Effective treatment of patients with triple-negative (ER-negative, PR-negative, HER2-negative) breast cancer remains a challenge. Although PARP inhibitors are being evaluated in clinical trials, biomarkers are needed to identify patients who will most benefit from anti-PARP therapy. We determined the responses of three PARP inhibitors (veliparib, olaparib, and talazoparib) in a panel of eight triple-negative breast cancer cell lines. Therapeutic responses and cellular phenotypes were elucidated using high-content imaging and quantitative immunofluorescence to assess markers of DNA damage (53BP1) and apoptosis (cleaved PARP). We determined the pharmacodynamic changes as percentage of cells positive for 53BP1, mean number of 53BP1 foci per cell, and percentage of cells positive for cleaved PARP. Inspired by traditional dose–response measures of cell viability, an EC50 value was calculated for each cellular phenotype and each PARP inhibitor. The EC50 values for both 53BP1 metrics strongly correlated with IC50 values for each PARP inhibitor. Pathway enrichment analysis identified a set of DNA repair and cell cycle–associated genes that were associated with 53BP1 response following PARP inhibition. The overall accuracy of our 63 gene set in predicting response to olaparib in seven breast cancer patient-derived xenograft tumors was 86%. In triple-negative breast cancer patients who had not received anti-PARP therapy, the predicted response rate of our gene signature was 45%. These results indicate that 53BP1 is a biomarker of response to anti-PARP therapy in the laboratory, and our DNA damage response gene signature may be used to identify patients who are most likely to respond to PARP inhibition. Mol Cancer Ther; 16(12); 2892–901. ©2017 AACR.

Although the overall survival of patients with breast cancer has improved over the past two decades (1), patients with triple-negative breast cancer (TNBC) have a poor prognosis with shorter disease-free survival and overall survival (2). Lacking expression of estrogen receptor (ER), progesterone receptor (PR), and HER2, triple-negative breast tumors constitute 15% to 20% of all breast cancers, are genomically and phenotypically heterogeneous, and have few effective therapeutic options (3). One of the strongest risk factors associated with development of TNBC is a deleterious mutation in the BRCA1 gene, which is present in 10% to 15% of patients with TNBC (4). PARP inhibitors have shown promise for patients with BRCA1/2 mutations and TNBC (5).

PARP inhibitors have two main mechanisms of action: synthetic lethality and PARP–DNA trapping. The underlying premise for synthetic lethality is that of a two-hit theory: PARP inhibition in combination with defective BRCA1/2 function results in complex chromatid rearrangements and ultimately, cell death (6, 7). PARP inhibitors target PARP1, an enzyme that, when recruited to single-strand breaks, binds to DNA and catalyzes the synthesis of PARP chains onto a series of protein substrates (PARylation). In this process, PARP1 recruits DNA repair proteins and eventually autoPARylates, leading to its release from damaged DNA. PARP inhibitors also have been shown to trap PARP1/2 enzymes on damaged DNA, creating trapped PARP–DNA complexes that induce cytotoxicity (8, 9).

Several PARP inhibitors are currently being tested preclinically and in clinical trials. We focus here on three PARP inhibitors: veliparib (ABT-888, Abbvie), olaparib (AZD2281, AstraZeneca), and talazoparib (Pfizer, formerly called BMN 673). Although all three PARP inhibitors are orally available and have been shown to target PARP1/2 activity, talazoparib has demonstrated the greatest potency in trapping PARP–DNA complexes (8–10). Over 100 clinical trials have been undertaken with PARP inhibitors, and many of these focus on patients with BRCA1/2 mutations (10). Of these three PARP inhibitors, olaparib has the most advanced clinical development, and has been granted FDA approval for use in ovarian cancer (11). Current clinical trials are testing PARP inhibitors as a single agent and in combination with other therapeutic agents in patients with TNBC and other types of cancers (10, 12).

Our goal in this study was to identify pharmacodynamic biomarkers of response, and genes that could predict response to PARP inhibitors. We accomplished this by assessment of PARP responses in a panel of well-characterized TNBC cell lines, and correlation of these responses with pre-treatment molecular features. In our study, we used high-content imaging (13, 14) to measure cellular changes in DNA damage and cell death in response to PARP inhibition. We found the DNA damage response to correlate strongly with IC50 values and identified the genes and critical pathways associated with DNA damage response to PARP inhibition. Finally, we validated the predictive value of our gene signature in a publicly available dataset of patient-derived xenografts (PDXs) and identified the clinical relevance of these genes in breast cancer patients with triple-negative disease.

In vitro drug sensitivity assay

We first performed experiments to identify optimal cell seeding densities to ensure an average of 75% confluence of control cells at the end of the assay. Cells were assessed in 96-well plate format, where each plate tested 9 concentrations of two drugs, distributed in a randomized layout. Perimeter wells were not used. We tested each concentration in triplicate wells and in one to three replicate assays. Cells were plated and allowed to adhere for 24 hours, followed by drug treatment (Fig. 1). Media and drug were changed after 4 to 5 days. For all experiments, cells were treated for a total of 10 days. Cells were then fixed and permeabilized with 4% paraformaldehyde, diluted from stock paraformaldehyde 32% solution, EM grade (cat no. 15714, Electron Microscopy Sciences), and 0.3% Triton X-100 (cat no. T9284, Sigma-Aldrich).

Figure 1.

Workflow used to identify genes and pathways associated with 53BP1 response to PARP inhibition. Abbreviation: PARPi, PARP inhibitor.

Figure 1.

Workflow used to identify genes and pathways associated with 53BP1 response to PARP inhibition. Abbreviation: PARPi, PARP inhibitor.

Close modal

We tested three PARP inhibitors: veliparib, olaparib, and talazoparib (Selleckchem, cat nos. S1004, S1060, S7048, respectively). Chemical formulations can be found in Fig. 2A–C. We used 1:5 serial dilutions, with concentrations optimized for each compound: veliparib and olaparib tested at 0.25 nmol/L to 100 μmol/L; and talazoparib, tested at 0.0128 nmol/L to 5 μmol/L.

Figure 2.

Chemical structures of veliparib (A), olaparib (B), and talazoparib (C). D, IC50 values for veliparib, olaparib, and talazoparib compiled for 8 breast cancer cell lines. Average values for triplicate wells and replicate assays were used. Error bars indicate SEM of replicate assays. Light gray bars indicate BRCA-mutant cell lines, while dark gray bars indicate BRCA wild-type cell lines. Black arrows indicate plasma concentrations of the PARP inhibitors achieved in patients. NA, not available.

Figure 2.

Chemical structures of veliparib (A), olaparib (B), and talazoparib (C). D, IC50 values for veliparib, olaparib, and talazoparib compiled for 8 breast cancer cell lines. Average values for triplicate wells and replicate assays were used. Error bars indicate SEM of replicate assays. Light gray bars indicate BRCA-mutant cell lines, while dark gray bars indicate BRCA wild-type cell lines. Black arrows indicate plasma concentrations of the PARP inhibitors achieved in patients. NA, not available.

Close modal

We studied eight molecularly characterized TNBC cell lines from our laboratory (15): MDAMB436, MDAMB231, MDAMB453, MDAMB468, HCC1143, HCC1937, HCC1806, and HCC1395 (Supplementary Table S1). Short tandem repeat DNA profiling (Genetica DNA Laboratories), performed in October 2014 prior to conduction of chemosensitivity experiments, confirmed cell line authenticity, and PCR analysis verified the absence of mycoplasma. Molecular features of the cell lines, including gene cluster information (15, 16), breast cancer subtype (15, 17), mutational status for BRCA1/2, ATM, ATR (18, 19), and PTEN deficiency status (20), are summarized in Supplementary Table S1.

Immunofluorescence

We prepared a primary and secondary antibody solution using 2% BSA (cat no. 001-000-162, Jackson ImmunoResearch). We used the following primary antibodies: cleaved PARP (cl-PARP; 1:200, cat no. 9546, Cell Signaling Technology) and 53BP1 antibody (1:500, cat no. NB100-904, Novus Biologicals). Secondary antibodies used were Alexa 488 donkey anti-mouse (1:300, cat no. A21202, Life Technologies) and Alexa 647 donkey anti-rabbit (1:300, cat no. A31573, Life Technologies). We used HCS Nuclear Mask (1:2,000, cat no. H10325, Life Technologies) to stain the nucleus, which was added at the time of the secondary antibody solution.

High-content imaging

We performed wide-field microscopy using a scan⁁R microscope (Olympus) alongside an ORCA-R2 CCD Digital Camera (21) with a 10× objective and filter sets for Alexa 488 and Alexa 647. We scanned 25 images per well and performed image analysis with spot identification using scan⁁R analysis software version 2.4.1.1.

Statistical analysis for immunofluorescence

We analyzed the DMSO control wells to identify the baseline level of 53BP1 foci per nucleus and cl-PARP intensity for each cell line.

We calculated two metrics for 53BP1: (i) percentage of cells positive for 53BP1 foci formation; and (ii) mean number of 53BP1 foci per nucleus. In the control wells, we identified the number of foci at the 95th percentile, such that 5% of cells were considered positive for 53BP1 foci formation. We used these thresholds to identify the percentage of positive cells for 53BP1 foci formation for each drug concentration of each cell line. We also determined the mean number of 53BP1 foci per nucleus for each drug concentration of each cell line.

We used a similar approach to analyze the cl-PARP intensity signal. Here, the 99th percentile of the intensity cl-PARP in the DMSO control cells was used as a threshold to divide cells into positive and negative bins. Cells positive for cl-PARP expression were interpreted to be apoptotic. The threshold of 1% was chosen based on what was previously reported (22) and was kept constant across all cell lines. We also calculated the percentage of apoptotic cells for each drug concentration. All single-cell analysis was performed using STATA SE (version 13.1, StataCorp).

Data visualization

For each PARP inhibitor concentration, we created heatmaps to visualize three metrics: (i) percentage of cells positive for 53BP1; (ii) mean number of 53BP1 foci per nucleus; and (iii) percentage of cells positive for cl-PARP. In all cases, data from each drug treatment were normalized to the DMSO control. Values were scaled to the maximum value to compare across cell lines. Heatmaps were created using Multi Experiment Viewer (MeV) software (23). A double gradient color scheme was used, where 10% of the elements were set to the lowest and highest levels of color saturation.

Calculation of EC50 curves

We calculated EC50 metrics for each of the following cellular phenotypes: percentage of cells positive for 53BP1, mean number of 53BP1 foci per cell, and percentage of cells positive for cl-PARP. EC50 is defined as the drug concentration required to induce a response halfway between the baseline and maximum response. For each phenotype, we normalized the values for each well of each drug concentration by subtracting the mean value of triplicate wells for the DMSO control. We divided the normalized values for each drug concentration by the maximal value (mean of triplicate wells). These normalized metrics were plotted against the log-transformed molar drug concentrations with a top constraint of 100. We fitted a sigmoidal curve to these data and interpolated to identify the drug concentration at 50% of the maximal response. Spearman rank correlations were used to correlate EC50 values with IC50 values. EC50 values and correlations were calculated using GraphPad Prism (version 6.0d for Mac OS X, GraphPad Software).

Cell-cycle analysis

We determined the cell-cycle distribution of DNA content by analysis of total DAPI intensity histograms. We first set up gates in the control (DMSO) population of each plate, and then applied these to each treatment well. Constraints were applied to the mean peak population of 2N and 4N, and to the coefficient of variation, such that the coefficient of variation of 4N was set at the same as 2N. We reported the results of cell-cycle analysis when cell numbers per well were greater than 200. We performed all cell-cycle analysis using FlowJo (v10.1r5).

Gene association analysis

We identified genes associated with the 53BP1 response and assessed their clinical significance using the pipeline described in Supplementary Fig. S1. For this analysis, we stratified the cell lines into two groups: sensitive and resistant to PARP inhibition using the percentage of cells positive for 53BP1 response. The PARP inhibitors were grouped to determine the three common sensitive cell lines and three common resistant cell lines. We used previously published gene expression data from untreated cell lines (15) and calculated a log fold change metric by dividing the average gene expression of the sensitive cell lines by the average gene expression of the resistant cell lines to create a rank list. We also created a curated list of gene sets from gene sets previously shown to be associated with response to olaparib (20, 24), talazoparib (25), BRCAness (26), BRCA1 mutation (27), BRCA2 mutation (27), homologous recombination deficiency (HRD; ref. 28), and DNA damage response pathways (29). We used our rank gene list and the curated gene set list to perform a preranked gene set enrichment analysis (GSEA) to identify enriched gene sets and core enriched genes. There are four statistical parameters that are used to describe and interpret the GSEA output. The normalized enrichment score (NES) reflects the degree to which a gene set is overrepresented at the top or bottom of a ranked list of genes, taking into account differences in gene set size. The nominal P value represents the statistical significance of the enrichment score for a single gene set. The FDR is the estimated probability for a false-positive finding of a gene set with a given NES. FDR is adjusted for gene set size and multiple hypothesis testing. The rank metric score is the position of the gene in the ranked list of genes. GSEA was performed using GenePattern (Version 3.9.8, Build Id: 140; ref. 30).

To determine which pathways were statistically associated with the core genes, Reactome Pathway Enrichment analysis was performed within Cytoscape v3.3.0. We identified enriched pathways and the genes associated with these pathways, with an FDR q-value <0.1. The pathways were organized into their hierarchy of major, minor, subpathways, and components, based on the hierarchical organization provided by Reactome within Cytoscape. We considered a pathway to be significant only if the major or minor pathways were statistically significant (FDR q-value <0.1).

We evaluated the clinical significance of our 63 pathway-enriched genes by assessing their frequency of mutation in TNBC patients (n = 82), and ER-positive patients (n = 594) from The Cancer Genome Atlas (TCGA dataset; ref. 31) available in cBioPortal (32). Mutation frequency was also assessed in breast cancer patients with the following subtypes: basal (n = 81), luminal B (n = 133), and luminal A (n = 235; ref. 33).

We tested the predictive potential of our gene signature using previously published expression data of seven breast cancer tumors that were treated with olaparib in a patient-derived xenograft model (34). We normalized the gene expression data as previously described (20) and used a weighted voting algorithm (30). Sensitivity and resistance to olaparib in the PDX model was defined as per Bruna and colleagues (34). We also determined the performance of previously published gene signatures predictive of response to PARP inhibition or BRCAness (20, 24–28) with treatment to olaparib in PDXs. For each of the gene signatures, our 53BP1 cell line response was used to train the response to PARP inhibition, and a weighted voting algorithm was used to predict response in mice. Parameters that were used to compare performance of gene signatures include: overall accuracy [(true positives + true negatives)/(true positives + false positives + true negatives + false negatives)], sensitivity, specificity, positive predictive value, negative predictive value, and a positive test, referring to the sum of tumors wherein the test was sensitive to olaparib.

We also determined the predictive potential of our gene signature and those previously published in a cohort of 82 TNBC patients from TCGA. We obtained FPKM-upper quartile normalized gene expression data from the NCI Genomic Data Commons portal (35) and processed the data as described previously (20). Prediction of response to PARP inhibition was calculated using the same weighted voting algorithm that was used to predict response in the PDX model (30, 36). The prediction of response or sensitivity to PARP inhibition was defined as a positive test.

Talazoparib has greater potency in TNBC cell lines than veliparib or olaparib

We used a 10-day assay to measure responses of eight TNBC cell lines to three PARP inhibitors (Fig. 2; Supplementary Fig. S2; Supplementary Tables S1 and S2). As expected, the two most sensitive cell lines, MDAMB436 and HCC1395, were BRCAMUT (BRCA mutant) and the most resistant cell line, HCC1143, was BRCAWT (BRCA wild type; Fig. 2D; Supplementary Fig. S2). Talazoparib demonstrated the greatest potency, with mean IC50 values ranging between 0.20 and 28.0 nmol/L. Olaparib IC50 values ranged from 0.003 to 3.8 μmol/L, while IC50 values for veliparib varied between 0.03 and 67.1 μmol/L. Olaparib is 5- to 50-fold more potent than veliparib, and talazoparib is 15- to170-fold more potent than olaparib, depending on the cell line (Supplementary Table S3).

To better understand the significance of the IC50 values in the context of patients, we annotated Fig. 2D with the plasma concentrations achieved in patients for each compound. We obtained the peak plasma concentrations of each compound within 24 hours of drug administration from clinical trials. The average plasma concentrations described in the literature are 9.9 μmol/L for veliparib (37, 38), 14 μmol/L for olaparib, and 25 nmol/L for talazoparib (39, 40). Therefore, the IC50 values for olaparib and talazoparib were mainly at physiologic concentrations, whereas the IC50 values for veliparib exceeded the maximum plasma concentrations for three of the eight cell lines.

Variability in DNA damage response and apoptosis across cell lines and PARP inhibitors

We examined the DNA damage response, represented by 53BP1 foci formation following treatment with each PARP inhibitor (Fig. 3A–C; Supplementary Fig. S3). We observed variability in the number of 53BP1 foci across cell lines, with greater number of foci in HCC1806, and the fewest foci in HCC1143. Treatment with olaparib or talazoparib also resulted in cells with larger nuclei in HCC1806 and MDAMB231 cell lines, as compared with HCC1143. Our imaging approach allowed us to quantitatively compare the effect of PARP inhibition on cell cycle and hyperploid cell populations (Supplementary Figs. S4 and S5). In comparison with HCC1143, increasing concentrations of all three PARP inhibitors were associated with an increasingly greater proportion of hyperploid cells in MDAMB231 and HCC1806 cell lines. As HCC1806 is much more sensitive to PARP inhibition than MDAMB231 and HCC1143, it is plausible that an increase in DNA content is associated with response to PARP inhibition. Treatment with olaparib or talazoparib was also associated with a greater proportion of cells undergoing apoptosis, in comparison with veliparib (Fig. 3D and E; Supplementary Fig. S6).

Figure 3.

Cleaved PARP expression and 53BP1 foci. Representative images with 10× objective from high-content imaging of untreated control cells in the left column, and cells treated with veliparib, olaparib, and talazoparib in three right columns. 53BP1 foci are seen in HCC1143 (A), MDAMB231 (B), and HCC1806 (C) cell lines. Blue represents nuclear staining and pink represents 53BP1 foci. Cleaved PARP expression is seen in HCC1143 (D) and HCC1806 (E). Blue, nuclear staining; green, cleaved PARP expression.

Figure 3.

Cleaved PARP expression and 53BP1 foci. Representative images with 10× objective from high-content imaging of untreated control cells in the left column, and cells treated with veliparib, olaparib, and talazoparib in three right columns. 53BP1 foci are seen in HCC1143 (A), MDAMB231 (B), and HCC1806 (C) cell lines. Blue represents nuclear staining and pink represents 53BP1 foci. Cleaved PARP expression is seen in HCC1143 (D) and HCC1806 (E). Blue, nuclear staining; green, cleaved PARP expression.

Close modal

Strong correlation between EC50 values for 53BP1 response and IC50 values

We calculated EC50 values for each of the phenotypic endpoints: percentage of cells positive for 53BP1, number of 53BP1 foci per cell, and percentage of cells positive for cl-PARP (Supplementary Figs. S7 and S8). Overall, the EC50 values for 53BP1, computed from both percentage of cells and mean number of foci, demonstrated a similar trend across all cell lines. The EC50 values for percentage of cells positive for cl-PARP were generally higher than the IC50 values. We performed correlations between IC50 and EC50 values for each PARP inhibitor. Supplementary Figure S9 shows statistically significant correlations for each of the PARP inhibitors between IC50 and EC50 values for percentage of cells positive for 53BP1 for veliparib (r = 0.83, P = 0.01), olaparib (r = 0.96, P = 0.003), and talazoparib (r = 0.93, P = 0.002). We also identified positive correlations between IC50 and EC50 values for 53BP1 foci for veliparib (r = 0.95, P = 0.001), olaparib (r = 0.93, P = 0.007), and talazoparib (r = 0.83, P = 0.01). There were no statistically significant correlations between IC50 values and EC50 values for percentage of cells positive for cl-PARP.

We determined the dose-dependent effect of veliparib, olaparib, and talazoparib upon percentage of cells positive for 53BP1, mean number of 53BP1 foci per cell, and percentage of cells positive for cl-PARP (Fig. 4). For the two 53BP1 metrics (Fig. 4A and B), there is a shift in the response from veliparib to olaparib to talazoparib, with increased 53BP1 foci formation or percentage of cells positive for 53BP1 at progressively lower concentrations for all cell lines. As the concentration range tested was in the same micromolar range for veliparib and olaparib, while the concentration range for talazoparib was mainly in the nanomolar range, this suggests that the 53BP1 response is similar across all three PARP inhibitors, but the differences observed may be attributed to differences in affinity between the PARP inhibitors. Figure 4C shows the dose response of each PARP inhibitor upon apoptosis. The effect is more dichotomous, and the induction of apoptosis occurs near the peak plasma concentrations observed in patients.

Figure 4.

Heatmap of cellular phenotypes as a function of PARP inhibition. Along the x-axis are nine increasing concentrations of each PARP inhibitor, and along the y-axis are eight different cell lines. Color indicates percentage of cells positive for 53BP1 (A), mean number of 53BP1 foci per cell (B), and percentage of cells positive for cleaved PARP (C). Gray arrows refer to peak plasma concentrations of the PARP inhibitors achieved in patients. Gray in the heatmap represets missing values.

Figure 4.

Heatmap of cellular phenotypes as a function of PARP inhibition. Along the x-axis are nine increasing concentrations of each PARP inhibitor, and along the y-axis are eight different cell lines. Color indicates percentage of cells positive for 53BP1 (A), mean number of 53BP1 foci per cell (B), and percentage of cells positive for cleaved PARP (C). Gray arrows refer to peak plasma concentrations of the PARP inhibitors achieved in patients. Gray in the heatmap represets missing values.

Close modal

Gene association analysis of 53BP1 response

One of our aims was to identify a set of genes that was associated with response to all three PARP inhibitors (Supplementary Fig. S1). We created a transcriptome-wide ranked list of differential gene expression between the sensitive and resistant cell lines (Supplementary File S1). This ranked gene list was used in a GSEA analysis to assess whether gene sets previously shown to be involved with response to PARP inhibition, BRCAness, BRCA1/2 mutation status, HRD, or DNA damage repair (20, 24–28) are differentially expressed in sensitive and resistant cell lines (see Supplementary File S2 for gene set lists). The total number of genes from these lists is 1,091, of which there are 919 unique genes. Six of the eight gene sets were significantly differentially expressed between sensitive and resistant cell lines at q-value <0.25 (Supplementary Figs. S10A and S11). Significant gene sets include the two olaparib-associated gene sets, BRCA1 and BRCA2 signatures, BRCAness, and the HRD signature. These gene sets have a negative enrichment score, which indicates that their downregulation is associated with sensitivity to PARP inhibition. We identified 189 core-enriched genes, (Supplementary File S3), of which, 12 of these genes were present in more than two or three gene sets: MCM2, RAD51C, HELLS, ESCO2, TIMELESS, NBN, BRCA1, MCM3, ATAD5, ANLN, FAM83D, and SHCBP1 (Supplementary Fig. S10B). As these genes were derived from two or three independent methods, we consider them to be of high interest.

We were also interested in determining the pathways enriched in the 176 core genes associated with response to PARP inhibition. Using Reactome pathway enrichment analysis, we identified three major pathways: DNA repair, cell cycle, and programmed cell death (Table 1; Supplementary File S4). Major and minor pathways are defined in the Materials and Methods section. Within the major pathway of DNA repair, several minor pathways involved in single-strand break and double-strand break repair were identified including base excision repair, nucleotide excision repair, mismatch repair, and homology-directed repair. Of note, genes implicated in DNA damage bypass, such as translesion synthesis, were also enriched. Cell-cycle genes were also enriched, including checkpoint factors, as well as genes involved in DNA replication, chromosome maintenance, and telomere maintenance.

Table 1.

Summary of enriched pathways associated with 53BP1 response

Reactome pathway (major, minor)Pathway proteinsGene set proteinsPFDRHit genes
Cell cycle 500 40 1.11E−16 1.47E−14  
 Cell-cycle checkpoints 149 18 1.98E−12 9.52E−11 RFC4, PSMA6, RFC2, MRE11A, ANAPC10, CHEK1, RNF168, BRCA1, NBN, TP53, MCM2, MCM3, MCM5, MCM6, SUMO1, PSMD14, PSMD12, BARD1 
 Cell cycle, mitotic 399 25 9.39E−11 2.44E−09 KNTC1, ESCO2, RFC4, PSMA6, RFC2, ZWINT, ANAPC10, POLE2, GINS2, GINS3, GINS4, DHFR, PCNA, SEH1L, PRKCA, MCM2, MCM3, MCM5, MCM6, RRM2, TYMS, BUB1, FEN1, PSMD14, PSMD12 
 Chromosome maintenance 64 2.56E−04 2.44E−03 RFC4, RFC2, POLE2, PCNA, FEN1, TERF1 
 Meiosis 63 10 1.80E−08 3.75E−07 RAD51C, MND1, BLM, MRE11A, BRCA1, NBN, PSMC3IP, RAD51, TOP3A, TERF1 
DNA repair 258 32 1.11E−16 1.47E−14  
 Base excision repair 31 1.30E−08 2.86E−07 LIG3, POLB, RFC4, RFC2, TDG, PCNA, UNG, FEN1 
 DNA damage bypass 44 3.18E−03 0.0191 USP1, RFC4, RFC2, PCNA 
 DNA double-strand break repair 44 2.68E−06 4.29E−05 MRE11A, RNF168, BRCA1, NBN, TP53, SUMO1, BARD1 
 Nucleotide excision repair 102 8.15E−05 8.66E−04 LIG3, RFC4, RFC2, PCNA, INO80D, XPA, POLR2D, SUMO1 
 Mismatch repair 15 1.18E−03 8.62E−03 PCNA, EXO1, MSH2 
 Fanconi anemia pathway 36 0.0133 0.0624 USP1, FANCE, FANCC 
Programmed cell death 154 0.0189 0.0756 MLKL, PSMA6, TP53, DAPK1, PSMD14, PSMD12 
Reactome pathway (major, minor)Pathway proteinsGene set proteinsPFDRHit genes
Cell cycle 500 40 1.11E−16 1.47E−14  
 Cell-cycle checkpoints 149 18 1.98E−12 9.52E−11 RFC4, PSMA6, RFC2, MRE11A, ANAPC10, CHEK1, RNF168, BRCA1, NBN, TP53, MCM2, MCM3, MCM5, MCM6, SUMO1, PSMD14, PSMD12, BARD1 
 Cell cycle, mitotic 399 25 9.39E−11 2.44E−09 KNTC1, ESCO2, RFC4, PSMA6, RFC2, ZWINT, ANAPC10, POLE2, GINS2, GINS3, GINS4, DHFR, PCNA, SEH1L, PRKCA, MCM2, MCM3, MCM5, MCM6, RRM2, TYMS, BUB1, FEN1, PSMD14, PSMD12 
 Chromosome maintenance 64 2.56E−04 2.44E−03 RFC4, RFC2, POLE2, PCNA, FEN1, TERF1 
 Meiosis 63 10 1.80E−08 3.75E−07 RAD51C, MND1, BLM, MRE11A, BRCA1, NBN, PSMC3IP, RAD51, TOP3A, TERF1 
DNA repair 258 32 1.11E−16 1.47E−14  
 Base excision repair 31 1.30E−08 2.86E−07 LIG3, POLB, RFC4, RFC2, TDG, PCNA, UNG, FEN1 
 DNA damage bypass 44 3.18E−03 0.0191 USP1, RFC4, RFC2, PCNA 
 DNA double-strand break repair 44 2.68E−06 4.29E−05 MRE11A, RNF168, BRCA1, NBN, TP53, SUMO1, BARD1 
 Nucleotide excision repair 102 8.15E−05 8.66E−04 LIG3, RFC4, RFC2, PCNA, INO80D, XPA, POLR2D, SUMO1 
 Mismatch repair 15 1.18E−03 8.62E−03 PCNA, EXO1, MSH2 
 Fanconi anemia pathway 36 0.0133 0.0624 USP1, FANCE, FANCC 
Programmed cell death 154 0.0189 0.0756 MLKL, PSMA6, TP53, DAPK1, PSMD14, PSMD12 

Clinical significance of pathway-enriched genes

From the list of enriched major and minor pathways, we identified a unique set of 63 genes. Because inactivation of DNA damage response pathways typically leads to increased genomic instability (29), we analyzed the frequency with which these genes are mutated (including truncating, inframe, and missense mutations) in different breast cancer subtypes using patient datasets from TCGA (31, 33) within cBioPortal (Supplementary Fig. S12A; ref. 32). As expected, the mutational frequency of these genes was enriched in patients with triple-negative, basal breast cancer or ER-negative breast cancers (Supplementary Fig. S12B). Eighty-five percent of the patients with TNBC demonstrated a mutation in these genes, and in total, we observed mutations in 21 of the 63 genes. This contrasts with a frequency of 31% for patients with ER-positive disease (31). The trend for this mutational frequency was also present in the intrinsic subtypes: 90% for basal, 38% for luminal B, and 17% for luminal A subtypes (33).

It is plausible that synergy with anti-PARP therapy may be achieved by targeting members of DNA repair or other DNA damage response pathways. Possible druggable targets of the DNA damage response pathways was extensively reviewed using several approaches including targets with druggable structures, ligand-based approach, network-based approach, and based on the availability of compounds of submicromolar activity or affinity (29, 41). We found that 19 of the 21 mutated genes in these pathways demonstrated druggable potential and are candidates for cotreatment with PARP inhibitors (Supplementary Table S4; ref. 29, 41).

Predictive performance of our gene signature

We determined the predictive value of our gene signature in previously published data of seven patient breast tumors treated with olaparib in a xenograft model (34). Three of these tumors were BRCA1MUT, of which 2 were sensitive to olaparib and 1 was resistant. The clinical characteristics of these tumors were reported previously (34). We found that our combined PARP inhibitor gene signature correctly predicted response in 6 of 7 tumors. Our signature performed comparably with olaparib (24), talazoparib (25), and HRD (28) gene signatures, which predicted response in 5 of 7 tumors, and better than the gene signatures associated with BRCA1/2 mutations (27), which demonstrated a poorer specificity (Table 2).

Table 2.

Performance of gene signatures in a PDX model and TNBC patients

Response to olaparib in PDX (n = 7)TCGA (n = 82)
Gene signaturesAccSensSpecPPVNPVPositive testPositive test
Combined PARP inhibitor 0.86 0.75 1.00 1.00 0.75 0.43 0.45 
Olaparib (Bajrami et al.; ref. 24) 0.71 0.50 1.00 0.50 0.60 0.29 0.49 
Talazoparib (Shen et al.; ref. 25) 0.71 0.75 0.67 0.75 0.67 0.43 0.41 
HRD Deficiency (Peng et al.; ref. 28) 0.71 0.50 1.00 0.50 0.60 0.29 0.29 
Olaparib (Daemen et al.; ref. 20) 0.57 0.25 1.00 1.00 0.50 0.14 0.48 
BRCAness (Konstantinopoulos et al.; ref. 26) 0.57 0.75 0.33 0.60 0.50 0.71 0.78 
BRCA2MUT (Larsen et al.; ref. 27) 0.43 0.50 0.33 0.50 0.33 0.57 0.68 
BRCA1MUT (Larsen et al.; ref. 27) 0.29 0.50 0.00 0.40 0.00 0.71 0.51 
Response to olaparib in PDX (n = 7)TCGA (n = 82)
Gene signaturesAccSensSpecPPVNPVPositive testPositive test
Combined PARP inhibitor 0.86 0.75 1.00 1.00 0.75 0.43 0.45 
Olaparib (Bajrami et al.; ref. 24) 0.71 0.50 1.00 0.50 0.60 0.29 0.49 
Talazoparib (Shen et al.; ref. 25) 0.71 0.75 0.67 0.75 0.67 0.43 0.41 
HRD Deficiency (Peng et al.; ref. 28) 0.71 0.50 1.00 0.50 0.60 0.29 0.29 
Olaparib (Daemen et al.; ref. 20) 0.57 0.25 1.00 1.00 0.50 0.14 0.48 
BRCAness (Konstantinopoulos et al.; ref. 26) 0.57 0.75 0.33 0.60 0.50 0.71 0.78 
BRCA2MUT (Larsen et al.; ref. 27) 0.43 0.50 0.33 0.50 0.33 0.57 0.68 
BRCA1MUT (Larsen et al.; ref. 27) 0.29 0.50 0.00 0.40 0.00 0.71 0.51 

Abbreviations: Acc, overall accuracy; Sens, sensitivity; Spec, specificity; PPV, positive predictive value, NPV, negative predictive value.

We also determined the predictive value of the gene signatures in 82 TNBC patients from TCGA that were not reported to receive anti-PARP therapy (Table 2). Our combined PARP inhibitor gene signature predicted that 45% of these TNBC patients would respond to anti-PARP therapy. Overall, the predicted response rate in TNBC patients was similar to the frequency of a positive test identified in vivo. Because the patient TCGA cohort did not receive anti-PARP therapy, false positives could not be identified, and so our prediction of response rate is not a reflection of overall accuracy.

The efficacy of PARP inhibitors in cell lines has been previously published using different approaches, with varying assay lengths, from 72 hours to 15 days, and different measurements of cell viability, such as sulforhodamine B or AlamarBlue (resazurin; refs. 17, 20, 42). We used an automated approach to measure nuclear counts as a rapid and more direct means of determining therapeutic response after 10 days of treatment. In the context of TNBC, we found that talazoparib had the greatest potency, with IC50 values in the nanomolar range, followed by olaparib and veliparib, with IC50 values in the micromolar range. We also found that talazoparib was about 100-fold more potent than olaparib, in terms of IC50 values, in most of the cell lines. As talazoparib was previously shown to be 100-fold more potent than olaparib at trapping PARP–DNA complexes (9), it is plausible that the PARP-trapping mechanism is mainly responsible for talazoparib's greater potency in therapeutic response.

Our study is the first to use high-content imaging to demonstrate heterogeneity in expression of 53BP1 and apoptosis across multiple breast cancer cell lines. Although semiquantitative and manual approaches have been previously used to measure the level of double-strand or single-strand breaks (43), and to identify the percentage of cells that express γ-H2AX or 53BP1 in response to PARP inhibition (25, 44), high-content imaging allows screening of several cell lines and drug concentrations in a high-throughput manner, followed by single-cell analyses. We found that the EC50 values of 53BP1 foci formation or percentage of cells positive for 53BP1 strongly correlated with IC50 values, suggesting the significance of the DNA damage response as a phenotypic endpoint. Of note, we did not identify a statistically significant correlation between the EC50 values for apoptosis and any of the three PARP inhibitors. This could be due to the presence of alternative mechanisms of cell death, such as mitotic catastrophe, which may result from an accumulation of chromatid aberrations.

Although clinical trials have focused on patients with BRCA1/2 mutations, the search for predictors of BRCAness for breast cancer tumors is ongoing (10). A vast array of methodologies has been used to identify gene signature predictors of response to PARP inhibitors or BRCAness. These include genetic screens using siRNA/shRNA libraries (24, 25, 28), and computational approaches using in vitro response and genes that were previously known to be involved in DNA repair (20). Gene signatures associated with BRCAness, BRCA1 and BRCA2 mutations, have also been derived from breast cancer or ovarian cancer patients without any prior selection for genes involved with DNA repair (26, 27). Our novel approach of comparing the 53BP1 response in sensitive and resistant cell lines provides insight into the pathways associated with response to the three PARP inhibitors.

Our core gene set enrichment analysis identified some of the critical genes previously found to be important in determining response to PARP inhibition. For example, Daemen and colleagues (20) identified 5 genes associated with response to olaparib, namely, BRCA1, NBN, TDG, XPA, and MRE11A. Interestingly, we also identified CDK12 to be a core-enriched gene from Bajrami and colleagues' gene set (24), which has been shown to play a role in resistance to PARP inhibition (45).

We identified key pathways associated with response to PARP inhibition using pathway enrichment analysis. In addition to DNA repair pathways, we also found pathways involved in translesion synthesis, telomere maintenance, as well as cell-cycle and checkpoint factors associated with response to PARP inhibition. This is not surprising as genes associated with transcription, chromatin modification, mitosis, and apoptosis have previously been reported to be associated with PARP function (10). Furthermore, pathways involving translesion synthesis, telomere maintenance, and checkpoint factors have all been shown to be important components of the DNA damage response pathways (29).

We determined the clinical significance of the pathway-enriched genes in breast cancer patients. We found an enrichment in the mutational frequency of our 63 genes in basal and TNBCs in comparison with luminal A, and ER-positive breast cancers, suggesting that our panel of breast cancer cell lines is representative of the genetic aberrations in TNBC patients. This is concordant with what we and others previously demonstrated: panels of breast cancer cell lines capture much of the genomic, transcriptomic, and biological heterogeneity of primary breast tumors (16, 46), and can be used to demonstrate a differential response to therapy (15). We also identified druggable potential of most of the mutated genes, suggesting the possibility for identification of novel therapeutic agents that could be used in combination with anti-PARP therapy.

We further validated our combined PARP inhibitor gene signature on patient-derived breast cancer xenografts and found that the overall accuracy of our gene signature in predicting response to olaparib was 86% in seven tumors. Although there were a small number of tumors in this validation cohort, we still compared the performance of other BRCAness gene signatures. We found that our gene signature was one of the higher performing signatures. We also determined the predicted response rate of our combined PARP inhibitor gene signature to be 45% in TNBC patients.

In summary, we used high-content cell imaging to determine chemosensitivity of PARP inhibitors in a panel of eight breast cancer cell lines. We identified a novel approach to characterize the DNA damage and cell death response. Using gene set and pathway enrichment analysis, we identified gene predictors of 53BP1 response to PARP inhibition. When mutated, these genes are prevalent in TNBC patients and are suggestive of druggable targets that could be used in combination with anti-PARP therapy. The high overall accuracy of our gene signature in PDXs and predicted response rate in TNBC patients lead the way for clinical studies to validate the predictive potential of our gene signature in TNBC patients.

No potential conflicts of interest were disclosed by all authors.

Conception and design: S. Hassan, L.M. Heiser

Development of methodology: S. Hassan, A. Esch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Hassan, A. Esch, J.W. Gray

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Hassan, L.M. Heiser

Writing, review, and/or revision of the manuscript: S. Hassan, J.W. Gray, L.M. Heiser

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

Study supervision: L.M. Heiser

S. Hassan would like to acknowledge her mentor, Dr. André Robidoux, members of her division of Surgical Oncology at the CHUM, and l'Institut de Cancer de Montréal.

This study was conducted with salary support from the TELUS-Canadian Breast Cancer Foundation National Fellowship, Banting Postdoctoral Fellowships Program, administered by the Canadian Institutes of Health Research (CIHR) by the Government of Canada, Ontario Institute for Cancer Research (OICR) by the Government of Ontario, and grant support from Young Investigator Award by the Conquer Cancer Foundation of ASCO, the Evelyn H. Lauder Family, and the Breast Cancer Research Foundation (to S. Hassan), and by the NIH, NCI grant U54 CA 112970 (to J.W. Gray).

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.

1.
Howlader
N
,
Noone
AM
,
Krapcho
M
,
Garshell
J
,
Miller
D
,
Altekreuse
SF
, et al
SEER Cancer Statistics Review, 1975–2012
.
Bethesda, MD
:
NCI
; 
2015
.
2.
Anders
CK
,
Carey
LA
. 
Biology, metastatic patterns, and treatment of patients with triple-negative breast cancer
.
Clin Breast Cancer
2009
;
9
Suppl 2
:
S73
81
.
3.
Metzger-Filho
O
,
Tutt
A
,
de Azambuja
E
,
Saini
KS
,
Viale
G
,
Loi
S
, et al
Dissecting the heterogeneity of triple-negative breast cancer
.
J Clin Oncol
2012
;
30
:
1879
87
.
4.
Gonzalez-Angulo
AM
,
Timms
KM
,
Liu
S
,
Chen
H
,
Litton
JK
,
Potter
J
, et al
Incidence and outcome of BRCA mutations in unselected patients with triple receptor-negative breast cancer
.
Clin Cancer Res
2011
;
17
:
1082
9
.
5.
Rugo
HS
,
Olopade
OI
,
DeMichele
A
,
Yau
C
,
van 't Veer
LJ
,
Buxton
MB
, et al
Adaptive randomization of veliparib-carboplatin treatment in breast cancer
.
N Engl J Med
2016
;
375
:
23
34
.
6.
Lord
CJ
,
Ashworth
A
. 
Mechanisms of resistance to therapies targeting BRCA-mutant cancers
.
Nat Med
2013
;
19
:
1381
8
.
7.
Lord
CJ
,
Tutt
ANJ
,
Ashworth
A
. 
Synthetic lethality and cancer therapy: lessons learned from the development of PARP inhibitors
.
Annu Rev Med
2015
;
66
:
455
70
.
8.
Murai
J
,
Huang
SY
,
Das
BB
,
Renaud
A
,
Zhang
Y
,
Doroshow
JH
, et al
Trapping of PARP1 and PARP2 by clinical PARP inhibitors
.
Cancer Res
2012
;
72
:
5588
99
.
9.
Murai
J
,
Huang
S-YN
,
Renaud
A
,
Zhang
Y
,
Ji
J
,
Takeda
S
, et al
Stereospecific PARP trapping by BMN 673 and comparison with olaparib and rucaparib
.
Mol Cancer Ther
2014
;
13
:
433
43
.
10.
Sonnenblick
A
,
de Azambuja
E
,
Azim
HA
 Jr
,
Piccart
M
. 
An update on PARP inhibitors—moving to the adjuvant setting
.
Nat Rev Clin Oncol
2015
;
12
:
27
41
.
11.
U.S. Food and Drug Administration. Olaparib.
Silver Spring, MD
:
FDA
. Available from: http://www.fda.gov/Drugs/InformationOnDrugs/ApprovedDrugs/ucm427598.htm.
12.
Mateo
J
,
Carreira
S
,
Sandhu
S
,
Miranda
S
,
Mossop
H
,
Perez-Lopez
R
, et al
DNA-repair defects and olaparib in metastatic prostate cancer
.
N Engl J Med
2015
;
373
:
1697
708
.
13.
Pepperkok
R
,
Ellenberg
J
. 
High-throughput fluorescence microscopy for systems biology
.
Nat Rev Mol Cell Biol
2006
;
7
:
690
6
.
14.
Perlman
ZE
,
Slack
MD
,
Feng
Y
,
Mitchison
TJ
,
Wu
LF
,
Altschuler
SJ
. 
Multidimensional drug profiling by automated microscopy
.
Science
2004
;
306
:
1194
8
.
15.
Heiser
LM
,
Sadanandam
A
,
Kuo
WL
,
Benz
SC
,
Goldstein
TC
,
Ng
S
, et al
Subtype and pathway specific responses to anticancer compounds in breast cancer
.
Proc Natl Acad Sci U S A
2012
;
109
:
2724
9
.
16.
Neve
RM
,
Chin
K
,
Fridlyand
J
,
Yeh
J
,
Baehner
FL
,
Fevr
T
, et al
A collection of breast cancer cell lines for the study of functionally distinct cancer subtypes
.
Cancer Cell
2006
;
10
:
515
27
.
17.
Lehmann
BD
,
Bauer
JA
,
Chen
X
,
Sanders
ME
,
Chakravarthy
AB
,
Shyr
Y
, et al
Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies
.
J Clin Invest
2011
;
121
:
2750
67
.
18.
Wellcome Trust Sanger Institute
. 
COSMIC, Catologue of Somatic Mutations in Cancer
.
Cambridge
,
United Kingdom
:
Wellcome Trust Sanger Institute
.
Available from
: http://cancer.sanger.ac.uk/cell_lines.
19.
Tomlinson
GE
,
Chen
TT
,
Stastny
VA
,
Virmani
AK
,
Spillman
MA
,
Tonk
V
, et al
Characterization of a breast cancer cell line derived from a germ-line BRCA1 mutation carrier
.
Cancer Res
1998
;
58
:
3237
42
.
20.
Daemen
A
,
Wolf
DM
,
Korkola
JE
,
Griffith
OL
,
Frankum
JR
,
Brough
R
, et al
Cross-platform pathway-based analysis identifies markers of response to the PARP inhibitor olaparib
.
Breast Cancer Res Treat
2012
;
135
:
505
17
.
21.
Rantala
J
,
Kwon
S
,
Korkola
J
,
Gray
J
. 
Expanding the diversity of imaging-based RNAi screen applications using cell spot microarrays
.
Microarrays
2013
;
2
:
97
114
.
22.
Liu
T
,
Yacoub
R
,
Taliaferro-Smith
LD
,
Sun
S-Y
,
Graham
TR
,
Dolan
R
, et al
Combinatorial effects of lapatinib and rapamycin in triple-negative breast cancer cells
.
Mol Cancer Ther
2011
;
10
:
1460
9
.
23.
Saeed
AI
,
Sharov
V
,
White
J
,
Li
J
,
Liang
W
,
Bhagabati
N
, et al
TM4: a free, open-source system for microarray data management and analysis
.
Biotechniques
2003
;
34
:
374
8
.
24.
Bajrami
I
,
Frankum
JR
,
Konde
A
,
Miller
RE
,
Rehman
FL
,
Brough
R
, et al
Genome-wide profiling of genetic synthetic lethality identifies CDK12 as a novel determinant of PARP1/2 inhibitor sensitivity
.
Cancer Res
2014
;
74
:
287
97
.
25.
Shen
Y
,
Rehman
FL
,
Feng
Y
,
Boshuizen
J
,
Bajrami
I
,
Elliott
R
, et al
BMN 673, a novel and highly potent PARP1/2 inhibitor for the treatment of human cancers with DNA repair deficiency
.
Clin Cancer Res
2013
;
19
:
5003
15
.
26.
Konstantinopoulos
PA
,
Spentzos
D
,
Karlan
BY
,
Taniguchi
T
,
Fountzilas
E
,
Francoeur
N
, et al
Gene expression profile of BRCAness that correlates with responsiveness to chemotherapy and with outcome in patients with epithelial ovarian cancer
.
J Clin Oncol
2010
;
28
:
3555
61
.
27.
Larsen
MJ
,
Kruse
TA
,
Tan
Q
,
Laenkholm
AV
,
Bak
M
,
Lykkesfeldt
AE
, et al
Classifications within molecular subtypes enables identification of BRCA1/BRCA2 mutation carriers by RNA tumor profiling
.
PLoS One
2013
;
8
:
e64268
.
28.
Peng
G
,
Chun-Jen Lin
C
,
Mo
W
,
Dai
H
,
Park
Y-Y
,
Kim
SM
, et al
Genome-wide transcriptome profiling of homologous recombination DNA repair
.
Nat Commun
2014
;
5
:
3361
.
29.
Pearl
LH
,
Schierz
AC
,
Ward
SE
,
Al-Lazikani
B
,
Pearl
FM
. 
Therapeutic opportunities within the DNA damage response
.
Nat Rev Cancer
2015
;
15
:
166
80
.
30.
Reich
M
,
Liefeld
T
,
Gould
J
,
Lerner
J
,
Tamayo
P
,
Mesirov
JP
. 
GenePattern 2.0
.
Nat Genet
2006
;
38
:
500
1
.
31.
Ciriello
G
,
Gatza
ML
,
Beck
AH
,
Wilkerson
MD
,
Rhie
SK
,
Pastore
A
, et al
Comprehensive molecular portraits of invasive lobular breast cancer
.
Cell
2015
;
163
:
506
19
.
32.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
33.
The Cancer Genome Atlas Network
. 
Comprehensive molecular portraits of human breast tumours
.
Nature
2012
;
490
:
61
70
.
34.
Bruna
A
,
Rueda
OM
,
Greenwood
W
,
Batra
AS
,
Callari
M
,
Batra
RN
, et al
A biobank of breast cancer explants with preserved intra-tumor heterogeneity to screen anticancer compounds
.
Cell
2016
;
167
:
260
74
.
35.
Grossman
RL
,
Heath
AP
,
Ferretti
V
,
Varmus
HE
,
Lowy
DR
,
Kibbe
WA
, et al
Toward a shared vision for cancer genomic data
.
N Engl J Med
2016
;
375
:
1109
12
.
36.
Golub
TR
,
Slonim
DK
,
Tamayo
P
,
Huard
C
,
Gaasenbeek
M
,
Mesirov
JP
, et al
Molecular classification of cancer: class discovery and class prediction by gene expression monitoring
.
Science
1999
;
286
:
531
7
.
37.
Kummar
S
,
Kinders
R
,
Gutierrez
ME
,
Rubinstein
L
,
Parchment
RE
,
Phillips
LR
, et al
Phase 0 clinical trial of the poly (ADP-ribose) polymerase inhibitor ABT-888 in patients with advanced malignancies
.
J Clin Oncol
2009
;
27
:
2705
11
.
38.
Puhalla
S
,
Beumer
JH
,
Pahuja
S
,
Appleman
LJ
,
Tawbi
HA-H
,
Stoller
RG
, et al
Final results of a phase 1 study of single-agent veliparib (V) in patients (pts) with either BRCA1/2-mutated cancer (BRCA+), platinum-refractory ovarian, or basal-like breast cancer (BRCA-wt)
.
J Clin Oncol
32
:
5s
, 
2014
(
suppl; abstr 2570
).
39.
Fong
PC
,
Boss
DS
,
Yap
TA
,
Tutt
A
,
Wu
P
,
Mergui-Roelvink
M
, et al
Inhibition of poly(ADP-ribose) polymerase in tumors from BRCA mutation carriers
.
N Engl J Med
2009
;
361
:
123
34
.
40.
de Bono
JS
,
Mina
LA
,
Gonzalez
M
,
Curtin
N
,
Wang
E
,
Henshaw
JW
, et al
First-in-human trial of novel oral PARP inhibitor BMN 673 in patients with solid tumors
.
J Clin Oncol
31
, 
2013
(
suppl; abstr 2580
).
41.
Tym
JE
,
Mitsopoulos
C
,
Coker
EA
,
Razaz
P
,
Schierz
AC
,
Antolin
AA
, et al
canSAR: an updated cancer research and drug discovery knowledgebase.
Nucleic Acids Res
2016
;
44
:
D938
43
.
42.
Yang
W
,
Soares
J
,
Greninger
P
,
Edelman
EJ
,
Lightfoot
H
,
Forbes
S
, et al
Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells
.
Nucleic Acids Res
2013
;
41
:
D955
D61
.
43.
Tice
RR
,
Agurell
E
,
Anderson
D
,
Burlinson
B
,
Hartmann
A
,
Kobayashi
H
, et al
Single cell gel/comet assay: guidelines for in vitro and in vivo genetic toxicology testing
.
Environ Mol Mutagen
2000
;
35
:
206
21
.
44.
Shen
J
,
Peng
Y
,
Wei
L
,
Zhang
W
,
Yang
L
,
Lan
L
, et al
ARID1A deficiency impairs the DNA damage checkpoint and sensitizes cells to PARP inhibitors
.
Cancer Discov
2015
;
5
:
752
67
.
45.
Johnson
SF
,
Cruz
C
,
Greifenberg
AK
,
Dust
S
,
Stover
DG
,
Chi
D
, et al
CDK12 inhibition reverses de novo and acquired PARP inhibitor resistance in BRCA wild-type and mutated models of triple-negative breast cancer
.
Cell Rep
2016
;
17
:
2367
81
.
46.
Prat
A
,
Karginova
O
,
Parker
JS
,
Fan
C
,
He
X
,
Bixby
L
, et al
Characterization of cell lines derived from breast cancers and normal mammary tissues for the study of the intrinsic molecular subtypes
.
Breast Cancer Res Treat
2013
;
142
:
237
55
.

Supplementary data