Identification of novel drug-specific PARP inhibitor resistance mechanisms in ovarian cancer–implications for clinical practice

Maintenance PARP inhibitor (olaparib or niraparib) treatment is commonly prescribed following carboplatin/paclitaxel chemotherapy in ovarian cancer patients, but response is compromised by adaptive drug resistance [ 1 ]. We have shown that P-gp/ ABCB1 influences resistance to paclitaxel and olaparib, but similar niraparib resistance mechanisms have not been described [ 2 , 3 ].

We used qRT-PCR, Western blot, RNASeq and LC-MS/MS proteomics analysis to compare drug transporter expression in sensitive and resistant immortalised and primary patient-derived cell lines. ABCB1 and ABCG2 expression was modified by shRNA-mediated knockdown and heterologous expression, with chemosensitivity changes assessed by MTT and clonogenic assays. Substrate specificity of P-gp and BCRP was assessed by efflux assays in polarised cells.

P-gp/ ABCB1 expression was not increased in A2780nirapR cells, which alternatively up-regulated BCRP/ ABCG2 . ABCG2 was consistently induced in niraparib-resistant patients, but ABCB1 only in patients pre-treated with paclitaxel. sh ABCG2 re-sensitised A2780nirapR cells, while heterologous expression in A2780 cells induced drug resistance. Efflux assays confirmed that olaparib and niraparib are both P-gp and BCRP substrates, suggesting that resistance results from transcriptional regulation of efflux transporters not substrate specificity.

Treatment-induced BCRP/ ABCG2 induction is a novel clinically relevant niraparib resistance biomarker. Routine inclusion of paclitaxel in first-line chemotherapy regimens may promote efflux transporter-mediated resistance, compromising response to PARPi maintenance treatment.

High-grade serous ovarian cancer, the most common and aggressive histotype, is frequently diagnosed when already advanced (FIGO Stage III/IV, https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/ovarian-cancer#ovarian_stats4 ). Surgery is not always possible in patients with advanced disease, and response to neo-adjuvant or adjuvant chemotherapy, most commonly combining carboplatin and paclitaxel, influences both progression-free and overall survival [ 4 ]. First-line chemotherapy is often extended by maintenance PARP inhibitor (PARPi, olaparib or niraparib) treatment, with eligible patients selected following assessment of germline and/or somatic BRCA1/2 mutation status or by extended homologous recombination deficiency (HRD) panel testing [ 5 , 6 ]. HRD status is an important biomarker of PARPi response, as PARPi-mediated inhibition of poly (ADP-ribose) polymerase 1 (PARP1) limits the ability of cells to repair single strand DNA breaks [ 7 ] and leads to PARP trapping and replication fork stalling [ 8 ]. Resulting double strand breaks are preferentially repaired by homologous recombination-mediated repair mechanisms, but this process is compromised in BRCA1/2 mutant or HRD deficient cancers, which exhibit synthetic lethality with DNA damaging platinum drugs and PARPis [ 9 ].

Although many patients initially respond well to treatment, the development of chemotherapy-induced resistance is a common, treatment-limiting complication [ 10 ]. Various resistance mechanisms including reduced drug bioavailability resulting from down-regulation of the copper import transporter 1 (CTR1, [ 11 ] and up-regulation of the detoxifying glutathione S-transferase GSTP1 [ 12 ] limit carboplatin bioavailability, while resistance to paclitaxel results from altered binding target β-tubulin isoform expression [ 13 ] or, more commonly, chemotherapy-induced increased expression of P-glycoprotein, a membrane-bound ATP-dependent efflux pump, encoded by ABCB1 (MDR1) [ 14 ]. The development of multidrug resistance is a particular clinical challenge as it not only limits response to initially prescribed drugs, but to structurally diverse alternatives which are also P-gp substrates [ 15 ]. We have previously shown that A2780 ovarian cancer cells made resistant to paclitaxel are cross-resistant to olaparib, with increased P-gp expression promoting resistance to both drugs [ 16 ]. Importantly, in contrast to previously reported mechanisms of PARPi resistance including acquisition of reversion mutations to restore HR pathway function [ 17 ], our data highlights that the impact of increased P-gp expression is PARPi-specific, with olaparib and rucaparib cross-resistant in A2780pacR cells, while sensitivity to veliparib and AZD2461 is not influenced by P-gp expression.

An alternative PARPi, niraparib (Zejula™), was licensed for first-line maintenance treatment of ovarian cancer patients, following the NOVA [ 18 ] and PRIMA [ 19 ] clinical trials. Niraparib is licensed for all platinum-sensitive patients, independent of BRCA1/2 mutation status although, like olaparib, exhibits greater efficacy in HRD patients. Our previous data highlights the importance of assessing whether niraparib, like olaparib, is a P-gp substrate and therefore also influenced by paclitaxel-induced increased drug efflux. Importantly, identification of additional novel, niraparib-specific resistance mechanisms may facilitate extension of current biomarker-guided prescribing protocols to limit the development of drug resistance in ovarian cancer patients.

The Dundee Ovarian Cancer Study (DOCS) was approved by the East of Scotland Research Ethics Service REC1 Committee (13/ES/0035). Informed consent was obtained from all participants, and the study performed in accordance with the Declaration of Helsinki.

A2780 cells were obtained from the European Collection of Authenticated Cell Cultures (ECACC). Novel niraparib-resistant (A2780nirapR) and olaparib-resistant (A2780olapR, [ 16 ]) A2780 derivatives were created following continuous incremental drug selection (0.1 µM–12.8 µM niraparib; 1 µM–20 µM olaparib), to mimic typical patient peak plasma concentrations [ 20 ]. Cell lines were authenticated by short tandem repeat profiling (Centre for Life, Newcastle, UK) and routinely tested for mycoplasma contamination (Lonza Biologics, Slough, UK).

Cell lines were cultured in RPMI-1640 medium supplemented with 10% v/v foetal bovine serum (Thermo Fisher, Renfrewshire, UK) and 1% v/v penicillin/streptomycin (Sigma Aldrich, Dorset, UK). Media for resistant cell lines was supplemented with 12.8 µM niraparib or 20 µM olaparib every third passage to maintain resistance.

Ascites samples were collected from consented DOCS study patients following paracentesis, transferred to a Category II cell culture facility and mixed 1:1 with ascites media (1:1 MCDB (Sigma Aldrich, Dorset, UK)/Media 199 (Gibco, Renfrewshire, UK), 10% v/v FBS, 1% v/v penicillin/streptomycin. Cells were incubated for 3–5 days before differential trypsinisation to limit fibroblast contamination, then primary cells incubated for an additional 3 days before use in subsequent experiments.

MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) chemosensitivity assays [ 21 ] were used (3 independent experimental repeats) to compare the sensitivity of A2780, A2780nirapR and A2780olapR cells to niraparib, olaparib, carboplatin, paclitaxel, camptothecin, irinotecan and the ABCG2 inhibitor Ko143. Cells were seeded at 5000 cells/well and treated in triplicate with 2-fold serial dilutions of each drug (all sourced from Selleckchem, Hertfordshire, UK) - 1.25 µM–40 µM niraparib, 1.56 µM–50 µM olaparib, 2.66 µM–85.12 µM carboplatin, 1 µM–31.85 µM paclitaxel, 0.16 µM–5 µM camptothecin and 0.31 µM–10 µM irinotecan. Ko143 concentrations (0.06 µM–2 µM) were initially tested in A2780nirapR cells before subsequent experiments combined 1 µM Ko143 with serial dilutions of either niraparib or olaparib. Cells were incubated with drug or DMSO vehicle control for either 144 h (niraparib and olaparib) or 72 h (carboplatin, paclitaxel, camptothecin and irinotecan), before removal of media and incubation with 100 µL 0.5 mg/mL MTT solution in phenol red-free RPMI for 3 h at 37 °C. MTT solution was removed and 100 µL DMSO added to solubilise formazan crystals, before absorbance was quantified at 570 nm using a spectrophotometer (MultiSkan, ThermoFisher Scientific). Percentage of viable cells remaining following drug treatment was calculated relative to vehicle-treated control cells and associated EC 50 values estimated from log dose-response curves using Prism 10 software (GraphPad Software, Inc., La Jolla, CA, USA).

Total RNA was extracted from 1 × 10 6 cells using RNeasy Mini Kits (Qiagen, Manchester, UK), following the Manufacturer’s instructions, including an on-column DNase digestion (RNase free DNase Kit, Qiagen, Manchester, UK). RNA yield and integrity were confirmed using a Nanodrop ND1000 spectrophotometer (Thermo Fisher, Renfrewshire, UK).

RNA (200 ng) was reverse transcribed into cDNA using TaqMan Reverse Transcription Reagents (Thermo Fisher, Renfrewshire, UK) according to the Manufacturer’s instructions, replacing oligo dT with random hexamers. ABCG2 (Hs01053790_m1), ABCB1 (Hs00184500_m1) and 18S ribosomal RNA expression (438839) was assessed in 20 µL reactions, combining 10 µL TaqMan universal master mix (Thermo Fisher, Renfrewshire, UK), 1 µL gene-specific probe, 1 µL cDNA and 8 µL nuclease-free water. Each reaction was performed in triplicate and run on the standard PCR programme (50 °C 2 min, 95 °C 10 min, and 40 cycles of 95 °C 15 s, 60 °C 1 min) on a QuantStudio5 qRT-PCR instrument (Thermo Fisher, Renfrewshire, UK). Baseline and threshold values were calculated automatically and gene expression quantified by cycle threshold (Ct) values, with relative gene expression comparing Ct values in target gene and 18S rRNA control (ΔCt), as previously described [ 16 ]. Compound errors (s) were calculated using the formula s = ((standard deviation target gene) 2 + (standard deviation 18S rRNA) 2 ) ½ ( https://assets.thermofisher.com/TFS-Assets/LSG/manuals/cms_042380.pdf ).

shRNA constructs packaged in pLKO.1 vector were purchased as bacterial glycerol stocks (SigmaAldrich), and plasmid DNA extracted using a QIAGEN HiSpeed maxi kit, according to the Manufacturer’s instructions. 2.5 × 10 5 cells were transfected with 2 µg ABCG2 shRNA construct (TRC0000059802) or empty-vector control (pLKO1 EV ), using 5 µL Lipofectamine 3000 and 4 µL P3000 reagent, in serum-free OptiMEM medium (all sourced from Thermo Fisher, Renfrewshire, UK). After 24 h, cells were re-seeded into 10 mm cell culture dishes, using fresh media containing 8 µg/mL puromycin for 120 h. Surviving colonies were selected using 150 µL colony chambers (Merck, Dorset, UK) and grown in selective puromycin-containing media until confluent, before harvest for mRNA and protein confirmation of gene knockdown.

RNA was extracted in duplicate from A2780, A2780olapR and A2780nirapR cells and from paired drug-sensitive and drug-resistant DOCS study patients, as previously described and sent to the Genetics Core, Wellcome Trust Clinical Research Facility, Edinburgh ( https://clinical-research-facility.ed.ac.uk/core-services/genetics/sequencing ) for confirmation of RNA integrity using an Agilent 2100 Bioanalyser, with subsequent library preparation and RNA sequencing analysis performed using Ion Ampliseq Transcriptome Human Gene expression technology (ThermoFisher Scientific). Subsequent bioinformatics analysis was performed by FIOS Genomics ( https://www.fiosgenomics.com ). Quality of sequencing reads was assessed using FastQC and reads aligned to the human genome using STAR aligner. Sample data was normalised using trimmed mean of M-values normalisation and expression values transformed using voom, with Principal Component Analysis (PCA) plots used to illustrate global gene expression profiles. Significant differentially expressed genes (DEGs; fold change threshold≥2) were identified in pairwise comparisons, with p -values corrected for multiple testing using the Benjamini-Hochberg false discovery rate (FDR) adjustment (FDR-adjusted p ≤ 0.05; [ 22 ]). DEGs were represented in volcano plots, plotting log 2 (fold-change) against -log10 (adjusted p -value). Functional enrichment analysis was performed on pairwise comparisons of all significantly expressed genes, comparing KEGG pathways and GO terms.

2.5 × 10 5 A2780 cells were seeded per well of a 6-well plate and incubated for 24 h, before transfection with 2 µg pCMV3.1- ABCG2 plasmid or pCMV3.1 EV control plasmid (both sourced from SinoBiological, Eschborn, Germany) using 5 µL Lipofectamine 300 and 4 µL P3000 reagent in serum-free OptiMEM medium (all sourced from Thermo Fisher, Renfrewshire, UK). Cells were incubated for 24 h before re-seeding into 10 cm 2 dishes using fresh media containing 50 µg/mL hygromycin. Single cells were grown into individual colonies and expanded separately until confluent, prior to confirmation of ABCG2 mRNA and protein expression.

Protein was extracted from 1 × 10 6 cells by scraping in 150 µl RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1% NP-40, 2 mM EDTA) containing protease inhibitor cocktail (SigmaAldrich). Lysates were centrifuged at 2000 × g for 5 min to pellet cellular debris and the resulting supernatant used in a Detergent Compatible (DC) assay (BioRad, Hertfordshire, UK) to determine protein concentration relative to a bovine serum album standard curve (0–2 mg/mL).