Cisplatin is a crucial agent in the treatment of many solid tumors, yet many tumors have either acquired or intrinsic resistance to the drug. We have used the homozygous diploid deletion pool of Saccharomyces cerevisiae, containing 4728 strains with individual deletion of all nonessential genes, to systematically identify genes that when deleted confer sensitivity to the anticancer agents cisplatin, oxaliplatin, and mitomycin C. We found that deletions of genes involved in nucleotide excision repair, recombinational repair, postreplication repair including translesional synthesis, and DNA interstrand cross-link repair resulted in sensitivity to all three of the agents, although with some differences between the platinum drugs and mitomycin C in the spectrum of required translesional polymerases. Putative defective repair of oxidative damage (imp2′Δ strain) also resulted in sensitivity to platinum and oxaliplatin, but not to mitomycin C. Surprisingly in light of their different profiles of clinical activity, cisplatin and oxaliplatin have very similar sensitivity profiles. Finally, we identified three novel genes (PSY1–3, “platinum sensitivity”) that, when deleted, demonstrate sensitivity to cisplatin and oxaliplatin, but not to mitomycin C. Our results emphasize the importance of multiple DNA repair pathways responsible for normal cellular resistance to all three of the agents. Also, the similarity of the sensitivity profiles of the platinum agents with that of the known DNA interstrand cross-linking agent mitomycin C, and the importance of the gene PSO2 known to be involved in DNA interstrand cross-link repair strongly suggests that interstrand cross-links are important toxic lesions for cisplatin and oxaliplatin, at least in yeast.
Platinum-based chemotherapy is a vital component in the treatment of many tumors including lung, testicular, ovarian, and head and neck cancers. However, despite the high efficacy of these anticancer agents, drug resistance, either acquired or intrinsic, remains problematic, and there is no way of predicting which individual tumors will respond to treatment. One possible method to predict individual tumor sensitivity would be to measure the protein (or mRNA levels) of genes known to affect sensitivity to platinum-based drugs. In support of such an approach is the fact that clinical resistance to cisplatin of ovarian, lung, and gastric cancers has been reported for tumors with increased levels of expression of the ERCC1 (excision repair cross-complementing 1) gene (1, 2, 3) . The ERCC1 protein forms a heterodimer with XPF (xeroderma pigmentosum complementation group F), which performs the 5′ strand incision during nucleotide excision repair (NER; Refs. 4 , 5 ) and in the removal of interstrand cross-links (6) . Expression levels of another gene involved in NER, xeroderma pigmentosum complementation group A (XPA), have also been associated with cisplatin-resistant ovarian cancers (1 , 7) and testicular cancer cell lines (8) . In addition to the above defects in the NER pathway, disruption of other repair pathways also affect the sensitivity of cells to cisplatin, including recombinational repair (9, 10, 11) , mismatch repair (MMR; Refs. 12, 13, 14 ), and repair by translesional synthesis (a subset of postreplication repair; Ref. 15 ). However, for genetic profiling to be maximally effective a complete list of the genes affecting cisplatin sensitivity is required. The goal of the present study was to help provide such a list by performing a screen for cisplatin sensitivity with the budding yeast Saccharomyces cerevisiae.
S. cerevisiae has proven to be a powerful genetic system for the study of DNA repair because of its ease of genetic manipulation and the high degree of functional conservation between S. cerevisiae and human DNA repair genes (16 , 17) . As in humans, all three of the major complementation groups in yeast DNA repair influence platinum sensitivity. Mutation of genes within NER (RAD1, RAD10, and RAD2), recombinational repair (RAD51 and RAD52), and postreplication repair (RAD6 and RAD18) confer platinum sensitivity (18, 19, 20) . As in humans, the loss of MMR leads to platinum resistance in yeast (21) , and mutations in genes in the error prone pathway in translesional synthesis (REV1, REV3, and REV7) are sensitive to cisplatin (20 , 22) . Finally, another repair pathway contributing to platinum resistance is the HMG-domain proteins, which bind DNA intrastrand cross-links (23 , 24) . Within the HMG-domain family, the absence of the yeast IXR1 gene has been shown to cause cisplatin resistance (25) , and human HMG1 has been shown to inhibit removal of platinum-DNA intrastrand cross-links (26) . Genetic screens for cisplatin resistance in yeast have also found that overexpression of the transcription factors Cin5 and Yap6 and disruption of the SKY1 gene confer resistance to cisplatin (27 , 28) . However, these screens have identified genes for which the overexpression or disruption produced resistance, and they did not identify the known genes for which deletion causes cisplatin sensitivity.
The recent completion of a systematic deletion of all of the open reading frames (ORFs) in yeast (29) has provided a powerful new tool for screening for genes for which deletion produces sensitivity (or resistance) to cytotoxic agents. We and others have used this collection of mutants to identify novel genes for which deletion confers sensitivity to UV (30) , ionizing radiation (31 , 32) , and methyl methane sulfonate (MMS) (33) . One of the advantages of this resource is that the gene replacement cassette contains two molecular “bar code” tags, or unique 20-base oligonucleotide sequences, which allow for unique identification of the strain in a pool of all of the deletion mutants by PCR amplification of the tags and subsequent hybridization to a high-density oligonucleotide array containing the corresponding complementary sequences (29) .
Here we have used a pool of 4728 homozygous deletion strains, representing deletion of all of the nonessential genes, to detect genes for which deletion confers sensitivity to cisplatin, oxaliplatin, and mitomycin C. We chose to perform the screen with oxaliplatin, because it is an analog of cisplatin that has a different clinical spectrum of activity and for which there is evidence for some differences in basic mechanisms of action (34) . We chose mitomycin C because of the known mechanism of action of this drug in producing interstrand cross-links, which is one of the suggested mechanisms by which cisplatin kills cells.
Our results corroborate the importance of multiple DNA repair pathways responsible for normal cellular resistance to platinum and mitomycin C, demonstrate the importance of the repair of DNA interstrand cross-links for cisplatin and oxaliplatin sensitivity, and identify several new genes for which deletion produces sensitivity to these platinum drugs.
MATERIALS AND METHODS
Genotypes of the parental yeast strain BY4743, construction of the homozygous diploid deletion strains, and construction of the homozygous diploid deletion pool have been described previously (35) . All of the completed deletion strains are available through Research Genetics (Huntsville, AL) or EUROSCARF (Frankfurt, Germany). We used a mutant pool of 4728 nonessential homozygous diploid deletion strains. The parental diploid strain BY4743 was used as control in survival and complementation assays.
Cisplatin and mitomycin C were obtained from Sigma. Oxaliplatin was obtained from Sanofi-Sythelabo (Great Valley, PA). Cisplatin stock solutions were prepared in yeast extract peptone dextrose (YPD), stored as aliquots at −20°C, and used once; oxaliplatin and mitomycin C stock solutions were freshly prepared in YPD for each experiment.
Chemotherapy Treatment Assays.
Deletion pool aliquots were resuspended in YPD medium, shaken at 300 rpm at 30°C, and grown to early log-phase (A600 0.2–0.3). Equivalent numbers of cells (6 × 106) were then mock treated or treated with cisplatin, oxaliplatin, or mitomycin C for 1 or 4 h, and shaken at 300 rpm at 30°C. Cells were then harvested, washed once with buffer (10 mm Tris and 5 mm EDTA), and resuspended in YPD. A fraction of the cell resuspension was added to 250 ml of YPD and grown an additional 16 h at 30°C. Cells were then harvested by centrifugation and stored at −80°C.
Genomic DNA, Probe Production, Chip Hybridization, and Postscanning Analysis.
Genomic DNA preparation, PCR, and hybridization of DNA from the treated and control cultures was performed as described previously (36) . Each deletion strain is associated with four hybridization signals on the high-density oligonucleotide array generated in two separate PCR labeling reactions, UPTAG (sense and antisense) and DNTAG (sense and antisense). Equal numbers of cells were harvested in both the control and treated pools to produce equal pool label intensities. We normalized the data generated in the experimental array to that of the control array to eliminate any bias created during the PCR amplification reaction. In brief, a separate UPTAG normalization factor was calculated such that the total signal intensity of the UPTAGs was equal in the control and experimental arrays. We also calculated and applied a separate DNTAG normalization factor. We calculated the background intensity of each array to identify those tags that fail to generate a sufficient signal above the background to be meaningful. Of the 13,842 unassigned tags, 10,000 were chosen as a representative sampling to statistically model the nonspecific binding and background signal of an individual array (see Supplemental Data). 3 A small subset of the unassigned tags consistently reported high levels of signal intensity indicating a high degree of specific cross-hybridization and were not used in the background calculation. An experimental:control intensity ratio was including in the data analysis only if the signal generated in the untreated control array was at least twice the background signal. This arbitrary limit was chosen so that a maximum value of 0.5 for the ratio of treated:control (T:C) hybridizations would be obtained if the treated signal was at the background level. In addition, each treated tag that fell within 2 SDs of the background was flagged indicating that the measured value may be an overestimation of the true ratio (i.e., the true value would likely be lower indicating greater sensitivity). To yield a more stable estimate of the average T:C ratio with the skewed distribution of values from each experiment (Fig. 1) ⇓ we averaged the logs of the T:C ratios for each of the four tags assigned to an individual strain. Those strains that failed to have at least two of the tags significantly above background were not called for an individual experiment. In addition only those strains that were called in at least two of the replicate experiments were included in further analyses. This produced an average rejection rate of 9.3% of the strains across all of the experiments in this series.
For each drug, we tested 4728 different deletion strains in several experiments per drug. Statistical procedures that simultaneously perform large numbers of tests have the potential to generate many false-positive results, if done in the same manner as a single test. To avoid this problem, we calculated an adjusted P for each strain, which is the probability that any of the 4728 strains would have a statistically significant result by chance alone. We obtained Ps for simultaneous two-sided tests of both sensitivity and resistance.
To compute the one-sided sensitivity Ps, we ranked the geometric means for all of the deletion strains within each experiment from lowest to highest and computed Rj, the maximum rank achieved by deletion strain j across all of the experiments for a given drug. Low values of Rj correspond to genes that consistently exhibited sensitivity to the drug. We defined our adjusted P to be the probability that one or more nonsensitive strains would consistently rank at or below Rj by chance (i.e., have no rank above Rj). The Ps are computed under a nonsensitivity model in which every possible permutation of the deletion strains is equally likely. The probabilities can be computed explicitly (see Supplementary data for details of the derivation and formulae). 3 We computed analogous Ps for resistance genes by reversing the order in which the genes were ranked. The final adjusted P for each gene is twice the lesser of the sensitivity and resistance Ps. The adjusted P so defined is a conservative one, because it is calculated on the basis of having zero false positives.
In every experiment, some deletion strains fail to yield “good” ratios due to levels in the control being near background, as described above. As noted above we only ranked genes with 2 or more “good” values in two or more experiments. We modified the Ps for Rj, because strains with fewer good values are more likely to have small values of Rj by chance. We replaced the total number of genes tested with the smallest number of good ratios in any one experiment, which yields a conservative P. For each gene, we also replaced the total number of experiments with the number of experiments in which the gene yielded a good ratio. Consequently, genes with bad ratios in some experiments can still reach a statistically significant result but must have smaller values of Rj to do so.
The adjusted Ps are valid under more general assumptions than used for the calculations. Suppose n is the total number of genes tested. The model yields conservative adjusted Ps if, roughly, the probability a nonsensitive gene falls in the bottom k genes by chance is no greater than k/n, for k ≤ Rj.
The parent strain BY4743 and selected homozygous diploid deletion strains were tested for clonogenic survival to cisplatin treatment. Equivalent numbers (6 × 106) of early log-phase cells were suspended in YPD and treated with cisplatin (0.25–1.5 mm) for 4 h while being shaken at 300 rpm at 30°C. Cells were then pelleted, washed once with buffer (10 mm Tris and 5 mm EDTA), pelleted, and resuspended in the same buffer. Cells were then plated at appropriate dilutions onto YPD solid medium to allow for accurate counting of surviving colonies (range, 50–250). Plates were incubated for 3–4 days at 30°C before counting colonies. Full survival curves were performed at least three times for each selected deletion strain.
ORFs of selected deletion strains were generated by PCR and then subcloned into the vector pRS416.GAL1 (gift of Dr. Pat Brown, Department of Biochemistry, Stanford University). Genomic clones of MMS4, MMS2, and MUS81 were generous gifts from Dr. Steven Brill, (Department of Molecular Biology and Biochemistry, Rutgers University, Piscataway, NJ). Strains were transformed by the lithium acetate method with vector alone or with the construct containing the deleted ORF. Expression of PCR clones was achieved by use of the galactose promoter, whereas genomic clones were under control of their native promoters. Spots were grown on the appropriate solid medium (ura-/galactose or leu-/dextrose) and assessed after incubation for 2 days at 30°C.
Sensitivity Profiles of Cisplatin, Oxaliplatin, and Mitomycin C.
In preliminary experiments with wild-type cells we established the concentrations of cisplatin, oxaliplatin, and mitomycin C to obtain ∼50% killing with a 1-h exposure to cisplatin and a 4-h exposure to oxaliplatin and mitomycin C (1 mm, 10 mm, and 0.5 mm for the three drugs, respectively). In each experiment with the deletion pool we split a pool of 4728 deletion mutants into two and treated one with the relevant concentration of each of the three drugs, the other serving as the untreated control. After the drug exposure the cells were centrifuged, resuspended in fresh YPD, diluted, and grown for an additional 16 h before preparation of the genomic DNA. After PCR amplification of the tags and hybridization of the products to the oligonucleotide arrays, we obtained signal intensities from the sense and antisense tags for both the UPTAGS and DOWNTAGS for each deletion strain in the pool. Ratios of T:C were then calculated for each of these four tags and combined with those from replicate experiments to give an overall geometric mean T:C for each strain for each of the drugs.
We performed 6 identical experiments with cisplatin, 3 with oxaliplatin, and 5 with mitomycin C. We also reanalyzed our previous data (30) for UV irradiation using the same rigorous criteria (“Materials and Methods”) for comparison with the drug data. As we have reported previously for both UV and ionizing radiation exposures (30 , 32) , there was a high degree of correlation from experiment to experiment in the sensitivity rankings for each of the drugs (data not shown). This is reflected in the large number of genes that were classified as significantly sensitive (130) or resistant (100) to cisplatin according to our rigorous statistical test. Table 1 ⇓ shows a listing of the rankings for the top 50 most sensitive strains for cisplatin with the rankings for oxaliplatin, mitomycin C, and UV also shown.
Fig. 1 ⇓ shows a distribution of the ratios for the four treatments for all of the strains. Because our experiment was designed to identify sensitive strains, we gave a dose that produced only ∼50% cell kill in the wild-type cells. Thus, resistant strains would not be expected to deviate from a T:C ratio of 1.0 as much as sensitive strains. This is demonstrated in the “Results,” which show a skewing of the distribution of ratios to low rather than high values. Nonetheless, as noted we found 100 strains that were resistant (P < 0.05). Of the top 50 most resistant, 30 were involved in protein biosynthesis and 15 were of unknown function. None were involved in DNA repair. We did not pursue individual characterization of any of these resistant strains, but their identities are shown in the supplementary material.
Table 1 ⇓ shows that many of the genes in the top 50 ranking are known to be involved in DNA repair (including 21 of the top 25 genes). It is also apparent that there is a close similarity between the rankings for cisplatin and oxaliplatin, as well as a similarity with mitomycin C and UV irradiation.
Because of the clear importance of repair genes, we have listed (Table 2) ⇓ all of the nonessential genes involved in DNA repair and damage checkpoint response to show which of these genes, when deleted, produced sensitivity to the three drugs and UV radiation. This table illustrates the importance to cisplatin and oxaliplatin sensitivity of the NER factors 1, 2, and 3 (but not 4 or transcription-coupled repair), DNA interstrand cross-link repair (as shown by the importance of PSO2), postreplication repair, and to a lesser extent recombinational repair. Deletion of genes involved in double-strand break repair by nonhomologous end-joining, in base excision repair, and (interestingly) in the DNA-damage checkpoint response did not produce sensitivity to cisplatin or to oxaliplatin, although deletion of the checkpoint genes clearly produced sensitivity to UV and to mitomycin C. The absence of genes known to be involved in DNA repair from this table is either because they are essential or because (as in the case of RAD6, which is known to produce sensitivity to cisplatin when deleted; Ref. 20 ) their hybridization signals in the untreated sample were within the background range.
Confirmation of Individual Sensitivities.
We selected several of the strains that appeared sensitive to cisplatin by the hybridization assay and analyzed these by clonogenic survival after exposure to different concentrations of cisplatin (Fig. 2) ⇓ . In this assay we used a 4-h exposure to the drug to get high levels of cell kill at soluble concentrations of cisplatin. In other tests we found that the ranking of genes to a 1-h and a 4-h exposure to cisplatin was essentially identical (Supplementary Data). 3 In all of the cases strains that were found to be sensitive by the hybridization assay were also confirmed as sensitive by clonogenic survival. The strains showing the greatest sensitivity were those with deletions in RAD2, RAD10, and PSO2 followed by RAD5 (postreplication repair) and genes involved in recombinational repair (MUS81, MMS4, RAD57, and SAE2). Four genes previously unknown to confer sensitivity to cisplatin when deleted were confirmed in the clonogenic assay as moderately sensitive to cisplatin.
Confirmation of Gene Deletions in Causing Sensitivity.
To confirm that the sensitivities of selected strains to cisplatin were the result of deletion of the ORF and not to other genetic alterations, we reintroduced and expressed the gene products in a plasmid-based, inducible expression vector, pRS416 Gal-1. The strains we selected for this were those not known previously to be sensitive to cisplatin (imp2’, pph3, yk1076, ynl201c, yil132c, and ylr376c), as well as a selection of those in different pathways including recombination repair (mms4 and mus81), postreplication repair (mms2), and interstrand cross-link repair (pso2). Fig. 3 ⇓ shows spot tests of cisplatin sensitivity of these strains with the same number of cells spotted for wild-type, the deletion strain, and the deletion strain with the reintroduced ORF. In all of the cases the reintroduction of the appropriate ORF complemented the sensitivity of the deletion strain back to wild-type sensitivity.
Role of Polymerase ζ in Cisplatin and Oxaliplatin Sensitivity.
Polymerase ζ, a heterodimer of Rev3p and Rev7p, is a DNA polymerase involved in translesional synthesis (37) . Because our sensitivity rankings suggested a greater effect of deletion of REV3 than of REV7 on sensitivity to cisplatin (ranks 10 and 24, respectively) and especially to oxaliplatin (ranks 9 and 132, respectively), we performed additional experiments on these mutant strains using clonogenic assays to test their sensitivities to the two platinum drugs (Fig. 4) ⇓ . These data confirm the conclusions of the hybridization assay that deletion of REV3 produces greater sensitivity than does that of deletion of REV7, particularly for oxaliplatin.
Use of Genomic Screens in Yeast in the Study of Anticancer Agents
We performed the current studies to identify as many genes as possible for which the deletion produced sensitivity to cisplatin, oxaliplatin, and mitomycin C. The Yeast Deletion Project, an undertaking by an international consortium to delete all of the known ORFs in yeast, has produced a powerful resource to enable such a screen to be performed (29) . The value of such a screen in yeast relies on the high conservation of genes from yeast to humans, particularly those involved in repairing DNA (16 , 17) . The other major advantage of this resource is the fact that DNA extracted from a pool of all of the deletion mutants can be hybridized to a high density oligonucleotide array to allow the relative abundances of each strain in the pool, and, hence, their relative sensitivities, to be determined rapidly and quantitatively (38) . We have previously used this technique to identify genes involved in the resistance of cells to UV and X-irradiation (30 , 32) . In the present study we introduced a new method for data analysis that allowed us to better identify the strains where low abundance in the pool could affect their apparent sensitivity analyzed by the DNA hybridization technique. Our analysis of multiple experiments confirmed the high reproducibility of this system in identifying sensitive strains. On the basis of this close agreement between multiple experiments we performed a rigorous statistical analysis to identify sensitive and resistant stains and found 130 deletion strains that were sensitive to cisplatin with the criterion of zero false positives at the 95% level of confidence and 100 deletion strains that were resistant at the same level of confidence. However, because there is a continuum of sensitivities to any agent, it is somewhat artificial to categorize strains into having reduced, normal, or increased sensitivities. For this reason we have highlighted the top 50 most sensitive strains as determined by the hybridization assay (Table 1) ⇓ , because these have the most dramatic impact on drug resistance. However, all of the results can be accessed in the Supplementary Data. 3
We selected 18 strains for individual testing in cell survival assays identified as sensitive in the hybridization assay, and we confirmed the sensitivities of all of these strains. We were also able to show in 10 of 10 of these sensitive strains that the sensitivity of the strain could be attributed to deletion of the specific gene, because introduction of the gene in the deletion strain abrogated the sensitivity of that strain. These data provide a measure of confidence that the hybridization assay with the deletion pool can identify with some accuracy genes involved in cisplatin sensitivity.
One limitation of the present system is that essential genes cannot be interrogated. Although inactivation of the human orthologs of these genes is unlikely in human cancers (because the cells would not be expected to be viable), hypomorphic mutations in these genes could lead to a phenotype. One way to investigate this would be to examine the sensitivity of strains with heterozygous deletion of these genes.
At the present time there is no comparable resource to perform such a screen with mammalian cells. However, it has been argued that gene expression profiling after exposure to DNA-damaging agents could identify the genes that might be important for the survival of the cells to that agent (39 , 40) . Unfortunately, we and others have shown using yeast that there is no relationship between the genes that are induced after a range of DNA-damaging agents and the genes that are necessary for survival to those same agents (33 , 36 , 41) .
Multiple DNA-Repair Pathways Participate in Platinum and Mitomycin C Resistance.
Many of the strains with deletions of genes involved in NER were found to be sensitive to cisplatin, oxaliplatin, and mitomycin C in the screen. This was particularly the case for the endonucleases performing the incisions on both sides of the lesion, the Rad1p-Rad10p complex, and Rad2p (XPF-ERCC1 and XPG in humans). Strains with deletions of the genes involved in damage recognition, RAD14 and RAD4, were also equally sensitive to all three of the drugs and UV irradiation. Deletion of RAD23, which encodes a protein that forms a complex with Rad4p, produced somewhat less sensitivity to cisplatin than to UV and less still for oxaliplatin and mitomycin C. These data suggest less importance of Rad23p than Rad4p in recognizing the bulky adducts caused by cisplatin, oxaliplatin, or mitomycin C compared with UV. An even greater loss of sensitivity compared with UV was seen for strains with deletion of RAD7 and RAD16. Deletion of RAD26 and RAD28, the S. cerevisiae homologs of the Cockayne’s syndrome B and A genes, did not produce any sensitivity to cisplatin or to UV, consistent with earlier reports (42 , 43) .
The involvement of NER in the response of cells to these three cross-linking anticancer drugs very likely reflects the role of intrastrand cross-links in killing cells by these agents. However, for some of the proteins it could also reflect their involvement in interstrand cross-link repair (see below).
Repair of DNA Interstrand Cross-Links.
Repair of DNA interstrand cross-links is a complex process that involves genes that are in other DNA repair pathways, namely NER, homologous recombination, and postreplication repair (22 , 44) . Thus, the fact that yeast with deletions in any of these three pathways are sensitive to cisplatin, oxaliplatin, and mitomycin C does not constitute evidence that interstrand cross-links are important for the killing to these agents. However, one nonessential gene, PSO2/SNM1, has been demonstrated to be involved in interstrand cross-link repair without being involved in NER, postreplication repair, or recombination repair. We found that deletion of this gene conferred exquisite sensitivity to cisplatin, oxaliplatin, and mitomycin C but little or no sensitivity to UV as reported earlier (22) . In clonogenic survival assays of the individual strains the pso2 deletion was equally sensitive as strains with deletion of either of the DNA endonucleases encoding genes involved in NER (RAD2 and RAD10). This is evidence for the involvement of DNA interstrand cross-links in the killing of yeast cells by these three chemotherapeutic drugs. A human homologue of PSO2, hSNM1, has been identified and characterized (45) . Mice with homozygous deletion of the gene (mSNM1−/− mice) are viable and are sensitive to mitomycin C. A similar sensitivity of mSNM1−/− ES cells to mitomycin C but not to cisplatin was also observed (45) .
Deletion mutants in all of the components of postreplication repair were sensitive to cisplatin, although not equally so for oxaliplatin and mitomycin C. The two major proteins in postreplication repair, Rad6p and Rad18p, form a heterodimer that modifies proteins by conjugation to ubiquitin. The pathway then divides into two error-free branches, both involving the Ubc13p/Mms2p heterodimer and a third pathway, which is error prone and involves translational synthesis by Rev1p and the heterodimer Rev3p/Rev7p. Because their hybridization intensities were at background levels we could not obtain data for the sensitivity of the rad6 and ubc13 deletion mutants, but rad6 deletion strains have been reported to be as sensitive to cisplatin as rad18 strains (20) . Of interest is the sensitivity of the rev1 and rev3 deletion strains to cisplatin and oxaliplatin but not to mitomycin C. This is in agreement with earlier studies (20) . One of the few strains that showed a different sensitivity to cisplatin and oxaliplatin was the rev7 mutant, which was sensitive to cisplatin but much less so for oxaliplatin. This is somewhat surprising, because Rev3p and Rev7p form a heterodimer making up polymerase ζ (37) . This suggests that Rev3p may be able to function independently of Rev7p in repairing cisplatin lesions. The strains with deletion of components of the two error-free pathways of postreplication repair (rad5, pol32, and MMS2) showed equal sensitivity to cisplatin, oxaliplatin, and mitomycin C. In agreement with earlier studies (22) we found that yeast deficient in DNA polymerase pol-eta (Rad30p), which is involved in translesional synthesis past UV-induced photoproducts, was not sensitive to cisplatin.
Mutants in recombinational repair that showed sensitivity to both cisplatin and oxaliplatin included the deletion strains sae2, mus81, mms4, slx4, rad55, and rad59. Strains with deletions of RAD54 and RAD57 showed sensitivity to cisplatin but only modest sensitivity to oxaliplatin. Several members of this group had hybridization signals too close to background levels to be able to assess their sensitivity including rad50, Mre11, Xrs2, and rad51.
Nonhomologous End-Joining and Base Excision Repair.
As expected from the literature, mutants in double-strand break repair by nonhomologous end-joining and mutants in base excision repair showed no sensitivity to any of the drugs or to UV radiation.
DNA Damage Checkpoints.
A major difference between the two platinum drugs and mitomycin C was the sensitivity of mutants with deletion of genes in the DNA-damage checkpoint response including RAD9, RAD24, RAD17, MEC3, and DCC1. All of these genes when deleted produced sensitivity to mitomycin C but not to either cisplatin or oxaliplatin. This is somewhat in contrast to a previous report (20) showing similar, although marginal, sensitivities of rad9, rad17, and mec3 mutants to cisplatin and mitomycin C. However, it is in agreement with the work of Grossmann et al. (22) , who showed that cisplatin did not cause an S-phase arrest and who found no sensitivity of the checkpoint mutants rad9, rad17, and rad24 and slight sensitivity to cisplatin of the mec3 mutant.
Other Genes Conferring Sensitivity to Cisplatin.
The deletion strain imp2’ showed sensitivity by the hybridization assay to cisplatin, which we confirmed by clonogenic survival. This gene has been reported to encode a transcriptional activator that is involved in the protection of yeast against oxidative damage produced by bleomycin and other oxidants (46) . Although the sensitivity of imp2’ mutants to platinum drugs has not been reported previously, there are data in the literature showing that some of the toxic side effects of cisplatin can be abrogated using antioxidants such as α-tocopherol (47 , 48) .
The deletion strain pph3 was sensitive to cisplatin, oxaliplatin, and mitomycin C. Again, this is a novel finding, although this strain has been reported previously to be sensitive to MMS (41) . This gene encodes a protein serine/threonine phosphatase related to PP2A phosphatases. These are involved in many signal transduction pathways and have also been reported to be involved in DNA repair (49) .
We also identified three unknown genes that were sensitive both to cisplatin and oxaliplatin and less so to mitomycin C and UV (YKL076C, YNL201C, and YLR376C, which we have named PSY1, 2, and 3, respectively). All three were shown to be moderately sensitive to cisplatin by clonogenic assay with the psy2 strain being the most sensitive. Experiments are under way to determine the function of these genes.
Cisplatin and Oxaliplatin Have Similar but Not Identical Sensitivity Profiles
Oxaliplatin is a third generation platinum anticancer drug with differences from cisplatin that include a different toxicity profile as well as activity against platinum-resistant cells and tumors (50 , 51) . Despite that fact that oxaliplatin forms covalent adducts with DNA that have a similar sequence and region specificity to those formed by cisplatin (52) they are more cytotoxic than the adducts formed by cisplatin (51) . However, the mechanisms for the higher specific toxicity of the DNA adducts and for the lack of cross-resistance with cisplatin are not understood, but there are differences between the two drugs both in translesional synthesis (53 , 54) and in MMR (55 , 56) . In our analysis we found that loss of the stimulatory component of the replicative bypass polymerase ζ, Rev7p, conferred greater sensitivity to cisplatin than it did to oxaliplatin (Table 2 ⇓ ; Fig. 4 ⇓ ). However, the significance of this is unclear, because loss of the catalytic subunit of polymerase ζ, Rev3p, produced equal sensitivity to the two drugs (Fig. 4) ⇓ . The possibility that defective MMR does not result in oxaliplatin resistance is especially intriguing, given the activity of oxaliplatin in colon cancers, where aberrant MMR is common and cisplatin is ineffective. However, because our study was designed to detect sensitivity and not resistance, we were not able to detect any MMR gene deletions that differentiated the two drugs.
In this screen using a pool of homozygous deletion mutants of all nonessential genes in diploid yeast we identified >130 and 100 genes that, when deleted, produced sensitivity and resistance, respectively, to killing by cisplatin. However, because sensitivity is a continuum, we focused on those genes for which deletion produced the greatest sensitivity to cisplatin, and these were largely genes encoding proteins involved in multiple pathways of DNA repair, including NER, interstrand cross-link repair, postreplication repair, and recombinational repair. However, deletion mutants in double-strand break repair by nonhomologous end-joining and base excision repair were not sensitive to cisplatin. Of interest was the fact that strains deleted in genes involved in DNA damage checkpoints, including RAD9, RAD24, RAD17, MEC3, and DCC1, were sensitive neither to cisplatin nor to oxaliplatin but were sensitive to mitomycin C. This was the major difference in the spectrum of strains sensitive to the two platinum drugs and mitomycin C. The sensitivity of the strain deleted in PSO2, a gene known to be involved specifically in the repair of DNA interstrand cross-links, to both the platinum drugs and to mitomycin C provides strong evidence that, at least in yeast, DNA interstrand cross-links are important in cisplatin killing.
In addition, we discovered novel genes for which deletion produced sensitivity to cisplatin including IMP2’, PPH3, and three previously unnamed genes that we have named PSY1, 2, and 3.
Because there is strong conservation of genes between yeast and humans, particularly in DNA repair (16) , many of the yeast genes identified in this screen have human orthologs that would be expected to confer sensitivity to platinum drugs when mutated. However, such a screen in yeast cannot be expected to provide a comprehensive catalogue of mammalian genes involved in sensitivity to platinum drugs, because processes other than DNA repair are likely to be involved, and these could be considerably different between yeast and mammalian cells. These include genes encoding proteins in drug uptake/efflux, in sulfhydril metabolism, and in some oncogene pathways (57) . Nonetheless, the present screen in yeast provides useful and novel insights into the mechanisms involved in resistance to platinum containing anticancer drugs.
Grant support: Grant P01 CA67166 (to J. M. Brown) and Training Grant CA09302 (to J. A. Brown) from the National Cancer Institute, Department of Health and Human Services, and Grant R01 GM62628 from the National Institute of General Medical Sciences, Department of Health and Human Services (to L. Lazzeroni).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: J. Martin Brown, Division of Radiation and Cancer Biology, 269 Campus Drive, CCSR South Room 1255, Stanford University Medical Center, Stanford, CA 94305-5152. Phone: (650) 723-5881; Fax: (650) 723-7382; E-mail:
↵3 Supplemental data available at http://cbrl.stanford.edu/mbrown/Cisptresis.html.
- Received October 2, 2003.
- Revision received February 17, 2004.
- Accepted March 26, 2004.
- ©2004 American Association for Cancer Research.