Cross-Talk between Nucleotide Excision and Homologous Recombination DNA Repair Pathways in the Mechanism of Action of Antitumor Trabectedin

Trabectedin (Yondelis) is a novel antitumor agent originally isolated from the Caribbean marine tunicate, Ecteinascidia turbinata, which was selected for clinical investigation due to its potent cytotoxic activity against a variety of tumor cell lines in vitro and human tumor xenografts in vivo (1). Currently, trabectedin is undergoing phase II/III clinical trials both in Europe and the U.S., with promising results for the treatment of soft tissue sarcomas, and breast and ovarian cancers (2–4).

Despite its advanced stage in clinical investigation, the precise mechanism of action of trabectedin remains poorly understood. DNA seems to be an important target because the drug has been shown to form, in vitro, covalent adducts with the N2 of guanines located in the DNA minor groove (5, 6). The most prominent characteristic that makes its mechanism of action unique as compared with classic DNA-binding antitumor drugs is the atypical response detected in cells affected in the DNA repair mechanism nucleotide excision repair (NER). Whereas all known DNA-interacting drugs are either more effective or equally effective in NER-proficient and NER-deficient cells, trabectedin has been shown to be less effective in NER-deficient cells (7, 8). A hypothesis to explain this particular behavior has been proposed; i.e., that the drug may interact with the NER machinery to induce lethal DNA strand breaks (9).

To gain some insight into the in vivo mechanism of action of trabectedin and to explore the contribution of the NER system in such a mechanism, we used the fission yeast, Schizosaccharomyces pombe, as a model organism. Using this yeast, we observed that trabectedin causes DNA damage in vivo because it activates the S phase and G₂-M DNA damage checkpoint responses. Moreover, cells deficient in homologous recombination repair (HRR) are extremely sensitive to the drug, indicating that trabectedin probably causes double-strand breaks (DSB). As also found in mammalian cells, NER mediates trabectedin's cytotoxicity in S. pombe cells. In particular, cells deficient in Rad13 (the fission yeast orthologue to human XPG) are less sensitive to the drug. By constructing different mutants in the Rad13 protein, we have shown that (a) the ability of Rad13 to produce single-strand breaks (SSB) is not required for trabectedin's cytotoxicity, and (b) that a short region in the COOH terminus of the protein is involved in NER-dependent cell killing. The involvement of NER in the formation of trabectedin-DNA lethal complexes and HRR in its resolution are discussed.

Drugs, strains, media, and growth conditions. Trabectedin was obtained from PharmaMar and prepared as a 1 mmol/L stock solution in ethanol that was kept at −20°C. Methyl methanesulfonate (MMS) was supplied by Fluka (Buchs, Switzerland) and 4-nitroquinoline 1-oxide (4NQO) was supplied by Sigma (St. Louis, MO). The S. pombe strains used in this study are listed in Table S1 (supplemental data). Standard S. pombe molecular genetic techniques and the media were used as described previously (10).

Spot assays for analyzing sensitivity to trabectedin and MMS. Yeast strains were grown in YES medium until they reached the exponential phase. Cells were harvested by centrifugation and resuspended in YES medium to an absorbance of 2 (OD₅₉₅), corresponding to 2 × 10⁷ cells/mL. Five microliters of undiluted cell culture and 1/10 serial dilutions of each cell culture were spotted onto YES plates containing trabectedin (1-10 μmol/L), MMS (0.005-0.01%), or 4NQO (0.05 μg/mL).

Fluorescence-activated cell sorting analysis. Approximately 10⁷ cells were spun down, fixed in 70% ethanol, and processed for flow cytometry as described previously (10). A Becton Dickinson (Mountain View, CA) FACSCalibur was used. Cell size measurements were calculated with the forward-light scatter data of fluorescence-activated cell sorting (FACS), considering 100 as the size of the wild-type control.

4′,6-Diamidino-2-phenylindole staining and microscopy. Ethanol-fixed cells were rehydrated and stained with 4′,6-diamidino-2-phenylindole at a final concentration of 0.05 mg/mL. Pictures were taken with a Leica DM6000 B microscope, using a Hamamatsu ORCA-ER camera and Openlab 4.0.3 software.

Protein extracts and Western blots. About 3 × 10⁸ cells growing in YES medium were collected and processed for total protein extraction, as described previously (10). For Chk1HA detection, 50 μg of total protein extract was run on an 8% SDS-PAGE gel, transferred to nitrocellulose, and incubated with anti-HA monoclonal antibodies, 12CA5 (0.16 μg/mL), for 2 hours at room temperature. Sheep anti-mouse antibodies conjugated to horseradish peroxidase (1/2,000; Amersham) were used as secondary antibodies in 40-minute incubations at room temperature. Protein extracts for Rad13HA detection were obtained by TCA precipitation, as described previously (11), and resuspended in 200 μL of Laemmli buffer. Aliquots (25 μL) of the extracts were run on a 7% SDS-PAGE gel and Western blotted as described above. The immunoblots were developed using Super Signal (Pierce, Rockford, IL).

For details regarding the cloning of rad13 and construction of rad13 mutants, as well as the construction, refinement, and molecular dynamics simulations of a Rad13/DNA-trabectedin ternary complex, see supplemental data.

Trabectedin causes cell cycle arrest in G₂-M in fission yeast. To assess whether trabectedin elicits DNA damage, we first looked at the phenotype induced by this drug in S. pombe wild-type cells. When fission yeast cells were grown on plates containing trabectedin, we observed two different phenotypes depending on the concentration of the drug. At low concentrations (5-25 μmol/L), the cells became elongated with a characteristic cell cycle arrest phenotype. At high concentrations (25-100 μmol/L), the cells became small and round, and stopped growth after a few divisions. The cell cycle arrest phenotype consisted of elongated cells with a 2C DNA content, suggesting a possible delay in either the G₂-M transition or in mitosis (Fig. 1). Nuclear staining revealed that these cells arrested with an interphase nucleus (Fig. 1), indicating that the arrest was produced at G₂-M. Further evidence of the G₂-M arrest phenotype is presented in the supplemental data.

Trabectedin activates the intra-S phase and the G₂-M DNA damage checkpoints. Because trabectedin is able to bind to DNA in vitro (5, 6, 12), and causes a cell cycle delay in G₂, it was reasonable to speculate that this drug was activating the G₂-M DNA damage checkpoint. To test this possibility, we examined the sensitivity to trabectedin of a variety of checkpoint mutants in serial dilution experiments. First, we analyzed a set of mutants involved in the recognition of the DNA lesion: rad1Δ, rad9Δ, and rad17Δ, and a mutant involved in the transduction of the signal, rad3Δ (13). All of these mutants were hypersensitive to trabectedin (Fig. 2A), indicating that cells that are unable to activate the DNA damage checkpoint response are less tolerant to the drug. This hypersensitivity was observed even at 1 μmol/L of trabectedin, a concentration at which wild-type cells did not show any phenotype on plates.

Chk1 and Cds1 are two effector kinases activated by the sensors/transducer described above (14, 15). Cds1 is specific for the DNA damage induced during S phase, and Chk1 is activated in response to DNA damage induced in G₂. chk1Δ cells were also more sensitive than the wild-type to trabectedin treatment. Cell survival was reduced and cells barely elongated in the presence of the drug (Fig. 2A and B). Surprisingly, we observed increased resistance to trabectedin in cds1Δ cells as compared with wild-type cells (Fig. 2A). The same behavior has been reported for cds1 mutant cells treated with the bifunctional (alkylating and interstrand cross-linking) antitumor agents, nitrogen mustard and mitomycin C (16), possibly reflecting the fact that the delay in S phase and the attempt to repair the DNA damage caused by these drugs may be detrimental for cell survival. To study the possible effect of trabectedin during S phase, we synchronized cells in G₁ by nitrogen depletion and allowed them to reenter the cell cycle with or without the drug. Wild-type cells treated with trabectedin entered S phase at the same time as untreated cells (Fig. 2C, 2-3 hours). However, they were delayed in their progression through S phase (Fig. 2C,, arrows). This delay was specifically abolished in a cds1Δ strain but not in a chk1Δ strain (Fig. 2C), corroborating the idea that trabectedin produces lesions in the DNA (9, 17), and showing that it elicits DNA damage response in both S and G₂ phases.

The DNA damage response elicited by trabectedin recruits proteins from both the NER and homologous recombination machineries. In order to understand the type of DNA damage response evoked by the drug, we tested the sensitivity to trabectedin of a variety of mutants in different DNA damage repair pathways [NER, UV excision repair and homologous recombination (HR) repair]. We used the following NER mutants: rph14Δ (an orthologue to the mammalian XPA and likely involved in lesion recognition), swi10Δ and rad16Δ [orthologues to mammalian ERCC1 and XPF, endonucleases involved in single-stranded DNA (ssDNA) incision 5′ to the DNA lesion], and rad13Δ (an endonuclease orthologue to mammalian XPG that is involved in ssDNA incision 3′ to the lesion; ref. 18). In the UV excision repair (UVER) pathway, we used mutant strains, uve1Δ and rad2Δ, which lack ssDNA 5′ and 3′ endonucleases for this repair pathway, respectively (19). Finally, in the HRR pathway, the rad51Δ, rad22Δ, and rad54Δ strains were used. Rad51 is required in the initial steps of HRR for homologous pairing and DNA strand exchange (20). Rad22 is the fission yeast counterpart of the mammalian DNA-binding protein, Rad52, that helps to load Rad51 onto the recombination site and also promotes annealing of complementary ssDNA (21). Rad54 is a dsDNA-stimulated ATPase homologue of mammalian Rad54 (22) that is involved in DNA remodeling and whose activity is stimulated by Rad51 (23).

Strains disrupted for rad2 or uve1, both required for UVER, were as sensitive to the drug as wild-type cells (Fig. 3A). However, mutants impaired in HRR (rad51Δ and rad54Δ) were extremely sensitive to trabectedin (Fig. 3A). They were hypersensitive even at 1 μmol/L of trabectedin, a concentration at which wild-type cells do not show any phenotype on plates. The hypersensitivity of these mutants to the drug suggests that trabectedin might give rise to DSB. In striking contrast, mutants in genes involved in NER, except rhp14Δ, were more resistant to the drug. In particular, the rad13Δ mutant (lacking the XPG orthologue) was the most resistant (Fig. 3A). Sensitivity to trabectedin was recovered when rad13Δ cells were transformed with a multicopy plasmid carrying the rad13 gene (Fig. 3B). These effects were specific for trabectedin because they were not observed for MMS (Fig. 3B). Whereas the rad13 deletion conferred resistance to trabectedin, rad13 overexpression induced hypersensitivity to the drug (Fig. 3C). The decreased sensitivity of this particular NER mutant to trabectedin has also been reported in mammalian cells (9) and in Saccharomyces cerevisiae (17), and suggests that Rad13 (XPG) is somehow involved in generating the DNA damage signal elicited upon trabectedin treatment.

Trabectedin requires rad13 to induce DNA damage and to activate the DNA damage checkpoint. The partial resistance of NER mutant cells to trabectedin indicates that it is more deleterious for drug-treated cells to have a functional rad13 gene. Thus, Rad13 could be involved in generating the DNA damage and/or in checkpoint activation. To test this hypothesis, we first analyzed changes in cell size as an indication of the cell cycle delay induced by the drug during a time course experiment comparing wild-type and rad13Δ strains. rad13Δ mutant cells elongated much less and showed a better appearance under the microscope than trabectedin-treated wild-type cells, (Fig. 4A and B). This result indicates that the cell cycle delay induced by trabectedin is almost completely abolished when rad13 is not present in the cell. We then measured the activation of the DNA damage checkpoint following the phosphorylation status of the Chk1 kinase. Chk1 phosphorylation was visualized by Western blot as a new band running above the unphosphorylated form when cells were treated with MMS for 3 hours (Fig. 4C). Deletion of rad13 had no effect on the activation of Chk1 because the same mobility shift was detected when cells were treated with MMS for 3 hours (Fig. 4C, compare lanes 2 and 5). However, a reduction in Chk1 activation was observed in the rad13Δ mutant in the presence of trabectedin (Fig. 4C, compare lanes 3 and 6).

To further study the role of Rad13, we analyzed the sensitivity to trabectedin of the double mutant rad51Δ rad13Δ. If Rad13 is somehow involved in generating DNA damage in cells treated with the drug, deletion of rad13 should suppress the hypersensitivity of rad51Δ mutant cells because the damage that Rad51 repairs is eliminated. rad51Δ rad13Δ double mutant cells were found to be more resistant to trabectedin than the single rad51Δ mutants (Fig. 4D). However, the double mutant was still more sensitive to the drug than the single rad13Δ (Fig. 4D). Similar to what happens in a wild-type background, the deletion of rad13 abolished the cellular phenotypes (elongation and lysis) induced by trabectedin in the rad51Δ background (data not shown). This result indicates that Rad13 is essential for the induction of DNA damage, although in its absence, there are still some DNA lesions that require repair by HR.

Trabectedin requires Rad13, but not its nuclease activity, to induce DNA damage. The NER 3′ endonucleases Rad13, human XPG, and S. cerevisiae Rad2 are members of a large nuclease family that also includes the structure-specific endonuclease-1 (FEN-1), an enzyme that specifically recognizes the 3′ DNA flap structure that has been proposed to exist during replication, repair, and recombination (24). All these proteins possess three conserved regions: two of them highly conserved at the NH₂ terminus (N) and the internal region (I), and a moderately conserved COOH-terminal region (C) (Fig. 5A). Different authors have shown that a mutation to alanine of a glutamate residue in the highly conserved sequence EAEA in region I completely abolishes the catalytic activity of XPG, Rad2, and FEN-1 (25, 26). Because it has been proposed that DNA damage induced by trabectedin in mammalian cells results from NER endonuclease activity, producing lethal SSBs (9), we decided to mutagenize the S. pombe Rad13E779 residue to alanine (equivalent to the human XPGE791A, Rad2E794A, and FEN-1E160A mutations) and analyze the effect of the mutation on sensitivity to trabectedin. The rad13E779A mutant allele was used to replace rad13 in the fission yeast chromosome and the sensitivity to trabectedin was analyzed (Fig. 5C). In the same experiment, we also checked the sensitivity of this S. pombe strain to MMS. NER deletion mutants were more sensitive to MMS (Fig. 5C), as previously reported (27). The fact that the strain carrying the mutant version rad13E779A (Fig. 5C , lane 3) was more sensitive to both, MMS and 4NQO, a drug that mimics the effect of UV light, as compared with the wild-type strain (lane 1) indicated that the mutation of E779 was affecting the ability of Rad13 to excise in 3′, as expected. However, this mutation did not confer resistance to trabectedin (plates + trabectedin, lane 3) unlike the deletion of rad13 (lane 2), indicating that the endonuclease activity of Rad13 is not required for the DNA damage induced by trabectedin. Moreover, we found that the E779A mutation increased the sensitivity of the cells to the drug. This observation suggests that the endonuclease activity of Rad13 could actually contribute to the repair of the DNA damage induced by trabectedin.

Rad13 COOH terminus is a target for trabectedin. Several lines of evidence have indicated that the COOH-terminal part of XPG exerts an endonuclease-independent function (28, 29); therefore, we decided to explore whether the COOH terminus of Rad13 was somehow contributing to trabectedin sensitivity. For this purpose, a S. pombe mutant strain with a Rad13 COOH-terminal truncation was constructed, and sensitivity to trabectedin was analyzed. Because XPG has been reported to contain in its COOH terminus nuclear localization signals (NLS) that mediate nuclear localization (30), and because NLSs also exist in FEN-1, Rad2, and Rad13 (Fig. 5A), we decided to delete most of the COOH-terminal domain of Rad13, except for a short region (amino acids 1,047-1,112) that contained the NLSs KRRR, RRRK and the bipartite signal RKTKLSTSLLKKPSKRR. We found that the rad13CΔ strain (Fig. 5C,, lane 4) was as resistant to trabectedin as the rad13 null mutant (Fig. 5C , lane 2), indicating that this part of the protein is essential for trabectedin cytotoxicity.

In order to identify which region within the COOH-terminal part of Rad13 was responsible for the phenotype exhibited by rad13CΔ, we searched for functional domains that had been previously characterized in this family of nucleases. This search gave us two possible candidates (Fig. 5A), (a) a domain previously described in FEN-1 and XPG to bind proliferating cell nuclear antigen (PCNA), a ring-shaped homotrimeric protein that encircles DNA and acts as a “sliding clamp” which links the polymerase to the DNA template (31, 32), and (b) a domain reported in FEN-1 to form two α-helices involved in binding and recognizing double-stranded DNA associated with a 3′ flap, as observed in the X-ray crystal structure of a FEN-1/DNA complex (32). To analyze the contribution of each of these two domains, we constructed the following S. pombe strains: rad13PΔ, lacking amino acids 970 to 1000 of Rad13, which include the putative PCNA binding site, and rad13DΔ, lacking amino acids 925 to 969, encompassing the putative DNA binding domain (Fig. 5A and B). When the rad13PΔ mutants were assayed against trabectedin, we clearly observed that the presence of the mutated protein did not increase trabectedin resistance (Fig. 5C,, lane 5). Moreover, as shown above for the rad13E779A strain, the expression of Rad13PΔ lead to a loss of endonuclease activity (detected as increased sensitivity to MMS and 4NQO as compared with the wild-type strain), which resulted in increased sensitivity to trabectedin as compared with the wild-type strain (Fig. 5C,, lane 1). On the other hand, the rad13DΔ strain (Fig. 5C , lane 6) was found to be as resistant to trabectedin as the rad13Δ mutant strain. In conclusion, the PCNA binding region of Rad13 is dispensable, whereas a short COOH-terminal region (amino acids 925-969) is essential for trabectedin-induced cytotoxicity.

Pinpointing a conserved arginine in Rad13 as an important residue for resistance to trabectedin. To understand the molecular basis through which the COOH-terminal region of Rad13 is involved in the mechanism of resistance to trabectedin, we built a model of Rad13, in complex with a double-stranded DNA molecule containing a 3′ flap, on the basis of its homology to FEN-1 (32). The model revealed that Arg961 from α15 in Rad13 occupies a position equivalent to that of Arg314 in FEN-1, which is the only residue that points toward the minor groove. Because trabectedin binds covalently to guanines in the DNA minor groove, we speculated that Arg961 might be interacting with the drug, thereby stabilizing a Rad13/DNA-trabectedin ternary complex. To test this hypothesis, the trabectedin molecule, which is composed of three fused tetrahydroisoquinoline rings (Fig. 6A), was incorporated into the Rad13/DNA complex, as described in Materials and Methods. A suitable location for the drug was found three bases upstream from the flap. In this ternary complex (Fig. 6B), we observed that a hydrogen bond could indeed be formed between Arg961 and subunit C of trabectedin, and that it was later maintained during the ensuing molecular dynamics simulation in aqueous solution. In the equilibrated complex, Rad13 is seen to recognize the widened minor groove of the DNA molecule and simultaneously interact with the guanine-bonded drug. One side of the C subunit of trabectedin makes extensive van der Waals contacts with the sugar-phosphate backbone of the two nucleotides downstream from the covalently bonded nucleotide, whereas the other side is exposed to the solvent and remains close to the guanidinium group of Arg961. Hence, it seems feasible that the side chain of this amino acid would establish good hydrogen-bonding interactions with both the hydroxyl and methoxy oxygens present in this subunit, which protrudes out of the minor groove, thereby contributing to ternary complex stabilization.

To analyze the effect of the Arg961 residue on the cytotoxicity exerted by trabectedin, a mutant strain carrying the R961A mutation was created, and sensitivity to trabectedin and MMS was analyzed. The fact that the rad13R961A strain was found to be as resistant to MMS and 4NQO (Fig. 5C,, lane 7) as the wild-type strain (Fig. 5C,, lane 1) clearly indicated that the mutated protein is functional and does not affect the efficiency of DNA repair. However, we found that this point mutation conferred resistance to trabectedin (Fig. 5C , plates + trabectedin, lane 7), indicating that the drug requires this particular arginine residue of Rad13 (Arg961) to induce its cytotoxic effect.

To verify that the Rad13 mutant proteins had been expressed, we constructed COOH-terminal HA-tagged versions of Rad13 wild-type, point mutants, and COOH-terminal deleted proteins. We then analyzed the expression of these proteins by Western blotting and their functionality by assaying the sensitivity of the strains to trabectedin and MMS (Supplemental Fig. S2). HA-tagged proteins were detected in all the strains constructed (Fig. 5D), and the strains behaved like their non-tagged counterparts in terms of sensitivity to both trabectedin and MMS.

In the present study, we took advantage of S. pombe genetics to clarify important aspects regarding the mechanisms of action of the antitumor drug trabectedin. Here, we report that trabectedin activates the G₂-M and S phase DNA damage checkpoints mediated, respectively, by the effector kinases Chk1 (Fig. 2A and B) and Cds1 (Fig. 2C), in good agreement with the S phase delay and G₂-M block reported in human cells (8, 33–35). Furthermore, treatment of S. pombe cells with trabectedin also produces concentration-dependent phenotypes: at low concentrations, the drug induced cell cycle delay in S phase and a G₂-M block, whereas at higher concentrations, the cells treated with the drug stopped growing and died (Figs. 1 and 2A; data not shown). Similar cytostatic (at low doses) and proapoptotic (at high doses) effects have been described for cancer cells treated with trabectedin (34).

Role of NER in trabectedin cytotoxicity. NER proteins are involved in repairing the DNA damage caused by many DNA-binding drugs commonly used in cancer treatment, such as nitrogen mustard, cisplatin, or mitomycin C (36). As a consequence, defects in the NER repair system usually increase drug sensitivity (1, 37). The widely reported finding that defects in NER decrease the sensitivity to trabectedin in both Chinese hamster ovary (7) and human cell lines (9) is somewhat paradoxical, whereas the proposal that trabectedin causes DNA breaks remains controversial (8, 9). By analyzing the sensitivity to trabectedin of a panel of NER mutants, we now show that rad13 mutant cells are much more resistant to trabectedin than the wild-type strain, in good agreement with previous results. This mutant is more resistant than strains deficient in the endonucleases that cut on the 5′ side of the lesion, Swi10 and Rad16, which are orthologues of ERCC1 and XPF, respectively (Fig. 3). This result indicates that although the three endonucleases could play a role in the cell-killing mechanism induced by trabectedin, Rad13 is the most important one. In the absence of this structure-specific endonuclease, the cells underwent much less DNA damage (as measured by the requirement for Rad51, cell cycle delay, and Chk1 activation) and survived much better than the wild-type cells exposed to the drug (Fig. 4). In contrast, increasing levels of Rad13 led to a higher sensitivity to the drug (Fig. 3C). However, the finding that rad13E779A nuclease-deficient cells were not resistant (Fig. 5C) clearly indicates that the ability of Rad13 to produce SSBs is not required for the DNA damage induced by trabectedin, in contrast to a previous proposal (9). Instead, we found that it was the COOH-terminal region of the protein that confers resistance to trabectedin (Fig. 5) and—more specifically—part of its DNA-binding domain, which is highly homologous to one of the several DNA-binding domains described for the nuclease FEN-1 (ref. 32; Fig. 5). By homology modeling a Rad13/DNA-trabectedin ternary complex, we found that Arg961 (positionally and functionally equivalent to Arg314 of FEN-1) was likely to contribute to the stabilization of the Rad13/DNA-trabectedin complex (Fig. 6B). In agreement with this hypothesis, an S. pombe strain carrying a R961A point mutation in rad13 that did not affect its nuclease activity (Fig. 5C,, plate + MMS and plate + 4NQO, lane 7) was found to be strongly resistant to the drug (Fig. 5C).

The formation of a putative Rad13/DNA-trabectedin cytotoxic ternary complex is consistent with earlier proposals that a protein/DNA-trabectedin intermediate in the NER processing of trabectedin-DNA adducts could be trapped and give rise to the formation of cytotoxic complexes (17, 38). DNA repair inhibition, rather than excision of the lesion, due to direct immobilization of NER factors by another type of adduct, has also been reported (39). However, the formation of other complexes between one or more DNA-trabectedin adducts and DNA repair proteins other than Rad13 cannot be ruled out.

Role of HR in the repair of the DNA damage caused by trabectedin. DSBs are serious DNA lesions, and if not repaired properly, may lead to important DNA aberrations. Two main pathways repair DSBs in eukaryotic cells: HR and nonhomologous end joining. HR is carried out by the RAD52 group of genes (40). Mutation of any of these genes results in sensitivity to DSB-inducing agents.

Here, we show the extraordinary sensitivity to trabectedin (Fig. 3) of deletion mutants in the RAD52 epistasis group. Because the proteins encoded by these genes are responsible for most DSB repair in eukaryotic cells, this finding can be taken as a clear indication that the drug could give rise (directly or indirectly) to DSBs. These results are in agreement with previous reports showing that trabectedin was more effective in cells lacking DNA-dependent protein kinase (involved in nonhomologous end joining). The fact that the absence of rad13 partially rescued the great sensitivity to trabectedin of rad51Δ cells (Fig. 4D) indicates that Rad13 is somehow involved in the induction of DSBs, which could be an indirect effect like the one described for the simple alkylating agent MMS (41, 42). HR can also be induced by spontaneous SSBs probably induced by collapsed replication forks during normal DNA replication (43). Against the possibility that, in the absence of a proficient NER system (rad13Δ), other repair systems might be activated to compensate for this absence, thereby accounting for the better survival of rad51Δ rad13Δ cells, is the finding that the double mutant rad51Δ rad13Δ is more sensitive to MMS than the single rad51Δ mutant (44). This indicates that, at least in the case of MMS, no additional repair system compensates for the lack of a proficient NER.

Model for the mechanism of action of trabectedin in eukaryotic cells and relevance to cancer. Taking together the results reported here and others published in the literature, we suggest the following sequence of events for the action of trabectedin in eukaryotic cells: (a) trabectedin binds covalently to the DNA minor groove and the adduct is recognized by the NER system; (b) the recruited Rad13 (XPG) protein binds to DNA and simultaneously interacts with the minor groove-bound drug by means of Arg961; (c) the proteins making up the Swi10-Rad16 (ERCC1-XPF) complex are the last NER factors to arrive and, together with proteins from other DNA repair pathways that try to correct the DNA lesions, are hijacked at the sites of damage, creating stronger cytotoxic complexes; (d) during S phase, these complexes give rise to DNA lesions that need to be repaired by HR (e.g., DSBs) so that HR-proficient cells can repair the damage, whereas those defective in one or more HR proteins, a common observation in several solid tumors (40), will be especially sensitive to the action of the drug. In view of this likely scenario, our results may have important implications for the optimal use of trabectedin in cancer therapy because patients harboring tumor cells with proficient NER and deficient HR systems would be expected to respond best to the treatment. This hypothesis is currently being tested in clinical trials.

Grant support: PharmaMar (Madrid, Spain).

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 Antony Carr, Albert Pastink, Gerald Smith, and Oliver Fleck for providing several S. pombe strains; as well as Jose María Fernández-Sousa, Carmen Cuevas, Juan Carlos Tercero, José Jimeno, and the members of our lab for interesting discussions during the course of this work.