and homozygous diploid strains) and since its completion almost 6 years ago, it has proven to be a remarkably useful toolbox. These collections permit comprehensive and systematic genetic screens to be performed by any academic laboratory. In one experimental scenario, a fitness profile of a given compound, generated by measuring the effect of the compound on 6000 different genetic backgrounds, generates a global measure of the compound at the cellular level. This allows an unbiased approach to group compounds based on the similarity of their fitness profiles [38, 23]. Fitness profiles can also be used to identify relationships between genes, e.g. genes

CHAPTER

1.

INTRODUCTION

3

that shaxe similar fitness profiles tend to share common function, in other words they behave as if "co-fit", and this co-fitness can help uncover novel functions of characterized and uncharacterized genes[34, 23]. The inclusion of unique DNA barcode in each deletion strain allows fitness assays to be performed in a pooled fashion and individual growth measurements to be resolved in parallel using a microarray[50, 15]. Based on these pioneering efforts, a number of these resources and approaches have since been replicated in other model organisms. Several groups have published chemical genomic studies using S.cerevisiae as well as other yeast species like Schizosaccharomyces pombe and Candida albicans [27, 44, 51, 7]. Mechanism of drug action can be defined at different levels of resolution from a phenotypic descriptor (e.g. apoptosis), to a,process (e.g. the identification of cellular pathways), and finally, to the in vivo drug target. Developments in chemical genomics in yeast have focused on identifying mechanism of action using the latter two definitions, as the primary phenotypic descriptor for yeast studies is most often growth or fitness. The genetic tractability of yeast, combined with increased access to automation and high-density experimental platforms (e.g. DNA microarrays) has facilitated the development of many genome-wide assays that measure the effect of altering gene dosage on drug sensitivity. For example, drug-induced Haploinsufficency Profiling (HIP) was developed for the identification of small-molecules that target gene products essential for growth by measuring the drug sensitivity of heterozygous deletion strains[16, 17, 36, 10]. Based on a similar principle, multicopy suppression profiling identifies small-molecule targets via genes that confer resistance to drug when over-expressed [35, 6, 43]. Both assays can identify direct targets of compounds as well as genes that modify drug resistance indirectly. Homozygous profiling uses complete loss-of-function alleles as homozygotes to identify pathways that buffer drug sensitivity[34, 4, 11]. A similar approach makes use of haploid deletion strains [38]. Homozygous profiling could also be used to identify the direct target of small-molecules by detecting gene deletions that result in small-molecule resistance. For example, FK506 binds to the FKBP12 prolyl isomerase and together the FK506/FKBP12 complex inhibits the phosphatase calcineurin. Deletion of FKBP12 eliminates this interaction and results in resistance to FK506 [22]. Another approach, which I term Toxicity Suppression Profiling (TSP), identifies compounds that alleviate the growth inhibition caused by over-expression of toxic proteins[29, 12, 40, 49, 1]. Though in its present form, TSP is a "candidate approach" for target identification and only applicable in the case when overexpression of the protein is toxic, these screens could be scaled to interrogate pools of strains expressing different toxic proteins, with the results resolved by DNA microarrays. An advantage of this approach is that heterologous proteins from other species can be used provided that expression of these proteins cause a growth defect or are toxic to yeast [49, 40]. This may be especially useful for identifying drugs that target aberrant human proteins. Also, because these screens identify compounds that restore growth, any such chemical suppressors would have the additional desired property of being

CHAPTER

1.

INTRODUCTION

4

cell-permeable and not cytotoxic. Other successful genome-wide drug discovery strategies have relied on 'guilt-by-association', inferring targets from a compendium of reference profiles (e.g. genetic interactions and gene expression)[37, 38, 26, 30]. This approach requires the accumulation of accessible knowledgebases; it also depends on the availability of genome-wide, diverse datasets as well as the availability of standardized protocols.

1.4

Organization

In this dissqrtation, I describe 4 projects that sought to advance the study of chemical and gene function using yeast as a model organism. In Chapter 2, I describe a cost-effective, extensible platform that integrates three distinct assays with the goal of improving the characterization of the bioactivity of small-molecules and making the technology accessible. This platform is built on the robust TAG4 barcode microarray [42] and simultaneously resolves the fitness of strains derived from pools of 1) homozygous deletion mutants, 2) heterozygous deletion mutants, and 3) genomic library transformants. This approach was used to characterize several different reference compounds and a library of uncharacterized small molecules. Combining the results of both haploinsufficiency profiling and multicopy suppressor profiling allows the direct molecular target to be distinguished from genes that interact with the target. In addition, the homozygous deletion profiles reveal additional effects that compounds exert on cells, further narrowing down the candidates of the potential drug target. A promising new application of chemogenomic profiles is to advance our understanding of multiple-perturbations, e.g. multi-component therapies. In Chapter 3, I describe the methods I have developed in yeast to screen drug combinations systematically and to identify genetic mutants that modifies the cellular response to drug combinations thereby allowing the mechanistic basis for the interaction to be inferred. This work described in Chapter 2 and 3 has been published[25]. Glyoxal and methyglyoxal are reactive carbonyls, formed as by-products of metabolism and found in a large number of environmental sources such as cigarette smoke, automobile exhaust and thermal processing of food. In Chapter 4,1 describe the results of using multiple genomewide assays to identify the genetic basis for carbonyl stress resistance. This work is in the process of being submitted for publication. Gain-of-function (GOF) genetic screens are a powerful way of generating phenotypes and have been used with great success in multiple biological systems. With an estimated 2/3 of all genes having no loss-of-function phenotypes in Saccharomyces cerevisiae or Caenorhabditis elegans [15, 28], GOF approaches are therefore complementary to genetic knockout or knockdown screens. This has direct relevance to the study of human diseases caused by gene hyperactivation or amplification. In Chapter 5, I describe the generation of a yeast pool expressing 12,000 different human genes, demonstrate that toxic human genes can be identified via a competitive growth assay and discuss ways in which these results could be used for future chemogenomics studies. Finally in Chapter 6,1 summarize the

CHAPTER

1.

INTRODUCTION

5

major findings in this thesis and discuss some possible avenues for future studies.

1.5

Bibliography

[1] A. Arnoldo, J. Curak, S. Kittanakom, I. Chevelev, V. T. Lee, M. Sahebol-Amri, B. Koscik, L. Ljuma, P. J. Roy, A. Bedalov, G. Giaever, C. Nislow, R. A. Merrill, S. Lory, and I. Stagljar. Identification of small molecule inhibitors of pseudomonas aeruginosa exoenzyme s using a yeast phenotypic screen. PLoS genetics, 4(2):el000005, Feb 2008. [2] D. Botstein, S. A. Chervitz, and J. M. Cherry. Yeast as a model organism. Science; Science, 277(5330):1259-1260, Aug 29 1997. [3] H. Breithaupt. The costs of reach, reach is largely welcomed, but the requirement to test existing chemicals for adverse effects is not good news for all. EMBO reports, 7(10):968-971, Oct 2006. [4] J. A. Brown, G. Sherlock, C. L. Myers, N. M. Burrows, C. Deng, H. I. Wu, K. E. McCann, O. G. Troyanskaya, and J. M. Brown. Global analysis of gene function in yeast by quantitative phenotypic profiling. Molecular systems biology, 2:2006.0001, 2006. [5] V. J. Brown. Reaching for chemical safety. Environmental health perspectives, 111(14) :A766-9, Nov 2003. LR: 20041117; PUBM: Print; JID: 0330411; 0 (Hazardous Substances); CIN: Environ Health Perspect. 2004 Apr;112(5):A269. PMID: 15064184; ppublish. [6] R. A. Butcher, B. S. Bhullar, E. O. Perlstein, G. Marsischky, J. LaBaer, and S. L. Schreiber. Microarray-based method for monitoring yeast overexpression strains reveals small-molecule targets in tor pathway. Nat.Chem.Biol,

2(2):103-109, Feb 2006.

[7] K. S. Chung, Y. J. Jang, N. S. Kim, S. Y. Park, S. J. Choi, J. Y. Kim, J. H. Ahn, H. J. Lee, J. H. Lim, J. H. Song, J. H. Ji, J. H. Oh, K. B. Song, H. S. Yoo, and M. Won. Rapid screen of human genes for relevance to cancer using fission yeast. Journal of Biomolecular Screening : the official journal of the Society for Biomolecular Screening, 12(4):568-577, Jun 2007. [8] F. S. Collins, G. M. Gray, and J. R. Bucher. Toxicology, transforming environmental health protection. Science (New York, N.Y.), 319(5865):906-907, Feb 15 2008. [9] L. David, W. Huber, M. Granovskaia, J. Toedling, C. J. Palm, L. Bofkin, T. Jones, R. W. Davis, and L. M. Steinmetz. A high-resolution map of transcription in the yeast genome. Proceedings of the National Academy of Sciences of the United States of America, 103(14):5320-5325, Apr 4 2006.

CHAPTER

1.

INTRODUCTION

6

[10] R. K. Dorer, S. Zhong, J. A. Tallarico, W. H. Wong, T. J. Mitchison, and A. W. Murray. A small-molecule inhibitor of mpsl blocks the spindle-checkpoint response to a lack of tension on mitotic chromosomes. Current biology : CB, 15(11):1070-1076, Jun 7 2005. [11] A. M. Dudley, D. M. Janse, A. Tanay, R. Shamir, and G. M. Church. A global view of pleiotropy and phenotypically derived gene function in yeast. Molecular systems biology, 1:2005.0001, 2005. [12] J. Fleming, T. F. Outeiro, M. Slack, S. L. Lindquist, and C. E. Bulawa. Detection of compounds that rescue rabl-synuclein toxicity. Methods in enzymology, 439:339-351, 2008. [13] S. Frantz. Pharma faces major challenges after a year of failures and heated battles. Nature reviews.Drug discovery, 6(l):5-7, Jan 2007. [14] A. P. Gasch, P. T. Spellman, C. M. Kao, O. Carmel-Harel, M. B. Eisen, G. Storz, D. Botstein, and P. O. Brown. Genomic expression programs in the response of yeast cells to environmental changes. Molecular biology of the cell, ll(12):4241-4257, Dec 2000. [15] G. Giaever, A. M. Chu, L. Ni, G. Connelly, L. Riles, S. Veronneau, S. Dow, A. Lucau-Danila, K. Anderson, B. Andre, A. P. Arkin, A. Astromoff, M. El-Bakkoury, R. Bangham, R. Benito, S. Brachat, S. Campanaro, M. Curtiss, K. Davis, A. Deutschbauer, K. D. Entian, P. Flaherty, F. Foury, D. J. Garfmkel, M. Gerstein, D. Gotte, U. Guldener, J. H. Hegemann, S. Hempel, Z. Herman, D. F. Jaramillo, D. E. Kelly, S. L. Kelly, P. Kotter, D. LaBonte, D. C. Lamb, N. Lan, H. Liang, H. Liao, L. Liu, C. Luo, M. Lussier, R. Mao, P. Menard, S. L. Ooi, J. L. Revuelta, C. J. Roberts, M. Rose, P. Ross-Macdonald, B. Scherens, G. Schimmack, B. Shafer, D. D. Shoemaker, S. Sookhai-Mahadeo, R. K. Storms, J. N. Strathern, G. Valle, M. Voet, G. Volckaert, C. Y. Wang, T. R. Ward, J. Wilhelmy, E. A. Winzeler, Y. Yang, G. Yen, E. Youngman, K. Yu, H. Bussey, J. D. Boeke, M. Snyder, P. Philippsen, R. W. Davis, and M. Johnston. Functional profiling of the saccharomyces cerevisiae genome. Nature, 418(6896):387-391, Jul 25 2002. [16] G. Giaever, P. Flaherty, J. Kumm, M. Proctor, C. Nislow, D. F. Jaramillo, A. M. Chu, M. I. Jordan, A. P. Arkin, and R. W. Davis. Chemogenomic profiling: identifying the functional interactions of small molecules in yeast. Proceedings of the National Academy of Sciences of the United States of America, 101(3):793-798, Jan 20 2004. [17] G. Giaever, D. D. Shoemaker, T. W. Jones, H. Liang, E. A. Winzeler, A. Astromoff, and R. W. Davis. Genomic profiling of drug sensitivities via induced haploinsufficiency. Nature genetics; Nature genetics, 21(3):278-283, Mar 1999. [18] A. Goffeau, B. G. Barrell, H. Bussey, R. W. Davis, B. Dujon, H. Feldmann, F. Galibert, J. D. Hoheisel, C. Jacq, M. Johnston, E. J. Louis, H. W. Mewes, Y. Murakami, P. Philippsen, H. Tettelin, and S. G. Oliver. Life with 6000 genes. Science (New York, N.Y.), 563-7, Oct 25 1996.

274(5287):546,

CHAPTER

1.

INTRODUCTION

7

[19] D. Gresham, D. M. Ruderfer, S. C. Pratt, J. Schacherer, M. J. Dunham, D. Botstein, and L. Kruglyak. Genome-wide detection of polymorphisms at nucleotide resolution with a single dna microarray. Science (New York, N. Y.), 311(5769):1932-1936, Mar 31 2006. [20] L. H. Hartwell. Saccharomyces cerevisiae cell cycle. Bacteriological Reviews, 38(2):164-198, Jun 1974. [21] L. H. Hartwell, J. Culotti, J. R. Pringle, and B. J. Reid. Genetic control of the cell division cycle in yeast. Science (New York, N.Y.), 183(120):46-51, Jan 11 1974. [22] J. Heitman, N. R. Movva, P. C. Hiestand, and M. N. Hall. Fk 506-binding protein proline rotamase is a target for the immunosuppressive agent fk 506 in saccharomyces cerevisiae. Proceedings of the National Academy of Sciences of the United States of America, 88(5):1948-1952, Mar 1 1991. [23] M. E. Hillenmeyer, E. Fung, J. Wildenhain, S. E. Pierce, S. Hoon, W. Lee, M. Proctor, R. P. St Onge, M. Tyers, D. Roller, R. B. Altman, R. W. Davis, C. Nislow, and G. Giaever. The chemical genomic portrait of yeast: uncovering a phenotype for all genes. Science (New York, N. Y.), 320(5874):362-365, Apr 18 2008. [24] T. Hofer, I. Gerner, U. Gundert-Remy, M. Liebsch, A. Schulte, H. Spielmann, R. Vogel, and K. Wettig. Animal testing and alternative approaches for the human health risk assessment under the proposed new european chemicals regulation. Archives of Toxicology, 78(10):549-564, Oct 2004. [25] S. Hoon, A. M. Smith, I. M. Wallace, S. Suresh, M. Miranda, E. Fung, M. Proctor, K. M. Shokat, C. Zhang, R. W. Davis, G. Giaever, R. P. StOnge, and C. Nislow. An integrated platform of genomic assays reveals small-molecule bioactivities. Nature chemical biology, 4(8):498-506, Aug 2008. [26] T. R. Hughes, M. J. Marton, A. R. Jones, C. J. Roberts, R. Stoughton, C. D. Armour, H. A. Bennett, E. Coffey, H. Dai, Y. D. He, M. J. Kidd, A. M. King, M. R. Meyer, D. Slade, P. Y. Lum, S. B. Stepaniants, D. D. Shoemaker, D. Gachotte, K. Chakraburtty, J. Simon, M. Bard, and S. H. Friend. Functional discovery via a compendium of expression profiles. Cell, 102(1):109126, Jul 7 2000. [27] B. Jiang, D. Xu, J. Allocco, C. Parish, J. Davison, K. Veillette, S. Sillaots, W. Hu, R. RodriguezSuarez, S. Trosok, L. Zhang, Y. Li, F. Rahkhoodaee, T. Ransom, N. Martel, H. Wang, D. Gauvin, J. Wiltsie, D. Wisniewski, S. Salowe, J. N. Kahn, M. J. Hsu, R. Giacobbe, G. Abruzzo, A. Flattery, C. Gill, P. Youngman, K. Wilson, G. Bills, G. Platas, F. Pelaez, M. T. Diez,

CHAPTER 1. INTRODUCTION

8

S. Kauffman, J. Becker, G. Harris, P. Liberator, and T. Roemer. Pap inhibitor with in vivo efficacy identified by Candida albicans genetic profiling of natural products. Chemistry & biology, 15(4):363-374, Apr 2008. [28] R. S. Kamath, A. G. Eraser, Y. Dong, G. Poulin, R. Durbin, M. Gotta, A. Kanapin, N. Le Bot, S. Moreno, M. Sohrmann, D. P. Welchman, P. Zipperlen, and J. Ahringer. Systematic functional analysis of the caenorhabditis elegans genome using rnai. Nature, 421(6920):231-237, Jan 16 2003. [29] V. Krishnan, L. Dirick, H. H. Lim, T. S. Lim, S. L. Si-Hoe, C. S. Cheng, K. L. Yap, A. Ting, E. Schwob, and U. Surana. A novel cell cycle inhibitor stalls replication forks and activates s phase checkpoint. Cell cycle (Georgetown, Tex.), 6(13):1621-1630, Jul 1 2007. [30] C. Kung, D. M. Kenski, S. H. Dickerson, R. W. Howson, L. F. Kuyper, H. D. Madhani, and K. M. Shokat. Chemical genomic profiling to identify intracellular targets of a multiplex kinase inhibitor. Proceedings of the National Academy of Sciences of the United States of America; Proceedings of the National Academy of Sciences of the United States of America, 102(10):35873592, Mar 8 2005. [31] C. R. Landry, J. P. Townsend, D. L. Hartl, and D. Cavalieri. Ecological and evolutionary genomics of saccharomyces cerevisiae. Molecular ecology, 15(3):575-591, Mar 2006. [32] D. A. Lashkari, J. L. DeRisi, J. H. McCusker, A. F. Namath, C. Gentile, S. Y. Hwang, P. O. Brown, and R. W. Davis. Yeast microarrays for genome wide parallel genetic and gene expression analysis. Proceedings of the National Academy of Sciences of the United States of America, 94(24):13057-13062, Nov 25 1997. [33] W. Lee, D. Tillo, N. Bray, R. H. Morse, R. W. Davis, T. R. Hughes, and C. Nislow. A highresolution atlas of nucleosome occupancy in yeast. Nature genetics, 39(10):1235-1244, Oct 2007. [34] William Lee, Robert P. St Onge, Michael Proctor, Patrick Flaherty, Michael I. Jordan, Adam P. Arkin, Ronald W. Davis, Corey Nislow, and Guri Giaever. Genome-wide requirements for resistance to functionally distinct dna-damaging agents. PLoS Genet, 1(2), 08:19: 2005. [35] H. Luesch, T. Y. Wu, P. Ren, N. S. Gray, P. G. Schultz, and F. Supek. A genome-wide overexpression screen in yeast for small-molecule target identification.

Chemistry & biology,

12(l):55-63, Jan 2005. [36] P. Y. Lum, C. D. Armour, S. B. Stepaniants, G. Cavet, M. K. Wolf, J. S. Butler, J. C. Hinshaw, P. Gamier, G. D. Prestwich, A. Leonardson, P. Garrett-Engele, C. M. Rush, M. Bard, G. Schimmack, J. W. Phillips, C. J. Roberts, and D. D. Shoemaker. Discovering modes of

CHAPTER

1.

INTRODUCTION

9

action for therapeutic compounds using a genome-wide screen of yeast heterozygotes. Cell, 116(1):121-137, Jan 9 2004. [37] A. B. Parsons, R. L. Brost, H. Ding, Z. Li, C. Zhang, B. Sheikh, G. W. Brown, P. M. Kane, T. R. Hughes, and C. Boone. Integration of chemical-genetic and genetic interaction data links bioactive compounds to cellular target pathways. Nature biotechnology, 22(l):62-69, Jan 2004. [38] A. B. Parsons, A. Lopez, I. E. Givoni, D. E. Williams, C. A. Gray, J. Porter, G. Chua, R. Sopko, R. L. Brost, C. H. Ho, J. Wang, T. Ketela, C. Brenner, J. A. Brill, G. E. Fernandez, T. C. Lorenz, G. S. Payne, S. Ishihara, Y. Ohya, B. Andrews, T. R. Hughes, B. J. Prey, T. R. Graham, R. J. Andersen, and C. Boone. Exploring the mode-of-action of bioactive compounds by chemical-genetic profiling in yeast. Cell, 126(3):611-625, Aug 11 2006. [39] L. Pena-Castillo and T. R. Hughes. Why are there still over 1000 uncharacterized yeast genes? Genetics, 176(1):7-14, May 2007. [40] E. Perkins, D. Sun, A. Nguyen, S. Tulac, M. Francesco, H. Tavana, H. Nguyen, S. Tugendreich, P. Barthmaier, J. Couto, E. Yeh, S. Thode, K. Jarnagin, A. Jain, D. Morgans, and T. Melese. Novel inhibitors of poly(adp-ribose) polymerase:parpl and parp2 identified using a cell-based screen in yeast. Cancer research, 61(10):4175-4183, May 15 2001. [41] E. O. Perlstein, D. M. Ruderfer, G. Ramachandran, S. J. Haggarty, L. Kruglyak, and S. L. Schreiber. Revealing complex traits with small molecules and naturally recombinant yeast strains. Chemistry & biology, 13(3):319-327, Mar 2006. [42] S. E. Pierce, E. L. Fung, D. F. Jaramillo, A. M. Chu, R. W. Davis, C. Nislow, and G. Giaever. A unique and universal molecular barcode array. Nat.Methods, 3(8):601-603, Aug 2006. [43] J. Rine, W. Hansen, E. Hardeman, and R. W. Davis. Targeted selection of recombinant clones through' gene dosage effects. Proceedings of the National Academy of Sciences of the United States of America, 80(22):6750-6754, Nov 1983. [44] R. Rodriguez-Suarez, D. Xu, K. Veillette, J. Davison, S. Sillaots, S. Kauffman, W. Hu, J. Bowman, N. Martel, S. Trosok, H. Wang, L. Zhang, L. Y. Huang, Y. Li, F. Rahkhoodaee, T. Ransom, D. Gauvin, C. Douglas, P. Youngman, J. Becker, B. Jiang, and T. Roemer.

Mechanism-of-

action determination of gmp synthase inhibitors and target validation in Candida albicans and aspergillus fumigatus. Chemistry & biology, 14(10):1163-1175, Oct 2007. [45] D. R. Scannell, G. Butler, and K. H. Wolfe. Yeast genome evolution-the origin of the species. Yeast (Chichester, England), 24(ll):929-942, Nov 2007. [46] T. Stearns. Lewis-sigler institute seminar: Yeast genomics - an ames test for the 21st century?, June 09 2005.

CHAPTER

1.

INTRODUCTION

10

[47] E. J. Topol. Failing the public health-rofecoxib, merck, and the fda. The New England journal of medicine, 351(17):1707-1709, Oct 21 2004. [48] I. J. Tsai, D. Bensasson, A. Burt, and V. Koufopanou. Population genomics of the wild yeast saccharomyces paradoxus: Quantifying the life cycle. Proceedings of the National Academy of Sciences of the United States of America, 105(12):4957-4962, Mar 25 2008. [49] E. Vottero, A. Balgi, K. Woods, S. Tugendreich, T. Melese, R. J. Andersen, A. G. Mauk, and M. Roberge. Inhibitors of human indoleamine 2,3-dioxygenase identified with a target-based screen in yeast. Biotechnology journal, l(3):282-288, Mar 2006. [50] E. A. Winzeler, D. D. Shoemaker, A. Astromoff, H. Liang, K. Anderson, B. Andre, R. Bangham, R. Benito, J. D. Boeke, H. Bussey, A. M. Chu, C. Connelly, K. Davis, F. Dietrich, S. W. Dow, M. El Bakkoury, F. Foury, S. H. Friend, E. Gentalen, G. Giaever, J. H. Hegemann, T. Jones, M. Laub, H. Liao, N. Liebundguth, D. J. Lockhart, A. Lucau-Danila, M. Lussier, N. M'Rabet, P. Menard, M. Mittmann, C. Pai, C. Rebischung, J. L. Revuelta, L. Riles, C. J. Roberts, P. Ross-MacDonald, B. Scherens, M. Snyder, S. Sookhai-Mahadeo, R. K. Storms, S. Veronneau, M. Voet, G. Volckaert, T. R. Ward, R. Wysocki, G. S. Yen, K. Yu, K. Zimmermann, P. Philippsen, M. Johnston, and R. W. Davis. Functional characterization of the s. cerevisiae genome by gene deletion and parallel analysis. Science (New York, N.Y.),

285(5429):901-906,

Aug 6 1999. [51] D. Xu, B. Jiang, T. Ketela, S. Lemieux, K. Veillette, N. Martel, J. Davison, S. Sillaots, S. Trosok, C. Bachewich, H. Bussey, P. Youngman, and T. Roemer. Genome-wide fitness test and mechanism-of-action studies of inhibitory compounds in Candida albicans. PLoS pathogens, 3(6):e92, Jun 2007.

Chapter 2

A n integrated platform of genomic assays reveals small molecule bioactivities 2.1

Abstract

Bioactive compounds are widely used to modulate protein function and can serve as important leads for drug development. Identifying the in vivo targets of these compounds remains a challenge. Using yeast, we integrated three genomewide gene-dosage assays to measure the effect of small molecules in vivo. A single TAG microarray is used to resolve the fitness of strains derived from pools of 1) homozygous deletion mutants, 2) heterozygous deletion mutants, and 3) genomic library transformants. We demonstrated, with eight diverse reference compounds, that integration of these three chemogenomic profiles improves the sensitivity and specificity of small-molecule target identification. We further dissected the mechanism of action of two protein phosphatase inhibitors. Finally, we applied this platform to 188 novel synthetic chemical compounds identifying both potential targets and structure-activity relationships. The work described here have been published[19]. Supplementary Data is available at http://chemogenomics.stanford.edu/supplements/05chemo/ I was involved in this study from the beginning and was a primary contributor in all aspects of the study.

11

CHAPTER 2. AN INTEGRATED

2.2

CHEMOGENOMICS

PLATFORM

12

Introduction

Identifying the mechanism of action of bioactive compounds in vivo is essential for chemical biology and drug discovery. To realize this goal, a number of approaches using cell-based assays to carry out forward chemical-genetic screens in yeast have been developed. For example, drug-induced Haplolnsufficency Profiling (HIP) was developed to identify small-molecules that target essential genes[13, 33]. Similarly, Homozygous Profiling (HOP), uses the yeast homozygous (or haploid) deletion collection to identify genetic modifiers of drug resistance[29, 40]. Diverse multicopy suppression strategies have been described that identify genes that confer resistance to chemical treatment when over-expressed [43, 32, 9]. Other strategies rely on 'guilt-by-association', inferring targets from a compendium of reference profiles (e.g. genetic interactions and gene expression) [39, 21, 26, 27]. Each method has limitations and here we present a cost-effective method that integrates three distinct assays with the goal of improving the characterization of the bioactivity of small-molecules and making this technology accessible. Using a miniaturized screening procedure in conjunction with a common barcode TAG microarray[42, 41], this platform systematically examines the effect of both increasing and decreasing gene dosage on chemical sensitivity. We demonstrate its utility for small-molecule characterization with 9 well-studied chemical compounds with distinct mechanism of actions (Table 2.4). We successfully identified both known and novel molecular targets for several of these compounds and note that examining the effect of both increasing and decreasing gene dosage was often required to distinguish the bona fide target from a longer list of potential candidates. This multi-pronged approach, combined with genome-wide drug combination screens, uncovered important in vivo differences between the phosphatase inhibitors cantharidin and calyculin A. We also identified a previously uncharacterized gene, which we dub CRGl (Cantharidin Resistance Gene 1), as a dose-dependent regulator of resistance to cantharidin, a natural product isolated from blister beetles. Extension of multicopy suppression profiling identified a potential CRGl ortholog in the fungal pathogen Candida albicans (orfl9.63S),

underscoring the flexibility of this assay and the

utility of small molecules for annotating gene function. Finally, we applied this platform to forward chemogenomics by screening 188 synthetic chemical compounds with no previously known biological activity. This discovery effort uncovered several potential protein targets and revealed new structureactivity relationships for these compounds.

2.3

Results

2.3.1

A Chemogenomics Platform

The yeast deletion collection is a proven resource for chemical-genetics [8]. Each strain in these collections-contains a unique 20-bp DNA tag that enables the fitness of individual strains (from a heterogeneous pool), to be resolved simultaneously using an oligonucleotide array. Employing

CHAPTER

2. AN INTEGRATED

CHEMOGENOMICS

PLATFORM

13

a pooled strategy has the distinct advantage over strain-by-strain experiments because it drastically reduces compound usage and simplifies sample processing. Exploiting an array that contains both tag and ORF probes [42], we developed a chemogenomics platform to resolve the fitness of strains derived from pools of 1) homozygous deletion mutants, 2) heterozygous deletion mutants, and 3) genomic library transformants (Fig. 2.1a). We refer to those screens that use the deletion pools as Deletion Sensitivity Profiling (DSP) and screens that use genomic library transformants as Multicopy Suppression Profiling (MSP). MSP is a novel genome-wide assay that screens DNA clone libraries competitively to identify genes that confer resistance to compounds when overrepresented. Traditional multicopy suppressor screens involve cumbersome plating techniques and clone characterization [43] but more recently, a microarray-based approach was used to characterize smallmolecule modulators of the TOR pathway[9]. Here, we apply a variation of this concept, by using a high-copy, random genomic library (rather than an inducible ORF library) and a simplified means to amplify library clones (Fig. 2.2a). A genomic library avoids the gene-dose differential seen for different ORFs and circumvents the need to optimize carbon-source induction parameters. Perhaps more importantly, approximately 15% of yeast genes are known to be toxic when their expression are induced by galactose [46]. Genomic libraries are also widely available and easy to construct to high levels of genomic-coverage. We have also miniaturized the assay such that full-genome screens are performed in only 700/xl volume thus greatly reducing compound usage, an important consideration for compounds that are limited in supply or prohibitively expensive. Coverage of the pool was estimated by counting the number of probes for each ORF that were above background chip intensity (Fig. 2.2b). To demonstrate the efficacy of MSP, we performed a genetic complementation screen using a well-characterized, highly-specific kinase inhibitor. The mutation of the highly conserved 'gatekeeper' residue in the ATP binding pocket of the CDC28 kinase renders the allele specifically sensitive to inhibition by the ATP analog 1-NM-PP1[5]. A multicopy suppressor pool was generated in a heterozygous diploid strain that contained a single cdc28 analog sensitive allele (and no other endogenous source of CDC28) and screened with 1-NM-PP1 (Fig. 2.2c). CDC28, and three other genes that flank the CDC28 locus and were likely amplified from the same genomic clone (CSH1, IFA38 and TOS1), were identified as suppressors of 1-NM-PP1 (Fig. 2.2d). Even though the genomic library insert (~5kb in size) may often contain several genes from the same linkage group, integrating deletion sensitivity data (see Section 2.3.2) can often resolve the true genetic modifier of drug resistance. We also used the engineered target, cdc28-AS, to assess the noise in the DSP assay. The cdc28AS strain described above, containing a unique barcode, was added to the heterozygous deletion pool and screened with 1-NM-PP1. Because only one strain in the pool contains the analog sensitive allele, we expected only this strain to be sensitive to the inhibitor and by extension, any other strain identified would be spurious. As predicted, only the cdc28-AS strain was highly sensitive to the inhibitor (Fig. 2.2e). In contrast, the inhibitor had no effect on the heterozygous deletion strain


CHEMOGENOMICS

PLATFORM

14

containing the wildtype copy of CDC28 or any of the other 6000 heterozygous strains in the pool. Thus we conclude that both assays have very low technical noise.

2.3.2

Integrating gene-dose assays improves target identification

Several studies have demonstrated that increasing or decreasing the abundance of a small molecule's target will directly affect the sensitivity of the cell to that small-molecule[40, 14, 32, 30, 43, 9]. We applied both DSP and MSP to eight other reference compounds that have distinct mechanisms of action (Table 2.4). The results are presented such that each ORF is represented as a vector in a Cartesian plane whereby the y-coordinate represents the level of multicopy suppression and the x-coordinate represents the level of deletion sensitivity in response to chemical exposure (Fig. 2.1b). Vector analysis [7] was used to identify strains (colored red or blue) that were significant in both assays (see Methods). The plots are divided into eight different sectors to represent different response patterns; in this study we focused on three sectors in particular. The green sector identifies genes that are either bona fide suppressors or genes that are linked to these suppressors (genomic clones often contain multiple genes), but do not cause sensitivity when deleted. The red sector identifies genes that are sensitive to the chemical when deleted (homozygous or heterozygous) but are not multi-copy suppressors. The yellow sector contains genes that result in sensitivity when deleted and result in suppression when overexpressed. Essential genes (red) in this sector represent potential targets while non-essential genes (blue) represent genetic modifiers of drug resistance. Methotrexate, fluconazole and rapamycin are widely prescribed drugs with well-characterized targets and complex mechanisms of action. Methotrexate inhibits folic acid biosynthesis by targeting dihydrofolate reductase (encoded by DFR1) [36]. Fluconazole inhibits ergosterol biosynthesis by inhibiting the cytochrome P450 lanosterol 14a-demethylase encoded by ERGll[24]. acts by binding to the immunophilin FKBP12 and together, the FKBP12-rapamycin

Rapamycin complex binds

to and inhibits the TOR proteins [50]. Consistent with these studies, our experiments identified DFR1, ERG11 and TOR2 as both haploinsufficient and multicopy suppressors of methotrexate, fluconazole and rapamycin, respectively (Fig. 2.3a-c). We note that in all cases, integrating deletion sensitivity with multicopy suppression greatly refined the list of potential targets. For example, in each case, the most sensitive deletion strain determined by DSP did not represent the respective target (although the bona fide target was always one of the top ten most sensitive strains). Genomewide results for MMS, latrunculin A and nocodazole are presented in Fig. 2.4a-c. Surprisingly, the amplicons produced following a screen with rapamycin, the major DNA fragment obtained contained not the full-length TOR2 gene but rather only the C-terminal portion of the genes, EAP1 and TOR2 (Fig. 2.4d). Because this C-terminal fragment of the TOR2 gene included the FifBi^-rapamycin binding domain (FRB), this result implied that overexpression of this domain alone was sufficient to confer rapamycin resistance. We tested this hypothesis directly by expressing a fragment of TOR2 lacking the FRB domain and showed that loss of the FRB domain

CHAPTER

2. AN INTEGRATED

CHEMOGENOMICS

PLATFORM

15

abolished rapamycin resistance (Fig. 2.4d). While this observation is not surprising, it underscores a benefit of performing MSP with DNA fragments, i.e. to define the ligand-binding sites of proteins.

2.3.3


Cantharidin and calyculin A are naturally occurring toxins that each inhibit both type I (PP1) and type IIA (PP2A) protein phosphatases[18, 31, 22]. These inhibitors possess modest in vitro specificity for each protein with cantharidin being more selective for PP2A[12]. We screened both compounds using our integrated platform and found that Glc7, the only essential type I protein phosphatase in yeast, was haploinsufficient in both cantharidin and calyculin A. Strikingly, overexpression of GLC7 conferred resistance to calyculin A but not cantharidin (Fig. 2.5a-b). These results were confirmed using isogenic cultures (Fig. 2.5c). Next, we used an in vitro assay to measure direct binding of each inhibitor to Glc7 by testing the ability of each inhibitor to disrupt the interaction between Glc7 and microcystin-LR, a non-cell permeable inhibitor of Glc7[34]. These results revealed that calyculin A, but not cantharidin was an effective competitor for Glc7 binding. (Fig. 2.5d). Collectively, these results indicate that calyculin A is a more effective inhibitor of the yeast Glc7 protein in vivo than cantharidin. Cantharidin and calyculin A represent the most mechanistically similar compounds in our reference set (Table 2.4). This is not, however, supported by the deletion sensitivity profiles which revealed that the cantharidin and calyculin A profiles are among the most dissimilar in our dataset (Fig. 2.6a). The difference between the two profiles is also reflected by the functional enrichment of strains that are significantly sensitive to each compound (Fig. 2.6b). These results underscore significant differences between the in vivo effects of these two phosphatase inhibitors. Further inspection of their deletion sensitivity profiles revealed that each compound identified different GLC7-related genes. For example, strains heterozygous for components of the APT complex (PTAl, SWD2), an essential subcomplex of the multicomplex holo-CPF containing the yeast cleavage and polyadenylation factor [48, 25] that includes Glc7, were sensitive to both compounds. In contrast, GLC7 is also known to promote cell wall integrity, bud morphology and polarization of the actin cytoskeleton through the PKC-regulated SLT2 MAP kinase pathway[l] and strains deficient in various components of PKC signaling (BCK1, SLT2, RLM1) were uniquely sensitive to cantharidin. In an independent confirmation of the observed genome-wide differences, we found that three glc7 alleles, each known to affect different GLC7-dependent processes [4, 20, 6], were differentially sensitive to each inhibitor (Fig. 2.6c). glc7 alleles defective in nuclear pathways (glc7-127, glc7-129) were more sensitive to either compound compared to a glc7 allele that is defective in cytoplasmic function (glc7-109) (Fig. 2.6c). This allele-specificity is consistent with the enrichment of nuclear functions (mRNA processing and chromatin modification) that we observed in the genome-wide deletion sensitivity screens. Importantly, glc7-129 was more sensitive to cantharidin than glc7-127, while the opposite was true for calyculin A (Fig. 2.6c).

CHAPTER 2. AN INTEGRATED CHEMOGENOMICS PLATFORM

DSP of cantharidin showed that homozygous deletions in YHR209W

16

(CRG1), a non-essential

gene encoding a putative S-adenosylmethionine(SAM)-dependent methyltransferase of unknown function, sensitized yeast to cantharidin (Fig.

2.5a).

This gene was also haploinsufficient in

cantharidin[16] and was an effective multicopy suppressor of cantharidin sensitivity (Fig. 2.5a,c). Homology comparisons revealed that CRGl, like other SAM-dependent methlytransferases is poorly conserved outside of a highly conserved structural fold [35]. To identify a functional homolog of CRGl, we modified MSP by introducing a C. albicans genomic library into S. cerevisiae to generate a pool of multi-copy transformants.

This pool was grown in the presence of calyculin A

and cantharidin and resistant clones were identified by hybridization to C. albicans expression arrays. Consistent with our screens using a S. cerevisiae library, the C. albicans ortholog of GLC7 (orfl9.6285)

was identified as the major suppressor of calyculin A-induced sensitivity (Fig. 2.7b).

The predominant suppressor clone identified in the presence of cantharidin contained four genes of unknown function (Fig. 2.7b). The highest scoring gene in this clone (and indeed the entire experiment) was orfl9.633, annotated as a putative methyltransferase with no ortholog in S. cerevisiae based on sequence similarity [2]. Nonetheless, the methyltransferase domain shared between CRGl and orf 19.633 suggests they are functional homologs, and indeed a clone that contained only orf 19.633 conferred even greater cantharidin resistance than a clone containing S.cerevisiae CRGl (Fig. 2.7c).

2.3.4


To demonstrate the utility of these integrated assays in understanding the bioactivity of a more challenging chemical set, we screened 188 synthetic chemical compounds with uncharacterized activities. These compounds were selected from a library of compounds that have inhibitory activity against yeast. An activity score (See Methods), designed to identify genes for which copy number changes conferred potent and specific sensitivity in DSP and potent and specific resistance in MSP, was calculated for each compound-gene pair yielding >10 6 activity scores. Many of the known interactions from our reference compounds were among the highest scores in this list (Table 2.5) (suggesting that the reference compounds were specific for their targets). The highest activity score among the synthetic compounds was for 4130-1278 and the gene SEC14, which encodes a conserved phosphatidylinositol/phosphatidylcholine transfer protein (Table 2.5, Fig. 2.8a). These screen results were confirmed with isogenic cultures and suggest that Secl4 is a target of 4130-1278 (Fig. 2.8b). Moreover, SEC14 did not score highly in both MSP or DSP with the other 195 compounds we screened, indicating that 4130-1278 was unique in its ability to target SEC14 (Fig. 2.8c). We also identified a potential novel inhibitor of Ergll, 4513-0042 (Table 2.5, Fig. 2.9a). 45130042 contains an azole ring, a hallmark of many characterized E r g l l inhibitors, and specifically identified ERGll

in both MSP and DSP assays (Fig. 2.9a). Comparing the DSP and MSP activity


17

score for ERG 11 across all 196 compounds revealed that a number of compounds displayed antiE r g l l activity in either MSP or DSP alone but not in both. Only 4513-0042 and fluconazole displayed activity against E r g l l in both assays (Fig. 2.9b).

2.3.5

Comparison of profiles to compound structure

Giaever et al. showed that compounds that share a common substructure had similar chemogenomic profiles[13]. More recently, a large-scale study showed that phenotype-based structure-activity relationships (SARs) could be derived using cytological phenotypes from high-content screening [49]. We therefore asked to what extent chemical structural similarities are reflected in our DSP and MSP results. For each compound, a circular molecular fingerprint was defined using ECFP.4 descriptors. A similarity matrix, based on Tanimoto scores, was used to describe the pair-wise relationships between the 196 compounds screened. We also generated chemogenomic profile similarity matrices, one for DSP and one for MSP, by calculating the Pearson's correlation between each pair of experiments. Each matrix is presented as a heatmap with compounds ordered by hierarchical clustering of the DSP profiles (Fig. 2.10a). We found the relationship between chemical structure and profiling results to be significant for both DSP and MSP, with DSP exhibiting a higher correlation with structure than MSP (Fig. 2.10b). Nevertheless, it is evident from the heatmaps that while some structurally similar chemicals produced similar DSP and/or MSP results (red boxes), other structurally dissimilar compounds also produced highly correlated profiles (green boxes). To examine how small changes in chemical structure can affect function or bioactivity, we compared Tanimoto similarities with DSP profile similarities between each compound pair. Using a Tanimoto structural similarity score cutoff of > 0.3, we found that ~80% of the compound pairs with similar structures showed concordance in their chemogenomic profiles while 20% of the compound pairs showed discordance (Fig. 2.11a). Four exemplar discordant compound pairs are shown in Figure 2.11a. Compounds k064-0027 and k064-0035 for example, differ only in one position, but differ greatly in their respective DSP profiles. These compounds are structurally similar to a family of 9-anilinoacridine derivatives that include m-AMSA (and its analogues) which have been extensively studied as potential antitumor agents [3]. m-AMSA has been shown to interact with DNA and also to inhibit topoisomerase II [37]. k064-0027 shared the greatest structural similarity with m-AMSA among these derivatives (Table 2.3) and comparison of DSP results for k064-0027, k0640035, and three other derivatives in our dataset, showed that strains deficient in DNA repair were sensitive only to k064-0027 (Fig. 2.11b). This observation is consistent with the methyl ester group unique to k064-0027, as QSAR analysis of 643 members of the 9-anilinoacridine family showed that electron-donating groups were favored on the aniline ring [10]. Two of the compounds that do not induce a DNA damage response (k064-0012, k064-0035) both contain two halogen substituents, and it is known from the QSAR study[10] that electron-withdrawing groups reduce antitumor activity


CHEMOGENOMICS

PLATFORM

18

for this family of compounds. As well as having an electron-donating group instead of an electronwithdrawing group, k064-0027 is also less lipophilic, having a calculated LogP of 5.5 compared to 6.14 for k064-0035. For comparison, m-AMSA has a calculated LogP of 4.00.

2.4

Discussion

Bioactive chemical probes are useful biological tools, and can be as effective as mutants or antibodies for studying the functions of genes and pathways relevant in human health. Here we describe an integrated, miniaturized platform that comprehensively interrogates the effect of increasing and decreasing individual gene dosage on drug resistance using a single microarray to measure cellular fitness. Our experiments with 8 reference compounds and 188 compounds of unknown activity validated the use of this platform for identifying small molecule-target interactions. From a practical perspective, extensibility is an essential feature of genome-wide assays and accordingly we designed this suite of assays to be compatible with existing protocols, reagents and devices. Additionally, miniaturization is important for high throughput screens where compounds, in particular natural products, are limited in supply. Much of the current cost of drug development can be attributed to a poor understanding of the effects of potential drugs in vivo. The practice of target-based drug discovery has indeed produced many promising drug candidates that ultimately fail clinical trials due to unanticipated side effects [47]. Systems-level approaches, like those described here, that measure small molecule effects in an in vivo context, are better suited to define the biological effects of potential drugs. For example, the protein phosphatase inhibitors cantharidin and calyculinA possess similar activities in vitro, however their deletion sensitivity and multicopy suppression profiles are surprisingly dissimilar (Fig. 2.5). An additional benefit of our approach is that all potential targets are interrogated simultaneously and without bias. This approach readily identified the therapeutic targets of several drugs (Fig. 2.3). Our efforts with 188 previously uncharacterized synthetic compounds, which we expect will be a valuable resource for future experiments (and analysis), identified several potential interactions with various cellular targets (Table 2.5). While further chemical modifications would be needed to confer the pharmacokinetic properties required for a therapeutic, compounds identified using this methodology could prove excellent starting points for developing analogs with improved properties. In principle these modifications could be directed in part by SAR studies similar to those illustrated in Fig. 2.11. Using a randomly generated genomic library for MSP has several key benefits. One is that it is readily adaptable to testing the genomes of other organisms, provided an appropriate expression array is available. We demonstrated this by identifying a putative functional homolog of an uncharacterized and poorly conserved S. cerevisiae gene (CRGl) in the fungal pathogen C. albicans (Fig. 2.7b). Given the conservation between yeast and human genes, expression of human cDNAs in yeast


could prove useful in identifying human drug targets. In summary, integration of multiple genomewide assays is an effective way to employ smallmolecules to dissect fundamental cellular processes. Development of scalable technologies in addition to those described here, will be important as will the development of analysis tools that can integrate these genome-wide datasets to piece together the complete spectrum of chemical-genetic interactions.

2.5 2.5.1

Materials and Methods Reagents

Methotrexate, rapamycin, MMS, nocodazole, cantharidin and fluconazole were purchased from Sigma-Aldrich (St. Louis, Missouri, United States). Calyculin A and latrunculin A were purchased from BioMol (Plymouth Meeting, Pennsylvania, United States). 1-NM-PP1 was chemically synthesized as previously described[5]. The chemical diversity library was obtained from ChemDiv (San Diego, USA). Each compound was dissolved in DMSO, with the exception of rapamycin which was dissolved in 90% ethanol, 10% Tween 20. All compounds were stored at -20° C until use. Strains and Media. Yeast were maintained in YPD media[44, 45] at 30° C unless stated otherwise. Strains and plasmids used for individual analysis in this study listed in Table 2.2 and Table 2.3. Deletion strains were obtained from the yeast deletion collection or constructed de novo using PCR-based gene replacement [11]. Multicopy suppressor ORFs are listed below and were cloned by gap-repair as described by Oldenburg K.R. et al. [38]. Primers used for cloning are listed Table 2.1.

19


20

Table 2.1: Primers for cloning ORFs by gap-repair

ORF

Left Primer

5'-TAGTGGATCCCC CGG GCT GCA GGA ATT CGA TAT CAA GCT TCA GTG GCT GTT TGC TGA CAT - 3' 5' - TAG TGG ATC CCC CRG1 CGG GCT GCA GGA ATT CGA TAT CAA GCT TAA TGA ATG CGG CAA GAT ACC - 3' 5'-TAG TGG ATC CCC HEM1 CGG GCT GCA GGA ATT CGA TAT CAA GCT TGT TGTTGCTGCTGCTTT TGA-3' orfl9.633 5'-TAG TGG ATC CCC CGG GCT GCA GGA ATT CGA TAT CAA GCT TCA AAA TGT CCA TGT GAT GCC-3'

GLC7

Right Primer 5'-GCGTAATACGAC TCA CTA TAG GGC GAA TTG GGT ACC GGG CCC AAA GGA AGA CGT GAC CAT - 3' 5' - GCG TAA TAC GAC TCA CTA TAG GGC GAA TTG GGT ACC GGG CCC GGA AAC AGC TTT CTG AAG - 3' 5'-GCGTAA TAC GAC TCA CTA TAG GGC GAA TTG GGT ACC GGG CAC TTC TAA GTT GGC CGC TGA-3' 5'-GCG TAA TAC GAC TCA CTA TAG GGC GAA TTG GGT ACC GGG CAT TCC AAT TTG CCA TAC CCA

Amplicon Size 2620

2083

2806

2197


Table 2.2: Individual strains used in this study

Strain BY4743

Source Giaever et al. 2002'

HHY100 HHY101 HHY102 KT1112 KT1638 KT1621

Genotype MATa/a his3Al/his3Al leu2AO/leu2AO lys2AO/LYS2 MET15/metl5AO uraiAOIuraSAO BY4743; cdc28A::Kan/cdc28-as;Nat BY4743; glc7A::Kao/GLC7 BY4743; crglA::Kw/CRGl MATa leu2 ura3-52 his3 MATa leu2 ura3-52 his3 glc7-109 MATa leu2 ura3-52 his3 glc7-129

KT1623

MATa leu2 ura3-52 his3 glc7-127s

Baker etal.1997 2

This study* Giaever et al. 20021 Giaever et al. 20021 Baker etal.1997 2 Baker etal. 19972 Baker etal. 19972

"Plasmid pJAUl containing the cdc28-as allele harboring a F88G as described by Bishop A.C. et al. 3 was used to integrate the cdc28-as allele into the wild-type CDC28 locus using a pop-in-pop-out strategy4. 1. Giaever, G. etal. Functional profiling of the Saccharomyces cerevisiae genome. Nature 418,387-391 (2002). 2. Baker, S. H., Frederick, D. L., Bloecher, A. & Tatchell, K. Alanine-scanning mutagenesis of protein phosphatase type 1 in the yeast Saccharomyces cerevisiae. Genetics 145, 615-626 (1997). 3. Bishop, A. C. etal. A chemical switch for inhibitor-sensitive alleles of any protein kinase. Nature 407, 395-401 (2000). 4. Sherman, F., Fink, G. R., Hicks, J. B. & Cold Spring Harbor Laboratory, in Laboratory course manualfor methods in yeast genetics 186 (Cold Spring Harbor Laboratory, New York, N.Y., 1986).

21


Table 2.3: Individual plasmids used in this study

Plasmids pRS426 YEplacl95 pRS426-CRG/

Charateristics Source SikorskiR.S.etal.' High copy number, URA3 marker Gietz R.D. et al.2 High copy number, URA3 marker High copy number, URA3 marker, CRG1-0RF This study5 plus regulatory regions pRS426-#EA// High copy number, URA3 marker, HEMI-OKF This study" plus regulatory regions High copy number, URA3 marker, GIC7-ORF This study" pRS426-GZC7 plus regulatory regions pRS426-w/79./-rO#2-trunc 1. Sikorski, R. S. & Hieter, P. A system of shuttle vectors and yeast host strains designed for efficient manipulation of DNA in Saccharomyces cerevisiae. Genetics 122, 19-27 (1989). 2. Gietz, R. D. & Sugino, A. New yeast-Escherichia coli shuttle vectors constructed with in vitro mutagenized yeast genes lacking six-base pair restriction sites. Gene 74,527-534 (1988). 3. Oldenburg, K. R., Vo, K. T., Michaelis, S. & Paddon, C. Recombination-mediated PCR-directed plasmid construction in vivo in yeast. Nucleic Acids Res. 25, 451-452 (1997).

22


2.5.2

23

Individual strain growth analysis

For individual strain growth, yeast strains were grown to saturation (~20 h). Cells were then diluted to an ODeoo of 0.02 in a final volume of lOOul. Normalized cultures were grown in the presence of drug or diluent control in 96-well plates (Nunc, Rochester, New York, United States) using Tecan GENios microplate readers (Tecan, Austria) for up to 30 h. The growth rate of each culture was monitored by measuring the ODgoo every 15 minutes and the average doubling time (AvgG) was calculated as previously described[29]. Growth experiments for deletion strains were performed in YPD while those for multicopy suppressors were performed in synthetic complete media lacking uracil (SCM URA-).

2.5.3

Deletion-pool construction and screening conditions

Deletion pool construction was carried out as previously described[29, 42] with the following modification. Only strains identified as essential for growth in rich media, sterile or deficient in mating were used to create the heterozygous deletion pool. Pooled growth of the homozygous deletion pool and the heterozygous deletion pool were carried out for 5 and 20 generations, respectively. Both pools were inoculated at an ODeoo of 0.02, grown in 48-well microtiter plates (Nunc, Rochester, New York, United States) in a volume of 700/d and in compound concentrations that inhibited pool growth by ~10%. Cells were harvested automatically by a Packard Multiprobe II four-probe liquid-handling system (PerkinElmer, Wellesley, California, United States). For 20-generation experiments, cells were maintained in logarithmic phase by robotically diluting cultures every five doublings.

2.5.4

Multicopy Pool Construction and screening conditions

A S. cerevisiae random genomic library (gift from Martha Cyert) constructed in a high-copy 2fiM expression vector (YEplacl95) with an average insert size of approximately 5kb was transformed into yeast (cdc28-as or BY4743) by a standard lithium acetate method [15] and selected on media lacking uracil (URA-). After three days of growth ~10 6 transformants were pooled into media containing 7% DMSO, aliquoted, and stored at -80°C until use. For the C. albicans pool experiments, a random genomic library constructed in a high-copy 2/JM expression vector (pRS426) [23] was purchased from Open Biosystems (Hunstville, Alabama). For both pools, frozen aliquots were thawed and inoculated directly into URA- media to an OD600 of 0.02 and a volume of 700/ul. Compound was added and the pool was grown for 5 generations in 48-well microtiter plates (Nunc, Rochester, New York, United States) at inhibitory concentrations of at least 50% (IC50) and harvested as in the same way as described above.


2.5.5

24

Genomic D N A preparation, TAG PCR, and microarray hybridization

Genomic DNA preparation, PCR amplification of molecular tags, and microarray hybridization were performed as previously described [42] with the following modification. Cells collected from the essential heterozygous deletion pool and the homozygous deletion pool were combined in 1:2 ratio before genomic DNA preparation, TAG PCR and microarray hybridization.

2.5.6

Plasmid isolation and insert P C R amplification and microarray hybridization

For MSP screens, plasmids were isolated using the Zymoprep II plasmid isolation kit (Zymoresearch; Catalog number D2004). The inserts were amplified by PCR with the FailSafe™ PCR System (EPICENTRE Biotechnologies) using common M13 primers (M13 forward primer: 5' - GTT GTA AAA CGA CGG CCA GT - 3'; M13 reverse primer: 5' - CAG GAA ACA GCT ATG ACC - 3'). PCR cycling conditions were: 95°C 2min; 95°C 0.5min; 58°C 0.5min; 68°C lOmin; 30 cycles; 68°C 15 min. The PCR products were purified using QIAquick PCR purification kit (Qiagen; Catalog number 28104) and labeled with biotin using the BioPrime (Invitrogen; Catalog number 18094-011) labeling kit. Labeled products were hybridized to Affymetrix TAG4 arrays using the same protocols as described for TAG hybridizations [42].

2.5.7

Microarray Analysis

Both DSP and MSP were analyzed using a high-density oligonucleotide tag array manufactured by Affymetrix.

MSP for C. albicans was analyzed using a custom high-density oligonucleotide

Genechip(c) also manufactured by Affymetrix[28] (PN=510556). For Deletion Sensitivity Profiling (DSP), barcode probe intensities were extracted and processed as previously described[42]. Each array was mean normalized and fold change (log2 control/treatment) was calculated by comparing to a set of control arrays. Tags from the homozygous pool were normalized separately from tags from the heterozygous essential pool, as were the upstream tags (uptag) and downstream tags (downtag). At least two biological replicates were carried out for each treatment condition. The log2 ratios of both tags were averaged to generate a single score for each gene. For multicopy suppression profiling (MSP), ORF probe intensities were extracted and processed in the same way as the barcode probes. Each ORF is represented by at least 2 probes and the log2 ratios of each probe were averaged to generate a single score for each gene. To identify each suppressor locus, the log2 ratio of intensities were ordered by each ORF's genomic location and analyzed using a sliding window to identify locus that have at least two adjacent ORFs with log2 ratios >— 1.6. The same analysis was applied to the C. albicans expression array. Vector analysis was used to identify strains that were significant in both assays [7]. Z-scores of log2 ratios from both assays are represented jointly as a vector in


25

a Cartesian plane and various sectors of the plane correspond to different response patterns. An average vector VREP was calculated and significance was determined as previously described[7]. Array labels were shuffled and vector analysis was carried out. This was repeated 1000 times to generate a null distribution of the average vectors VREP. We then counted the number of simulated VREP vectors, x(VREP), that have lengths greater than or equal to a given experimental VREP length and calculated the average expected value E(VREP) = x(VREP)/1000. Subsequently for each gene, we calculated an estimate of the percentage of false-positives if this gene was considered as significantly represented in both assays qg = E(VREP)/rank(g) where rank(g) represents the position of ORF g in a list of all genes sorted by decreasing \VREP\

values. Activity scores for

DSP and MSP in Table 2.5 were calculated independently using log ratios of array intensities. For each gene i with compound j , two activity scores were calculated: Aitjt£>sp = Ditj • Ditj/ 5 for both DSP and MSP were reported. All analyses were performed in MATLAB (MathWorks). GO-slfm analysis of deletion sensitivity profiles in Fig. 2.6 was performed using the GO-slim tool at the Saccharomyces Genome Database[17] website. Enrichment difference between both compounds was calculated by taking the difference between the fraction of genes mapped to each GO-slim annotation.

2.5.8

Chemical descriptor a n n o t a t i o n

ECFP_4 fingerprints were calculated using Pipeline Pilot V6.1.1 (Scitegic) and LogP values were calculated using (Molecular Operating Environment version 2007.09, Chemical Computing Group Inc)

2.5.9

Acknowledgements

We thank K. Tatchell for sharing the glc-7 alleles and M. Cyert for providing the S. cerevisiae genomic library;


2.6

Tables Table 2.4: Reference compounds used in this study Compound

DSP Concentration

MSP Concentration

1

1-NM-PP1

500nM

500nM

2

Methotrexate

200pM

810uM

3

Fluconazole

33.516 uM

37.5 uM

4

Rapamycin

0.35nM

lOnM

5

Cantharidin

lOOpM

100pM

6

Calyculin A

2.5uM

10(iM

7

Methyl methanesulfonate (MMS)

0.002%

0.015%

8

Latrunculin A

2.5uM

5uM

9

Nocodazole

15pM

33.2uM

Known mechanism of action Inhibits analog sensitive kinase cdc28-AS' Inhibitor of folic acid biosynthesis2 Inhibitior of Lanosterol 14alphademethylase M Inhibitor of TOR signaling5 Protein phosphatase inhibitor 6 ' 7 Protein phosphatase inhibitor 8 DNA-alkylating agent9 Actindepolymerizing agent' 0 Microtubuledepolymerizing agent' 1

Protein Target cdc28-AS Dihydrofolate Reductase

Erg 11

Torl,Tor2 PP1, PP2a PP1, PP2a Unknown Actin monomer Alpha-beta tubulin dimer

1. Bishop, A. C. el al. A chemical switch for inhibitor-sensitive alleles of any protein kinase. Nature 407, 395-401 (2000). 2. Myers, C. E., Lippman, M. E., Elliot, H. M. & Chabner, B. A. Competitive protein binding assay for methotrexate. Proc. Nail. Acad. Sci. U. S. A. 72,3683-3686(1975). 3. White, T. C, Marr, K. A. & Bowden, R. A. Clinical, cellular, and molecular factors that contribute to antifungal drugresistance.Clin. Microbiol. Rev. 11,382-402 (1998). 4. Kontoyiannis, D. P., Sagar, N. & Hirschi, K. D. Overexpression of Ergl lp by the regulatable GAL1 promoter confersfluconazoleresistancein Saccharomyces cerevisiae. Antimtcrob. Agents Chemother. 43, 2798-2800 (1999). 5. Crespo, J. L. & Hall, M. N. Elucidating TOR signaling andrapamycinaction: lessons from Saccharomyces cerevisiae. Microbiol. Mol. Biol. Rev. 66,579-91, table of contents (2002). 6. Honkanen, R. E. Cantharidin, another natural toxin that inhibits the activity of serine/threonine protein phosphatases types 1 and 2A. FEBSLeu. 330,283-286 (1993). 7. Li. Y. M. & Casida, J. E. Cantharidin-binding protein: identification as protein phosphatase 2A. Proc. Natl. Acad. Set. U. S. A. 89,11867-11870 (1992). 8. Ishihara, H. el al. Calyculin A and okadaic acid: inhibitors of protein phosphatase activity. Btochem. Btophy.1. Res. Cammtm. 159,871-877 (1989). 9. Beranek, D. T. Distribution of methyl and ethyl adducts following alkylation with monofunctional attylating agents. Mutat. Res. 231,11-30(1990). 10. Ayscough, K.R.elnl. High rates of actin filament turnover in budding yeast and roles for actin in establishment and maintenance of cell polarity revealed using die actin inhibitor latrunoulin-A. / Cell Biol. 137,399-416(1997). 11. Kunkel, W. Effects of the antimicrotubular cancerostatic drug nocodazole on the yeast Saccharomyces cerevisiae. Z. Allg. Mikrobiol. 20, 315-324 (1980).

26


Table 2.5: List of top 20 gene-compound pairs with the highest combined DSP-MSP activity

Strain

Compound

1 2 3 4 5 6 7

CRG1 TOR2 ERG11 SEC14 ERG11 GLC7 ERG24

8

Cantharidin (8) Rapamycin (4) Fluconazole (3) 4130-1278(11) 4513-0042(13) Calyculin A (9) 4092-0821 (18)

Activity Score 2082.2 1106.9 1089.1 1011.7 702.6 527.9 463.4

horn het het het het het horn

PDC1

0086-0128 (19)

354.4

horn

9

SOD2

3013-0144 (20)

324.9

horn

10

HEM1

0986-0246(21)

316.5

het

11

PDR1

3937-0236 (22)

314.4

horn

12

TRP2

1486-1328 (23)

306.5

horn

13

DFR1

Methotrexate (2)

280.3

het

14

ERG3

0958-0271 (24)

212.1

horn

15

HEM15

1120-0019(25)

201.5

het

16

YAP1

0987-0079 (26)

179.7

horn

17

RPN4

0988-0037 (27)

173.5

horn

18

HYP2

4466-0038 (28)

168.4

het

19

STP1

0109-0045 (29)

161.6

horn

20

PAB1

1326-1318(30)

152.5

het

Strain


Table 2.6: ECFP_4 Tanimoto Coefficient Similarity of 9-anilinoacridine derivatives to m-AMSA

Name 0986-0246 k064-0012 k064-0020 k064-0027 k064-0035 m-AMSA

0986-0246 k064-0012 k064-0020 k064-0027 k064-0035 /M-AMSA 0.62 1.00 0.43 0.57 0.67 0.38 0.43 1.00 0.38 0.43 0.42 1.00 0.47 0.54 0.36 1.00 0.69 0.53 1.00 0.41 1.00

28


2.7

29

Figures Heterozygous deletion strain (essential gene)

Compound X

f AX/AX f Homozygous deletion pool

_ Heterozygous deletion pool

Multi-copy pool o. o. 3

Deletion barcode

Genomic library insert

•Putative target

"(Homozygous deletion strain (non-essential gene)

CO

Sensitivity TAG4 Array

Deletion sensitivity profile

Multi-copy suppressor profile

Figure 2.1: An integrated chemogenomics screening platform, (a) Illustration of the chemogenomic platform that interrogates three different yeast pools with a single TAG4 array. A homozygous deletion pool (n=4990), a heterozygous deletion pool (n=1145), and a pool of genomic library transformants (n~4700), are each challenged with a compound of interest (X). The heterozygous deletion pool represents only those genes that are essential for viability or for which homozygous deletions could not be constructed systematically (see Methods). Barcode sequences are isolated and amplified from deletion strain pools and genomic DNA inserts are isolated and amplified from the library transformants. Labeled products are then hybridized sequentially to the same TAG4 array, (b) Genes directly related to a drug's mechanism of action are predicted to alter drug resistance when deleted or overexpressed. Homozygous deletion strains (blue) and heterozygous deletion strains (red) identified in both assays represent genes that are more likely to be directly related to the drug mechanism of action.

CHAPTER 2. AN INTEGRATED CHEM0GEN0M1CS PLATFORM

Genomicllbrarywi|dlyp60rmutant

-•fft) Transformation

Compoum

Stock Pool

Screen

o _>

o

Plasmid Isolation

SO

=

==

PCR amplification of insert and labeling

DNA hybridization and analysis

Multicopy suppressor stock pool generation (once)

b

c

Pool Coverage

d , CDC28-AS MSP Pool

i*-~

*CDC2S IFA38

-TOS1

!,„ s—-I

1-NM-PP1 (1)

Intensity cutoff {fold above background)

DMSO

^ \™

|y -i.. "t

~"-rte»*

'BSSS

ssos"

agar

lass*

tw>

GENE

Figure 2.2: Multicopy Suppression Profiling (a) Schematic of MSP. A multicopy pool is generated by transforming a cloned genomic library into a yeast strain of choice. Following the growth of pooled transformants, plasmids are isolated and the genomic inserts are amplified by PCR using common primers and hybridized to a TAG4 array, (b) Distribution showing the number of ORFs present in the multicopy suppressor pool as a function of chip intensity cutoff. At 2-fold above a conservative estimate of background intensity (500), at least 4307 ORFs are estimated to be in the pool, (c) A genetic complementation screen using the cdc28-AS analog sensitive kinase. A diploid strain harboring the cdc28-AS allele as the sole source of CDC28 was transformed with a genomic library as described in text. Transformants were pooled in 700/d of selective media and grown in the presence or absence of 500nM l-NM-PPl(l) which resulted in a ~ 5 0 % inhibition of growth rate with inhibitor. The chemical structure of 1-NM-PP1 is shown on the left, (d) The fold change (log2(l-NM-PPI-treated/untreated)) in microarray signal intensity is plotted on the y-axis for ~6000 genes (arranged alphabeti cally on the x-axis). CDC28 and three genes flanking the GDC28 locus (CSH1, IFAS8, TOS1) were overrepresented in the 1-NM-PP1 treated pool, (insert) Genomic location of the 4 genes highlighted. Specificity of DSP. (left)The specificity of 1-NM-PPl was assessed by screening a pool containing ~6000 heterozygous deletion strains and a single cdc28-AS strain with 1-NM-PPl. (right)The fold change in tag hybridization intensity (log2(control/l-NM-PPl)) is plotted on the y-axis for each strain (arranged alphabetically on the x-axis). Only the cdc28-AS strain exhibited sensitivity to the inhibitor.


Rapamycin (4)

31

Sensitivity

Figure 2.3: Data integration improves small-molecule target prediction, (a-c) Chemical structures (left) and MSP versus DSP plots (right) for methotrexate (a), fluconazole (b) and rapamycin (c). For each ORF, the average z-score of fold change derived from MSP (representing suppression) and DSP (representing sensitivity) are plotted on the y and x-axis, respectively. Strains that meet a false discovery rate threshold of 0.05 are colored red for heterozygous deletion strains and blue for homozygous deletion strains.


32

Latrunculln A

MMS

^

10

10

/ c o

^

Q.

• ^

Q.

1

L 5

\

to

/

: 10

*EM1

" ^ j

to -10

"U*-"

c o in °

^L /jMxie -^

1

-10 0

5

10

Sensitivity'

Sensitivity

-10

-5

0

5

10

Sensitivity

PI3PI4 Kinase FRB - C-term. TOR! C-term.TOR2AfRBAPBTOKinase

C-term. 70R2

Time (h)

Time (h)

C-lerm. 70R2 AWB AP/3 PM Kinase • Diluent • 10nM Rapamycin

Time (h)

Figure 2.4: (a-c) Vector analysis results for (a) latrunculin A, (b) nocodazole and (c) MMS are presented. For latrunculin A, the only gene identified in both assays was MAK10, a N-terminal acetyltransferase. For nocodazole, only TUB3, one of 2 genes encoding alpha-tubulin was identified from both assays. TUB1, the other alph-tubulin, was filtered from the analysis because of poor tag intensity. TUB2, the beta-tublin was neither identified in DSP nor MSP. For MMS, HEM1, an essential gene encoding 5-aminolevulinate synthase was identified in both assays, (d) Genomic fragment (black line) that suppresses rapamycin sensitivity contains both the PI3/PI4 kinase domain and the binding domain (FRB) of TOR2 and is sufficient to suppress rapamycin sensitivity in an isogenic culture (bottom middle), relative to the vector control (bottom left). A truncated fragment (red line) lacking the FRB domain fails to suppress rapamycin sensitivity (bottom right).


3

U

liSSVi

"

™5>

' w5>

"

QH O

™?(J~T

Figure 2.5: Cantharidin and Calyculin A have distinct effects in vivo, (a-b) Chemical structure (left) and MSP versus DSP plots (right) for cantharidin and calyculin A respectively, (c) (top) BY4743 (wild-type) and a GLC7 heterozygous deletion were grown in the presence of DMSO, cantharidin and calyculin A in YPD. In the presence of either compound, the deletion strain grew slower than wild-type, (bottom) BY4743 (wild-type) containing either vector (pRS426) or a plasmid containing GLC7 (pRS426-G£C7) was grown in the presence of DMSO, cantharidin and calyculin A in URA- media. The strain harboring multiple copies of GLC7 conferred resistance to calyculin A but not cantharidin. (d) The relative affinity of calyculin A and cantharidin for Glc7 was measured using a microcystin competition assay. Cell lysates from yeast expressing Glc7 m j, c were pre-incubated with buffer, calyculin A, microcystin, cantharidin or okadaic acid before adding microcystin-conjugated agarose beads. Proteins bound to the beads were isolated, separated by SDS-PAGE and immunoblotted with an anti-myc mouse monoclonal antibody and HRP-labelled goat anti-mouse secondary antibody.

33


16 Methotrexate Methotrexate Methotrexate MMS MMS Nocodazole Nocodazole Latrunculin A Latrunculin A Rapamycln Rapamycin Calyculin A Calyculin A Cantharidin Cantharidin

to 14 S 12

34

IWT |glc7-109 aglc7-129 CDglc7-127

o> | 8

*" 6

?

5 4 8 2 0

DMSO Cantharidin Calyculin A 25(iM 2.5jiM

• RNA metabolic process • ONA metabolic process • response to stress • anatomical structure morphogenesis • organdie organization and biogenesis • biological process unknown IS ribosome biogenesis and assembly • protein modification process • cytoskeleton organization and biogenesis ffi transcription ^ pseudohyphal growth M amino acid and derivative metabolic process W melosls m ceH cycle

£ nuclear organization end biogenesis i cytokinesis conjugation ' cell budding carbohydrate metabolic process % cellular homeostasis I sporulatJon I 8 I 8

vitamin metabolic process generation of precursor metabolites and energy lipid metabolic process protein cataboltc process

membrane organization and biogenesis I transport % vesicle-mediated transport

,:. cell wall organization and biogenesis SI signal transduction

-04

0

0.1

0.2

GO-slim Enrichment Difference

Figure 2.6: (a) Clustergram of deletion sensitivity profiles of reference compounds used in this study. Log2 ratios from deletion sensitivity experiments were hierarchically clustered using Pearson's correlation with aver-age linkage, (b) Functional enrichment, determined by GO-slim analysis, of the top 50 strains sensitive to cantharidin and calyculin A were compared. Enrichment difference between both compounds was calculated by taking the difference between the fraction of genes mapped to each GO-slim annotation, (c) Three different glc7 alleles, defective in nuclear (glc7-127, glc7-129) and cytoplasmic (glc7-109) pathways, were grown in the presence of cantharidin and calyculin A and their doubling times are presented ±s.d. (n=3).


1 DMSO

2.5UM »pRS426 mpRS426-CR0.8 $0.7

35

PRS426 pRS426-CRG1 pRS426-orf19.633[

jfo.6 Time (Hours) DMSO

25UM

J\ \M \JA \J-. Calyculin A

a>0.5 •.§0.4 ^0.3 ^0.2 0.1 0

Cantharidin 100fiM

Cantharidin c o 'Jo,

prfl 9.633 0rf19 1808/af rf19.634 orf19.180" • - "

It)

2

a. o-, 3

Location

CO-

Location

Figure 2.7: (a) CRG1 is a major suppressor of cantharidin sensitivity. Expression of CRG1 from a high-copy vector under its own promoter confers resistance to cantharidin but not calyculin A sensitivity, (b) A C. albicans genomic library was transformed into S. cerevisiae (BY4743). Transformants were pooled and grown in the presence of calyculin A and cantharidin. Recovered inserts were hybridized to a C. albicans expression array. The fold ratio (log2 (treatment/control)) of each ORF is plotted on the y-axis and arranged by its annotated genomic location on the x-axis. The red, dashed-line is the fold-change cutoff (log2 1.6) used to identify each significant suppressor locus. Area highlighted in green corresponds to the identified suppressor locus, (c) or}19.633 and CRG1 were cloned into the high-copy plasmid pRS426, transformed into BY4743 and grown in the presence and absence of cantharidin (lOOpM). Doubling times were normalized to growth of each strain in DMSO control ( s.d. (n=3).


a

F 0=U*

DMSO

36

4130-1278 (125|*M)

CI

o >

N N

\

• PRS426 • pRS«6-SEC14

O

H

-5

4130-1278(11)

0

5

Sensitivity

Time(Hours)

Time(Hours)

.4130-1278

30 25

.4534-1311 8 15 CO .•5*10

Is

§5° -10 -5

jr:' * 5

10

15

20

25

30

35

DSP activity score Figure 2.8: Characterization of novel bioactive compounds, (a) Chemical structure of 4130-1278 (middle) MSP versus DSP plot for 4130-1278 (b) Confirmation growth curves for 4130-1278. (Top row) Growth of HO strain compared to secl4+/- in the presence of DMSO or 4130-1278. (Bottom row) Growth of wild type strain with plasmids containing either empty vector or SEC14 under control of endogenous promoter, (c) MSP activity score versus DSP activity score for SEC14 for all 196 compounds. Datum for 4130-1278 is highlighted in red.


J3

30

£25

O

8 W20 2
>

as £|0.5

•pllfrv.-

55|

$*•>•

jgf

IS 'g>8 o j^B'vv; o CO

0 -0.5,

39

Concordance

-... 80.9%

i

• ,

*.* " k084-0035(15)

9

k084-0027(14)

Or

|

k060-OOB7(31)

k072-0108(32)

18.1% Discordance

0.5

1

Structure Similarity (ECFP_4fingerprintTanimoto coerWent) k015-0008(32)

kO48-O0S2(34)

k01&-0014

k048-0028(35)

• •

2

. RA B 55 .«AD 5 4 RAD5* (RAD51

fil1 * •2* Sa.0 a

~\' -

.».!&£t • • m (P*.*

.« |

WQ* •V

m 2

-

1

0

1

2

k064-0G12

;xr DSP log^ratio

NÂ-^k^

^

3

- 1 0

1 2 k064-0020 DSP log,ratio

LA>-V

3

(37)

Figure 2.11: Chemogenomics profiles reveal structure activity relationships. (a)Structure activity concordance was assessed by comparing Tanimoto similarities between each pair of compound to their DSP profile similarity. Compound pairs corresponding to the green and red boxes in Fig. 2.10a are plotted in green and red respectively. Compound pairs with high structural similarity (Tanimoto score > 0.3) and high profile correlation (correlation > 0.2) are considered concordant while compound pairs with high structural similarity and low profile correlation (correlation > 0.2) are considered discordant. Representative discordant compound pairs (circled red, cyan, green and black) and their corresponding compound structures are shown (matching colored boxes), (b) Pairwise comparison of DSP profiles (log2 ratios) of k064-0027 with 4 other structurally similar compounds. Genes involved in homologous recombination-mediated DNA repair (RAD51, RAD54, RAD55, RAD57, RAD59) are highlighted in red.


2.8

CHEMOGENOMICS

PLATFORM

40

Bibliography

[1] P. D. Andrews and M. J. Stark. Type 1 protein phosphatase is required for maintenance of cell wall integrity, morphogenesis and cell cycle progression in saccharomyces cerevisiae. Journal of cell science, 113 ( Pt 3)(Pt 3):507-520, Feb 2000. [2] M. B. Arnaud, M. C. Costanzo, M. S. Skrzypek, P. Shah, G. Binkley, C. Lane, S. R. Miyasato, and G. Sherlock. Sequence resources at the Candida genome database. Nucleic acids research, 35(Database issue):D452-6, Jan 2007. [3] B. C. Baguley and R. Nash. Antitumour activity of substituted 9-anilinoacridines-comparison of in vivo and in vitro testing systems. European journal of cancer, 17(6):671-679, Jun 1981. [4] S. H. Baker, D. L. Frederick, A. Bloecher, and K. Tatchell. Alanine-scanning mutagenesis of protein phosphatase type 1 in the yeast saccharomyces cerevisiae. Genetics, 145(3):615-626, Mar 1997. [5] A. C. Bishop, J. A. Ubersax, D. T. Petsch, D. P. Matheos, N. S. Gray, J. Blethrow, E. Shimizu, J. Z. Tsien, P. G. Schultz, M. D. Rose, J. L. Wood, D. O. Morgan, and K. M. Shokat. A chemical switch for inhibitor-sensitive alleles of any protein kinase. Nature, 407(6802):395-401, Sep 21 2000. [6] A. Bloecher and K. Tatchell. Defects in saccharomyces cerevisiae protein phosphatase type i activate the spindle:kinetochore checkpoint. Genes & development, 13(5):517-522, Mar 1 1999. [7] R. Breitling, P. Armengaud, and A. Amtmann. Vector analysis as a fast and easy method to compare gene expression responses between different experimental backgrounds. BMC bioinformatics, 6:181, Jul 19 2005. [8] C. Brenner. Chemical genomics in yeast. Genome biology, 5(9):240, 2004. [9] R. A. Butcher, B. S. Bhullar, E. O. Perlstein, G. Marsischky, J. LaBaer, and S. L. Schreiber. Microarray-based method for monitoring yeast overexpression strains reveals small-molecule targets in tor pathway. Nat.Chem.Biol,

2(2):103-109, Feb 2006.

[10] W. A. Denny, B. F. Cain, G. J. Atwell, C. Hansch, A. Panthananickal, and A. Leo. Potential antitumor agents. 36. quantitative relationships between experimental antitumor activity, toxicity, and structure for the general class of 9-anilinoacridine antitumor agents. Journal of medicinal chemistry, 25(3):276-315, Mar 1982. [11] N. Erdeniz, U. H. Mortensen, and R. Rothstein. Cloning-free pcr-based allele replacement methods. Genome research, 7(12):1174-1183, Dec 1997.


CHEMOGENOMICS

PLATFORM

41

[12] H. Fujiki and M. Suganuma. Tumor promotion by inhibitors of protein phosphatases 1 and 2a: the okadaic acid class of compounds. Advances in Cancer Research, 61:143-194, 1993. [13] G. Giaever, P. Flaherty, J. Kumm, M. Proctor, C. Nislow, D. F. Jaramillo, A. M. Chu, M. I. Jordan, A. P. Arkin, and R. W. Davis. Chemogenomic profiling: identifying the functional interactions of small molecules in yeast. Proceedings of the National Academy of Sciences of the United States of America, 101(3):793-798, Jan 20 2004. [14] G. Giaever, D. D. Shoemaker, T. W. Jones, H. Liang, E. A. Winzeler, A. Astromoff, and R. W. Davis. Genomic profiling of drug sensitivities via induced haploinsufficiency. Nature genetics; Nature genetics, 21(3):278-283, Mar 1999. [15] R. D. Gietz, R. H. Schiestl, A. R. Willems, and R. A. Woods. Studies on the transformation of intact yeast cells by the liac:ss-dna:peg procedure. Yeast (Chichester, England), ll(4):355-360, Apr 15 1995. [16] M. E. Hillenmeyer, E. Fung, J. Wildenhain, S. E. Pierce, S. Hoon, W. Lee, M. Proctor, R. P. St Onge, M. Tyers, D. Koller, R. B. Altman, R. W. Davis, C. Nislow, and G. Giaever.

The

chemical genomic portrait of yeast: uncovering a phenotype for all genes. Science (New York, N.Y.), 320(5874) :362-365, Apr 18 2008. [17] J. E. Hirschman, R. Balakrishnan, K. R. Christie, M. C. Costanzo, S. S. Dwight, S. R. Engel, D. G. Fisk, E. L. Hong, M. S. Livstone, R. Nash, J. Park, R. Oughtred, M. Skrzypek, B. Starr, C. L. Theesfeld, J. Williams, R. Andrada, G. Binkley, Q. Dong, C. Lane, S. Miyasato, A. Sethuraman, M. Schroeder, M. K. Thanawala, S. Weng, K. Dolinski, D. Botstein, and J. M. Cherry. Genome snapshot: a new resource at the saccharomyces genome database (sgd) presenting an overview of the saccharomyces cerevisiae genome. Nucleic acids research, 34(Database issue):D442-5, Jan 1 2006. [18] R. E. Honkanen. Cantharidin, another natural toxin that inhibits the activity of serine:threonine protein phosphatases types 1 and 2a. FEBS letters; FEBS letters, 330(3) :283-286, Sep 20 1993. [19] S. Hoon, A. M. Smith, I. M. Wallace, S. Suresh, M. Miranda, E. Fung, M. Proctor, K. M. Shokat, C. Zhang, R. W. Davis, G. Giaever, R. P. StOnge, and C. Nislow. An integrated platform of genomic assays reveals small-molecule bioactivities. Nature chemical biology, 4(8):498-506, Aug 2008. [20] J. Y. Hsu, Z. W. Sun, X. Li, M. Reuben, K. Tatchell, D. K. Bishop, J. M. Grushcow, C. J. Brame, J. A. Caldwell, D. F. Hunt, R. Lin, M. M. Smith, and C. D. Allis. Mitotic phosphorylation of histone h3 is governed by ipll:aurora kinase and glc7:ppl phosphatase in budding yeast and nematodes. Cell, 102(3):279-291, Aug 4 2000.


CHEMOGENOMICS

PLATFORM

42

[21] T. R. Hughes, M. J. Marton, A. R. Jones, C. J. Roberts, R. Stoughton, C. D. Armour, H. A. Bennett, E. Coffey, H. Dai, Y. D. He, M. J. Kidd, A. M. King, M. R. Meyer, D. Slade, P. Y. Lum, S. B. Stepaniants, D. D. Shoemaker, D. Gachotte, K. Chakraburtty, J. Simon, M. Bard, and S. H. Friend. Functional discovery via a compendium of expression profiles. Cell, 102(1):109126, Jul 7 2000. [22] H. Ishihara, B. L. Martin, D. L. Brautigan, H. Karaki, H. Ozaki, Y. Kato, N. Fusetani, S. Watabe, K. Hashimoto, and D. Uemura. Calyculin a and okadaic acid: inhibitors of protein phosphatase activity. Biochemical and biophysical research communications; Biochemical and biophysical research communications, 159(3) :871-877, Mar 31 1989. [23] D. Kadosh and A. D. Johnson. Rfgl, a protein related to the saccharomyces cerevisiae hypoxic regulator roxl, controls filamentous growth and virulence in Candida albicans. Molecular and cellular biology, 21(7):2496-2505, Apr 2001. [24] D. P. Kontoyiannis, N. Sagar, and K. D. Hirschi. Overexpression of ergllp by the regulatable gall promoter confers fluconazole resistance in saccharomyces cerevisiae. Antimicrobial

Agents

and Chemotherapy, 43(ll):2798-2800, Nov 1999. [25] N. J. Krogan, W. T. Peng, G. Cagney, M. D. Robinson, R. Haw, G. Zhong, X. Guo, X. Zhang, V. Canadien, D. P. Richards, B. K. Beattie, A. Lalev, W. Zhang, A. P. Davierwala, S. Mnaimneh, A. Starostine, A. P. Tikuisis, J. Grigull, N. Datta, J. E. Bray, T. R. Hughes, A. Emili, and J. F. Greenblatt. High-definition macromolecular composition of yeast rna-processing complexes. Molecular cell, 13(2):225-239, Jan 30 2004. [26] C. Kung, D. M. Kenski, S. H. Dickerson, R. W. Howson, L. F. Kuyper, H. D. Madhani, and K. M. Shokat. Chemical genomic profiling to identify intracellular targets of a multiplex kinase inhibitor. Proceedings of the National Academy of Sciences of the United States of America; Proceedings of the National Academy of Sciences of the United States of America, 102(10):35873592, Mar 8 2005. [27] J. Lamb, E. D. Crawford, D. Peck, J. W. Modell, I. C. Blat, M. J. Wrobel, J. Lerner, J. P. Brunet, A. Subramanian, K. N. Ross, M. Reich, H. Hieronymus, G. Wei, S. A. Armstrong, S. J. Haggarty, P. A. demons, R. Wei, S. A. Carr, E. S. Lander, and T. R. Golub. The connectivity map: using gene-expression signatures to connect small molecules, genes, and disease. Science, 313(5795):1929-1935, Sep 29 2006. [28] C. Y. Lan, G. Newport, L. A. Murillo, T. Jones, S. Scherer, R. W. Davis, and N. Agabian. Metabolic specialization associated with phenotypic switching in candidaalbicans. Proceedings of the National Academy of Sciences of the United States of America, 99(23):14907-14912, Nov 12 2002.

CHAPTER

2. AN INTEGRATED

CHEMOGENOMICS

PLATFORM

43

[29] William Lee, Robert P. St Onge, Michael Proctor, Patrick Flaherty, Michael I. Jordan, Adam P. Arkin, Ronald W. Davis, Corey Nislow, and Guri Giaever. Genome-wide requirements for resistance to functionally distinct dna-damaging agents. PLoS Genet, 1(2), 08:19: 2005. [30] X. Li, M. Zolli-Juran, J. D. Cechetto, D. M. Daigle, G. D. Wright, and E. D. Brown. Multicopy suppressors for novel antibacterial compounds reveal targets and drug efflux susceptibility. Chemistry & biology, 11(10):1423-1430, Oct 2004. [31] Y. M. Li and J. E. Casida.

Cantharidin-binding protein: identification as protein phos-

phatase 2a. Proceedings of the National Academy of Sciences of the United States of America, 89(24): 11867-11870, Dec 15 1992. [32] H. Luesch, T. Y. Wu, P. Ren, N. S. Gray, P. G. Schultz, and F. Supek. A genome-wide overexpression screen in yeast for small-molecule target identification.

Chemistry & biology,

12(l):55-63, Jan 2005. [33] P. Y. Lum, C. D. Armour, S. B. Stepaniants, G. Cavet, M. K. Wolf, J. S. Butler, J. C. Hinshaw, P. Gamier, G. D. Prestwich, A. Leonardson, P. Garrett-Engele, C. M. Rush, M. Bard, G. Schimmack, J. W. Phillips, C. J. Roberts, and D. D. Shoemaker. Discovering modes of action for therapeutic compounds using a genome-wide screen of yeast heterozygotes.

Cell,

116(1):121-137, Jan 9 2004. [34] C. MacKintosh, K. A. Beattie, S. Klumpp, P. Cohen, and G. A. Codd.

Cyanobacterial

microcystin-lr is a potent and specific inhibitor of protein phosphatases 1 and 2a from both mammals and higher plants. FEBS letters, 264(2): 187-192, May 21 1990. [35] J. L. Martin and F. M. McMillan. Sam (dependent) i am: the s-adenosylmethionine-dependent methyltransferase fold. Current opinion in structural biology, 12(6):783-793, Dec 2002. [36] C. E. Myers, M. E. Lippman, H. M. Elliot, and B. A. Chabner. Competitive protein binding assay for methotrexate. Proceedings of the National Academy of Sciences of the United States of America, 72(9):3683-3686, Sep 1975. [37] E. M. Nelson, K. M. Tewey, and L. F. Liu. Mechanism of antitumor drug action: poisoning of mammalian dna topoisomerase ii on dna by 4'-(9-acridinylamino)-methanesulfon-m-anisidide. Proceedings of the National Academy of Sciences of the United States of America, 81(5):13611365, Mar 1984. [38] K. R. Oldenburg, K. T. Vo, S. Michaelis, and C. Paddon. Recombination-mediated pcr-directed plasmid construction in vivo in yeast. Nucleic acids research, 25(2):451-452, Jan 15 1997. [39] A. B. Parsons, R. L. Brost, H. Ding, Z. Li, C. Zhang, B. Sheikh, G. W. Brown, P. M. Kane, T. R. Hughes, and C. Boone. Integration of chemical-genetic and genetic interaction data links bioactive compounds to cellular target pathways. Nature biotechnology, 22(l):62-69, Jan 2004.

CHAPTER

2. AN INTEGRATED

CHEMOGENOMICS

PLATFORM

44

[40] A. B. Parsons, A. Lopez, I. E. Givoni, D. E. Williams, C. A. Gray, J. Porter, G. Chua, R. Sopko, R. L. Brost, C. H. Ho, J. Wang, T. Ketela, C. Brenner, J. A. Brill, G. E. Fernandez, T. C. Lorenz, G. S. Payne, S. Ishihara, Y. Ohya, B. Andrews, T. R. Hughes, B. J. Frey, T. R. Graham, R. J. Andersen, and C. Boone. Exploring the mode-of-action of bioactive compounds by chemical-genetic profiling in yeast. Cell, 126(3):611-625, Aug 11 2006. [41] S. E. Pierce, R. W. Davis, C. Nislow, and G. Giaever. Genome-wide analysis of barcoded saccharomyces cerevisiae gene-deletion mutants in pooled cultures. Nature protocols, 2(ll):2958-2974, 2007. [42] S. E. Pierce, E. L. Fung, D. F. Jaramillo, A. M. Chu, R. W. Davis, C. Nislow, and G. Giaever. A unique and universal molecular barcode array. Nat.Methods, 3(8):601-603, Aug 2006. [43] J. Rine, W. Hansen, E. Hardeman, and R. W. Davis. Targeted selection of recombinant clones through gene dosage effects. Proceedings of the National Academy of Sciences of the United States of America, 80(22) :6750-6754, Nov 1983. [44] Mark David Rose, Fred Marshall Winston, and Philip Hieter. Methods in yeast genetics : a laboratory course manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y., 1990. [45] Fred Sherman, Gerald R. Fink, James B. Hicks, and Cold Spring Harbor Laboratory. Laboratory course manual for methods in yeast genetics. Cold Spring Harbor Laboratory, New York, N.Y., 1986. [46] R. Sopko, D. Huang, N. Preston, G. Chua, B. Papp, K. Kafadar, M. Snyder, S. G. Oliver, M. Cyert, T. R. Hughes, C. Boone, and B. Andrews. Mapping pathways and phenotypes by systematic gene overexpression. Molecular cell; Molecular cell, 21(3):319-330, Feb 3 2006. [47] G. C. Terstappen, C. Schlupen, R. Raggiaschi, and G. Gaviraghi. Target deconvolution strategies in drug discovery. Nature reviews.Drug discovery, 6(ll):891-903, Nov 2007. [48] E. P. Walsh, D. J. Lamont, K. A. Beattie, and M. J. Stark. Novel interactions of saccharomyces cerevisiae type 1 protein phosphatase identified by single-step affinity purification and mass spectrometry. Biochemistry, 41(7):2409-2420, Feb 19 2002. [49] D. W. Young, A. Bender, J. Hoyt, E. McWhinnie, G. W. Chirn, C. Y. Tao, J. A. Tallarico, M. Labow, J. L. Jenkins, T. J. Mitchison, and Y. Feng. Integrating high-content screening and ligand-target prediction to identify mechanism of action. Nature chemical biology, 4(l):59-68, Jan 2008. [50] X. F. Zheng, D. Florentino, J. Chen, G. R. Crabtree, and S. L. Schreiber. Tor kinase domains are required for two distinct functions, only one of which is inhibited by rapamycin. Cell; Cell, 82(1):121-130, Jul 14 1995.

Chapter 3

A framework for analyzing drug-drug interactions 3.1

Abstract

A promising new application of chemogenomic profiles is to aid our understanding of multi-component therapies. Multi-component therapies are the standard of care in the treatment of various diseases such as cancer, diabetes and AIDs. One rationale for multi-component approaches is by combining agents that act independently, the propensity for adaptive resistance is lowered. Additionally, use of multiple drugs may boost therapeutic potency by combining agents with non-additive toxicities. Despite their promise, most multi-component drug regimens are developed ad hoc, by physicians on a patient-by-patient basis, because of the staggering costs of full-scale clinical trials. As a result, the number of drug combinations is quite limited. It is thus increasingly desirable to develop assays for the identification and characterization of multi-component drugs early in the drug discovery pipeline. This chapter describes methods I have developed in yeast to screen drug combinations systematically and identify genetic mutations that sensitize the cell to drug combinations thereby allowing the mechanistic basis for drug interactions to be inferred. The major finding in this study is that treatment with both cantharidin and calyculin A was shown to inhibit GLC7 related pathways more selectively.

3.2

Introduction

Since the completion of the Saccharomyces cerevisiae sequencing project in 1995, and the subsequent completion of the yeast knockout collection in 2002 [8], there have been a large number of studies using chemogenomic profiling for understanding drug mechanism of action and to identify 1) novel

45

CHAPTER

3. ANALYZING

DRUG-DRUG

INTERACTIONS

46

drug targets and 2) pathways that mediates the cellular response to drug[9, 10, 14, 12, 5, 16]. One of the goals of personalized medicine is to match the knowledge of one's genetics and disease to the appropriate therapeutic. Despite their promise, most multi-component drug regimens are developed ad hoc, by physicians on a patient-by-patient basis, because of the staggering costs of full-scale clinical trials of compound combinations[22]. As a result, the number of approved drug combinations is quite limited. It is thus increasingly desirable to develop assays for the identification and characterization of multi-component drugs early in the drug discovery pipeline. Large-scale efforts to map the genetic network of the cell provide global maps of functional relationships between genes and pathways [6] and the challenge is to see whether this information can lead to better drug design that target genes and pathways for different genotypes. Recent high throughput drug combination studies have produced interesting results. One recent study showed that in Escherichia coli, drugs with a common mechanism of action tend to interact similarly with other drugs[20] suggesting a way for drug classification. In addition, both modeling and experimental analysis have shown that the outcome of drug combinations is highly dependent on how the targeted signaling pathway is wired [7]. Two other studies [3, 13] have reported systematic screening of combinations of small molecules in drug-resistant Candida albicans, primary human blood cells, and human tumor cell lines. Saccharomyces cerevisiae is the most well-studied eukaryote in terms of genetic and chemicalgenetic relationships. The goal in this study was to demonstrate that we can use yeast as a model to test strategies for designing drug combinations. The methods developed in this study are distinctive for previous studies in two ways. First, growth measurements are measured over the entire course of the experiment allowing subtle drug interactions to be detected. Second, chemogenomic profiling was used to identify genetic modifiers of the observed drug interaction.

3.3

Results

3.3.1

Growth assay for generating dose response surfaces

We screened 10 different drugs pairwise using an automated liquid handling robot for assay preparation. Data was collected using a Tecan Genois reader in 96-well format, analyzed and stored in a mySQL database. For each pair of drug, 6 difference concentrations were screened in duplicate yielding a 6x6 dose response surface. Each surface was then modeled against multiple drug interaction models including: Highest Single Agent (HSA), Loewe Additivity, Bliss Boosting, and Potentiation (See Material and Methods). The best fit was determined following the method of Lehar et al. [13] and assigned to that drug pair. A visual summary of our results is shown in Fig. 3.1. As expected, most drug pairs do not interact as showed by their assignment to an additive or HSA model. One drug, miconazole, increases the potency of most other drugs. Miconazole targets lanosterol 14-alpha-demethylase (encoded by ERG11), an enzyme that catalyzes a key step in the

CHAPTER 3. ANALYZING DRUG-DRUG INTERACTIONS

47

production of ergosterol [15], an essential component of the plasma membrane. Perturbation of the plasma membrane may affect general drug uptake and/or efflux thereby increasing the propensity for drug interactions and this suggests a way to sensitize drug screens. Indeed, a drug-sensitized screen was successful in identifying drugs that interact synergistically with low-doses of ketoconazole, another E r g l l inhibitor[21].

3.3.2

Cantharidin and Calyculin A synergize to inhibit yeast growth

It was shown in chapter 2 that cantharidin and calyculin A disrupt Glc7 function in distinct ways. To examine the effect of cantharidin and calyculin A when applied in combination, I quantified the degree of interaction across a dose response matrix surface using a Bliss interaction model [2] and found that the combination of cantharidin and calyculin A was highly synergistic compared to the combination of cantharidin and seven other compounds (Fig. 3.2). Of four previously described shape models [13], potentiation, which describes a combination where one drug's curve is shifted with a power-law slope p above an enhancer concentration

YPOT[13],

best described the relationship

between cantharidin and calyculin A (Fig. 3.2a). In this case, cantharidin, strongly potentiated calyculin A with p = 0.84 and

3.3.3

YPOT

= 41/xM (See Materials and Methods).


To characterize the nature of this interaction, we first compared the chemogenomic profiles, for a given inhibitory concentration (ICio), of a cocktail of cantharidin and calyculin A to the profiles of both drugs when used singly. At the same significance cutoff (false discovery rate < 0.1), fewer strains were sensitive to the cocktail (n=114) compared to the combined number of strains sensitive to either compound used singly at the equivalent inhibitory concentration (n=149) (Fig. 3.3). Strikingly, the cocktail screen also identified additional strains that were not identified by either single-drug treatment (n=26). This observation suggests that treatment with lower doses of both compounds in the cocktail experiment may reduce the off-target effects of the treatment. The additional strains that were unique to the cocktail screen may reflect increased perturbation of certain pathways that are targeted by both compounds.

3.3.4


To explore this further, we compared the strain sensitivities of each deletion strain in the cocktail and single drug treatment (at the same concentration). We wanted to identify deletion strains that are not sensitive in either cantharidin or calyculin A single treatments but are highly sensitive in the cocktail treatment.

Deletion Sensitivity Profiling (DSP) was performed in: diluent only

(DMSO), cantharidin, calyculin A, and a cocktail of cantharidin and calyculin A (Fig. 3.4a). Next we calculated the deviation of the observed fitness from the expectation (of the Bliss model), e, for

CHAPTER 3. ANALYZING

DRUG-DRUG

INTERACTIONS

48

each deletion strain using the array results for single and double-drug treatments (see Methods). Consistent with the effect on wild-type at the selected concentrations, the distribution of e values is centered at zero, indicating that the effects of each drug combine predictably for most deletion strains (Fig. 3.4b). The twenty heterozygous and twenty homozygous deletion strains with the smallest (most negative) e scores are listed in Fig. 3.4c. These represent gene deletions that sensitize the cell to the cantharidin-calyculin A cocktail. We found that the top 1% of sensitized strains (n=56) were significantly enriched for strains that are sensitive to calyculin A alone, compared to cantharidin alone (chi-square p < 6.3e-10, Fig. 3.4d). In agreement with our dose response surface analysis, this suggests that cantharidin acts by potentiating the effect of calyculin A. We next mapped the e scores of strains described in Fig. 3.4b (sensitive to either compound or the cocktail) to a network comprising fifteen types of protein interactions as denned by BioGRID[18](Fig. 3.5a). We found that particular neighborhoods of the network were enriched with genes having negative e scores. In addition, searching this network for highly connected modules of proteinprotein interactions (see methods) identified several modules with an average e score of less than zero (Fig. 3.4b). Components of four major processes were found to be major contributors to the observed drug synergy: the chaperonin CCT ring complex (CCT2, CCTS, CCT4, CCT5, CCT8, TCP1), the mRNA cleavage and polyadenylation (APT) complex (PTAl,

SWD2,

CCT7, PTI1,

SSU72), the RSC chromatin remodeling complex (RSC58, ARP7, RSC8, STH1) and golgi-related vesicular transport consisting oftheretromer complex (PEP8, VPS5, VPS29, VPS35) and the COPI vesicle coat complex (SEC26, SEC21, RET2)(3.5b).

At least one known Glc7-related process, the

APT complex, was perturbed more effectively with the use of both drugs. It is at present unclear whether Glc7 regulates the other processes identified. Together, these results illustrate that drug cocktails can specifically target genetically unique cells within a heterogeneous population.

3.4

Discussion

This study demonstrated the utility of chemogenomic profiles for the study of drug combinations. Understanding how drug combinations affect cell physiology is essential for three reasons: 1) predict adverse drug interactions 2) develop better strategies for designing effective drug combination treatment regimens and 3) develop chemical-genetic strategies for studying cellular processes. Recent advances in 'combination high-throughput screening' (cHTS)[22, 3] promise the identification of many drug combinations that result in non-additive interactions. However, determining the links between the observed drug interaction and the underlying molecular mechanism is not straightforward. In this study, we showed that the application of cantharidin and calyculin A at lower concentrations increased the selectivity of the inhibitors and integration with genetic networks can be used to map the nature of this selectivity. Since Glc7 is an essential protein and whose functions can be partitioned by different mutant alleles (Chapter 2), cantharidin and calyculin A may

CHAPTER

3. ANALYZING

DRUG-DRUG

INTERACTIONS

49

inhibit different aspects of Glc7 function that when used in combination, inhibits Glc7 activity more effectively while lowering off-target effects. One application of this work is to use this approach to iteratively optimize the use of inhibitors that work similarly in order to perturb cellular pathways more specifically.

3.5 3.5.1

Materials and Methods Dose response surface analysis

For detailed methods and theoretical explanations of the models, the reader is kindly referred to the paper by Lehdr et al. [13] from which I based the analysis described herein. Inhibition screens: For each drug, 1.25-fold serial drug dilutions were performed in DMSO. Wildtype yeast (BY4743) were grown overnight to saturation and diluted into YPD media to an ODeoo of 0.2 and aliquoted in an 8x8 matrix (96-well plate). Each drug combination pair was then added to yield the dose response matrix and grown using Tecan GENios microplate readers for up to 30h. At least two replicates were conducted for each growth condition. We used area under the growth curve (AUGC) as a metric to capture both defects in growth rate and carrying capacity. For each well, we measured AUGC and normalized it to the AUGC of the no-drug control to produce a 8 x 8 growth fitness matrix (available as supplementary data on the authors' website). For the analysis described here, this matrix is converted into an inhibition matrix using the equation 1 = 1 - D where I is the inhibition matrix and D the growth fitness matrix. Response Surface Modeling: Sigmoidal dose response I = E • Ca/(Sa + Ca) as a function of concentration C were fitted to single drug data where E is the limiting response at high concentration, S is the effective concentration, and a is the Hill coefficient determining the steepness of transition. We used these fitted curves for the surface model analysis. The HSA model is based on the concept that an interaction exists when an effect of the combination exceeds the maximum effect exerted by any of the single components, i.e. IHSA — max(lx,

Iy) where Ix and Iy are effects of the

single agents at their respective concentrations of X and Y. Loewe additivity is the concept that zero interaction occurs when the response produced by the combination is the additive response of the single components. An iterative approach was used to find the inhibition Ix,0etue that satisfied the Loewe additivity combination index (CI) equation: (X/Xj) + (Y/Yj) = 1 where Xj and Yj are the single-agent effective concentrations. For each concentration pair, the single agent curves were interpolated to find Xj and Yj that produced / and the corresponding combination index CI was calculated. The equation was solved iteratively using nonlinear least squares (lsqnonlin) with MATLAB to converge on a value of I that produced CI = 1. We used the Lehdr et al.[13] 'Bliss boosting' model for boost levels other than multiplicative. Dose response surfaces were fit to the equation:

IBLISS

= Ix +

IY

+ (/? —

Emin)(IxlY/ExEy),

where E m j n = min(Ex,Ey) and E x and Ey are the limiting single agent reponse. 0 is the amount


50

of boosting above EM AX, the greater of the single agent response. The type of boosting can be further classified by using different reference levels: (3 = —Emax is 'canceling', /? = Emin — Emax is 'suppressing', /? = 0 is 'masking', /3 = Emin(l and j3 — (1 — Emax)

- Emax)

is multiplicative (Bliss independence)

is 'saturating'. The reference level that most closely matches the /? value

obtained from the fit is assigned. Since Bliss boosting can only be used provided that the single agent responses increases monotonically, nocodazole was excluded from this analysis as inhibition for this compound decreases for concentrations > 8.2/ug/ml. Finally, a power-law potentiation model was used to describe a combination where one drug's curve is shifted with a power-law slope p above an enhancer concentration Ypor[13]. The inhibition I.POT

= max(Ix(C), \Y) where I x (C) is the single agent response curve of the potentiated

compound at a shifted concentration C = X[l + (Y'/YPOT)\P\]st3n^P) where sign(p) is a unit sign with values (-1, 0, +1). The free parameters of this model p and

YPOT

are estimated using the

Nelder-Mead method (the fminsearch function in MATLAB was used). For Bliss boosting and potentiation models, parameters were estimated by minimizing sum of the squared errors between the model and the observed dose response surface: SS = 2_^(Idata — Imodei)2- For each compound pair, data

all four model surfaces were evaluated and the model with the overall best fit with the least sum of squared errors was assigned to that combination. A best fit was chosen from HSA, Loewe, Bliss boosting, and potentiation in that order depending on which was first consistent with SS m j„, the minimum sum-of-squares error. Consistency was defined using a tolerance threshold defined as: SS - SSmjn < 0.1. The results from these analysis are summarized in Table 3.1.

3.5.2

Quantifying deletion strain drug interactions.

Deletion sensitivity profiling screens were performed as described in Chapter 2. We used rankproduct analysis[4] to identify strains that were sensitive to each compound and the cocktail (3.4b) . Briefly, log2 ratios were calculated pairwise for each control and treatment array and rank sorted. The product of the ranks for each of pair of arrays was calculated. To estimate significance, the procedure was repeated (n=1000) but this time the array values were randomized to generate a null distribution. We then counted the number of simulated rankproduct values that are smaller than a given experimental rankproduct value. For each condition, a false discovery rate of 0.1 was used to identify the strains mapped to the interaction network shown in Fig. 3.5a. To quantify the degree of interaction between cantharidin and calyculin A (Fig. 3.4b), we calculated a z-score (W x ) for each strain x from the log2 ratios of treatment versus control for each condition: cantharidin (33/zM) alone, calyculin A (1.6/JM)

alone and the cocktail (cantharidin 33/itM, calyculin A

1.6(JM).

This was performed sepa-

rately for the heterozygous and homozygous pools to normalize for differences in the ratios between the two pools as the homozygous pool screens were performed for 5 generations while the heterozygous essential pool screens were performed for 20 generations. For each strain x, the z-scores for each Condition w e r e t h e n Used t o c a l c u l a t e (ex = Wx,ca,ntharidin,calyculinA

~ Wxcantharidin

•

WXtcalyculinA

CHAPTER

3. ANALYZING

DRUG-DRUG

INTERACTIONS

51

to characterize the degree of drug interaction in that strain. We filtered strains for which e could not be accurately assigned because the strain was highly sensitive to either single drug condition or had low overall TAG intensity in the control. To identify these strains, we calculated for each strain, the maximum possible e value by calculating: emaXtX = Bgd - WXtCantharidin • Wx,calyculinA where Bgd is the background intensity (~60) simulating that the strain is highly sensitive to the cocktail. Strains for which emax is negative were excluded from the analysis and are represented as gray nodes in Fig 3.5a.

3.5.3

Yeast Interaction Network

Genetic, physical interaction data for S. cerevisiae were obtained from BioGRID[18]. For this analysis, physical interactions included interactions identified by affinity capture, biochemical activity, co-crystal, co-fractionation, co-purification, reconstituted complex. Genetic interactions included interactions identified by synthetic lethality, synthetic rescue, synthetic growth defect, dosage rescue, dosage lethality and dosage growth defect. Genetic congruency interactions were determined from synthetic lethal interactions using a congruency score > 10 as described by Ye et al. [19]. Cofitness interactions were obtained by calculating the correlations of strain fitness in a compendium of chemical-genetic fitness profiles[ll]. For the purpose of this analysis, we combined all interactions to create a merged network to contextualize the effect of the drug cocktail (Fig. 3.4a). Networks were visualized in Cytoscape[17] and the MCODE plugin[l] was used to identify highly connected modules using the default parameters. The search for modules was restricted to the set of physical interactions to identify complexes. For each module identified, we calculated an average drug intern

action score Si = jj ^ J e» where e» represents the drug interaction value for gene i in module j and n represents the number of nodes in that module. Nodes without an assigned ( (gray nodes) were excluded from this analysis.


3.6

52

Tables Table 3.1: Fit results for drug surface modeling

Combination

Model Assigned

SS

Canth-Canth

Loewe Additivity

0.079 -

Canth-Caly. A Potentiation

0.66

Parameter/its

p = 0.84, YPOT = 41//M

Canth-Rapa

Bliss Boost (suppressing) 1.16

-0.66

Canth-Meth

Loewe

0.20

-

Canth-Noc

Not assigned

-

-

Canth-Mech

Loewe

0.23

-

Canth-Lat. A

Loewe

0.13

-

Canth-MMS

Bliss Boost (saturating)

0.20

P = 0.28

Caly. A - Calyculin A Rapa - Rapamycin Meth - Methorexate Noc - Nocodazole Mech - Mechlorethamine Lat. A - Latrunculin A MMS - Methyl methanesulfonate


3.7

53

Figures

HSA j § Potentiation Loewe (Additive) Bliss Boost

•K

o

nic; mycin

o. c

£ |

lico azole

Io

i—

rexate

JE

rop morph

aridin

c

:yurea

cc

omyl

I1

apaimycin

tunicamycin

£

sz E

E

2

!=

m

Figure 3.1: Best fit of dose response surfaces for 10 compound pairs. Each square in the matrix (pair of compound) represents a 6x6 dose response surface and the color indicates the model that best describes the surface(See Materials and Methods).


Cantharidin(iiM)

Cantharidin (fxM)

Cantharidin(nM)

26 33 41 51 64 80 100

0

86 33 41 51 64 80 100

54

Antagonism • 0.5

o 8

J

-0.5

Synergy ' OMSO ' Cantharidin ' CalyculinA Cantharidin+Caiyculin A

Cantharidin(fiM)

• • -

OMSO Cantharidin Cantharidin Cantharidin + Cantharidin

Cantharidin(nM)

OMSO Cantharidin MMS Cantharidin+ MMS

e

Cantharidin(nM) 0

Not Assigned

Loewe

Cantharidin(M-M)

Cantharidin(jiM)

26 33 41 61 64 80 100

Loewe

26 33 41 61 64 80 100

Loewe

Bliss Boost (suppressing)

Figure 3.2: Cantharidin and calyculin A are synergistic in inhibiting growth. Dose response matrix of cantharidin treatment with eight other compounds in addition to mocktreatment with itself. The matrices are colored according to e using a Bliss model of independence with yellow (synergistic) for e < 0 and blue for e > 0 (antagonistic). Representative growth curves are shown for different pairs of drug concentrations. Treatment of cantharidin against itself is a control for a purely additive interaction. The best-fit shape model assigned to this combination by surface response analysis is indicated.


55

©Single drug (IC10) and cocktail (IC10) sensitive ©Cocktail (IC10) sensitive only Osingle drug (IC10) sensitive only

Figure 3.3: Chemical combination exert distinct effects. Venn diagram illustrating the number of strains identified as sensitive in i) both single drug and cocktail screens ii) sensitive only in the cocktail screen iii) sensitive only in the single drug screen. All screens were performed at the same inhibitory concentration of ICio- A strain was classified as sensitive to the single drug treatment if it was sensitive to either cantharidin(100/xM) or calyculin A (2.5/iM). A strain was classified as sensitive to the cocktail drug treatment if it was sensitive to the combination of cantharidin 33/uM, calyculin A 1.6/xM.


Epsilon = Wcanth.calyA - (WcanthWcalyA)

DMSO

Cantharidin

Calvculin A

Cantharidin + Calyculin A

55

c '

Epsilon Heterozygous Deletion Homozygous Deletion Strains Strains

m

SEN2 5CT4 RETT

m

RSCS8

PHI

Synergistic

-7-

t< 6.36-10

Epulen

Antagonistic

YDR167C

tSm ESP1 RPL25 YJL009W

cantharidin

calyculin A

calyculin A

Figure 3.4: Mechanistic insights into drug interactions, (a) Four different pairs of concentrations for cantharidin and calyculin A ([OfiM, QfM], [33/uM, 0/iM], [0/xM, 1.6/xM], [33/uM, 1.6/xM]) were screened using DSP and the degree of drug interaction for each strain was quantified, (b) Distribution of epsilon «•:, a metric that quantifies the degree of interaction between cantharidin and calyculin A for each strain (see Methods), (c) Heat map representing strain sensitivity (blue to yellow) and drug interaction (yellow to red) for the top 20 deletion strains exhibiting synergistic interactions, (d) Cantharidin boosts calyculin A sensitivity. Strains identified as synergistic or antagonistic in the cocktail experiment (Cantharidin 33/tM;calyculin A 1.6/iM) were used in calculating the average sensitivity to cantharidin and calyculin A at the single drug IC10 concentrations (Cantharidin 100/xM; calyculin A 2.5//M). Boxplot of the distribution of scores are shown. Synergistic strains are significantly more sensitive to calyculin A than cantharidin.


57

Figure 3.5: Identifying biological processes targeted by drug combinations, (a) An interaction network with nodes colored (white to red) by degree of drug synergy as quantified by epsilon e. Grey nodes represent strains for which e: values could not be accurately assigned (See Methods), (b) Modules representing protein complexes with the highest average synergy score S.

CHAPTER

3.8

3. ANALYZING

DRUG-DRUG

INTERACTIONS

58

Bibliography

[1] G. D. Bader and C. W. Hogue. An automated method for finding molecular complexes in large protein interaction networks. BMC bioinformatics, 4:2, Jan 13 2003. [2] CI Bliss. The toxicity of poisons applied jointly. Ann Appl. Biol, 26:585-615, 1939. [3] A. A. Borisy, P. J. Elliott, N. W. Hurst, M. S. Lee, J. Lehar, E. R. Price, G. Serbedzija, G. R. Zimmermann, M. A. Foley, B. R. Stockwell, and C. T. Keith. Systematic discovery of multicomponent therapeutics. Proceedings of the National Academy of Sciences of the United States of America, 100(13):7977-7982, Jun 24 2003. [4] R. Breitling, P. Armengaud, A. Amtmann, and P. Herzyk. Rank products: a simple, yet powerful, new method to detect differentially regulated genes in replicated microarray experiments. FEBS letters, 573(l-3):83-92, Aug 27 2004. [5] J. A. Brown, G. Sherlock, C. L. Myers, N. M. Burrows, C. Deng, H. I. Wu, K. E. McCann, O. G. Troyanskaya, and J. M. Brown. Global analysis of gene function in yeast by quantitative phenotypic profiling. Molecular systems biology, 2:2006.0001, 2006. [6] M. Costanzo, G. Giaever, C. Nislow, and B. Andrews. Experimental approaches to identify genetic networks. Current opinion in biotechnology, 17(5):472-480, Oct 2006. [7] J. B. Fitzgerald, B. Schoeberl, U. B. Nielsen, and P. K. Sorger. Systems biology and combination therapy in the quest for clinical efficacy. Nature chemical biology, 2(9):458-466, Sep 2006. [8] G. Giaever, A. M. Chu, L. Ni, C. Connelly, L. Riles, S. Veronneau, S. Dow, A. Lucau-Danila, K. Anderson, B. Andre, A. P. Arkin, A. Astromoff, M. El-Bakkoury, R. Bangham, R. Benito, S. Brachat, S. Campanaro, M. Curtiss, K. Davis, A. Deutschbauer, K. D. Entian, P. Flaherty, F. Foury, D. J. Garfmkel, M. Gerstein, D. Gotte, U. Guldener, J. H. Hegemann, S. Hempel, Z. Herman, D. F. Jaramillo, D. E. Kelly, S. L. Kelly, P. Kotter, D. LaBonte, D. C. Lamb, N. Lan, H. Liang, H. Liao, L. Liu, C. Luo, M. Lussier, R. Mao, P. Menard, S. L. Ooi, J. L. Revuelta, C. J. Roberts, M. Rose, P. Ross-Macdonald, B. Scherens, G. Schimmack, B. Shafer, D. D. Shoemaker, S. Sookhai-Mahadeo, R. K. Storms, J. N. Strathern, G. Valle, M. Voet, G. Volckaert, C. Y. Wang, T. R. Ward, J. Wilhelmy, E. A. Winzeler, Y. Yang, G. Yen, E. Youngman, K. Yu, H. Bussey, J. D. Boeke, M. Snyder, P. Philippsen, R. W. Davis, and M. J o h n s t o n . Functional

profiling of the saccharomyces cerevisiae genome. Nature, 418(6896):387-391, Jul 25 2002. [9] G. Giaever, P. Flaherty, J. Kumm, M. Proctor, C. Nislow, D. F. Jaramillo, A. M. Chu, M. I. Jordan, A. P. Arkin, and R. W. Davis. Chemogenomic profiling: identifying the functional interactions of small molecules in yeast. Proceedings of the National Academy of Sciences of the United States of America, 101(3):793-798, Jan 20 2004.

CHAPTER

3. ANALYZING

DRUG-DRUG

INTERACTIONS

59

[10] G. Giaever, D. D. Shoemaker, T. W. Jones, H. Liang, E. A. Winzeler, A. Astromoff, and R. W. Davis. Genomic profiling of drug sensitivities via induced haploinsufficiency. Nature genetics; Nature genetics, 21(3):278-283, Mar 1999. [11] M. E. Hillenmeyer, E. Fung, J. Wildenhain, S. E. Pierce, S. Hoon, W. Lee, M. Proctor, R. P. St Onge, M. Tyers, D. Koller, R. B. Altman, R. W. Davis, C. Nislow, and G. Giaever.

The

chemical genomic portrait of yeast: uncovering a phenotype for all genes. Science (New York, N. Y), 320(5874):362-365, Apr 18 2008. [12] William Lee, Robert P. St Onge, Michael Proctor, Patrick Flaherty, Michael I. Jordan, Adam P. Arkin, Ronald W. Davis, Corey Nislow, and Guri Giaever. Genome-wide requirements for resistance to functionally distinct dna-damaging agents. PLoS Genet, 1(2), 08:19: 2005. [13] J. Lehar, G. R. Zimmermann, A. S. Krueger, R. A. Molnar, J. T. Ledell, A. M. Heilbut, G. F. Short 3rd, L. C. Giusti, G. P. Nolan, O. A. Magid, M. S. Lee, A. A. Borisy, B. R. Stockwell, and C. T. Keith. Chemical combination effects predict connectivity in biological systems. Molecular systems biology, 3:80, 2007. [14] P. Y. Lum, C. D. Armour, S. B. Stepaniants, G. Cavet, M. K. Wolf, J. S. Butler, J. C. Hinshaw, P. Garnier, G. D. Prestwich, A. Leonardson, P. Garrett-Engele, C. M. Rush, M. Bard, G. Schimmack, J. W. Phillips, C. J. Roberts, and D. D. Shoemaker. Discovering modes of action for therapeutic compounds using a genome-wide screen of yeast heterozygotes.

Cell,

116(1):121-137, Jan 9 2004. [15] L. W. Parks, S. J. Smith, and J. H. Crowley. Biochemical and physiological effects of sterol alterations in yeast-a review. Lipids, 30(3):227-230, Mar 1995. [16] A. B. Parsons, A. Lopez, I. E. Givoni, D. E. Williams, C. A. Gray, J. Porter, G. Chua, R. Sopko, R. L. Brost, C. H. Ho, J. Wang, T. Ketela, C. Brenner, J. A. Brill, G. E. Fernandez, T. C. Lorenz, G. S. Payne, S. Ishihara, Y. Ohya, B. Andrews, T. R. Hughes, B. J. Prey, T. R. Graham, R. J. Andersen, and C. Boone. Exploring the mode-of-action of bioactive compounds by chemical-genetic profiling in yeast. Cell, 126(3):611-625, Aug 11 2006. [17] P. Shannon, A. Markiel, O. Ozier, N. S. Baliga, J. T. Wang, D. Ramage, N. Amin, B. Schwikowski, and T. Ideker. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome research, 13(ll):2498-2504, Nov 2003. [18] C. Stark, B. J. Breitkreutz, T. Reguly, L. Boucher, A. Breitkreutz, and M. Tyers. Biogrid: a general repository for interaction datasets. Nucleic acids research, 34(Database issue) :D535-9, Jan 1 2006.

CHAPTER

3. ANALYZING

DRUG-DRUG

INTERACTIONS

60

[19] P. Ye, B. D. Peyser, X. Pan, J. D. Boeke, F. A. Spencer, and J. S. Bader. Gene function prediction from congruent synthetic lethal interactions in yeast. Molecular systems biology, 1:2005.0026, 2005. [20] P. Yeh, A. I. Tschumi, and R. Kishony. Functional classification of drugs by properties of their pairwise interactions. Nature genetics, 38(4):489-494, Apr 2006. [21] L. Zhang, K. Yan, Y. Zhang, R. Huang, J. Bian, C. Zheng, H. Sun, Z. Chen, N. Sun, R. An, F. Min, W. Zhao, Y. Zhuo, J. You, Y. Song, Z. Yu, Z. Liu, K. Yang, H. Gao, H. Dai, X. Zhang, J. Wang, C. Fu, G. Pei, J. Liu, S. Zhang, M. Goodfellow, Y. Jiang, J. Kuai, G. Zhou, and X. Chen. High-throughput synergy screening identifies microbial metabolites as combination agents for the treatment of fungal infections. Proceedings of the National Academy of Sciences of the United States of America, 104(11):4606-4611, Mar 13 2007. [22] G. R. Zimmermann, J. Lehar, and C. T. Keith. Multi-target therapeutics: when the whole is greater than the sum of the parts. Drug discovery today, 12(l-2):34-42, Jan 2007.

Chapter 4

A global perspective of t h e genetic basis for carbonyl stress resistance 4.1

Abstract

The accumulation of protein adducts caused by carbonyl stress is a hallmark of cellular aging and other diseases. Understanding the effects of carbonyl stress will provide better insight into disease mechanisms and development of therapeutics to ameliorate their effects. Multiple genomewide genetic screens were performed to identify cellular functions involved in resistance to carbonyl stress. We found that the genetic requirements for MG and GLY resistance are distinct and that different carbonyl stresses require distinct genetic cohorts for cell survival. We demonstrate the importance of macromolecule catabolism processes for resistance to MG, confirming and extending known mechanisms of MG toxicity by modification of DNA, RNA and proteins. The effect of glyoxal on the cell is poorly understood, but by using multiple genomewide approach to dissect the effect of glyoxal in vivo, we provide a comprehensive view of carbonyl stress and provide a resource for future studies in other cell types. When combined with ROS, we found that a diverse range of cellular functions are affected by oxidative and carbonyl stress. Furthermore, a quantitative epistasis analysis allowed us to uncover novel glyoxal-dependent genetic interactions.

4.2

Introduction

A hallmark of aging and its attendant ailments is the accumulation of molecular defects, including DNA lesions , oxidized proteins and protein carbonylation.[61, 44, 23, 5, 33]. Identifying the cellular functions necessary for providing resistance to these toxic molecules will provide insight into the molecular mechanisms that underlie age-related diseases and could suggest a means to ameliorate

61

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

62

them or manage their consequences. It is well documented that normal cellular metabolism produces reactive molecules. For example, reactive oxidative species (ROS), comprising superoxide anions (0 2 ~), hydrogen peroxide (H2O2) and hydroxy radicals (-OH), are by-products of aerobic metabolism. They have been well-studied with respect to their roles in causing DNA damage, protein oxidation and lipid peroxidation [7, 60, 62, 4]. Studies with yeast have proven useful for understanding diverse of environmental stresses, including metabolism-induced stress[18]. Cellular components are susceptible to modifications (and damage) by reactive carbonyl compounds (RCC) that are produced during lipid peroxidation and glycoxidation[52, 12] yet there is no comprehensive analysis of the consequences of these modifications. On the cellular level, carbonyl stress (CS) leads to the formation of adducts resulting in the formation of cross-links within and between macromolecules, resulting in both protein dysfunction and genotoxicity. At the organismal level, RCCs are causally related to the development of vascular complications that characterize diabetes, cell aging and Alzheimer's and diverse inflammatory diseases such as chronic obstructive pulmonary disease and ischemia/reperfusion [14]. RCC-induced pathologies develop when glucose reacts with macromolecules to form advanced glycation end products (AGEs) that cause tissue damage. Oxidation of these glycation products leads to the formation of a-oxoaldehdyes such as glyoxal (GLY), 3-deoxyglucosone (3-DG) and methylglyoxal (MG). Each of these highly reactive aldehydes are potent glycation agents that react with proteins, further exacerbating the production of AGE and a-oxoaldehdyes. Both methylglyoxal and glyoxal are readily detectable in human plasma and urine[15, 67]. Glyoxal can be formed as a major lipid and DNA oxidative degradation product. It can also be formed via autoxidation of glycolaldehyde[6]. Methylglyoxal is formed as a by-product of several metabolic processes that includes threonine catabolism[39] and lipid peroxidation [49]. Methylglyoxal can also arise enzymatically during glycolysis from triose phosphate intermediates glyceraldehyde 3-phophate and dihydroxyacetone phosphate[66, 42]. In addition to intracellular production of glyoxal and MG, there are a large number of environmental sources such as cigarette smoke and automobile exhaust, both abundant sources of carbonyls[54, 78, 42]. The thermal processing of food during the Maillard/browning reaction can also lead to formation of methylglyoxal and other aldehydes[41, 74]. The cell's principal defense against AGE's and aldehydes, the glyoxalase system, is a wellconserved pathway that detoxifies methylglyoxal and glyoxal in the presence of glutathione (GSH). In this reaction, glyoxal is converted to glycolate and MG to D-lactate[67]. Glyoxal and MG can also be detoxified by NADPH-dependent aldose reductases [29]. Previous work using yeast as a model to study the effects of a-oxoaldehydes have focused on methylglyoxal, and have shown that the conserved HOG MAP kinase pathway is important for the induction of methylglyoxal-responsive genes[2, 37]. Here, we take an unbiased, genome-wide approach to greatly expand our understanding of the cellular response to carbonyl stress. Systematic genetic studies have been useful for identifying synthetic sick or synthetic lethal

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

63

interactions[72, 57, 77]. More recently, refined analysis tools have permitted the detection and classification of alleviating interactions in multi-mutant strains. St. Onge et al. [45] showed that the number of aggravating and alleviating interactions that could be identified was increased when performed in the presence of a perturbation that affects the process under consideration, such as the alkylating agent MMS for understanding DNA repair. We reasoned that, given the pleiotropic effects of CS stress, epistasis-based approaches could be used to dissect how cells respond to these perturbations. Here we used glyoxal to perturb double mutant strains deleted for genes suspected to be involved in the response to carbonyl stress. Our experiments identified the cohort of genes involved in CS and their epistatic relationships. The result is a comprehensive view of the physiology that underlies the cellular response to normal and pathological CS.

4.3

Results

4.3.1

Methyglyoxal and glyoxal inhibit yeast growth

Studies in yeast have shown that methylglyoxal is a potent inhibitor of yeast growth[2, 37, 1, 28, 8, 27]. Using a quantitative fitness growth assay, we found that both methyglyoxal and glyoxal (Fig. 4.1a) inhibit yeast growth with methylglyoxal being approximately ten-fold more potent than glyoxal (Fig. 4.1b). We also found that aminoguanidine (AG), a clinically tested a, /3-dicarbonyl scavenging agent[ll], effectively rescues the growth inhibition caused by both compounds (Fig. 4.1b). We note that treatment with AG (20mM) in the absence of methylglyoxal and glyoxal inhibits yeast growth, likely due to the reaction of AG with pyruvate to form a hydrazone adduct at high concentrations[68].

4.3.2

Different H O G pathway components are required for resistance t o M G a n d Glyoxal

In yeast, the high-osmolarity glycerol (HOG) mitogen-activated protein kinase (MAPK) pathway regulates osmotic homeostasis[46]. Previous work has determined that the HOG pathway is important for mediating methylglyoxal resistance presumably via its induction of genes involved in methylglyoxal metabolism[2, 28]. Among these induced genes is GL01, which codes for glyoxalase I, an enzyme that converts MG into S-D-lactoylgluthathione in the presence of glutathione [27]. We found that strains deficient in various components of HOG signaling showed differential sensitivity to glyoxal and methylglyoxal relative to mutants deficient in glyoxalase activity. For example, strains deleted for GL01, GL02, GSH1 were more sensitive to methylglyoxal than strains deleted for HOG1, PBS2, SSK1 and SSK2 while in contrast, the opposite was observed under glyoxal stress (Fig. 4.1c). These results suggest HOG pathway mediates resistance to GLY and MG in distinct ways.

CHAPTER 4. CARBONYL STRESS RESISTANCE

4.3.3

64

Genomewide fitness profiling identifies differences between t h e cellular effects of M G and GLY

MG and GLY exert different effects in vivo e.g. MG and GLY induce distinct signals for MAP family kinases in human endothelial cells[3]. To systematically identify additional differences, we screened both compounds by Deletion Sensitivity Profiling (DSP), against a collection of 4,700 S. cerevisiae homozygous deletion strains [26, 25, 36, 20, 19], and identified 458 deletion strains that were significantly sensitive (false discovery rate < 0.05) to at least one carbonyl compound. Gene Ontology (GO) analysis identified cellular functions that mediate resistance to each compound (Fig. 4.Id). For comparison, we performed the same analysis for strains sensitive to other oxidants identified by a previous study[69] which found that the effects of oxidative stress are pleiotropic and specific to each oxidant treatment. From this study, I found additional functions not previously associated with ROS were important for carbonyl stress resistance. As expected from our results above, the genome-wide fitness profiles of GLY and MG were also very dissimilar. MG sensitive genes were enriched for protein, mRNA and DNA metabolic processes. This agrees with the known mechanism of action of MG in forming DNA, RNA and protein adducts[31]. In contrast, for glyoxal, we observed functional enrichment for genes involved in glucose metabolism, peroxisomal and signal transduction processes. Taken together, these results demonstrate that ROS and carbonyl stress affect a diverse range of cellular processes. These pathways may constitute important targets for therapeutic intervention.

4.3.4

Carbonyl adduct detoxification

GLY and MG are potent arginine-directed glycating agents and exert one of their effects through covalent cross-linking of proteins and via modification of arginine residues leading to protein dysfunction [52]. Protein modification by MG has been shown to activate ubiquitin/proteasome-dependent proteolysis [13]. In agreement with these observations, we found multiple strains deleted for genes involved in ubiquitin-dependent protein degradation (shplA,

bsd2A, stp22A, doaJÂ, swmlA,

rad6A, epslA,

doalA, snflA,

rttlOlA,

grrlA,

vps25A, deflA,

bstlA,

cdc26A,

srn2A, vps36A, ubx2A, ubxJÂ,

vps20A, ydjlA, snf8A, vps28A, brolA) were sensitive to methylglyoxal. These results demonstrated that protein catabolism play a major role in detoxifying glycated proteins. Besides producing protein adducts, methylglyoxal also reacts with guanine bases in RNA and DNA. Because Messenger RNA (mRNA) quality control is important for maintaining t h e fidelity of gene expression, we expected

MG to induce defects in strains involved in mRNA metabolism. For example, the first step in mRNA decay is deadenylation followed by mRNA degradation by decapping and either 1) 5'—>3' decay or 2) 3'—>5' decay[17]. Consistent with this, several strains deficient in deadenylation (ccrÂ, not5A), decapping and 5'—>3' decay(dhhlA,patlA,lsm6A,

pop2A,

IsmlA, lsm7A) and 3'—>5' decay (ski8A,

ski7A, skiSA, ski2A) were sensitive to methylglyoxal. Furthermore, multiple strains involved in


DNA repair (mms4A, radl8A, rpn4A, rad59A, rad57A, rad55A, hprlA, mus8lA,

65

radSlA, radJÂ,

rad6A, rpb9A, rad54A, srs2A, deflA, doalA, rad5A, mms22A, rad52A, radUA, mrellA,

tho2A,

rad50A, snf2A, radlA, mmslA) were also sensitive to methylglyoxal. In contrast, genes involved in DNA repair and mRNA decay pathways were not required for glyoxal resistance. For glyoxal, the requirement for catabolic functions was less pronounced: fewer strains involved in protein degradation (reglA,doa4A,

4.3.5

ubc8A, jemlA,

ydjlA) were identified.

Identifying glyoxal resistant strains

Pooled fitness profiling assays have, to date, focused on identifying deletion strains with reduced fitness in the presence of a chemical perturbation. This is partly due to technical difficulties in assessing resistance when low doses (e.g. ICio) of compound are applied during a typical 'drop-out' screen. For example, in these cases, many strains identified as significantly resistant are slow growers in the absence of perturbation and appear artificially resistant because these slow growing strains have lower tag intensity distribution leading to unreliable results[20]. To circumvent this problem, we inverted and optimized the HOP assay to identify strains resistant to glyoxal. Here we designed an 'enrichment screen' where the entire deletion collection is treated with a high dose (IC50) and maintained at > 50% inhibition over 20 generations of growth (Fig. 4.2a). This strong selection revealed strains that are resistant to the compound - this change in composition of the pool is reflected by the small number of tags with intensities above background following tag hybridization (Fig. 4.2a). Multiple strains were significantly resistant to glyoxal and we confirmed most of these strains by individual strain growth analysis (Fig. 4.2b). Interestingly, one of these genes was PTC1, a type 2C protein phosphatase that negatively regulates the HOG pathway by dephosphorylating Hogl providing further evidence that upregulation of the HOG pathway is required for glyoxal resistance. Gene Ontology enrichment of the glyoxal-resistant strains showed enrichment for functions such as conjugation, vesicle-mediated transport, transport, cellular homeostasis and membrane organization and biogenesis (Fig. 4.2c).

4.3.6

A sensitized suppressor screen to understand glyoxal stress

To further characterize the effect of glyoxal, we performed Multicopy Suppressors Profiling (MSP), a genome-wide assay that screens DNA clone libraries competitively to identify genes that confer resistance to compounds when overrepresented[26]. In the presence of high doses of glyoxal, we found that SIR2 deletion mutants displayed resistance to glyoxal. We also found that nicotinamide, an endogenous inhibitor of Sir2[9], conferred resistance to glyoxal (Fig. 4.3a) while, treatment with isonicotinamide, an antagonist of nicotinamide inhibition and an activator of Sir2 deacetylase activity[56], conferred sensitivity to glyoxal (Fig. 4.3a). To improve the MSP selection, we used isonicotinamide to sensitive the screen with glyoxal. A multicopy suppressor pool was grown in the presence of glyoxal or glyoxal supplemented with isonicotinamide (Fig. 4.3b). We identified


66

multiple suppressor loci where multiple adjacent ORFs were overrepresented in glyoxal compared to control (Table 4.1). By sensitizing the screen with isonicotinamide, more suppressor loci were identified. One of the most significant loci identified was GLOl, an enzyme that metabolizes glyoxal; an observation that validates our experimental approach. Multiple suppressors derived from this screen were confirmed by isogenic tests with singly cloned ORFs (Fig. 4.4).

4.3.7

Overexpression of aldehyde reductases suppress glyoxal toxicity

We identified several aldehyde reductases that, when overexpressed, conferred resistance to glyoxal. GRE3 is an aldehyde reductase that is capable of metabolizing methylglyoxal and is regulated by the HOG pathway[1, 2]. YPR1 is also an aldehyde reductase that has high specificity activity for 2methylbutyraldehyde[16]. ADH6 is an alcohol dehydrogenase able to detoxify aldehydes in vivo[34]. All three genes were strong suppressors of glyoxal and were confirmed by single strain analysis (Fig. 4.3c). Interestingly, although each of these genes were identified by overexpression profiling, none were sensitive to glyoxal as single-gene deletions - reflecting potential functional redundancy between these genes - and underscoring the benefit of using MSP in additional to DSP.

4.3.8

N A D P H generating pathways buffer glyoxal resistance

Thioredoxin and GSH are important cellular antioxidants that are maintained in their reduced state by thioredoxin reductase and glutathioine reductase using NADPH as a reductant[22]. Cells have several metabolic sources of NADPH, including glucose (via the pentose-phosphate pathway), the isocitrate dehydrogenase pathway and the conversion of acetaldehyde to acetate by acetaldehyde dehydrogenase encoded by ALD6. Mutants in genes involved in all three sources of these antioxidants were sensitive to glyoxal including: ALD6, IDP1, RPE1 and TKL1. This result corroborates the importance of NADPH-dependent enzymes Gre3 and Adh6, both of which were identified as multicopy suppressors of glyoxal resistance in this study (Fig. 4.3c). Sensitivity to glyoxal varies among mutants in the pentose phosphate pathway. Sensitivity of zwfl and tall mutants is lower than sensitivity of tkll and rpel mutants, in agreement with observations made with hydrogen perxoxide[30].

4.3.9

Role of P K A Signaling

The cAMP-protein kinase A (PKA) pathway plays a central role in regulating a number of metabolic processes in S. cerevisiae. It plays a key role in nutrient sensing, in particular, glucose signaling[55, 65]. Regulation of cAMP levels is also important for intrinsic stress tolerance of yeast cells [47]. We identified multiple components of this pathway in our glyoxal screen. Loss of Ira2, a GTPaseactivating protein that negatively regulates Ras by promoting the conversion from an active GTPbound form to the inactive GDP-bound form, led to glyoxal sensitivity. Pde2 and Rpil are proteins

CHAPTER 4. CARBONYL

STRESS

RESISTANCE

67

that down-regulate RAS/cAMP/PKA pathway[32, 75] and both were critical for glyoxal resistance. cAMP is hydrolyzed by Pde2, a high-affinity cMP phosphodiesterase, and deletion of PDE2 enhanced glyoxal sensitivity while overexpression of PDE2 enhanced glyoxal resistance. Similarly, Rpil, an inhibitor of the Ras-cAMP pathway, was identified in both MSP and DSP assays. Moreover, loss of TPK1 and TPK2 (encoding isoforms of the catalytic subunit of PKA), enhanced glyoxal sensitivity. Together, these results demonstrate that down-regulation of the RAS/cAMP/PKA pathway is required for glyoxal resistance. This result agrees with previous studies that have shown that the Ras-cAMP pathway negatively regulates resistance H2O2 stress[24, 63]. We also found that a heterozygous mutant of Cdc20, a cell-cycle regulated activator of the anaphase-promoting complex or cyclosome (APC/C), was hypersensitive to glyoxal (Fig. 4.5). It has been shown that Tpk2 inhibits APC/C function via phosphorylation of Cdc20[58, 10] and it will be of interest to further investigate the link between PKA signaling and APC/C function in this context.

4.3.10

The HOG pathway suppresses glyoxal sensitivity

We compared the degree of suppression conferred by each gene with the DSP screen to identify genes that confer sensitivity when deleted and resistance when overexpressed (Fig. 4.3d). This analysis identified multiple components of the HOG pathway (PBS2, SSK1, SSK2) thereby verifying that the HOG pathway is critical for resistance to glyoxal. Overexpression of HOG1, while not identified by MSP, also conferred resistance to glyoxal (Fig. 4.3c). Because mutants of GLOl did not demonstrate the same degree of sensitivity as mutants of components of the HOG pathway (Fig. lc), we surmised that the cell contains other effectors of glyoxal resistance regulated by the HOG pathway.

4.3.11


We sought to further identify effectors of glyoxal resistance by epistasis analysis. We assessed glyoxaldependent genetic interactions by systematically generating multiple combinations of double-deletion strains for quantitative fitness analysis[45]. ~800 double-deletions between 15 single deletion strains sensitive to glyoxal and 24 deletion strains resistant to glyoxal were constructed (Fig. 4.6a.) and grown in the presence and absence of glyoxal. Each double mutant was constructured twice independently and replicate gene pairs were highly correlated(Fig. 4.6b) demonstrating the robustness of our approach. Using fitness values for double and single-deletion strains, we quantified the genetic interaction e between gene pairs where exy = Wxv — Wx • Wv and Wxy

is the fitness of the

double mutant strain and W x and Wj, are the fitness of the single mutant strains. Interactions with negative e values are classified as aggravating and interactions with positive e values are classified as alleviating. By performing our screens at multiple drug doses with both resistant and sensitive mutant strains, we could identify subtle genetic interactions that ranged from aggravating to complete suppression. For example, in 5mM glyoxal, loss of RPE1 severely aggravated glyoxal sensitivity of a pbs2A strain where in contrast, in lOmM glyoxal, loss of ERV14 completely suppressed glyoxal


68

sensitivity of a pbs2A strain (Fig. 4.6c). In agreement with a previous study[45], a greater number of genetic interactions where uncovered when the screen was performed under stress (Fig. 4.6d). To provide a comprehensive summary of our results, we present them as a heatmap, with strains ordered by hierarchical clustering of their e values (Fig. 4.7) and found that genes of related function clustered together. For example, H0G1 and PBS2 were highly correlated in both YPD and glyoxal (r = 0.98). In contrast, other pairs of genes shared genetic interactions only in the presence of glyoxal. For example, RPN10 codes for the 19S regulatory particle (RP) of the 26S proteasome. It clusters with SPT3 and SPT8, members of the SAGA transcriptional regulatory complex. It was recently shown that the 19S RP alters SAGA to stimulate its interaction with transcription activators[35]. We also note that a strain deleted for GCN5, the catalytic subunit of SAGA, was also resistant to glyoxal (data not shown). RPE1, a D-ribulose-5-phosphate 3-epimerase, functions in the non-oxidative part of the pentose phosphate pathway and is shown to display increased sensitivity to hydrogen peroxide[30]. It clusters in the presence of glyoxal with TKL1, a transketolase that also functions in the pentose phosphate pathway[73]. Both are involved in NADPH production, important for protection against reactive oxidative stress. As expected, RPE1 and TKL1 showed an alleviating interaction (e = 0.15) demonstrating that they operate in the same pathway. In contrast, mutants of both RPE1 and TKL1 shared aggravating interactions with mutants of IRA2, SSD1, PPZ1, HOGl and PBS2 indicating that they buffer glyoxal resistance via distinct pathways. Our analysis provides the first comprehensive gene-dose analysis of carbonyl stress resistance.

4.3.12

Loss of F p s l and E r v l 4 abolishes requirement for HOG pathway

Ervl4 is a transmembrane protein found in the ER and the early Golgi compartment. It is involved in COPII cargo selection and is required for the delivery of bud-site selection protein Axl2 to the cell surface[51, 50, 40]. Our genetic interaction analysis showed that Ervl4 clustered with Fpsl, a membrane channel involved in glycerol export.

Deletions of both genes resulted in complete

suppression of glyoxal sensitivity of hogl A and pbs2A mutants demonstrating that both FPS1 and ERV14 function downstream of the HOG pathway (Fig. 4.8a). Fpsl, a plasma membrane channel regulating efflux of glycerol, mediates uptake of acetic acid, arsenite and antimonite and downregulation of Fpsl activity via Hogl was shown to confer resistance to these toxins[70, 76, 38]. In the case of acetic acid, phosphorylation of Fpsl by Hogl targets it for endocytic degradation[38]. Based on the degree of resistance conferred by loss of FPS1 to both hogl A and pbs2A mutants it is likely that glyoxal enters the cell via the same channel. It is also likely that Ervl4 regulates transport of Fpsl to the plasma membrane and loss of Ervl4 disrupts localization of Fpsl to the plasma membrane thereby rendering cells resistant to glyoxal. Indeed, we found that localization of Fpsl to the plasma membrane was disrupted in ervl4A mutants (Fig. 4.8b).

CHAPTER

4.4

4. CARBONYL

STRESS

RESISTANCE

69

Discussion

This study has focused on using a systems biological approach to identify the effects of carbonyl stress on the cell. We show that the genetic requirements for MG and GLY resistance are distinct and that different carbonyl stresses require distinct genetic cohorts for cell survival. We demonstrate the importance of macromolecule catabolism processes for resistance to MG, confirming and extending known mechanisms of MG toxicity by modification of DNA, RNA and proteins. The effect of glyoxal on the cell is poorly understood, but by using multiple genomewide approach to dissect the effect of glyoxal in vivo, we provide a comprehensive view of carbonyl stress and provide a resource for future studies in other cell types. Based on our observations, we suggest the following model (Fig. 4.9) for glyoxal resistance in yeast. Glyoxal detoxification is mediated via multiple aldehyde reductases (GRE3, ADH6, YPR1 and GL01). The pentose-phosphate pathway (RPE1, TKL1), the isocitrate dehydrogenase pathway (IDP1) and the conversion of acetaldehyde to acetate by acetaldehyde dehydrogenase (ALD6) are necessary for replenishing the cell's pool of NADPH. Maintaining the supply of NADPH is important for NADPH-dependent enzymes Gre3 and Adh6. A functioning HOG pathway is also critical for glyoxal resistance. Genetic evidence in this study demonstrated that besides GL01, there exist other downstream effectors that mediate glyoxal resistance. One such possible effector is FPS1, a known target of the HOG pathway. Loss of FPS1 abolishes sensitivity of HOG pathway mutants to glyoxal, an observation that was also made previously for acetic acid, arsenite and antimonite. Multiple mutants responsible for ER and early Golgi transport were observed to be resistant to glyoxal. In particular, loss of ERV14 also abolishes sensitivity of HOG pathway mutants. It is possible this is due to defective transport of Fpsl to the cell surface. We also found that the RAS/cAMP/PKA pathway negatively regulates glyoxal resistance. Cdc20, an activator of the anaphase-promoting complex, was haploinsufficient in glyoxal. It is currently unclear if the PKA pathway regulates Cdc20 in this case as has been previously shown[58, 10]. In summary, we demonstrated the broad range of cellular functions important for carbonyl stress resistance. By taking multiple genomewide approaches, we were able to comprehensively identify multiple genetic requirements for glyoxal stress resistance. Furthermore, performing a quantitative epistasis analysis allowed us to uncover glyoxal-dependent genetic interactions. Aging and other stress-related diseases are complex with multiple etiologies and this study reinforces this idea by demonstrating that oxidative stress by the different oxidants affect diverse cellular processes and require multiple detoxification pathways.

CHAPTER

4.5 4.5.1

4. CARBONYL

STRESS

RESISTANCE

70

Materials and Methods Strains and Reagents

Yeast were maintained in YPD media[53, 59] at 30°C unless stated otherwise. Strains and plasmids used for individual analysis in this study are listed below were obtained from the yeast deletion collection[19]. Overexpression strains were generated by transformation of 2/JM plasmids (provided generously by Dr. Charlie Boone) containing cloned ORFs into BY4743. Methylglyoxal, glyoxal, nicotinamide, isonicotinamie were purchased from Sigma-Aldrich (St. Louis, Missouri, United States). Nicotinamide and isonicotinamide were dissolved in sterile H20 and filtered sterilized.

4.5.2

Plasmids

Plasmids used in this study are shown in Table 2.2. GLOl was cloned into pRS426 and FPS1 was cloned into pUG23 by yeast gap repair [43].

4.5.3

Growth Assay

Individual deletion strains were inoculated into 100/xl of YPD and grown to saturation for ~20 h at 30°C. Overnight cultures were resuspended by shaking for 15 min and diluted into 100/xl of media in 96-well plates and grown in Tecan GENios microplate readers for 24 h. The growth rate of each culture was monitored by measuring the ODeoo every 15 min as previously described [20].

4.5.4

Genomewide screening procedure

Deletion pool construction was carried out as previously described[36, 48]. Pooled growth of the homozygous deletion pool and multicopy suppression pool were carried out for 5 and 20 generations respectively. Both pools were inoculated at an ODgoo of 0.02, grown in 48-well microtiter plates (Nunc, Rochester, New York, United States) in a volume of 700/il, and harvested automatically by a Packard Multiprobe II four-probe liquid-handling system (PerkinElmer, Wellesley, California, United States). For 20-generation experiments, cells were maintained in logarithmic phase by robotically diluting cultures every five doublings. A S. cerevisiae random genomic library (gift from Martha Cyert) constructed in a high-copy 2/xM expression vector (YEplacl95) with an average insert size of approximately 5kb was transformed into yeast (cdc28-as or BY4743) by lithium acetate transformation[21] and selected on media lacking uracil (URA-). After three days of growth ~10 6 transformants were pooled into media containing 7% DMSO, aliquoted, and stored at -80°C. Frozen aliquots were thawed and inoculated directly into URA- media to an ODgoo of 0.02 in a volume of 700/Ltl. Compound was added and the pool was grown for 20 generations in 48-well microtiter plates (Nunc, Rochester, New York, United States) at inhibitory concentrations of at least 50% (IC50) and harvested in the same way described above.


4.5.5

71

Microarray Analysis

Both DSP and MSP were analyzed using a high-density oligonucleotide TAG array manufactured by Affymetrix[48]. For Deletion Sensitivity Profiling (DSP), barcode probe intensities were extracted and processed as previously described [48]. Each array was mean normalized and the fold change (log2 control/treatment) was calculated by comparing to a set of control arrays. Tags from the homozygous pool were normalized separately from tags from the heterozygous essential pool, as were the upstream tags (uptag) and downstream tags (downtag). At least two biological replicates were carried out for each treatment condition. The log2 ratios of both tags were averaged to generate a single score for each gene. For multicopy suppression profiling (MSP), ORF probe intensities were extracted and processed in the same way as the barcode probes. Each ORF is represented by at least 2 probes and the log2 ratios of each probe were averaged to generate a single score for each gene. To identify each suppressor locus, the log2 ratio of intensities were ordered by each ORF's genomic location and analyzed using a sliding window to identify locus that have at least 2 adjacent ORFs with log2 ratios > 1.6.

4.5.6

Double mutant Strain Construction

We assessed genetic interactions among a subset of genes that confer resistance to the compound glyoxal by generating multiple combinations of double-deletion strains for quantitative fitness analysis [45] MATa haploids of single-deletion strains were obtained from the yeast deletion collection and MATa haploids of single-deletion strains were obtained from the SGA query collection (gift of Charlie Boone). In MATa haploids, genes were replaced with a kanamycin resistance marker gene (Kan r ) and in MATa haploids, genes were replaced with a nourseothricin resistance marker gene (Nat r ). Double-deletion strains were constructed by the synthetic genetic array (SGA) protocol with minor modifications[71, 64]. ~800 double-deletions between 15 deletion strains that were sensitive to glyoxal and 24 deletion strains that were resistant to glyoxal were constructed. This allows each double-deletion gene pair to be constructed independently twice (Kan r - Nat r and Nat r -Kan r ). Single deletion strains with the same genotype (Kan r -Nat r and Nat r -Kan r ) as double genotype were constructed by using the HO-deletion strain as a query strain. The fitness of individual strains was determined in the presence and absence of glyoxal. Two different concentrations of glyoxal were used to improve the dynamic range of the assay. Strains that were synthetic lethal were assigned a fitness of zero. Fitness values between each gene pair were highly correlated (Fig. 4.6b, R 2 = 0.92) demonstrating the reproducibility of the assay.

4.5.7

Growth assay for epistasis analysis

Deletion strains arrayed on YPD agar were inoculated into 96-well plates containing 100/zl of YPD using a Singer RoToR HDA (Singer Instruments). Cultures were grown to saturation for 20h at


72

30° C and stored at 4°C for 24-48h. The cells were then resuspended by shaking for 15 min and cultures were diluted into 100/xl volumes in 96-well plates using a Singer RoToR HDA and grown in Tecan GENios microplate readers for 30h. The growth rate of each culture was monitored by measuring the OD60o every 15 min. The doubling time (D) was calculated exactly as previously described [45]. The fitness of each deletion strain was calculated as the ratio of the doubling time of the parental wild-type strain divided by that of the mutant. We quantified the genetic interaction between each gene pair using a multiplicative model. If a strain deleted for gene x has a fitness of Wx and a strain deleted for gene y has a fitness of "Wy then the double mutant strain is expected to have a fitness (Wxy) of Wx x Wy. We measured the deviation exy from this expectation where eXy = Wxy — Wx • Wy. For each gene pair, the exy values for each Kan r -Nat r and Nat r -Kan r were averaged and used to generate the heatmap Fig. 4.6d.

4.5.8

Fluorescence Microscopy

For detection of GFP, cells were grown overnight in HIS-MET- media and images were acquired with a Zeiss mRM Axiocam on a Zeiss Axiovert 40CFL microscope (Carl Zeiss, Thornwood, NY).


4.6

73

Tables

Table 4.1: Multicopy suppressors identified by MSP. ND: Not tested. Table lists syntenic genes identified by MSP screen. Singletons are genes that do not have neighboring genes with high log2 ratios. ORFs in brackets \\ are not present on the microarray. Locus Glyoial GLOl;YML003W YDR366C;YDR367W;YPRl;XRS2 YMR317W;ADH6 PBN1;IRE1 Gryoxal + Isonicounamide PBN1;LRE1 YDR366C;YDR367W;YPR1;XRS2 YMR317W;ADH6 MRPL2S-.STP1 MED8;S0YJ;MSU;PGI1 YML007C-A;YAP1:GIS4: [IS(AGA)MJ; TRM12;GLOl;YML001W THRl;PPAl;RPNl ZDS2;YML108W;PML39 NSTI;RH02 YGR125W;YGM26W FBP26;VPS35 CCC2;rYDR27ICl;GL02;DONl YGL036WMG1 ljCP5;YER128W SSL1;SSKI M1G2.S1P2 GIS3;lOC2 BUD22;ERG5:SOK2 STBl;KRlt YKR023W;DBP7 TRR2;CDCJ2 NIP7;SRP72 ALG12;SSK2 RBAS0;[mR841:HlRl:OCR7:APA2 YGR016W.YGR017W BSPl;YPRI72W:VPS4 DDP1.YORI64C CBSI;/YDL06SW1;COX9 Singletons DAM1 MTH1 CDC34 BUD6 PBS2 RP11 SEC.6 YPT7 DNA2 \ GRE3

Logi Ratio

Avg Ratio

Confirmed Suppressor

2.52.3.25 3.29,2.07,2.20,2.37 2.86,2.04 2.13,2.35

2.88 2.48 2.45 2.24

GLOI YPRJ ADH6 ND

4.81,4.74 5.05,4.14,4.53,4.73 4.73,4.49 3.07,4.24 4.25,4.09,3.68,2.60 2.40,2.46,2.33,0.00,4.90,5.80,6.43

4.78 4.61 4.61 3.66 3.65 3.47

ND YPRI ADH6 STP1 MEDS GLOI

3.16,3.83,3.11 3.60,3.48,2 86 2.88,3.58 2.42,3.66 2.95,2.96 2.94,0.00,4.38,4.04 3.38,2.26 2.94,2.65 2.50,3.09 2.56,3.00 2.14,3.39 2.40,2.76,2.91 2.91,2.40 2.73.2.45 2.55,2.52 2.57,2.44 2.00,2.79 3.36,0.00,3,30,3.04,2.19 2.42,2.18 2.14,2.38,2.39 2.27,2.29 2.00,0.00.2.48

3.37 3.31 3.23 3.04 2.96 2.84 2.82 2.79 2.79 2.78 2.77 2.69 2.66 2.59 2.54 2.50 2.40 2.38 2.30 2.30 2.28 1.49

THR1 ZDS2 RH02 YGR126W ND GL02 ND ND SSK1 MG2 GIS3 ND ND ND ND ND ND HI.R1 ND ND ND ND

3.55 3.10 3.00 2.90 2.90 2.87 2.85 2.83 2.82 2.70

ND ND ND ND PBS2 ND ND ND ND GRE3


74

Table 4.2:Plasmids used in this study.

Plasmids

Characteristics

Source

P5476 P5476-ZDS2

High copy number LEU2 marker High copy number LEU2 marker, ZDS2-OSF plus regulatory regions High copy number LEU2 marker GL02-ORF plus regulatory regions High copy number LEU2 marker THR1-OKF plus regulatory regions High copy number LEU2 marker YPR1-ORF plus regulatory regions High copy number LEU2 marker SLGl-ORF plus regulatory regions High copy number LEU2 marker PDE2-OPJF plus regulatory regions High copy number LEU2 marker GRE3-OKF plus regulatory regions High copy number LEU2 marker PBS2-OKF plus regulatory regions High copy number LEU2 marker RH02-ORF plus regulatory regions High copy number LEU2 marker MRPL28-OKF plus regulatory regions High copy number LEU2 marker STP1-ORF plus regulatory regions High copy number LEU2 marker HLR1-ORF plus regulatory regions High copy number LEU2 marker M/Gi-ORF plus regulatory regions High copy number LETJ2 marker MG2-ORF plus regulatory regions High copy number LEU2 marker ADH6-ORF plus regulatory regions High copy number LEU2 marker YGRI26W-OPF plus regulatory regions High copy number LEU2 marker SSKI-OKF plus regulatory regions High copy number LEU2 marker G/S3-QRF plus regulatory regions High copy number LEU2 marker HOG1-ORF plus regulatory regions High copy number LEU2 marker SSK2-ORF plus regulatory regions High copy number LEU2 marker ERG5-ORF plus regulatory regions High copy number LEU2 marker G/52-ORF plus regulatory regions High copy number LEU2 marker MED8-ORF plus regulatory regions High copy number LEU2 marker PBN1-ORF plus regulatory regions High copy number URA3 marker GLOl-OKF plus regulatory regions Low copy number HIS3 marker

Charlie Boone Charlie Boone

P5476-GL02 P5476-THR1 PS476-YPR1 P5476-SLG1 P5476-PDE2 P5476-GRE3 P5476-PBS2 P5476-RH02 P5476-MRPL28 P5476-STPJ PS476-HLR1 P5476-MIG3 P5476-MIG2 P5476-ADH6 P5476-YGR126W P5476-SSK1 P5476-GIS3 P5476-HOG1 P5476-SSK2 P5476-ERG5 P5476-GIS2 P5476-MED8 P5476-PBN1 pRS426-GLO! p\JG23-FPSlGFP

Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone Charlie Boone This study This study


4.7

75

Figures

H Methylgiyoxal

° 0 Glyoxal

Jl Methylgiyoxal

glucose metabolic process peroxisomal transport intracellular signaling cascade pyruvate metabolic process pentose-phosphate shunt generation of precursor metabolites and energy mitochondrion organization and biogenesis amino acid activation tRNA aminoacylation lor protein translation

-log 10 (P--value)

mitochondrial genome maintenance

3

energy derivation by oxidation of organic compounds aerobic respiration response to drug response to oxidative stress vacuole organization and biogenesis response to endogenous stimulus response to DNA damage stimulus mRNA catabolic process mitotic recombination DNA repair RNA catabolic process ubiquitin-dependent protein catabolic process via the multivesicular body pathway vacuolar transport

Figure 4.1: Methylgloxal and glyoxal inhibit yeast growth, (a) Chemical structure of methylgiyoxal and glyoxal. (b) Fitness of wildtype (HO) strain grown in the presence of glyoxal (left) and methylgiyoxal (right) in the presence (black bars) and absence of aminoguanidine (white bars). Fitness was quantified using area under growth curve and normalized to growth in YPD. (c) Fitness of different deletion strains in methylgiyoxal and glyoxal. (d) Heatmap representing significant Gene Ontology terms identified to be significantly sensitive (FDR < 0.05) to methylgiyoxal and glyoxal. For comparison, we included strains identified to be sensitive to oxidants hydrogen peroxide (H2O2), cumene hydroperoxide (CHP), linoleic acid 13-hydroperoxide (LoaOOH) 3 menadione and diamide by a previous study[69]. The heatmap is colored according to the p-value of each Gene Ontology term.


76

a)

20 generations

HHHH

'Log^nWu,-

Figure 4.2: Identifying deletion strains resistant to glyoxal stress, (a) Screen schematic for identifying deletion strains resistant to glyoxal stress. A homozygous deletion pool was grown over 20 generations in 80mM glyoxal. Barcodes were amplified and hybridized to TAG4 arrays. Strong selection in the presence of glyoxal resulted in selection for strains highly resistant to glyoxal leaving only a few tags with a signal on the array, (b) (left) Log2 fold ratio of control/treatment of TAG array results. Negative fold ratios indicate tags that are over-represented in glyoxal selection. Genes mentioned in the text are highlighted, (right) Confirmation growth curves of deletion strains resistant to glyoxal. (c) Radar plot showing gene ontology fold enrichment of strains resistant (log2 fold ratio < -3.5) to glyoxal.


77

PCR-amplified library inserts hybridized to arrays

••HO iHASir2 ^ • H O + Nam HiAsir2 + Nam ^ B H O + IsoNam C_]Asir2 + IsoNam

Control y Glyoxal

Identify suppressors

Glyoxal Isonam LeggU

Glyoxal Vector

YPR1

ADH6

GRE3

PBS2

SSK1

SSK2

HOG1

Control

*U-

hzL~^,.~.~~~>

;lbrl.—™J

0fc^L_

™~.

Jlme(H)

/-\

Glyoxal 40mM

-y\ s • •

' 2

py Suppiressn Glyoxal

g

o

o aa

• •

•• «

••

*»• .

£L01

(jLCn

> THR1YDL133W

•

-4

* H fJHBMBCT'^* VkTswSr • ^^*^B^r : * • •" •\ . 4

..,

2

-""" -S

/ Deletion sensitivity Glyoxal

Deletion Sensitivity Glyoxal

Figure 4.3: (a) Fitness of wildtype and S1R2 mutants grown hi different glyoxal concentrations with and without supplementing nicotinamide and isomcotinamide. (b) Schematic of multicopy suppressor screen with glyoxal and isonicotinamide. A pool of yeast strains haboring a genomic library was grown competitively in the presence of control, glyoxal and glyoxal with isonicotinamide. Plasmkis were isolated and inserts were amplified by PCR and hybridized to ORF probes present on the TAG4 array. Over-represented ORFs in treatment versus control are identified as candidate suppressors, (c) Growth curves of singly cloned suppressors identified from MSP. (d) MSP versus DSP plot for glyoxal (left) and glyoxal with isonicotinaide (right). For each ORF, the average z-seore of fold change derived from MSP (representing suppression) and DSP (representing sensitivity) are plotted on the y and x-axis, respectively. ORFs found to be significant in either assay (fake discovery rate < 0.05) are colored red.


Vwtor

a o

Veetof

\tectw

VMtor

ZDS2

GL02

78

GL02

THAI

THRi

YPR1

-LJ °*j-j °*±1 asLL= aJj "JJ -JJ "JJ -JJ °*J Time (H)

mmmwmwmmmm mw'W'jj-'iimwmmm mmmmmmm Glyoxal 40mM Control

Figure 4.4: Confirmation growth curves of multicopy suppressors identified from MSP screen with glyoxal and/or isonicotinamide. Singly cloned ORFs under the control of native promoters on 2fjM plasmids were grown in Leu- synthetic media in the presence of absence or glyoxal.


Control

79

Glyoxal 40 mM

Glyoxal 80 mM

WT

0

10

20

Time

cdc20+l-

Figure 4.5: Growth curves of a heterozygous mutant of CDC20, a cell-cycle regulated activator of the anaphase-promoting complex or cyclosome (APC/C) in glyoxal.


80

y = 0.06698-tO.92087x . . R2=0.92 f\ 24 resistant strains^

. Kanr

^T

fitness (W) kanr-natr

s r

Aggravating Control

8

Time (hours) Glyoxal 5mM

Alleviating Control

T-~~—7,—;—

Glyoxal 10mM

£? 8 to

c

;/,

• WT • rpetA • pbs2A • pbs2&.,rpe1&

•WT • Apos2 # &erv14 • Sipbs2,Aerv14

Figure 4.6: (a) Double deletion strains were constructed between 15 glyoxal sensitive strains and 24 glyoxal resistant strains. Each double mutant was constructed independently twice using two different markers Kan' and Nat r . (b) Fitness correlation between reciprocal double-deletion mutants for each gene pair in the presence and absence of glyoxal. The correlation coefficient (R) and the best fitting line are shown, (c) Examples of double-deletion mutants displaying aggravating and alleviating genetic interaction in the presence of glyoxal. pbsSArpelA mutants were strongly aggravating in 5mM glyoxal with e = -0.66 and pbs2Aervl4A were strongly alleviating (suppression) in lOmM glyoxal with e = 0.44. (d) Distribution of t value for double mutants pairs grown in YPD (blue) or lOmM glyoxal (red).


fM^Ftex

81

r:-

IRA2 DDI3 TKL1 TMA108 YAP3 H0G1 PBS2 KIN3 NHX1 VPS5 SSD1 GPD2 RPE1 RPN4 PPZ1 SDS3 GL01 ECM1 FPS1 GIM4 NBP2 AAT2 PIN2 MET32 PFK1 RPN10 RTF1 SGF73 SIS2 MOT3 ERV14 CCW12 0ST3 SAC3 SPT8 SPT3 YNL295W SAP30 VAM10

-0.3 Aggravating

IRA2 SSD1 PPZ1 H0G1 PBS2 SAP30 ECM1 GL01 TMA108 VAM10 RPE1 TKL1 VPS5 AAT2 YAP3 GPD2 M0T3 YNL295W MET32 PIN2 RPN4 ERV14 FPS1 GIM4 SDS3 RPN10 SPT3 SPT8 NBP2 KIN3 NHX1 CCW12 PFK1 DDI3 RTF1 SGF73 OST3 SIS2 SAC3

|0.3 Alleviating

Figure 4.7: Genetic interaction profiles predict pathways. Genes hierarchically clustered (Pearson correlation) according to similar patterns of genetic interactions (e) for growth in YPD (left) and in lOmM glyoxal (right). Genes identified as sensitive to glyoxal when deleted are highlighted in red and genes identified as resistant to glyoxal when deleted are highlighted in green.


82

Glyoxal(10mM)

YPD 1.0

•wr

2 1 o o co

Q

• hoglA • e/v*4A ® hog1&erv14&

^gUifll

^

"—

0.5

O 10

30

20

Time (hours) 1.5

•

• WT • P&S2A

^

• ervUts, • pbs2Amvf4A

*

^^ffl£^®^ w s * M ** B B *™***^* l i r f ™

f 0.5

/ ' 10

20

30

Hoechst

Fps1-GFP

Merge

WT

erv14A

Figure 4.8: Ervl4 deletion abolishes requirement for the HOG pathway (a) Growth curve of double mutants (hoglAervl4A (top row) and (pbs2Aervl4A (bottom row) in YPD and glyoxal (10mM). (b) Wild-type (BY4741) and ervlÂ cells expressing C-terminally GFP-tagged Fpsl from the pUG23-FPSlGFP plasmid were grown in synthetic media lacking methionine. DNA staining was performed with Hoescht stain 33258.


83

Figure 4.9: Model for glyoxal resistance. We highlight four main pathways necessary for glyoxal resistance. The HOG pathway positively regulates the expression of GLOl and negatively regulates FPS1 in conferring glyoxal resistance. We hypothesize that the HOG pathway targets Fpsl for degradation thereby reducing glyoxal import. Moreover we hypothesize that transport of Fpsl to the cell surface is mediated by Ervl4 and associated proteins. We also identified three other mechanisms by which glyoxal resistance is managed: 1) downregulation of the Ras/cAMP/PKA pathway 2) NADPH generation 3) glyoxal metabolism.

CHAPTER 4. CARBONYL

4.8

STRESS

RESISTANCE

84

Bibliography

[1] J. Aguilera and J. A. Prieto. The saccharomyces cerevisiae aldose reductase is implied in the metabolism of methylglyoxal in response to stress conditions. Curr Genet, 39(5-6):273-283, 07:: 2001. [2] Jaime Aguilera, Sonia Rodriguez-Vargas, and Jose A. Prieto. The hog map kinase pathway is required for the induction of methylglyoxal-responsive genes and determines methylglyoxal resistance in saccharomyces cerevisiae. Mol Microbiol, 56(l):228-239, 04:: 2005. [3] A. A. Akhand, K. Hossain, H. Mitsui, M. Kato, T. Miyata, R. Inagi, J. Du, K. Takeda, Y. Kawamoto, H. Suzuki, K. Kurokawa, and I. Nakashima. Glyoxal and methylglyoxal trigger distinct signals for map family kinases and caspase activation in human endothelial cells. Free radical biology & medicine, 31(l):20-30, Jul 1 2001. [4] G. Barrera, S. Pizzimenti, and M. U. Dianzani. Lipid peroxidation: control of cell proliferation, cell differentiation and cell death. Molecular aspects of medicine, 29(l-2):l-8, Feb-Apr 2008. [5] K. B. Beckman and B. N. Ames. The free radical theory of aging matures.

Physiological

Reviews, 78(2):547-581, Apr 1998. [6] L. Benov and I. Fridovich. Superoxide dependence of the toxicity of short chain sugars. The Journal of biological chemistry, 273(40):25741-25744, Oct 2 1998. [7] C. Bertram and R. Hass. Cellular responses to reactive oxygen species-induced dna damage and aging. Biological chemistry, 389(3):211-220, Mar 2008. [8] A. Bito, M. Haider, I. Hadler, and M. Breitenbach. Identification and phenotypic analysis of two glyoxalase ii encoding genes from saccharomyces cerevisiae, glo2 and glo4, and intracellular localization of the corresponding proteins. J Biol Chem, 272(34):21509-21519, 08:22: 1997. [9] K. J. Bitterman, R. M. Anderson, H. Y. Cohen, M. Latorre-Esteves, and D. A. Sinclair. Inhibition of silencing and accelerated aging by nicotinamide, a putative negative regulator of yeast sir2 and human sirtl. The Journal of biological chemistry, 277(47):45099-45107, Nov 22 2002. [10] M. Bolte, P. Dieckhoff, C. Krause, G. H. Braus, and S. Irniger. Synergistic inhibition of apc:c by glucose and activated ras proteins can be mediated by each of the tpkl-3 proteins in saccharomyces cerevisiae. Microbiology (Reading, England), 149(Pt 5):1205-1216, May 2003. [11] N. E. Cameron, T. M. Gibson, M. R. Nangle, and M. A. Cotter. Inhibitors of advanced glycation end product formation and neurovascular dysfunction in experimental diabetes. Annals of the New York Academy of Sciences, 1043:784-792, Jun 2005.

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

85

[12] I. Dalle-Donne, D. Giustaxini, R. Colombo, R. Rossi, and A. Milzani. Protein carbonylation in human diseases. Trends in molecular medicine, 9(4):169-176, Apr 2003. [13] J. Du, J. Zeng, X. Ou, X. Ren, and S. Cai. Methylglyoxal downregulates raf-1 protein through a ubiquitination-mediated mechanism. The international journal of biochemistry & cell biology, 38(7)=1084-1091, 2006. [14] E. M. Ellis. Reactive carbonyls and oxidative stress: potential for therapeutic intervention. Pharmacology & therapeutics, 115(1):13—24, Jul 2007. [15] A. Espinosa-Mansilla,

I. Duran-Meras,

and

F.

Salinas.

High-performance

liquid

chromatographic-fluorometric determination of glyoxal, methylglyoxal, and diacetyl in urine by prederivatization to pteridinic rings. Analytical Biochemistry, 255(2):263-273, Jan 15 1998. [16] G. Ford and E. M. Ellis. Characterization of yprlp from saccharomyces cerevisiae as a 2methylbutyraldehyde reductase. Yeast (Chichester, England), 19(12):1087-1096, Sep 15 2002. [17] N. L. Garneau, J. Wilusz, and C. J. Wilusz. The highways and byways of mrna decay. Nature reviews.Molecular cell biology, 8(2):113-126, Feb 2007. [18] A. P. Gasch, P. T. Spellman, C. M. Kao, O. Carmel-Harel, M. B. Eisen, G. Storz, D. Botstein, and P. O. Brown. Genomic expression programs in the response of yeast cells to environmental changes. Molecular biology of the cell, ll(12):4241-4257, Dec 2000. [19] G. Giaever, A. M. Chu, L. Ni, C. Connelly, L. Riles, S. Veronneau, S. Dow, A. Lucau-Danila, K. Anderson, B. Andre, A. P. Arkin, A. Astromoff, M. El-Bakkoury, R. Bangham, R. Benito, S. Brachat, S. Campanaro, M. Curtiss, K. Davis, A. Deutschbauer, K. D. Entian, P. Flaherty, F. Foury, D. J. Garfinkel, M. Gerstein, D. Gotte, U. Guldener, J. H. Hegemann, S. Hempel, Z. Herman, D. F. Jaramillo, D. E. Kelly, S. L. Kelly, P. Kotter, D. LaBonte, D. C. Lamb, N. Lan, H. Liang, H. Liao, L. Liu, C. Luo, M. Lussier, R. Mao, P. Menard, S. L. Ooi, J. L. Revuelta, C. J. Roberts, M. Rose, P. Ross-Macdonald, B. Scherens, G. Schimmack, B. Shafer, D. D. Shoemaker, S. Sookhai-Mahadeo, R. K. Storms, J. N. Strathern, G. Valle, M. Voet, G. Volckaert, C. Y. Wang, T. R. Ward, J. Wilhelmy, E. A. Winzeler, Y. Yang, G. Yen, E. Youngman, K. Yu, H. Bussey, J. D. Boeke, M. Snyder, P. Philippsen, R. W. Davis, and M. Johnston. Functional profiling of the saccharomyces cerevisiae genome. Nature, 418(6896):387-391, Jul 25 2002. [20] G. Giaever, P. Flaherty, J. Kumm, M. Proctor, C. Nislow, D. F. Jaramillo, A. M. Chu, M. I. Jordan, A. P. Arkin, and R. W. Davis. Chemogenomic profiling: identifying the functional interactions of small molecules in yeast. Proceedings of the National Academy of Sciences of the United States of America, 101(3):793-798, Jan 20 2004.

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

86

[21] R. D. Gietz, R. H. Schiestl, A. R. Willems, and R. A. Woods. Studies on the transformation of intact yeast cells by the liac:ss-dna:peg procedure. Yeast (Chichester, England), ll(4):355-360, Apr 15 1995. [22] C. M. Grant. Role of the glutathione:glutaredoxin and thioredoxin systems in yeast growth and response to stress conditions. Molecular microbiology, 39(3):533-541, Feb 2001. [23] D. Harman. The aging process. Proceedings of the National Academy of Sciences of the United States of America, 78(11):7124-7128, Nov 1981. [24] R. Hasan, C. Leroy, A. D. Isnard, J. Labarre, E. Boy-Marcotte, and M. B. Toledano. The control of the yeast h2o2 response by the msn2:4 transcription factors. Molecular microbiology, 45(1):233-241, Jul 2002. [25] M. E. Hillenmeyer, E. Fung, J. Wildenhain, S. E. Pierce, S. Hoon, W. Lee, M. Proctor, R. P. St Onge, M. Tyers, D. Roller, R. B. Altman, R. W. Davis, C. Nislow, and G. Giaever. The chemical genomic portrait of yeast: uncovering a phenotype for all genes. Science (New York, N. Y.), 320(5874):362-365, Apr 18 2008. [26] S. Hoon, A. M. Smith, I. M. Wallace, S. Suresh, M. Miranda, E. Fung, M. Proctor, K. M. Shokat, C. Zhang, R. W. Davis, G. Giaever, R. P. StOnge, and C. Nislow. An integrated platform of genomic assays reveals small-molecule bioactivities. Nature chemical biology, 4(8):498-506, Aug 2008. [27] Y. Inoue and A. Kimura. Identification of the structural gene for glyoxalase i from saccharomyces cerevisiae. J Biol Chem, 271(42):25958-25965, 10:18: 1996. [28] Y. Inoue, Y. Tsujimoto, and A. Kimura. Expression of the glyoxalase i gene of saccharomyces cerevisiae is regulated by high osmolarity glycerol mitogen-activated protein kinase pathway in osmotic stress response. The Journal of biological chemistry, 273(5):2977-2983, Jan 30 1998. [29] D. L. Vander Jagt, B. Robinson, K. K. Taylor, and L. A. Hunsaker. Reduction of trioses by nadph-dependent aldo-keto reductases, aldose reductase, methylglyoxal, and diabetic complications. J Biol Chem, 267(7):4364-4369, 03:05: 1992. [30] H. Juhnke, B. Krems, P. Kotter, and K. D. Entian. Mutants that show increased sensitivity to hydrogen peroxide reveal an important role for the pentose phosphate pathway in protection of yeast against oxidative stress. Molecular & general genetics : MGG, 252(4):456-464, Sep 25 1996. [31] M. P. Kalapos. Methylglyoxal in living organisms: chemistry, biochemistry, toxicology and biological implications. Toxicology letters, 110(3):145-175, Nov 22 1999.

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

87

[32] J. H. Kim and S. Powers. Overexpression of rpil, a novel inhibitor of the yeast ras-cyclic amp pathway, down-regulates normal but not mutationally activated ras function. Molecular and cellular biology, ll(8):3894-3904, Aug 1991. [33] G. C. Kujoth, A. Hiona, T. D. Pugh, S. Someya, K. Panzer, S. E. Wohlgemuth, T. Hofer, A. Y. Seo, R. Sullivan, W. A. Jobling, J. D. Morrow, H. Van Remmen, J. M. Sedivy, T. Yamasoba, M. Tanokura, R. Weindruch, C. Leeuwenburgh, and T. A. Prolla. Mitochondrial dna mutations, oxidative stress, and apoptosis in mammalian aging. Science (New York, N. Y.), 309(5733) :481484, Jul 15 2005. [34] C. Larroy, M. R. Fernandez, E. Gonzalez, X. Pares, and J. A. Biosca. Characterization of the saccharomyces cerevisiae ymr318c (adh6) gene product as a broad specificity nadph-dependent alcohol dehydrogenase: relevance in aldehyde reduction.

The Biochemical journal, 361 (Pt

1):163-172, Jan 1 2002. [35] D. Lee, E. Ezhkova, B. Li, S. G. Pattenden, W. P. Tansey, and J. L. Workman. The proteasome regulatory particle alters the saga coactivator to enhance its interactions with transcriptional activators. Cell, 123(3):423-436, Nov 4 2005. [36] William Lee, Robert P. St Onge, Michael Proctor, Patrick Flaherty, Michael I. Jordan, Adam P. Arkin, Ronald W. Davis, Corey Nislow, and Guri Giaever. Genome-wide requirements for resistance to functionally distinct dna-damaging agents. PLoS Genet, 1(2), 08:19: 2005. [37] Kazuhiro Maeta, Shingo Izawa, and Yoshiharu Inoue. Methylglyoxal, a metabolite derived from glycolysis, functions as a signal initiator of the high osmolarity glycerol-mitogen-activated protein kinase cascade and calcineurimcrzl-mediated pathway in saccharomyces cerevisiae. J Biol Chem, 280(l):253-260, 01:07: 2005. [38] M. Mollapour and P. W. Piper. Hogl mitogen-activated protein kinase phosphorylation targets the yeast fpsl aquaglyceroporin for endocytosis, thereby rendering cells resistant to acetic acid. Molecular and cellular biology, 27(18):6446-6456, Sep 2007. [39] K. Murata, T. Saikusa, Y. Fukuda, K. Watanabe, Y. Inoue, M. Shimosaka, and A. Kimura. Metabolism of 2-oxoaldehydes in yeasts, possible role of glycolytic bypath as a detoxification system in 1-threonine catabolism by saccharomyces cerevisiae. European journal of biochemistry : FEBS, 157(2):297-301, Jun 2 1986. [40] H. Nakanishi, Y. Suda, and A. M. Neiman. Ervl4 family cargo receptors are necessary for er exit during sporulation in saccharomyces cerevisiae. Journal of cell science, 120(Pt 5):908-916, Mar 1 2007. [41] I. Nemet, L. Varga-Defterdarovic, and Z. Turk. Methylglyoxal in food and living organisms. Molecular nutrition & food research, 50(12):1105-1117, Dec 2006.

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

88

[42] P. J. O'Brien, A. G. Siraki, and N. Shangari. Aldehyde sources, metabolism, molecular toxicity mechanisms, and possible effects on human health. Critical reviews in toxicology, 35(7):609-662, Aug 2005. [43] K. R. Oldenburg, K. T. Vo, S. Michaelis, and C. Paddon. Recombination-mediated pcr-directed plasmid construction in vivo in yeast. Nucleic acids research, 25(2) :451-452, Jan 15 1997. [44] C. N. Oliver, B. W. Ahn, E. J. Moerman, S. Goldstein, and E. R. Stadtman. Age-related changes in oxidized proteins. The Journal of biological chemistry, 262(12):5488-5491, Apr 25 1987. [45] R. P. St Onge, R. Mani, J. Oh, M. Proctor, E. Fung, R. W. Davis, C. Nislow, F. P. Roth, and G. Giaever. Systematic pathway analysis using high-resolution fitness profiling of combinatorial gene deletions. Nature genetics, 39(2):199-206, Feb 2007. [46] S. M. O'Rourke, I. Herskowitz, and E. K. O'Shea. Yeast go the whole hog for the hyperosmotic response. Trends in genetics : TIG, 18(8):405-412, Aug 2002. [47] J. I. Park, E. J. Collinson, C. M. Grant, and I. W. Dawes. Rom2p, the rhol gtp:gdp exchange factor of saccharomyces cerevisiae, can mediate stress responses via the ras-camp pathway. The Journal of biological chemistry, 280(4):2529-2535, Jan 28 2005. [48] S. E. Pierce, E. L. Fung, D. F. Jaramillo, A. M. Chu, R. W. Davis, C. Nislow, and G. Giaever. A unique and universal molecular barcode array. Nat.Methods, 3(8):601-603, Aug 2006. [49] G. Poli, M. U. Dianzani, K. H. Cheeseman, T. F. Slater, J. Lang, and H. Esterbauer. Separation and characterization of the aldehydic products of lipid peroxidation stimulated by carbon tetrachloride or adp-iron in isolated rat hepatocytes and rat liver microsomal suspensions. The Biochemical journal, 227(2):629-638, Apr 15 1985. [50] J. Powers and C. Barlowe. Transport of axl2p depends on ervl4p, an er-vesicle protein related to the drosophila cornichon gene product. The Journal of cell biology, 142(5):1209-1222, Sep 7 1998. [51] J. Powers and C. Barlowe. Ervl4p directs a transmembrane secretory protein into copii-coated transport vesicles. Molecular biology of the cell, 13(3):880-891, Mar 2002. [52] N. Rabbani and P. J. Thornalley. The dicarbonyl proteome: proteins susceptible to dicarbonyl glycation at functional sites in health, aging, and disease. Annals of the New York Academy of Sciences, 1126:124-127, Apr 2008. [53] Mark David Rose, Fred Marshall Winston, and Philip Hieter. Methods in yeast genetics : a laboratory course manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y., 1990.

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

89

[54] Y. Saint-Jalm and P. Moree-Testa. Study of nitrogen-containing compounds in cigarette smoke by gas chromatography-mass spectrometry. Journal of chromatography, 198(2):188-192, Oct 3 1980. [55] G. M. Santangelo. Glucose signaling in saccharomyces cerevisiae. Microbiology and molecular biology reviews : MMBR, 70(l):253-282, Mar 2006. [56] A. A. Sauve and V. L. Schramm. Sir2: the biochemical mechanism of nad(+)-dependent protein deacetylation and adp-ribosyl enzyme intermediates. Current medicinal chemistry, 11(7):807826, Apr 2004. [57] M. Schuldiner, S. R. Collins, N. J. Thompson, V. Denic, A. Bhamidipati, T. Punna, J. Ihmels, B. Andrews, C. Boone, J. F. Greenblatt, J. S. Weissman, and N. J. Krogan. Exploration of the function and organization of the yeast early secretory pathway through an epistatic miniarray profile. Cell, 123(3):507-519, Nov 4 2005. [58] J. S. Searle, K. L. Schollaert, B. J. Wilkins, and Y. Sanchez. The dna damage checkpoint and pka pathways converge on ape substrates and cdc20 to regulate mitotic progression. Nature cell biology, 6(2):138-145, Feb 2004. [59] Fred Sherman, Gerald R. Fink, James B. Hicks, and Cold Spring Harbor Laboratory. Laboratory course manual for methods in yeast genetics. Cold Spring Harbor Laboratory, New York, N.Y., 1986. [60] R. Shringarpure and K. J. Davies. Protein turnover by the proteasome in aging and disease. Free radical biology & medicine, 32(11):1084-1089, Jun 1 2002. [61] E. R. Stadtman. Protein oxidation and aging. Science (New York, N.Y.), 257(5074):1220-1224, Aug 28 1992. [62] E. R. Stadtman. Protein oxidation and aging. Free radical research, 40(12):1250-1258, Dec 2006. [63] A. Stanhill, N. Schick, and D. Engelberg. The yeast ras:cyclic amp pathway induces invasive growth by suppressing the cellular stress response. Molecular and cellular biology, 19(11):75297538, Nov 1999. [64] Ian Stansfield and Michael J. R. Stark. Yeast gene analysis, volume ol. 36. Elsevier:Academic Press, Amsterdam ; Boston, 2007. [65] J. M. Thevelein and J. H. de Winde. Novel sensing mechanisms and targets for the camp-protein kinase a pathway in the yeast saccharomyces cerevisiae. Molecular microbiology, 33(5):904-918, Sep 1999.

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

90

[66] P. J. Thornalley. Pharmacology of methylglyoxal: formation, modification of proteins and nucleic acids, and enzymatic detoxification-a role in pathogenesis and antiproliferative chemotherapy. General pharmacology, 27(4):565-573, Jun 1996. [67] P. J. Thornalley. Glutathione-dependent detoxification of alpha-oxoaldehydes by the glyoxalase system: involvement in disease mechanisms and antiproliferative activity of glyoxalase i inhibitors. Chem Biol Interact, 111-112:137-151, 04:24: 1998. [68] P. J. Thornalley. Use of aminoguanidine (pimagedine) to prevent the formation of advanced glycation endproducts. Archives of Biochemistry and Biophysics, 419(l):31-40, Nov 1 2003. [69] Geoffrey W. Thorpe, Chii S. Fong, Nazif Alic, Vincent J. Higgins, and Ian W. Dawes. Cells have distinct mechanisms to maintain protection against different reactive oxygen species: oxidativestress-response genes. Proc Natl Acad Sci USA,

101(17):6564-6569, 04:27: 2004.

[70] M. Thorsen, Y. Di, C. Tangemo, M. Morillas, D. Ahmadpour, C. Van der Does, A. Wagner, E. Johansson, J. Boman, F. Posas, R. Wysocki, and M. J. Tamas. The mapk hoglp modulates fpslp-dependent arsenite uptake and tolerance in yeast. Molecular biology of the cell, 17(10):4400-4410, Oct 2006. [71] A. H. Tong and C. Boone. Synthetic genetic array analysis in saccharomyces cerevisiae. Methods in molecular biology (Clifton, N.J.), 313:171-192, 2006. [72] A. H. Tong, G. Lesage, G. D. Bader, H. Ding, H. Xu, X. Xin, J. Young, G. F. Berriz, R. L. Brost, M. Chang, Y. Chen, X. Cheng, G. Chua, H. Friesen, D. S. Goldberg, J. Haynes, C. Humphries, G. He, S. Hussein, L. Ke, N. Krogan, Z. Li, J. N. Levinson, H. Lu, P. Menard, C. Munyana, A. B. Parsons, O. Ryan, R. Tonikian, T. Roberts, A. M. Sdicu, J. Shapiro, B. Sheikh, B. Suter, S. L. Wong, L. V. Zhang, H. Zhu, C. G. Burd, S. Munro, C. Sander, J. Rine, J. Greenblatt, M. Peter, A. Bretscher, G. Bell, F. P. Roth, G. W. Brown, B. Andrews, H. Bussey, and C. Boone. Global mapping of the yeast genetic interaction network. Science, 303(5659):808-813, Feb 6 2004. [73] M. Walfridsson, J. Hallborn, M. Penttila, S. Keranen, and B. Hahn-Hagerdal.

Xylose-

metabolizing saccharomyces cerevisiae strains overexpressing the tkll and tall genes encoding the pentose phosphate pathway enzymes transketolase and transaldolase. Applied and Environmental Microbiology, 61(12):4184-4190, Dec 1995. [74] K. J. Wells-Knecht, D. V. Zyzak, J. E. Litchfield, S. R. Thorpe, and J. W. Baynes. Mechanism of autoxidative glycosylation: identification of glyoxal and arabinose as intermediates in the autoxidative modification of proteins by glucose. Biochemistry, 34(ll):3702-3709, Mar 21 1995. [75] R. B. Wilson, G. Renault, M. Jacquet, and K. Tatchell. The pde2 gene of saccharomyces cerevisiae is allelic to real and encodes a phosphodiesterase which protects the cell from extracellular camp. FEBS letters, 325(3):191-195, Jul 5 1993.

CHAPTER

4. CARBONYL

STRESS

RESISTANCE

91

[76] R. Wysocki, C. C. Chery, D. Wawrzycka, M. Van Hulle, R. Cornells, J. M. Thevelein, and M. J. Tamas. The glycerol channel fpslp mediates the uptake of arsenite and antimonite in saccharomyces cerevisiae. Molecular microbiology, 40(6):1391-1401, Jun 2001. [77] P. Ye, B. D. Peyser, X. Pan, J. D. Boeke, F. A. Spencer, and J. S. Bader. Gene function prediction from congruent synthetic lethal interactions in yeast. Molecular systems biology, 1:2005.0026, 2005. [78] E. Zervas, X. Montagne, and J. Lahaye. Emission of alcohols and carbonyl compounds from a spark ignition engine, influence of fuel and ainfuel equivalence ratio. Environmental science & technology, 36(11):2414-2421, Jun 1 2002.

Chapter 5

Identifying human proteins toxic to yeast 5.1

Abstract

5.2

Introduction

Assigning function to genes is and will be one of the preeminent challenges facing modern biological research [38]. The development of high-throughput sequencing and microarray technology has advanced our ability to sequence genomes and embark on novel efforts like environmental sequencing and characterization of human genetic variation. While sequence analysis has made great strides in identifying genes and regulatory sequences, its ability to assign gene function is still limited. Even with Saccharomyces cerevisiae, one of the most well-studied eukaryote to date, approximately 21% of the genes are still uncharacterized[33] making experimental approaches to assign gene function ever more pressing. Gain-of-function (GOF) genetic screens are a powerful way of generating phenotypes and have been used with great success in multiple biological systems. With an estimated 2/3 of all genes having no loss-of-function phenotypes in Saccharomyces cerevisiae or Caenorhabditis elegans [10, 20], GOF approaches are therefore complementary to genetic knockout or knockdown screens. This has direct relevance to the study of human diseases caused by gene hyperactivation or amplification. Typically, GOF screens are performed by introducing cDNA expression libraries into cells followed by a selection procedure to identify clones that confer the phenotype of interest. Another approach commonly used in Drosophila developmental studies is transposon-mediated insertion of regulatory sequences that mediate GAL4-dependent transcription [37]. A number of studies have performed GOF screens in yeast using genomic or cDNA libraries[l, 3, 7, 27, 42, 51, 41]. Recently, GOF

92

CHAPTER

5. IDENTIFYING

HUMAN PROTEINS

TOXIC TO

YEAST

93

studies have been combined with micro-arrays to perform high-throughput functional analyses with mouse embryonic stem cells [36]. Cell-based assays using model organisms such as yeast allows human proteins to be studied in a cellular context. A number of studies have designed assays that couple mammalian proteins to yeast signaling pathways [22]. Genetic complementation has been used successfully to isolate orthologous proteins from higher organisms. A classic example is the isolation of the human homolog of CDK1J CDC2 by selecting for clones that can complement a mutant of CDC2[24] in fission yeast. Moreover, this method can also be used to study heterologous proteins with no yeast homologs. For example, inhibitors of mammalian apoptosis inducer Bax was identified by screening for clones that can suppress Bax-mediated growth inhibition [48, 13]. Recently, a-synuclein, which is associated with neurodegenerative disease, was expressed in yeast and used to establish and dissect a-synuclein-mediated toxicity[31]. This approach was further extended by performing genome-wide screens to identify genes that enhance or suppress the toxicity of mutant huntingtin fragment and asynuclein[47]. Further characterization using genome-wide approaches in yeast identified ER-Golgi trafficking as a key process in mediating a-synuclein resistance [6]. A similar approach was also taken by screening yeast deletion strains to identify proteins necessary for the functional expression of a mammalian Kir channel at the cell surface[15]. A number of studies have extended this approach by screening for compounds that rescue the growth of yeast expressing toxic human proteins [9, 46, 34, 44]. Screening for chemical suppressor is advantageous because it identifies compounds that are cell-permeable and yet not cytotoxic. The expression of human proteins in yeast is also a rapid way of characterizing human genetic variation. Using an in vivo yeast-based assay,14 nonsynonymous alleles of methylenetetrahydrofolate reductase (MTHFR) were tested for complementation of MET3, which encodes yeast MTHFR[28\. The authors discovered four low-frequency alleles that affected enzyme function of which three could be rescued by supplementing with folate. In another study from the same lab, polymorphisms of human ABCB1 (MDRl)-encoded

human multidrug transporter P-glycoprotein (P-gp) was studied

by testing in a yeast-based assay for its ability to confer drug resistance[19]. The goals of this project were to 1) generate a pool of yeast expressing the human orfeome (HORF pool). This pool will be useful for performing multicopy suppression screens (Chapter 2) as well as performing human toxicity screens and 2) demonstrate that human toxic genes can be identified via a competitive growth assay.

5.3

Results

5.3.1

Generating a yeast pool expressing human yeast proteins

I used the Human Orfeome v3.1[23], containing 12,212 distinct ORFs and transferred these ORFs, via the Gateway® reaction, to a yeast expression vector pAG426-GAL[2] that places the ORFs

CHAPTER 5. IDENTIFYING HUMAN PROTEINS TOXIC TO YEAST

94

under the control of the galactose inducible GAL1 promoter (Fig. 5.1a). The Gateway® reaction was performed in pools (n=34) of ~ 374 clones per pool to save cost. I assessed the quality of the pools by size analysis of the PCR fragments (Fig. 5.1b). The size distribution of ORFs transferred correlated well with the size distribution of the pool. Pools for which transfer failed were readily detected and repeated. Next individual pools of plasmids were transformed into yeast (BY4743) and colonies for each individual transformation were pooled to yield 34 subpools. Finally, the subpools were pooled to yield the final HORF pool.

5.3.2

Human Toxicity Screen

The HORF pooled was inoculated into inducing and repressing media and grown for 20 generations (Fig. 5.2). Plasmids following 20 generations of growth were isolated. ORF inserts were amplified, labeled and hybridized to the Human Gene 1.0 ST Array. This screen was performed twice independently. Plasmids harboring toxic human proteins are depleted following growth in galactose and were identified by comparing intensity ratios of probes for ORFs between inducing and repressing conditions (Fig. 5.3a). Significance analysis of microarray (SAM)[45] identified 759 distinct ORFs (FDR < 0.1) that were significantly depleted following growth for twenty generations in galactose media.

5.3.3

Confirmation of Toxic human genes

To validate the results of the screen, individual ORFs were randomly selected for isogenic growth (Fig. 5.3b) and 11/11 tested were shown to inhibit yeast growth. Overexpression of CDK2 also resulted in a growth defect (Fig. 5.3c). CDC2 and CDK2 are two related proteins that have been shown to complement CDC28 in yeast [30]. It was previously shown that overproduction of human CDK2 blocks cell cycle progression through G2/M in fission yeast [32]. Similarly, I observed cells arresting with elongated buds following CDK2 induction (Fig. 5.3c) suggesting a similar effect in budding yeast.

5.3.4

Function Enrichment of toxic human genes

Ingenuity Pathway Analysis (IPA, Ingenuity® Systems, www.ingenuity.com) identified multiple disease associated biological functions enriched (Fig. 5.4a) among the 759 toxic human ORFs. The category with the most number of HORF toxic genes assigned was cancer (n=130). Further subclassification of the category shows that a number of genes are associated with tumorigenesis and neoplasia (Fig. 5.4b) Another biological function identified was cell death in yeast (P-value < 2.05 x l 0 ~ 0 4 - 4.79 x 10 - 0 2 ). Among the genes assigned to this category are genes involved in apoptosis (BAK1,

BAX,

CHAPTER

BCL2L1).

5. IDENTIFYING

HUMAN PROTEINS

TOXIC TO

YEAST

95

Apoptosis is a form of programmed cell death regulated by the conserved family of Bcl-

2-related proteins. This family includes proteins that promote (Bax, Bak) and antagonize (Bcl2, Bcl211) apoptosis. Overexpression of Bax has been shown to inhibit yeast growth caused by cell death and this could be suppressed by overexpression of Bcl-2 and Bcl211[16, 43,14, 50, 26]. Overexpression of BAK1 was also shown to induce cell death[43, 18]. Interestingly, BCL2L1 was found to be toxic even though it was previously shown to suppress Bakl and Bax toxicity[43].

5.3.5

Structural variants within the toxic ORF set

The HORFeome collection contains 1563 ORFs that are represented by at least two clones. These different clones represent polymorphic ORFs and splice isoforms. The ORFs were divided into three groups: 1) ORFs with only single nucleotide polymorphisms (SNPs) 2) ORFs containing larger insertions/deletions (INDELs) and splice isoforms and 3) ORFs containing SNPs, INDELs and splice isoforms. 195/759 ORFs from the toxic set were found to have more than one clone represented (Table 5.2). This is an underestimate because with a microarray-based approach, the presence of one nontoxic allele may mask the ability to detect the effect of all other toxic alleles. The Human Gene ST 1.0 array uses a subset of probes from the Human Exon ST 1.0 array that are tiled across the entire transcript. It should to a certain extent be able to detect different splice forms but insufficient coverage across the entire gene did not permit robust identification of splice variants in this study. Nevertheless, these 195 ORFs represent a set of genes for which structural variation can be studied in a context of the growth inhibition phenotype.

5.3.6

ANG nuclear targeting signal modulates toxicity

To explore this idea further, I chose to test the effect of a SNP on growth inhibition by Angiogenin, encoded by ANG, a small polypetide with the ability to promote angiogenesis in turmors[8]. It is an enzyme with sequence similarity to bovine pancreatic ribonuclease (RNase A) and differs in activity from RNase A in its activity and specificity for its substrates[40, 40, 25]. Nuclear translocation of ANG is necessary for ANG-induced angiogenesis and inhibition of this process by inhibiting endocytosis or mutating its nuclear localization sequence abolishes its angiogenic acitivity[29]. ANG has previously been shown to be cytotoxic when injected into Xenopus oocytes by degrading tRNA[39] and inhibiting protein synthesis[5, 4]. There were two clones of ANG present in the Human ORFeome collection and sequence comparison revealed that these ORFs differed by a single amino acid within the nucleus targeting signal (Fig. 5.5a). Surprisingly, the two alleles conferred different levels of toxicity when these clones were tested individually (Fig. 5.5b). This suggest that proper localization of ANG modulates its toxicity to yeast. GFP fusions of both alleles did not reveal marked differences in localization although this could be due to reduced expression of Ang cause by inhibition of protein synthesis (data not shown).

CHAPTER

5.4

5. IDENTIFYING

HUMAN PROTEINS

TOXIC TO

YEAST

96

Discussion

In this study, 759 ORF human ORF were shown to be toxic to yeast using a competitive growth assay. This is a conservative estimate because of the pooled microarray-based approach that was taken. Analysis by using isogenic cultures will most certainly identify additional ORFs. Multiple ORFs previously found to inhibit yeast growth were identified, validating the screen. These toxic ORFs were associated with multiple functional categories and a number of ORFs were found to be associated with diseases like cancer. We also showed that a liquid-based growth assay was also sensitive enough to distinguish subtle growth defects in different alleles as shown for the ANG protein. The generation of the HORF pool in this study will be useful for future studies looking to uncover novel human gene function. I see a few different approaches that this could take. The list of toxic human ORFs identified in this study represents a starting point from which further analysis can be performed. The mechanism of toxicity can be assessed in a number of different ways. First, localization of the toxic protein can be assessed by fluorescent protein fusions. Secondly, the effects of gene expression on the cell cycle can be assessed by flow cytometry. Thirdly, expression of the toxic protein can be introduced into the yeast deletion or overexpression strains to identify genetic modifiers of protein toxicity [47, 6]. Naturally, these findings will have to be validated in higher model organisms but yeast-based approaches such as these will greatly accelerate hypothesis generation and experimental design. Another avenue for future research will be to include a greater number of human genetic variants in this study as well as disease associated alleles. Transfer of these clones into barcoded strains[49] or onto barcoded plasmids will permit these different variants to be resolved unambiguously and the use of the cost-effective TAG arrays[35]. Finally, the identification of these toxic ORFs can be used to identify novel inhibitors. The design of screens for which chemicals can be used to suppress ORF toxicity[34, 44, 9] will greatly complement other drug screen approaches that rely on growth inhibition[17, 11].

5.5 5.5.1

Materials and Methods Pool Construction

Human ORFeome v3.1[23] comprising of 12,212 clones was obtained from Invitrogen. Clones were transferred in pools of 374 clones to the yeast expression vector pAG426GAL-ccdB[2] by Gateway® reaction (Gateway® LR Clonase II enzyme mix, Invitrogen cat. no. 11791-100) to yield 34 pools. Individual pools from each Gateway® reaction were transformed into E. coli (TransforMaxTM EC100TM cat. no. EC10010) by electroporation and plated onto LB media containing carbenicillin (50/xg/ml). Colonies were pooled and plasmids were isolated (QIAprep Spin Miniprep Kit cat. no. 27104). Each pool was then transformed into yeast (BY4743) by standard lithium acetate


97

procedure[12] and selected on URA- selection media. Finally, yeast colonies were pooled to create final pool containing all clones and stored at -80° C until use.

5.5.2

Pool Growth

Overexpression pools were thawed into 10ml synthetic URA- media containing 2% rafiinose to an ODeoo of 0.5 and incubated for 3 hours with shaking. The culture was then diluted into synthetic URA- media containing 2% dextrose or 2% galactose to an ODeoo of 0.0625 and 0.7 ml was pipetted into a well of a 48-well microplate. Cells were grown in a Tecan (Durham, NC) GENios microplate reader and every 5 generations, cells were automatically pipetted into 0.7 ml of fresh media using a Packard Multiprobe II four-probe liquid handling system (Perkin-Elmer Life Sciences, Norwalk, CT) controlled by custom LABVIEW software (National Instruments, Austin, TX). Over the course of 20 generations of growth, cells from three independent cultures were pooled and saved every 5 generations and frozen at -20° C for subsequent preparation of plasmid DNA. Each screen was performed independently twice.

5.5.3

Plasmid purification, clone amplification and D N A labeling

Plasmid was purified from each cell pellets using Zymoprep II kit (Zymo Research cat. no. D2004). Plasmid was eluted in 20/d of 0.1X TE buffer and further diluted 1:20 in water to be amplified by PCR using primers 5'-gcgaagcgatgatttttgat-3' and 5'-cttttcggttagagcggatg-3'. Amplified products were purified using QIAquick PCR Purification Kit (Qiagen cat. no. 28104) and eluted in 30/d of EB buffer. Purified PCR products were labeled with biotin using the BioPrime® DNA labeling system (cat. no. 18094-011). Labeled products were purified using QIAquick Nucleotide Removal Kit (cat. no. 28304) and eluted in 60/d of EB buffer.

5.5.4

Chip hybridization

Biotin-labeled products were hybridized to the Human Gene 1.0 ST Array. 120 /d of hybridization cocktail (75 [A 2X Hyb buffer, 2.5 /A b213 control oligo, 3 /A of 50X Denhardt's and 55 fA of biotinlabeled PCR product) was denatured in boiling water for 5 minutes and chilled on ice for 5 minutes and hybridized to Human Gene 1.0 ST Array at 45°C for 16 hours with rotation at 60 rpm. Following hybridization, chips were stained once without anti-SAPE antibody amplification (Staining Mix: 2X MES staining buffer 300 fA, BSA 20 mg/ml 60/d, Streptavidin-Phycoerythrin (SAPE) lmg/ml 6/Ltl, water 234 (A; Blocking Mix: 2X MES staining buffer 300 (A, BSA 20mg/ml 60 //l, water 240 fA) using the Affymetrix fluidics station 450. Chips were scanned with GeneChip® Scanner 3000 7G.


5.5.5

C h i p Analysis

As the cloned ORFs found in the Human Orfeome exclude non-coding sequences (e.g. UTRs), we needed to exclude probes that did not correspond to the ORF sequences in the colleciton. HumanGene St 1.0 Array probe sequences were mapped to coding sequences of the Human ORFeome v3.1 (http://horfdb.dfci.harvard.edu/ ) using Blat[21] with standard parameters. Customized probe group files (pgf) were created using this mapping by including only probes that had a perfect match. Expression probe data were extracted from CEL files using apt-probeset-summarize (Affymetrix Power Tools) with rma-sketch normalization.

Finally, log2 ratios were calculated in MATLAB

(Mathworks) between pools grown in dextrose and galactose media. These ratios were then analyzed with significance analysis of microarrays (SAM) [45] with 2 replicates for each condition and ORFs meeting a false discovery rate of < 0.1 were classified as toxic.

98


5.6

Tables Table 5.1: Biological Functions Over-represented in Toxic Gene set.

Category Neurological Disease Cell Death

P-valtie 1.51E-04-3.ME02 2.05E-044.79E02

Ccll-ToCell Signaling and Interaction

3.96E-04-I.66E02

Hematological System Development and Function

3.96E-04-4.77E02

Connective Tissue Disorders

4.06E-04-4.66E02 4.06E-04-4.67E02 4.06E-04-4.66E02 4.06E-04-4.66E02

Immunological Disease Inflammatory Disease Skeletal and Muscular Disorders

Cancer

4.52E*M.79E02

Cellular Development

9.62E^M-5E-02

Immune and Lymphatic System Development and Function

9.62E-04-1.77E02

Reproductive System Development and Function Cell Morphology

1.54E^)3-3.88E02 U3rW>3-3.95E02 1.55E-03-2.73E02

Cellular Assembly and Organization

Molecules HLA-DQB1, HIGD1A, ANG, BCL2L1, NF2. HTR2B. EPL15, CSN3, CPT1A. BAK1, TMED10, BAX, FGG, MS4A1. GABRA4, APOD, DO/73. GAME. H3F3B. HNRPDL. MOG. IL1S. PDE4D. MLH1, IL1S. XIAP. SLC25A6. CHRNA6, RAN. K1F1B, IFNG, GRIA2, THYI ILIA HSPD1, EME1, CDC45L, TRA%, BCUll, CHEK2. KLRDI. TNFRSF9, HTAJ1P2, ST3GAI.3. BAK1. RTN4. FANCC. MXDI, DD1T3, IL2KG, MNATt, SSRP1, HCK 11.18, TTCS, CD9, HLF, CD247, THYI, BNIP3 (includesEG-.664), TDGF1, ISARl. GCG, SATBt, GZMA, BNIP3L, HTR2B, NCR!, BAX. OGDH. MS4A1. BVES. BAG1, PRG2, SNA12.1RF2,1L21, PRPFI9. CDK2, MLH1, YWHAE. SIRTI.1L1S. ARL11. XIAP. PIMA. 1FNG. EPHA7, ICOS ILIA APCS, HSPD1, OSTF1. TRA%, ANG, BCL2U, GCG, GATA3. SATB1, GZMA, KLRDI. TNFRSF9. A1CDA. NCR1, AFP, FGG, PRG2. LTF. FANCC. C.PB2, ACSLS, IL2RG. IL2I, MNAT1. MOG, GC. HCK IL18. IL1S. HBPI. CD9, FPR1. IFNG. CD247, PBX3,

icos, mri, PRKCI

HSPD1, ILIA, APCS. 7JM®. BCI.2L1, CHF.K2, MED1. GATA3. KLRDI. TNFRSF9. AICDA, FGF7, BAK1, FANCC, MXDI, CPB2, IL2RG, NFYA TOX.MOG, GC, HCK IL18, CD9, HLF, CD247, THYI, SAIB1, GZMA, CXCL11. C1SS, TCF12.NCR1, BAX BCHE, MS4A1, PRG2, LTF, IRF2, SNA/2, IL21, POV2F1, PDE4D, SOCSS, MLH1,1LIS, XIAP, HOXA10, IFNG, FPR1, ICOS ILIA APCS BCL2L1. ALOX5AP, TNFRSF9, CDK2, TREX1, CTSS, 1L18,71/5, BAX, FGG. MS4A1, 17A1. PIMA. LTF, IFNG, ICOS APCS, MLPH, HU-DQB1. BCL2L1. CHEK2. SATB1. KLRDI, GZMA, TNFRSF9, BNIP3L, AICDA. TREX1. NCR1, BAK1. BAX, MS4A1, LTF, FANCC. DDIT3. IRF2.1L2RG, IL21. HCK IL18. CDK2. IL15. XIAP, CD9, IFNG, FPR1. ICOS, CD247, DCLRE1C ILIA APCS, HSPD1, TRA®, SAT1, BCL2L1. TNFRSF9, BNIP3U CTSS, TREX1, BAX, FGG. MS4A1, T1A1. PRG2. LTF, FANCC, CPB2, AOAH, IRF2, ALOX5AP, 11.18, HCK, CDK2, PDE4D, 11.15, PTMA, FPR1, IFNG, ICOS APCS. ILIA. HSPD1. BCULl, CHEK2, TNFRSF9. BNIP3L, HTR2B, 1REX1 (includes EG: 11277), CTSS, BAX, FOG, FOXM1, MS4A1. TIA1, BAG1, LTF. IRF2, IL2RG, EXT1. FNTB. ALOXSAP, MNAT1, CDK2, ILI8, SIRTI. 1L15, XIAP, PIMA. IFNG. ICOS. BNIP3 /includes EG-.664) ILIA CDC4SL, TRA%. AMY2A. BACH1. CGA PIP. BCL2L1. CHEK2, GATA3, NF2, RACGAT1. TNFRSF9. ERG. DNAL11, BAK1, SCGBIAI. RTN4. MXDI, ENC1, SI00A11 (includes EG:S282). SULT1C2, EFEMP1, MNAT1. HNRPDL. SSRPI, GC, IL18, GNL3. CDS, SDPR. BNIP3 (Includes EG:664), TDGF1, LPAR1, SAT1, AMY2B. FOXM1, UBE3A, LTF, APOD, IRF2. CD82. FNTB. 1121. SDHC. PRPF19, CDH6, PDE4D, SIRTI, MGP, HBPI, ARIA I, PTMA, HFIB, FPR1, POLL, FOXN2. HSPD1, APCS, EME1, SLC39A6, CDC2SC, HIGD1A, ATPSJ. FGF7, HTATIP2, DUS2L. GABRA4, CACNG3, FANCC, DD1T3. SERP1ND1. ACVRI, GABRE ESRRG. II.2RG, FRAME (Includes EG.-23S32), LYZ, PSMA2, TTCS, BEX2, NDVFC2, EMP1, HLF (Includes EG:313l), RAN, GPM6A, CD247. MIX DCIJtEIC, BED, MREUA. MLPH, IMBRD1, ANG, PTPRA. GCG, GZMA, BNIP3L, CPA4, SERP1NAS, CTSS. RUVBL1, NCR1, BAX, FGG, MS4AI, MGMT, BAG1, CASP10, LACRT, MCM3, ACSL5, SNAI2, EXTl, DLEU1, UBE2I, CDK2. MLH1, YWHAE, IL1S. XIAP. BVB3. TCEAL4. CHRNA6. HOXA10. IFNG. GRIA2. ICOS. PRKCI, RADS4B ILIA HSPD1, TRA%. BCULl, CGA, MED1, GATA3, TNFRSF9. RACGAP1, TNP1, FGF7, CRISP2. STRBP, FANCC, ACVRI, IL2RG, TOX. NFYA, MOG, SFXN1, CPEBl, IL18, CD9. ADAM18, CD247, DCLRE1C, P1WIL2, CAPZA3, SATB1, ODF2. SERPINA5. CXCLIl. RUVBLI, MAX, TCF12, BAX BCHE, SPANXA1, IRF2.11.21, VBE2I, CDK2, MLHl, SOCSS, SIRTI, 1L1S, XIAP, HOXAI0. UG3, IFNG, ICOS, PRKCI APCS, ILIA. HSPD1. 7JM@. BCL2L1, MEDI, SATB1. GATA3, KLRDI, GZMA, TNFRSF9. AICDA, CTSS. FGF7. NCR1, TCF12, AFP, BAX MS4A1,MGMT, PRG2(includesEG.S5S3). LTF.MXDI, FANCC, CPB2, IRF2,IL2RG, TOX 1L21,MOG, GC, IL18. HCK PDE4D, SOCSS, MIM1, HBPI, 1I.1S, XIAP. CD9, FPRI, IFNG. THYI, C.D247, ICOS CATSPER2, BCL2L1, CGA, CAPZA3, TNP1. ODF2, SERPINA5, FGF7, ERMPI, RUVBLI, DNAIJ1, AFP, BAKI, BAX, CRISP2, UBE3A, SPANXA1, STRBP. FANCC. ACVRI, CDK2. PDE4D, MLHl, SIRTI, CD9, HOXA10, UG3, ADAMI8, PIW1L2, HEXA ILIA TPM3. BCULl, GATA3, TNFRSFV, 11,21.IL18, BAKI, 1L15, BAX CD9, FOXM1, RTN4, IFNG, BNIP3 (includesEG-.664) BAKI, BAX BCULl. BN1P3 (Includes EG:664)

99


Gastrointestinal Disease Immune Response Skeletal and Muscular System Development and Function Hematological Disease Cellular Growth and Proliferation Connective Tissue Development and Function Tumor Morphology Amino Acid Metabolism Carbohydrate Metabolism Cell Cycle

Cellular Compromise Cellular Function and Maintenance Cellular Movement DNA Replication, Recombination, and Repair Hair and Skin Development and Function Molecular Transport Organismal Functions Small Molecule Biochemistry Tissue Morphology Renal and Urological Disease Genetic Disorder Drug Metabolism Cardiovascular Disease Cell Signaling Endocrine System Development and Function Endocrine System Disorders Gene Expression Hepatic System Disease Lipid Metabolism Metabolic Disease Nutritionat Disease Organ Development RN A Post-Transcripdonal Modification Respiratory Disease

Tissue Development Vitamin and Mineral Metabolism Post-Translational Modification Organismal Development Infectious Disease Organ Morphology Dermatological Diseases and Conditions Embryonic Development Hepatic System Development and Function

1.55E-03-4.79E1)2 1.58E-03-4.66E02 224E-03-4.08E02 2.84E-03-4.66E02 2.97E^)3-4.66E02 2.97E-03-4.08E02 3.05E-03-3.05E03 3.05E-03-2.73E02 3.05E-O3-3.18E02 3.05E-O3-4.89E02 3.05E-03-4.66E02 3.0SE-03-4.84E02 3.05E-03-1.7E02 3.05E-03-4.77E02 3.05E-O3-2.73E02 3.05E-03-4.19B02 3.05E-O3-3.9SE02 3.05E-03-4.66E02 3.38E-03-3.95E02 4.98E-03-1.99E02 5.9JE-03-I.99E02 6,29E-03-rJ,29E03 ti.48E-03-3.95E02 8.83E-03-4.19E02 8.83E-03-2.73B02 8.83E-03-4.65E02 8.83E-03-4.66E02 8.83E-03-3.95E02 8.83E-03-4.6oE02 8.83E-03-4.6JE02 8.83E-03-8.83E03 8.83E-03-3.75E02 8.83E-03-1.7E02 8.83E-03-4.66E02 8.83E-03-3.95E02 8.83E-03-4.19E02 9.9E-03-2.73E02 1.1E-02-1.IE-O2

100

ILIA. 11SPD1, CPB2, CI1EK2, BC12L1, MEDl, GCG, MLHI, YWHAE, BAK1, XIAP, BAX,IFNG, DDIT3 HSPDI, ILIA, APCS, TRA%, BCL2L1, GATA3. SATBI, GZMA, KLRD1, TNFRSF9, A1CDA, FGF7, NCR1, AFP, BAX, PRC2 (Includes EG:SSS3), LTF, CPB2, IRF2,1L2RG, TOXIL2I, MOG, GCILIS. HCK, MLHI, HBPl, ILIS, FPR1, IFNG, CD247, THYI, ICOS ILIA OSTFl. ANG, TPM3. BAX. FOXMI, NF2, MNATl, IFNG, HTR2B, CDK2, ILIS ILIA BCL2LI, CHEK2. GOG. SATBI, KLRDI, GIMA, TNFRSP9, BN1P3L CPTIA, BAKI. BAX, MS4AI. MXDI. FANCC, IRF2, IL2I, GC, ILIS, HCK, CDK2. PDE4D, SIRTI. ILIS, XW. CD9, FPR1. IFNG, ICOS. CD247 ILIA OSTFl, IRF2, CGA, IL2RG, GOG, TOX. NF2, MOG. TNFRSF9. IL21, BNIP3L, IITR2B, ILIS, TCFI2,11.15, DUS2L, IIOXAI0, IFNG. FANCC, PRKCI, ICOS, DD1T3 ILIA, OSTFl, SATl, NF2, IFNG, HTR2B, ILI8 ILIS, BCI.2LI, TNFRSF9. IL2I. IFNG, ILIS HGD, CPB2, SATl, GCG, SLC3SA2. IFNG, GRIA2 ILIA APCS, TRA% AOAII, EXTI, ILI7F, GCG, FGF7, ILIS, BPI, IFNG. DYRK2, CD247, DSE. HEXA MREIIA, ILIA. ORC2L, CDC2SC, CIIEK2. BCL2LI, PTPRA, MTCII2, BAKI, BAX FOXMI, MGMT, FANCC, MXDI, RBXI (includesEG:997S), DDIT3, ESRRG, MNATl, SSRPI, PIAS2, CDK2, MLHI. SIRTI, YWHAE.ILIS, GNU, BUBS, IIOXAI0, LIG3, PTMA. IFNG. CD247, RAD54B 1RF2, BCL2LI. MOG. TNFRSF9, HCK, ILIS, BAKI, ILIS. BAX. XIAP. MS4AI, IFNG. FPRI, GRIA2, ICOS HSPDI. TRA% 1RF2, CHEK2, SATl, PTPRA, NF2, RACGAPI, PCYOX1, VBE21, HCK, ILIS, TTCS, HBPl, BAX PTMA UBE3A, IFNG, LTF, MXDI, FANCC, ICOS CDS2, CW. NF2, TNFRSF9, IFNG MRE1IA, CDC4SL, ORC2L. CDC25C, CHEK2, BCL2LI, MEDl, GCG, NF2, GZMA C25S, BAKI, BAX, XRN2, FOXMI, KPNAI, TIAl. MGMT. MXDI, FANCC, MCM3, DDIT3, IRF2, SI00AII (includesEG-.62S2), PAX1P1, MNATl, PRPFI9, C.DK2. ILIS, POU2F1. ASF1A, M1MI, YWHAE, BUB3, PTMS, POLA2, LIG3, R4N, IFNG, POLL, DCLRE1C, BNIP3 (includes EGM4). RAD54B CDC2SC. CHEK2 ILIA. CGA, BCL2L1, SATl, BEST1. MEDl. GCG. KLRD1. SLC3SA2, RACGAPI, BAKI, BAY, LPAR4, FANCC, DDIT3. GABRE, CDS2, NVPS0, PCrOXI, SWI2AI. CD9, RAN, IFNG. GR1A2, CD247, MIX GCG, FGF7 HSPDI. ILIA APCS, CPB2, TRA%, SATl, CDS2, MEDl. GCG, SLC3SA2, KLRD1, MOG, FGF7, ILIS, BPI, HOD, ILIS, BAX. CD). IFNG, GRIA2. FANCC. CD247, DD1T3 ILIA, TRA®, SATl. BCL2L1, MEDl. GATA3, SATBI, TNFRSF9, HTR2B. FGF7, CTSS, TCF12, BAKI, BAX, MS4A1. MGMT. MXDI. FANCC. IRF2.1L2RG. TOX, MOG, 11.21, ILIS, CDK2, POU2F1, PDE4D, MU11, ILIS, XIAP, COP. IFNG, CD247. ICOS BAKI, B4X, BC1.2LI, CDS2, Cm, NF2, IFNG BAKI, AMY2A, BAX, BCL2L1, IL2RG, /fiVO. CD247, DCLREIC, AMY2B ILIA BAX CDS2, MEDl, IFNG, DD1T3 HSPDI, S1RT1, XIAP, BCL2L1, GUCYIA3, BAG1, IFNG, BNIP3L, HTR2B, BN1P3 (includesEG-,664), PDE4D HSPDI. ILIA, BAKI. CGA. BAX BCL21.1, GCG. KLRDI, LPAR4, IFNG, QR1A2 BCL2L1, CGA MEDl, IFNG ILIA, HLA-DQB1, BCL2L1, CGA SATl. GCG, GC, ILIS, CDK2, CPTIA, SIRTI. BAKI. XIAP, B.iX, FOXMI, IFNG, PRKCI, BN1P3 (Includes EG-.664). DDIT3 MAX, ERG, ILIA TCF12, BESTl, MEDl. BAG1, NFYA, IFNG, P1AS2. POU2F1 ILIA CPB2, AMY2A MEDl. IFNG, AMY2B ILIA, TRA% ILIS, CDS2, Cm, CD247. ILIS, FGF7 HLA-DQB1, CPTIA. SIRTI, AMY2A, GCG, IFNG, ILIS, AMY2B AMY2A, AMY2B BAKI, CGA, BAX, BCL2L1, FGF7, ERMPl ANG, EXOSCS, UBE21, SENP2 APCS, ANG, BCL2LI, CHEK2, HTR2B, CXCL11, CTSS, BPI. HTATIP2, BAKI. BAX DUS2L, MS4AI, FOXMI, GABRA4, SCGB1A1, PRG2 (Includes EG.SSS3), LTF, MXDI. DDTI3, GABRE, CPB2, IL2I, SSRPI. SLC12A1, HCK ILIS, PDE4D, YWHAE, MGP, CLEC4E. ILIS. XIAP. ARl.ll, CD9, PUFII (Includes EG:SI13I). FPRI, IfNG, POLL, PBX3. CD247 ILIA, ACVfU. ILIS, IRF2. IL21, MOG, IFNG. ILIS, CNBP BAKI, CGA. BAX BCL2LI, GCG, KLRDI, LPAR4, GRIA2 SIRTI. BAKI, BAX, IMMP2L, GZMA CASP10, PHB (Includes EG:524S) BAKI, BAX, IFTS2

I.24E-02-3.95E02 1.24E-02-1.65E02 1.7E-02-1.7E-02

ILIS, GUCY1A3, LTF, IFNG, FANCC, ICOS, PDE4D

1.7E-02-4.60E02 1.7E4B-3.95E02

AC.VR1, EXTI, CD9, IFNG, OLD, CNBP

TRA®, NRP2, CGA 1L2RG. BC1.2L1, GCG, SATBI, ALOXSAP, CDK2, MLHI, ILIS, MGMT, UBE3A. KIF1B, IFNG, MXDI ILIS, ILIS

ILIA GCG, IFNG


Table 5.2:Distribution of polymorphic and splice variants

SNPs HORFeomeV3.1

535

INDEL and SNPs, INDEL Splice isoforms and Splice isoforms 298 730

Toxic Set

62

29

104

Total 1563 195

101


5.7

102

Figures

Human Expression Pool in BY4743

Figure 5.1: Human Pool Construction, (a) Clones from the Human Orfeome v3.1 collection were transferred to the yeast expression vector pAG426GAL-ccdB by G a t e w a y ^ cloning. 34 LR Gateway^-* reactions were performed with pools of plasmid DNA from the Human Orfeome v3.1 collection. Approximately 374 ORFs were present in each pool. Purified DNA was transformed into BY4743 and selected on URA- selective media. Yeast clones were pooled to create one final human expression pool, (b) Example of electrophoresis gel result for PCR amplicons of inserts following Gateway® transfer. In this example, pools 2108 and 2110 were identified as failed reactions and repeated.


/"^tjA-jsT

oo.

n Em

-> F ^ nn

«

CD

URA-Galactose (Inducing Media) Plasmld Isolation

ooo.

I

103

ORF amplification and labeling

Hybridization to Human Gene 1.0 ST Array

Human Orfeome Pool

URA- Dextrose (Repressing Media)

Figure 5.2: Human Toxicity Screen outline, (a) The human orfeome pool was grown in URA- media containing galactose (inducing media) and dextrose (repressing media) for twenty generations. Following growth, plasmids were isolated and amplified by PCR using primer common to the vector backbone. Inserts were labeled by random primer extension and hybridized to a Human Gene 1.0 ST array. Two independent replicates were performed.


a)

b)

Dextrose

Galactose

Dextrose

Galactose

104

Dextrose

Galactose

Figure 5.3: Identifying toxic human genes, (a) Average fold difference in microarray signal (log2( dextrose/galactose treatment)) for 2 independent biological replicates is plotted on the y-axis for 11,000 genes. ORFs meeting a false discovery rate (FDR) cutoff of < 0.1 as determined by significant analysis of microarrays (SAM) are highlighted in red. ORFs that were picked for isogenic growth confirmation are highlighted in green, (b) Isogenic growth curves for 11 randomly selected ORFs in URA- selection media containing 2% dextrose or 2% galactose, (c) (left) Growth curve of CDK2 strains in dextrose and galactose, (right) Morphology of pAG426GAL-CDK2 containing strains in dextrose and galactose media. Cells with aberrant cell morphology are highlighted with red arrows.


Q)

Biological Function • 2.05E-04-4.79E-02

Celt Death

| 3.96E~04~468E~02

Cell-To-Cell Signaling and Interaction

13.98E-04-4.77E-02

Hematological System Development and Function

|4.06E-04-4.e7E-02

Immunological Disease

• 4.0SE-04-4.66E-02

Inflammatory Disease

• 4.52E-04-4.79E-02

Cancer 9.62E-04-5E-02

Cellular Development |

Immune and Lymphatic System Development and Function

P

H |

Reproductive System Development and Function

9.62E-04-4.77E-O2

1.54E~O3~3.88E-02

1.55&-03-3.95E-03 CeR Morphology Cellular Assembly and Organization | 1.5SE-03-2.73E-02 • j 1.58E-03-4.7&E-O8 Gastrointestinal Disease

Hematological Disease Cellular Growth and Proliferation Connective Tissue Development and Function

tumorigenesis 119

I 4 06E-04-4.66E-02

Skeletal and Muscular Disorders

Immune Response

G1 phase 21

|4.06E-04-4.WE~Q2

Connective Tissue Disorders

Skeletal and Muscular System Development and Function

b)

|l.51E-04-3,95E-02

NeurologicaJ Disease

105

^^^H1.58E-O3-4.68E-02 •

2.24E-03-4.08E-02

• M B ^ H

2.84E-03-4.66E-02

2.97E-03-4.66E-02

I2.97E-03-4.08E-02

neoplasia 104

apoptdsis 40 II death

w

Cancer 20

40

80

Figure 5.4: Features of the Human toxic gene set. (a) 759 toxic human genes were analyzed with Ingenuity Pathway Analysis and significantly enriched functions are shown. The number of genes located each category are plotted on the x-axis. P-values (right-tailed Fisher Exact test) are indicated for each category in a range for each sub-category, (b) Pie chart showing relative fraction of cancer-associated genes annotated into different functional categories.


a)

106

NLS ANG ANG L5 9P

MVMGLGVLLLVFVLGLGLTPPTLAQDNSRYTHFLTQHYDAKPQ6RDDRYCESIMRRRGLT MVMGLGV1LLVFVLGLGI.TPPTLAQDNSRYTHFI,TQHYDAKPQGRDDRYCESIMRRRGPT

********************************************************** * ANG ANG L59P

SPCKDINTFIHGNKRSIKAICENKNGNPHRENLRISKSSFQVTTCKLHGGSPWPPCQYRA SPCKDINTFIHGNKRSIKAICENKNGNPHRENLRISKSSFQVTTCKLHGGSPWPPCQYRA

************************************************************ ANG ANG L59P

TAGFRNWVACENGLPVHLDQSIFRRP TAGFRNVWACENGLPVHLDQSIFRRP

***************************

b)

Dextrose

Galactose ' Vector ANG ANGL59P

5

10

15

20

Time (Hours) Figure 5.5: Missense mutation in ANG modifies toxicity to yeast, (a) Multiple sequence alignment of ANG and ANG L59P. The L59P substitution located in the nuclear localization signal domain is highlighted, (b) Growth of ANG and ANG L59P containing strains in dextrose and galactose.

CHAPTER 5. IDENTIFYING

5.8

HUMAN PROTEINS

TOXIC TO

YEAST

107

Bibliography

[1] R. Akada, J. Yamamoto, and I. Yamashita. Screening and identification of yeast sequences that cause growth inhibition when over expressed. Molecular & general genetics : MGG, 254(3) :267274, Apr 16 1997. [2] S. Alberti, A. D. Gitler, and S. Lindquist. A suite of gateway cloning vectors for high-throughput genetic analysis in saccharomyces cerevisiae. Yeast (Chichester, England), 24(10):913-919, Oct 2007. [3] J. Boyer, G. Badis, C. Fairhead, E. Talla, F. Hantraye, E. Fabre, G. Fischer, C. Hennequin, R. Koszul, I. Lafontaine, O. Ozier-Kalogeropoulos, M. Ricchetti, G. F. Richard, A. Thierry, and B. Dujon. Large-scale exploration of growth inhibition caused by overexpression of genomic fragments in saccharomyces cerevisiae. Genome biology, 5(9):R72, 2004. [4] D. K. St Clair, S. M. Rybak, J. F. Riordan, and B. L. Vallee. Angiogenin abolishes cellfree protein synthesis by specific ribonucleolytic inactivation of ribosomes. Proceedings of the National Academy of Sciences of the United States of America, 84(23) :8330-8334, Dec 1987. [5] D. K. St Clair, S. M. Rybak, J. F. Riordan, and B. L. Vallee. Angiogenin abolishes cellfree protein synthesis by specific ribonucleolytic inactivation of 40s ribosomes.

Biochemistry,

27(19):7263-7268, Sep 20 1988. LR: 20061115; PUBM: Print; JID: 0370623; 0 (Angiogenesis Inducing Agents); 0 (Neoplasm Proteins); 0 (RNA, Ribosomal); 0 (RNA, Ribosomal, 18S); 0 (RNA, Ribosomal, 28S); 97950-81-7 (angiogenin); EC 3.1.27.5 (Ribonuclease, Pancreatic); ppublish. [6] A. A. Cooper, A. D. Gitler, A. Cashikar, C. M. Haynes, K. J. Hill, B. Bhullar, K. Liu, K. Xu, K. E. Strathearn, F. Liu, S. Cao, K. A. Caldwell, G. A. Caldwell, G. Marsischky, R. D. Kolodner, J. Labaer, J. C. Rochet, N. M. Bonini, and S. Lindquist. Alpha-synuclein blocks er-golgi traffic and rabl rescues neuron loss in parkinson's models. Science (New York, N. Y.), 313(5785) :324328, Jul 21 2006. [7] C. Espinet, M. A. de la Torre, M. Aldea, and E. Herrero. An efficient method to isolate yeast genes causing overexpression-mediated growth arrest. Yeast (Chichester, England), ll(l):25-32, Jan 1995. [8] J. W. Fett, D. J. Strydom, R. R. Lobb, E. M. Alderman, J. L. Bethune, J. F. Riordan, and B. L. Vallee. Isolation and characterization of angiogenin, an angiogenic protein from human carcinoma cells. Biochemistry, 24(20):5480-5486, Sep 24 1985. [9] J. Fleming, T. F. Outeiro, M. Slack, S. L. Lindquist, and C. E. Bulawa. Detection of compounds that rescue rabl-synuclein toxicity. Methods in enzymology, 439:339-351, 2008.

CHAPTER

5. IDENTIFYING

HUMAN PROTEINS

TOXIC TO

YEAST

108

[10] G. Giaever, A. M. Chu, L. Ni, C. Connelly, L. Riles, S. Veronneau, S. Dow, A. Lucau-Danila, K. Anderson, B. Andre, A. P. Arkin, A. Astromoff, M. El-Bakkoury, R. Bangham, R. Benito, S. Brachat, S. Campanaro, M. Curtiss, K. Davis, A. Deutschbauer, K. D. Entian, P. Flaherty, F. Foury, D. J. Garfinkel, M. Gerstein, D. Gotte, U. Guldener, J. H. Hegemann, S. Hempel, Z. Herman, D. F. Jaramillo, D. E. Kelly, S. L. Kelly, P. Kotter, D. LaBonte, D. C. Lamb, N. Lan, H. Liang, H. Liao, L. Liu, C. Luo, M. Lussier, R. Mao, P. Menard, S. L. Ooi, J. L. Revuelta, C. J. Roberts, M. Rose, P. Ross-Macdonald, B. Scherens, G. Schimmack, B. Shafer, D. D. Shoemaker, S. Sookhai-Mahadeo, R. K. Storms, J. N. Strathern, G. Valle, M. Voet, G. Volckaert, C. Y. Wang, T. R. Ward, J. Wilhelmy, E. A. Winzeler, Y. Yang, G. Yen, E. Youngman, K. Yu, H. Bussey, J. D. Boeke, M. Snyder, P. Philippsen, R. W. Davis, and M. Johnston. Functional profiling of the saccharomyces cerevisiae genome. Nature, 418(6896):387-391, Jul 25 2002. [11] G. Giaever, P. Flaherty, J. Kumm, M. Proctor, C. Nislow, D. F. Jaramillo, A. M. Chu, M. I. Jordan, A. P. Arkin, and R. W. Davis. Chemogenomic profiling: identifying the functional interactions of small molecules in yeast. Proceedings of the National Academy of Sciences of the United States of America, 101(3):793-798, Jan 20 2004. [12] R. D. Gietz, R. H. Schiestl, A. R. Willems, and R. A. Woods. Studies on the transformation of intact yeast cells by the liac:ss-dna:peg procedure. Yeast (Chichester, England), ll(4):355-360, Apr 15 1995. [13] W. Greenhalf, J. Lee, and B. Chaudhuri. A selection system for human apoptosis inhibitors using yeast. Yeast (Chichester, England), 15(13):1307-1321, Sep 30 1999. [14] W. Greenhalf, C. Stephan, and B. Chaudhuri. Role of mitochondria and c-terminal membrane anchor of bcl-2 in bax induced growth arrest and mortality in saccharomyces cerevisiae. FEBS letters, 380(1-2):169-175, Feb 12 1996. [15] F. A. Haass, M. Jonikas, P. Walter, J. S. Weissman, Y. N. Jan, L. Y. Jan, and M. Schuldiner. Identification of yeast proteins necessary for cell-surface function of a potassium channel. Proceedings of the National Academy of Sciences of the United States of America, 104(46):1807918084, Nov 13 2007. [16] M. Hanada, C. Aime-Sempe, T. Sato, and J. C. Reed. Structure-function analysis of bcl-2 protein, identification of conserved domains important for homodimerization with bcl-2 and heterodimerization with bax. The Journal of biological chemistry, 270(20):11962-11969, May 19 1995. [17] S. Hoon, A. M. Smith, I. M. Wallace, S. Suresh, M. Miranda, E. Fung, M. Proctor, K. M. Shokat, C. Zhang, R. W. Davis, G. Giaever, R. P. StOnge, and C. Nislow. An integrated platform of genomic assays reveals small-molecule bioactivities. Nature chemical biology, 4(8):498-506, Aug 2008.

CHAPTER

5. IDENTIFYING

HUMAN PROTEINS

TOXIC TO

YEAST

109

[18] B. Ink, M. Zornig, B. Baum, N. Hajibagheri, C. James, T. Chittenden, and G. Evan. Human bak induces cell death in schizosaccharomyces pombe with morphological changes similar to those with apoptosis in mammalian cells. Molecular and cellular biology, 17(5):2468-2474, May 1997. [19] H. Jeong, I. Herskowitz, D. L. Kroetz, and J. Rine. Function-altering snps in the human multidrug transporter gene abcbl identified using a saccharomyces-based assay. PLoS genetics, 3(3):e39, Mar 9 2007. [20] R. S. Kamath, A. G. Eraser, Y. Dong, G. Poulin, R. Durbin, M. Gotta, A. Kanapin, N. Le Bot, S. Moreno, M. Sohrmann, D. P. Welchman, P. Zipperlen, and J. Ahringer. Systematic functional analysis of the caenorhabditis elegans genome using rnai. Nature, 421(6920):231-237, Jan 16 2003. [21] W. J. Kent. Blat-the blast-like alignment tool. Genome research, 12(4):656-664, Apr 2002. [22] G. Ladds, A. Goddard, and J. Davey. Functional analysis of heterologous gpcr signalling pathways in yeast. Trends in biotechnology, 23(7):367-373, Jul 2005. [23] P. Lamesch, N. Li, S. Milstein, C. Fan, T. Hao, G. Szabo, Z. Hu, K. Venkatesan, G. Bethel, P. Martin, J. Rogers, S. Lawlor, S. McLaren, A. Dricot, H. Borick, M. E. Cusick, J. Vandenhaute, I. Dunham, D. E. Hill, and M. Vidal. horfeome v3.1: a resource of human open reading frames representing over 10,000 human genes. Genomics, 89(3):307-315, Mar 2007. [24] M. G. Lee and P. Nurse. Complementation used to clone a human homologue of the fission yeast cell cycle control gene cdc2. Nature, 327(6117):31-35, May 7-13 1987. [25] P. A. Leland, K. E. Staniszewski, C. Park, B. R. Kelemen, and R. T. Raines. The ribonucleolytic activity of angiogenin. Biochemistry, 41(4):1343-1350, Jan 29 2002. [26] M. Ligr, F. Madeo, E. Frohlich, W. Hilt, K. U. Frohlich, and D. H. Wolf. Mammalian bax triggers apoptotic changes in yeast. FEBS letters, 438(1-2):61-65, Oct 30 1998. [27] H. Liu, J. Krizek, and A. Bretscher. Construction of a gall-regulated yeast cdna expression library and its application to the identification of genes whose overexpression causes lethality in yeast. Genetics, 132(3):665-673, Nov 1992. [28] N. J. Marini, J. Gin, J. Ziegle, K. H. Keho, D. Ginzinger, D. A. Gilbert, and J. Rine. The prevalence of folate-remedial mthfr enzyme variants in humans. Proceedings of the National Academy of Sciences of the United States of America, 105(23):8055-8060, Jun 10 2008. [29] J. Moroianu and J. F. Riordan. Nuclear translocation of angiogenin in proliferating endothelial cells is essential to its angiogenic activity. Proceedings of the National Academy of Sciences of the United States of America, 91(5):1677-1681, Mar 1 1994.

CHAPTER

5. IDENTIFYING

HUMAN PROTEINS

TOXIC TO

YEAST

110

[30] J. Ninomiya-Tsuji, S. Nomoto, H. Yasuda, S. I. Reed, and K. Matsumoto. Cloning of a human cdna encoding a cdc2-related kinase by complementation of a budding yeast cdc28 mutation. Proceedings of the National Academy of Sciences of the United States of America, 88(20) :90069010, Oct 15 1991. [31] T. F. Outeiro and S. Lindquist. Yeast cells provide insight into alpha-synuclein biology and pathobiology. Science (New York, N.Y.), 302(5651):1772-1775, Dec 5 2003. [32] J. Paris, P. Leplatois, and P. Nurse. Study of the higher eukaryotic gene function cdk2 using fission yeast. Journal of cell science, 107 ( Pt 3)(Pt 3):615-623, Mar 1994. [33] L. Pena-Castillo and T. R. Hughes. Why are there still over 1000 uncharacterized yeast genes? Genetics, 176(1):7-14, May 2007. [34] E. Perkins, D. Sun, A. Nguyen, S. Tulac, M. Francesco, H. Tavana, H. Nguyen, S. Tugendreich, P. Barthmaier, J. Couto, E. Yeh, S. Thode, K. Jarnagin, A. Jain, D. Morgans, and T. Melese. Novel inhibitors of poly(adp-ribose) polymerase:parpl and parp2 identified using a cell-based screen in yeast. Cancer research, 61(10):4175-4183, May 15 2001. [35] S. E. Pierce, E. L. Fung, D. F. Jaramillo, A. M. Chu, R. W. Davis, C. Nislow, and G. Giaever. A unique and universal molecular barcode array. Nat.Methods, 3(8):601-603, Aug 2006. [36] M. Pritsker, N. R. Ford, H. T. Jenq, and I. R. Lemischka. Genomewide gain-of-function genetic screen identifies functionally active genes in mouse embryonic stem cells. Proceedings of the National Academy of Sciences of the United States of America, 103(18):6946-6951, May 2 2006. [37] P. Rorth, K. Szabo, A. Bailey, T. Laverty, J. Rehm, G. M. Rubin, K. Weigmann, M. Milan, V. Benes, W. Ansorge, and S. M. Cohen. Systematic gain-of-function genetics in drosophila. Development (Cambridge, England), 125(6):1049-1057, Mar 1998. [38] A. Saghatelian and B. F. Cravatt. Assignment of protein function in the postgenomic era. Nature chemical biology, 1(3):130-142, Aug 2005. [39] S. K. Saxena, S. M. Rybak, R. T. Davey Jr, R. J. Youle, and E. J. Ackerman. Angiogenin is a cytotoxic, trna-specific ribonuclease in the rnase a superfamily. The Journal of biological chemistry, 267(30):21982-21986, Oct 25 1992. [40] R. Shapiro, J. F. Riordan, and B. L. Vallee. Characteristic ribonucleolytic activity of human angiogenin. Biochemistry, 25(12):3527-3532, Jun 17 1986. [41] R. Sopko, D. Huang, N. Preston, G. Chua, B. Papp, K. Kafadar, M. Snyder, S. G. Oliver, M. Cyert, T. R. Hughes, C. Boone, and B. Andrews. Mapping pathways and phenotypes by systematic gene over expression. Molecular cell; Molecular cell, 21(3):319-330, Feb 3 2006.

CHAPTER

5. IDENTIFYING

HUMAN PROTEINS

TOXIC TO

YEAST

111

[42] L. F. Stevenson, B. K. Kennedy, and E. Harlow. A large-scale overexpression screen in saccharomyces cerevisiae identifies previously uncharacterized cell cycle genes. Proceedings of the National Academy of Sciences of the United States of America, 98(7):3946-3951, Mar 27 2001. [43] W. Tao, C. Kurschner, and J. I. Morgan. Modulation of cell death in yeast by the bcl-2 family of proteins. The Journal of biological chemistry, 272(24):15547-15552, Jun 13 1997. [44] S. Tugendreich, E. Perkins, J. Couto, P. Barthmaier, D. Sun, S. Tang, S. Tulac, A. Nguyen, E. Yeh, A. Mays, E. Wallace, T. Lila, D. Shivak, M. Prichard, L. Andrejka, R. Kim, and T. Melese. A streamlined process to phenotypically profile heterologous cdnas in parallel using yeast cell-based assays. Genome research, 11(11):1899-1912, Nov 2001. [45] V. G. Tusher, R. Tibshirani, and G. Chu. Significance analysis of microarrays applied to the ionizing radiation response. Proceedings of the National Academy of Sciences of the United States of America, 98(9):5116-5121, Apr 24 2001. [46] E. Vottero, A. Balgi, K. Woods, S. Tugendreich, T. Melese, R. J. Andersen, A. G. Mauk, and M. Roberge. Inhibitors of human indoleamine 2,3-dioxygenase identified with a target-based screen in yeast. Biotechnology journal, l(3):282-288, Mar 2006. [47] S. Willingham, T. F. Outeiro, M. J. DeVit, S. L. Lindquist, and P. J. Muchowski. Yeast genes that enhance the toxicity of a mutant huntingtin fragment or alpha-synuclein. Science (New York, N.Y.), 302(5651):1769-1772, Dec 5 2003. [48] Q. Xu and J. C. Reed. Bax inhibitor-1, a mammalian apoptosis suppressor identified by functional screening in yeast. Molecular cell, l(3):337-346, Feb 1998. [49] Z. Yan, M. Costanzo, L. E. Heisler, J. Paw, F. Kaper, B. J. Andrews, C. Boone, G. Giaever, and C. Nislow. Yeast barcoders: a chemogenomic application of a universal donor-strain collection carrying bar-code identifiers. Nature methods, 5(8):719-725, Aug 2008. [50] H. Zha, H. A. Fisk, M. P. Yaffe, N. Mahajan, B. Herman, and J. C. Reed. Structure-function comparisons of the proapoptotic protein bax in yeast and mammalian cells. Molecular and cellular biology, 16(ll):6494-6508, Nov 1996. [51] N. Zhang, T. Ismail, J. Wu, K. C. Woodwark, D. C. Gardner, R. M. Walmsley, and S. G. Oliver. Disruption of six novel orfs on the left arm of chromosome xii reveals one gene essential for vegetative growth of saccharomyces cerevisiae. Yeast (Chichester, England), 15(12):12871296, Sep 15 1999.

Chapter 6

Conclusion and perspectives 6.1

Conclusion

In summary, the major accomplishments of this dissertation are (1) the development of an integrated and cost-effective assay for chemical characterization and target identification and the first largescale demonstration of using chemogenomics profiles for identifying structure-activity relationships; (2) the development of a framework for the defining the mechanistic basis of drug interactions by integrating genetic interaction networks with drug interaction modeling; (3) a rigorous study of the genetic requirements for carbonyl stress resistance; and (4) a global survey of human proteins that are toxic to yeast. In this last chapter, I discusses the major findings and future directions that this work can be taken.

6.2

Perspectives

6.2.1

Small-molecule screening

Testing the effect of increasing and decreasing individual gene dosage on drug resistance was shown to improve drug target identification. Several new yeast knockout collections are being generated in diverse Saccharomyces species, including wild strains. Additionally, genome-wide overexpression collections, conditional essential collections (e.g. temperature sensitive and promoter shut-off alleles) as well as diverse epitope-tagged collections are now available[3, 4, 15, 16, 6, 19]. Extending the integrated screening platform described here to incorporate these new collections and technologies will certainly open up new experimental possibilities. As part of a large screening effort, the integrated DSP and MSP assays are currently being screened against a set of small-molecules from a larger chemical diversity library. To date, more than 1000 DSP assays and 300 MSP assays have been performed. Several refinements to the assays

112

CHAPTER 6. CONCLUSION AND PERSPECTIVES

113

are actively pursued. Singly cloned ORFs (generated by Charlie Boone) in both CEN and 2/xM plasmids are currently being incorporated into the MSP assay. We are also exploring the use of a sensitized yeast strain that is deleted for multiple drug efflux pumps in order to improve the sensitivity of the assays. One of the main questions with a yeast-based assay is whether the smallmolecule inhibiting the yeast protein will also inhibit the corresponding human homolog. One way to directly address this problem is to refine the DSP assay by replacing each essential yeast gene with its human homolog in a heterozygous deletion background. This will require that each human homolog for each yeast essential gene to be identified which could be performed via complementation screens with the human expression pool generated in Chapter 5. One of the most surprising results was that chemogenomics profiles were able to distinguish the effects of highly similar chemical structures. The ability to identify structure activity relationships and mechanism of action within the same experiment makes chemogenomics extremely powerful. Further development of chemogenomics in yeast will continue to advance our understanding of the interplay between small-molecules and biological systems. For example, a systematic effort to assess environmental or small molecule effects will, when combined with genetic lesions, augment our understanding of genetic interactions[8, 9]. Another challenging area of research that chemical genomics could be applied is towards the development of assays that identify inhibitors of proteinprotein and protein-DNA interactions [5, 1, 17].

6.2.2

Drug interactions

The use of chemogenomic profiles in Chapter 3 for characterizing drug interactions can be used to further classify drug mechanism of action and design better drug combinations. An interesting observation observed in Chapter 2 was that a number of small-molecules with diverse structures showed similar mechanism of action. Thus an interesting question is whether these different structures target the same protein in the same or different way or different proteins in the same pathway. The nature of the interaction (synergistic or additive) between these compounds when used in combination may shed light on this question. Ultimately, prediction of drug interactions will require better signaling maps as well as kinetic models and their parameters. As whole-genome efforts to map these pathways begin to bear fruit, the selection of drug combinations can be informed by integrating drug mechanism of action with gene interaction and signaling maps. Experimentally testing multiple drug combinations at multiple doses results is cost-prohibitive due to a combinatorial explosion of the permutations that is required. This challenge will be aided by better computational models of cell signaling [2].

6.2.3

Characterization of toxic compounds

The ability to comprehensively characterize the effects of a toxic compound was demonstrated using multiple genomewide assays. In this case, quantitative epistasis analysis in the presence of glyoxal

CHAPTER 6. CONCLUSION AND PERSPECTIVES

114

enabled multiple pathways to be distinguished as well as discover novel genetic relationships. Approaches like these can be used to further characterize toxic compounds. An interesting application of this work is towards characterizing the toxicity of chemical mixtures. Most chemicals that exist in the environment are found as mixtures. Just like synthetic lethal phenotypes that only manifest with multiple mutations, acute toxicity may only occur in the presence of chemical mixtures. High content assays such at those described in this work may have an advantage over traditional testing procedures.

6.2.4

Yeast a n d Beyond

Chemical genomic advances in budding yeast have begun to inspire development of similar approaches in other model organisms. For example, systematic construction of barcoded deletion collections in Schizosaccharomyces pombe and Candida albicans have recently been reported [11, 10]. Haploinsufficiency profiling in C. albicans was recently used to identify mechanism of action of novel antifungal agents [18, 10]. Similar screens in mammalian cells, where RNA interference is used to decrease gene dose, has recently been reduced to practice in a series of elegant studies[13, 14, 7, 12]. One simple way that chemical genomics can be extended to other organisms is by adapting the classical genetic method of cross-species complementation to a genome wide assay as was performed in this study with a C. albicans genomic library. Another approach is to identify proteins that inhibit growth of yeast. The toxic human proteins identified in Chapter 5 represent only a starting point. Each of these will ultimately have to be verified experimentally both in yeast and in human cell lines. One of the limitations using a pooled approach is the possibility of cross-hybridization of probes between homologous proteins leading to false negative errors. Thus a systematic effort that measures the effect of overexpressing each ORF individually will certainly improve this study. Further phenotypic characterization of the growth defects caused by these toxic ORFs will be valuable. Identifying genetic modifiers of toxicity will certainly provide greater insight into the pathological effects of these proteins. These functional assays should also be useful for prioritizing and evaluating the effects of SNPs and splicing variants on protein function. Finally, a practical application of discovery of these toxic human ORF will be for drug discovery. Yeast strains expressing these toxic human ORFs could be pooled and systematically screened for compounds that can rescue toxicity. In summary, the lessons learned from yeast-based technologies have been vital to the development of the approaches in other model systems. I expect that more sophisticated yeast screens, building on those developed in this dissertation, will continue to inspire better experimental design in other organisms.

CHAPTER

6.3

6. CONCLUSION AND

PERSPECTIVES

115

Bibliography

[1] M. R. Arkin and J. A. Wells. Small-molecule inhibitors of protein-protein interactions: progressing towards the dream. Nature reviews.Drug discovery, 3(4):301-317, Apr 2004. [2] J. B. Fitzgerald, B. Schoeberl, U. B. Nielsen, and P. K. Sorger. Systems biology and combination therapy in the quest for clinical efficacy. Nature chemical biology, 2(9):458-466, Sep 2006. [3] D. M. Gelperin, M. A. White, M. L. Wilkinson, Y. Kon, L. A. Kung, K. J. Wise, N. LopezHoyo, L. Jiang, S. Piccirillo, H. Yu, M. Gerstein, M. E. Dumont, E. M. Phizicky, M. Snyder, and E. J. Grayhack. Biochemical and genetic analysis of the yeast proteome with a movable orf collection. Genes & development, 19(23) :2816-2826, Dec 1 2005. [4] S. Ghaemmaghami, W. K. Huh, K. Bower, R. W. Howson, A. Belle, N. Dephoure, E. K. O'Shea, and J. S. Weissman. Global analysis of protein expression in yeast. Nature, 425(6959):737-741, Oct 16 2003. [5] J. Huang and S. L. Schreiber. A yeast genetic system for selecting small molecule inhibitors of protein-protein interactions in nanodroplets. Proceedings of the National Academy of Sciences of the United States of America, 94(25): 13396-13401, Dec 9 1997. [6] W. K. Huh, J. V. Falvo, L. C. Gerke, A. S. Carroll, R. W. Howson, J. S. Weissman, and E. K. O'Shea. Global analysis of protein localization in budding yeast. Nature, 425(6959):686-691, Oct 16 2003. [7] J. Moffat, D. A. Grueneberg, X. Yang, S. Y. Kim, A. M. Kloepfer, G. Hinkle, B. Piqani, T. M. Eisenhaure, B. Luo, J. K. Grenier, A. E. Carpenter, S. Y. Foo, S. A. Stewart, B. R. Stockwell, N. Hacohen, W. C. Hahn, E. S. Lander, D. M. Sabatini, and D. E. Root. A lentiviral rnai library for human and mouse genes applied to an arrayed viral high-content screen. Cell, 124(6):1283-1298, Mar 24 2006. [8] R. P. St Onge, R. Mani, J. Oh, M. Proctor, E. Fung, R. W. Davis, C. Nislow, F. P. Roth, and G. Giaever. Systematic pathway analysis using high-resolution fitness profiling of combinatorial gene deletions. Nature genetics, 39(2):199-206, Feb 2007. [9] A. B. Parsons, A. Lopez, I. E. Givoni, D. E. Williams, C. A. Gray, J. Porter, G. Chua, R. Sopko, R. L. Brost, C. H. Ho, J. Wang, T. Ketela, C. Brenner, J. A. Brill, G. E. Fernandez, T. C. Lorenz, G. S. Payne, S. Ishihara, Y. Ohya, B. Andrews, T. R. Hughes, B. J. Prey, T. R. Graham, R. J. Andersen, and C. Boone. Exploring the mode-of-action of bioactive compounds by chemical-genetic profiling in yeast. Cell, 126(3):611-625, Aug 11 2006. [10] R. Rodriguez-Suarez, D. Xu, K. Veillette, J. Davison, S. Sillaots, S. Kauffman, W. Hu, J. Bowman, N. Martel, S. Trosok, H. Wang, L. Zhang, L. Y. Huang, Y. Li, F. Rahkhoodaee, T. Ransom,

CHAPTER

6. CONCLUSION AND

PERSPECTIVES

116

D. Gauvin, C. Douglas, P. Youngman, J. Becker, B. Jiang, and T. Roemer.

Mechanism-of-

action determination of gmp synthase inhibitors and target validation in Candida albicans and aspergillus fumigatus. Chemistry & biology, 14(10):1163-1175, Oct 2007. [11] A. Roguev, M. Wiren, J. S. Weissman, and N. J. Krogan. High-throughput genetic interaction mapping in the fission yeast schizosaccharomyces pombe. Nature methods, 4(10):861-866, Oct 2007. [12] D. E. Root, N. Hacohen, W. C. Hahn, E. S. Lander, and D. M. Sabatini. Genome-scale lossof-function screening with a lentiviral rnai library. Nature methods, 3(9):715-719, Sep 2006. [13] M. R. Schlabach, J. Luo, N. L. Solimini, G. Hu, Q. Xu, M. Z. Li, Z. Zhao, A. Smogorzewska, M. E. Sowa, X. L. Ang, T. F. Westbrook, A. C. Liang, K. Chang, J. A. Hackett, J. W. Harper, G. J. Hannon, and S. J. Elledge. Cancer proliferation gene discovery through functional genomics. Science (New York, N.Y.), 319(5863):620-624, Feb 1 2008. [14] J. M. Silva, K. Marran, J. S. Parker, J. Silva, M. Golding, M. R. Schlabach, S. J. Elledge, G. J. Hannon, and K. Chang. Profiling essential genes in human mammary cells by multiplex rnai screening. Science (New York, N.Y.), 319(5863):617-620, Feb 1 2008. [15] R. Sopko, D. Huang, N. Preston, G. Chua, B. Papp, K. Kafadar, M. Snyder, S. G. Oliver, M. Cyert, T. R. Hughes, C. Boone, and B. Andrews. Mapping pathways and phenotypes by systematic gene overexpression. Molecular cell; Molecular cell, 21(3):319-330, Feb 3 2006. [16] B. Suter, M. J. Fetchko, R. Imhof, C. I. Graham, I. Stoffel-Studer, C. Zbinden, M. Raghavan, L. Lopez, L. Beneti, J. Hort, J. Fillingham, J. F. Greenblatt, G. Giaever, C. Nislow, and I. Stagljar. Examining protein protein interactions using endogenously tagged yeast arrays: The cross-and-capture system. Genome research, 17, Nov 7 2007. [17] A. Whitty and G. Kumaravel. Between a rock and a hard place?

Nature chemical biology,

2(3):112-118, Mar 2006. [18] D. Xu, B. Jiang, T. Ketela, S. Lemieux, K. Veillette, N. Martel, J. Davison, S. Sillaots, S. Trosok, C. Bachewich, H. Bussey, P. Youngman, and T. Roemer. Genome-wide fitness test and mechanism-of-action studies of inhibitory compounds in Candida albicans. PLoS pathogens, 3(6):e92, Jun 2007. [19] Z. Yan, M. Costanzo, L. E. Heisler, J. Paw, F. Kaper, B. J. Andrews, C. Boone, G. Giaever, and C. Nislow. Yeast barcoders: a chemogenomic application of a universal donor-strain collection carrying bar-code identifiers. Nature methods, 5(8):719-725, Aug 2008.

I o

I o

Suggest Documents

Distributed I/O-I/O Module/Digital I/O Module/Remote I/O-Industrial I/O

I/O System UNIX I/O System

i"'"'O

< oOO?~-T~---=-o;---:-:::.I--=--,----~-:-:O-;:!

V O C A C I O N M I S I O N

U n n u o v o i n i z i o - Veganzetta online

CMMD I/O : A Parallel Unix I/O - Semantic Scholar

F i l o s o f i e

Optima - Data I/O

File I/O

O~~() ~;~I OOGo18:rooo

I/O Brush - CiteSeerX

+ . vi o. '" ("r* I'

1 I - O FASCISMO:

I & O Manual

PICAXE I/O Routines

Proza polska i o

7"o: ,;i-' ha

SMART I/O Manual

I/O Compatibility Guide

I/O Devices

S i O Espresso

7"o: ,;i-' ha

(I -~) rJ.+ O(rJ.Z)