Multiple Repair Pathways Mediate Tolerance to Chemotherapeutic Cross-linking Agents in Vertebrate Cells Free

Cross-linking agents that induce DNA interstrand cross-links (ICL) are widely used in anticancer chemotherapy. Yeast genetic studies show that nucleotide excision repair (NER), Rad6/Rad18-dependent postreplication repair, homologous recombination, and cell cycle checkpoint pathway are involved in ICL repair. To study the contribution of DNA damage response pathways in tolerance to cross-linking agents in vertebrates, we made a panel of gene-disrupted clones from chicken DT40 cells, each defective in a particular DNA repair or checkpoint pathway, and measured the sensitivities to cross-linking agents, including cis-diamminedichloroplatinum (II) (cisplatin), mitomycin C, and melphalan. We found that cells harboring defects in translesion DNA synthesis (TLS), Fanconi anemia complementation groups (FANC), or homologous recombination displayed marked hypersensitivity to all the cross-linking agents, whereas NER seemed to play only a minor role. This effect of replication-dependent repair pathways is distinctively different from the situation in yeast, where NER seems to play a major role in dealing with ICL. Cells deficient in Rev3, the catalytic subunit of TLS polymerase Polζ, showed the highest sensitivity to cisplatin followed by fanc-c. Furthermore, epistasis analysis revealed that these two mutants work in the same pathway. Our genetic comprehensive study reveals a critical role for DNA repair pathways that release DNA replication block at ICLs in cellular tolerance to cross-linking agents and could be directly exploited in designing an effective chemotherapy. (Cancer Res 2005; 65(24): 11704-11)

cis-Diamminedichloroplatinum (II) (cisplatin) can form intrastrand cross-links, interstrand cross-links (ICL), DNA-protein cross-links, and monoadducts with DNA (reviewed in ref. 1). Among these lesions, ICLs are believed to be the main determinant of the toxicity of cross-linking agents presumably due to the difficulty of eliminating ICLs. However, intrastrand cross-links also seem to contribute to cisplatin toxicity (2). The toxic effects of ICLs as well as other types of cross-links are thought to be a result of blocked transcription and DNA replication (reviewed in refs. 3, 4). In particular, replication fork blocks in cycling cells might be crucial for cytotoxicity of ICLs because of the resulting single-strand breaks (SSB) and double-strand breaks (DSB). A single unrepaired DSB can stimulate the DNA damage checkpoint and trigger apoptosis in chicken B-cell line DT40 as well as in mammalian cells.

Comprehensive study of yeast mutants reveals that nucleotide excision repair (NER), homologous recombination, and postreplication repair (PRR) play a major role in repairing DNA damage caused by cross-linking agents (5–7). Cisplatin adducts are eliminated by the NER pathway, which uses Xpc/Hr23B and Xpa to detect the lesions. These proteins recruit two endonucleases, Ercc1-Xpf and Xpg, leading to incision at the 5′ and 3′ sides of a DNA lesion and to removal of the lesion. PRR helps restarting replication stalled at lesions that are not repaired before the S phase of cell cycle. This pathway includes two major branches: translesion DNA synthesis (TLS) and homologous recombination (8, 9). TLS can fill the strand gap by employing the Rad6/Rad18 complex, an ubiquitin-conjugating enzyme, and specialized TLS polymerases, such as Polη, Polκ, and Polζ (8). Although Rad18 is required for the function of each TLS polymerase in yeast, this may not the case in vertebrate cells, as Rad18-deficient DT40 cells have a less severe phenotype than do Polζ-deficient DT40 cells (9, 10). Homologous recombination can fill the strand gap by using the other intact sister DNA as a template (reviewed in refs. 11, 12) or could be required to repair replication-induced DSB, and nonhomologous end-joining (NHEJ) might also be employed to religate DNA strands that are broken as a result of replication fork stalling. Finally, RecQ helicases, such as yeast Sgs1 and mammalian Bloom helicase (BLM), also help stabilizing stalled forks by suppressing “promiscuous” recombination (13).

Although the primary structure of genes involved in DNA repair and checkpoint pathways are well conserved in the eukaryotes, the relative role for each pathway is distinctly different between yeast and mammalian cells. A good example is the differential contribution of homologous recombination and NHEJ to ionizing radiation (IR)–induced DSB repair, with homologous recombination being the dominant pathway in yeast and with NHEJ being of major importance in mammalian cells (14). In addition, the vertebrate repair machinery seems to be more complex than its yeast counterpart. There are several homologous recombination genes that are present in vertebrates but not in the budding yeast. Such genes include XRCC2, XRCC3, and tumor suppressor genes for familial breast and ovarian cancer, BRCA1 and BRCA2. Furthermore, Fanconi anemia (FA) complementation group (FANC) proteins and poly(ADP-ribose) polymerases (PARP) are also present only in higher eukaryotes. The mutations in FANC genes are responsible for the highly cancer-prone genetic disease FA, and the relevant mutants show hypersensitivity specifically to cross-linking agents (reviewed in ref. 15). There seems to be a cross-talk between FA and BRCA2 molecules, as some of the FA patients are found to carry mutations in the BRCA2 gene (16). Parp-1 is rapidly activated by SSB and DSB and can modify various components of DNA replication and repair machinery by poly(ADP-ribosyl)ation (17). The role for damage checkpoints in cellular survival following DNA damage also seems to be different between yeast and vertebrate cells. In multicellular organisms, the damage checkpoint not only transiently arrests the cell cycle, as it is the case in yeast, but also triggers apoptosis. Thus, vertebrate damage checkpoints have both positive and negative effects on cellular survival. Moreover, recent studies suggest that mammalian and chicken damage checkpoint proteins can directly facilitate DNA repair, including stimulation of Artemis-dependent DSB repair by ATM (18) and the phosphorylation of Rad51 by Chk1 (19) as well as the contribution of Nbs1 to homologous recombination (20).

The ultimate goal of chemotherapy is to selectively eliminate cancer cells. To predict the effectiveness of each chemotherapeutic treatment, one needs to understand the correlation between vulnerability to chemotherapeutic agents and activity of each cellular DNA damage response pathway. To this end, pioneering studies in yeast might be a starting point but are surely not sufficient to allow a systematic assessment of the effects of chemotherapeutic agents on mutants in different DNA damage response pathways that are relevant to humans. Such a genetic approach using a vertebrate cell line has been difficult to perform because of the lack of a good cellular model system for reverse genetic studies. Due to its high gene targeting efficiency and its stable karyotype, the chicken B lymphocyte line DT40 provides such a unique collection of isogenic mutant clones of DNA repair and damage checkpoint pathways (12). Moreover, because DT40 cells do not carry functional p53 and proliferate an extremely high rate, DT40 and malignant cancer cells may share comparable characteristics in cellular responses to cross-linking agents. Here, we describe for the first time a systematic analysis of the effects of DNA cross-linking agents on this panel of vertebrate mutant cell lines. Our study reveals an important role for Polζ presumably in both TLS- and Fanc-dependent cross-link repair to render the cells resistant to cisplatin.

Cell lines and cell culture. All mutants were generated from DT40 cells. The following mutants were kindly provided: NBS1 gene-disrupted DT40 mutant cells (nbs1; Dr. Tauchi, Ibaraki University, Ibaraki, Japan, and Dr. Komatsu, Kyoto University, Kyoto, Japan), fen1 and ligase IV (Dr. Koyama, Yokohama City University, Yokohama, Japan), blm and wrn (Dr. Matsumoto, AGENE Research Institute, Kanagawa, Japan), blm/ku70 and blm/rad54 Drs. Seki and Enomoto (Tohoku University, Tohoku, Japan), and atm by Dr. Yamamoto (Kanazawa University, Kanazawa, Japan). Disruption of the gene in each mutant was confirmed by the disappearance of either target mRNA or protein (references in Table 1 and data not shown). rev3/fanc-c double mutants were generated by introducing a fanc-c knockout construct containing a puromycin-resistant cassette (21) into rev3 mutant (10). Note that FANC-C is encoded in a sex chromosome. Cells were cultured as described previously (10).

Colony formation assay. Cells (1 × 10⁵) were incubated for 1 hour with complete medium containing cisplatin (Nihon-Kayaku, Tokyo, Japan), mitomycin C (Mitomycin Kyowa S, Kyowa Hakko, Tokyo, Japan), melphalan (Alkeran, GlaxoSmithKline, Tokyo, Japan), trans-diamminedichloroplatinum (II) (transplatin; Sigma, St. Louis, MO), methylmethan sulfonate (MMS; Nakalai, Kyoto, Japan), or vincristine (Oncovin, Nihon-Kayaku). Serially diluted cells were plated in triplicate onto 1.5% (w/v) methylcellulose semisolid medium as described previously (22).

Measurement of cisplatin sensitivity in liquid culture. Cells (5 × 10⁶) were incubated with complete medium containing cisplatin for 1 hour. The cells were washed and incubated for ∼10 days in cisplatin-free complete medium. Using flow cytometry, the number of cells was monitored (22) daily up to 10 days, when a population of the cells started to grow exponentially. This exponential growth curve was extrapolated to a zero time by a numerical fit to calculate the percentage of the cells that reacquired proliferation capability after the exposure to cisplatin. Alternatively, cells were continuously exposed to various concentrations of cisplatin for 2 days, and the number of cells was measured at 48 hours. The sensitivity was calculated by dividing the numbers of cells treated with cisplatin by those of untreated cells.

Karyotype analysis. For karyotype analysis, cells were treated with 5 μmol/L cisplatin for 1 hour, washed twice with PBS, and incubated with complete medium containing 0.1 μg/mL colcemid for another 3 hours. Cells were then fixed and metaphase spreads were prepared as described previously (23).

Experimental design. To study the contribution of each DNA damage response pathway to the repair of cross-linking agents in vertebrate cells, we have made a panel of isogenic DT40 mutants, each defective in a particular DNA repair or checkpoint pathway (Table 1). When gene-disrupted clones from DT40 and murine embryonic stem cells were studied previously, their sensitivity to cross-linking agents was not necessarily evaluated under the same experimental condition. In this study, to understand the relative contribution of each repair pathway, we examined cellular response to cross-linking agents using colony survival assay after 1-hour exposure. The following three cross-linking agents were studied, cisplatin (Fig. 1), melphalan, a derivative of nitrogen mustard (Fig. 2A), and mitomycin C, a natural antitumor antibiotic (Fig. 2B) and transplatin (Fig. 2C). The sensitivity of each mutant was assessed by D₁ values (i.e., the dose that reduces the cell survival to 1%). Furthermore, to accurately compare the sensitivity of each clone, we normalized the sensitivities of mutant cells according to the D₁ values of their parental wild-type DT40 clone, because three parental clones used in different laboratories showed up to ±18% variation in the D₁ value. To examine the cause of cell death following cisplatin treatment, we scored chromosomal aberrations following 1-hour exposure of representative repair-deficient mutants to cisplatin.

Sensitivity profiles of repair and damage checkpoint-defective mutants to cross-linking agents. Figure 1 shows the sensitivity profile of all analyzed mutants to cisplatin. Cells deficient in rev3, fanc-c, and rad18 displayed the highest sensitivity in this order. The xrcc2, xrcc3, brca2^truncated (involved in homologous recombination), brca1, atm (checkpoint), snm1a, snm1b (ICL repair), rad30 (TLS), xpa, xpg (NER), blm, wrn (RecQ helicase), msh3 (mismatch repair), and parp-1 [base excision repair (BER), and SSB repair] mutants also showed elevated sensitivity to cisplatin. To confirm that the sensitivity to killing by cisplatin is caused by DNA cross-linking, we exposed representative cisplatin-sensitive mutants to other cross-linking agents, melphalan and mitomycin C. The sensitivity profiles to the three different cross-linking agents were very similar (Fig. 2A and B), including the order of sensitivity among rev3, fanc-c, and rad18 cells. These observations revealed the critical roles of TLS and FANC pathway in processing DNA damages induced by a variety of cross-linking agents.

To further study the role for TLS in cellular tolerance to cisplatin, some hypersensitive mutants were exposed to transplatin. Transplatin has a much lower cytotoxicity than cisplatin presumably due to a slow conversion rate of transplatin monoadducts into ICLs [at least 10 times slower (t_1/2 > 17 hours) than that of cisplatin] and due to lack of the toxic d(GpG) intrastrand cross-links (24). We treated cells with transplatin only for 1 hour, so that most lesions should be monofunctional adducts. In agreement with the cisplatin sensitivity profile, rev3 and rad18 cells were more sensitive to killing by transplatin than any other cross-linker-sensitive mutants (Fig. 2C). To further support the notion that rev3 and rad18 play a critical role in bypassing monoadducts, we exposed the mutants to MMS, a monofunctional alkylating agent. rev3 and rad18 were sensitive to MMS (Fig. 2D), whereas fanc-c did not show a significant sensitivity (see Discussion). Thus, Rev3 and Rad18 may play an important role in the tolerance to platinum monoadducts in addition to tolerance to ICLs.

Besides a defect in DNA repair, the observed sensitivity of rev3 and rad18 mutants to cross-linking agents and MMS could also reflect compromised cell growth in these mutants (e.g., spontaneous cell death or low plating efficiency). However, rad18 cells show normal growth property and 100% plating efficiency (9). To further verify that the sensitivity is due to DNA repair defects in the mutant cells, we measured the sensitivity of representative mutants to the anticancer agent vincristine, which kills cancer cells by disrupting microtubules but does not form covalent lesions on DNA. rev3 and rad18 were not hypersensitive to vincristine (Supplementary Data), suggesting that the sensitivity to cross-linking agents is indeed due to compromised DNA repair and does not reflect altered cell growth properties.

To further confirm if the cisplatin toxicity is mediated by DNA insults, we scored chromosome aberrations of representative mutants after cisplatin treatment. As expected, cells sensitive to cisplatin showed high levels of chromosomal aberrations, suggesting that unrepaired DNA damage is a main cause of cell death. rev3 showed highest levels of aberrations followed by fanc-c, xrcc2, and rad18 (Table 2). We found that TLS mutants exhibited predominantly chromatid-type breaks, where one of the sisters is broken. This observation is consistent with the role for TLS in releasing replication block in a sister chromatid. Interestingly, xrcc2 exhibited higher percentage of chromosome-type breaks, where two sisters are broken at the same site. This might be the consequence of collapsed recombination structures, such as Holliday junctions.

Role for excision repair in cross-link repair. One of the major differences in cross-link repair between yeast and mammal is the extent of contribution by NER. All NER mutants in yeast show a relatively severe sensitivity to cross-linking agents. Thus, NER may be essential for initiation of ICL repair presumably by introducing incisions in one strand and thereby unhooking ICLs. However, rodent Chinese hamster ovary (CHO) cell mutants defective in NER show only a mild sensitivity to cross-linking agents. In agreement with data from the CHO cell mutants, DT40 deficient in Xpa and Xpg exhibited only mild sensitivity to cisplatin (Fig. 1). On the other hand, we failed to detect any sensitivity of xpa to melphalan or mitomycin C (Fig. 2A and B). Thus, NER may have a minor role in cross-link repair in higher eukaryotic cells when compared with yeast. We were not able to evaluate the role for XPF, because xpf-null mutation seems to be lethal in DT40 cells.⁶

We also assessed the involvement of BER by studying cells deficient in Polβ and Fen1. As expected, a defect in BER did not result in hypersensitivity to cisplatin. On the other hand, cells deficient in Parp-1, which is involved in a variety of DNA repair pathways, including BER, SSB, stabilization of replication forks, and perhaps DSB repair, showed high sensitivity to cisplatin as well as IR.⁷ Taken together, these results suggest that the two excision repair pathways play a minor role, if any, in tolerance to cross-linking agents in chicken as well as in mammalian cells.

Role for homologous recombination in cross-link repair. According to yeast genetic studies, homologous recombination–mediated DSB repair plays a critical role in cross-link repair. Most of the DT40 mutants defective in homologous recombination showed a significant increase in sensitivities to cisplatin, melphalan, and mitomycin C (Figs. 1 and 2A and B). Interestingly, the analyzed homologous recombination mutants displayed marked variation in cellular response to cisplatin. Cells deficient in Rad51 paralogues (xrcc2, xrcc3, rad51b, rad51c, and rad51d; Fig. 1; refs. 25, 26) consistently showed prominent hypersensitivity; brca1, brca2^truncated, and rad54 showed only mild sensitivity; and nbs1 and rad52 showed no increase (Fig. 1). In contrast, previous data indicate that Rad54 and Nbs1 have a critical role in DSB repair following X-rays. These findings reveal that vertebrate homologous recombination may consist of different subpathways that are employed at different types of damage.

Cells deficient in Blm and Werner DNA helicase (WRN), both of which are members of the RecQ helicase family, were also less resistant to killing by cisplatin. This moderate sensitivity of blm may be attributed to its increased frequency of aberrant homologous recombination associated with crossover as manifested by dramatic elevation of sister chromatid exchange (SCE) in blm mutants (27, 28). Paradoxically, although Blm and Rad54 have opposite effect on the level of SCE, the sensitivity of blm, rad54, and blm/rad54 double mutants were similar. This result implies complex functional interactions between Blm and Rad54; for example, enhancement of homologous recombination in blm and decrease in homologous recombination efficiency in rad54 may be balanced in homologous recombination–dependent cross-link repair. The wrn mutant also showed mild sensitivity, whereas blm/wrn double mutations had an additive effect on the sensitivity. Thus, the two helicases may differentially contribute to cross-link repair. Recent data suggest that BLM can associate with FA complex (21, 29) and is recruited to damaged chromatin with the FA protein FancD2 (21), whereas WRN may be directly involved in homologous recombination (reviewed in ref. 30).

Interrelation of Rev3- and Fanc-c–dependent interstrand cross-link repair pathway. The highest sensitivity of rev3 and fanc-c mutants led us to investigate the relation of these pathways in cross-link repair as has been done previously (31). We disrupted the FANC-C gene in rev3 cells and found in agreement with the previous study that the double mutant was not more sensitive than the rev3 single mutant (Fig. 3, top right). Whereas Niedzwiedz et al. evaluated cisplatin sensitivity by counting cells during continuous exposure of cells to the agent for ∼48 hours in liquid culture, we employed additional two methods because cellular viability at a short period might not correlate with the colony survival (32). As shown in Fig. 3, we verified this result by using pulse exposure to cisplatin and monitoring survival by both colony formation assay (top left) and growth recovery in liquid culture (bottom).

Nonhomologous end-joining contributes only little to cross-link repair and its component, Ku, may interfere with other cross-link repair pathways. Although NHEJ is preferentially used to repair IR-induced DSBs in mammalian cells, mutations in NHEJ have little effect on the sensitivity to cross-linking agents (33). Similarly, we found that all the NHEJ mutants, including ku70, ligase IV, dna-pkcs, and artemis did not show hypersensitivity to killing by cisplatin (Fig. 1). The gene family that contains the Artemis endonuclease includes two additional paralogue genes, SNM1A and SNM1B, whose yeast homologue is SNM1 specifically involved in cross-link repair (34–36). The snm1a and snm1b DT40 mutants showed moderate and slight, respectively, sensitivities to cisplatin, and their double mutations had an additive effect as published before (36).

During analysis, we noticed that a defect in Ku70, which binds DSBs and initiates NHEJ, had opposite effects on the cisplatin sensitivity. The cisplatin sensitivity of rad54/ku70 was higher than that of either mutant clone, whereas a defect in Ku70 in blm and rev3 had little effect (Figs. 1 and 4). Remarkably, deletion of the KU70 gene in parp-1 reversed the cisplatin hypersensitivity nearly to wild-type level (Fig. 4, top right) This suppression is reminiscent of the higher efficiency of DSB-induced homologous recombination and resulting IR tolerance caused by the deletion of the KU70 gene in DT40 and mammalian cells (37–39). The present data shed light on complex cross-talk between Ku, homologous recombination, and Parp-1 in cellular response to replication block at cisplatin cross-links.

Here, we describe the contribution of multiple DNA repair and damage response pathway in cross-link repair in vertebrate cells. The sensitivity profiles of chicken DT40 cells revealed that the contribution to cross-link repair of each repair pathway seems to be different between yeast and vertebrates. Clearly, in DT40 cells, PRR, homologous recombination, and FANC pathways are mainly responsible for cross-link repair, whereas NHEJ, NER, and BER play only a minor role. As previous studies suggested, our data show that PRR is closely connected with the FANC pathway in ICL repair. We also reveal dual actions of Ku70, which can partially substitute for the lack of Rad54-dependent homologous recombination but also interferes with cross-link repair as shown in the example of the parp-1/ku70 double mutant.

Rev3 contributes to cellular tolerance to cisplatin presumably through Fanc-dependent homologous recombination pathway. Of all mutants tested in this study, Rev3, the catalytic subunit of Polζ, showed the highest sensitivity to cisplatin. This finding implies several possible explanations. Firstly, Rev3 might simply work as a TLS polymerase, allowing replication to continue through the DNA cross-link. Biochemically, this is hard to imagine especially for ICLs, and to date, such an activity has not been observed in vitro. Secondly, Rev3 may be involved in other repair pathways, such as homologous recombination. Both rev3 and rad18 mutants show decreased gene targeting frequencies and might be defective in specific homologous recombination reactions (9, 10). Thirdly, the finding that rev3 and rev3/fanc-c to cisplatin showed very similar sensitivity implies that these two genes might act in the same pathway (Fig. 3; ref. 31). Thus, vertebrate Rev3 might have acquired a new function, besides its role in TLS, in dealing with ICL in combination with the FANC group of proteins. Indeed, Fanc- and Rev3-deficient DT40 cells have generally very different phenotypes; however, they seem to be epistatic in their sensitivity to cisplatin tolerance. Considering that vertebrate rev3 is more sensitive to cisplatin than rad18 or fanc mutants, Rev3-dependent response to cisplatin in vertebrate may be independently regulated by two ubiquitin-associated pathways (i.e., FA/BRCA and Rad6/Rad18 pathways).

It is unclear whether Fanc-dependent activation of Polζ facilitates TLS past monoadduct and intrastrand cross-link or facilitates ICL repair. We favor the idea that the Fanc-dependent pathway mainly affects ICL repair but not TLS for the following reason. Defective TLS can be substituted at least partially by more frequent usage of homologous recombination, as spontaneous SCE is increased in TLS mutant cells, including polκ, polη, and polζ. In contrast, a defect in Fanc-c or Rev3 dramatically decreases the level of cisplatin-induced SCE compared with wild-type cells (31). Moreover, mammalian FA as well as fanc-c DT40 cells show virtually no sensitivity to MMS or UV (Fig. 2D; refs. 21, 40), and fanc-c had only moderate sensitivity to transplatin. Thus, the Fanc pathway may have a minor role, if any, in Polζ-dependent TLS past MMS or UV lesions. Taking the role for both the Fanc pathway and Rev3 in homologous recombination into account, it is tempting to speculate that this unknown mechanism for ICL repair might involve a specialized form of homologous recombination. The role of Rev3 in this ICL-induced homologous recombination could involve the start of strand elongation from the damaged structure. In summary, Polζ may have dual function in cellular tolerance to cross-linking agents, one in TLS past monoadducts and intrastrand cross-links and the other in ICL repair possibly involving FA signal transduction pathway as well as homologous recombination. This dual action might explain the extremely high sensitivity of the rev mutants to cross-linking agents. Taking the low fidelity of TLS polymerases into account, a significant contribution of Rev3 to bypassing a large number of cisplatin lesions implies that the use of cisplatin in chemotherapy could result in mutations through error-prone DNA synthesis.

Application to clinical research. DT40 is a unique cell line that offers a panel of isogenic mutants derived from a stable parental line. DT40 cells have the following characteristics that might affect the cellular responses to genotoxic stresses. First, DT40 seems, for unknown reasons, to possess significantly higher homologous recombination efficiency than any mammalian cell line. This special feature might substitute for a defect of known homologous recombination gene, as a defect in Xrcc2 has a more profound effect on rodent cells than on DT40 (25, 41). Second, DT40 divide thrice daily and spend only a short time in the G₁ phase of the cell cycle. Thus, an asynchronous population of DT40 cells consists only of a small G₁ fraction (∼15%) and a much larger S-phase fraction (70%). Third, like many cancer cells, DT40 lacks the functional p53 and as a result has no G₁-S damage checkpoint (42). Thus, DNA damage at any phase of the cell cycle may have direct effect on DNA replication. These DT40-specific characteristics suggest that a defect in the DNA repair that is associated with DNA replication may display a more prominent phenotype in DT40 cells than in other cell lines that have longer G₁ phase and/or normal G₁-S checkpoint. Bearing these DT40-specific characteristics in mind, DT40 is a valuable tool and has been used extensively to explore relevant cellular pathways responsible for cancer therapy. In general, most phenotypes, especially those in homologous recombination mutants, observed in DT40 correlate well with those in mammalian model systems. Moreover, genes that are essential for embryonic development, such as Rev3, are more readily analyzed in cellular model systems, such as DT40. Thus, the findings in this study could prove very useful for improvements in the chemotherapeutic application of cisplatin and related compounds. A novel strategy, for example, could be a screen to identify chemical compounds that inhibit the TLS pathway. According to our findings, inhibitors of Rev3 or Rad18 could be highly efficient in sensitizing tumor cells to killing by cisplatin. Moreover, inhibition of Rev3 may reduce unwanted mutations induced by its error-prone activity. Similarly, selective inhibitors of Parp, which are already available, in combination with cross-linking agents could be useful to further sensitize cancer cells to chemotherapy. The data presented in this study thus provide a rational approach to measure the importance of individual of repair pathways as targets in chemotherapy.

Grant support: Core Research for Evolutional Science and Technology from Japan Science and Technology Corp., Center of Excellence Grant for Scientific Research and Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology of Japan, The Uehara Memorial Foundation, and The Naito Foundation.

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.

We thank Y. Sato for her technical assistance and Dr. T. Enomoto, Dr. R. Kanaar (Erasmus University, Rotterdam, the Netherlands), Dr. A.D. D'Andrea (Dana-Farber Cancer Institute, Boston, MA), and L.J. Niedernhofer (University of Pittsburgh Cancer Institute, Pittsburgh, PA) for critical reading.