Uncovering Genetic Relationships using SmallMolecules that Selectively Target Yeast CellCycle Mutants

Michael T. Nehil1, Craig M. Tamble2, DavidJ. Combs3, Douglas R. Kellogg2, R. ScottLokey2,*

1University of California, San Francisco, CA, USA2University of California, Santa Cruz, CA, USA3University of Pennsylvania, PA, USA*Corresponding author: R. Scott Lokey,lokey@chemistry.ucsc.edu

Genetic analysis in budding yeast has shown thatmultiple G1 cyclins and cyclin-dependent kinasescontrol cell cycle entry, polarized growth, andspindle pole duplication. The G1 cyclins Cln1 andCln2 associate with the cyclin-dependent kinaseCdc28 to facilitate cell cycle progression anddevelopment of the cleavage apparatus. We havedeveloped a chemical genetic approach towardthe discovery of compounds that target G1 controlpathways by screening for compounds that selec-tively kill a yeast strain lacking the G1 cyclinsCln1 and Cln2. A class of small molecules wasidentified that is highly toxic toward thecln1Dcln2D double mutant and has relatively littleeffect on wild-type yeast. We call these com-pounds ’clinostatins’ for their selectivity towardthe cln1/2 deletion strain. Clinostatins were usedin a genome-wide chemical synthetic lethalityscreen to identify other genes required for growthin the presence of the drug. Other deletions thatwere sensitive to the drug include members of theprotein kinase C(PKC)-dependent MAP kinasepathway. These results suggest an approach forcombining chemical synthetic lethality and chem-ical genomic screens to uncover novel geneticinteractions that can be applied to other eukaryot-ic pathways of interest.

Key words: cell cycle, chemical genetics, chemical genomics, high-throughput screening, yeast

Received 19 February 2007, revised and accepted for publication 7March 2007

A classic genetic method for elucidating biologic pathways is toscreen for mutants that enhance the effects of a pre-existing muta-tion. Mutations in two different genes are said to be synthetically

lethal if either mutation is viable in a wild-type background, but thecombination of both mutations results in lethality. Genes related bysynthetic lethality tend to operate in similar or redundant pathways(1), and systematic genome-wide synthetic lethality screens haverevealed new relationships between genes and pathways (2).

Late G1 signaling in yeast is dependent on two main cyclin-cyclin-dependent kinase (CDK) pathways. One is defined by the interactionbetween Cln1 and Cln2 with Cdc28. Another pathway involves thecyclins Pcl1 and Pcl2, which interact with the CDK Pho85 (3). Acln1/2Dpcl1/2D null mutant shows a deficiency in bud emergenceand is unviable, while pcl1/2D and cln1/2D cells are viable (4),suggesting that Cln–Cdc28 and Pcl–Pho85 complexes represent atleast partially redundant G1 signaling pathways (Figure 1). Late G1CDK activity stimulates the degradation of the S-phase inhibitorSic1 (5) and is required for the establishment of cell polarity andbud emergence (4).

Extending the concept of synthetic lethality to small molecules, acompound that exhibits selective toxicity to a particular deletionmutant is likely to target the same pathway, or a pathway that isparallel to or linked by epistasis to the deleted gene. The conceptof 'chemical epistasis' has been used to study the effects of drugson specific yeast deletion mutants in the context of DNA repair andcheckpoint processes (2,6–10), and chemical modifier screens havebeen performed in yeast leading to small molecule enhancers andsuppressors of the drugs FK506 (11) and rapamycin (12). Here, wedemonstrate the application of chemical screening toward the studythe cln branch of G1 cell cycle signaling in yeast. Our results pointto the potential generality of targeting a pathway of interest usingchemical epistasis and chemical genomics screening to uncovernew genetic relationships.

Results and Discussion

We employed a high-throughput screening approach to identifycompounds that were selectively toxic to either a cln1Dcln2D dele-tion strain (cln1/2D) or a pcl1Dpcl2D deletion strain (pcl1/2D).About 3099 compounds from the National Cancer Institute (NCI)diversity, mechanistic, and natural product libraries were screenedfor compounds that inhibited growth in liquid culture of eachmutant strain relative to wild type. Compounds were screened in384-well plates, and the OD544 for each well was measured as afunction of time using a standard plate reader. Although several

258

Chem Biol Drug Des 2007; 69: 258–264

Research Letter

ª 2007 The AuthorsJournal compilation ª 2007 Blackwell Munksgaard

doi: 10.1111/j.1747-0285.2007.00496.x

compounds showed moderate selective toxicity, only one compound(1, Figure 2) exhibited a >10-fold inhibition of mutant growth relat-ive to wild type. Compound 1 inhibited growth of the cln1/2D yeastmutant with an IC50 of 0.7 lM (Figure 2) and exhibited only slighttoxicity to both the pcl1/2D and wild-type strains (IC50 >50 lM).The sensitivity of cln1/2D appeared to be specific to compound 1

and not due to a general sensitivity of cln1/2D to toxic compounds,as other known cytotoxic compounds (wortmanin, benomyl, andrapamycin) were found to be equally growth inhibitory to cln1/2D,pcl1/2D, and wild type (data not shown).

To monitor the lethal phenotype in the presence of compound, cln1/2D cells were grown in the presence of 1 at 20 lM on agar plates.After 24 h, the cells arrested as microcolonies of about 4–20 cells.These cells were dramatically enlarged over untreated cln1/2D cells,and all cells arrested in an unbudded state (Figure 3). Many cellsunderwent lysis after becoming enlarged. This phenotype was distinctfrom observations in untreated cln1/2D cells, where a small popula-tion of cells delayed in G1 and became slightly enlarged, but eventu-ally progressed in the cell cycle. No obvious phenotype was observedat the same concentration of 1 for either wild-type or pcl1/2D cells.

Compound 1 is a derivative of naphthazarin (5,8-dihydroxy-1,4-naph-thoquinone) with a 2-pyridylpiperazine group linked to the naphtho-quinone ring. To probe the structural basis for its biologicselectivity, a series of derivatives was synthesized in which the pyr-idine ring was substituted with other aryl substituents (Figure 4).Derivatives in which the pyridyl group was replaced with either apyrimidine (2) or a trifluorylmethylpyridine ring (3) were nearlyequal in potency and selectivity as the original compound 1. Place-ment of a 4-cyano group on the pyridine ring provided 4, a com-pound with significantly increased potency against the cln1/2Dcells (IC50 ¼ 150 nM). Like 1–3, compound 4 was highly selectivefor the cln1/2D deletion mutant and had no discernable effect oneither wild type or pcl1/2D cells up to 50 lM. Remarkably, com-pound 5, in which the pyridine ring was replaced with a benzenering, was approximately 50-fold less potent than 1 against thecln1/2D mutant. As the inactive phenyl derivative 5 differed from

the original compound 1 by only one nitrogen atom, compound 5

was used as a negative control in subsequent experiments. We callthis new class of compounds 'clinostatins' for their ability to selec-tively kill cln yeast deletion mutants.

In order to further characterize 1 with respect to its targeted path-way(s) in yeast, a chemical genomics screen was performed tomeasure the effect of 1 on growth rates of 4819 viable haploiddeletion mutants (Figure 5). At 20 lM, 1 completely inhibited thegrowth of cln1/2D cells but had negligible effect on wild type orpcl1/2D cells (Figure 4). Compound 5 had little effect on thegrowth of cln1/2D cells at 20 lM and was therefore used as thenegative control. Thus, cell growth was measured as a function oftime for each deletion mutant in the presence of 1 or 5, and theresulting ratios were normalized to the sensitivity of the cln1/2Ddeletion mutant (Figure 5A). Remarkably, only two strains, slt2Dand ume6D, were as sensitive to 1 as cln1/2D. Slt2 is a proteinkinase C (Pkc1)-regulated MAP kinase that is implicated in stressresponse, maintenance of cell integrity, nutrient signaling, and cellcycle control (13). Ume6 is a key transcription factor that regulatesearly meiotic genes and is a negative regulator of unscheduled mei-osis during vegetative growth (14). We confirmed the sensitivitiesof slt2D and ume6D to compound 1 individually using spot assays(Figure 5B).

In addition to cln1/2D, slt2D, and ume6D, among the top five mostsensitive strains to 1 was bck1D. Bck1 is the MAP kinase kinase kin-ase (MAPKKK) that is upstream of Slt2 in the arm of the yeast MAPkinase cascade that is dependent on Pkc1. As, like pcl1/2D, 1 is syn-thetically lethal with the cln1/2D double mutant, we hypothesizedthat 1 might act on components of the Pcl1/2-Pho85 pathway thatcommunicate with the MAP kinase cascade. To assess this hypothesisgenetically, we created the triple null mutant pcl1Dpcl2Dslt2D. Thepcl1Dpcl2Dslt2D triple mutant exhibited a synthetic growth defectand was unviable at 37 �C (Figure 5C), with gross defects in cytokin-esis including widened bud necks and a failure to separate at cytokin-esis (Figure 5D), supporting a role for the MAP kinase cascade duringbud neck formation that is at least partially functionally redundantwith Pcl1/2 signaling. Neither ume6 nor bck1 showed a syntheticinteraction with pcl1/2 (data not shown).

A PubChem search of compound 1 shows reported activity inthe NCI Yeast Anticancer Drug Screen (http://dtp.nci.nih.gov/ya-cds/index.html) and is cytotoxic to human breast and melanomacancer cell lines in the low micromolar range. The yeast strainsthat were sensitive to 1 in the NCI screen are DNA damage-related mutants, including the double mutants mlh1Drad18D andsgs1Dmgt1D, and single mutants mec2-1D and rad50D. Theabsence of rad50 as a hit in our genome-wide sensitivity screenmay be due to our use of 5 as a control, in which only thosedeletions related to the observed cln1/2 sensitivity would beexpected to score as sensitive. Nonetheless, the suggested con-nection between 1 and DNA repair pathways may be real. Cellintegrity genes slt2 and bck1 are synthetic lethal with rad50, aswell as the PAK family kinase cla4, which is involved in septinring assembly. The identification of the molecular target(s) of 1

in yeast may be useful in the subsequent identification of itsrelevant target(s) in human cancer cells.

A

B

Figure 1: (A) Parallel G1-S cyclin-CDK pathways in yeast. (B)Concept of chemical synthetic lethality screening, in which com-pounds that selectively kill a cln1,2D deletion mutant might act bytargeting the parallel pathway via Pcl1/2-Pho85.

Chemical Genomic Screening in Yeast Cell Cycle Mutants

Chem Biol Drug Des 2007; 69: 258–264 259

The G1 arrest induced by compound 1 supports the hypothesis thatthe compound is affecting pathways prior to bud emergence. Thegenomic evidence points to a compound target that is functionallyredundant with components of the Pkc1-dependent MAP kinase cas-cade. In an slt2 (MAPKKK) or bck1 (MAPK) deletion background, thecompound showed a significant toxic effect. To extrapolate fromthe hypothesis that 1 targets a PCL1/2-dependent pathway, wereasoned that we should see a genetic interaction between PCL1/2and the most sensitive strains from the genomic analysis. Thepcl1Dpcl2Dslt2D deletion displayed a significant synthetic growthdefect, having gross abnormalities in cytokinesis and disorganizedbudding at 37�, while neither of the individual mutants, pcl1/2D orslt2D, exhibited a growth defect at 37�. These results support thepreviously established link between the PKC-dependent MAP kinase

pathway and the G1 cyclins. For example, slt2 mutants exhibitdefects in polarized cell growth (15) and augment the cell divisiondefect of a partially inactivated cdc28 allele, and both slt2 andbck1 were shown to be synthetically lethal with pho85 (16). In addi-tion, overexpression of PCL1 and PCL2 (but not CLN1 or CLN2) sup-presses some of the osmolytic sensitivity of an slt2 mutant strain(17). However, despite the observation that bck1 and ume6 werehighly sensitive to 1, neither gene showed a synthetic genetic inter-action with pcl1/2. This suggests that the target of 1 is not func-tionally redundant with Pcl1/2 and may operate downstream of thePcl cyclins or in a parallel pathway.

Given that pho85, the Pcl1/2-dependent CDK, is synthetically lethalwith slt2 and bck1, as well as the meiotic transcription factor ume6,

A

BE

C

D

Figure 2: (A) Structure of compound 1. (B–D) Yeast growth curves in the presence of 1. Yeast strains wild type (B); pcl1 ⁄ 2D (C); cln1 ⁄ 2D(D), were treated with different concentrations of 1 and growth was measured as a function of absorbance at 544 nm. (E) Spot assays ofeach yeast strain onto agar containing 1 at the indicated concentrations.

Nehil et al.

260 Chem Biol Drug Des 2007; 69: 258–264

we hypothesized that the clinostatins may target Pho85 directly. Inhi-bition of Pho85 activity by small molecules causes nuclear accumula-tion of a Pho4-GFP fusion protein (18), and we used this assay toask whether addition of clinostatins affect Pho85 activity in vivo.Clinostatins had no effect on Pho4-GFP localization (data not shown),suggesting that clinostatins do not target Pho85 in yeast. We arecurrently working to elucidate the target(s) of clinostatins using acombination of biochemical and genetic approaches.

Conclusions and Future Directions

The combined use of chemical synthetic lethal screening followedby chemical genomic profiling of hit compounds allows for the iden-tification of new genetic relationships even without prior knowledgeof the small molecule's biologic target(s). Although the genetic con-nection between the Pkc1-MAP kinase pathway and G1 cyclins wasknown previously, this study provides a proof-of-concept that suchconnections can be made independently of the knowledge of acompound's molecular target.

The ability to monitor lethal phenotypes as a function of time anddosage presents distinct advantages over classical genetic syntheticlethal combinations, which often provide only binary results. Thetemporal and dosage control of small molecules allow phenotypicdata to be added to the classical synthetic genetic interaction map,providing an extra layer of information in the analysis of pathwayrelationships. As the map of known synthetic lethal interactionscontinues to expand, linking chemical sensitivities to genetic inter-action data offers the possibility of 'triangulating' onto small mole-cule targets (7). In addition, clustering drugs according to theirgenomic sensitivity profiles in yeast can provide valuable insightsinto their mechanisms of action (19). The screening of larger chem-ical libraries against other yeast deletion mutants is likely touncover new chemical-genetic relationships relevant to many cellu-lar processes of interest.

Materials and Methods

Yeast strains and materialsThe following yeast strains were used in this study:

DK186: MATa his3-11 leu2-3, 112 trp1-1 ura3-52 ade2-1 can1-100GAL+ bar1

KA61: MATa his3-11 leu2-3, 112 trp1-1 ura3-52 ade2-1 can1-100GAL+ bar1 cln1::TRP1 cln2::LEU2

DK573: MATa his3-11 leu2-3, 112 trp1-1 ura3-52 ade2-1 can1-100GAL+ bar1 pcl1::NAT pcl2::KAN

Additional deletion strains were obtained from the Open BiosystemsMATa haploid deletion collection, and the Pho4-GFP strain wasobtained from Invitrogen. The strain SL010 (pcl1D pcl2D slt2D) wasgenerated by using standard genetic procedures to cross strainsDK573 and the MATa slt2 haploid deletion strain obtained from OpenBiosystems. After selection for diploids, sporulation was induced andtetrads were dissected. The triple mutant was identified by scoringtetrads for the nutritional markers associated with each deletion.

Chemical screeningThe NCI Diversity, Mechanistic, and Natural Products libraries wereobtained from the Drug Synthesis and Chemistry Branch, Develop-mental Therapeutics Program, Division of Cancer Treatment andDiagnosis, NCI. Compounds were reformatted into 384-well platesat 1 or 10 mM in DMSO prior to screening. The initial screen usedyeast diluted into YPD + adenine liquid media to OD544nm <0.01.Compound was added using the V&P Scientific Pin Tool RobotModel VP 903B (pin tool VP384FP3S100) to a final concentration ofapproximately 5 or 50 lM in 384-well polystyrene plates. Growthlevels were analyzed after 24 h for wild type and 36 h for cln1/2D,pcl1/2D using a plate reader (Wallac Victor2) measuring absorbanceat 544 nm. Hits were considered as those with a differential read-ing between wild type and the mutant of interest of >10-fold.

Spot assaysCompound at 10 mM in DMSO was infused into agar media to givea final concentration of 20 lM. A series of dilutions were per-formed on the yeast strains and plated at 2 lL per spot. Plateswere incubated at 30 �C for 36 h.

Phenotypic analysisYeast strains were grown on agar containing 20 lM compound orDMSO as a negative control at 30 �C. Thin agar sections wereremoved and placed on glass slides with coverslips. Cell phenotypeswere observed and photographed at various intervals over 24 h.

Chemical genomic screenCompound in DMSO was added to YPD to a final concentrationof 20 lM. This solution was then added at 80 lL per well into96-well plates. Yeast strains were transferred into compound/YPDplates using a 96-well metal pin array. Deletion plates weretransferred initially into 50 lL YPD in order to provide the appro-priate yeast dilution, and then immediately from this plate theyeast were pin-transferred into the assay plate containing com-

0 h

24 h

1 (20 µM) DMSO

Figure 3: Images (DIC) of microcolonies of cln1,2D cells grownin the presence of 1 at 20 lM on agar.

Chemical Genomic Screening in Yeast Cell Cycle Mutants

Chem Biol Drug Des 2007; 69: 258–264 261

A B

Figure 4: (A) Concentration dependence data for compounds 1–5 against cln1,2D (solid bars), wild type (hashed bars), and pcl1,2D (yellowbars) yeast strains. Yeast growth was measured after 24 h as a function of absorbance at 544 nm. (B) Spot assays comparing the effects ofcompounds 1 and 5 against the three yeast strains.

Nehil et al.

262 Chem Biol Drug Des 2007; 69: 258–264

pound. The plates were agitated in the plate reader and absorb-ance was measured at 544 nm after 24–48 h of growth. Mul-tiple time-points were used in order to account for slow-growingstrains. In each set of experiments, control plates containing wildtype and cln1/2D yeast were run in parallel. Growth inhibitionof different strains was normalized to these control plates, i.e.the ratio of the absorbance of the cln1/2D strain in the pres-ence of 1 versus 5 was set equal to 100% inhibition, andgrowth ratios of all deletion strains were plotted as a fractionof this. The top 50 most sensitive strains were verified byrepeating the drug sensitivity screen on these strains.

Acknowledgments

This work was supported by the National Institutes of Health (1R01 CA104569-03) and California Institute for Quantitative Biomedi-cal Research (QB3). We thank William Sullivan for insightful discus-sions and Patrick Cleveland (V&P Scientific) for technical support.We also thank the Developmental Therapeutics Program at theNational Cancer Institute for supplying libraries and reagents.

References

1. Hartman J.L.T., Garvik B., Hartwell L. (2001) Principles for thebuffering of genetic variation. Science;291:1001–1004.

2. Tong A.H., Lesage G., Bader G.D., Ding H., Xu H., Xin X., YoungJ. et al. (2004) Global mapping of the yeast genetic interactionnetwork. Science;303:808–813.

3. Lenburg M.E., O'Shea E.K. (2001) Genetic evidence for a mor-phogenetic function of the Saccharomyces cerevisiae Pho85cyclin-dependent kinase. Genetics;157:39–51.

4. Moffat J., Andrews B. (2004) Late-G1 cyclin-CDK activity isessential for control of cell morphogenesis in budding yeast.Nat Cell Biol;6:59–66.

5. Nash P., Tang X., Orlicky S., Chen Q., Gertler F.B., MendenhallM.D., Sicheri F., Pawson T., Tyers M. (2001) Multisite phosphory-lation of a CDK inhibitor sets a threshold for the onset of DNAreplication. Nature;414:514–521.

6. Simon J.A., Szankasi P., Nguyen D.K., Ludlow C., Dunstan H.M.,Roberts C.J., Jensen E.L., Hartwell L.H., Friend S.H. (2000) Dif-ferential toxicities of anticancer agents among DNA repair and

A B

C D

Figure 5: (A) Results of chemical genomics experiment. Yeast deletion mutants that showed significant sensitivity to 1 relative to controlcompound 5 are shown in red. (B) Spot assays for two strains, ume6D and slt2D identified in the chemical genomics experiment as beingsensitive to 1 relative to 5. Compounds were added to the plates at 20 lM. (C) Temperature sensitivity of pcl1/2Dslt2D triple mutant. (D)pcl1/2Dslt2D triple mutant shows widened and diffuse septin distribution at bud necks. Septins were visualized using anti-cdc11 antibody(Santa Cruz Biotech) conjugated to Alexa-fluor.

Chemical Genomic Screening in Yeast Cell Cycle Mutants

Chem Biol Drug Des 2007; 69: 258–264 263

checkpoint mutants of Saccharomyces cerevisiae. CancerRes;60:328–333.

7. Parsons A.B., Brost R.L., Ding H., Li Z., Zhang C., Sheikh B.,Brown G.W., Kane P.M., Hughes T.R., Boone C. (2004) Integrationof chemical-genetic and genetic interaction data links bioactivecompounds to cellular target pathways. Nat Biotechnol;22:62–69.

8. Simon J.A., Yen T.J. (2003) Novel approaches to screen for anti-cancer drugs using Saccharomyces cerevisiae. Methods MolBiol;223:555–576.

9. Dunstan H.M., Ludlow C., Goehle S., Cronk M., Szankasi P.,Evans D.R., Simon J.A., Lamb J.R. (2002) Cell-based assays foridentification of novel double-strand break-inducing agents.J Natl Cancer Inst;94:88–94.

10. Lee W., St Onge R.P., Proctor M., Flaherty P., Jordan M.I., ArkinA.P., Davis R.W., Nislow C., Giaever G. (2005) Genome-widerequirements for resistance to functionally distinct DNA-dam-aging agents. PLoS Genet;1:e24.

11. Butcher R.A., Schreiber S.L. (2003) A small molecule suppressorof FK506 that targets the mitochondria and modulates ionic bal-ance in Saccharomyces cerevisiae. Chem Biol;10:521–531.

12. Huang J., Zhu H., Haggarty S.J., Spring D.R., Hwang H., Jin F.,Snyder M., Schreiber S.L. (2004) Finding new components of thetarget of rapamycin (TOR) signaling network through chemicalgenetics and proteome chips. Proc Natl Acad Sci USA;101:16594–16599.

13. Heinisch J.J., Lorberg A., Schmitz H.P., Jacoby J.J. (1999) Theprotein kinase C-mediated MAP kinase pathway involved in themaintenance of cellular integrity in Saccharomyces cerevisiae.Mol Microbiol;32:671–680.

14. Kassir Y., Adir N., Boger-Nadjar E., Raviv N.G., Rubin-Bejerano I.,Sagee S., Shenhar G. (2003) Transcriptional regulation of mei-osis in budding yeast. Int Rev Cytol;224:111–171.

15. Mazzoni C., Zarov P., Rambourg A., Mann C. (1993) The SLT2(MPK1) MAP kinase homolog is involved in polarized cell growthin Saccharomyces cerevisiae. J Cell Biol;123:1821–1833.

16. Huang D., Moffat J., Andrews B. (2002) Dissection of a complexphenotype by functional genomics reveals roles for the yeastcyclin-dependent protein kinase Pho85 in stress adaptation andcell integrity. Mol Cell Biol;22:5076–5088.

17. Madden K., Sheu Y.J., Baetz K., Andrews B., Snyder M. (1997)SBF cell cycle regulator as a target of the yeast PKC-MAPkinase pathway. Science;275:1781–1784.

18. Kung C., Kenski D.M., Krukenberg K., Madhani H.D., ShokatK.M. (2006) Selective kinase inhibition by exploiting differentialpathway sensitivity. Chem Biol;13:399–407.

19. Parsons A.B., Lopez A., Givoni I.E., Williams D.E., Gray C.A.,Porter J., Chua G. et al. (2006) Exploring the mode-of-action ofbioactive compounds by chemical-genetic profiling in yeast.Cell;126:611–625.

Supplementary Material

The following supplementary material is available for this article:

General procedure for synthesis of compounds 1–6. 1H- and 13C-NMR spectral data and ESI mass spectral data on compounds 1–6.

This material is available as part of the online article from: http://www.blackwell-synergy.com/doi/abs/10.1111/j.1747-0285.2007.00496.x (This link will take you to the article abstract).

Please note: Blackwell Publishing are not responsible for the con-tent or functionality of any supplementary materials supplied by theauthors. Any queries (other than missing material) should be direc-ted to the corresponding author for the article.

Nehil et al.

264 Chem Biol Drug Des 2007; 69: 258–264