Cell, Vol. 116, 121–137, January 9, 2004, Copyright 2004 by Cell Press

Discovering Modes of Actionfor Therapeutic Compounds Usinga Genome-Wide Screen of Yeast Heterozygotes

analyzing the modes of action of clinically relevantcompounds.

Introduction

Pek Yee Lum,1,6,* Christopher D. Armour,1,6

Sergey B. Stepaniants,1 Guy Cavet,1

Maria K. Wolf,3 J. Scott Butler,3 Jerald C. Hinshaw,5

Philippe Garnier,5 Glenn D. Prestwich,5

Amy Leonardson,1 Philip Garrett-Engele,1

Christopher M. Rush,4 Martin Bard,4 The sequencing of the human genome has revealedGreg Schimmack,1 John W. Phillips,2 thousands of putative proteins that have the potentialChristopher J. Roberts,1 and Daniel D. Shoemaker1 to revolutionize modern medicine (Lander et al., 2001).1Rosetta Inpharmatics LLC, a wholly-owned A critical step in translating genomic information into

subsidiary of Merck & Co., Inc. new therapeutics will be to develop approaches that12040 115th Avenue N.E. can rapidly distinguish high-quality drug targets amongKirkland, Washington 98034 the thousands of potential candidates. Identifying the2 Merck & Co., Inc. cellular targets of clinically proven small molecules is a126 East Lincoln Avenue powerful approach to identify proteins that can be safelyR80Y-255 and effectively targeted in humans. Interestingly, manyRahway, New Jersey 07065 of the therapeutic compounds on the market today were3 Department of Microbiology and Immunology discovered fortuitously with no a priori knowledge ofJ.P. Wilmot Cancer Center the target or mechanism of action. For example, theUniversity of Rochester School of Medicine and beneficial effects of niacin on lipid levels in humans have

Dentistry been known since 1955, while identification of its cellular601 Elmwood Avenue target remained elusive until just recently (Knopp, 1999;Rochester, New York 14642 Tunaru et al., 2003). The importance of the identification4 Department of Biology of the target of a clinically proven molecule such asIndiana University-Purdue University Indianapolis niacin shows that these molecules provide a rich source723 West Michigan Street of chemical probes that can be used to mine the humanIndianapolis, Indiana 46202 genome for the next generation of high-quality drug5 Department of Medicinal Chemistry targets.The University of Utah A major challenge facing this type of approach is the419 Wakara Way, Suite 205 need for new tools that can rapidly identify the proteinSalt Lake City, Utah 84108 targets of small molecules with interesting biological

properties. The emergence of new technologies such asprotein arrays (MacBeath and Schreiber, 2000; Service,2000), reverse transfection (Phizicky et al., 2003; Ziaud-Summarydin and Sabatini, 2001), and DNA microarrays (Huels etal., 2002) has accelerated this type of discovery effort.Modern medicine faces the challenge of developingModel organisms such as Saccharomyces cerevisiaesafer and more effective therapies to treat human dis-have also proven to be powerful tools for mechanisticeases. Many drugs currently in use were discoveredstudies of clinically relevant compounds (Heitman etwithout knowledge of their underlying molecularal., 1991; Schreiber and Crabtree, 1992), which is mademechanisms. Understanding their biological targetspossible by the fact that many human disease-associ-and modes of action will be essential to design im-ated genes have highly conserved yeast counterpartsproved second-generation compounds. Here, we(Foury, 1997; Steinmetz et al., 2002).describe the use of a genome-wide pool of tagged

A study by Giaever et al. (1999) demonstrated thatheterozygotes to assess the cellular effects of 78 com-parallel analysis of yeast strains with heterozygous dele-pounds in Saccharomyces cerevisiae. Specifically, la-tions of drug target genes can be used to monitor com-nosterol synthase in the sterol biosynthetic pathwaypound activities in vivo. Specifically, they showed thatwas identified as a target of the antianginal drug molsi-reducing the gene copy number of drug targets in adomine, which may explain its cholesterol-loweringdiploid cell can result in sensitization to the drug ofeffects. Further, the rRNA processing exosome wasinterest. We have extended this approach to analyzeidentified as a potential target of the cell growth inhibi-the activities of 78 chemical entities, most of which aretor 5-fluorouracil. This genome-wide screen validatedmedically relevant. In addition, we increased the numberpreviously characterized targets or helped identify po-of mutant strains to represent over half of the yeasttentially new modes of action for over half of the com-genome and used high-density oligonucleotide arrayspounds tested, providing proof of this principle forwith a two-color labeling strategy to monitor growthrates. Finally, a strain-specific error model was used toidentify drug-dependent fitness defects that are statisti-*Correspondence: pek_lum@merck.com; shoemakd@stanford.cally significant. In this study, we correctly identified thealumni.org

6 These authors contributed equally to this work reported targets for many well-characterized com-

Cell122

Figure 1. Schematic Representation of the Fitness Profiling Experimental Strategy

(A) Growth of tagged heterozygous deletion pool grown in the presence of drug.(B) Isolation of genomic DNA from cultures before (G-0) and after (G-20) outgrowth.(C) Amplification and labeling of barcode tags by PCR.(D) Hybridization of labeled barcode tags to DNA microarrays. Signal intensities resulting from microarray hybridizations are proportional tothe relative tag abundances in the pool population. The inset shows an enlargement of a microarray following hybridization with examples ofan enriched strain (red spot), a depleted strain (green spot), and a strain that was unchanged (yellow spot).

pounds in addition to discovering many potentially novel cedures). Competitive growth of the mutant pool wascarried out for 20 generations in media containing se-drug targets.lected compounds (Figure 1A). Strain abundance wasmeasured before and after outgrowth by hybridizationResultsof differentially labeled (Cy3/Cy5) barcode tags to DNAmicroarrays (Figures 1B–1D). Deletion strains that wereStudy Design Rationale

The experimental strategy used to interrogate drug ac- sensitive to a given compound were outcompeted bythousands of unaffected strains in the pool. Compoundstivities in this study is outlined in Figure 1. A mixture of

isogenic yeast mutants was generated by pooling 3503 were administered at concentrations resulting in moder-ate growth inhibition to ensure that the activities of theheterozygous deletion strains engineered with strain-

specific molecular barcode tags (see Experimental Pro- putative drug target(s) were rate limiting for growth. Un-

Discovering Drug Activities in Yeast123

Figure 2. Identification of Drug-SpecificGrowth Defects

(A) Compilation of signal intensity data for theERG1 heterozygote barcode tag across all 78drug treatments. The Cy3 (green) or Cy5 (red)signal intensities were obtained from tag-specific oligonucleotide features representedin triplicate on the microarray. In most drugtreatments, the Cy3 and Cy5 intensities wereroughly equal indicating that the strain abun-dance remained unchanged over the courseof most drug treatments. In the terbinafinetreatment, however, the Cy5 intensity is muchlower than that of the Cy3 for the ERG1 het-erozygote tag. This indicates that the relativeabundance of the ERG1 heterozygote wasdepleted during outgrowth in the presence ofterbinafine, which is known to inhibit Erg1pactivity.(B) Statistical significance of terbinafine-spe-cific fitness changes in the pool. Hybridiza-tion data across all drug treatments wereused to determine the average performanceof each strain during outgrowth. A modifiedStudent’s t test was used to calculate confi-dence levels by comparing the fitness levelsfrom individual treatments to the average per-formance across all treatments. The resultingP values for each strain in the terbinafinetreatment are plotted on the x axis and thenumber of strains with a given P value areplotted on the y axis. Notice that the erg1/ERG1 strain is the single outlier in the terbi-nafine treatment.

der these conditions, strains that are heterozygous for ments as indicated by the similarity of hybridization in-tensities before and after each of the competitive growthdrug target genes have the greatest chance of displaying

hypersensitivity phenotypes (Giaever et al., 1999). experiments. However, the log10 intensities of the ampli-fied tags showed a significant reduction following treat-ment with terbinafine suggesting that the erg1/ERG1Identifying Drug-Specific Growth Defectsstrain had been depleted from the population. FigureA strain-specific error model was developed to distin-2B shows the results of log10 P values for all 3503 strainsguish between general and drug-dependent growth de-in the pool following terbinafine treatment. The ERG1fects. To accomplish this, a reference set was generatedheterozygote was the most significant outlier with a Pby calculating the mean performance of each strainvalue of less than 1 � 10�85.across a large number of competitive growth experi-

ments. Next, the scatter error for these measurementswas calculated to determine the reproducibility of each Large-Scale Analysis of 78 Compounds

This screen was initiated by testing commercially avail-strain’s performance. To identify strains with drug-spe-cific changes in growth rates, the fitness values from a able compounds for the ability to inhibit growth in our

yeast strain background (see Experimental Proceduresgiven drug treatment were compared to the correspond-ing values in the reference set using a modified Stu- for description). Seventy-eight compounds that induced

moderate levels of growth inhibition (0.70 � pooldent’s t test (see Experimental Procedures). The re-sulting P values reflect the probability that the observed fitness � 0.96; see Experimental Procedures for defini-

tion of pool fitness) were selected for analysis (Table 1).growth rate in any given drug treatment could occur bychance based on the strain’s behavior in the refer- The compounds analyzed in this study included many

clinically used therapeutic agents and agriculturalence set.Figure 2A shows an example of the hybridization data chemicals. Figure 3 shows an overview of the P value

data for all 3503 strains across the 78 compound treat-collected for one of the heterozygotes (ERG1) before(Cy3) and after (Cy5) each of the 78 different competitive ments (also in Supplemental Table S1 available at http://

www.cell.com/cgi/content/full/116/1/121/DC1). Of thegrowth experiments. The 20 bp tags for each of thestrains in the pool were printed in triplicate on the mi- 78 compounds analyzed, 18 resulted in no drug-specific

fitness changes (Group I), 56 resulted in a small numbercroarrays that were used for these experiments. In thisexample, the relative growth rate of the ERG1 heterozy- of significant outliers (Group II), and 4 resulted in wide-

spread fitness changes (Group III) compared to the refer-gote was not affected by the majority of the drug treat-

Cell124

Tab

le1.

AC

om

pen

diu

mo

f78

Co

mp

oun

ds

CA

SR

egul

ato

ryP

oo

lN

umb

era

Co

mp

oun

dN

ame

Clin

ical

Ind

icat

ionb

Sta

tusc

Do

seT

este

dF

itnes

sS

igni

fican

tO

utlie

rS

trai

nsR

epo

rted

Tar

get

(s)

2022

-85-

75-

Flu

oro

cyto

sine

antif

ung

alF

DA

-ap

pro

ved

200

ng/m

L0.

80R

RP

6(�

20.6

8),

NO

P4

(�20

.08)

,C

DC

21(H

ard

man

etal

.,20

01)

(5F

C)

YP

R14

3W(�

12.4

2)50

-91-

95-

Flu

oro

deo

xyur

idin

ean

tivir

al;

antin

eop

last

icF

DA

-ap

pro

ved

300

�M

0.81

RR

P6

(�16

.21)

,H

OF

1(�

13.3

4),

CD

C21

(Har

dm

anet

al.,

2001

)(F

UD

R)

RS

B1

(�10

.65)

51-2

1-8

5-F

luo

rour

acil

(5F

U)

antin

eop

last

icF

DA

-ap

pro

ved

2�

M0.

92R

RP

6(�

17.7

9),

RR

P41

(�15

.9),

CD

C21

(Har

dm

anet

al.,

2001

)M

AK

21(�

12.1

)50

-76-

0A

ctin

om

ycin

Dan

tineo

pla

stic

FD

A-a

pp

rove

d20

�g

/mL

0.72

VM

A21

(�13

8.66

),Y

OR

114W

(�11

5.57

),P

RP

28(�

95.2

3)50

-48-

6A

mitr

ipty

line

antid

epre

ssan

tF

DA

-ap

pro

ved

125

�g

/mL

0.77

GF

A1

(�68

.37)

,N

EO

1(�

31.1

),Y

KL1

11C

(�29

.7)

1397

-89-

3A

mp

hote

rici

nB

antif

ung

alF

DA

-ap

pro

ved

600

ng/m

L0.

89A

CC

1(�

10.6

5),

YS

H1

(�8.

94),

TH

S1

(�6.

77)

147-

94-4

Ara

-CM

Pan

tineo

pla

stic

not

liste

d50

0�

g/m

L0.

88S

HY

1(�

46.9

8),

YG

R12

2W(�

35.7

3),

AV

T2

(�32

.98)

3337

-71-

1A

sula

mhe

rbic

ide

not

liste

d25

0�

g/m

L0.

95F

OL2

(�26

.16)

,C

DC

21(�

12.1

8),

FO

L1,

FO

L3(N

ard

ese

etal

.,G

IC2

(�3.

69)

1996

;P

atel

etal

.,20

03)

2912

2-68

-7A

teno

lol

antih

yper

tens

ive,

antia

ngin

al,

FD

A-a

pp

rove

d4.

0m

g/m

L0.

90P

OP

3(�

6.16

),R

IT1

(�5.

66),

TR

K1

antia

rrhy

thm

ic(�

4.21

)30

516-

87-1

Azi

do

thym

idin

e(A

ZT

)an

tivir

alF

DA

-ap

pro

ved

2.0

mg

/mL

0.96

YD

R23

0W(�

3.64

),C

DC

26(�

3.45

),Y

HR

029C

(�3.

32)

not

liste

dB

eno

xina

telo

cala

nest

hetic

not

liste

d90

0�

g/m

L0.

83E

RG

7(�

14.4

6),

RO

X3

(�11

.02)

,C

IS3

(�5.

03)

58-0

8-2

Caf

fein

eC

NS

stim

ulan

t;re

spir

ato

ryF

DA

-ap

pro

ved

700

�g

/mL

0.71

KO

G1

(�68

.88)

,T

OR

1(�

46.9

7),

stim

ulan

tS

HD

7(�

24.4

)76

89-0

3-4

Cam

pto

thec

inan

tineo

pla

stic

not

liste

d10

0�

g/m

L0.

96T

RX

2(�

11.7

),Y

PL2

51W

(�6.

77),

TO

P1

(Po

mm

ier

etal

.,19

98)

YLR

003C

(�6.

06)

1060

5-21

-7C

arb

end

azim

fung

icid

eno

tlis

ted

100

�g

/mL

0.81

GIM

2(�

109.

06),

GIM

3(�

90.4

6),

TU

B/T

UB

2(G

eiss

ler

etal

.,C

LN2

(�75

.36)

1998

;N

eff

etal

.,19

83)

50-5

3-3

Chl

orp

rom

azin

ean

tiem

etic

;an

tipsy

cho

ticF

DA

-ap

pro

ved

37.5

�g

/mL

0.93

DO

P1

(�23

.27)

,N

EO

1(�

21.7

6),

KR

E25

(�15

.71)

1566

3-27

-1C

isp

latin

antin

eop

last

icF

DA

-ap

pro

ved

600

�M

0.88

VP

S65

(�6.

38),

PS

Y2

(�5.

23),

AT

P4

(�4.

32)

303-

49-1

Clo

mip

ram

ine

antid

epre

ssan

t;F

DA

-ap

pro

ved

50�

g/m

L0.

88N

EO

1(�

32.9

1),

YF

R04

4C(�

20.5

3),

antio

bse

ssio

nala

gen

tC

AN

1(�

14.6

9)23

593-

75-1

Clo

trim

azo

lean

tifun

gal

FD

A-a

pp

rove

d37

.5ng

/mL

0.85

PD

R16

(�46

.66)

,E

RG

11(�

46.3

3),

ER

G11

(Tru

anet

al.,

1994

)P

DR

5(�

21.4

9)66

-81-

9C

yclo

hexi

mid

efu

ngic

ide;

pla

ntg

row

thno

tlis

ted

10ng

/mL

0.90

RP

P0

(�9.

3),

CIS

3(�

4.3)

,N

IP10

0R

PL4

1A/R

PL4

1B(S

tock

lein

reg

ulat

or

(�4.

01)

and

Pie

per

sber

g,

1980

)57

966-

95-7

Cym

oxa

nil

agri

cultu

ralf

ung

icid

eno

tlis

ted

1.25

�g

/mL

0.90

LCB

1(�

16.6

2),

YB

R22

6C(�

9.39

),U

RA

1(�

6.53

)20

830-

81-3

Dau

noru

bic

inan

tineo

pla

stic

FD

A-a

pp

rove

d40

ng/m

L0.

96LS

M6

(�11

.48)

,P

LO2

(�8.

56),

BS

C2

(�8.

55)

50-4

7-5

Des

ipra

min

ean

tidep

ress

ant

FD

A-a

pp

rove

d20

0�

g/m

L0.

87N

EO

1(�

33.2

3),

GO

S1

(�17

.34)

,K

ES

1(�

8.58

)

(co

ntin

ued

)

Discovering Drug Activities in Yeast125

Tab

le1.

Co

ntin

ued

CA

SR

egul

ato

ryP

oo

lN

umb

era

Co

mp

oun

dN

ame

Clin

ical

Ind

icat

ionb

Sta

tusc

Do

seT

este

dF

itnes

sS

igni

fican

tO

utlie

rS

trai

nsR

epo

rted

Tar

get

(s)

4239

9-41

-7D

iltia

zem

antia

ngin

al;

antih

yper

tens

ive;

FD

A-a

pp

rove

d2.

0m

g/m

L0.

86Y

MR

187C

(�4.

39),

YLR

016C

antia

rrhy

thm

ic(�

3.97

),P

MT

3(�

3.19

)67

-68-

5D

imet

hyls

ulfo

xid

eno

tlis

ted

FD

A-a

pp

rove

d2.

5%0.

93C

DC

31(�

32.0

3),

YO

R25

6C(D

MS

O)

(�16

.55)

,Y

OR

356W

(�9.

53)

2321

4-92

-8D

oxo

rub

icin

antin

eop

last

icF

DA

-ap

pro

ved

3�

g/m

L0.

91G

IC2

(�16

.31)

,Y

DR

319C

(�13

.8),

MIG

2(�

13.7

8)56

4-25

-0D

oxy

cycl

ine

antib

acte

rial

FD

A-a

pp

rove

d40

0�

g/m

L0.

86E

RV

1(�

9.63

),S

DS

3(�

6.88

),S

RB

5(�

5.24

)58

6-60

-7D

yclo

nine

loca

lane

sthe

ticF

DA

-ap

pro

ved

32�

g/m

L0.

85E

RG

2(�

13.1

6),

RP

S24

A(�

3.01

),S

SA

2(�

2.55

)E

RG

2(H

ughe

set

al.,

2000

)no

tlis

ted

Efa

vire

nzan

tivir

alF

DA

-ap

pro

ved

300

�g

/mL

0.91

YB

L086

C(�

3.75

),Y

KR

065C

(�3.

67),

SC

W10

(�3.

28)

1264

-62-

6E

ryth

rom

ycin

antib

acte

rial

FD

A-a

pp

rove

d2.

5m

g/m

L0.

89Y

OR

1(�

21.4

3),

YG

L069

C(�

11.1

5),

Eth

ylsu

ccin

ate

SE

C17

(�9.

28)

64-1

7-5

Eth

ano

lno

tlis

ted

FD

A-a

pp

rove

d1.

25%

0.79

CM

K1

(�5.

7),

MLH

1(�

5.53

),D

PM

1(�

4.91

)67

306-

03-0

Fen

pro

pim

orp

hsy

stem

icfu

ngic

ide

not

liste

d0.

0000

1%0.

82E

RG

2(�

18.5

5),

ISW

1(�

4),

BZ

Z1

(�3.

42)

ER

G2

(Mar

cire

auet

al.,

1990

)51

04-4

9-4

Flu

rbip

rofe

nan

ti-in

flam

mat

ory

;an

alg

esic

FD

A-a

pp

rove

d75

�g

/mL

0.83

YC

R01

6W(�

7.89

),S

EC

10(�

6.73

),A

TP

4(�

6.24

)25

812-

30-

Gem

fibro

zil

antih

yper

lipo

pro

tein

emic

FD

A-a

pp

rove

d30

�g

/mL

0.75

SP

C19

(�12

.06)

,Y

NL1

79C

(�8.

31),

SE

C13

(�7.

05)

67-9

9-2

Glio

toxi

nno

tlis

ted

not

liste

d3.

0�

g/m

L0.

91A

CC

1(�

6.98

),Y

MR

306C

-A(�

3.78

),E

RG

28(�

3.54

)52

-86-

8H

alo

per

ido

lan

tidys

kine

tic;

antip

sych

otic

FD

A-a

pp

rove

d62

.5�

g/m

L0.

85S

ET

6(�

59.4

5),

ER

G3

(�33

.89)

,O

LE1

(�25

.58)

ER

G2

(Mo

ebiu

set

al.,

1996

)51

-45-

6H

ista

min

eno

tlis

ted

FD

A-a

pp

rove

d3

mg

/mL

0.80

YG

L042

C(�

21.3

1),

LOT

6(�

8.76

),Y

LR16

4W(�

8.69

)12

7-07

-1H

ydro

xyur

eaan

tineo

pla

stic

;tr

eatm

ent

of

FD

A-a

pp

rove

d20

mM

0.91

RN

R4

(�12

.2),

RN

R2

(�8.

29),

RN

R2,

RN

R4

(Ritt

ber

gan

dsi

ckle

cell

anem

iaR

MS

1(�

7.04

)W

rig

ht,

1989

)15

687-

27-1

Ibup

rofe

nan

ti-in

flam

mat

ory

;an

alg

esic

;F

DA

-ap

pro

ved

10�

g/m

L0.

92B

CP

1(�

4.26

),Y

IR02

4C(�

3.94

),an

tipyr

etic

UB

A2

(�3.

25)

50-4

9-7

Imip

ram

ine

antid

epre

ssan

tF

DA

-ap

pro

ved

312.

5�

g/m

L0.

74D

OP

1(�

67.3

7),

KR

E25

(�31

.89)

,Y

KL1

11C

(�30

.7)

53-8

6-1

Ind

om

etha

cin

anti-

infla

mm

ato

ry,

FD

A-a

pp

rove

d10

0�

g/m

L0.

85F

AS

1(�

117.

96),

LCB

2(�

28.2

6),

antip

yret

ic,

anal

ges

icY

PL2

72C

(�23

.41)

54-8

5-3

Iso

niaz

idtu

ber

culo

stat

ican

tibac

teri

alF

DA

-ap

pro

ved

275

�g

/mL

0.89

HS

P82

(�9.

09),

YM

R15

8W-A

(�9.

07),

RS

C2

(�8.

66)

8063

-07-

8K

anam

ycin

antib

acte

rial

FD

A-a

pp

rove

d6.

0m

g/m

L0.

95Z

IP1

(�4.

18),

YP

R15

2C(�

3.23

),Y

NL1

84C

(�2.

59)

7533

0-75

-5Lo

vast

atin

antih

yper

lipo

pro

tein

emic

FD

A-a

pp

rove

d40

�g

/mL

0.80

HM

G1

(�9.

54),

VP

S8

(�5.

29),

HM

G1/

HM

G2

(Rin

eet

al.,

YP

L197

C(�

2.24

)19

83)

58-2

7-5

Men

adio

nep

roth

rom

bo

gen

icvi

tam

inF

DA

-ap

pro

ved

40ng

/mL

0.80

GIM

1(�

14.5

9),

RP

N10

(�6.

91),

YD

R10

7C(�

6.08

)

(co

ntin

ued

)

Cell126

Tab

le1.

Co

ntin

ued

CA

SR

egul

ato

ryP

oo

lN

umb

era

Co

mp

oun

dN

ame

Clin

ical

Ind

icat

ionb

Sta

tusc

Do

seT

este

dF

itnes

sS

igni

fican

tO

utlie

rS

trai

nsR

epo

rted

Tar

get

(s)

59-0

5-2

Met

hotr

exat

ean

tineo

pla

stic

;an

tirhe

umat

icF

DA

-ap

pro

ved

12.5

�g

/mL

0.90

CIN

1(�

9.7)

,Y

MR

247C

(�6.

16),

DF

R1

(Hua

nget

al.,

1992

)S

PP

1(�

5.91

)66

-27-

3M

ethy

lex

per

imen

talm

utag

en,

not

liste

d0.

002%

0.84

HE

M2

(�39

.97)

,H

EM

1(�

27.1

2),

Met

hane

sulfo

nate

tera

tog

en,

carc

ino

gen

KR

E21

(�14

.44)

(MM

S)

364-

62-5

Met

ocl

op

ram

ide

antie

met

icF

DA

-ap

pro

ved

7.0

mg

/mL

0.88

GF

A1

(�74

.5),

HIP

1(�

22.4

4),

YG

R19

0C(�

20.0

3)25

717-

80-0

Mo

lsid

om

ine

antia

ngin

alno

tlis

ted

2.0

mg

/mL

0.84

ER

G7

(�16

.08)

,Q

CR

10(�

5.15

),Y

PL0

71C

(�4.

65)

389-

08-2

Nal

idix

icac

idan

tibac

teri

alF

DA

-ap

pro

ved

325

�g

/mL

0.88

YLR

312C

(�15

.83)

,R

IT1

(�13

.07)

,N

AM

8(�

11.5

1)22

204-

53-1

Nap

roxe

nan

ti-in

flam

mat

ory

;an

alg

esic

;F

DA

-ap

pro

ved

40�

g/m

L0.

93Y

PR

148C

(�4.

31),

YLR

402W

antip

yret

ic(�

4.25

),R

IB5

(�4.

1)12

9618

-40-

2N

evir

apin

ean

tivir

alF

DA

-ap

pro

ved

250

�g

/mL

0.96

YP

L103

C(�

5.58

),R

IM21

(�5.

48),

YD

R31

4C(�

5.28

)55

985-

32-5

Nic

ard

ipin

ean

tiang

inal

;an

tihyp

erte

nsiv

eF

DA

-ap

pro

ved

1.0

mg

/mL

0.93

GF

A1

(�67

.99)

,Y

PL2

38C

(�4.

24),

HIP

1(�

3.9)

2182

9-25

-4N

ifed

ipin

ean

tiang

inal

;F

DA

-ap

pro

ved

200

�g

/mL

0.83

PM

C1

(�4.

82),

AA

D6

(�3.

7),

PP

E1

antih

yper

tens

ive.

(�3.

54)

67-2

0-9

Nitr

ofu

rant

oin

antib

acte

rial

FD

A-a

pp

rove

d50

0�

g/m

L0.

78Y

LR02

2C(�

3.31

),F

RT

1(�

3.17

),Y

LR24

1W(�

2.96

)73

590-

58-6

Om

epra

zole

antiu

lcer

ativ

eF

DA

-ap

pro

ved

50�

g/m

L0.

82Y

NL1

84C

(�4.

68),

YP

L206

CP

MA

1(M

onk

etal

.,19

95)

(�4.

55),

GC

R1

(�4.

51)

7542

-37-

2P

aro

mo

myc

inan

tibac

teri

al;

antia

meb

icF

DA

-ap

pro

ved

5.0

mg

/mL

0.95

SS

S1

(�4.

85),

RP

S13

(�1.

92),

MR

P21

(�1.

79)

100-

33-4

Pen

tam

idin

ean

tipro

tozo

al;

FD

A-a

pp

rove

d25

0�

g/m

L0.

90C

DH

1(�

43.0

4),

SD

H3

(�38

.56)

,an

tipne

umo

cyst

icM

RP

13(�

35.7

8)50

-33-

9P

heny

lbut

azo

nean

ti-in

flam

mat

ory

FD

A-

200

�g

/mL

0.89

YO

R35

6W(�

4.74

),H

IR1

(�3.

08),

dis

cont

inue

dY

DR

051C

(�2.

85)

140-

65-8

Pra

mo

xine

loca

lane

sthe

ticF

DA

-ap

pro

ved

16�

g/m

L0.

84E

RG

2(�

7.22

),S

TB

1(�

3.9)

,B

ZZ

1(�

3.63

)59

-46-

1P

roca

ine

loca

lane

sthe

ticF

DA

-ap

pro

ved

2.5

mg

/mL

0.90

GF

A1

(�44

.61)

,H

IP1

(�29

.93)

,Y

GR

190C

(�25

.53)

60-8

7-7

Pro

met

hazi

nean

tihis

tam

inic

FD

A-a

pp

rove

d10

0�

g/m

L0.

92N

EO

1(�

15.9

2),

YO

R26

3C(�

6.78

),F

TH

1(�

6.61

)77

84-4

6-5

So

diu

mA

rsen

iteto

pic

alac

aric

ide

not

liste

d20

�g

/mL

0.89

TR

K1

(�11

.03)

,R

IT1

(�7.

45),

GC

D2

(�6.

4)76

47-1

4-5

So

diu

mC

hlo

rid

eno

tlis

ted

not

liste

d25

0m

M0.

89Y

BL0

86C

(�13

.74)

,Y

DR

049W

(�9.

98),

GP

I1(�

8.5)

1375

5-38

-9S

od

ium

Nitr

op

russ

ide

antih

yper

tens

ive

FD

A-a

pp

rove

d66

7�

g/m

L0.

89H

EM

1(�

36.2

7),

NN

T1

(�11

.16)

,Y

LR28

0C(�

9.62

)62

996-

74-1

Sta

uro

spo

rine

antib

acte

rial

not

liste

d2.

0m

g/m

L0.

95R

HO

1(�

58.6

2),

CD

C12

(�41

.54)

,P

KC

1(Y

osh

ida

etal

.,19

92)

RO

T1

(�22

.28)

(co

ntin

ued

)

Discovering Drug Activities in Yeast127

Tab

le1.

Co

ntin

ued

CA

SR

egul

ato

ryP

oo

lN

umb

era

Co

mp

oun

dN

ame

Clin

ical

Ind

icat

ionb

Sta

tusc

Do

seT

este

dF

itnes

sS

igni

fican

tO

utlie

rS

trai

nsR

epo

rted

Tar

get

(s)

723-

46-6

Sul

fam

etho

xazo

lean

tibac

teri

al;

FD

A-a

pp

rove

d20

0�

g/m

L0.

85F

OL2

(�49

.11)

,M

ET

12(�

20.8

5),

FO

L1,

FO

L3(N

ard

ese

etal

.,an

tipne

umo

cyst

icC

DC

21(�

18.9

7)19

96;

Pat

elet

al.,

2003

)63

-74-

1S

ulfa

nila

mid

ean

tibac

teri

alF

DA

-ap

pro

ved

625

�g

/mL

0.86

FO

L2(�

37.4

2),

CD

C21

(�19

.53)

,F

OL1

,F

OL3

(Nar

des

eet

al.,

YB

L073

W(�

16.6

5)19

96;

Pat

elet

al.,

2003

)57

-96-

5S

ulfin

pyr

azo

neur

ico

suri

c;an

tithr

om

bo

ticF

DA

-ap

pro

ved

20�

g/m

L0.

92F

AS

1(�

29.3

7),

RIT

1(�

9.56

),T

IM18

(�7.

42)

7422

2-97

-2S

ulfo

met

uro

n-m

ethy

lhe

rbic

ide

not

liste

d1.

25�

g/m

L0.

90Y

OR

286W

(�11

.65)

,Y

PL2

67W

(�6.

52),

YP

L272

C(�

6.17

)10

540-

29-1

Tam

oxi

fen

antie

stro

gen

;ho

rmo

nal

FD

A-a

pp

rove

d3.

5�

g/m

L0.

86N

EO

1(�

31.4

4),

GC

D2

(�5.

2),

antin

eop

last

icY

NL1

79C

(�5.

18)

7862

8-80

-5T

erb

inaf

ine

antib

acte

rial

FD

A-a

pp

rove

d2.

5�

g/m

L0.

89E

RG

1(�

80.5

9),

YG

R17

6W(�

16.1

),E

RG

1(J

and

rosi

tzet

al.,

1991

)N

CP

1(�

8.64

)60

-54-

8T

etra

cycl

ine

antib

acte

rial

FD

A-a

pp

rove

d80

0�

g/m

L0.

96R

IM2

(�19

.18)

,R

PD

3(�

13.6

9),

SLS

1(�

8.84

)50

-35-

1T

halid

om

ide

imm

uno

mo

dul

ato

rF

DA

-ap

pro

ved

1.0

mg

/mL

0.93

YO

R31

8C(�

3.6)

,Y

OR

170W

(�3.

28),

YO

L083

W(�

3.19

)11

7-89

-5T

riflu

op

eraz

ine

antip

sych

otic

FD

A-a

pp

rove

d17

.5�

g/m

L0.

91N

EO

1(�

31.7

4),

NO

C2

(�5.

82),

YD

R33

9C(�

4.76

)97

322-

87-7

Tro

glit

azo

nean

tidia

bet

icF

DA

-30

0�

g/m

L0.

89P

OP

3(�

5),

PP

T1

(�3.

49),

YE

L001

Cd

isco

ntin

ued

(�3.

48)

1108

9-65

-9T

unic

amyc

inan

tibio

ticus

edin

not

liste

d15

0ng

/mL

0.83

YF

L032

W(�

67.7

7),

ALG

7(�

24),

ALG

7(B

arne

set

al.,

1984

)g

lyco

pro

tein

rese

arch

GF

A1

(�23

.81)

99-6

6-1

Val

pro

icA

cid

antic

onv

ulsa

nt;

antim

anic

;F

DA

-ap

pro

ved

300

�g

/mL

0.94

CIS

3(�

3.01

),Y

OL0

75C

(�2.

73),

antim

igra

ine

HC

S1

(�2.

18)

81-8

1-2

War

fari

nan

tico

agul

ant

FD

A-a

pp

rove

d50

0�

g/m

L0.

85Y

EN

1(�

7.75

),B

UL1

(�6.

62),

RIT

1(�

5.49

)

The

tab

lelis

tsea

chco

mp

oun

dth

atw

asan

alyz

edin

this

stud

y.T

hecl

inic

alin

dic

atio

n(M

erck

Ind

ex,

Thi

rtee

nth

Ed

itio

n),

reg

ulat

ory

stat

us(F

DA

Ora

nge

Bo

ok,

15S

ept,

2003

),d

ose

test

ed,

po

ol

fitne

ssle

vela

rep

rovi

ded

for

each

com

po

und

.In

add

itio

n,th

eth

ree

mo

stsi

gni

fican

to

utlie

rst

rain

sw

ithlo

g10

Pva

lues

(par

enth

eses

)ar

elis

ted

fro

mle

ftto

rig

htin

ord

ero

fsi

gni

fican

ce.

Rep

ort

edd

rug

targ

ets

are

acco

mp

anie

db

yre

leva

ntlit

erat

ure

cita

tions

.aC

AS

Reg

istr

y;S

epte

mb

er15

,20

03bT

heM

erck

Ind

ex,

Thi

rtee

nth

Ed

itio

ncF

DA

Ora

nge

Bo

ok:

http

://w

ww

.fd

a.g

ov/

cder

/ob

/def

ault.

htm

(15

Sep

tem

ber

,20

03)

Cell128

Figure 3. Comprehensive View of Fitness Profiles for 78 Compounds

Dot plots were calculated by comparing drug-specific growth rates of each heterozygous deletion strain to its average behavior from thereference set. Statistical significance was determined for each strain’s fitness value and the resulting P values were plotted for strains withdrug-specific growth deficits. Each data point represents one heterozygous deletion strain with a growth deficit. The confidence level of eachstrain’s growth rate is plotted along the x axis as the log10 P value. The P value distributions are most useful for identifying outlier strainswithin drug treatments. Compounds are sorted by the overall distribution of the heterozygous deletion pool with compounds causing relativelyfew fitness changes at the top and those causing widespread fitness changes at the bottom. Group I includes compounds with no drug-specific fitness changes. Compounds in Group II resulted in a small number of highly significant outliers. Group III compounds inducedwidespread fitness changes in the pool.

ence set. Hypersensitive heterozygotes resulting from Fitness Profiling Identifies Known Activitiesof Well-Characterized Compoundsthe compound treatments in Group II represent the most

attractive candidates for further analysis due to the rela- To determine the sensitivity of this method, we analyzedthe results for 20 compounds with reported protein tar-tively small number of highly significant outliers.

Discovering Drug Activities in Yeast129

Figure 4. High Resolution View of Selected Fitness Profiles from Group II Compounds

(A) Compounds with previously reported protein targets in yeast.(B) Selected compounds with no reported protein targets in yeast.In Figures 4A and 4B, the three most significant outliers are listed on the right side of each P value plot and the x axis has been specificallyscaled to the P value distribution of each drug treatment. Outliers are listed from left to right in order of decreasing log10 P values.

gets in yeast (Table 1). The correct targets or constit- ergosterol pathway. These included lovastatin (HMG1)(Rine et al., 1983), terbinafine (ERG1) (Jandrositz et al.,uents of target complexes were identified for the major-

ity of these well-characterized compounds (Figure 4A). 1991), clotrimazole (ERG11) (Truan et al., 1994), fenpro-pimorph (ERG2) (Marcireau et al., 1990), and dyclonineSix of the correctly identified targets involved com-

pounds that inhibit distinct biochemical steps along the (ERG2) (Hughes et al., 2000). Haloperidol affected ERG3

Cell130

and ERG24 heterozygotes, which are adjacent to the Inhibition of Lanosterol Synthase (Erg7p)proposed target (ERG2) in the pathway (Moebius et al., by Molsidomine1996). In addition, we found that three different sulfa- Molsidomine (N-ethoxycarbonyl-3-morpholino-sydnon-based drugs (asulam, sulfamethoxazole, and sulfanil- imine) is a potent vasodilator used clinically to treatamide) affected folic acid biosynthesis as reported in angina for the past twenty years. Enzymatic modificationthe literature (Nardese et al., 1996; Patel et al., 2003). of molsidomine in the human liver results in derivativesAlthough broad-spectrum lethality resulted from treat- (Figure 5A) that are effective nitric oxide donors, anment with carbendazim, a benomyl-analog known to activity that accounts for the vasodilatory propertiestarget tubulin, the most affected strains were heterozy- observed in mammals (Reden, 1990). Fitness profilinggotes of genes involved in microtubule formation (GIM2 revealed significant molsidomine-induced haploinsuffi-and GIM3) (Geissler et al., 1998; Neff et al., 1983). The ciency for only the ERG7 (lanosterol synthase) heter-targets were also correctly identified for compounds ozygote (Figure 5B). Lanosterol synthase is a highlythat inhibit the small unit of ribonucleotide reductase conserved and essential component of ergosterol bio-(hydroxyurea) (Rittberg and Wright, 1989); the 60s ribo- synthesis (Corey et al., 1994; Lees et al., 1995). Thissomal subunit (cycloheximide) (Stocklein and Piepers- finding was further supported by the observation thatberg, 1980); the protein kinase C complex (stauro- overexpression of ERG7 in a wild-type yeast strain con-sporine) (Yoshida et al., 1992); and protein glycosylation fers resistance to molsidomine (Figure 5C) and that mol-(tunicamycin) (Barnes et al., 1984). These examples sidomine causes an accumulation of the Erg7p sub-demonstrate that fitness profiling can accurately detect strate, 2,3-oxidosqualene (data not shown), as isthe cellular effects of drugs in a variety of biochemical observed with other Erg7p inhibitors and in an erg7pathways. deletion strain (Milla et al., 2002). All of these data are

Some compounds with known yeast targets yielded indicative of Erg7p inhibition.unexpected results. Treatment with methotrexate, for Next, we asked whether molsidomine could block theexample, resulted in identification of CIN1 but not the activity of lanosterol synthase from other organisms.gene (DFR1) encoding the known target dihydrofolate Specifically, we measured the effect of molsidomine andreductase (Huang et al., 1992). Given the nonessential its metabolites on mammalian and bacterial homologsrole of Cin1p in microtubule formation, it is likely that of ERG7. We found that lanosterol synthase purifiedthe observed drug sensitivity resulted from a synthetic from rat liver (Abe and Prestwich, 1995) was potentlylethal interaction with the known target. In another ex- inhibited in vitro by SIN-1, the first metabolic derivativeample, camptothecin unexpectedly induced a growth of molsidomine (Figure 5D). Likewise, in vitro inhibitiondefect in the trx2/TRX2 strain, while the heterozygote of recombinant lanosterol synthase from Alicyclobacil-of the known target gene (TOP1) was unaffected (Pom- lus acidocaldarius (Ochs et al., 1992) was observed formier et al., 1998). Identification of TRX2, a nonessential SIN-1 (Figure 5E). Together these results indicate thatgene involved in the tolerance of oxidative stress, is molsidomine effectively blocks sterol synthesis throughsupported by a recent report in plants showing that the inhibitory effect of its metabolite SIN-1 on lanosterolcamptothecin induces oxidative stress via the mito- synthase. Interestingly, previous reports have shownchondrial respiratory chain (Weir et al., 2003). The het- that molsidomine lowers cholesterol levels in both ratserozygote for PMA1, which encodes the H�-ATPase tar- and humans (Chassoux, 1989; Granzer and Ostrowski,geted by omeprazole (Monk et al., 1995), was not 1984). As the metabolism of molsidomine in humansaffected by drug treatment. This may be due to comple- and rats is quite similar, providing SIN-1 as an activementation of the drug-induced inhibition of Pma1p by intermediate (Tanayama et al., 1974; Wilson et al., 1987),the functionally overlapping PMA2 protein. Finally, three the results from this study identify the likely molecularcompounds (5-fluorocytosine, 5-fluorouracil, and FUDR) mechanism for the cholesterol lowering properties ofknown to inhibit Cdc21p induced growth defects in

this drug.members of the rRNA processing complex in additionto the expected target (CDC21 was one of the top 10

Disruption of Exosome-Specific rRNAmost affected strains) (Hardman et al., 2001). We haveProcessing by 5-Fluoruracilsubsequently demonstrated that the exosome is likelyThe antiproliferative antimetabolite 5-flourouracil (5FU)to be a major cellular target for 5FU in yeast (see follow-is one of the most successful and widely used chemo-up section on 5FU).therapeutics for the treatment of solid tumors in cancerA number of compounds with previously unknownpatients. It is thought that the major cytotoxic effects ofyeast targets induced fitness defects in specific hetero-5FU occur at the level of DNA synthesis as a competitivezygotes (Figure 4B). Among these findings was the dis-inhibitor of thymidylate synthetase (Parker and Cheng,covery that molsidomine blocked sterol biosynthesis at1990). However, a growing body of evidence suggestsa particular step in the pathway (ERG7). In addition,that the antiproliferation effects of 5FU may result fromwe found that tricyclic antidepressants and tamoxifeninhibition of RNA metabolism (Longley et al., 2003). Theaffected NEO1, a P-type ATPase with a number of hu-most convincing argument for the role of RNA metabo-man homologs. Caffeine was found to inhibit the TORlism as a target of 5FU comes from experiments showingprotein kinase complex. Other processes such as sphin-that cotreatment of cells with uridine, but not thymidine,golipid synthesis, fatty acid synthesis, cell wall metabo-relieves the cytotoxic and apoptotic effects of the druglism, and heme biosynthesis also appear to be targeted(Engelbrecht et al., 1984; Linke et al., 1996; Pritchard etby compounds in this study. The results from treatmental., 1997).with 5FU and molsidomine were chosen for follow-up

studies, which are described in the following sections. Treatment of the heterozygous deletion pool with 5FU

Discovering Drug Activities in Yeast131

Figure 5. Inhibition of Lanosterol Synthase by Molsidomine

(A) Chemical structures and the relationships between molsidomine and its metabolites, SIN-1, SIN-1A, and SIN-1C. Molsidomine is metabolizedin vivo by a series of transformations as illustrated. The vasodilatory properties of molsidomine occur through SIN-1C.(B) Fitness profile of molsidomine. The top four outliers are labeled in this dot plot. The ERG7 heterozygote displayed the most significantgrowth defect in response to molsidomine treatment (P value � 1 � 10�16).(C) Yeast overexpression of ERG7 confers resistance to molsidomine. A wild-type yeast strain was transformed with plasmids pDW394 (circle)or pDW394-ERG7 (square). Overnight cultures were diluted and grown in the presence of molsidomine (0–10 mg/mL) for 15 hr. The relativegrowth for each strain was calculated by dividing the optical density of the drug-treated culture by that of the untreated control. Each conditionwas carried out in triplicate.(D) Inhibition of purified rat liver oxidosqualene cyclase (OSC) activity. The in vitro conversion of [14C] 2,3-oxidosqualene by purified OSC wasmeasured after incubation at 37�C for 30 min with purified OSC. The conversions were analyzed by radio-TLC. The percent activity wascalculated by dividing each value from the drug treatment by that of the untreated control.(E) Inhibition of recombinant bacterial squalene:hopene cyclase (SHC) activity. The in vitro conversion of [14C]-squalene by recombinant SHCwas measured after incubation at 30�C for 30 min. The conversions were analyzed by radio-TLC. The percent activity was calculated bydividing each value from the drug treatment by that of the untreated control.

revealed drug-induced haploinsufficiency for eight (Venema and Tollervey, 1999). We monitored the re-moval of internal transcribed spacers, ITS1 and ITS2,genes that play roles in ribosome biogenesis in addition

to the previously reported target CDC21 (Figure 6A). by Northern blot analysis using ITS-specific oligonucleo-tide probes (Figure 6C). We found that 5FU treatmentFour of the seven genes (RRP6, RRP41, RRP44, and

RRP46) encode components of the exosome, a con- causes the accumulation of an rRNA processing productwe refer to as A2-C2 (Figure 6C). In addition, 5FU treat-served complex required for 5.8S rRNA and snoRNA 3�

end processing (Butler, 2002; Mitchell and Tollervey, ment causes a transient increase in 35S pre-rRNA levelssimilar to the effects of depletion of individual exosome2000). The other four genes, NOP4, MAK21, SSF1, and

YPR143W, play roles in either 25S rRNA processing, 60S components, including Rrp6p, Rrp41p, and Rrp44p (All-mang et al., 2000). Interestingly, the untreated rrp6ribosome production, 27S rRNA processing, or associ-

ate with other nucleolar components (Edskes et al., strain displays a phenocopy of the 5FU-induced accu-mulation of the A2-C2 rRNA processing product (Fig-1998; Fatica et al., 2002; Ho et al., 2002; Sun and Wool-

ford, 1994). ure 6C).Next, we asked whether 5FU treatment enhanced theTo further characterize the effect of 5FU, we moni-

tored rRNA intermediates in normal and mutant yeast accumulation of A2-C2 in specific heterozygous deletionstrains identified in this study. The wild-type control andstrains following drug treatment. Figure 6B outlines the

major events in rRNA processing, which begins with each of the heterozygotes were grown in the presenceof 5FU followed by Northern blot analysis to measuresynthesis of the 35S pre-rRNA by RNA polymerase I

and proceeds by a series of endo- and exonucleolytic A2-C2 levels (Figure 6D). Strains that were heterozygousfor components of the exosome (RRP6, RRP41, RRP44,cleavages that yield three of the four mature rRNAs

Cell132

Figure 6. Disruption of Exosome Specific rRNA Processing by 5FU

(A) Fitness Profile of 5FU. Eight of the top ten outliers (RRP6, RRP41, MAK21, YPR143W, RRP46, SSF1, RRP44, NOP4) play roles in ribosomebiogenesis. CDC21, the reported target of 5FU, is also among the top ten outliers.(B) Diagram of the major events of rRNA processing in S. cerevisiae. The positions complementary to oligonucleotide probes used are indicatedbelow the 35S pre-rRNA. Processing begins with cleavage of the 35S precursor rRNA (35S pre-rRNA) at sites A0 and A1 in the 5� externaltranscribed spacer (5� ETS) and at B2 to yield the 32S pre-rRNA. Next, cleavage of 32S pre-rRNA yields 27S pre-rRNA and 20S pre-rRNA,which is converted to mature 18S rRNA by further cleavage. The 27S pre-rRNA follows either of two parallel pathways each of which yieldsmature 25S rRNA. The pathways differ in the mechanism that generates the mature 5� end of 5.8S rRNA. The exosome and Rrp6p form the3� end of 5.8S rRNA by exonucleolytic removal of the 3� end of 7S pre-rRNA: the exosome removes all but the last 30 nucleotides, which isthen trimmed off by Rrp6p.(C) Northern blot analysis of rRNA processing intermediates. A wild-type diploid was treated in culture with 200 �M 5FU and RNA was preparedfrom samples taken at the times indicated at the top of the figure (lanes 1–3). RNA was also prepared from an untreated wild-type haploid(lane 4) and an isogenic rrp6::KAN mutant (lane 5). The blot was probed with radiolabeled oligonucleotides (indicted at the left of each image)complementary to specific pre-rRNA intermediates and mature 5.8S rRNA. Notation to the right of each image indicates the identity of themajor pre-rRNA intermediates.

Discovering Drug Activities in Yeast133

and RRP46) produced 2- to 10-fold more A2-C2 than the Another advantage of the fitness assay is that biologi-cal processes that are affected by a given compoundwild-type strain treated with 5FU (Figure 6D). Although

there is evidence that 5FU may result in general RNA are identified in addition to the precise protein target(s).For example, treatment of the heterozygous deletiondamage, our results reveal a specific molecular mecha-

nism by which 5FU may inhibit cell growth through per- pool with 5FU resulted in the identification of a groupof genes that encode components of the exosome com-turbation of rRNA processing by the exosome complex.

In fact, preliminary data suggest that 5FU prolongs the plex. Another example involved the drug tunicamycin,which is known to affect protein glycosylation. Besidestime yeast cells spend in G2/M and that this effect

is exacerbated by the exosome mutations (data not identifying the known target (ALG7p), heterozygotes in-volved in either the endoplasmic reticulum (ER) stressshown).response (HAC1) or cell wall glycosylation (GFA1) alsodisplayed sensitivity to tunicamycin. Finally, identifica-Discussiontion of the fol2/FOL2 strain in response to the sulfa-based drugs or the cdc21/CDC21 and rho1/RHO1Use of Fitness Profiling For Understanding

Drug Activities strains following staurosporine treatment also confirmsthe effectiveness of this assay in detecting affectedWe have demonstrated that fitness profiling is a power-

ful approach for analyzing the modes of action of clini- pathways.The dataset presented here has yet to be fully minedcally relevant compounds. Identifying the protein targets

of existing drugs can reveal sources of pharmacologi- and offers opportunities for many follow-up experi-ments. Of the 56 compounds yielding clear outliers, wecally desirable side effects in addition to the mecha-

nisms for unwanted toxicities. Understanding the mo- chose two examples for in-depth characterization byconventional experimental techniques. In the first exam-lecular mechanisms for therapeutic compounds is also

critical because it allows researchers to develop sec- ple, fitness results suggesting that lanosterol synthase(Erg7p) may be a target of molsidomine were confirmedond-generation compounds with improved pharmaco-

logical properties. Knowing the cellular targets for clini- in yeast by overexpression resistance and biochemicalanalyses of pathway intermediates. This work was fur-cally proven small molecules can also reveal drug

targets that are already validated in humans. ther validated by monitoring the inhibitory activity of thiscompound on the bacterial and mammalian enzymesThis highly parallel and sensitive yeast assay offers

several key advantages over traditional approaches for in vitro. Interestingly, the cholesterol-lowering effect ofmolsidomine appears to be distinct from the nitric oxideanalyzing drug activities. First, this approach requires

no prior knowledge of compound mode of action, which releasing activity, which accounts for the drug’s vasodi-latory properties. This new information provides poten-allows truly novel drug activities to be uncovered in a

systematic and unbiased fashion. For example, we tial clinical validation for lanosterol synthase as a safeand effective target for the development of new choles-found that the tricyclic antidepressants (amitriptyline,

imipramine, desipramine, clomipramine, chlorproma- terol-lowering drugs.In the second example, we performed a set of experi-zine, and trifluoperazine) affected the growth of the

NEO1 heterozygote (Figure 3C). NEO1 encodes an inte- ments to confirm the hypothesis that 5FU blocks rRNAprocessing by the exosome, which may account forgral membrane protein in the P-type ATPase superfamily

(Axelsen and Palmgren, 1998). Although the primary some of this drug’s antiproliferation activities. At presentit remains unclear how inhibition of the exosome bymode of antidepressive action of these compounds oc-

curs by the nonselective inhibition of neurotransmitter 5FU inhibits cellular growth. Defects in rRNA processingcould potentially cause inhibition of protein synthesis.reuptake, these compounds confer a number of deleteri-

ous side effects by mechanisms that are poorly under- Alternatively, recent evidence suggests that tumor sup-pressor proteins such as nucleolar ARF may mediatestood (Hardman et al., 2001). Evidence for TCA inhibition

of ATPase activity has been documented in human crosstalk between the rRNA processing pathway andcomponents that regulate cell cycle progression (Ita-erythrocyte plasma membranes and skeletal sarcoplas-

mic reticulum membranes (Plenge-Tellechea et al., hana et al., 2003; Ruggero and Pandolfi, 2003). Our pre-liminary data (not shown) also suggest that 5FU affects1999; Soler et al., 2000). We found that the growth rate

of the NEO1 heterozygote was also impeded by tamoxi- cell cycle timing in yeast that is exacerbated by muta-tions in exosome components. Hence, these findingsfen, a potent antiestrogen drug for the treatment of

breast cancer. The chemical structures of the tricyclics are significant in light of a growing body of researchindicating a critical relationship between ribosome bio-and tamoxifen interestingly share a common lipophilic

moiety characterized by an amine tail. Additional studies genesis and cell cycle control (Du and Stillman, 2002;Ruggero et al., 2003; Sugimoto et al., 2003; Volarevic etwill be needed to further characterize the putative ef-

fects of these compounds on ATPase activity in yeast al., 2000). Moreover, the cytotoxic effects of 5FU in hu-man cells may result from p53-mediated effects on RNAand humans.

(D) Analysis of A2-C2 rRNA processing product accumulation in wild-type (�/�) and mutant yeast strains treated for 30 min with the indicatedconcentration of 5FU. The top image is a Northern blot probed with OSB155 (see Figure 6B) and the central image shows the same blotprobed with an oligonucleotide complementary to SCR1 RNA (OSB151). The graph illustrates the level of A2-C2 rRNA processing productdetermined by storage phosphorimager analysis and normalized to the level of SCR1 for each strain and is the average of two independent exper-iments.

Cell134

metabolism. This is supported by data showing that that this gene encodes a potential target for this drug.However, the open reading frame of YML176C overlapsadministration of the drug causes apoptosis in p53�/�,with ERG1 (YML175C), the known target of terbinafine,but not in p53�/� intestinal epithelial cells. In addition,which suggests that the sensitivity of this heterozygotecoadministration of 5FU with uridine, but not thymidine,was likely due to the inadvertent disruption of the ERG1inhibited 5FU-induced apoptosis (Pritchard et al., 1997).open reading frame. Similarly, background mutationsThe dependence of 5FU action on p53 and connectionsthat are unlinked to the heterozygous locus or crosshy-between p53 and rRNA metabolism suggest that thebridizing tags may also result in false positives. Thep53 pathway may act as a sensor of defects in rRNAnumber of false positives can be minimized by confirm-processing thereby signaling cell cycle arrest and/oring the fitness results with the individual deletion strainsapoptosis. A clear understanding of the role of the exo-and testing whether overexpression of the candidatesome in the p53 response to 5FU may provide a newgene confers resistance to the drug of interest.opportunity for therapeutic intervention in the treatment

A critical component of this study was the use of aof cancer.strain-specific error model to identify drug-dependenteffects that are statistically significant. Since some het-Limitations and Technical Considerationserozygotes have nonspecific growth defects, it is neces-While the fitness assay is a powerful approach for identi-sary to understand the normal behavior of each strainfying drug targets, there are also a number of limitations.before calculating drug-induced effects. We determinedFirst, the compound of interest must be able to affectthat a reference set must be populated with a minimumthe growth rate of the cell. Failure of a compound toof 50 experiments and that chemical treatments areimpart a fitness defect may indicate that (1) the drugmore effective compared to nontreated controls for thetarget is not encoded in the yeast genome; (2) drugreference set. When performed as described here, thiseffects are masked by other proteins with redundantmethod will detect as little as 5–10% growth deficitsactivities; (3) the drug was not correctly metabolized; orin individual strains (data not shown). However, it is(4) the drug was not able to enter the cell. However, theimportant to note that each study must generate its ownability of a compound to affect the growth rate of yeastreference set due to the inherent variability of experi-does not guarantee that a target will be identified bymental conditions. Sensitivity could theoretically be in-this approach. In order for this approach to work, acreased by extending the pool outgrowth beyond 20heterozygote for the target gene must be present in thegenerations and by increasing the number of experi-mutant pool. Unfortunately, heterozygotes representingments included in the reference set. We have also foundonly about half of the yeast genome were available atthat P values from this approach provide an effective

the time of this study, which resulted in a lower detectionway to rank order the strains in a given drug treatment

rate than would be expected from a full genome pool.to identify the significant outliers. The ability to compare

Failure to detect the precise protein targets of stauro-P values across compounds is limited by the fact that

sporine (PKC1) and the sulfa-based drugs (FOL1 andcompounds have very diverse modes of action and the

FOL3) can be attributed to the fact that the target strainsdoses of different compounds cannot be considered

were not in the pool. Another requirement of this assay equivalent.is that the activity level of the targeted protein must beinfluenced by the dosage level of the corresponding Future Directionsgene under the conditions profiled. Finally, compounds We have demonstrated that fitness profiling is a power-that exert their effects through direct interaction with ful approach for studying the activities of clinically rele-nonprotein elements in the cell, such as DNA (actinomy- vant compounds in yeast. Tagged heterozygous dele-cin D) or ergosterol (amphotericin B), do not appear tion strains have been made available for virtually everysuitable for this approach. gene in the yeast genome since this study was com-

In some cases, the increased drug sensitivities of het- pleted. This resource coupled with commercially avail-erozygotes are caused by indirect effects on the cell able tag arrays make it possible to extend this type ofand do not represent primary drug targets. For instance, analysis to the full genome level. In addition, a similarmembers of the multidrug resistant (MDR) family of strategy using pools of homozygous deletion strainstransporters (e.g., PDR5) were implicated in a number could be used to reveal targets that are currently maskedof drug treatments. These nonessential efflux pumps by redundant activities. Growth conditions can also beact to reduce the effective intracellular concentration of modified to assess drug sensitivities of genes that arecertain structural classes of compounds. Deletion of one expressed only under certain physiological conditions.of the two copies of these genes in a diploid cell can Recently developed techniques to introduce specificresult in a reduced capacity to export drugs from the mutations into all strains of a deletion pool make it possi-cell, thereby causing a hypersensitivity phenotype of the ble to identify drug effects in a variety of different geneticheterozygotes. Comparison of gene expression profiles backgrounds (Ooi et al., 2001). Finally, fitness profilingfrom putative target deletion strains with those of drug results can be combined with data from other large-treatments can help distinguish actual drug targets from scale functional genomics efforts to enhance our under-proteins involved in an indirect cellular drug response standing of compound activities.(Hughes et al., 2000).

Experimental ProceduresSome of the candidate targets identified with this ap-proach are due to artifacts that were introduced during

Supplemental Datathe construction of the deletion strains. For example, A comprehensive list of P values for all experiments discussed herethe heterozygote for YML176C displayed increased sen- is available on the Cell website. Log ratio data is available upon re-

quest.sitivity to terbinafine (P value 1 � 10�16) indicating

Discovering Drug Activities in Yeast135

Yeast Strains and Plasmids rays consisting of either 5873 UP tag probes or 5873 DOWN tagprobes. DNA microarrays consisting of 20-mer probes (see strainIsogenic heterozygous deletion strains were obtained from the Sac-

charomyces Genome Deletion Consortium (Winzeler et al., 1999). A information for probe sequences) were synthesized as described inHughes et al. (2000). Hybridizations were done in 3.5 ml of 1� SSTEplasmid expressing yeast ERG7 was constructed by PCR amplifica-

tion of the open reading frame from yeast genomic DNA. The ORF hybridization buffer (1 M NaCl, 10 mM Tris [pH 7.0], 0.5% TritonX-100) containing 10 pM gridline DNA for 3 hr at 40�C. Arrays werewas cloned by in vivo recombination into pDW394, a high copy

URA3 vector with expression driven by the HOR7 promoter. first washed in 6� SSPE � 0.05% Triton-X and then in ice-cold0.06� SSPE. Slides were then scanned on a GMS 418 scanner(Genetic Microsystems, Woburn, MA).Construction of Heterozygous Deletion Pool

The collection of 3503 heterozygous deletion strains was obtainedfrom the Saccharomyces Deletion Consortium (see website for Data Analysisstrain and tag information: http://www-sequence.stanford.edu/ Descriptions of formulas used for array quantitation and P valuegroup/yeast_deletion_project/Enter_function.html) as saturated glyc- derivation are available on the Cell website.erol stocks in 96-well plates. Strains were transferred to YEPD agarplates with a 96-well colony replicator (V & P Scientific, Inc., San

Northern Blot AnalysisDiego, CA) and were grown at 30�C for 72 hr. All colony-formingrRNA processing was analyzed by Northern blot analysis of totalunits were collected with a rubber spatula and combined in a 50 mlRNA samples as described by Briggs et al. (1998). Transfer mem-conical tube. The cell mixture was washed in 30 ml of YEPD andbranes were hybridized with 5� 32P-labeled deoxyribonucleotidesdiluted to 1 � 1010 cells per mL in YEPD with 15% glycerol. SingleOSB151 (5�-gtctagccgcgaggaagg-3�); OSB155 (5�-tccagttacgaaaatuse 2 ml aliquots were flash-frozen in a dry ice/ethanol bath andtcttgtttttgacaa-3�); OSB154 (5�-tcttgcccagtaaaagctctcatgc-3�);stored at �80�C until use.OSB158 (5�-gttcgcctagacgctctcttc-3�); OSB157 (5�-ggtgaccaatttcaagtta-3�); OSB156 (5�-cgctgcgttcttcatcgatgcg-3�); or OSB138 (5�-

Drug Treatmentstcagagatcttggtgataat-3�), which are complementary to SCR1 RNA,

To determine the appropriate drug treatment concentrations, aITS1, ITS2, and 5.8S, respectively (Figure 6B). Blots were analyzed

5-fold dilution series of each compound was made in syntheticby storage phosphorimager analysis.

complete medium with 2% glucose and 2% (w/v) casamino acids(SCC) for eleven 2� drug concentrations and one vehicle control.

Enzyme Activity AssayA 100 �l aliquot of each dose was distributed to a 96-well tissueChemicalsculture plate in triplicate. Subsequently, a 2� cell suspension was(3S)-[24,30]2,3-oxidosqualene was synthesized as described by Baimade by thawing an aliquot of the frozen deletion pool on ice beforeet al. (1998). Triton X-100 was purchased from Sigma. Molsidomine,harvesting by centrifugation, washing twice with SCC, and dilutingSIN-1-hydrochloride, and SIN-1A/�CD complex (10%, w/w SIN-1Ato 2.65 � 104 cells/ml in SCC at room temperature. A 100 �l aliquotcontent) were obtained from Alexis Biochemicals.of the 2� cell mixture was then added to each of the 96 wells inBiological Materialsthe drug-containing tissue culture plate. Plates were then incubatedRat liver lanosterol cyclase was purified according to the procedureat 30�C with gentle agitation on a rotary shaker for 15 hr reachingdescribed by Abe and Prestwich (1995).a density of 1.3 � 107 cells per mL (vehicle control). The pool fitnessEnzyme Assay for OSC(PF) was calculated (PF 1 � [(log(OD600treated/OD600untreated))/The reaction mixture contained 100 mM sodium citrate, [pH 7.4],(log(2)g )], where g is the number of generations assumed to be 10)0.1% (v/v) Triton X-100, 5 �M [14C]2,3-oxidosqualene (52.6 mCi/for each drug concentration and doses that resulted in 0.70 � PF �

mmol), and 5 �g of purified enzyme in a final volume of 1 mL. After0.96 were selected for fitness profiling experiments.incubation at 37�C for 30 min, the reaction was stopped by additionof 1 ml CH2Cl2. The extract was concentrated using a Speed-Vac,Fitness Profilingand then subjected to silica gel TLC (Whatmann LK6D). The TLCAn aliquot of the frozen pool was thawed at room temperature andplates were developed with CH2Cl2. The conversions were then ana-diluted in 100 ml of SCC to a final density of 1.3 � 107 cells/ml. Alyzed by radio-TLC (Bio-Scan, System 200 Imaging Scanner). All5 ml aliquot was harvested immediately (G-0) by centrifugation andassays were carried out in triplicate.frozen at �80�C until DNA isolation. Dilutions of 1.3 � 104 cells/mlEnzyme Assay for SHC Bacterial Lanosterol Cyclasein 20 ml aliquots were grown at 30�C for 10 generations to a densityRecombinant A. acidocaldarius lanosterol cyclase was expressedof 1.3 � 107 cells/ml. Cultures were then diluted once again to 1.3 �in E. coli and purified as outlined by Ochs et al (1992). To 50 �l of104 cells/ml in fresh media or fresh media containing drug. After5 mM sodium citrate, [pH 6.0], 0.1% (v/v) Triton X-100 was addedanother ten population doublings, 1 � 108 cells from each culture4 �l of a solution of SHC (600 �g/ml) and 3 �l (0.021 �Ci) of [14C]-were harvested (G-20) as described above. All harvested cell pelletssqualene (1 mM in ethanol, 7 mCi/mmole). After incubation at 60�Cwere thawed on ice and subjected to a 30 s, 50 krpm mechanicalfor 30 min, the reaction was stopped by addition of 1 ml CH2Cl2. Thelysis with 0.5 mm glass beads in a Mini-Beadbeater (Biospec Prod-extract was concentrated using a Speed-Vac and then subjected toucts, Bartlesville, OK). Genomic DNA was subsequently isolatedsilica gel TLC. The TLC plates were developed to a distance of 5using the MasterPure Yeast DNA Purification Kit (Epicentrecm in chloroform and after drying for 15 cm in hexane. The conver-#mpy80200), quantified, and normalized to 15 ng/�l.sions were then analyzed by radio-TLC.

Tag Amplification and Hybridization to DNA MicroarraysFor each treatment, genomic DNA isolated from cultures harvested Acknowledgmentsat G-0 and G-20 culture was labeled with Cy3 and Cy5, respectively.The UP tag and DOWN tag barcodes were amplified and labeled in We thank H. Dai, R. Stoughton, A. Adams, A. Sachs, and S. Friend

for many helpful discussions; M. Kuo for literature research on com-separate 50 �l reactions with 1 �l (15 ng) of genomic DNA, 5 ul ofUP tag or DOWN tag primer mix, and 44 �l of PCR SuperMix (PCR pounds profiled in this study; J. Koch, T. Ward, J. Burchard, and E.

Coffey for technical assistance; and C. Raymond and L. Lim forPlatinum SuperMix, Gibco #11306-016). The UP tag primer mix con-tained 5 �M of unlabeled UP tag forward primer (5�-gatgtccac comments on the manuscript. We are grateful to all members of the

Saccharomyces Deletion Consortium for their efforts. We apologizegaggtctct-3� and Cy3- or Cy5-labeled reverse UP tag primer (5�-Cy-gtcgacctgcagcgtacg-3�). The DOWN tag primer mix contained 5 �M to the many researchers whose work could not be cited due to

space considerations and refer the reader to the Saccharomycesof Cy3- or Cy5-labeled DOWN tag forward primer (5�-Cy-cgagctcgaattcatcg-3�) and 5 �M of unlabeled DOWN tag reverse primer (5�- Genome Database (http://www.yeastgenome.org) for information

and references for specific gene products. This work was supportedcggtgtcggtctcgtag-3�). PCR conditions were as follows: 5 min at94�C; then 35 cycles of 30 s at 94�C; 30 s at 50�C; and 30 s at 72�C by Rosetta Inpharmatics, LLC; two grants from the Public Health

Service (GM59898 and CA95913) to J.S.B.; a grant from the Nationalfollowed by 7 min at 72�C to ensure product elongation. Blockingprimers complementary to the common priming sites were annealed Institutes of Health (GM62104) to M.B.; and a grant (GM 44836)

to G.D.P.prior to hybridization. Labeled tags were hybridized to DNA microar-

Cell136

Received: October 30, 2003 ings, P.J. (1992). The phenotype of a dihydrofolate reductase mutantof Saccharomyces cerevisiae. Gene 121, 167–171.Revised: November 24, 2003

Accepted: December 16, 2003 Huels, C., Muellner, S., Meyer, H.E., and Cahill, D.J. (2002). ThePublished online: December 24, 2003 impact of protein biochips and microarrays on the drug development

process. Drug Discov. Today 7, S119–S124.References Hughes, T.R., Marton, M.J., Jones, A.R., Roberts, C.J., Stoughton,

R., Armour, C.D., Bennett, H.A., Coffey, E., Dai, H., He, Y.D., etAbe, I., and Prestwich, G.D. (1995). Molecular cloning, characteriza- al. (2000). Functional discovery via a compendium of expressiontion, and functional expression of rat oxidosqualene cyclase cDNA. profiles. Cell 102, 109–126.Proc. Natl. Acad. Sci. USA 95, 9274–9278.

Itahana, K., Bhat, K.P., Jin, A., Itahana, Y., Hawke, D., Kobayashi,Allmang, C., Mitchell, P., Petfalski, E., and Tollervey, D. (2000). Deg- R., and Zhang, Y. (2003). Tumor suppressor ARF degrades B23, aradation of ribosomal RNA precursors by the exosome. Nucleic nucleolar protein involved in ribosome biogenesis and cell prolifera-Acids Res. 28, 1684–1691. tion. Mol. Cell 12, 1151–1164.Axelsen, K.B., and Palmgren, M.G. (1998). Evolution of substrate Jandrositz, A., Turnowsky, F., and Hogenauer, G. (1991). The genespecificities in the P-type ATPase superfamily. J. Mol. Evol. 46, encoding squalene epoxidase from Saccharomyces cerevisiae:84–101. cloning and characterization. Gene 107, 155–160.Bai, M., Xiao, X.Y., and Prestwich, G.D. (1998). Epoxidation of 2,3

Knopp, R.H. (1999). Drug treatment of lipid disorders. N. Engl. J.oxidosqualene to 2,3;22,23-squalene dioxide by squalene epoxi-

Med. 341, 498–511.dase. Biochem. Biophys. Res. Commun. 185, 323–329.

Lander, E.S., Linton, L.M., Birren, B., Nusbaum, C., Zody, M.C.,Barnes, G., Hansen, W.J., Holcomb, C.L., and Rine, J. (1984). Aspar-

Baldwin, J., Devon, K., Dewar, K., Doyle, M., FitzHugh, W., et al.agine-linked glycosylation in Saccharomyces cerevisiae: genetic

(2001). Initial sequencing and analysis of the human genome. Natureanalysis of an early step. Mol. Cell. Biol. 4, 2381–2388.

409, 860–921.Briggs, M.W., Burkard, K.T., and Butler, J.S. (1998). Rrp6p, the yeast

Lees, N.D., Skaggs, B., Kirsch, D.R., and Bard, M. (1995). Cloninghomologue of the human PM-Sc1 100-kDa autoantigen, is essential

of the late genes in the ergosterol biosynthetic pathway of Saccharo-for efficient 5.8 rRNA 3� end formation. J. Biol. Chem. 273, 13255–

myces cerevisiae–a review. Lipids 30, 221–226.13263.

Linke, S.P., Clarkin, K.C., Di Leonardo, A., Tsou, A., and Wahl, G.M.Butler, J.S. (2002). The yin and yang of the exosome. Trends Cell

(1996). A reversible, p53-dependent G0/G1 cell cycle arrest inducedBiol. 12, 90–96.

by ribonucleotide depletion in the absence of detectable DNA dam-Chassoux, G. (1989). Molsidomine and lipid metabolism. J. Cardio- age. Genes Dev. 10, 934–947.vasc. Pharmacol. 14 (Suppl 11), S137–S138.

Longley, D.B., Harkin, D.P., and Johnston, P.G. (2003). 5-fluorouracil:Corey, E.J., Matsuda, S.P., and Bartel, B. (1994). Molecular cloning, mechanisms of action and clinical strategies. Nat. Rev. Cancer 3,characterization, and overexpression of ERG7, the Saccharomyces 330–338.cerevisiae gene encoding lanosterol synthase. Proc. Natl. Acad. Sci.

MacBeath, G., and Schreiber, S.L. (2000). Printing proteins as mi-USA 91, 2211–2215.

croarrays for high-throughput function determination. ScienceDu, Y.C., and Stillman, B. (2002). Yph1p, an ORC-interacting protein: 289, 1760–1763.potential links between cell proliferation control, DNA replication,

Marcireau, C., Guilloton, M., and Karst, F. (1990). In vivo effects ofand ribosome biogenesis. Cell 109, 835–848.

fenpropimorph on the yeast Saccharomyces cerevisiae and deter-Edskes, H.K., Ohtake, Y., and Wickner, R.B. (1998). Mak21p of Sac- mination of the molecular basis of the antifungal property. Antimi-charomyces cerevisiae, a homolog of human CAATT-binding pro- crob. Agents Chemother. 34, 989–993.tein, is essential for 60 S ribosomal subunit biogenesis. J. Biol.

Milla, P., Athenstaedt, K., Viola, F., Oliaro-Bosso, S., Kohlwein, S.D.,Chem. 273, 28912–28920.

Daum, G., and Balliano, G. (2002). Yeast oxidosqualene cyclaseEngelbrecht, C., Ljungquist, I., Lewan, L., and Yngner, T. (1984). (Erg7p) is a major component of lipid particles. J. Biol. Chem.Modulation of 5-fluorouracil metabolism by thymidine. In vivo and 277, 2406–2412.in vitro studies on RNA-directed effects in rat liver and hepatoma.

Mitchell, P., and Tollervey, D. (2000). Musing on the structural organi-Biochem. Pharmacol. 33, 745–750.

zation of the exosome complex. Nat. Struct. Biol. 7, 843–846.Fatica, A., Cronshaw, A.D., Dlakic, M., and Tollervey, D. (2002). Ssf1p

Moebius, F.F., Bermoser, K., Reiter, R.J., Hanner, M., andprevents premature processing of an early pre-60S ribosomal parti-

Glossmann, H. (1996). Yeast sterol C8–C7 isomerase: identificationcle. Mol. Cell 9, 341–351.

and characterization of a high-affinity binding site for enzyme inhibi-Foury, F. (1997). Human genetic diseases: a cross-talk between man tors. Biochemistry 35, 16871–16878.and yeast. Gene 195, 1–10.

Monk, B.C., Mason, A.B., Abramochkin, G., Haber, J.E., Seto-Young,Geissler, S., Siegers, K., and Schiebel, E. (1998). A novel protein D., and Perlin, D.S. (1995). The yeast plasma membrane protoncomplex promoting formation of functional alpha- and gamma- pumping ATPase is a viable antifungal target. I. Effects of the cys-tubulin. EMBO J. 17, 952–966. teine-modifying reagent omeprazole. Biochim. Biophys. Acta

1239, 81–90.Giaever, G., Shoemaker, D.D., Jones, T.W., Liang, H., Winzeler, E.A.,Astromoff, A., and Davis, R.W. (1999). Genomic profiling of drug Nardese, V., Gutlich, M., Brambilla, A., and Carbone, M.L. (1996).sensitivities via induced haploinsufficiency. Nat. Genet. 21, 278–283. Disruption of the GTP-cyclohydrolase I gene in Saccharomyces cer-

evisiae. Biochem. Biophys. Res. Commun. 218, 273–279.Granzer, E., and Ostrowski, J. (1984). Lipoprotein regulatory effectof molsodomine in rats. Arzneimittelforschung 34, 191–193. Neff, N.F., Thomas, J.H., Grisafi, P., and Botstein, D. (1983). Isolation

of the beta-tubulin gene from yeast and demonstration of its essen-Hardman, J.G., Limbard, L.E., and Goodman Gilman, A. (2001).Goodman and Gilman’s The Pharmacological Basis of Therapeutics, tial function in vivo. Cell 33, 211–219.10th ed. (New York, NY: McGraw-Hill). Ochs, D., Kaletta, C., Entian, K.D., Beck-Sickinger, A., and Poralla,

K. (1992). Cloning, expression, and sequencing of squalene-hopeneHeitman, J., Movva, N.R., and Hall, M.N. (1991). Targets for cellcycle arrest by the immunosuppressant rapamycin in yeast. Science cyclase, a key enzyme in triterpenoid metabolism. J. Bacteriol.

174, 298–302.253, 905–909.

Ho, Y., Gruhler, A., Heilbut, A., Bader, G.D., Moore, L., Adams, S.L., Ooi, S.L., Shoemaker, D.D., and Boeke, J.D. (2001). A DNA microar-ray-based genetic screen for nonhomologous end-joining mutantsMillar, A., Taylor, P., Bennett, K., Boutilier, K., et al. (2002). System-

atic identification of protein complexes in Saccharomyces cerevis- in Saccharomyces cerevisiae. Science 35, 2552–2556.iae by mass spectrometry. Nature 415, 180–183. Parker, W.B., and Cheng, Y.C. (1990). Metabolism and mechanism

of action of 5-fluorouracil. Pharmacol. Ther. 48, 381–395.Huang, T., Barclay, B.J., Kalman, T.I., von Borstel, R.C., and Hast-

Discovering Drug Activities in Yeast137

Patel, O., Satchell, J., Baell, J., Fernley, R., Coloe, P., and Macreadie, (2000). Proliferation, but not growth, blocked by conditional deletionof 40S ribosomal protein S6. Science 288, 2045–2047.I. (2003). Inhibition studies of sulfonamide-containing folate analogs

in yeast. Microb. Drug Resist. 9, 139–146. Weir, I.E., Pham, N.A., and Hedley, D.W. (2003). Oxidative stress isgenerated via the mitochondrial respiratory chain during plant cellPhizicky, E., Bastiaens, P.I., Zhu, H., Snyder, M., and Fields, S.

(2003). Protein analysis on a proteomic scale. Nature 422, 208–215. apoptosis. Cytometry 54A, 109–117.

White, T.C., Marr, K.A., and Bowden, R.A. (1998). Clinical, cellular,Plenge-Tellechea, F., Soler, F., and Fernandez-Belda, F. (1999). Tri-cyclic antidepressants inhibit the Ca(2�)-dependent ATPase activity and molecular factors that contribute to antifungal drug resistance.

Clin. Microbiol. Rev. 11, 382–402.from plasma membrane. Arch. Biochem. Biophys. 370, 119–125.

Pommier, Y., Pourquier, P., Fan, Y., and Strumberg, D. (1998). Mech- Wilson, I.D., Fromson, J.M., Illing, H.P., and Schraven, E. (1987). Themetabolism of [14C]N-ethoxycarbonyl-3-morpholinosydnonimineanism of action of eukaryotic DNA topoisomerase I and drugs tar-

geted to the enzyme. Biochim. Biophys. Acta 1400, 83–105. (molsidomine) in man. Xenobiotica 17, 93–104.

Winzeler, E.A., Shoemaker, D.D., Astromoff, A., Liang, H., Anderson,Pritchard, D.M., Watson, A.J., Potten, C.S., Jackman, A.L., and Hick-man, J.A. (1997). Inhibition by uridine but not thymidine of p53- K., Andre, B., Bangham, R., Benito, R., Boeke, J.D., Bussey, H., et

al. (1999). Functional characterization of the S. cerevisiae genomedependent intestinal apoptosis initiated by 5-fluorouracil: evidencefor the involvement of RNA perturbation. Proc. Natl. Acad. Sci. USA by gene deletion and parallel analysis. Science 285, 901–906.94, 1795–1799. Yoshida, S., Ikeda, E., Uno, I., and Mitsuzawa, H. (1992). Character-

ization of a staurosporine- and temperature-sensitive mutant, stt1,Reden, J. (1990). Molsidomine. Blood Vessels 27, 282–294.of Saccharomyces cerevisiae: STT1 is allelic to PKC1. Mol. Gen.Rine, J., Hansen, W., Hardeman, E., and Davis, R.W. (1983). TargetedGenet. 231, 337–344.selection of recombinant clones through gene dosage effects. Proc.

Natl. Acad. Sci. USA 80, 6750–6754. Ziauddin, J., and Sabatini, D.M. (2001). Microarrays of cells express-ing defined cDNAs. Nature 411, 107–110.Rittberg, D.A., and Wright, J.A. (1989). Relationships between sensi-

tivity to hydroxyurea and 4-methyl-5-amino-1-formylisoquinolinethiosemicarbazone (MAIO) and ribonucleotide reductase RNR2mRNA levels in strains of Saccharomyces cerevisiae. Biochem. CellBiol. 67, 352–357.

Ruggero, D., and Pandolfi, P.P. (2003). Does the ribosome translatecancer? Nat. Rev. Cancer 3, 179–192.

Ruggero, D., Grisendi, S., Piazza, F., Rego, E., Mari, F., Rao, P.H.,Cordon-Cardo, C., and Pandolfi, P.P. (2003). Dyskeratosis congenitaand cancer in mice deficient in ribosomal RNA modification. Science299, 259–262.

Schreiber, S.L., and Crabtree, G.R. (1992). The mechanism of actionof cyclosporin A and FK506. Immunol. Today 13, 136–142.

Service, R.F. (2000). Biochemistry. Protein arrays step out of DNA’sshadow. Science 289, 1673.

Soler, F., Plenge-Tellechea, F., Fortea, I., and Fernandez-Belda, F.(2000). Clomipramine and related structures as inhibitors of the skel-etal sarcoplasmic reticulum Ca2� pump. J. Bionerg. Biomembr.32, 133–142.

Steinmetz, L.M., Scharfe, C., Deutschbauer, A.M., Mokranjac, D.,Herman, Z.S., Jones, T., Chu, A.M., Giaever, G., Prokisch, H., Oefner,P.J., and Davis, R.W. (2002). Systematic screen for human diseasegenes in yeast. Nat. Genet. 31, 400–404.

Stocklein, W., and Piepersberg, W. (1980). Binding of cycloheximideto ribosomes from wild-type and mutant strains of Saccharomycescerevisiae. Antimicrob. Agents Chemother. 18, 863–867.

Sugimoto, M., Kuo, M.L., Roussel, M.F., and Sherr, C.J. (2003). Nu-cleolar Arf tumor suppressor inhibits ribosomal RNA processing.Mol. Cell 11, 415–424.

Sun, C., and Woolford, J.L., Jr. (1994). The yeast NOP4 gene productis an essential nucleolar protein required for pre-rRNA processingand accumulation of 60S ribosomal subunits. EMBO J. 13, 3127–3135.

Tanayama, S., Nakai, Y., Fujita, T., Sukuoki, Z., Imashiro, Y., andMasuda, K. (1974). Biotransformation of molsidomine (N-ethoxycar-bonyl-3-morpholinosyndonimine), a new antianginal agent in rats.Xenobiotica 4, 175–191.

Truan, G., Epinat, J.C., Rougeulle, C., Cullin, C., and Pompon, D.(1994). Cloning and characterization of a yeast cytochrome b5-encoding gene which suppresses ketoconazole hypersensitivity ina NADPH-P-450 reductase-deficient strain. Gene 142, 123–127.

Tunaru, S., Kero, J., Schaub, A., Wufka, C., Blaukat, A., Pfeffer, K.,and Offermanns, S. (2003). PUMA-G and HM74 are receptors fornicotinic acid and mediate its anti-lipolytic effect. Nat. Med. 9,352–355.

Venema, J., and Tollervey, D. (1999). Ribosome synthesis in Sac-charomyces cerevisiae. Annu. Rev. Genet. 33, 261–311.

Volarevic, S., Stewart, M.J., Ledermann, B., Zilberman, F., Terracci-ano, L., Montini, E., Grompe, M., Kozma, S.C., and Thomas, G.