Introduction
Chemo- and immunotherapy resistance is, besides incomplete cytoreductive surgery and tumor metastasis, a major cause for recurrence and cancer therapy failure. Therapy resistance can be either intrinsic where tumors are not sensitive to therapy due to existing resistance-causing factors or acquired where initially sensitive tumors become resistant during therapy. Major causes for drug resistance have been extensively studied in the past: elevated expression of drug efflux transporters, increased or decreased DNA-damage repair, altered drug metabolism and detoxification, mutated drug targets, altered survival/death signaling, hypoxia, and presence of cancer stem cells (1,2).
Poly (ADP-ribose) polymerase (PARP)-inhibitors (PARPi) such as olaparib and niraparib are successfully used as a treatment of platinum-sensitive Breast Cancer Susceptibility Gene (BRCA)-deficient tumors of the breast and the ovaries. They are also effective in tumors with a 'BRCAness'-phenotype, that is tumors with no mutations in BRCA genes but with loss-of-function mutations in genes encoding other key players in the homologous recombination repair (HRR) pathway. BRCA1 and BRCA2 are crucial for the HRR process. PARPi are small-molecule inhibitors of PARP enzymes 1, 2, and 3, which are involved in detection and repair of single-strand breaks (SSB). PARPi block PARP catalytic activity or 'trap' PARP molecules onto the DNA, resulting in SSB persistence, replication fork stalling, and accumulation of DNA double-strand breaks (DSB), eventually leading to a 'synthetic lethality' phenotype in an HRR-deficient (HRD) background (3-9). However, emerging evidence indicates that PARPi are even effective in cells with functional HRR (10).
Although BRCA-mutations are a favorable predictor for PARPi sensitivity, a significant number of BRCA-mutated cancer nevertheless fail to respond to PARPi (11-14), either due to intrinsic resistance or resistance acquisition during the treatment and possibly enforced by the 'mutator phenotype' of HRD cells. Mechanisms of PARPi resistance include the restoration of HRR through reversion mutations in BRCA and RAD51, through BRCA1 promotor alterations, or through reconstitution of end-resection; the occurrence of alterations of PARP1 that abolish PARPi-trapping; the stabilization of the replication fork via depletion of chromatin remodelers; and the increased PARPi efflux via overexpression of multidrug resistance (MDR) pumps (15-17).
Reversion mutations (insertions or deletions) 'stamp out' pathogenic BRCA or RAD51 mutations and thus re-establish the reading frame, remove the deleterious mutation, or cause a synonymous mutation, restoring functional BRCA and RAD51 (18-20). Restoration of homologous recombination (HR) with coinciding BRCA protein re-expression due to de-methylation of BRCA1 was demonstrated in patient-derived xenograft (PDX) models of PARPi-resistant breast cancer (21) as well as in patients with ovarian cancer (22). HR restoration can also occur by loss of the end-resection repressor 53BP1 (23) or depletion of the shielding complex component REV7 (24).
The occurrence of mutations in the PARP1 gene is the most obvious cause for abolished PARPi-trapping: These mutations may result in a PARP protein unable to become trapped in response to PARPi. This has been found in an ovarian cancer patient with olaparib de novo resistance: she had a PARP1 mutation affecting a region critical for the cross-talk between the DNA-binding and catalytic domains (25). Phosphorylation of PARP1 was also found to reduce binding by PARPi, increase PARP1 enzymatic activity, and thus confers resistance to PARPi (26).
The dysregulation or collapse of replication forks, where for instance a variety of fork- and chromatin-remodeling proteins promote MRE11-dependent nascent DNA strand degradation at stalled replication forks, is one important feature of PARPi cytotoxicity. Depletion of these remodelers prevented strand degradation by MRE11, leading to the stabilization of the replication fork, and resistance to olaparib in BRCA1/2 deficient cells (27,28).
MDR1 overexpression is a common feature for MDR and the negative impact of MDR1 overexpression on olaparibsensitivity was shown in a mouse model (29) and also in breast and ovarian cancer cases where PARPi-resistance could be attributed to fusions and rearrangements of genes located near ABCB1 (30).
The issue whether PARPi exposure can generate an acquired resistance phenotype in cancer cells has only been poorly addressed, with controversial results been reported. Two studies reported development of acquired PARPi resistance in ovarian cancer cells by olaparib (31) and in breast cancer cells by niraparib (32) whereas another study reported that niraparib failed to induce mutations responsible for treatment resistance (33).
Thus, it remains unclear whether PARPi-imposed resistance acquisition is a common or a rather rare phenomenon, whether it is cell line-dependent, whether it depends on the mutational profile of genes implicated in DNA damage pathways (e.g. BRCA, TP53) or on the genetic background in general, whether it occurs preferably in cells that are intrinsically PARPi-sensitive, and whether it is a permanent or rather a transient (resistance phenotype is lost after removal of the selection pressure) phenomenon. In addition, it is unclear whether cells with an acquired PARPi resistance are cross-resistant to PARPi-unrelated compounds such as platinum salts, anthracyclines and even compounds not directly interacting with the DNA like Paclitaxel.
To address these topics, the intrinsic sensitivity of a panel of fourteen ovarian cancer cell lines to olaparib and niraparib as well as to PARPi-unrelated compounds was first determined, and then a subpanel of seven cell lines (six with and one without TP53-mutations, and two with and five without BRCA-mutations) was selected to investigate whether exposure to escalating olaparib concentrations generates subcells with acquired resistance to PARPi and/or PARPi-unrelated compounds.
Materials and methods
Cell culture and drugs
The following fourteen parental cell lines were used: A2780, BG1, CaOV3, COV362, ES-2, IGROV1, Kuramochi, OVCAR3, OVCAR4, OVCAR5, OVCAR8, OVSAHO, SKOV3, and TYK-nu. Nine of them are TP53-mutated and five BRCA-mutated (Fig. 1). Cisplatin-resistant A2780/CP were obtained from parental A2780 cells by stepwise exposure to increasing cisplatin concentrations (35). The cisplatin-resistant TYK-nu(R) cells were developed from parental TYK-nu cells by stepwise exposure to cisplatin and obtained from JCRB Japan Cell Bank (36). Paclitaxel-resistant IGROV1-PXL were generated from parental IGROV1 cells through stepwise exposure to escalating concentrations of paclitaxel in our lab (37). All cell lines were cultured in RPMI-1640 (cat. no. R8758) supplemented with 10% fetal bovine serum (FBS; both from Sigma-Aldrich; Merck KGaA), penicillin/streptomycin (100 U/ml/100 µg/ml; Sigma-Aldrich; Merck KGaA) at 37°C in a 95% humidified atmosphere containing 5% CO2. All cell lines are STR-profiled and routinely tested for mycoplasma infection.
Chemotherapeutic drugs were obtained from various suppliers: Gloucester Pharmaceuticals Inc. (romidepsin); Labatec (carboplatin); MedChemExpress (OTS167); Sigma-Aldrich; Merck KGaA [cisplatin, doxorubicin, epirubicin, paclitaxel and suberoylanilide hydroxamine (SAHA)]. They were dissolved in DMSO (niraparib, olaparib and SAHA), methanol (paclitaxel), or water (carboplatin, cisplatin, doxorubicin, epirubicin, OTS167 and romidepsin) and stored as aliquots at −20°C.
Drug sensitivity and cell proliferation rate
Drug sensitivity was determined by the MTT-assay and the colony formation assay (CFA). For the MTT-assay, cells (5,000-7,000 in 200 ml medium: density depends on the cell line) were seeded into 96-well plates and treated with each drug for 72 h: carboplatin (range: 3-500 µM), cisplatin (0.5-50 µM), doxorubicin (3-3,000 nM), epirubicin (0.1-10 µM), niraparib (0.5-100 µM), olaparib (3-1,000 µM), OTS167 (10-1,000 nM), paclitaxel (0.5-300 nM), romidepsin (0.1-100 nM), SAHA (0.5-300 µM).
Then MTT-dye (cat. no. M2128; Sigma-Aldrich; Merck KGaA; final concentration: 0.5 mg/ml) was added for 3 h, followed by removal of the medium and dissolution of the purple formazan crystals with DMSO. The optical density (OD; absorbance at 540 nm) was measured using the SynergyH1 Hybrid Reader (BioTek Instruments, Inc.). Data (mean ± SD of at least four independent experiments performed in quadruplets) are presented as the relative proliferation as a function of time after seeding. IC50-values were calculated by linear extrapolation. The ratio of the IC50-values of the matched subcells and the parental cells was calculated. Subcells were considered resistant if the ratio was ≥2.0 or hypersensitive if ≤0.5.
For the CFA, 1,000 cells in 2-ml culture medium were seeded into 12-well plates and exposed to olaparib on the next day for 8-10 days: 0.1, 0.2, 0.5, 1 µM olaparib for A2780 cells; 4, 8, 16, 32 µM for IGROV1; 0.3, 1, 3, 10 µM for OVCAR3; 1, 2, 5, 10 µM for OVCAR8. Then the medium was removed, the colonies were fixed at room temperature for 30 min and stained with 0.05% crystal violet (Sigma-Aldrich; Merck KGaA) in 4% formalin (Formafix AG), the plates were then rinsed 3-4 times with water and dried and images of the plates were captured (Fusion FX7 Edge Imaging System; Witek AG).
Cell proliferation rate was calculated from cell counts at the seeding day and the harvesting day by the following formula for exponential growth: Td = T1 × log(2)/log(N1/N0): where N0 is the number of cells seeded at time T0, N1 the number of cells harvested after time T1 and Td the doubling-time (proliferation rate).
Western blotting
Western blot analysis was used to assess the expression of MDR1. Cell lysates were obtained from subconfluent cultures at the time of harvest. Cells were lysed with RIPA buffer (cat no. 9806; Cell Signaling Technology Europe). Protein concentration was determined by the BCA Protein Assay (cat. no. 23227; Pierce; Thermo Fisher Scientific, Inc.). A total of 20 µg of protein were loaded and separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (10% SDS-PAGE), followed by blotting onto a polyvinylidene difluoride (PVDF) membrane (cat. no. 162-0177; BioRad Laboratories, Inc.) according to standard protocols. Membranes were blocked at room temperature for 60 min in TBST/milk [Tris-buffered saline with 0.1% Tween® 20 (Sigma-Aldrich; Merck KGaA)] and containing 3% (w/v) fat-free milk powder purchased from a local grocer). Cas9, EGFP, GAPDH, MDR1, PARP1 and tubulin were detected with specific primary antibodies: mouse anti-Cas9 (cat. no. 14697; Cell Signaling Technology, Inc.), mouse anti-EGFP (cat. no. sc-9996; Santa Cruz Biotechnology, Inc.), mouse anti-GAPDH (cat. no. sc-47724; Santa Cruz Biotechnology, Inc.), mouse anti-MDR1 (cat. no. sc-13131; Santa Cruz Biotechnology, Inc.), rabbit anti-PARP1 (cat. no. 9542; Cell Signaling Technology, Inc.) and rabbit anti-tubulin antibody (cat. no. 2148; Cell Signaling Technology, Inc.). All primary antibodies were diluted 1:1,000 in TBST/milk, except MDR1 which was diluted 1:500). Following the primary incubation (at 4°C for overnight), the membrane was incubated at room temperature for 3 h with the matched secondary antibodyeither horseradish peroxidase (HRP)-conjugated anti-mouse (cat. no. 7076) or HRPO-conjugated anti-rabbit (cat. no. 7074; both from Cell Signaling Technology, Inc.) antibody (both diluted 1:2,000 in TBST/milk) Complexes were visualized by enhanced chemiluminescence (Dura West; Pierce; Thermo Fisher Scientific, Inc.) and autoradiography (Fusion FX7 Edge Imaging System, Witek AG).
Generation of cell lines with acquired resistance
The following protocol was used to generate cells with acquired resistance to olaparib. It is based on the selection principle of clonal growth by repeated exposure of cells to stepwise escalating drug concentrations, assuming that cells acquire new features in an irreversible fashion by chronic drug exposure, as previously described (38).
The seven cell lines subjected to this protocol were selected according the following criteria. They exhibited different status for TP53, BRCA1 and BRCA2 (Fig. 1) and displayed different intrinsic olaparib sensitivity, ranging from relatively sensitive to relatively resistant (Fig. 2A): A2780, ES-2, IGROV1, OVCAR3, OVCAR4, OVCAR8, and TYK-nu). Briefly, 50,000 cells were seeded into six-well plates and exposed to olaparib for 48 h, followed by replacement of the olaparib-containing medium by olaparib-free medium in order to allow viable cells to recover and expand to confluency. After recovery, cells were re-seeded and exposed to a higher concentration of olaparib for 48 h, again followed by replacement of the olaparib-containing medium and recovery and expansion of viable cells. These cycles of exposure with escalating olaparib concentrations were repeated until no viable cells were left anymore in the last step of the protocol. This protocol and in particular the 48 h-exposure was selected because it was able to produce histone deacetylase inhibitor resistance acquisition in our previous studies (38-40). However, it is considered worthwhile trying different experimental protocols and conditions to induce PARPi resistance in future studies. The following matched subcells were obtained: A2780-OLA, ES-2-OLA, IGROV1-OLA, OVCAR3-OLA, OVCAR4-OLA, OVCAR8-OLA, TYK-nu-OLA. Protocol details are summarized in Table I (the start and end concentration of olaparib, the number of cycles and the total olaparib escalation for each cell line). For each cell line the protocol was performed twice.
In order to monitor potential olaparib resistance acquisition, matched subcells and parental cells were subjected to MTT assays after each cycle. Resistance was defined if the ratio of the IC50-values of the subcells and the parental cells is ≥2.0.
Generation of CRISPR-Cas9-mediated PARP1-knockout ovarian cancer cell lines
For molecular cloning, single guide RNAs (sgRNA) targeting protein-coding genomic DNA sequences of PARP1 gene, exon 1 sg1_PARP1_5′-CGA GTC GAG TAC GCC AAG AG-3′, exon 2 sg2_PARP1 5′-TGG GTT CTC TGA GCT TCG GT-3′, and exon 1 sg3_PARP1_5′-GCA TCC CCA AGG ACT CGC TC-3′, were designed using Benchling (Biology Software, 2021, retrieved from https://benchling.com). Single strand oligonucleotides were purchased from Sigma-Aldrich; Merck KGaA and cloned into LRG2.1 (cat. no. 108098; Addgene, Inc.) plasmid using the BsmBI endonuclease restriction site. Annealed oligonucleotides were ligated into the desired plasmid using the T4-DNA ligase (Promega Corporation) for subsequent expression of the sgRNA together with EGFP fluorescent protein. Ligations were transformed into Stbl3 E. coli following ampicillin selection using ZR Plasmid Miniprep-Classic plasmid purification (Zymo Research Corp.) and Sanger DNA sequencing (Microsynth AG) was used to confirm insertion of respective sgRNA using the human U6 primer (5′-GAG GGC CTA TTT CCC ATG ATT-3′).
Constitutively Cas9+ expressing ovarian cancer cell lines were generated by lentiviral transduction and subsequent puromycin selection. In brief, 293T cells (kindly provided by Dr Neutzner, Department of Biomedicine, University Hospital Basel) were seeded in a T75 flask at 50% confluency one day before transfection for preparation of lentiviral particles. A total of 4 µg of LRG2.1 (cat. no. 108098) or pLenti-Cas9-P2A-Puro (ca. no. 110837), 2 µg of pMD2.G (cat. no. 12259) and 2 µg of pCMVR8.74 (cat. no. 22036; all from Addgene, Inc.) were co-transfected using 24 µl of jetPEI reagent in 1 ml of 150 mM NaCl solution (Polyplus-transfection; Chemie Brunschwig AG). Growth medium was changed 24 h after transfection. Supernatant containing-lentivirus particles was collected 48 h later and filtered with a 0.45-µm PVDF filter (Sartorius AG), aliquoted in cryotubes and stored at −80°C until further use. OVCAR3, OVCAR5, and OVCAR8 cells were transduced with 1 ml of supernatant containing pLenti-Cas9-P2A-Puro lentivirus particles and further selected with 1-3 µg/ml puromycin for one week. Selected Cas9+ cell lines were kept in media containing 1 µg/ml puromycin. Cells lines stably expressing Cas9 protein were then lentiviral-transduced either with sgRNAs targeting the AAVS1 loci (mock) (41); or PARP1 followed by sorting enrichment of EGFP+ cells using the BD FACSAria Cell Sorter (BD Biosciences).
To analyze and confirm CRISPR-mediated mutagenesis, genomic DNA of Cas9+ non-gRNA transduced (control) or transduced cells with sgRNAs targeting AAVS1 (mock) sg1_ AAVS1_5′-ACT GTT GAC GGC GGC GAT GT-3′, sg2_AAVS1_ 5′-GCT GAT ACC GTC GGC GTT GG-3′; or PARP1 sg1_ PARP1_5′-CGA GTC GAG TAC GCC AAG AG-3, sg2_PARP1 5′-TGG GTT CTC TGA GCT TCG GT-3′, sg3_PARP1_5′-GCA TCC CCA AGG ACT CGC TC-3′ was extracted using the DNeasy Blood & tissue Kit (cat. no. 69504 Qiagen) 3 and 6 days after transduction. The genomic locus targeted by PARP1 was amplified using the forward 5′-GGG GGA GGG GTT GGG GGT AAA A-3′ and reverse 5′-GCC TTC AAG CCC ACC ACC TCA C-3′ primers. PCRs were performed using 2xGoTAq green Master Mix (Promega Corporation), 200 nM of each primer and 100 ng of genomic DNA. PCR conditions were as follows: Initial DNA denaturation at 94°C for 5 min followed by 32 cycles of 95°C for 20 sec, 62°C for 15 sec and 72°C for 1 min and 30 sec with a final extension at 72°C for 5 min. Amplicons were visualized on 1% agarose gel and purified by Wizard SV gel and PCR Clean/up System (Promega Corporation). Amplicons were analyzed using the Tracking of Indels by Decomposition (TIDE) assay (42).
Statistical analysis
For all comparisons, the mean ± SD values were calculated and statistical analysis was performed using the paired, two-tailed Student's t-test (Microsoft Excel, version 2016). P<0.05 was considered to indicate a statistically significant difference. For correlation analyses, the Spearman's rank correlation was calculated (Microsoft Excel).
Results
PARPi-sensitive ovarian cancer cells are not only sensitive to platinum but also to other chemotherapeutic compounds
At first the sensitivity of fourteen cell lines (TP53-mutated, n=9; BRCA-mutated, n=5) (Fig. 1), to olaparib and niraparib was determined. The results demonstrated that these cell lines display a wide spectrum of olaparib and niraparib sensitivity, with ES-2 and BG1 as the most sensitive and COV362 and OVSAHO as the least sensitive cells to olaparib (Fig. 2A), and with BG1 and TYK-nu the most sensitive and OVCAR8 and COV362 the least sensitive cells to niraparib (Fig. 2B). Spearman's rank correlation analysis (Fig. 2C) showed that olaparib-sensitive cells were commonly also niraparibsensitive (rs=0.582). They also indicated that among the fourteen cell lines, those with BRCA-mutations (COV362, Kuramochi, OVCAR4, IGROV1) tended to be less sensitive to both PARPi than those without BRCA-mutations (Fig. 2A-C). No correlation was found between the proliferation rate of ovarian cancer cells and their sensitivity to olaparib or niraparib (Fig. 2D).
Then it was determined whether PARPi-sensitive cells were, in addition to expectedly being sensitive to platinum compounds, also sensitive to other classes of chemotherapeutic compounds, such as representatives of the classes of anthracyclines, taxanes, histone deacetylase (HDAC)-inhibitors (HDACi), and maternal embryonic leucine zipper kinase (MELK)-inhibitor. The latter three are not known to interact with the DNA. MELK expression has recently been shown to correlate with poor outcome in ovarian cancer and its inhibition by OTS167 abrogates proliferation and viability of ovarian cancer cells (43). HDACi act epigenetically, that is without directly interacting with the DNA, and induce acetylation of histones and non-histone proteins and thus control gene transcription, protein function, proliferation and apoptosis (44). They have been shown to inhibit the growth and spread of ovarian tumors and synergize with platinum-based chemotherapy drugs (45), although their clinical usefulness remains unclear (46). PARPi sensitivity was not only identified in cells sensitive to carboplatin and cisplatin (rs platinum compounds=0.542) but in cells sensitive to doxorubicin and epirubicin (rs anthracyclines=0.456), to paclitaxel (rs=0.667), and to romidsepsin and SAHA (rs HDACi= 0.607). An association was also found between PARPi and compounds that interact with the DNA (rs=0.587) and compounds that do not interact with the DNA (rs=0.633). No correlation was found for OTS167 (rs=0.248) (Fig. 3). Details are summarized in Table II.
Next, the sensitivity to olaparib and niraparib of cells with acquired cisplatin-resistance [TYK-nu(R) and A2780/CP cells] or acquired paclitaxel-resistance (IGROV1-PXL cells) was determined. TYK-nu(R) are 4-fold and A2780/CP are >9-fold resistant to cisplatin, and IGROV1-PXL are 9.3-fold resistant to paclitaxel (37). The results (Fig. 4A) demonstrated that TYK-nu(R) cells were cross-resistant to both PARPi (4-fold to olaparib and 6-fold to niraparib), while A2780/CP cells were cross-resistant to niraparib only (3.1-fold). By contrast, IGROV1-PXL cells retained sensitivity to both olaparib and niraparib. IGROV1-PXL cells express MDR1, but cisplatin-resistant A2780/CP and TYK-nu(R) cells do not (Fig. 4B). Details are shown in Table III.
Olaparib does not induce acquired resistance to PARPi or acquired cross-chemoresistance
Expanding on the previous studies which reported opposing results on PARPi-imposed resistance acquisition (31-33), seven cell lines differing in their status for TP53, BRCA1 and BRCA2, and also in their intrinsic sensitivity to olaparib (from relatively sensitive to relatively resistant) were selected (Figs. 1 and 2A). These cell lines were subjected to the protocol of 'repeated exposure with olaparib concentration escalation' as described in 'Materials and methods', yielding the following subcell lines: A2780-OLA, ES-2-OLA, IGROV1-OLA, OVCAR3-OLA, OVCAR4-OLA, OVCAR8-OLA and TYK-nu-OLA. The proliferation rate and the sensitivity to olaparib, niraparib, carboplatin, cisplatin, doxorubicin and paclitaxel were also determined, demonstrating that the parental cells and their matched subcells have comparable proliferation rates (Table IV).
The MTT-assay results demonstrated that in none of these seven cell lines, stepwise 48 h-exposure of cells to escalating olaparib produced acquired PARPi resistance or an acquired cross-resistance to platinum compounds, doxorubicin, and paclitaxel (Fig. 5A), meaning that resistance factors (ratio from the IC50-values of the resistant subcells vs. the parental cells) >2.0 were not found. Detailed data are summarized in Table V. The failure to produce acquired PARPi-resistance was confirmed by CFAs for A2780, IGROV1, OVCAR3 and OVCAR8 (cell lines that form distinct colonies rather than proliferating as a confluent monolayer), where the clonogenic potential of the-OLA cells was not different from that of their parental cells in response to olaparib (Fig. 5B). The absence of an acquired resistance phenotype in all cells in this setup also indicated that it was not relevant whether or not TP53 and/or BRCA were mutated. OVCAR8 cells were also continuously (instead of 48 h as for olaparib) exposed to three cycles of escalating Niraparib and resistance acquisition to Niraparib, Olaparib, Carboplatin and Doxorubicin was not observed (data not shown).
Genomic deletion of PARP1 does not affect sensitivity to PARPi and other chemotherapeutic compounds
To determine whether the abundance of PARP1 protein expression affects drug sensitivity, PARP1-knockouts of three ovarian cancer cell lines with differential olaparib-sensitivity (OCVAR3 > OVCAR8 > OVCAR5) were produced using the CRISPR-Cas9 technology. To this aim, three different sgRNAs targeting exons 1 and 2 were designed in order to perform CRISPR-Cas9-mediated mutagenesis of PARP1 in these three cell lines (Fig. 6A). Downstream western blot analysis demonstrated reduction of PARP1 protein expression in all three PARP1-knockout cell lines (PARP1) in comparison with the non-transduced (control) and the AAVS1-transduced (mock) cells (Fig. 6B). The gene-editing of the PARP1 loci was further confirmed by the TIDE assay (42) (Fig. 6C).
Sensitivity of the OVCAR3, OVCAR5 and OVCAR8 PARP1-knockout cells and their respective mock cells to olaparib and niraparib, carboplatin and cisplatin, doxorubicin and paclitaxel was determined by MTT-assays. The results revealed that loss of PARP1 in these cell lines does not affect sensitivity to these drugs (Table VI).
Discussion
In the present study, the potential of olaparib to induce acquired resistance to PARPi and to PARPi-unrelated compounds was investigated. The results demonstrated i) that olaparib exposure did neither induce acquired resistance to PARPi nor induce cross-resistance to PARPi-unrelated compounds such as platinum salts, paclitaxel and doxorubicin; ii) that intrinsic PARPi resistance not only associates with resistance to platinum salts but also with resistance to other chemotherapeutic compounds like doxorubicin and epirubicin, paclitaxel and romidepsin and SAHA; and that iii) cells with acquired cisplatin-resistance are PARPi cross-resistant.
The key finding of the present study is the failure to generate subcells with a detectable acquired resistance phenotype both to PARPi and to PARPi-unrelated compounds after escalating olaparib (Table V). This is consistent with a previous study reporting that long-term treatment with niraparib did not cause genetic alterations and did not increase the mutation load in wildtype and BRCA1-defective breast cancer cells to allow the genetic evolution of resistance (33).
The present findings are, however, opposed to two other studies reporting the occurrence of acquired PARPi resistance after long-term exposure of ovarian cancer cells to olaparib (31) and in high-grade serous ovarian carcinoma, patient-derived xenograft modes following treatment with niraparib (32). In the former study, olaparib-imposed acquired resistance to olaparib and niraparib was associated with the activation of the Wnt-signaling pathway. This was found in PEO1 cells, which are mutated in both TP53 and BRCA2 (and hence were cells representative for ovarian cancer) as well as in OVCA433 cells, which were TP53-mutated but BRCA2-wildt-ype. In the latter study, acquired PARPi resistance was associated with RAD51C promoter methylation loss, indicating that PARPi treatment can cause demethylation of RAD51C and that a single alteration was sufficient to confer PARPi resistance (32). Whether olaparib-induced acquired PARPi resistance also associated with cross-resistance to PARPi-unrelated compounds was not reported in these studies (31,32).
The study by Yamamoto et al (31), also suggested that PARPi resistance acquisition can occur regardless of the BRCA2-status. It was also considered that if the status of BRCA1, BRCA2, and TP53 in our subcell panel is relevant for resistance acquisition, but failed to produce acquired resistance in all cell lines investigated, that is regardless of whether they are mutated in TP53 (ES-2, OVCAR3, OVCAR8, TYK-nu) or TP53 and BRCA2 (IGROV1, OVCAR4) or whether they are wild-type for both (A2780). Intriguingly, not even the triple TP53/BRCA1/BRCA2-mutant IGROV1 cells, which carry also mutations in numerous DNA repair-associated genes including ARID1A, ATR, BLM, MRE11, MLH1, MSH2, MSH3, MSH6, RAD50 and RAD52, developed acquired resistance after olaparib escalation. It thus appears that whether or not olaparib-imposed resistance acquisition occurs does not necessarily depend on the TP53 and BRCA mutation status and that, if it occurs, it may rather be a matter of cell line dependence and/or of the mutational background of potential oncogenic genes.
There are other important findings. One is that intrinsically PARPi-sensitive ovarian cancer cells were not only cross-sensitive to platinum salts as expected (7,47,48) but notably tended to be also cross-sensitive to anthracyclines and even to DNA-unrelated compounds like HDACi, and paclitaxel (Table II). These findings suggested an association between PARPi sensitivity and a general chemo-sensitivity of cancer cells, that is beyond platinum sensitivity.
Anthracyclines intercalate into DNA and directly interfere with topoisomerase II, resulting in DSB and eventually in cell death (49). BRCA-deficient cells have been revealed to be sensitive to doxorubicin (50) and doxorubicin may reduce PARP activity and PARP1 expression (51). It appears that doxorubicin can mimic the effect of PARPi by reducing the abundance and function of PARP. Similarly, HDACi may acetylate PARP and increase PARP binding to chromatin, resembling PARP-trapping to DNA and hence mimicking the effect of PARPi, which eventually leads to decreased repair of cytotoxic DSBs (52) or HDACi induce hyperacetylation of the nuclear HSP90 and cause depletion of HR-related proteins, thus conferring BRCAness and defective DNA damage and HR response in wild-type BRCA1 breast cancer cells (53). Paclitaxel interacts with microtubules and inhibits cytokinesis and is not known to be implicated in DNA-repair (54). PARP has also various functions in mitosis, and PARP inhibition may give rise to various mitotic defects (55), possibly imitating the effect of the microtubule poison.
Similarly, cells with acquired platinum-resistance were cross-resistant to PARPi, whereas cells with acquired paclitaxel resistance remain PARPi-sensitive (Table III). The absence of PARPi cross-resistance in cells with acquired paclitaxel-resistance is obvious as PARPi and paclitaxel act by different mechanisms. Moreover, the observed PARPi cross-resistance in cells with acquired cisplatin-resistance is predictable since resistance to platinum-based chemotherapy is a strong predictor for PARPi resistance (12). However, it is unclear why A2780/CP cells were cross-resistant to niraparib but not olaparib, whereas TYK-nu(R) cells are cross-resistant to both. Although olaparib was reported to be a MDR1 substrate (56), an involvement of MDR1 does not explain the present observations. Cisplatin-resistant A2780/CP and TYK-nu(R) cells and their respective parental cells do not express MDR1, indicating that PARPi cross-resistance in A2780/CP and TYK-nu(R) cells is MDR1-independent. Similarly, MDR1-expressing paclitaxel-resistant IGROV1- PXL cells retain PARPi sensitivity, also indicating that MDR1 is not involved.
Another interesting observation was that at least in our cell line panel cells with BRCA-mutations tended to be PARPi-resistant rather than PARPi-sensitive as compared with cell lines with no BRCA mutations, which is opposed to the synthetic lethality concept (3,4).
Taken together, our in vitro results argue against olaparib as a likely inducer of acquired PARPi resistance and cross-resistance to other chemotherapeutic compounds. Not ignoring that the in vivo and in vitro situation may be different, but nevertheless assuming certain transferability into a clinical context, the present results would mean that an olaparib-based therapy would not produce PARPi- or chemotherapy-resistant cells and that 'any other' chemotherapy or even a therapy with a different type of PARPi could follow a PARPi-based therapy. They also suggested to extend the current view of PARPi efficacy into a broader context, that is beyond BRCAness, meaning that PARPi can be an option to treat cancers regardless of a BRCAness phenotype. However, this is rather speculative and should be evaluated in clinical trials. Moreover, although not observed, an acquired hypersensitivity phenotype would be even more intriguing in this respect and perhaps be an add-on to the idea that so-called 'exceptional responders' may be an alternative strategy to better identify novel molecular determinants of (hyper)sensitivity to these agents PARPi (57).
The failure to induce acquired PARPi resistance seems predictable, because PARPi are, in contrast to platinum compounds (58), unlikely mutagenic. On the other hand, it cannot be ruled out that other genetic changes or cellular events may occur (33) which in our experimental setup remain phenotypically undetectable, maybe because the present experimental protocol was not sufficiently stringent to produce or to select for subcells with acquired resistance. Similarly, it appears too simple to just consider BRCA- and TP53-mutations or the HRD-status (10) to delineate the sensitivity of different drugs to PARPi and to other chemotherapeutics. Rather, the entire mutational profile of the potential oncogenes for each individual cell line should be taken into consideration.