The Antitumor Drugs Trabectedin and Lurbinectedin Induce Transcription-Dependent Replication Stress and Genome Instability

Cancer therapy greatly relies on the use of small molecules that directly perturb the DNA metabolism. Trabectedin (ET743) is one of such molecules. It has been initially derived from the sea squirt Ecteinascidia turbinata and is now produced synthetically, being currently used in therapy of patients with advanced soft tissue sarcoma (1) and platinum-sensitive ovarian cancer (2). Different derivatives of trabectedin have been synthesized, among them lurbinectedin (PM01183), which is presently under evaluation in phase II and III clinical trials.

One of the most important features of the mechanism of action of trabectedin and lurbinectedin is their inhibitory effect on transcription of protein-coding genes, which is mediated by different means. First, they preferentially bind to specific DNA triplets that are often present on transcription recognition sites. In this way they could directly prevent loading of transcription factors onto chromatin (3). Moreover, both drugs were shown to induce degradation of the RNA polymerase II (RNAPII) through the ubiquitin–proteasome pathway (4–8). Of note, these effects are also observed in tumor-associated macrophages (TAM). Indeed, both drugs inhibit the transcription of selected cytokines (e.g., CCL2, IL6, IL8, PTX3) by TAMs abrogating their protumoral properties and modifying the tumor microenvironment (9, 10).

Tumor cells are commonly associated with genome instability. One important natural source of instability is represented by R-loops, a three-stranded nucleic acid structure consisting of an RNA–DNA hybrid and the displaced single-stranded DNA of the original DNA duplex (11). This structure could fulfill physiologic roles as it occurs during immunoglobulin class switching recombination, plasmid or mitochondrial DNA replication, and transcription regulation. However, they are an important source of DNA damage, genome instability, and replication stress, which relates R-loops with a number of neurodegenerative disorders and cancer (12, 13). Interestingly, the loss of function of the breast cancer susceptibility genes BRCA1 and BRCA2, which have been shown to increase R-loops and R-loop-dependent DNA damage (14, 15), is also associated with an increased sensitivity to trabectedin (16). In a similar way, dysfunctions of the Fanconi anemia (FA) pathway that increased sensitivity to such agent (5) also cause accumulation of R-loops (17, 18).

In addition, there are some similarities between R-loops and trabectedin adducts worth being considered. The head-to-tail binding of three trabectedin molecules to three adjacent optimal DNA binding sites changes the DNA duplex conformation from the B-form to an intermediate between the A- and the B-form, which strongly resembles the conformation of an RNA–DNA hybrid (19). Moreover, R-loops that accumulate in the absence of the putative RNA–DNA helicases AQR or SETX require the action of nucleotide excision repair (NER) to be processed into double-strand breaks (DSB; ref. 20) as well as trabectedin adducts (21).

For these reasons, in this study we have further explored the mechanism of action of trabectedin and lurbinectedin, in particular whether it could be partially mediated or affected by R-loops. We show that the DNA damage and replicative stress caused by trabectedin and lurbinectedin is indeed partially dependent on R-loops, as demonstrated by molecular and genetic assays using both human cell lines and the yeast Saccharomyces cerevisiae as a model eukaryotic system. Our findings suggest that R-loops could be targets of anticancer agents and given the evidence that R-loops occur all over the genome (22), this opens the possibility that tumoral cells that accumulate R-loops may be more suitable for trabectedin and lurbinectedin treatment.

HeLa cells were purchased from European Collection of Cell Cultures (ECACC), HeLa-TR (23), and HeLa HB-GFP (14) were previously generated in our laboratory. All cell lines were maintained in DMEM medium, supplemented with 10% heat-inactivated FCS, cultured at 37°C in a humidified atmosphere containing 5% CO₂, and routinely tested for mycoplasma using MycoAlert Mycoplasma Detection Kit (Lonza). HeLa HB-GFP cells medium was supplemented with doxycycline at the final concentration of 2 μg/mL to induce HB-GFP expression. Specific genes were knocked down using ON-TARGET SMARTpool siRNA from Dharmacon. Transient transfection of siRNA was performed using DharmaFECT 1 (Dharmacon), according to the manufacturer's instructions. The following plasmids were used for transfection: pcDNA3 (Invitrogen) (RNH1−) and pcDNA3-RNaseH1 (RNH1+), containing the full-length RNase H1 cloned into pcDNA3 (24). Lipofectamine 2000 (Invitrogen) was used for plasmid transfection. Assays were performed 48 or 72 hours after transfection. Trabectedin (ET743) and lurbinectedin (PM01183) were provided from Pharmamar. Cordycepin (C3394) and neocarzinostatin (NCZ; N9162) were purchased from Sigma.

Single-cell electrophoresis or comet assays were performed as described using a commercial kit (Trevigen) following the manufacturer's instructions. Comet tail moments were analyzed using Open Comet software. At least 100 cells were scored in each experiment to calculate the median of the tail moment, as reported (23).

HeLa cells were siRNA transfected to knock down the specified genes and plated in 96-well plate. Following 72 hours, cells were treated with increasing doses of trabectedin (ET743) for 24 hours and cell proliferation was measured by the WST-1 reagent (Roche) following the manufacturer's instructions. Absorbance was measured using microplate (ELISA) reader VARIOSKAN FLASH (Thermo) at 450 nm with a reference wavelength of 690 nm. Absorbance values were normalized to the value of the untreated and represented as arbitrary units (A.U.).

Antibodies anti-RNASEH1 (15606-1-AP) and anti-FANCD2 (sc-20022) anti-53BP1 (NB100-304) were purchased from Proteintech, Santa Cruz Biotechnology, and Novus Biologicals, respectively. The S9.6 antibody was purified from the hybridoma cell line HB-8730.

Immunofluorescence was performed as previously described (25). Briefly, cells were fixed with 3.7% formaldehyde in PBS for 15 minutes, permeabilized with 0.5% Triton X-100, and blocked with 3% BSA in PBS for 1 hour. The coverslips were then incubated with primary antibodies diluted 1:500 in 3% BSA in PBS for 2 hours followed by 1 hour incubation with 1:1,000 diluted secondary antibodies and nuclei counterstaining with 1 μg/mL DAPI in PBS. For FANCD2 immunolabeling, before fixation cells were pre-permeabilized with 0.25% Triton X-100 in PBS for 1 minute on ice. S9.6 immunofluorescence was performed as previously described (20). For the EdU labeling, cells were pulse-labeled with 10 μmol/L EdU for 20 minutes before fixation. EdU staining was performed with a Click-iT EdU Alexa Fluor 555 Imaging Kit (Invitrogen) according to manufacturer's instructions. Random images were acquired with a 63× or 40× objective and foci or nuclear intensity were scored using the MetaMorph software. At least 100 cells were scored for each condition.

DNA combing was performed as previously described (23). Briefly, DNA fibers were extracted from cells in agarose plugs immediately after CldU labeling and were stretched on silanized coverslips. DNA molecules were counterstained with an autoanti-ssDNA antibody (DSHB, 1:500) and an anti-mouse IgG coupled to Alexa Fluor 647 (A21241; Molecular Probes; 1:50). CldU and IdU were detected with BU1/75 (AbCys; 1:20) and an anti-rat IgG coupled to Alexa Fluor 488 (A21470; Molecular Probes; 1:50) or B44 (Becton Dickinson; 1:20) anti-BrdU antibodies and an anti-mouse IgG coupled to Alexa Fluor 546 (A21123; Molecular Probes; 1:50), respectively. DNA fibers were analyzed on a Leica DM6000 microscope equipped with a DFC390 camera (Leica). Data acquisition was performed with LAS AF (Leica). Fork velocity was calculated as previously described (26). Replication asymmetry was calculated by dividing (longest green tract –shortest green tract) by the longest tract in divergent CIdU tracks.

Yeast cells of the wild-type W303 or the rad52Δ strains were plated in presence of 0.003% SDS in order to increase yeast cell permeability. Afterwards, paper filter disks were placed on the agar plates and 3 μL of increasing amounts of trabectedin or lurbinectedin (or DMSO as carrier control) were spotted onto them. After 3 to 6 days cells growth was monitored, plates were scanned, and the diameter of the growth inhibition halo was measured.

Wild-type W303 yeast cells were transformed in plates containing doxycycline (at the final concentration of 10 μg/mL in order to repress the expression of ectopic RNase H1) with pWJ1344 plasmid coding the RAD52–YFP fusion protein (27) and with pCM189-RNH1 with the RNH1 gene under the tet promoter (28) or the empty vector pCM189 (29). Cells were grown overnight with medium SC-L-U (Complete medium SC: 0.17% yeast nitrogen base and 0.5% ammonium sulfate, 2% glucose, supplemented with amino acids; –L and –U indicate the absence of leucine and uracil, respectively) containing 0.003% SDS and without doxycyclin. Mid-log cultures overexpressing RNase H1 or not, were treated with increasing doses of trabectedin or lurbinectedin during 2 hours at 30°C with 200 rpm shaking. Following fixation with 2.5% formaldehyde in 0.1M KHPO at pH 6.4 for 10 minutes, the cells were washed twice with KHPO 0.1M pH 6.6 and once with KHPO 0.1M pH 7.4. Cells were permeabilized with ethanol 80% for 10 minutes and after centrifugation were resuspended in water containing DAPI at the final concentration of 1 μg/mL to counterstain the nuclei. Rad52–YFP foci were scored in more than 200 nuclei from S-G2 mid-log cells, using a Leica DC 350F microscope.

Statistical significance was determined by t tests or the Mann–Whitney test as specified. All tests performed were two-tailed unless otherwise specified and three levels of statistical significance were considered: *, P ≤ 0.05; **, P ≤ 0.01 and ***, P ≤ 0.001.

Trabectedin and lurbinectedin covalently bind to DNA, physically impeding transcription and other DNA metabolic processes. To explore the possibility that the DNA adducts formed by these drugs could induce transcription-dependent genome instability, we analyzed DNA damage foci accumulation in HeLa cells in which transcription was abrogated by treatment with cordycepin, a potent transcription inhibitor that halts RNA chain elongation. HeLa cells treated during 4 hours with 25 nmol/L of trabectedin or lurbinectedin accumulated 53BP1 foci, a DSB marker (Fig. 1A). However, when transcription was previously shut off with cordycepin (added 1 hour earlier), 53BP1 foci were significantly reduced, suggesting an important role of transcription in the mechanism of action of these drugs.

Next, we assayed whether RNA–DNA hybrids could be involved in the transcription-dependent DNA damage observed. Indeed, overexpression of RNase H1, which removes the RNA moiety of the hybrids, partially reduced the 53BP1 foci observed after trabectedin treatment, suggesting that part of the role of transcription in the mechanism of action of these drugs was associated with R-loops (Supplementary Fig. S1). In this case, we exposed cells to a lower dose of trabectedin for longer (10 nmol/L during 6 hours) to better see the increase of 53BP1 foci. This longer treatment was not possible with cordycepin-mediated transcription shut off.

Provided the known negative effect of transcription and RNA–DNA hybrids on replication fork (RF) progression, we decided to analyze the accumulation of FANCD2 foci, a central player of the FA DNA repair pathway, as a way to measure repair events occurring at potential stalled forks. FA is a DNA repair pathway that works on stalled RFs and cells deficient in this pathway have also been shown to be more sensitive to trabectedin (5). The experiment was performed as described above for 53BP1 foci. Trabectedin and lurbinectedin caused a significant increase of FANCD2 foci that was completely reversed by cordycepin treatment (Fig. 1B), suggesting a key role of these drugs in transcription–replication (T–R) conflicts. Because both drugs caused very similar phenotypes, consistent with an expected common mechanism of action, we focused the rest of the study mainly on trabectedin.

FANCD2 foci accumulation approximately doubled after a longer exposure to trabectedin at lower dose and this increase was entirely rescued by RNase H1 overexpression (Fig. 1C). To assure that this suppression by RNase H1 was specific for trabectedin, we analyzed the effect of a different and well-establish genotoxic and antitumoral drug, NCZ (30), and we found that it also increased FANCD2 foci but this phenotype was not suppressed by RNase H1 overexpression (Supplementary Fig. S2). Given the observed connection of RNA–DNA hybrids and DNA damage with trabectedin and lurbinectedin treatment, we tested directly for the presence of RNA–DNA hybrids using the S9.6 antibody, which specifically recognizes these structures, by in situ immunofluorescence (IF). As can be seen in Supplementary Fig. S3, trabectedin and lurbinectedin did not cause a significant increase in the nuclear intensity of the S9.6 signal. Because these drugs significantly decrease transcription and therefore the overall RNA levels (8, 31) it is unlikely that they concomitantly increase R-loop levels. Nonetheless, they could act by exacerbating their impact on replication and genome integrity. Consistently, both drugs affected nucleolus integrity, as deduced from a perturbed localization of the nucleolar marker nucleolin and by a smaller or fragmented appearance of nucleoli detected with the S9.6 antibody.

Trabectedin directly interacts with DNA, forming adducts and producing distortions of the double helix. Because R-loops impair replication progression (32–36), we analyzed the impact of trabectedin on DNA replication and whether this would be associated with the presence of RNA–DNA hybrids. For this, we first assessed DNA replication levels after trabectedin treatment, by measuring the amount of incorporated EdU, a thymidine analog. Trabectedin treatment strongly affected the rate of DNA synthesis in HeLa cells, leading to lower levels of EdU incorporation compared with untreated cells (Fig. 2A). However, the percentage of cells in S phase (EdU-incorporating cells) was not affected (Supplementary Fig. S4A). Importantly, overexpression of RNase H1, which by itself caused a slight decrease in EdU incorporation, partially rescued the phenotype of reduced DNA synthesis (Fig. 2A).

Next, we used DNA combing to check for impairment of replication fork progression in single molecules. Using low trabectedin concentrations at which the replication fork velocity was not perturbed, the drug caused an increase in fork stalling, as determined by fork asymmetry, which was significantly higher in trabectedin-treated versus nontreated cells (Fig. 2B). To know whether this effect was dependent on transcription, required for the cotranscriptional formation of RNA–DNA hybrids, we inhibited transcription using cordycepin. With this approach, all cells were exposed to the same treatment, something that was not possible after transfection with an RNase H1-overexpressing plasmid. We observed that treatment of cells with the transcription inhibitor cordycepin abolished the difference in fork asymmetry between trabectedin-treated and untreated samples. This result suggests that fork-stalling caused by trabectedin is linked to transcription. Inhibition of transcription by cordycepin was detrimental for replication at higher doses of trabectedin, which per se provoked a reduction in the replication rate (Supplementary Fig. S4B). Thus, it was not surprising that the concomitant treatment with the two chemicals reduced further the DNA replication rate as assessed by EdU incorporation. However, because trabectedin can also inhibit transcription (31), it is possible that the effect on DNA replication has multiple causes that would require further investigation.

Based on these observations, we infer that the replication obstacles posed by trabectedin treatment are partially dependent on RNA–DNA hybrids.

To better establish a functional link between RNA–DNA hybrids and trabectedin, we assayed the sensitivity to trabectedin of cells in which RNA–DNA hybrids were stabilized by a hybrid-binding protein. For this purpose we took advantage of a stable cell line that expresses the chimeric protein HB-GFP, a fusion protein constituted by the hybrid binding (HB) domain of the RNase H1 enzyme and the GFP. Under these conditions RNA–DNA hybrids are likely “stabilized” by the binding of the HB domain, thus retaining the R-loops in the cell, consistent with previously reported results from BRCA2-depleted cells (14). Interestingly, cells expressing the HB–GFP fusion protein were more sensitive to trabectedin as assessed by cell proliferation assays (Fig. 3, left), suggesting that cells with higher levels of R-loops can be made more susceptible to this anticancer agent. We next knocked down two cellular factors, THOC1 and BRCA2, previously shown to enhance RNA–DNA hybrids by either preventing their formation or by promoting their removal. Depletion of any of the two genes is associated with an increase in RNA–DNA hybrids formation (14). The lack of BRCA2 or THOC1 increased cellular sensitivity to trabectedin both in the absence of HB-GFP and in cells expressing HB–GFP (Fig. 3, middle and right), having THOC1 depletion a stronger effect.

Next, we analyzed the effect of trabectedin treatment in cell proliferation and genome integrity in human cells in relation to R-loop accumulation. For this, HeLa cells were first transfected with a plasmid that leads to the overexpression of RNase H1 before being treated with trabectedin or lurbinectedin and analyzed by alkaline single-cell electrophoresis (comet assay) for the accumulation of DNA breaks. As shown in Fig. 4, trabectedin and lurbinectedin increased DNA breaks, as expected. However, when cells overexpressed RNase H1, breaks were significantly reduced in cells treated with 50 nmol/L lurbinectedin for 2 hour. We evaluated γH2AX foci formation, a molecular marker that in addition to DNA damage is also associated with replicative stress signaling. Trabectedin treatment increased γH2AX foci accumulation (Supplementary Fig. S5A). In this case, however, RNase H1 treatment increased further the γH2AX foci, suggesting that a double treatment with trabectedin and RNase H1 leads to an additive replicative stress and DNA damage. Indeed, RNA–DNA hybrids are also important intermediates of normal DNA replication (for Okazaki fragment synthesis) and, consistently, a clonogenic assay revealed that survival of cells overexpressing RNase H1 at the time of treatment with increasing doses of trabectedin was reduced (Supplementary Fig. S5B).

To investigate whether trabectedin and lurbinectedin interact with RNA–DNA hybrids regardless of cell type and organism as a way to confirm that they target directly DNA metabolism, we asked whether similar DDR phenotypes could be reproduced in the eukaryotic model organism Saccharomyces cerevisiae. For this we first analyzed the effect of trabectedin. Replication impairment causes genome instability and Rad52 is used as a marker to detect instability in the form of the accumulation of DNA damage. To assess sensitivity to trabectedin and lurbinectedin, we used the “halo” assay. Different yeast strains were homogenously plated on YEPD-rich medium plates, and then Whatman-paper disks soaked with increasing concentrations of the tested drug were placed on top of the plate. After growth, halos of growth inhibition around the disks were formed with a diameter that correlates with the levels of toxicity of the drug. As can be seen in Fig. 5A, increasing concentrations of trabectedin lead to increased yeast sensitivity, as detected by the size of the halo of growth inhibition around the trabectedin spot. Importantly, this halo was largely increased in rad52Δ yeast strains inactivated for homologous recombination DSB repair in comparison with the wild-type strain, even at concentrations at which trabectedin had no effect on wild-type cells (Fig. 5A). These result suggest that recombinogenic DNA breaks accumulated after trabectedin treatment, consistent with previous reports (21). To confirm this, we analyzed recombinogenic DSBs by Rad52 foci, used as a way to detect DSB repair centers (27). Consistently, trabectedin caused a significant increase of Rad52 foci. Cells treated with 250 μmol/L trabectedin, showed a three-fold increase in the percentage of cells with Rad52 foci compared with the untreated control (Fig. 5B). To assay whether Rad52 foci were dependent on RNA–DNA hybrids, RNAse H1 was overexpressed in yeast cells by transformation with an RNase H1-overexpressing plasmid. Notably, the increase in Rad52 foci was suppressed by RNase H1 overexpression (Fig. 5B; Supplementary Fig. S6A). These results indicate that trabectedin-induced breaks are linked to the accumulation of RNA–DNA hybrids also in yeast cells.

Next, we assayed the effect of lurbinectedin on cell growth and genome instability using the assays described above for trabectedin. As can be seen in Fig. 5C, lurbinectedin affected yeast growth as determined by the size of the growth inhibition halo formed around the increasing concentrations of the chemical spotted on Whatman filter papers. Again, this effect was significantly enhanced in rad52Δ strains, consistent with the idea that cell growth was impaired due to the accumulation of unrepaired DSBs. Accordingly, lurbinectedin treatment caused a two- to three-fold increase of cells with Rad52 foci compared with untreated cells and this phenotype was fully suppressed by RNase H1 overexpression (Fig. 5D; Supplementary Fig. S6A). The decreased amount of Rad52 foci after overexpression of RNase H1 in trabectedin- or lurbinectedin-treated cells was not due to the reduction of the protein levels, as confirmed by Western blot analysis (Supplementary Fig. S6B).

Consequently, lurbinectedin, as well as trabectedin, caused R-loop–dependent genome instability both in yeast and in human cells. This finding suggests that their action is most likely common to many different cell types, because they target molecular pathways that are well conserved across evolution.

Increasing evidence points out the role of RNA–DNA hybrids in the boost of genome instability and replication stress that are major hallmarks of cancer cells. This opens the possibility that such hybrids could occur at high levels in some tumor cells and consequently could be used as therapeutic targets. For this reason, it is of growing importance to address the relationship between R-loops and the action of anticancer agents. Trabectedin and lurbinectedin are two powerful drugs used in cancer treatment whose ability to block cell proliferation is linked to its potential to cause DNA breaks and inhibit transcription. In this study, we show that trabectedin and lurbinectedin activities are favored by transcription and, at least partially, they might be enhanced by R-loops. Trabectedin and lurbinectedin treatment enhanced transcription-dependent genome instability and replication stress (Figs. 1 and 2). Exposure to trabectedin or lurbinectedin did not cause a global increase in R-loops as assayed by S9.6 immunofluorescence (Supplementary Fig. S3). However, these drugs caused genome instability and replicative stress that were dependent on transcription and RNA–DNA hybrids both in yeast and human cells. These findings suggest that trabectedin and lurbinectedin toxicity is higher at sites undergoing active transcription and R-loop accumulation.

It was previously demonstrated that trabectedin and lurbinectedin directly bind DNA, forming adducts (6) and distort the double helix. Here we showed that these DNA adducts compromise replication fork progression; indeed, we found that trabectedin caused a profound impairment of DNA synthesis (Fig. 2A). To some extent, we can attribute this effect to R-loop formation, because RNase H1 overexpression partially suppressed this DNA synthesis defect. RNase H1 specifically cleaves the RNA strand of RNA–DNA hybrids therefore, overexpression of this endoribonuclease counteracts the accumulation of RNA–DNA hybrids. However, the rescue of the EdU incorporation defect by RNase H1 overexpression was mild, which suggests that only a subset of the trabectedin-induced replicative problems were mediated by RNA-DNA hybrids. Analysis of DNA replication at the single molecule level by DNA combing revealed that transcription inhibition with cordycepin reduces the increase in fork asymmetry provoked by trabectedin exposure (Fig. 2B), consistent with the conclusion that R-loops contribute to fork stalling in trabectedin treated cells. In any case, we cannot exclude the possibility that the inhibition of transcription by trabectedin could contribute to RF asymmetry.

R-loop–dependent replication problems caused by trabectedin are possibly counteracted by the FA pathway. The FA pathway has a key role in balancing replication stress, in particular when interstrand crosslinking agents such as Mitomycin C (MMC) cause replication fork blockage (37) and has been shown to limit R-loop accumulation and R-loop–dependent genome instability (17, 18). Interestingly, FA is critical for the integrity of common fragile sites (CFS), because FANCD2, a central player of the FA pathway, is required for a safe replication through R-loop–accumulating CFSs (38). Here, we found an increased accumulation of FANCD2 foci after trabectedin treatment that, importantly, was fully rescued by RNase H1 overexpression (Fig. 1C). This phenotype was different to that caused by other genotoxic agents. Thus, contrary to trabectedin, overexpression of RNase H1 in cells treated with the radiomimetic agent NCZ exacerbated FANCD2 foci accumulation (Supplementary Fig. S2). This may be explained if FANCD2 foci induced by NCZ are not RNA–DNA hybrid-dependent, and RNase H1 overexpression causes an additional type of replication stress. This finding adds further support to the conclusion that trabectedin has a prominent and specific action on R-loops where it preferentially causes DNA damage and breaks able to block replication fork progression. Consistently, using a cell line system where R-loops were stabilized by the ectopic expression of the RNA–DNA hybrid binding domain of the RNase H1 enzyme fused to GFP, and additionally increasing the basal levels of R-loop through the siRNA-mediated depletion of THOC1 or BRCA2, we determined that cells with higher levels of R-loops were more sensitive to trabectedin treatment (Fig. 3).

Importantly, we were able to recapitulate the phenotype of R-loop–dependent genome instability caused by trabectedin and lurbinectedin in yeast cells. Rad52 foci induced after treatment of either trabectedin or lurbinectedin were rescued by RNase H1 overexpression (Fig. 5). This result strongly suggests that the preferred action of these antitumoral agents on DNA is independent of the eukaryotic system or cell type, arguing strongly in favor of a preferential action on the R-loop–containing chromatin. In HeLa cells exposed to these agents, the RNA–DNA hybrid-dependent DNA damage phenotype was not so clear as it was in yeast cells. However, we were able to show by comet assay and 53BP1 immunofluorescence that the DNA damage induced by trabectedin or lurbinectedin was partially rescued by RNase H1 overexpression (Fig. 4; Supplementary Fig. S1). Besides, we found that γH2AX foci accumulation after trabectedin treatment was further increased in cells overexpressing RNase H1 (Supplementary Fig. S5A). However, even if phosphorylated γH2AX on Ser139 is used as a DSB marker, it also denotes replication stress, because it is extended through a large chromatin domain as an early response to replication fork stalling in an ATR-dependent process (39). RNase H1 overexpression could cause replication stress and instability, likely because it hampers replication by altering Okazaki fragment dynamics. Accordingly, DNA synthesis was impaired in cells overexpressing RNase H1 (Fig. 2A). The double treatment with RNase H1 overexpression and trabectedin might have an additive effect by their impact on replication, enhancing the probability of ssDNA gaps or breaks that would lead to increased cell death (Supplementary Fig. S5B).

The novel finding of an interplay between R-loops and trabectedin in genome instability induction is of particular interest because it could be exploited for cancer cells in which R-loops might be a common feature. The observation that trabectedin toxicity is augmented by the dysfunction in pathway that have been shown to protect both from cancer and from R-loops accumulation, such as the FA pathway and the BRCA1 and BRCA2 tumor suppressor genes (14, 15, 17, 18), supports the idea that R-loops might be a hallmark of specific types of cancer (40). It has been shown that cancer patients with BRCA mutations exhibit improved clinical response to trabectedin (16). It would be interesting in the future to check if R-loops are increased in tumoral cells derived from such patients to strengthen our hypothesis that a subset of tumors characterized by an higher accumulation of R-loops are more suitable to be treated with trabectedin.

No potential conflicts of interest were disclosed.

Conception and design: E. Tumini, C.M. Galmarini, A. Aguilera

Development of methodology: E. Tumini, E. Herrera-Moyano, M.S. Martín-Alonso, S. Barroso

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Tumini, E. Herrera-Moyano, M.S. Martín-Alonso, S. Barroso,

Writing, review, and/or revision of the manuscript: E. Tumini, C.M. Galmarini, A. Aguilera

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Aguilera

Study supervision: A. Aguilera

The authors thank Hélène Gaillard and Oksana Brehey for assistance in setting up yeast experiments. This work was supported with funding by the European Research Council (ERC2014 AdG669898 TARLOOP), the Spanish Ministry of Economy and Competitiveness (BFU2016-75058-P), the European Union (FEDER; BFU2016-75058-P), and PharmaMar (PRJ201402213).

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.