Drug-Driven Synthetic Lethality: Bypassing Tumor Cell Genetics with a Combination of AsiDNA and PARP Inhibitors

DNA double-strand breaks (DSBs) are the most lethal of the DNA insults and, if left unrepaired, result in genomic instability and ultimately cell death (1). Therefore, targeted therapies increasing the frequency or the persistence of spontaneous DSBs or DSBs induced by treatments such as radiotherapy or chemotherapy have been extensively studied during the last 2 decades. The most advanced drugs in this field are the poly(adenosine diphosphate [ADP]-ribose) polymerase (PARP) inhibitors, with clinical trials showing significant benefits in patients with BRCA-mutated ovarian cancer (2). Essentially, cells deficient in BRCA1 or 2 are 100- to 1,000-fold more sensitive to PARP inhibitors (PARPi) than BRCA1/2 heterozygote or wild-type cell lines (3, 4). PARP is rapidly recruited at the site of damage where it strongly auto-modified. The polymers of poly(adenosine diphosphate [ADP]-ribose) formed by PARP are used as a platform for the recruitment of many enzymes involved in Base Excision Repair (BER; ref. 5) and in Microhomology Mediated End Joining (MMEJ) repair of DSBs (6). PARP inhibition prevents BER repair enzymes from being recruited at damage sites (7) and leads to the accumulation of DNA single-strand breaks (SSB) that result in unrepaired stalled replication forks and consequent DSBs. These DSBs are mainly repaired by the homologous recombination (HR) repair pathway. Cells with BRCA1/2 mutations are defective in HR (so-called BRCAness) and die directly or indirectly from unrepaired DSBs (1). Cells with functional HR accurately and efficiently repair DSBs and are not sensitive to PARP inhibition. Although PARPi monotherapy showed promising efficacy and safety profiles in the clinic (8, 9), their major limitations are the necessity of HR deficiency (HRD) and the rapid emergence of resistance. Many tumors that initially responded to PARPi treatments finally relapsed through compensatory mutations restoring the HR activity or stimulating the activity of alternative repair pathways such as the nonhomologous end joining (NHEJ) pathway (10, 11).

We have recently developed an original class of DNA repair pathway inhibitor, Dbait (12). AsiDNA, a molecule of Dbait family, consists of a 32-base pair oligonucleotide forming a double helix that mimics a DSB. AsiDNA acts by hijacking and hyperactivating PARP1 (13) and the DNA-dependent protein kinase (DNA-PK; ref. 14) which modify the chromatin and consequently inhibit the recruitment of many proteins involved in the HR and NHEJ pathways at the damage sites (14). This strategy sensitizes tumors to DNA-damaging therapies such as radiotherapy and chemotherapy (15–18). The first-in-human phase I trial, combining AsiDNA to radiotherapy, to treat patients with skin metastases from melanoma showed encouraging results, with 30% of complete responses (19). We anticipated that AsiDNA could potentiate PARPi activity in BRCA-proficient cells by inhibiting HR and establishing a transient state of BRCAness. However, as both drugs act differently on DNA damage response, the inhibitory activities and the efficacy of the association had to be demonstrated.

To test this combined treatment, we first analyzed the effects in DNA repair of the PARPi olaparib (Ola) and AsiDNA, to check that each drug does not interfere with the DNA repair inhibition activity of the other drug. These analyses were performed in the breast cancer (BC) model. BC is the most common female malignancy, with more than 1.7 million new cases diagnosed each year worldwide (20). Inactivating mutations of BRCA are observed in 8.8 % of all sporadic BC tumors (21) with a prevalence of 30% in the basal-like/triple-negative subgroup (22). We studied the sensitivities to the two drugs alone or in combination in 21 tumor cell lines including BC cell lines with different BRCA status. We observed a synergistic effect of Ola and AsiDNA in all the tested models regardless of BRCA status. Analysis of multi-level omics data from BC cell lines in the context of comprehensive signaling network maps identified different molecular profiles associated with the sensitivity to AsiDNA or Ola, especially in DNA repair and cell-cycle mechanisms, highlighting the rationale of combining these two drugs. We also demonstrated that this combination is effective using different PARPi, with no toxicity in nontumor cells.

Cell cultures were performed with four BRCA-deficient BC cell lines (BC227 from Institut Curie, HCC1937, HCC38, and MDAMB436 from the ATCC), eight BRCA-proficient BC cell lines (BC173 from Institut Curie, BT20, HCC1143, HCC1187, HCC70, MCF7, MDAMB231, and MDAMB468 from the ATCC), three nontumor mammary cell lines (184B5, MCF10A, and MCF12A from the ATCC), five human cervical cancer HeLa cell lines silenced for BRCA1 (HelaBRCA1SX, Tebu-Bio referenced as 00301-00041), for BRCA2 (HelaBRCA2SX, Tebu-Bio referenced as 00301–00028), for PARP1 (HeLaPARP1KD, a kind gift of Vincent Pennanaech, Institut Curie, France) and controls (HeLaCTLSX, Tebu-Bio01-00001, and HeLaCTLKD a kind gift of Vincent Pennanaech, Institut Curie, France), human glioblastoma cell lines MO59K and MO59J (DNA-PKcs deficient), human melanoma cell lines SK28LshCTL and SK28 LshDNA-PKcs, human colorectal cancer cell lines HCT116 WT and HCT116 KU70^+/- (heterozygote for KU70 gene), human head and neck cancer cell line Hep2, hematologic cancer cell lines Hut78, IM9, and Jurkat. Cells were grown according to the supplier's instructions. Cell lines were maintained at 37°C in a humidified atmosphere at 5% CO₂.

DT40 Burkitt lymphoma cells are chicken cells that have been knocked out for different genes as previously described in ref. 23. For this study, we used DT40 wild-type cells as control (DT40WT), and four cell lines respectively knocked out for BRCA1, KU70, TDP1, and PARP1 genes (DT40BRCA1^KO, DT40KU70^KO, DT40TDP1^KO, and DT40PARP1^KO). The DT40 cells were cultured at 37°C with 5% CO2 in Roswell Park Memorial Institute (RPMI-1640) medium supplemented with 1% chicken serum (Life Technologies), 10⁻⁵ mol/L β-mercaptoethanol, penicillin, streptomycin, and 10% FBS. Reagents for cell cultivation were obtained from Gibco Invitrogen.

All PARPi, AZD-2281 (olaparib), AZD-2461, ABT888 (veliparib), MK-4827 (niraparib), BSI-201 (iniparib), BMN673 (talazoparib), and AG-014699 (rucaparib), were purchased from Medchem express and diluted on DMSO to a stock concentration of 10 mmol/L.

AsiDNA is a new chemical entity, a 64-nucleotide (nt) oligodeoxyribonucleotide consisting of two 32-nt strands of complementary sequence connected through a 1,19-bis(phospho)-8-hydraza-2-hydroxy-4-oxa-9-oxo-nonadecane linker with a cholesterol at the 5´-end and three phosphorothioate internucleotide linkages at each of the 5´ and the 3´ ends (Agilent). The sequence is: 5′-X GsCsTs GTG CCC ACA ACC CAG CAA ACA AGC CTA GA L - C_LTCT AGG CTT GTT TGC TGG GTT GTG GGC AC sAsGsC-3′, where L is an amino linker, X a Cholesteryl tetraethyleneglycol, C_L a Carboxylic (Hydroxyundecanoic) Acid Linker, and s a phosphorothioate linkage.

AsiDNA or PARPi cytotoxicity was measured by relative survival and cell death quantification. Adherent cells were seeded in 24-well culture plates at appropriate densities and incubated 24 hours at 37°C before AsiDNA and/or PARPi addition. Cells were harvested day 6 after treatment, stained with 0.4% trypan blue (Sigma Aldrich), and counted with a Burker chamber. Cell survival was calculated as ratio of living treated cells to living mock-treated cells. Cell death was calculated as the number of dead cells on the total number of counted cells. Additivity of the toxicity was calculated by the product of cell survivals to AsiDNA and cell survivals to PARPi.

To measure cytotoxicity in DT40 chicken lymphoma repair mutants (23), 750 cells were seeded in 96-well white plate (final volume, 150 μL/well) from Perkin Elmer Life Sciences in media with or without the indicated concentrations of the drugs (AsiDNA and/or veliparib) at 37°C. After 72 hours, cells were assayed in triplicates with the ATPlite 1-step Kit (PerkinElmer). Briefly, ATPlite solution was added to each well (150 μL for DT40 cells). After 5-minute treatment, luminescence intensity was measured by Envision 2104 Multilabel Reader from Perkin Elmer Life Sciences. Signal intensities of untreated cells were set as 100%.

For immunostaining, cells are seeded on cover slips (Menzel) at a concentration of 5 × 10⁵ cells and incubated at 37°C during 1 day. Cells are then treated with 16 μmol/L AsiDNA +/−1 μmol/L Ola. Twenty-four hours after treatment, cells are fixed for 20 minutes in 4% paraformaldehyde/PBS 1x, permeabilized in 0.5% Triton X-100 for 10 minutes, blocked with 2% BSA/PBS 1x, and incubated with primary antibody for 1 hour at 4°C. All secondary antibodies were used at a dilution of 1/200 for 45 minutes at room temperature (RT), and DNA was stained with 4′, 6-diamidino-2-phenylindole (DAPI). The following antibodies were used: primary monoclonal mouse anti–phospho-H2AX (Millipore), anti-53BP1 rabbit antibody (Cell Signaling Technology), anti-Rad51 rabbit antibody (Merk Millipore), secondary goat anti-mouse IgG conjugated with Alexa-633 (Molecular Probes), and secondary goat anti-rabbit IgG conjugated with Alexa-488 (Molecular Probes).

Cells treated with AsiDNA (16 μmol/L), Ola (1 μmol/L), or both were suspended in 0.5% low melting point agarose in culture medium and transferred onto a frosted glass microscope slide precoated with a layer of 0.5% normal melting point agarose. Slides were immersed in lysis solution [2.5 mol/L NaCl, 100 mmol/L EDTA, 10 mmol/L Tris, 1% sodium lauryl sarcosinate, 10% DMSO, 1% Triton X-100 (pH 10)] at 4°C for 1 hour, placed in a electrophoresis tank containing 0.3 mol/L NaOH (pH 13) and 1 mmol/L EDTA for 40 minutes, electrophoresis for 25 minutes at 25 V (300 mA), washed with neutral buffer [400 mmol/L Tris-HCl (pH 7.5)], and stained with 20 Ag/mL ethidium bromide. The variables of the “comets” were quantified with the software Comet Assay 2 (Perceptive Instrument). Triplicate slides were processed for each experimental point. The tail moment is defined as the product of the percentage of DNA in the tail and the displacement between the head and the tail of the comet.

These experiments were performed with a Leica SP5 confocal system, attached to a DMI6000 stand using a 63/1.4 objective, under a controlled environment (37°C, 5% CO₂). All records were made using the appropriate sampling frequency (512_512 images, line average of four and zooming set to eight) and an argon laser line (514 nm for YFP) adapted to the fluorescent protein of interest X-ray repair cross-complementing protein 1 (XRCC1)-eYFP. In the first step, two images were acquired within a period of 2 to 3 seconds at a laser energy setting sufficiently low not to induce any photodynamic damage. The 405-nm laser line (diode) was then set to maximum output for 100 ms and focused onto a single spot of constant size (176 nm) within the nucleus to cause a point of photodamage with a reproducible amount of energy. Recruitment of XRCC1-eYFP was then monitored by fluorescence using the same setting as for the predamage sequence. Laser damage was induced 24 hours after treatment with AsiDNA (16 μmol/L), Ola (1 μmol/L), or both. Images were captured at 2-second intervals for the following 52 seconds. All images were processed using the freely available software ImageJ complemented with the LOCI bioformat plugin (http://www.loci.wisc.edu/ome/formats.html) to open images generated by the Leica SP5 confocal system. A macro was written to automate data extraction from images. Briefly, it consisted of retrieving two regions of interest (ROI), namely the photodamage spot and the nucleus area excluding the spot, and quantifying the total intensity within these ROIs. The latter was used to correct fluorescence intensity for the observational photobleaching. Intensity within the former ROI was normalized to 1, based on quantifications before photodamage, then plotted against time to get the recruitment kinetics.

mRNA expression data for BC cell lines were produced using Human Exon 1.0 ST Affymetrix microarrays. Raw data were RMA normalized and summarized with FAST DB annotation (version 2013_1; refs. 24, 25). Gene expressions were log₂ transformed and mean centred over all the cell line samples and then grouped into the four groups (AsiDNA sensitive, AsiDNA resistant, Ola sensitive, Ola resistant). Each gene was assigned with a score using median expression level across samples of the same group, and the data were visualized on Atlas of Cancer Signaling Network (ACSN) map.

Mutation data sets for BC cell lines were retrieved from COSMIC database v71 (http://cancer.sanger.ac.uk; ref. 26). The frequency of mutations for each gene across cell lines in the same groups (AsiDNA sensitive, AsiDNA resistant, Ola sensitive, Ola resistant) was calculated, and the data were visualized on ACSN map.

The copy-number (CN) values for each gene over the cell lines were assessed by GAP analysis of the data generated on Affymetrix Genome Wide SNP Array 6.0 (27) and corrected for ploidy (considering four CN as “normal”), considering CN less than 3 as loss and CN more than 5 as gain. Then, each gene was assigned with a score using average copy number across samples of the cell lines in the same group (AsiDNA sensitive, AsiDNA resistant, Ola sensitive, Ola resistant), and the CN variation data were visualized on ACSN map.

The correlations were assessed by a leave-one-out (LOO) Spearman rank correlation (26, 28). Multiple correction testing was done with the Benjamin–Hochberg method (doi = 10.2307/2346101). All tests were considered as two-sided. For each treatment, a list of correlated genes is ranked by correlation P value. In order to discover the most correlated genes with one treatment, a stepwise P value selection was used so that no gene whose correlation P value under the selected P value were retrieved in both treatment. The unique, nonoverlapping set of gene robustly correlated with survival to each one of the drugs is provided. The selected P value (threshold P value) determined for the ranked genes included in ACSN (29) is 0.005. Supplementary Fig. S3A illustrates the relationship between the number of overlapping genes and the P value from LOO Spearman correlation analysis.

ACSN (29) uses NaviCell environment to navigate maps (30) and to analyze and visualize data in the context of maps (31). The enrichment of the ACSN modules with unique genes robustly correlated with survival to each one of the drugs is calculated in the NaviCell toolbox using the standard hypergeometric test, computing the enrichment P values (P value <0.02) for ACSN modules.

All statistical analysis was performed with a two-tailed Student t test.

As Ola and AsiDNA are both DNA repair inhibitors acting by inhibiting the recruitment of repair enzymes at damage sites, we first checked that each molecule does not impair the capacity of the other to inhibit recruitment of its targeted repair enzymes. One of the first enzymes to be recruited at damage site after auto-modification of PARP is the XRCC1. As expected, Ola significantly delayed the XRCC1 foci recruitment while AsiDNA did not (Fig. 1A and B). The recruitment of XRCC1 was similarly delayed in cells treated with Ola in the presence as in the absence of AsiDNA (Fig. 1A and B). AsiDNA binds and activates both PARP and DNA-PK in cells. Activation of PARP revealed by the accumulation of Poly-ADP-Ribose polymers was observed in AsiDNA-treated cells but not in Ola-treated cells or Ola+AsiDNA-treated cells, indicating that Ola prevents PARP activation by AsiDNA. Activation of DNA-PK kinase activity by AsiDNA can be easily revealed by the pan-nuclear phosphorylation of the histone H2AX (14). This phosphorylation was observed in 80% of treated cells in the presence as in the absence of Ola (Fig. 1C). Pan-nuclear phosphorylation of H2AX is thought to be involved in the inhibition of HR and NHEJ repair enzyme recruitment by AsiDNA (14). Ola induces the accumulation of DSBs revealed by the formation of γH2AX foci that colocalize with 53BP1 and Rad51 foci (Fig. 1C). The addition of AsiDNA significantly reduced the formation of 53BP1 or Rad51 foci induced by Ola (Fig. 1C and D). To demonstrate that the reduction of Rad51 and 53BP1 foci after AsiDNA is induced by the inhibition of their recruitment at damage sites and not through a reduction of the number of DNA damage, we used single-cell alkalin comet assays to monitor the damage in MDAMB231 tumor cells after the different treatments. As suggested by γH2AX foci, Ola treatment induced accumulation of damage over 24 hours while AsiDNA did not (Fig. 1E). Combining AsiDNA to Ola resulted in a two-fold increase of DNA damage induced by Ola. In MCF10A nontumor cells, Ola induced formation of few foci of 53BP1 and Rad51, which decreased in cells receiving both Ola and AsiDNA (Fig. 1F and G). However in contrast to tumor cells, the nontumor mammary cells did not show any significant increase of spontaneous damage after treatments with single or combined drugs (Fig. 1H).

Efficacy of Ola and AsiDNA was assessed by measuring cell death and proliferation in 21 different cancer cell lines including glioblastoma, cervical cancer, colon cancer, blood cancer, melanoma, and breast cancer. The concentration of the drugs (0.1 μmol/L for Ola and 4.8 μmol/L for AsiDNA) was chosen based on the 65% to 75% survival in the BRCA-mutant cell lines (Table 1). All tumor models show supra-additive efficacy of the drug combination (Table 1). Moreover, analysis of isogenic pairs with DNA repair mutants to single and combined treatments indicates that AsiDNA is highly cytotoxic to all mutants with one repair defect (PARP1, BRCA1, BRCA2, Ku70, DNA-PKcs), whereas Ola sensitivity is essentially restricted to the BRCA mutants (Table 1). The sensitivity of PARP1, BRCA, and Ku70 mutants to AsiDNA was confirmed in an isogenic set of DT40 chicken lymphoma repair mutants (Supplementary Fig. S1A), where the highest sensitivity was observed in PARP1 knock-out DT40 cells. As expected, in these cells as in Hela-PARP1–silenced human cells, addition of PARPi did not increase sensitivity to AsiDNA (Table 1 and Supplementary Fig. S1B). All the other 17 tested solid tumor–derived cell lines show a supra-additive response to the combined treatment indicating a supra-additivity between both inhibitors. In contrast, two of three of the blood cancer cell lines, Hut78 and Jurkat, had a survival to combined treatment close to the calculated additive effect of both single treatments (Table 1). Taken together, these results indicate that AsiDNA sensitize most tumor cell lines to Ola independently of their BRCA status or other genetic defects.

Two immortalized mammary cell lines (MCF10A and MCF12A) were analyzed for their sensitivity to the Ola and AsiDNA drug combination. Interestingly, the survival to combined treatment was not decreased in the nontumor cells (Fig. 2). Increasing the dose of Ola to 1 μmol/L had no significant effect on the normal cells but increased the combined treatment efficacy in the breast cancer tumor cells (Fig. 2). The synergistic effect was high (three times higher than expected additivity) in the MDAMB231 cell line, which is insensitive to standalone treatment by AsiDNA (Fig. 2). In contrast, the nontumor cells were resistant to Ola and did not show increased sensitivity after addition of AsiDNA (Fig. 2; Table 1). In all cell lines, the decrease in the relative number of cells correlated with an increase in cell death (Fig. 2B), indicating that the number of living cells reflects a cytotoxic and not a cytostatic effect. Thus, the combined treatment AsiDNA+Ola is specific to tumor cells with no toxicity in normal cells.

All the tested cell lines were sensitive to the combined treatment with Ola and AsiDNA, suggesting that resistance to both drugs is a very rare event. To better understand this observation, we analyzed the genetic markers associated with resistance to Ola or AsiDNA in a set of 12 BC cell lines (including the six BC cell lines tested in combination). No significant correlation between response to AsiDNA and response to Ola was observed in BRCA-proficient tumor cell lines (Spearman coefficient r: 0.27 and P value: 0.14; Supplementary Fig. S2). Only the BRCA^−/− cell lines were sensitive to both Ola and AsiDNA single treatments. In order to determine how gene expression profiles could explain the differences in sensitivities to Ola or AsiDNA, we retrieved the “sensitivity” lists of 74 and 71 genes that significantly correlated with sensitivity to respectively AsiDNA or Ola (Supplementary Fig. S3A; Supplementary Table S1). Interestingly, these lists did not display any common gene. Among the genes correlated with sensitivities to AsiDNA or Ola, respectively nine and 14 genes were directly involved in DNA repair and cell-cycle pathways (Table 2).

As only the transcriptome was taken into account, the well-known BRCA gene mutations associated with Ola sensitivity were not shortlisted in this analysis. Therefore, we completed the analysis by a multilevel omics data assessment using an ASCN resource (29) to integrate mRNA expression, CN variations, and mutational profiles from the BC cell lines. Molecular profiles associated with resistance to AsiDNA or Ola were both quantitatively and qualitatively different and demonstrated that a number of nonoverlapping molecular mechanisms are associated with resistance to Ola and AsiDNA (Fig. 3; Supplementary Fig. S4). Interestingly, several molecular mechanisms such as MOMP regulation, Cytoskeleton and Polarity, WNT non-canonical pathway, etc. were implicated in response to both drugs, but regulated in an opposite manner (Supplementary Fig. S3B–S3D and Supplementary Tables S2 and S3). Cell lines resistant to AsiDNA are characterized by multiple perturbations as expression elevation, CN gains, and mutations in processes involved in cell proliferation, cell survival, epithelial-mesenchymal transition (EMT), and cell motility functional modules (Supplementary Fig. S4A). These data suggest that cells resistant to AsiDNA most likely have a higher proliferation status corresponding to an increase in DNA repair, especially through HR and Fanconi repair pathways (Fig. 3A and B). However, cells resistant to Ola show mostly CN losses in a number of DNA repair pathways, suggesting an opposing function with respect to the Ola response (Fig. 3C and D; Supplementary Fig. S4B). As expected, Ola-sensitive cell lines have an active BER pathway and a defect in HR. In contrast, cells resistant to Ola show multiple losses of CN in genes involved in the BER pathway, suggesting an inactivation of this pathway.

Taken together, omics analyses highlight different molecular mechanisms underlying the response to AsiDNA or Ola, and suggest that repair defects associated to resistance to one drug will increase sensitivity to the other drug making a double resistance very unlikely.

PARPi belong to at least two classes: the catalytic inhibitors that inhibit PARP enzyme activity and the dual inhibitors that block both PARP enzyme activity and trap PARP proteins on DNA damage sites (23). Ola belongs to the second group, whereas veliparib (Veli) is essentially a catalytic inhibitor, as it shows a PARP-trapping activity only at very high doses (23). We repeated the analysis of combination efficacy using Veli instead of Ola (Fig. 4A and B). As observed with Ola, the AsiDNA showed a synergistic effect with Veli in the three BC cell lines and not in nontumor cells. This indicates that trapping PARP on DNA is not essential for an efficient combination.

We also monitored the efficacy of the combined treatment in MDAMB231 cells with five other PARPi (rucaparib, iniparib, niraparib, AZD2461, and BMN673) developed for clinical applications (Fig. 4C and D). The applied doses of PARPi were chosen to give a sublethal effect and 50% survival (Supplementary Table S4). At both doses, the supra-additive efficacy of the combination of PARPi with AsiDNA was confirmed with all the inhibitors (Fig. 4C and D) independently of their mechanism of action. These results demonstrate that the observed synergy between AsiDNA and PARPi is a general mechanism and not only restricted to Ola.

Many cancer treatments exploit tumor cell weaknesses or dependencies that are absent or less pronounced, in normal cells. Genetic instability and associated DNA repair defects is one such factor. As some functions become essential during tumor development, molecular networks facilitate compensatory alterations to allow cell survival, a form of “pathway buffering” (32). The idea of targeting tumors with identified genes and proteins that are synthetically lethal with specific tumor suppressor genes (33) has been successfully illustrated by the development of the PARPi in the treatment of BRCA-mutated tumors. However, despite several attempts, only trials which mandate a pathogenic loss-of-function BRCA mutation as an inclusion criterion seemed to give sustained responses (9).

Double-strand DNA breaks are the most lethal DNA lesions, and their repair is guaranteed by at least three independent repair pathways that render unlikely the loss of all DSB repair: HR, NHEJ, and an alternative NHEJ pathway (alt-NHEJ) requiring PARP activity, which takes place when conventional NHEJ and HR fail (34). AsiDNA inhibition of repair enzyme recruitment weakens the ability of the cells to eliminate DSBs. However, as it does not increase the damage occurrence in the cells, AsiDNA toxicity is dependent on spontaneous DNA damage frequency or their induction provoked by associated treatments (14). Interestingly, we observed that Ola significantly increases spontaneous damage frequency. AsiDNA does not induce any damage by itself and therefore do not show any toxicity in cells that do not encounter frequent spontaneous accident or damage as it is the case for nontumor cells. In these cells, which do not have deregulated repair or cell-cycle functions, the AsiDNA addition does not increase the toxicity of PARPi.

Proteins involved in DNA repair and response to damage, such as DSS1, RAD51, NBS1, ATM, ATR, CHEK1, CHEK2 (35), the Fanconi anemia pathway (36), or more recently PTEN (37), have been implicated as possible predictive markers for tumor cell response to PARPi. Data analysis from BC cell lines in the context of ACSN maps, and especially in the context of the DNA repair map, confirms the role of these proteins and their related pathways in Ola sensitivity. Similar analysis on AsiDNA data did not highlight the same molecular mechanisms.

As tumors with “BRCAness,” a profile of tumor associated with deficiency in HR and sensitivity to platinum (38), are relatively rare (21), there is a need to develop new drugs that could recapitulate such features and allow a wider population of patients to benefit from PARPi treatment. The differences in the profile requirement and the lack of toxicity in nontumor cells make AsiDNA a good candidate for such an association. Essentially, AsiDNA inhibits HR and NHEJ via blinding DNA damage site recognition (12, 14). The general supra-additive effect of the combination of the two drugs indicates independent mechanisms of action. The persistence of the Ola-dependent inhibition of XRCC1 recruitment in the presence of AsiDNA and of the AsiDNA-dependent inhibition of 53BP1 and RAD51 recruitment in the presence of Ola indicates that both drugs act independently. Therefore, Ola inhibits BER resulting in an accumulation of unrepaired SSBs that are converted to DSBs during replication or transcription and cannot be repaired due to the inhibition of HR, NHEJ, and altNHEJ by AsiDNA (Fig. 4E). The toxicity of PARPi appears to depend upon the ability of the drug to block PARP at the damage site. We identified that they range in the order of BMN673> AZD2461> Niraparib> Rucaparib> Ola> Iniparib according to their IC₂₀ and IC₅₀ in MDAMB231 (Supplementary Table S4). However, all the PARPi showed an enhanced toxicity when administered with AsiDNA. Iniparib, which many believe to not be a bona fide PARPi, as it has very low PARP inhibition in vitro, was only slightly sensitized by AsiDNA. In contrast, Ola and BMN673 were highly potentiated by the addition of AsiDNA. The lack of effect of the combination in nontumor cells makes the AsiDNA and PARPi combination a potentially interesting treatment in tumors without BRCAness status.

J.-S. Sun holds ownership interest (including patents) in DNA Therapeutics. M. Dutreix is a consultant/advisory board member for DNA Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: W. Jdey, J.-S. Sun, Y. Pommier, M. Dutreix

Development of methodology: W. Jdey, S. Thierry, A. Zinovyev, Y. Pommier

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Jdey, S. Thierry, Y. Pommier

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Jdey, S. Thierry, C. Russo, F. Devun, E. Barillot, A. Zinovyev, I. Kuperstein, Y. Pommier, M. Dutreix

Writing, review, and/or revision of the manuscript: W. Jdey, C. Russo, F. Devun, J.-S. Sun, A. Zinovyev, I. Kuperstein, Y. Pommier, M. Dutreix

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Jdey, P. Noguiez-Hellin, Y. Pommier, M. Dutreix

Study supervision: S. Thierry, P. Noguiez-Hellin, M. Dutreix

Other (performed the experiments with DT40 cell lines): M. Al Abo

We thank Nathalie Berthault and Wendy Philippon for their technical participation in this project. We thank the platform RadExp for the comet assays and the LIP for providing the BC227 and BC173 cell lines. We also acknowledge the local CICS platform facilities.

This work was supported by the SIRIC-Curie, the program PIC SysBio of the Institut Curie, the Centre National de la Recherche Scientifique, and the Institut National de la Santé Et de la Recherche Médicale. W. Jdey was supported by a fellowship CIFFRE-ANRT (2013/0907). S. Thierry was supported by the Institut National du Cancer (TRANSLA13-081).

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.