Chemical genetic profiling of the microtubule-targeting agent peloruside A in budding yeast Saccharomyces cerevisiae

Gene 497 (2012) 140–146

Contents lists available at SciVerse ScienceDirect

Gene journal homepage: www.elsevier.com/locate/gene

Chemical genetic profiling of the microtubule-targeting agent peloruside A in budding yeast Saccharomyces cerevisiae Anja Wilmes a, b, 1, Reem Hanna a, b, 1, Rosemary W. Heathcott a, b, Peter T. Northcote a, c, Paul H. Atkinson a, b, David S. Bellows a, b, John H. Miller a, b,⁎ a b c

Centre for Biodiscovery, Victoria University of Wellington, PO Box 600, Wellington 6140, New Zealand School of Biological Sciences, Victoria University of Wellington, PO Box 600, Wellington 6140, New Zealand School of Chemical and Physical Sciences, Victoria University of Wellington, PO Box 600, Wellington 6140, New Zealand

a r t i c l e

i n f o

Article history: Accepted 25 January 2012 Available online 4 February 2012 Keywords: Peloruside Paclitaxel Chemical genetics Yeast MAD2 Saccharomyces cerevisiae

a b s t r a c t Peloruside A, a microtubule-stabilising agent from a New Zealand marine sponge, inhibits mammalian cell division by a similar mechanism to that of the anticancer drug paclitaxel. Wild type budding yeast Saccharomyces cerevisiae (haploid strain BY4741) showed growth sensitivity to peloruside A with an IC50 of 35 μM. Sensitivity was increased in a mad2Δ (Mitotic Arrest Deficient 2) deletion mutant (IC50 = 19 μM). Mad2 is a component of the spindle-assembly checkpoint complex that delays the onset of anaphase in cells with defects in mitotic spindle assembly. Haploid mad2Δ cells were much less sensitive to paclitaxel than to peloruside A, possibly because the peloruside binding site on yeast tubulin is more similar to mammalian tubulin than the taxoid site where paclitaxel binds. In order to obtain information on the primary and secondary targets of peloruside A in yeast, a microarray analysis of yeast heterozygous and homozygous deletion mutant sets was carried out. Haploinsufficiency profiling (HIP) failed to provide hits that could be validated, but homozygous profiling (HOP) generated twelve validated genes that interact with peloruside A in cells. Five of these were particularly significant: RTS1, SAC1, MAD1, MAD2, and LSM1. In addition to its known target tubulin, based on these microarray ‘hits’, peloruside A was seen to interact genetically with other cell proteins involved in the cell cycle, mitosis, RNA splicing, and membrane trafficking. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Determining the mode of action of new compounds is an important goal in drug development. Chemical genetics profiling in yeast provides a data set that allows chemical compounds and yeast genes to be organised into functionally relevant interacting groups or networks (Parsons et al., 2004, 2006). The availability of comprehensive deletion strain collections and gene libraries make yeast particularly well suited to large-scale genetic screens. Yeast deletion sets are available in which all non-essential genes have been knocked out by homologous recombination and insertion of an antibiotic resistance cassette. The cassette is flanked by two 20mer DNA sequence tags or “barcodes” (denoted as Up (UP) and Down (DN) tags), each unique to the deletion strain, allowing identification of the individual strains from a pool by PCR with common primers and hybridisation to a microarray of all tags (Giaever et al.,

Abbreviations: HIP, haploinsufficiency profiling; HOP, homozygous deletion profiling; MDR, multidrug resistance; PDR, pleiotropic drug resistance; PelA, peloruside A ⁎ Corresponding author at: School of Biological Sciences, Victoria University of Wellington, P.O. Box 600, Wellington 6140, New Zealand. Tel.: + 64 4 463 6082; fax: + 64 4 463 5331. E-mail address: [email protected] (J.H. Miller). 1 A. Wilmes and R. Hanna contributed equally to this research. 0378-1119/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.gene.2012.01.072

2002; Pierce et al., 2007; Winzeler et al., 1999). Microarray analysis can be used to determine which genes interact with a chemical of unknown function. One such microarray screen involves haploinsufficiency profiling (HIP) in which a heterozygous deletion set is used that has only one intact copy of each gene, with the other copy deleted. In the presence of a chemical that specifically inhibits the product of the heterozygous gene locus, growth is inhibited. A reduced copy number of a drug's target gene from two to one sensitises a diploid cell to the drug. The HIP microarray screen thus identifies the target of the drug (Giaever et al., 1999; Lum et al., 2004). HIP results provide comprehensive understanding of the genome-wide cellular response to chemical compounds with unknown mechanisms of action (Giaever et al., 2004; Lum et al., 2004). Homozygous deletion profiling (HOP) involves use of yeast haploid deletion sets in which no copy of a gene is present at a given locus or both copies of a diploid strain are deleted. Since no functional gene is present, only non-essential genes can be tested for interactions with the drug of interest. Following exposure to a drug, microarray HOP screens reveal functionally related or connected genes (‘friends of the target’) that buffer the cell against the cytotoxic or cytostatic effects of the drug. Yeast have an effective set of drug efflux pumps that provide resistance to the deleterious effects of drugs. The efflux systems establish a pleiotropic drug resistance (PDR) phenotype similar to the multidrug resistance (MDR) phenotype of mammalian cells (Jungwirt

A. Wilmes et al. / Gene 497 (2012) 140–146

and Kuchler, 2006). Typically, drugs that are effective in mammalian cells at nanomolar concentrations often are 100- to 1000-fold less potent in yeast, and this difference may be partly due to PDR mechanisms as well as differences in target affinity. Some yeast deletion sets are available that have the main PDR genes (PDR1, PDR3, PDR5) deleted, and these pump-deficient mutants often show increased sensitivity to test drugs. Peloruside A (PelA) was first described by West et al. (2000) after isolation and purification from a New Zealand marine sponge, and its mode of action as a microtubule-stabilising agent was determined a few years later by Hood et al. (2002). The classic microtubulestabilising agent, paclitaxel, is used clinically as an anticancer agent. Yeast has three tubulin isotypes. TUB1 is the major α-tubulin gene of yeast and TUB2 encodes β-tubulin. Both are essential for viability. The third yeast tubulin isotype gene, also α-tubulin (TUB3), is expressed at lower levels and is not essential for cell viability. Yeast tubulin shows about 75% homology to human brain tubulin (Barnes et al., 1992). Bode et al. (2002) showed that the tubulin targeting agents paclitaxel, colchicine, vinblastine, podophyllotoxin and crytophycin had no effect on yeast tubulin assembly in vitro; however, epothilone A and B were able to promote assembly of purified tubulin but concentrations up to 150 μM had little effect on yeast proliferation in vivo. By selective mutation of five amino acids (A19, T23, G26, N227, and Y270) in the yeast β-tubulin gene (TUB2) at the taxoid binding site, a functional paclitaxel binding site was constructed by Gupta et al. (2003), and this mutant tubulin could be polymerised by paclitaxel in vitro. Sensitivity of paclitaxel, however, in yeast in vivo required seven inactivations of ABC transporters in addition to these five mutations in β-tubulin (Foland et al., 2005). Amino acids in the proposed binding site for PelA (Gaitanos et al., 2004; Huzil et al., 2008; Kanakkanthara et al., 2011; Nguyen et al., 2010) are more highly conserved between yeast and humans than the taxoid binding site. We therefore hypothesised that PelA would be able to inhibit yeast growth in cells with wild type TUB2. The aim of the present study was to test whether yeast growth could be inhibited by PelA and to use yeast HIP and HOP microarray analyses to identify secondary targets and interactive networks of PelA in order to better understand its mode of action in cells. Previous studies have demonstrated that yeast can be used effectively as a model organism to study drug interactions in mammalian cells (Giaever et al., 2004; Lum et al., 2004; Menacho-Marquez and Murguia, 2007; Nislow and Giaever, 2007; Parsons et al., 2004, 2006). 2. Materials and methods 2.1. Materials Peloruside A (PelA) was purified from the marine sponge Mycale hentscheli as previously described (West et al., 2000), dissolved in DMSO at 10 mM, and stored at −80 °C. As the supply of natural PelA was strictly limited, stocks used in this study were conserved as much as possible, and this impacted on the number of replicates that could be carried out. Paclitaxel was purchased from LC Labs (Woburn, MA) and stored at − 80 °C in DMSO at 10 mM. 2.2. PelA inhibition of growth A concentration–response relationship for PelA in the haploid wild type yeast strain (BY4741) of Saccharomyces cerevisiae was performed to determine the optimal concentration of PelA for the microarray. A concentration was chosen that only affected growth to a minimal extent (IC10–IC20). In addition, individual growth inhibition studies were initially carried out with paclitaxel on several deletion mutants that had been previously shown to have lethal interactions with either α- or β-tubulin in yeast, including mad2Δ, mad3Δ, gim1Δ, gim4Δ, cin1Δ, cin4Δ, pac2Δ, mcm21Δ, and bem2Δ. The three most sensitive strains were then tested with PelA. Mad2 and Mad3

141

are part of the spindle assembly complex, Gim1 and Gim4 are part of the prochaperone prefolding complex, Cin1 and Cin4 are involved in tubulin folding, Pac2 is involved in tubulin heterodimer formation, Mcm21 is involved in minichromosome maintenance, and Bem2 is involved in bud emergence. 2.3. Yeast deletion pool growth for microarray experiments To make a pool of homozygous yeast deletion strains, the library (Open Biosystems) was grown on YPD agar (2% peptone, 1% yeast extract, 2% glucose, 0.012% adenine hemisulphate) containing 200 mg/ L G418 antibiotic (Geneticin, Gibco, Invitrogen). Colonies were scraped from the plates, pooled and aliquoted in YPD plus 15% glycerol and stored at −80 °C at a cell concentration of 1 × 108 cells/0.2 mL. The heterozygous knockout pool (Invitrogen) was stored in YPD/15% glycerol at a cell concentration of 0.5 × 108 cells/0.5 mL. An aliquot of the pool was grown overnight at 30 °C in a Bioline shaker incubator (Edwards Instrument Company, Australia) in 10 mL synthetic complete medium (SC medium) consisting of 6.7 g Bacto-yeast nitrogen, 136 g yeast nitrogen base (without amino acids), 0.8 g monosodium glutamate, and 1.6 g amino acid mixture in a total volume of 800 mL and supplemented with 2% glucose, 25 mM HEPES, and 0.1% G418 antibiotic. In order to have at least 1000 cells in the mixed suspension for each mutant, 5 × 10 6 cells were diluted into 10 mL SC medium containing 2% glucose and 25 mM HEPES in the presence of 10 μM PelA or 0.1% DMSO (control). After 15 h (~10 generations), cells were re-diluted to 5 × 106 cells/10 mL and re-treated with PelA or DMSO for an additional 15 h (~20 generations total). 2.4. Genomic DNA purification Genomic DNA was purified from a 1.5 mL aliquot of the treated and control cultures using a Master Pure™ yeast DNA purification kit (Epicentre Biotechnologies, Global Science & Technology, NZ), following the manufacturer's instructions, followed by RNA degradation by addition of 1 μL of RNase A (5 μg/μL) and incubation at 37 °C for 30 min. The DNA was purified by phenol–chloroform extraction, ethanol precipitated, and dissolved in 35 μL TE (10 mM Tris–HCl, 1 mM EDTA, pH 8.0). DNA was quantified using Hoechst 33258 dye (1 mg/mL) (DNA Quantification kit DNA-QF, Sigma-Aldrich) following the manufacturer's instructions. Fluorescence was measured in a SpectraMax Gemini plate reader (Molecular Dynamics, Sunnyvale, CA) at 360 nm excitation and 460 nm emission wavelengths. A typical DNA concentration of 300 ng/5 μL was obtained, and DNA aliquots were adjusted to 25 ng/μL. 2.5. Barcode amplification and Cy3/5 dye labelling by PCR For each sample, Up tags and DN tags were PCR amplified in separate reactions. DMSO-treated controls were amplified with one primer labelled with the fluorescent dye Cy3, and PelA-treated samples were amplified with a Cy5 labelled primer. The UP tag was amplified using primers Up1 (5′-GATGTCCACGAGGTCTCT) and Up2-Cy5 or Cy3 (5′GTCGACCTGCAGCGTACG). The DN tag was amplified by primers DN1 (5′-CGGTGTCGGTCTCGTAG) and DN2-Cy3 or Cy3 (5′-CGAGCTCGAATTCATCGAT). PCR Master Mix was prepared in a DNA-free laminar flow hood. The composition of the PCR Master Mix was (in a total of 51.2 μL): 10× platinum taq buffer (6 μL, 1× final concentration), 50 mM MgCl2 (1.8 μL, 1.5 mM final), 10 mM each of dNTPs (1.2 μL, 0.2 mM final), 5 U/μL platinum taq (0.2 μL, 1 U final), and distilled water (42 μL). In PCR tubes, the 51.2 μL of PCR Master Mix was mixed with 2.4 μL labelled primer (25 μM) and 2.4 μL unlabelled primer (25 μM). Final concentrations of primers were 1 μM. To all reactions, 4 μL of 25 ng/μL DNA was added, except for negative controls in which dH2O was added instead. The final volume of each PCR reaction was 60 μL. PCR was carried out on a T-Gradient PCR machine (Biometra®,

142

A. Wilmes et al. / Gene 497 (2012) 140–146

Labrepco, Horsham, PA) with PCR settings as follows: 94 °C 3 min, 94 °C 30 s, 50 °C 30 s, 72 °C 30 s for 38 cycles, 72 °C 5 min and pause at 10 °C. Following PCR, 5 μL of each PCR product was electrophoresed on a 4% MetaPhor agarose gel (Cambrex, East Rutherford, NJ) for 1 h at 100 V. The size of the amplified products for UP and DN tags was 56 bp. Tubes containing PCR products were wrapped in foil and stored at −20 °C (if hybridisation was not preformed on the same day). 2.6. Microarray hybridisation Hybridisation mixture was prepared by combining 50 μL of each of the four PCR products with 20 μL blocking solution (12.5 μM U1, 12.5 μM D1, 12.5 μM U2 block [5′-CGTACGCTGCAGGTCGAC], 12.5 μM D2 block [5′-ATCGATGAATTCGAGCTCG]). Two-times hybridisation buffer (2 M NaCl, 20 mM Tris HCl pH7.5, 1% Triton X-100) was filter-sterilised and stored at 4 °C. Dithiothreitol was added to 2 × hybridisation buffer to a final concentration of 1 mM to prevent Cydye oxidation. The hybridisation mixture was denatured at 95 °C for 2 min followed by cooling on ice. A microarray chamber was set up according to the Agilent microarray hybridisation chamber user guide (Agilent, Santa Clara, CA, USA). The chamber was placed into a preheated (42 °C) hybridisation oven (Agilent Technologies, Sheldon manufacturer, Cornelius, OR) at a rotation speed setting of #4 for 4 h. The microarray slide was separated from the gasket slide in 300 mL wash solution 1 (6× SSPE buffer consisting of 3 M NaCl, 0.2 M NaH2PO4, 20 mM EDTA, pH 7.4). The microarray slide was then transferred quickly to 50 mL of wash solution 2 (0.06× SSPE). Both wash solutions were filter-sterilised before use. The slide was then removed from the washing solution and centrifuged at room temperature for 2 min at 500 rpm in a 50 mL Falcon centrifuge tube with Cy-label at the bottom of the tube. The slides were then air-dried overnight and stored in the dark. Microarray slides were then scanned with an Axon GenePix 4000B Microarray Scanner at the Otago Genomics Facility, Biochemistry Department, University of Otago, Dunedin, New Zealand. 2.7. Microarray analysis GenePix Pro 6.0 (Molecular Devices, Sunnyvale, CA) was used to extract fluorescence data from the scanned microarray image. Only spots that had a Cy3 (control) signal that was at least 3× higher than the background were used to calculate a ratio representing PelA-treated sample/control sample (Cy5/Cy3). The ratios were converted to a log2 ratio, normalised so that the average log2 ratio was zero, and corrected for any intensity dependent biases using SNOMAD (Standardisation and Normalisation of MicroArray Data) http://pevsnerlab.kennedykrieger. org/snomadinput.html. UP and DN tags were analysed separately to allow for any PCR-specific biases. Corrected log2 ratios for tags represented by replicate spots were averaged, and z-scores were calculated to rank the corrected log2 ratios for UP and DN tags separately. A zscore of less than −3 (three SD below the average corrected log2 ratio) for either an UP or DN tag was considered to be significant and therefore sensitive to PelA. Deletion clones were identified as hits if a change in either or both tags was found and only if the clone was sensitive to the drug when grown individually as well (validation of a hit). 2.8. Database searches SGD Gene Ontology (GO) Slim Mapper web-based maps were used to annotate/classify a group of genes to more general groups and/or bin them into broad categories, for example, ‘cellular component’ (SGD Stanford). Moreover, to classify ORF genes based on their molecular function, ‘cellular component’ and ‘biological function’ of the ORF was subjected to Slim mapper, a web-based software programme that groups genes into the listed categories. Physical and genetic interactions of genes/proteins were searched with the “interactions” tool on the Saccharomyces Genome database (SGD, www.yeastgenome.org).

2.9. Concentration–response curve for TUB heterozygous genes A concentration–response inhibition curve was constructed for a set of heterozygous deletion mutants (HIP) that included TUB1, TUB2, and TUB3. The BY4743 diploid wild type strain was included as a wild type control. These strains were obtained from a −80 °C glycerol stock and streaked out on 10 cm diameter yeast extract peptone dextrose (YPD) plates (10 g yeast extract, 20 g peptone, 0.12 g adenine, and 20 g agar in a 1 L volume with 50 mL of 40% glucose added after autoclaving) plus 200 μg/mL G418. The plates were incubated at 30 °C (Contherm Digital Incubator, Contherm Scientific, Petone, NZ) for 48 h. Single colonies were inoculated into 2 mL SC medium supplemented with 2% glucose and grown overnight using a Glascol® rotator (Total Lab Systems, Auckland, NZ) at a speed setting of 50%. Cells were diluted the following day to 5 × 10 5 cells per well. Cells were treated with 0.008% SDS to facilitate the uptake of the drug and were grown in a half-log serial dilution of 10 μM PelA for 15 h. After 15 h, cells were re-diluted to 5 × 105 cells per well as before and then grown for an additional 15 h (~20 generations total). The absorbance in the wells was determined, and the % residual growth was calculated by dividing the absorbance value of treated cells by the absorbance value of the control. 2.10. Validation of HOP results To validate the HOP microarray analysis, homozygous deletion mutants were obtained from − 80 °C stocks and streaked onto plates. A single colony from each strain was inoculated into 2 mL SC medium, and the BY4743 diploid wild type strain was included as a positive control. The following day, 5 × 10 5 cells per well were plated on 96well plates. Cells from each strain were treated individually at 30 °C with 10 μM PelA or the appropriate concentration of DMSO for 15 h (~10 generations). After 15 h, cells were re-diluted in order to have a minimum of 1000 cells per well and re-treated with 10 μM PelA or DMSO for an additional 15 h (~ 20 generations total). The absorbance of the wells in the plates was determined and the % residual growth ([experimental absorbance/control absorbance] × 100) calculated. 3. Results 3.1. PelA and paclitaxel effects on yeast growth A concentration–response growth inhibition pre-screen of the haploid wild type strain (BY4741) was performed for PelA and paclitaxel to determine if a genome-wide microarray screen was feasible. Choosing nine potential tubulin-related, mitotic checkpoint deletion mutants (mad2Δ, mad3Δ, gim1Δ, gim4Δ, cin1Δ, cin4Δ, pac2Δ, mdm21Δ, and bem2Δ), yeast growth was assessed in these deletion mutants over a range of paclitaxel concentrations. As expected based on the known mutations in the tubulin binding site of yeast for paclitaxel (Gupta et al., 2003), neither the wild type nor the mutant strains were very sensitive to the drug at concentrations up to 20 μM. Residual growth over 18 h at 20 μM paclitaxel (n=4–5 independent experiments) was 92.4±1.8% of untreated in the wild type, 81.9±5.4% in mad2Δ, 80.3±10.4% in bem2Δ, and 86.3±6.5% in pac2Δ (Fig. 1). Preliminary tests in the other mutants (n=2) also gave little inhibition at 20 μM paclitaxel: gim4Δ 80.5%, cin1Δ 70.8%, cin4Δ 81.7%, mdm21Δ 70.2%, mad3Δ 69.9%, and gim1Δ 64.5%. We next tested the sensitivity of yeast growth to PelA. At 50 μM, residual growth was 30.5 ± 10.0% (n = 4) in the wild type haploid strain BY4741 compared to a single measurement in the wild type diploid strain BY4743 of 15% residual growth. To conserve drug, only three of the mutant mitotic checkpoint haploid strains (mad2Δ, bem2Δ, and pac2Δ) were then tested for growth inhibition with PelA. In the case of these mutants which had showed the most sensitivity to paclitaxel in early testing, the mad2Δ mutant strain proved to

A. Wilmes et al. / Gene 497 (2012) 140–146

143

with z-scores below −3 that were significant in two replicate microarrays (supplementary data Table S1). Attempts to validate the 13 gene hits individually, however, were unsuccessful, possibly because they were either false positives (noise) or because the microarray growth conditions were unable to be replicated exactly in the test tube. SGD Gene Ontology Slim Mapper search results for the 13 HIP gene hits are presented in Table S2, and the functions of the HIP genes are summarised in Table S4. 3.3. HOP microarray

Fig. 1. Growth inhibition with paclitaxel. Residual growth is presented for haploid strains treated with different concentrations of paclitaxel for 18 h compared to untreated controls. Data are the mean ± SEM; n = 5 independent experiments (using the average of duplicate measurements per experiment) for WT (BY4741), mad2Δ, and bem2Δ, n = 4 for pac2Δ, n = 2 for gim4Δ, cin1Δ, cin4Δ, mdm21Δ, and mad3Δ, and n = 1 for gim1Δ.

be the most highly sensitive to PelA (Fig. 2). At 20 μM PelA, residual growth of the mad2Δ mutant strain was 52% of the untreated mad2Δ (Fig. 2) compared to 82% with paclitaxel (Fig. 1). Thus, the haploid yeast mutant was much more sensitive to PelA than to paclitaxel. At 50 μM PelA, growth of mad2Δ was nearly completely inhibited with residual growth only 9.3 ± 3.7% of the untreated mad2Δ (n = 4) (Fig. 2). Using a four parameter logistic curve (SigmaPlot, Systat Software Inc.), the IC50 values for PelA in the BY4741 wild type and mad2Δ mutant strains were calculated to be 35 μM (R 2 = 0.8785) and 19 μM (R 2 = 0.9501), respectively. Thus, our results indicate that PelA has a significant synthetic lethal interaction with MAD2 in yeast. 3.2. HIP microarray At the time this study was undertaken, a library with a mad2Δ deletion background was not available; therefore, we used the complete heterozygous, diploid (BY4743) deletion mutant library for S. cerevisiae to carry out a HIP microarray screen for secondary targets of PelA. PelA was used at a concentration of 10 μM to generate a slight growth inhibition that would then be increased significantly for any deleted genes that had synthetic lethal interactions with PelA. We identified 13 strains

Fig. 2. Growth inhibition with PelA. Residual growth is presented for haploid strains treated with different concentrations of PelA for 18 h compared to untreated controls. Results are presented as the mean ± SEM; n = 4 independent experiments for the WT (BY4741) and mad2Δ mutant strain, and n = 3 for pac2Δ and bem2Δ.

To carry out a HOP microarray screen to identify ‘friends of the target’, the complete homozygous, diploid (BY4743) deletion mutant library for S. cerevisiae was used with 10 μM PelA over 20 generations of growth (n = 2 biological replicates). Significant HOP microarray hits are presented in Table 1, and the SGD Gene Ontology Search results are presented in Table S3, arranged into three categories: molecular function, biological process, and cellular component. In the two HOP replicates there were 68 and 69 genes respectively in which either one or both UP/DN tags were identified as a hit (zb −3), and 33 of these hits were found in common between the two replicates. Seven of these genes (in bold in Table 1) were either component genes of the PDR efflux pumps or multi-drug resistance genes and were not included in the hit list because they would give a non-specific increase in drug potency. This left twenty-six deletion mutant strains that were significantly affected by PelA (z-scores below −3) in both HOP microarrays, and 5 of these strains showed significant z-scores for both the UP and the DN tags (in both microarrays): RTS1, SAC1, MAD1, MAD2, and LSM1.

Table 1 Microarray HOP hits after 10 μM PelA treatment. The BY4743 homozygous deletion mutant set was treated for 20 generations with 10 μM PelA. Thirty-three gene deletion mutants returned z-scores of −3 or below in two experiments. Genes in bold are multi-drug resistance genes and were removed from the final hit list. Based on 2 biological replicates (z-score values shown here were obtained from averaging the results of the two replicates). Strain

Gene name

Ratio (Cy5/Cy3)

z-score

YNL323W YNL280C YCL029C YOR014W YBR189W YKL212W YBL065W YML008C YOR153W YPL187W YGL086W YJL030W YDL115C YNL054W YNL187W YDR264C YPL018W YOR292C YBR156C YBL054W YDR441C YER087W YJL124C YGL012W YGL170C YLR085C YHR025W YOR019W YGR106C YOR161C YMR202W YIL157C YDR017C

LEM3 ERG24 BIK1 RTS1 RPS9B SAC1 – ERG6 PDR5 MF(ALPHA)1 MAD1 MAD2 IWR1 VAC7 – AKR1 CTF19 YOR292C SLI15 – APT2 AIM10 LSM1 ERG4 SPO74 ARP6 THR1 – – PNS1 ERG2 COA1 KCS1

0.003 0.069 0.019 0.037 0.032 0.049 0.017 0.021 0.047 0.068 0.043 0.06 0.116 0.141 0.075 0.099 0.149 0.065 0.089 0.164 0.094 0.176 0.176 0.093 0.165 0.082 0.163 0.102 0.234 0.106 0.098 0.122

− 15.74 − 11.22 − 9.81 − 8.42 − 8.08 − 8.07 − 7.60 − 7.30 − 6.69 − 6.07 − 5.90 − 5.64 − 5.60 − 5.27 − 4.92 − 4.83 − 4.81 − 4.71 − 4.59 − 4.14 − 4.13 − 4.11 − 4.01 − 3.98 − 3.93 − 3.92 − 3.79 − 3.67 − 3.54 − 3.51 − 3.32 − 3.23 − 3.10

144

A. Wilmes et al. / Gene 497 (2012) 140–146

Amongst the 26 gene hits were 5 genes that were involved in mitosis (BIK1, MAD1, MAD2, SLI15, and CTF19), one in meiosis (SPO74), a yeast mating factor gene (MF(Alpha)1), 4 genes involved in RNA and ribosomal proteins (YBL054W, RPS9B, AIM10, and LSM1), 2 genes involved in cell wall shape and maintenance (SAC1 and AKRr1), 3 genes involved in ergosterol biosynthesis (RTS1, ERG6, and ERG24), a gene involved in the mitochondrial membrane (COA1), a gene needed for DNA repair (KCS1), and a gene involved in the vacuolar membrane (VAC7). The functions of the remaining 7 of the 26 genes were unknown. The detailed functions of the HOP gene hits are summarised in Table S4. Twenty-three of the 26 HOP microarray gene hits were tested individually (Table 2), and the 12 validated hits are presented in bold in Table 2. ERG24 and ERG6 were excluded from the validation process because cells lacking these genes exhibit pleiotropic growth phenotypes (Gaber et al., 1989). A similar pleiotropic phenotype effect has been suggested for SLI15 because of effects on kinetochore– chromosome interactions. Just over half (Hardwick and Murray, 1995) of the 23 microarray hits tested individually also showed significant inhibition of growth in test tube culture, thus, confirming their synthetic lethal interaction with PelA. 4. Discussion 4.1. Preliminary tests on PelA and paclitaxel Initial tests on the sensitivity of yeast to the microtubulestabilising drug paclitaxel, as expected (Bode et al., 2002; Gupta et al., 2003), failed to show significant or consistent growth inhibition. Thus, wild type yeast were relatively insensitive to paclitaxel, and the inhibition seen in the mitotic checkpoint mutant strains might have been a result of effects on secondary targets of paclitaxel rather than tubulin interactions, although some binding and stabilisation of yeast tubulin in cells by high concentrations of paclitaxel cannot be ruled out. Paclitaxel is also known to be a good substrate for drug efflux pumps in mammalian cells (Parekh et al., 1997), and the same may be Table 2 Individual validation of HOP-microarray hits. Twenty-three of the 26 homozygous deletion mutant strains identified in the HOP microarry of Table 1 were treated individually in test tubes with 10 μM PelA or DMSO (as control) for either 10 generations (incubation for 15 h) or 20 generations (incubation for 30 h) (n=one experiment). Twelve gene deletion strains (in bold) out of the 23 showed a significant PelA-induced decrease in growth after 20 generations (residual growth of less than 80% of the control was considered significant). Results are ranked according to percentage growth after 20 generations. Gene name

KCS1 MAD1 BIK1 RTS1 LSM1 SAC1 CTF19 MAD2 RPS9B VAC7 YBL054W MF(ALPHA)1 PNS1 YOR019W COA1 APT2 YNL187W YOR292C AKR1 AIM10 IWR1 SPO74 YBL065W

Residual growth

Residual growth

10 gen.

20 gen.

86% 65% 77% 71% 92% 62% 81% 85% 99% 77% 104% 95% 101% 103% 94% 98% 104% 102% 89% 104% 98% 99% 105%

26% 30% 31% 33% 36% 42% 45% 49% 71% 73% 78% 79% 82% 87% 88% 89% 93% 94% 96% 97% 98% 104% 110%

true in yeast. PelA, another microtubule-stabilising agent, was more effective at inhibiting yeast growth, although at concentrations much greater than those required in mammalian cells. The inhibition by PelA was greatly enhanced by deletion of the cell cycle protein Mad2, coded for by one of six non-essential mitotic checkpoint genes (Li and Benezra, 1996). The 16 μM IC50 in the mad2Δ haploid yeast strain is 1000 times greater than the IC50 values measured for PelA in mammalian cell lines (10–30 nM) (Hood et al., 2002).

4.2. Using a HIP screen to identify primary and secondary targets of PelA Although HIP screens in heterozygous yeast should be able to identify direct targets of a drug, the primary and well-known target of PelA, tubulin (Hood et al., 2002), was not identified in our microarray screen. Possible explanations for this are that the microarray was not sensitive enough to pick up the primary target or that the concentration of PelA was not high enough. On the other hand, the effect that PelA has on microtubules, namely stabilisation, may not be functionally equivalent to depletion of tubulin. Tubulin might be easier to identify as a target by HIP with a microtubule-depolymerising drug that functionally mimics tubulin loss of function. Another possible explanation for a lack of a hit on tubulin is that cytoskeletal gene expression is tightly regulated such that a single copy of TUB2 in the heterozygous diploid mutant strain is sufficient for normal tubulin expression levels. Although haploinsufficiency has been demonstrated for many genes and is the basis of the HIP screen, there are clear exceptions to the rule, including cytoskeletal proteins (Deutschbauer et al., 2005). Thus, in TUB2 heterozygous deletion strains, the common assumption that only half the amount of protein is present, may not be true. Our preliminary concentration–response screens confirmed that one copy of the two TUB2 tubulin genes was likely to be sufficient for normal function, since when these strains were treated individually with PelA, their growth inhibition curves were almost superimposed on each other and on wild type (data not presented). It is possible that structural proteins in general such as tubulin and actin are not easily identified as drug targets in HIP screens because one copy of the gene may not be equivalent to half the amount of protein. For α-tubulin, loss of one copy of TUB1, the main α-tubulin gene of yeast, seems to be compensated for by overexpression of TUB3, a second nonessential α-tubulin gene (Schatz et al., 1986). It is unknown whether PelA can directly bind to yeast tubulin, and this is another possible explanation for not identifying tubulin with the HIP screen. As reported previously, paclitaxel can only bind to yeast tubulin after five amino acids in the taxoid binding site have been converted to their human equivalents at those positions (Gupta et al., 2003). Interestingly, epothilone A and B on the other hand are able to bind to yeast tubulin, despite also binding to the taxoid site; however, they are, like paclitaxel, ineffective in preventing yeast growth even when given to intact cells at concentrations as high as 150 μM (Bode et al., 2002). The exact differences between amino acid sequences at the PelA binding sites on yeast and bovine or human tubulin are unknown, although two amino acid mutations that confer resistance to peloruside in mammalian cancer cells, R306H/C and A296T, are conserved between yeast TUB2 and mammalian βI-tubulin (Kanakkanthara et al., 2011). If PelA does not polymerise yeast tubulin, then the identified hits from the HIP microarray are likely to be secondary targets unrelated to tubulin polymerisation. However, other factors such as yeast ABC transporters may be responsible for the high IC50 value for PelA in yeast. The PDR drug efflux system is very efficient in yeast and probably contributes significantly to the high drug concentrations needed for effects in yeast. We showed that the haploid deletion strain mad2Δ has increased sensitivity to PelA. Since mad2Δ was chosen by us originally because of its known synthetic lethal interaction with various tubulin mutants, one could argue that PelA in fact does show a genetic connection with yeast tubulin. Given the lack of validation of any of

A. Wilmes et al. / Gene 497 (2012) 140–146

the hits from the HIP microarray, no direct secondary targets of PelA in yeast were identified. 4.3. Using a HOP screen for gene networks that interact with PelA The HOP microarray screen provides information about genes that interact with a drug targeting pathway. The hits in a HOP microarray are important for the survival of the cell in the presence of the drug (Nislow and Giaever, 2007) and identify functionally related pathways to the drug of interest (‘friends of the target’) but not the target itself. Hence, the HOP screen is a powerful complement to the HIP screen as it allows the identification of genes and pathways required for protection of the cell from inhibition of a particular target. After PelA treatment, 26 genes were identified, 7 of which had unknown function — IWR1, YNL187W, YOR019W, PNS1, YBL065W, APT2, and YOR292C. One of these genes, however, YORO19W, is associated with the cell cycle under GO biological process (Table S3). When these hits were validated individually, 12 of the 23 mutant strains tested were sensitive to the drug — KCS1, MAD1, BIK1, RTS1, LSM1, SAC1, CTF19, MAD2, RPS9B, VAC7, YBL054W, and MF(ALPHA)1, having residual growth values for PelA relative to control ranging from 26% to 79% after 20 generations of growth (Table 2). As expected, the mad2Δ yeast strain was identified in the HOP screen. It is a nonessential gene that encodes a component of the spindle checkpoint (Hardwick and Murray, 1995; Li and Murray, 1991). Mad2 forms a tight complex with another spindle checkpoint protein, Mad1, throughout the cell cycle. The HOP growth inhibition assay confirmed that mad1Δ and mad2Δ strains are quite sensitive to 10 μM PelA, with only 30% and 49% of wild type growth, respectively. A link between PelA and Mad2 in human cells was previously identified in our laboratory (Wilmes et al., 2011). PelA decreased protein levels of c-Myc, a transcription factor that activates the expression of Mad2 and other APC/C inhibitors. In addition aneuploid cells were seen at low concentrations of PelA that might be linked to the effects of Mad2 on the mitotic checkpoint pathway. A decrease of c-Myc and aneuploidy in response to paclitaxel has also been shown (Ikui et al., 2005; Wilmes et al., 2011). BIK1, a microtubule-associated protein gene, showed 31% of control growth, and CTF19, which codes for an outer kinetochore protein, gave 45% residual growth. Bik1 function is required for nuclear fusion, chromosome disjunction, and nuclear segregation during mitosis. Previous studies have shown that Bik1 protein colocalises with tubulin to the spindle pole body and mitotic spindle. Synthetic lethality observed in double mutant strains containing a mutation in the BIK1 gene and in the gene for α- or β-tubulin is consistent with a physical interaction between Bik1 and tubulin (Berlin et al., 1990). SAC1 codes for a transmembrane protein that is involved in protein trafficking, processing, secretion, and cell wall maintenance (Novick et al., 1989). The human homolog to SAC1 is human SAC1 (HSAC1). The remaining genes that were validated individually were involved in vacuolar structure and biogenesis (VAC7 and KCS1), RNA degradation and RNA splicing (LSM1), protein phosphatase activity (RTS1), and ribosome function (RPS9B). Therefore, the HOP microarray results indicate that the functionally related genes in the pathway of the action of PelA are genes that are involved in the regulation of mitosis and cell cycle, protein synthesis, transport, secretion, RNA processing, and steroid biosynthesis. 4.4. Other genome-wide or proteomic screens with microtubule-stabilising agents At present, only one genome-wide knock-down screen has been carried out for a microtubule-stabilising agent. Whitehurst et al. (2007) performed an RNA-mediated interference (RNAi) synthetic lethal screen with paclitaxel in human non-small-cell-lung cancer cells (NCIH1155). This approach in yeast would be more comparable to a HOP microarray screen than a HIP microarray screen, although,

145

unlike in the yeast mutation set which represents 100% loss of function, knock-down by siRNA generally varies from 70 to 95%. In their screen, siRNAs targeting each of the 21,127 human genes were used. Eighty-five “high-confidence hits” were identified, including many proteasome subunit genes (PSMA 6–8, PSMB1, PSMC3 and PSMD1) and three targets encoding proteins involved in microtubule function and dynamics (γ-TURC subunit, α-tubulin, and dynein heavy chain subunits). Other genes involved in post-translational modification and cell adhesion also showed up. There were no direct overlaps in identified hits between the paclitaxel RNAi screen of their study in lung cancer cells and our PelA HIP and HOP screens. In a separate study, Wilmes et al. (2011, 2012) used differential in-gel electrophoresis (DIGE) to look for protein abundance changes in a human promyelocytic leukemic cell line (HL-60). The overlap between our microarray analysis with PelA in yeast and the proteomic screens with PelA (Wilmes et al., 2011) and paclitaxel (Wilmes et al., 2012) was also relatively minor, although some of the proteomic changes, as with the yeast microarray screens, were in tubulinrelated proteins and proteins involved in cell division and mitosis, such as the link between Myc and aneuploidy. 5. Conclusions In the present study in yeast, we showed that PelA was more effective than paclitaxel in inhibiting yeast growth. To provide information on the mode of action of PelA, both HIP and HOP microarray screens were carried out. The HIP screen yielded no validated targets of the drug, but in the HOP screen, 12 validated hits were identified and 5 of these were found in both microarray replicates and by both UP and DN tag PCR amplification: RTS1, SAC1, MAD1, MAD2, and LSM1. Although there was little direct correlation between the genetic interactions identified in yeast and siRNA knock-down (Whitehurst et al., 2007) or proteomic hits (Wilmes et al., 2011, 2012) in mammalian cells, studies of this type can still provide information on the different effects of PelA and paclitaxel in cells and may reveal possible similarities between yeast and mammalian cells as analysis platforms for studies on the mode of action of drugs and the relevance of the yeast model to human disease (Mager and Winderickx, 2005). Acknowledgments The authors thank Cameron Jack, School of Biological Sciences, Victoria University of Wellington for designing the software for calculating the average corrected log2 ratios. This research was supported by grants to J.H.M from the Cancer Society of New Zealand, the Wellington Medical Research Foundation, and Victoria University of Wellington. Appendix A. Supplementary data Supplementary data to this article can be found online at doi:10. 1016/j.gene.2012.01.072. References Barnes, G., Louie, K.A., Botstein, D., 1992. Yeast proteins associated with microtubules in vitro and in vivo. Mol. Biol. Cell 3, 29–47. Berlin, V., Styles, C.A., Fink, G.R., 1990. BIK1, a protein required for microtubule function during mating and mitosis in Saccharomyces cerevisiae, colocalizes with tubulin. J. Cell Biol. 111, 2573–2586. Bode, C.J., Gupta, M.L., Reiff, E.A., Suprenant, K.A., Georg, G.I., Himes, R.H., 2002. Epothilone and paclitaxel: unexpected differences in promoting the assembly and stabilization of yeast microtubules. Biochemistry 41, 3870–3874. Deutschbauer, A.M., et al., 2005. Mechanisms of haploinsufficiency revealed by genome-wide profiling in yeast. Genetics 169, 1915–1925. Foland, T.B., Dentler, W.L., Suprenant, K.A., Gupta Jr., M.L., Himes, R.H., 2005. Paclitaxelinduced microtubule stabilization causes mitotic block and apoptotic-like cell death in a paclitaxel-sensitive strain of Saccharomyces cerevisiae. Yeast 22, 971–978. Gaber, R.F., Copple, D.M., Kennedy, B.K., Vidal, M., Bard, M., 1989. The yeast gene ERG6 is required for normal membrane function but is not essential for biosynthesis of the cell-cycle-sparking sterol. Mol. Cell. Biol. 9, 3447–3456.

146

A. Wilmes et al. / Gene 497 (2012) 140–146

Gaitanos, T.N., et al., 2004. Peloruside A does not bind to the taxoid site on β-tubulin and retains its activity in multidrug-resistant cell lines. Cancer Res. 64, 5063–5067. Giaever, G., et al., 1999. Genomic profiling of drug sensitivities via induced haploinsufficiency. Nat. Genet. 21, 278–283. Giaever, G., et al., 2002. Functional profiling of the Saccharomyces cerevisiae genome. Nature 418, 387–391. Giaever, G., et al., 2004. Chemogenomic profiling: identifying the functional interactions of small molecules in yeast. Proc. Natl. Acad. Sci. U. S. A. 101, 793–798. Gupta Jr., M.L., Bode, C.L., Georg, G.L., Himes, R.H., 2003. Understanding tubulin-Taxol interactions: mutations that impart Taxol binding to yeast tubulin. Proc. Natl. Acad. Sci. U. S. A. 100, 6394–6397. Hardwick, K.G., Murray, A.W., 1995. Mad1p, a phosphoprotein component of the spindle assembly checkpoint in budding yeast. J. Cell Biol. 131, 709–720. Hood, K.A., et al., 2002. Peloruside A, a novel antimitotic agent with paclitaxel-like microtubule-stabilizing activity. Cancer Res. 62, 3356–3360. Huzil, J.T., et al., 2008. A unique mode of microtubule stabilization induced by Peloruside A. J. Mol. Biol. 378, 1016–1030 31. Ikui, A.E., Yang, C.P.H., Matsumoto, T., Horwitz, S.B., 2005. Low concentrations of Taxol cause mitotic delay followed by premature dissociation of p55CDC from Mad2 and BubR1 and abrogation of the spindle checkpoint, leading to aneuploidy. Cell Cycle 4, 1385–1388. Jungwirt, H., Kuchler, K., 2006. Yeast ABC transporters — a tale of sex, stress, drugs and aging. FEBS Lett. 580, 1131–1138. Kanakkanthara, A., et al., 2011. Peloruside- and laulimalide-resistant human ovarian carcinoma cells have βI-tubulin mutations and altered expression of βII- and βIII-tubulin isotypes. Mol. Cancer Ther. 10, 1419–1429. Li, Y., Benezra, R., 1996. Identification of a human mitotic checkpoint gene: hsMAD2. Science 274, 246–248. Li, R., Murray, A.W., 1991. Feedback control of mitosis in budding yeast. Cell 66, 519–531. Lum, P.Y., et al., 2004. Discovering modes of action for therapeutic compounds using a genome-wide screen of yeast heterozygotes. Cell 116, 121–137. Mager, W.H., Winderickx, J., 2005. Yeast as a model for medical and medicinal research. Trends Pharmacol. Sci. 26, 265–273.

Menacho-Marquez, M., Murguia, J.R., 2007. Yeast on drugs: Saccharomyces cerevisiae as a tool for anticancer drug research. Clin. Transl. Oncol. 9, 221–228. Nguyen, T.L., Xu, X., Gussio, R., Ghosh, A.K., Hamel, E., 2010. The assembly-inducing laulimalide/peloruside A binding site on tubulin: molecular modeling and biochemical studies with [3H]peloruside A. J. Chem. Inf. Model. 50, 2019–2028. Nislow, C., Giaever, G., 2007. Chemical genomic tools for understanding gene function and drug action. Meth. Microbiol. 36, 387–414. Novick, P., Osmond, B.C., Botstein, D., 1989. Suppressors of yeast actin mutations. Genetics 121, 659–674. Parekh, H., Wiesen, K., Simpkins, H., 1997. Acquisition of Taxol resistance via P-glycoprotein- and non-P-glycoprotein-mediated mechanisms in human ovarian carcinoma cells. Biochem. Pharmacol. 53, 461–470. Parsons, A.B., et al., 2004. Integration of chemical-genetic and genetic interaction data links bioactive compounds to cellular target pathways. Nat. Biotechnol. 22, 62–69. Parsons, A.B., et al., 2006. Exploring the mode-of-action of bioactive compounds by chemical-genetic profiling in yeast. Cell 126, 611–625. Pierce, S.E., Davis, R.W., Nislow, C., Giaever, G., 2007. Genome-wide analysis of barcoded Saccharomyces cerevisiae gene-deletion mutants in pooled cultures. Nat. Protoc. 2, 2958–2974. Schatz, P.J., Solomon, F., Botstein, D., 1986. Genetically essential and nonessential α-tubulin genes specify functionally interchangeable proteins. Mol. Cell. Biol. 6, 3722–3733. West, L.M., Northcote, P.T., Battershill, C.N., 2000. Peloruside A: a potent cytotoxic macrolide isolated from the New Zealand sponge Mycale sp. J. Org. Chem. 65, 445–449. Whitehurst, A.W., et al., 2007. Synthetic lethal screen identification of chemosensitizer loci in cancer cells. Nature 446, 815–819. Wilmes, A., et al., 2011. Effects of the microtubule stabilizing agent peloruside A on the proteome of HL-60 cells. Invest. New Drugs 29, 544–553. Wilmes, A., Chan, A., Rawson, P., Jordan, T.W., Miller, J.H., 2012. Paclitaxel effects on the proteome of HL-60 promyelocytic leukemic cells: comparison to peloruside A. Invest. New Drugs 30, 121–129. Winzeler, E.A., et al., 1999. Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285, 901–906.