RAD51 Foci as a Biomarker Predictive of Platinum Chemotherapy Response in Ovarian Cancer

Abstract

Purpose: To determine the ability of RAD51 foci to predict platinum chemotherapy response in high-grade serous ovarian cancer (HGSOC) patient-derived samples.

Experimental design: RAD51 and γH2AX nuclear foci were evaluated by immunofluorescence in HGSOC patient-derived cell lines (n = 5), organoids (n = 11), and formalin-fixed, paraffin-embedded tumor samples (discovery n = 31, validation n = 148). Samples were defined as RAD51-High if >10% of geminin-positive cells had ≥5 RAD51 foci. Associations between RAD51 scores, platinum chemotherapy response, and survival were evaluated.

Results: RAD51 scores correlated with in vitro response to platinum chemotherapy in established and primary ovarian cancer cell lines (Pearson r = 0.96, P = 0.01). Organoids from platinum-nonresponsive tumors had significantly higher RAD51 scores than those from platinum-responsive tumors (P < 0.001). In a discovery cohort, RAD51-Low tumors were more likely to have a pathologic complete response (RR, 5.28; P < 0.001) and to be platinum-sensitive (RR, ∞; P = 0.05). The RAD51 score was predictive of chemotherapy response score [AUC, 0.90; 95% confidence interval (CI), 0.78-1.0; P < 0.001). A novel automatic quantification system accurately reflected the manual assay (92%). In a validation cohort, RAD51-Low tumors were more likely to be platinum-sensitive (RR, ∞; P < 0.001) than RAD51-High tumors. Moreover, RAD51-Low status predicted platinum sensitivity with 100% positive predictive value and was associated with better progression-free (HR, 0.53; 95% CI, 0.33-0.85; P < 0.001) and overall survival (HR, 0.43; 95% CI, 0.25-0.75; P = 0.003) than RAD51-High status.

Conclusions: RAD51 foci are a robust marker of platinum chemotherapy response and survival in ovarian cancer. The utility of RAD51 foci as a predictive biomarker for HGSOC should be tested in clinical trials.

Figure 1.

RAD51 score correlates with platinum chemotherapy response in established and patient-derived ovarian cancer cell lines and patient-derived organoids. A, Correlation between RAD51 score (percentage of geminin-positive cells with ≥5 RAD51 foci) and carboplatin sensitivity (IC₅₀) determined by MTS assay in HGSOC cell lines. B, Representative images of geminin, RAD51, and colocalization of geminin/RAD51 at ×10 with ×63 insets in two established ovarian cancer cell lines. All foci were counted in 2 technical replicates with n > 100 geminin-positive cells per experiment; scale bars, 50 μm. C, Correlation between RAD51 score and carboplatin sensitivity (IC₅₀) determined by MTS assay in patient-derived HGSOC cell lines. D, Representative images of geminin, RAD51, and colocalization of geminin/RAD51 at ×10 with ×63 insets in 1 POV cell line. All foci were counted in 2 technical replicates with n > 100 geminin-positive cells per experiment; scale bars, 50 μm. E, Organoid generation from ovarian cancer tumor or ascites and representative brightfield microscopic image of ovarian cancer organoids after 7 days of growth; scale bars, 100 μm. F, RAD51 scores in platinum-responsive and platinum-nonresponsive patient-derived ovarian cancer organoids before and after irradiation. Error bars indicate ± SD. *, P < 0.05 and ****, P < 0.0001 by Student two-tailed t test. All foci were counted in 2 technical replicates with n > 100 geminin-positive cells per experiment. G, Representative images of geminin, RAD51, and colocalization of geminin/RAD51 at ×10 with ×63 insets; scale bars, 10 μm.

Figure 2.

Validation of RAD51 immunofluorescence assay in formalin-fixed paraffin-embedded (FFPE) samples. A, RAD51 immunofluorescence assay in high-grade serous ovarian cancer (HGSOC) FFPE tumor samples. B, RAD51 scores in 2 HR-proficient HGSOC cell lines after transfection with 2 separate short-interfering RNA (siRNA) targeting RAD51 (siRAD51) or a noncoding region (siControl) and exposed to γ-irradiation. Cells were treated, fixed, embedded, and cut into 4-μm sections for evaluation. All foci were counted in 4 technical replicates with n > 100 geminin-positive cells per experiment. C, Western blot of the HGSOC cell lines after transfection with two siRNA (siRAD51–1: 121401; siRAD51–2: s531930) targeting RAD51. D, Dynamic range of RAD51 scores in 2 FFPE HR-proficient HGSOC cell lines after treatment with 0, 5, and 10 Gy of γ-irradiation. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 by the Student two-tailed t test. E, Representative images of geminin, RAD51, and colocalization of geminin/RAD51 at ×10 with ×63 insets in a patient-derived FFPE HGSOC sample. All foci were counted in n > 100 geminin-positive cells per experiment; scale bars, 50 μm. F, Correlation in HGSOC FFPE tumor samples between RAD51 scores obtained from 2 separate samples from the same patient.

Figure 3.

RAD51 score predicts platinum chemotherapy response and survival in ovarian cancer in a discovery cohort. A, RAD51 score and pathologic complete response (pCR) in patients with high-grade serous ovarian cancer before neoadjuvant chemotherapy. Dotted black line indicates manual quantification 10% cutoff value, which delineates RAD51-High and RAD51-Low tumors. All foci were counted in n > 100 geminin-positive cells per experiment. Technical replicates were performed for 30% of samples. B, Proportion of pretreatment RAD51-Low (n = 12) and RAD51-High (n = 19) tumors that had a pCR versus chemotherapy response score of 1–2 (RR, 5.28; 95% CI, 1.8–15.37; P < 0.001). C, Proportion of pretreatment RAD51-Low (n = 12) and RAD51-High (n = 19) tumors that were platinum-sensitive versus resistant (RR ∞, P = 0.05). *, P < 0.05 and ****, P < 0.0001 by the Student two-tailed t test D, Kaplan–Meier curves evaluating progression-free survival (left) and overall survival (right) in patients (n = 31) stratified by RAD51 scores. E, Receiver operating characteristic curve evaluating the predictive performance of RAD51 score and pathologic complete response. F, Correlation between manual quantification and novel automated quantification in patient-derived FFPE tumor samples.

Figure 4.

Automated RAD51 scores predict platinum chemotherapy response and survival in a validation cohort. A, RAD51 score in patients with high-grade serous ovarian cancer before (n = 126) or after (n = 22) 3 cycles of neoadjuvant chemotherapy. Primary tumors were scored when possible (n = 141) and metastatic tumors when primary samples were unavailable (n = 7). Dotted black line indicates automatic quantification 6% cutoff value, which delineates RAD51-High and RAD51-Low tumors. All foci were counted using automated software. B, Proportion of RAD51 RAD51-Low (n = 34) and RAD51-High (n = 114) tumors that were platinum-sensitive versus resistant (RR ∞, P < 0.001). ****, P < 0.0001 by Student two-tailed t test. C, Kaplan–Meier curves evaluating progression-free survival in patients (n = 141) stratified by RAD51 scores. D, Kaplan–Meier curves evaluating overall survival in patients (n = 147) stratified by RAD51 scores.

Figure 5.

RAD51-Low tumors predict platinum sensitivity. A, Forest plot with univariate hazard ratios and 95% confidence intervals (CIs) evaluating association between progression-free survival and RAD51-Low scores in different tumor samples. Analysis performed on the validation cohort. B, Forest plot with univariate hazard ratios and 95% CIs evaluating association between overall survival and RAD51-Low scores in different tumor samples. Analysis performed on the validation cohort. C, Sensitivity, specificity, PPV, and NPV for RAD51-Low tumors in predicting platinum sensitivity in both the discovery (n = 31) and validation cohorts (n = 148).