Purpose: Homologous recombination deficiency (HRD) correlates with platinum sensitivity in patients with ovarian cancer, which clinically is the most useful predictor of sensitivity to PARPi. To date, there are no reliable diagnostic tools to anticipate response to platinum-based chemotherapy, thus we aimed to develop an ex vivo functional HRD detection test that could predict both platinum-sensitivity and patient eligibility to targeted drug treatments.

Experimental Design: We obtained a functional HR score by quantifying homologous recombination (HR) repair after ionizing radiation-induced DNA damage in primary ovarian cancer samples (n = 32). Samples clustered in 3 categories: HR-deficient, HR-low, and HR-proficient. We analyzed the HR score association with platinum sensitivity and treatment response, platinum-free interval (PFI) and overall survival (OS), and compared it with other clinical parameters. In parallel, we performed DNA-sequencing of HR genes to assess if functional HRD can be predicted by currently offered genetic screening.

Results: Low HR scores predicted primary platinum sensitivity with high statistical significance (P = 0.0103), associated with longer PFI (HR-deficient vs. HR-proficient: 531 vs. 53 days), and significantly correlated with improved OS (HR score <35 vs. ≥35, hazard ratio = 0.08, P = 0.0116). At the genomic level, we identified a few unclear mutations in HR genes and the mutational signature associated with HRD, but, overall, genetic screening failed to predict functional HRD.

Conclusions: We developed an ex vivo assay that detects tumor functional HRD and an HR score able to predict platinum sensitivity, which holds the clinically relevant potential to become the routine companion diagnostic in the management of patients with ovarian cancer. Clin Cancer Res; 24(18); 4482–93. ©2018 AACR.

Translational Relevance

Patients with homologous recombination-deficient high-grade serous ovarian cancer (HGSOC) are sensitive to platinum-based chemotherapy and benefit from alternative treatments, such as PARP inhibitors. For this reason, there is clearly a critical need to identify homologous recombination deficiency (HRD) as early as possible. Currently, HRD is mostly diagnosed by genetic testing, which however fails to identify a large proportion of HR-deficient tumors and predict a patient's response to chemotherapy. In this study, we aimed to fill this gap by developing a clinically relevant diagnostic tool to detect functional HRD. We established an ex vivo test in primary HGSOC and quantified the tumor HRD in an HR score. We demonstrated that a low HR score significantly predicts platinum sensitivity and correlates with improved overall survival.

High-grade serous ovarian cancer (HGSOC) is the most common and aggressive type of ovarian cancer (OC) and the fifth most common malignancy occurring in women, meaning that 1 in 70 will develop OC during her lifetime (1).

The current first-line treatment for HGSOC patients consists of cytoreductive surgery and platinum/taxanes-based chemotherapy (2). The role of platinum compounds is to induce extensive DNA damage in the form of DNA crosslinks and, subsequently, double-strand breaks (DSB), that are usually lethal for rapidly proliferating cancer cells, but also cause severe adverse effects for the patients (3–5). Despite a good primary chemotherapy response, the majority of HGSOC patients will relapse within 6–12 months and develop platinum resistance (1). To date, there are no good clinical tools to predict either response to treatment, nor development of resistance.

Cancer cells with functional homologous recombination (HR) repair can overcome platinum-induced DSBs, but, paradoxically, this major DNA repair pathway is often defective in cancers (6). In fact, approximately 50% of HGSOCs are characterized by inactivation of genes required for HR, such as BRCA1 and BRCA2 (7). HR-deficient cancers are unable to properly repair chemotherapy-induced DNA damage, which results in cancer cell death and tumor shrinkage. Unfortunately, HR-deficient cells are known to rely on alternative pathways to repair DNA damage, such as base excision repair (BER), hence invalidating the efficacy of chemotherapy (8). In the last decades, new approaches have been developed to disrupt these alternative pathways and thus selectively treat HR-deficient cancers. PARP inhibitors' (PARPi) targets, PARPs, are enzymes required for BER. Inhibition of PARPs, in an already HR-deficient background, leads to cell death in a synergistic effect best known as “synthetic lethality” (9–12). Accordingly, PARPi have shown promising results in clinical trials when used after cytotoxic chemotherapy (13–15). In these studies, patients with either germline or somatic BRCA1/2 mutations responded better to the PARPi olaparib, niraparib and rucaparib than patients with wild-type BRCA. However, it was also found that clinical response to PARPi was not simply restricted to BRCA1/2-mutated tumors, but generally extended to patients with HR-deficient cancers (16).

To detect homologous recombination deficiency (HRD), targeted next-generation DNA sequencing of HR genes, such as BRCA1/2, has been used (17–19). Despite the sensitivity with which DNA sequencing has been able to predict HRD, a number of limitations exist. First, only half of HRD-positive HGSOCs harbor mutations in HR genes. For the remainder, epigenetic silencing or inactivation of HR-associated genes are mostly responsible for HRD (7). Second, gene sequencing often identifies variants of uncertain significance (VUS), for which there are no clear clinical guidelines. In addition, some tests try to measure the effects of HRD on genome stability, namely loss-of-heterozygosity, large-scale transitions and telomere allelic imbalance, collectively referred to as “genomic scars” (20–22). Still, the presence of genomic scars only indicates that, at some point during tumor formation, there was HRD, not that the same tumor is still HR-deficient at the time of testing. In fact, reversion mutations restoring BRCA1/2 have been reported in several cancers and are associated with resistance to chemotherapy (23–25). Ultimately, all clinically available genetic tests failed to predict response to PARPi (14–15). Efforts to assess functional tumor HRD by detecting the nuclear localization of RAD51, a key protein of HR, have also been made (23, 26–29). The fact that RAD51 becomes detectable only following DNA damage, has however precluded the use of traditional tumor specimens (i.e., paraffin-embedded tissues) collected from untreated patients, thus a few groups have attempted to induce DNA damage ex vivo, using primary cultures of fresh tumor biopsies and ascites (26, 30).

In this study, we set out to develop a robust ex vivo HRD functional assay capable of overcoming the limitations of currently offered genetic testing and predict a patient's response to platinum-based primary chemotherapy and her eligibility for alternative treatments, for example, PARPi. Briefly, we established an ionizing radiation-induced DNA damage test in primary OC cells and measured their ability to perform HR, which we quantified in an HR score. Our functional HR score was able to confidently and significantly predict primary chemotherapy response and, to some extent, even long-term survival. We believe that this assay can become a quick and reliable routine companion diagnostic tool for clinicians to help identify HGSOC patients who would benefit from alternative treatments.

Patient material and clinical data

Thirty-two primary OC and ascites samples, obtained from a cohort of 23 HGSOC patients, were included in the study. The cohort consisted of both platinum-sensitive and -resistant diseases with variable clinical outcome, as described in Table 1.

Table 1.

Clinicopathological characteristics of patients included in the study

No. of patients = 23
Age at diagnosis Median (range) 68 (54–81) 
 ≤68 11 
 ≥69 12 
FIGO stage IIB 
 IIIC 18 
 IVB 
Origin Ovary 20 
 Fallopian tube 
 Peritoneum 
Surgery strategy NACT 12 
 PDS 11 
Residual tumor after surgery No macroscopic 
 <1 cm 
 >1 cm 
 n.a. 
Response to primary therapy Complete response 14 
 Partial response 
 Progressive disease 
Primary chemotherapy regime Carboplatin/paclitaxel 18 
 Othera 
Follow-up time (months) Median (range) 17.5 (2.1–71.4) 
Previous cancers No 17 
 Yes 6 (5 breast, 1 lung) 
No. of patients = 23
Age at diagnosis Median (range) 68 (54–81) 
 ≤68 11 
 ≥69 12 
FIGO stage IIB 
 IIIC 18 
 IVB 
Origin Ovary 20 
 Fallopian tube 
 Peritoneum 
Surgery strategy NACT 12 
 PDS 11 
Residual tumor after surgery No macroscopic 
 <1 cm 
 >1 cm 
 n.a. 
Response to primary therapy Complete response 14 
 Partial response 
 Progressive disease 
Primary chemotherapy regime Carboplatin/paclitaxel 18 
 Othera 
Follow-up time (months) Median (range) 17.5 (2.1–71.4) 
Previous cancers No 17 
 Yes 6 (5 breast, 1 lung) 

Abbreviations: NACT, neoadjuvant chemotherapy; PDS, primary debulking surgery.

aCarboplatin/paclitaxel/bevacizumab.

HGSOCs biopsies were collected in DMEM/F12 medium (Life Technologies) at the Turku University Hospital and aseptically dissociated into single-cell suspension by overnight incubation in 1X collagenase/hyaluronidase solution (STEMCELL Technologies), followed by filtration through a 100 μm cell strainer. Ascites samples were treated with buffered ammonium chloride solution (STEMCELL Technologies) for red blood cell lysis, when necessary. Dissociated tumor cells and ascites were cultured in DMEM/F12 supplemented with glutamine, HEPES, penicillin/streptomycin, B-27 supplement (all Gibco), hEGF and bFGF (Sigma).

Cell lines

The OC cell lines OVCAR-3, OVCAR-8 and Kuramochi were maintained in RPMI-1640 medium (Lonza) supplemented with 10% FBS (Gibco) and 10 μg/mL insulin (only OVCAR cells; Sigma), whereas COV318 were cultured in DMEM medium (Lonza) + 10% FBS.

DNA damage assay and immunostaining

Low passage (passage 2–8) primary patient-derived tumor and ascites cells were seeded in a 96-well plate and allowed to grow for 48 hours. OVCAR-3 and COV318, and OVCAR-8 and Kuramochi cells were allowed to grow for 24 hours and served as HR-proficient and -deficient controls, respectively, in all experiments. To induce DNA damage, cells were exposed to 10 Grays (Gy) of ionizing radiation, whereas non-irradiated cells served as baseline references. Irradiated cells were allowed to recover for 4, 8, and 24 hours before fixation with 2% buffered paraformaldehyde. Fixed cells were incubated with primary antibodies against γH2Ax (ab22551, Abcam), RAD51 (sc-8349, Santa Cruz Biotechnology), cytokeratin 7 (ab9021, Abcam), cyclinA2 (ab16726, Abcam), and cleaved-caspase 3 (9664, Cell Signaling Technology) as previously described (31). Cells were then incubated with fluorescently labelled secondary antibodies (anti-mouse IgG-AlexaFluor 488, A21202; donkey anti-goat IgG-AlexaFluor 568, A31571; and anti-rabbit IgG-AlexaFluor 647, A11057; Life Technologies) and nuclei were counterstained with DAPI. Images were acquired on a CellInsight high-content imaging system (Thermo Scientific) and analyzed and quantified with the Cellomics software (Thermo Scientific).

HR score

Cytokeratin 7 (CK7) was used as a marker to identify epithelial cancer cells, as previously reported (32, 33). We then focused on cyclinA2-positive cells, as HR is restricted to S and G2 phases of the cell cycle (34), and scored cells as HR-positive when the number of RAD51 nuclear foci or the total nuclear immunofluorescence intensity was higher than in irradiated OVCAR-8 and Kuramochi cells (HR-deficient). Finally, the “HR score” was calculated as the percentage of RAD51-positive cells that were also positive for both cyclinA2 and CK7. The samples spontaneously clustered into three distinct groups (as illustrated in Fig. 3A): samples with HR scores similar to OVCAR-8 and Kuramochi (3% and 12% RAD51+ cells), samples with HR scores similar to OVCAR-3 and COV318 (52% and 65% RAD51+ cells), and samples with intermediate scores. We therefore assigned the samples to three categories, as follows: HR-deficient (samples where less than 20% of cyclinA2-positive epithelial cells displayed RAD51 foci), HR-low (20%–35% of triple-positive cells) and HR-proficient (triple-positive cells > 35%). For patients with multiple samples, the overall patient score was calculated as the average of individual samples' HR scores. We analyzed and quantified immunofluorescent images blinded to the patients' clinical response.

DNA sequencing, variants calling, and copy-number alterations analysis, LOH scores and mutational signatures

Detailed information is available in the online Supplementary Methods.

Statistical analysis

Statistical analysis was performed in R using the functions: fisher test, wilcox test, and Kruskal test. All statistical tests were two-sided. For survival analysis, the ‘survival' R package was used and P values were computed using the log-rank method.

Study approval

The study was conducted in accordance with the Declaration of Helsinki ethical guidelines and approved by the Hospital District of Southwest Finland ethics board. All patients recruited in the study signed an informed consent.

Development and validation of the functional HRD assay

Platinum agents induce DNA cross-links (5, 35), which removal generates temporary but potentially toxic DSBs that are then repaired by HR (36). However, the efficacy of platinum chemotherapy also depends on other factors such as cell cycle, as cross-links are usually tolerated in non-proliferating cells, and drug uptake (5). To overcome these issues, we decided to induce DNA damage in the form of DSBs directly using ionizing radiation (IR), rather than platinum agents. We chose 4 OC cell lines, OVCAR-3 and COV318, and OVCAR-8 and Kuramochi, as HR-proficient and HR-deficient controls, respectively (37). Although IR-induced DNA damage, detectable as nuclear γH2Ax foci, was substantial in all cell lines 8 hours post-irradiation (post-IR, Supplementary Fig. S1A–S1D), nuclear RAD51 foci were observed in 52% of OVCAR-3 and 65% of COV318 (Supplementary Fig. S1A and S1B), but only in 3% of OVCAR-8 and 12% of Kuramochi (Supplementary Fig. S1C and S1D). Moreover, when we introduced an S/G2-phase marker, cyclinA2, to identify the proliferating fraction of cells with the potential to perform HR, we found that 74% of OVCAR-3 but only 5% of OVCAR-8 cells displayed RAD51 foci, confirming that these cell lines can be used as HR-proficient and HR-deficient references (Supplementary Fig. S1E and S1F).

We then induced DNA damage in a total of 32 primary OC and ascites samples (Table 1). When we first set out to assess their HR proficiency, we noticed a high intra-sample heterogeneity in tumor versus stroma ratio (Supplementary Fig. S2A–S2E). Thus, using a marker known to be expressed by ovarian epithelial cancers, cytokeratin 7 (CK7), we determined the bona fide tumor population in each sample and then proceeded to measure the RAD51-positive cells among its G2 fraction (referred to as “HR score,” Fig. 1A). Samples were divided into three categories (Fig. 2A), based on their HR score, as follows: HR-deficient (HR score less than 20; Fig. 1B, Fig. 2A), HR-low (HR score 20–35; Fig. 1C, Fig. 2A) and HR-proficient (HR score >35; Fig. 1D, Fig. 2A). Two samples obtained from two different patients could not be scored using the current protocol (lack of G2 cells in the ascites sample, lack of CK7-positive cells in the ovarian tumor; Supplementary Fig. S3) and were therefore excluded from the study.

Figure 1.

Model describing HR score and HR proficiency categories. A, Step 1: Expression of cytokeratin 7 (CK7) is first used to discriminate epithelial cancer cells from stroma and other contaminants that can be found in primary ovarian cancer cultures. CK7-negative cells are excluded from the analysis. Step 2: CyclinA2 is used as a marker for cells in G2 phase, when HR can be used. CyclinA2-negative cells (non-G2) are excluded from the analysis. Step 3: Finally, RAD51 is used to identify cells that are repairing radiation-induced DNA damage via HR. Step 4: HR score is calculated as the percentage of RAD51-positive cells among epithelial cancer cells (CK7+) in G2 phase (CyclinA2+). B–D, Samples are divided into three categories, according to their HR score: HR-deficient (less than 20% of epithelial cells in G2 are positive for RAD51, example in B), HR-low (20%–35% of G2-epithelial cells express RAD51, example in C), and HR-proficient (>35% of RAD51-positivity among G2 epithelial cells, example in D). White arrows highlight RAD51-positive cancer cells in G2 (RAD51+/CK7+/CyclinA2+), whereas arrowheads depict RAD51-positive stromal cells (non-cancer, RAD51+/CyclinA2+, excluded from analysis). Note how the absolute number of RAD51+ cells in B and C is similar. However, the majority of RAD51+ cells in B are CK7-negative, indicating HR deficiency in cancer (epithelial) cells, but HR proficiency in stromal (normal) cells. The insets in the bottom show a higher magnification of RAD51 foci from the area marked with the dashed line; scale bars, 500 μm.

Figure 1.

Model describing HR score and HR proficiency categories. A, Step 1: Expression of cytokeratin 7 (CK7) is first used to discriminate epithelial cancer cells from stroma and other contaminants that can be found in primary ovarian cancer cultures. CK7-negative cells are excluded from the analysis. Step 2: CyclinA2 is used as a marker for cells in G2 phase, when HR can be used. CyclinA2-negative cells (non-G2) are excluded from the analysis. Step 3: Finally, RAD51 is used to identify cells that are repairing radiation-induced DNA damage via HR. Step 4: HR score is calculated as the percentage of RAD51-positive cells among epithelial cancer cells (CK7+) in G2 phase (CyclinA2+). B–D, Samples are divided into three categories, according to their HR score: HR-deficient (less than 20% of epithelial cells in G2 are positive for RAD51, example in B), HR-low (20%–35% of G2-epithelial cells express RAD51, example in C), and HR-proficient (>35% of RAD51-positivity among G2 epithelial cells, example in D). White arrows highlight RAD51-positive cancer cells in G2 (RAD51+/CK7+/CyclinA2+), whereas arrowheads depict RAD51-positive stromal cells (non-cancer, RAD51+/CyclinA2+, excluded from analysis). Note how the absolute number of RAD51+ cells in B and C is similar. However, the majority of RAD51+ cells in B are CK7-negative, indicating HR deficiency in cancer (epithelial) cells, but HR proficiency in stromal (normal) cells. The insets in the bottom show a higher magnification of RAD51 foci from the area marked with the dashed line; scale bars, 500 μm.

Close modal
Figure 2.

Low HR score correlates with platinum sensitivity, CR to primary chemotherapy and longer disease-free intervals. A, Patients with HR-deficient and HR-low tumors (HR-deficient: green, n = 6; HR-low: yellow, n = 10) are sensitive to platinum-based treatment, whereas those with HR-proficient tumors (red, n = 5) are platinum resistant. Low HR scores significantly correlate with platinum sensitivity (P = 0.012, Fisher test). B, Low HR score patients achieved CR/PR after primary chemotherapy (P = 0.006, Fisher test). All PD patients belonged to the HR-proficient category. C, In relapsing patients, low HR scores correlated with longer PFI (platinum-free intervals). D, Comparison of the HR score with other prognostic clinical parameters. HR scores lower than 35 (HR-deficient/low groups) significantly correlated with longer overall survival (HR, 0.081; 95% CI, 0.011–0.57, P = 0.0116), compared with HR scores above 35 (HR-proficient). Hazard ratios and confidence intervals, and P values were calculated in GraphPad Prism 6.0h using the Mantel–Haenszel and Mantel–Cox (logrank) test, respectively.

Figure 2.

Low HR score correlates with platinum sensitivity, CR to primary chemotherapy and longer disease-free intervals. A, Patients with HR-deficient and HR-low tumors (HR-deficient: green, n = 6; HR-low: yellow, n = 10) are sensitive to platinum-based treatment, whereas those with HR-proficient tumors (red, n = 5) are platinum resistant. Low HR scores significantly correlate with platinum sensitivity (P = 0.012, Fisher test). B, Low HR score patients achieved CR/PR after primary chemotherapy (P = 0.006, Fisher test). All PD patients belonged to the HR-proficient category. C, In relapsing patients, low HR scores correlated with longer PFI (platinum-free intervals). D, Comparison of the HR score with other prognostic clinical parameters. HR scores lower than 35 (HR-deficient/low groups) significantly correlated with longer overall survival (HR, 0.081; 95% CI, 0.011–0.57, P = 0.0116), compared with HR scores above 35 (HR-proficient). Hazard ratios and confidence intervals, and P values were calculated in GraphPad Prism 6.0h using the Mantel–Haenszel and Mantel–Cox (logrank) test, respectively.

Close modal

Low HR scores correlates with platinum sensitivity, longer time-to-progression, and improved overall survival

When we matched the HR scores with patients clinical data, we found a significant correlation between low HR scores and platinum sensitivity (P = 0.008, Table 2): all HR-deficient and 8/10 of HR-low samples were from platinum-sensitive patients. Conversely, all but one sample with HR score >35 (HR-proficient) were from patients who did not respond to platinum (Fig. 2A). 2/10 HR-low patients were also platinum-resistant. The HR score outperformed surgical strategy (primary debulking surgery, PDS, vs neo-adjuvant chemotherapy, NACT, P = 0.14), FIGO stage (IIIC vs. IVB, P = 1) and age at diagnosis (≤68 vs. ≥69, P = 0.064) in predicting primary platinum sensitivity. We also found a significant association when correlating the patient HR category to response to platinum (HR score <35 vs. ≥35, P = 0.014).

Table 2.

Correlations between HR score and treatment outcome

Correlations between HR score and treatment outcome
Correlations between HR score and treatment outcome

Despite achieving initial complete response (CR) to primary treatment, the majority of HGSOC patients will eventually relapse and die of cancer. In our cohort, 13/21 patients (62%) achieved CR (by RECIST/GCIG criteria; ref. 38) after primary chemotherapy, 5/21 (24%) obtained partial response (PR), whereas 3/21 (14%) were classified as having progressive disease (PD) and progressed right after or during treatment (Table 2), which is in line with other HGSOC studies (1, 39). The HR category strongly predicted therapy outcome (P = 0.006), and outperformed both choice of surgical strategy (P = 0.076) and FIGO stage (P = 0.0201).

Among patients who initially achieved CR (n = 13), 31% relapsed, and 15.5% died of the disease between 1.5 and 3 years from diagnosis. Of the PR patients (n = 5), 60% relapsed and one died after 6 years, while all PD patients (n = 3) relapsed within 2 and 4 months and died between 2 months and 2 years. All HR-deficient patients (n = 6) achieved CR and, whereas 2/6 relapsed, all were still alive after a follow-up of 2 and 4.5 years (Table 2, Supplementary Fig. S4A. Average follow-up: 21.7 months; range: 8–55 months). Of the HR-low patients (n = 10), 50% achieved initial CR and 50% PR. A CR patient progressed within one year from diagnosis, and died 6 months later, while a PR patient relapsed after 20 months, and died after 6 years from diagnosis. The average follow-up time was similar as for HR-deficient patients (23 months; range, 9–71 months. Table 2, Fig. 2D; Supplementary Fig. S4A), but the average time-to-progression was significantly shorter (12.6 vs. 22.2 months). In the HR-proficient category, all PD patients (3/5) progressed between 2 and 4 months from diagnosis, and died within 2 years (average time-to-progression: 4.6 months). Of the 2 HR-proficient patients who initially achieved CR, one progressed at 9 months, and died of the disease 3 years later, whereas the other was still alive at the 12 months follow-up. Because 80% of HR-proficient patients died, the follow-up time for this category was substantially shorter than for HR-deficient/-low patients (average: 17.4 months; range, 2–36 months. Table 2, Supplementary Fig. S4A).

Among patients who relapsed, the average platinum-free interval (PFI) for the HR-deficient category was 531 days, 179 days for HR-low, and 53 days for HR-proficient (Table 2; Fig. 2C). Essentially, when a patient relapsed, a low HR score correlated with longer PFI (Fig. 2C).

Ultimately, an HR score <35 was the only factor to significantly correlate with improved OS (HR score <35 vs. ≥35, HR, 0.081; 95% confidence interval (CI); 0.011–0.57; P = 0.0112. Fig. 2D). No correlation was found with choice of surgical strategy (PDS vs. NACT; HR, 0.173; 95% CI, 0.029–1.028; P = 0.0536), FIGO stage (IIIC vs. IVB; HR, 2.291; 95% CI, 0.333–15.74; P = 0.3993) or age at diagnosis (≤68 vs. ≥69; HR, 0.378; 95% CI, 0.06–2.34; P = 0.7589).

HR-deficient tumors display prolonged ionizing radiation-induced DNA damage

The efficacy of chemotherapy is the result of tumor cell death in response to toxic levels of DNA damage (5). On this basis, we hypothesized that DNA-damaging agents would have a more profound effect in HR-deficient than in HR-proficient tumors. We found that, regardless of the tumor HR proficiency, DNA damage was present in more than 80% of cells 4 and 8 hours post-IR (Fig. 3B; Supplementary Fig. S5B; 4 hours not shown). However, in cells from HR-low and HR-proficient tumors, we measured a highly significant decrease in DNA damage 24 hours post-IR (−18% and −28%, P < 0.0001). In contrast, the DNA damage decrease in HR-deficient samples was marginal (−7%, P = 0.0015). Compared with 8 and 24 hours post-IR HR-proficient tumors, particularly, cells from HR-deficient samples retained significantly higher levels of DNA damage (P < 0.0001, Fig. 3B; Supplementary Fig. S5B).

Figure 3.

HR-deficient tumors suffer from prolonged radiation-induced DNA damage. A, Distribution of individual samples, according to their HR score. Bars with a striped pattern show the percentage of RAD51+ cells in the HR-deficient OVCAR-8 and Kuramochi (striped in green, HR scores = 2.8 and 12.1) and HR-proficient OVCAR-3 and COV318 (striped in red, HR score = 52 and 65.5) ovarian cancer cell lines. RAD51-positivity for OVCAR-8, Kuramochi, OVCAR-3 and COV318 was calculated as RAD51+/total cell number. B, Quantification of basal DNA damage (grey) and IR-induced DNA damage in individual ovarian cancer samples, 8 and 24 hours post-IR, as percentage of phosphorylated histone H2Ax (γH2Ax) positive cells/total. HR-deficient tumors (green) displayed permanently high levels of DNA damage (−7% 24 hours post-IR, P = 0.0015). In contrast, HR-low (yellow) and HR-proficient (red) tumors showed a significant decrease in DNA damage levels 24 hours post-IR (−18% and −28%, respectively. P < 0.0001, Fisher test). At 24 hours post-IR, HR-deficient tumors retained considerable more DNA damage than HR-proficient samples (P < 0.0001). Plot shows means ± SD. B′–B″, DNA damage and HR-mediated repair were detected by immunostaining for γH2Ax (green) and RAD51 (magenta), respectively. The inserts show a higher magnification of both γH2Ax and RAD51 foci from the area marked with the dashed line. Representative examples for each HR category are shown for 8 (B′) and 24 hours (B″) post-IR. C, Quantification of basal levels of apoptosis (gray) and IR-induced apoptosis in individual ovarian cancer samples, 8 and 24 hours post-IR, represented as percentage of cleaved caspase 3 (cCasp3) positive cells/total. Note how apoptosis was elevated in all HR categories at 8 hours, but significantly reduced 24 hours post-IR in HR-low and -proficient tumors (−10% and −13%, P < 0.0001), while the decrease in HR-deficient samples was marginal (−6%, P = 0.0014). Plot shows means ± SD. C′–C″, Representative examples of staining for the apoptotic marker cCasp3 (magenta) in each HR category, detected 8 (C′) and 24 hours (C″) post-IR. F-actin (green) was used as a marker to visualize cellular structure and boundaries; scale bars, 500 μm. **, P < 0.005; ****, P < 0.0001. P values were calculated in GraphPad Prism 6.0h using a two-tailed Fisher test.

Figure 3.

HR-deficient tumors suffer from prolonged radiation-induced DNA damage. A, Distribution of individual samples, according to their HR score. Bars with a striped pattern show the percentage of RAD51+ cells in the HR-deficient OVCAR-8 and Kuramochi (striped in green, HR scores = 2.8 and 12.1) and HR-proficient OVCAR-3 and COV318 (striped in red, HR score = 52 and 65.5) ovarian cancer cell lines. RAD51-positivity for OVCAR-8, Kuramochi, OVCAR-3 and COV318 was calculated as RAD51+/total cell number. B, Quantification of basal DNA damage (grey) and IR-induced DNA damage in individual ovarian cancer samples, 8 and 24 hours post-IR, as percentage of phosphorylated histone H2Ax (γH2Ax) positive cells/total. HR-deficient tumors (green) displayed permanently high levels of DNA damage (−7% 24 hours post-IR, P = 0.0015). In contrast, HR-low (yellow) and HR-proficient (red) tumors showed a significant decrease in DNA damage levels 24 hours post-IR (−18% and −28%, respectively. P < 0.0001, Fisher test). At 24 hours post-IR, HR-deficient tumors retained considerable more DNA damage than HR-proficient samples (P < 0.0001). Plot shows means ± SD. B′–B″, DNA damage and HR-mediated repair were detected by immunostaining for γH2Ax (green) and RAD51 (magenta), respectively. The inserts show a higher magnification of both γH2Ax and RAD51 foci from the area marked with the dashed line. Representative examples for each HR category are shown for 8 (B′) and 24 hours (B″) post-IR. C, Quantification of basal levels of apoptosis (gray) and IR-induced apoptosis in individual ovarian cancer samples, 8 and 24 hours post-IR, represented as percentage of cleaved caspase 3 (cCasp3) positive cells/total. Note how apoptosis was elevated in all HR categories at 8 hours, but significantly reduced 24 hours post-IR in HR-low and -proficient tumors (−10% and −13%, P < 0.0001), while the decrease in HR-deficient samples was marginal (−6%, P = 0.0014). Plot shows means ± SD. C′–C″, Representative examples of staining for the apoptotic marker cCasp3 (magenta) in each HR category, detected 8 (C′) and 24 hours (C″) post-IR. F-actin (green) was used as a marker to visualize cellular structure and boundaries; scale bars, 500 μm. **, P < 0.005; ****, P < 0.0001. P values were calculated in GraphPad Prism 6.0h using a two-tailed Fisher test.

Close modal

Likewise, at 8 hours post-IR the fraction of apoptotic cells was similar among samples from different HR-proficiency categories (HR-deficient: 55% ± 7%; HR-low: 53% ± 5%; HR-proficient: 47% ± 7%, Fig. 3C). At 24 hours post-IR, HR-low, and HR-proficient samples showed a significant reduction in apoptosis (−10% and −13%, P < 0.0001), whereas the decrease was less pronounced in HR-deficient samples (−6%, P = 0.0014; Fig. 3C). This is likely due to the impaired DNA repair ability, which was also mirrored in a reduced cellular proliferation following DNA damage (Supplementary Fig. S5D).

Independent tumors from a single patient can display different HRD

BRCA sequencing and HRD genetic tests are usually performed on a single sample per patient (20, 40, 41). However, at the time of diagnosis, HGSOC typically affects several abdominal tissues or organs and, while probably deriving from the same primary tumor, each cancer possesses significant clonal, spatial and temporal heterogeneity, and can develop its independent properties and characteristics (16, 42, 43). DNA repair capacity can be one of these properties.

In our cohort, samples from different anatomical locations collected from the same patient, although having slightly different numerical HR scores, generally fell into the same HR-proficiency category with the exception of one patient, EOC415, from whom samples from five different locations were obtained during the same surgery (Table 2, Fig. 4). The patient was a good responder to chemotherapy and achieved CR after primary treatment. In the ex vivo HR assay, cancer cells isolated from omental metastasis and left ovary were classified as HR-deficient (HR scores 4.8 and 20), a peritoneum-derived tumor and ascites fell into the HR-low category (26.7 and 28.6), and a sample derived from the right ovary was allocated into the HR-proficiency category (HR score 48.7, Fig. 4). The resulting patient average HR score was 24 (HR-low). Among the four HR-deficient/-low samples, lower HR scores generally associated with elevated DNA damage and apoptosis, but low proliferation rate (Supplementary Fig. S6). Conversely, DNA damage and apoptosis were noticeably lower in the HR-proficient ovary sample, which also had the highest proliferation rate (Supplementary Fig. S5). It is very possible that these HR-proficient, highly proliferative tumors are accountable for relapses after initial CR to primary chemotherapy. This result also suggests that HRD screening should not focus on a single biopsy.

Figure 4.

Tumors from different anatomical locations of a single patient display different HR properties. A, Anatomical origins of five different tumors obtained from patient EOC415. B, Examples of immunostaining for cytokeratin 7 (CK7, green), CyclinA2 (orange), and RAD51 (magenta) that was used to calculate HR scores; scale, bars, 500 μm. C, Individual HR scores. HR scores range from HR-deficient (4.8 for omental metastasis and 20 for left ovary, green bars) to fully HR-proficient (48.7 for right ovary, red bar). The patient overall score (24, HR-low and depicted as yellow striped bar) shown is the average calculated from the five locations.

Figure 4.

Tumors from different anatomical locations of a single patient display different HR properties. A, Anatomical origins of five different tumors obtained from patient EOC415. B, Examples of immunostaining for cytokeratin 7 (CK7, green), CyclinA2 (orange), and RAD51 (magenta) that was used to calculate HR scores; scale, bars, 500 μm. C, Individual HR scores. HR scores range from HR-deficient (4.8 for omental metastasis and 20 for left ovary, green bars) to fully HR-proficient (48.7 for right ovary, red bar). The patient overall score (24, HR-low and depicted as yellow striped bar) shown is the average calculated from the five locations.

Close modal

Lack of mutations in BRCA1 and BRCA2 genes does not exclude functional HRD

Germline mutations in BRCA1 and BRCA2 occur in 8% and 6% of HGSOCs, respectively, whereas their somatic mutations are responsible for another 6% of cases, and BRCA1 promoter methylation is found in approximately 10% of HGSOCs (7). However, there are other genes involved in HR, as well as other players that regulate the expression of HR genes and/or interact with and modulate the function of HR proteins (7). Disruption of any of these leads to HRD and to a phenotype that resembles that of BRCA-mutated tumors (“BRCAness,” ref. 44).

The primary aim of this study was to develop a clinically relevant test to measure functional HRD. However, we also decided to analyze the same tumors at the genomic level as, to date, HRD screening in OC patients focuses heavily on genetic tests, rather than functional assays. We performed whole genome (n = 11), whole exome (n = 8), or targeted exome sequencing (n = 2) on tumors from 20 patients in our study cohort (Table 3). In addition, samples from 11 patients were also analyzed for expression of selected genes by RT-qPCR.

Table 3.

Mutational landscape of the samples included in the study

Mutational landscape of the samples included in the study
Mutational landscape of the samples included in the study

We detected a pathogenic BRCA1-truncating mutation (p.E402*) in the HR-deficient patient with the lowest HR score and 3 BRCA1 missense VUS in 3 other HR-deficient/-low patients. Possibly deleterious BRCA2 mutations were found in 3 patients (2 HR-low, 1 HR-proficient), whereas another HR-proficient patient carried a BRCA2 VUS. In the HR-low patient carrying the BRCA2 truncating mutation p.K3326*, we measured lower BRCA2 mRNA levels (∼40% of control, RT-qPCR data not shown), suggesting indeed a deleterious effect. We also found missense VUS in other HR-related genes (RAD51B, RAD51C, BRIP1, and CHEK1), and amplifications/upregulations of genes known to be deregulated in HGSOC (EMSY, CCNE1, and MYC). Besides TP53 mutations and the BRCA1 truncation, however, no clearly pathogenic/deleterious mutation in any HR- or HR-related gene could explain the HRD we measured at the functional level. Next, we assessed the tumors LOH status (loss-of-heterozygosity) and mutational signature, as HRD is associated with increased LOH and a specific mutational pattern (45–47). Generally, we obtained higher LOH scores in samples from HR-deficient/-low patients, but we could not find a strong correlation. However, we found that HR-deficient/low tumors were often characterized by the mutational signature 3 (Table 3; Supplementary Fig. S7), which is indeed associated with HRD (48–50). Oddly, the patient carrying the BRCA2 VUS (EOC933) displayed the genomic scars of HRD (high LOH and mutational signature 3), but functionally performed as HR-proficient.

Although this result could be influenced by a number of technical factors (e.g., tumor purity, sequencing coverage, reads depth and filtering criteria during analysis), it also highlights the fact that HRD-induced genomic instability can be troublesome and time consuming to identify and interpret, suggesting that a phenotype over genotype approach could simplify the diagnosis of clinically relevant HRD.

HGSOC remains one of the most lethal malignancy among women (1). The main obstacles hindering a more effective cancer management are late stage diagnosis, inadequate chemotherapy and lack of reliable methods to predict treatment response. As many HGSOCs are homologous recombination-defective, several groups have tried to develop predictive tools based on the detection of HR genes mutations, such as BRCA1 and BRCA2, and the distinct genomic signatures associated with defective HR (16). Although being relatively sensitive in detecting HRD, genetic testing has, however, its limitations. In a recent study, for example, HRD genetic screening failed to consistently predict clinical response to the PARPi niraparib (14), whereas Coleman and colleagues and Swisher and colleagues (15, 51) reported good response to rucaparib in platinum-sensitive, recurrent HGSOC patients, regardless of BRCA1/2 mutational status.

As HRD correlates with high platinum sensitivity, which clinically is the most useful predictor of sensitivity to PARPi (40), PARPi are now approved by both FDA and EMA for the management of relapsed platinum-sensitive HGSOC. However, platinum sensitivity can only be determined after completing chemotherapy and, therefore, cannot be used as a proxy to measure HRD during primary treatment.

Here, we demonstrate how, for the first time, a functional HRD test can predict real-life treatment response with high confidence. Our HR score outperformed other clinical and pathological parameters, such as choice of surgical strategy, FIGO stage and age at diagnosis, in predicting platinum sensitivity (P = 0.008), primary therapy outcome (P = 0.006) and OS (P = 0.0112). In addition, we showed that tumors can be HR-deficient and exhibit functional “BRCAness” phenotype (platinum-sensitivity, good response to treatment and improved OS) in the absence of mutations in BRCA1/2 or other HR-related genes. It is already known that inactivation of BRCA1/2 (by mutations or promoter inactivation) accounts only for up to 20% of HGSOCs, whereas another 25% to 30% of tumors are HR-defective because of alterations in other HR genes or HR-related factors (7). Possibly, this group could also benefit from PARPi treatment, as suggested by studies reporting clinical response in non-BRCA HGSOC patients (14, 15, 51). Although we found that HR-deficient tumors in our cohort were generally characterized by LOH and the HRD-related mutational signature 3, our results suggest that using a functional HRD test instead of the currently offered genetic screening could nevertheless identify more HR-deficient HGSOC patients who are eligible for other targeted drugs, such as PARPi, whereas being both quicker and more affordable than sequencing. In addition, a test able to predict treatment response before the completion of primary chemotherapy could help clinicians to identify poor responders, who could, wherever possible, be offered alternative or combinational treatments.

In its current state, however, our study has a few limitations. First, the predictive power of the functional HR score needs to be validated in a larger cohort. Second, the HR score cannot be calculated for samples that are non- or low-proliferating in culture, as HR is highly cell-cycle dependent. Furthermore, establishing a primary cell culture from solid tumors and ascites is too time consuming and laborious for routine clinical use. To begin to address these last two issues, we performed the functional HRD assay on a few solid HGSOC biopsies and found that their HR status was fully in line with the HR score of their respective primary cultured cells (Supplementary Fig. S8). Thus, our functional HRD assay can be readily adapted as a routine test in clinical laboratories. We also showed that, within one patient, tumors from independent anatomical locations can have different DNA repair proficiencies, as reported in other studies (52–53). This suggests that, within one patient, HR-proficient tumors might be responsible for relapse. To avoid misdiagnosis, this should be taken into consideration when performing both genetic and functional tests.

In conclusion, we have developed a robust assay to measure HRD at the functional level that has the potential of generating results rapidly, as quickly as 2 to 3 weeks (less than a week for solid biopsies), while overcoming, at the same time, the limitations of currently offered genetic testing (interpretation of uncertain variants, lack of information on epigenetic inactivation etc.). With the required further validations, we envision that our functional HR score can become the routine companion diagnostic tool in the management of HGSOC patients.

No potential conflicts of interest were disclosed.

Conception and design: M. Tumiati, S. Hietanen, S. Grénman, L. Kauppi

Development of methodology: M. Tumiati, S. Hietanen, L. Kauppi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Tumiati, S. Hietanen, J. Hynninen, E. Pietilä, A. Färkkilä, K. Kaipio, K. Huhtinen, K. Lehti, S. Grénman, O. Carpén, L. Kauppi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Tumiati, S. Hietanen, J. Hynninen, A. Alkodsi, Y. Li, L. Kauppi

Writing, review, and/or revision of the manuscript: M. Tumiati, S. Hietanen, J. Hynninen, A. Färkkilä, K. Huhtinen, A. Alkodsi, S.K. Hautaniemi, S. Grénman, O. Carpén, L. Kauppi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Tumiati, S. Hietanen, J. Hynninen, E. Pietilä, A. Färkkilä, P. Roering, K. Huhtinen, R. Lehtonen, E.P. Erkan, M.M. Tuominen, A. Vähärautio

Study supervision: M. Tumiati, S. Hietanen, R. Lehtonen, S. Grénman, L. Kauppi

The authors thank the Biomedicum Functional Genomics Unit (FuGU) and the Biomedicum Imaging Unit (BIU) at the University of Helsinki for excellent technical support.

Financial support for the project was provided by the University of Helsinki (to M. Tumiati, E. Pietilä, K. Lehti, and M.M. Tuominen), the Academy of Finland (to L. Kauppi, A. Vähärautio, and E.P. Erkan), the K. Albin Johansson Foundation (to E. Pietilä and K. Lehti), the Finnish Cancer Institute (to E. Pietilä and K. Lehti), the Sigrid Juselius Foundation (to E. Pietilä and K. Lehti), the Integrative Life Science Doctoral Program (to E. Pietilä), and the European Union's Horizon 2020 research and innovation programme under grant agreement No 667403 for HERCULES (to A. Alkodsi, Y. LI, and S.K. Hautaniemi).

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.
Siegel
RL
,
Miller
KD
,
Jemal
A
. 
Cancer statistics, 2017
.
CA Cancer J Clin
2017
;
67
:
7
30
.
2.
Morgan
RJ
,
Armstrong
DK
,
Alvarez
RD
,
Bakkum-Gamez
JN
,
Behbakht
K
,
Chen
LM
, et al
Ovarian cancer, version 1.2016, NCCN clinical practice guidelines in oncology
.
J Natl Compr Canc Netw
2016
;
14
:
1134
63
.
3.
Johnson
NP
,
Butour
J-L
,
Villani
G
,
Wimmer
FL
,
Defais
M
,
Pierson
V
, et al
Metal antitumor compounds: the mechanism of action of platinum complexes
.
Prog Clin Biochem Med
1989
:
1
24
.
4.
Siddik
ZH
. 
Cisplatin: mode of cytotoxic action and molecular basis of resistance
.
Oncogene
2003
;
22
:
7265
79
.
5.
Dasari
S
,
Bernard Tchounwou
P
. 
Cisplatin in cancer therapy: molecular mechanisms of action
.
Eur J Pharmacol
2014
;
740
:
364
78
.
6.
Dietlein
F
,
Thelen
L
,
Reinhardt
HC
. 
Cancer-specific defects in DNA repair pathways as targets for personalized therapeutic approaches
.
Trends Genet
2014
;
30
:
326
39
.
7.
Konstantinopoulos
PA
,
Ceccaldi
R
,
Shapiro
GI
,
D'Andrea
AD
. 
Homologous recombination deficiency: exploiting the fundamental vulnerability of ovarian cancer
.
Cancer Discov
2015
;
5
:
1137
54
.
8.
Hoeijmakers
JH
. 
Genome maintenance mechanisms for preventing cancer
.
Nature
2001
;
411
:
366
74
.
9.
Kaelin
WG
. 
The concept of synthetic lethality in the context of anticancer therapy
.
Nat Rev Cancer
2005
;
5
:
689
98
.
10.
O'Connor
MJ
,
Martin
NMB
,
Smith
GCM
. 
Targeted cancer therapies based on the inhibition of DNA strand break repair
.
Oncogene
2007
;
26
:
7816
24
.
11.
Helleday
T
,
Petermann
E
,
Lundin
C
,
Hodgson
B
,
Sharma
RA
. 
DNA repair pathways as targets for cancer therapy
.
Nat Rev Cancer
2008
;
8
:
193
204
.
12.
Moeller
BJ
,
Pasqualini
R
,
Arap
W
. 
Targeting cancer-specific synthetic lethality in double-strand DNA break repair
.
Cell Cycle
2009
;
8
:
1872
6
.
13.
Ledermann
JA
,
Harter
P
,
Gourley
C
,
Friedlander
M
,
Vergote
I
,
Rustin
G
, et al
Overall survival in patients with platinum-sensitive recurrent serous ovarian cancer receiving olaparib maintenance monotherapy: an updated analysis from a randomised, placebo-controlled, double-blind, phase 2 trial
.
Lancet Oncol
2016
;
17
:
1579
89
.
14.
Mirza
MR
,
Monk
BJ
,
Herrstedt
J
,
Oza
AM
,
Mahner
S
,
Redondo
A
, et al
Niraparib maintenance therapy in platinum-sensitive, recurrent ovarian cancer
.
N Engl J Med
2016
;
375
:
2154
64
.
15.
Coleman
RL
,
Oza
AM
,
Lorusso
D
,
Aghajanian
C
,
Oaknin
A
,
Dean
A
, et al
Rucaparib maintenance treatment for recurrent ovarian carcinoma after response to platinum therapy (ARIEL3): a randomised, double-blind, placebo-controlled, phase 3 trial
.
Lancet
2017
;
390
:
1949
61
.
16.
O'Donnell
RL
,
Kaufmann
A
,
Woodhouse
L
,
McCormick
A
,
Cross
PA
,
Edmondson
RJ
, et al
Advanced ovarian cancer displays functional intratumor heterogeneity that correlates to ex vivo drug sensitivity
.
Int J Gynecol Cancer
2016
;
26
:
1004
11
.
17.
Kraus
C
,
Hoyer
J
,
Vasileiou
G
,
Wunderle
M
,
Lux
MP
,
Fasching
PA
, et al
Gene panel sequencing in familial breast/ovarian cancer patients identifies multiple novel mutations also in genes others than BRCA1/2
.
Int J cancer
2017
;
140
:
95
102
.
18.
Walsh
T
,
Casadei
S
,
Lee
MK
,
Pennil
CC
,
Nord
AS
,
Thornton
AM
, et al
Mutations in 12 genes for inherited ovarian, fallopian tube, and peritoneal carcinoma identified by massively parallel sequencing
.
Proc Natl Acad Sci U S A
2011
;
108
:
18032
7
.
19.
Castéra
L
,
Krieger
S
,
Rousselin
A
,
Legros
A
,
Baumann
JJ
,
Bruet
O
, et al
Next-generation sequencing for the diagnosis of hereditary breast and ovarian cancer using genomic capture targeting multiple candidate genes
.
Eur J Hum Genet
2014
;
22
:
1305
13
.
20.
Birkbak
NJ
,
Wang
ZC
,
Kim
J-Y
,
Eklund
AC
,
Li
Q
,
Tian
R
, et al
Telomeric allelic imbalance indicates defective dna repair and sensitivity to DNA-damaging agents
.
Cancer Discov
2012
;
2
:
366
75
.
21.
Popova
T
,
Manie
E
,
Rieunier
G
,
Caux-Moncoutier
V
,
Tirapo
C
,
Dubois
T
, et al
Ploidy and large-scale genomic instability consistently identify basal-like breast carcinomas with BRCA1/2 inactivation
.
Cancer Res
2012
;
72
:
5454
62
.
22.
Wang
ZC
,
Birkbak
NJ
,
Culhane
AC
,
Drapkin
R
,
Fatima
A
,
Tian
R
, et al
Profiles of genomic instability in high-grade serous ovarian cancer predict treatment outcome
.
Clin Cancer Res
2012
;
18
:
5806
15
.
23.
Mutter
RW
,
Riaz
N
,
Ng
CKY
,
Delsite
R
,
Piscuoglio
S
,
Edelweiss
M
, et al
Bi-allelic alterations in DNA repair genes underpin homologous recombination DNA repair defects in breast cancer
.
J Pathol
2017
;
242
:
165
77
.
24.
Christie
EL
,
Fereday
S
,
Doig
K
,
Pattnaik
S
,
Dawson
SJ
,
Bowtell
DDL
. 
Reversion of BRCA1/2 germline mutations detected in circulating tumor DNA from patients with high-grade serous ovarian cancer
.
J Clin Oncol
2017
;
35
:
1274
80
.
25.
Weigelt
B
,
Comino-Méndez
I
,
De Bruijn
I
,
Tian
L
,
Meisel
JL
,
García-Murillas
I
, et al
Diverse BRCA1 and BRCA2 reversion mutations in circulating cell-free DNA of therapy-resistant breast or ovarian cancer
.
Clin Cancer Res
2017
;
23
:
6708
20
.
26.
Graeser
M
,
McCarthy
A
,
Lord
CJ
,
Savage
K
,
Hills
M
,
Salter
J
, et al
A marker of homologous recombination predicts pathologic complete response to neoadjuvant chemotherapy in primary breast cancer
.
Clin Cancer Res
2010
;
16
:
6159
68
.
27.
Mukhopadhyay
A
,
Elattar
A
,
Cerbinskaite
A
,
Wilkinson
SJ
,
Drew
Y
,
Kyle
S
, et al
Development of a functional assay for homologous recombination status in primary cultures of epithelial ovarian tumor and correlation with sensitivity to poly(ADP-Ribose) polymerase inhibitors
.
Clin Cancer Res
2010
;
16
:
2344
51
.
28.
Mukhopadhyay
A
,
Plummer
ER
,
Elattar
A
,
Soohoo
S
,
Uzir
B
,
Quinn
JE
, et al
Clinicopathological features of homologous recombination-deficient epithelial ovarian cancers: sensitivity to PARP inhibitors, platinum, and survival
.
Cancer Res
2012
;
72
:
5675
82
.
29.
Willers
H
,
Taghian
AG
,
Luo
C-M
,
Treszezamsky
A
,
Sgroi
DC
,
Powell
SN
. 
Utility of DNA repair protein foci for the detection of putative BRCA1 pathway defects in breast cancer biopsies
.
Mol Cancer Res
2009
;
7
:
1304
9
.
30.
Naipal
KAT
,
Verkaik
NS
,
Ameziane
N
,
van Deurzen
CH
,
Ter Brugge
P
,
Meijers
M
, et al
Functional ex vivo assay to select homologous recombination-deficient breast tumors for PARP inhibitor treatment
.
Clin Cancer Res
2014
;
20
:
4816
26
.
31.
Tumiati
M
,
Munne
PM
,
Edgren
H
,
Eldfors
S
,
Hemmes
A
,
Kuznetsov
SG
. 
Rad51c- and Trp53-double-mutant mouse model reveals common features of homologous recombination-deficient breast cancers
.
Oncogene
2016
;
35
:
4601
10
.
32.
Chu
P
,
Wu
E
,
Weiss
LM
. 
Cytokeratin 7 and cytokeratin 20 expression in epithelial neoplasms: a survey of 435 cases
.
Mod Pathol
2000
;
13
:
962
72
.
33.
Zhang
S
,
Balch
C
,
Chan
MW
,
Lai
HC
,
Matei
D
,
Schilder
JM
, et al
Identification and characterization of ovarian cancer-initiating cells from primary human tumors
.
Cancer Res
2008
;
68
:
4311
20
.
34.
San Filippo
J
,
Sung
P
,
Klein
H
. 
Mechanism of eukaryotic homologous recombination
.
Annu Rev Biochem
2008
;
77
:
229
57
.
35.
Deans
AJ
,
West
SC
. 
DNA interstrand crosslink repair and cancer
.
Nat Rev Cancer
2011
;
11
:
467
80
.
36.
Stingele
J
,
Bellelli
R
,
Boulton
SJ
. 
Mechanisms of DNA–protein crosslink repair
.
Nat Rev Mol Cell Biol
2017
;
18
:
563
73
.
37.
Stordal
B
,
Timms
K
,
Farrelly
A
,
Gallagher
D
,
Busschots
S
,
Renaud
M
, et al
BRCA1/2 mutation analysis in 41 ovarian cell lines reveals only one functionally deleterious BRCA1 mutation
.
Mol Oncol
2013
;
7
:
567
79
.
38.
Rustin
GJS
,
Vergote
I
,
Eisenhauer
E
,
Pujade-Lauraine
E
,
Quinn
M
,
Thigpen
T
, et al
Definitions for response and progression in ovarian cancer clinical trials incorporating RECIST 1.1 and CA 125 agreed by the gynecological cancer intergroup (GCIG)
.
Int J Gynecol Cancer
2011
;
21
:
419
23
.
39.
Cooke
SL
,
Brenton
JD
. 
Evolution of platinum resistance in high-grade serous ovarian cancer
.
Lancet Oncol
2011
;
12
:
1169
74
.
40.
Pennington
KP
,
Walsh
T
,
Harrell
MI
,
Lee
MK
,
Pennil
CC
,
Rendi
MH
, et al
Germline and somatic mutations in homologous recombination genes predict platinum response and survival in ovarian, fallopian tube, and peritoneal carcinomas
.
Clin Cancer Res
2014
;
20
:
764
75
.
41.
Kang
J
,
D'Andrea
AD
,
Kozono
D
. 
A DNA repair pathway—focused score for prediction of outcomes in ovarian cancer treated with platinum-based chemotherapy
.
JNCI J Natl Cancer Inst
2012
;
104
:
670
81
.
42.
Blagden
SP
. 
Harnessing pandemonium: the clinical implications of tumor heterogeneity in ovarian cancer
.
Front Oncol
2015
;
5
:
149
.
43.
Schwarz
RF
,
Ng
CKY
,
Cooke
SL
,
Newman
S
,
Temple
J
,
Piskorz
AM
, et al
Spatial and temporal heterogeneity in high-grade serous ovarian cancer: a phylogenetic analysis
.
PLOS Med
2015
;
12
:
e1001789
.
44.
Lord
CJ
,
Ashworth
A
. 
BRCAness revisited
.
Nat Rev Cancer
2016
;
16
:
110
20
.
45.
Telli
M
,
Timms
K
,
Reid
J
,
Hennessy
B
,
Mills
GB
,
Jensen
KC
, et al
Homologous recombination deficiency (HRD) score predicts response to platinum-containing neoadjuvant chemotherapy in patients with triple-negative breast cancer
.
Clin Cancer Res
2016
;
22
:
3764
73
.
46.
Abkevich
V
,
Timms
KM
,
Hennessy
BT
,
Potter
J
,
Carey
MS
,
Meyer
LA
, et al
Patterns of genomic loss of heterozygosity predict homologous recombination repair defects in epithelial ovarian cancer
.
Br J Cancer
2012
;
107
:
1776
82
.
47.
Timms
KM
,
Abkevich
V
,
Hughes
E
,
Chris Neff
C
,
Reid
J
,
Morris
B
, et al
Association of BRCA1/2 defects with genomic scores predictive of DNA damage repair deficiency among breast cancer subtypes
.
Breast Cancer Res
2014
;
16
:
475
.
48.
Alexandrov
LB
,
Nik-Zainal
S
,
Wedge
DC
,
Aparicio
SA
,
Behjati
S
,
Biankin
AV
, et al
Signatures of mutational processes in human cancer
.
Nature
2013
;
500
:
415
21
.
49.
Alexandrov
LB
,
Nik-Zainal
S
,
Wedge
DC
,
Campbell
PJ
,
Stratton
MR
. 
Deciphering signatures of mutational processes operative in human cancer
.
Cell Rep
2013
;
3
:
246
59
.
50.
Nik-Zainal
S
,
Alexandrov
LB
,
Wedge
DC
,
Van Loo
P
,
Greenman
CD
,
Raine
K
, et al
Mutational processes molding the genomes of 21 breast cancers
.
Cell
2012
;
149
:
979
93
.
51.
Swisher
EM
,
Lin
KK
,
Oza
AM
,
Scott
CL
,
Giordano
H
,
Sun
J
, et al
Rucaparib in relapsed, platinum-sensitive high-grade ovarian carcinoma (ARIEL2 Part 1): an international, multicentre, open-label, phase 2 trial
.
Lancet Oncol
2017
;
18
:
75
87
.
52.
Jiménez-Sánchez
A
,
Memon
D
,
Pourpe
S
,
Veeraraghavan
H
,
Li
Y
,
Vargas
HA
, et al
Heterogeneous tumor-immune microenvironments among differentially growing metastases in an ovarian cancer patient
.
Cell
2017
;
170
:
927
38
.
e20
.
53.
Patch
AM
,
Christie
EL
,
Etemadmoghadam
D
,
Garsed
DW
,
George
J
,
Fereday
S
, et al
Whole-genome characterization of chemoresistant ovarian cancer
.
Nature
2015
;
521
:
489
94
.