Quantitative imaging of RAD51 expression as a marker of platinum resistance in ovarian cancer

Abstract

Early relapse after platinum chemotherapy in epithelial ovarian cancer (EOC) portends poor survival. A-priori identification of platinum resistance is therefore crucial to improve on standard first-line carboplatin-paclitaxel treatment. The DNA repair pathway homologous recombination (HR) repairs platinum-induced damage, and the HR recombinase RAD51 is overexpressed in cancer. We therefore designed a REMARK-compliant study of pre-treatment RAD51 expression in EOC, using fluorescent quantitative immunohistochemistry (qIHC) to overcome challenges in quantitation of protein expression in situ. In a discovery cohort (n = 284), RAD51-High tumours had shorter progression-free and overall survival compared to RAD51-Low cases in univariate and multivariate analyses. The association of RAD51 with relapse/survival was validated in a carboplatin monotherapy SCOTROC4 clinical trial cohort (n = 264) and was predominantly noted in HR-proficient cancers (Myriad HRDscore < 42). Interestingly, overexpression of RAD51 modified expression of immune-regulatory pathways in vitro, while RAD51-High tumours showed exclusion of cytotoxic T cells in situ. Our findings highlight RAD51 expression as a determinant of platinum resistance and suggest possible roles for therapy to overcome immune exclusion in RAD51-High EOC. The qIHC approach is generalizable to other proteins with a continuum instead of discrete/bimodal expression.

Keywords: HRD; RAD51; immune exclusion; multiplexed IHC; ovarian cancer.

Figure EV1. Validation of a RAD51‐specific antibody

1. Immunoblot of RAD51 in HEY‐T30 control and knock‐down cell lines. The EPR4030(3) antibody reveals a single band of 37 kDa size in control cells (top band results from post‐translational modification of RAD51), which is not present in RAD51 knock‐down cells. Blot is a representation of three independent experiments.

2. Cells from (A) were used to create a FFPE cell block used for RAD51 fluorescent IHC (left). Cells from (A) in an FFPE block stained with an IgG isotype control (right). Scale bar is 50μm. All images are representative of three independent experiments.

3. Fluorescent IHC FFPE block of an ovarian PDX treated ex vivo with γ‐radiation and stained for RAD51 and p‐H2AX S139 (left). Scale bar is 50 μm. RAD51_NES score for seven ovarian PDXs before and after treatment with γ‐radiation (IR) (right). Paired t‐test.

4. RAD51 fluorescent IHC on normal FFPE tissues. Testis is shown as a positive control and liver a negative control. Scale bar is 50 μm. All images are representative of three independent biological samples.

Figure EV2. Correlation of Ki67 % extent and HRD phenotype with RAD51_NES

1. Correlation of Ki67 % extent and RAD51_NES in the BCC cohort. Spearman correlation (left) and one‐way ANOVA with Bonferroni correction (right). Median with interquartile range.

2. Kaplan–Meier plots for PFS (left) and OS (right) stratified according to Ki67 extent quartile in the BCC cohort. Q—quartile. Log‐rank test, shading denotes 95% confidence intervals.

3. Correlation of Ki67 extent and RAD51_NES in the SCOTROC4 cohort. Spearman correlation (left) and one‐way ANOVA with Bonferroni correction (right). Median with interquartile range.

4. Kaplan–Meier plots for PFS (left) and OS (right) stratified according to Ki67 extent quartile in the SCOTROC4 cohort. Q—quartile. Log‐rank test, shading denotes 95% confidence intervals.

5. Correlation of RAD51_NES with BRCA mutation status in EOC. One‐way ANOVA. Median with interquartile range.

6. Linear regression of “genomic scar” HRD score assay and RAD51_NES. Vertical dashed line denotes HRD positivity score of 42.

Figure EV3. Exogenous Flag‐RAD51 is functional

1. Immunoblot of RAD51 upon overexpression and subsequent RNAi‐mediated depletion of total (siRAD51‐CDS) or endogenous (siRAD51‐3’UTR) RAD51 mRNA. The exogenous stably overexpressed Flag‐RAD51 represents the top band. HGSOC cell utilized indicated below the blot. Blots are representative of two independent experiments.

2. Validation of exogenous Flag‐RAD51 functionality using a cell viability assay. Flag‐RAD51 was stably overexpressed in three HGSOC cell lines which were treated with increasing doses of carboplatin for 96 h. Flag‐RAD51 rescues carboplatin sensitivity upon depletion of endogenous RAD51 protein. Mean with standard deviation is shown of at least three biological replicates per point. Statistical comparison is performed at 1μM concentration of carboplatin, t‐test.

3. Immunofluorescence of TYK‐nu cells with stable Flag‐RAD51 overexpression treated with 10μM of carboplatin for 48 h. Cells were co‐stained for both RAD51 and Flag. DAPI serves as a nuclear counterstaining. Scale bar is 20 μm.

Figure EV4. Differential gene expression analysis of immune genes between RAD51‐High and ‐Low tumours

Differential gene expression analysis of immune genes between RAD51‐High (Q4) and ‐Low (Q1) across three independent mRNA cohorts of EOC (see also Fig 3E). t‐Test; dashed line denoted threshold of significance, Bonferroni corrected for multiple testing.

Figure EV5. Multiplexed fluorescent IHC staining for immune markers of the tumour microenvironment

1. Multiplexed fluorescent IHC staining for immune markers of the microenvironment in an EOC patient sample. Unmixed monochrome components are shown along with a false‐coloured merge image. Cytokeratin staining was used to differentiate between the tumour and stromal compartments of the sample. Scale bar is 50 μm.

2. Quantitation of immune populations in the BCC cohort. Results for RAD51‐High and ‐Low tumours are shown. T/S—tumour/stroma ratio. Bar is median. Mann–Whitney test.

3. Subset analysis of CD8⁺ cytotoxic T‐cell infiltration in the BCC cohort stratified according to BRCA mutation status. Absolute tumour CD8⁺ cytotoxic T‐cell infiltration numbers and tumour/stroma (T/S) cytotoxic T‐cell number ratio in RAD51‐High and ‐Low cases. Bar is median. Mann–Whitney test.

Figure 1. RAD51 assay development and testing on BCC cohort

1. Range of example RAD51 nuclear expression score (RAD51_NES) values with respective unmixed monochrome fluorescent IHC staining images. EpCAM is used as a tumour marker and an internal sample quality control. Scale bar is 50 μm.

2. Correlation of RAD51_NES with two independent pathologist H‐scores (top left and top right) and correlation of two pathologist with each other (bottom). Pearson correlation.

3. Distribution of RAD51 nuclear expression score (RAD51_NES) in the BCC cohort. The cohort is divided into RAD51‐Low expressing cases (first quartile, Q1—blue), intermediate cases (interquartile range, IQR—grey) and RAD51‐High expressing cases (fourth quartile, Q4—brown). Dashed line denotes the median RAD51_NES in this cohort.

4. Survival analysis of the BCC cohort. Kaplan–Meier plots for progression‐free survival (PFS) (left) and overall survival (OS) (right) stratified according to fourth quartile (Q4) and first quartile (Q1) of RAD51_NES. Log‐rank test, shading denotes 95% confidence intervals.

5. Number of cases with progression at 12 and 24 months. Chi‐square test.

Figure 2. Validation of assay and findings in SCOTROC4 cohort

1. Distribution of RAD51_NES in the SCOTROC4 cohort. Dashed line denotes the median RAD51_NES in this cohort.

2. Survival analysis of the SCOTROC4 cohort. Kaplan–Meier plots for PFS (left) and OS (right) stratified according to quartiles of RAD51_NES. Log‐rank test, shading denotes 95% confidence intervals.

3. Number of cases with progression at 12 and 24 months. Chi‐square test.

4. Survival analysis of HRD‐positive patients according to quartile of RAD51_NES. Log‐rank test, shading denotes 95% confidence intervals.

5. Survival analysis of HRD‐negative patients. Log‐rank test, shading denotes 95% confidence intervals.

Figure 3. RAD51 does not promote platinum resistance in vitro but modulates immune‐related gene expression

1. In vitro cell survival assay of HGSOC cell lines upon stable RAD51 overexpression. Mean with standard deviation is shown of at least three biological replicates per point. Extra‐sum‐of‐squares F‐test.

2. In vitro colony‐forming assays comparing RAD51‐overexpressing and control HGSOC cell line, Caov‐3, after treatment with increasing doses of carboplatin. Mean and SD of four biological replicates (left) a representative experiment (right). P‐value for a comparison between cell lines for each drug treatment condition is indicated above the bars. t‐Test.

3. Gene Set Enrichment Analysis of RAD51‐overexpression vs. control HGSOC cell lines (n = 4). The top ten enriched pathways in RAD51‐overexpression cell lines are listed on the right, ranked by the lowest false‐discovery rate (FDR). The corresponding Enrichment Score (ES) is shown. See Table EV1 for details on all data points. Colouring denotes the top two enriched pathways. GO—Gene Ontology.

4. Gene Set Enrichment Analysis of RAD51‐High vs. RAD51‐Low HGSOC tumours from a TCGA cohort. The top ten enriched pathways in RAD51‐High tumours are listed on the right, ranked by the lowest false‐discovery rate (FDR). See Table EV3 for details on all data points. Colouring denotes genesets related to immune response pathways. ES—Enrichment Score.

5. Volcano plot for fold changes of immune genes enriched or depleted in RAD51‐High tumours vs. RAD51‐Low tumours from TCGA. t‐Test; dashed line denoted threshold of significance, Bonferroni corrected for multiple testing.

6. Immune genes enriched in RAD51‐High and RAD51‐Low tumours across four EOC mRNA cohorts (TCGA, AOCS, MGH, Duke).

7. Example unmixed fluorescent IHC images demonstrating the presence of CD8⁺ T cells in the tumour (T) and stroma (S) in a RAD51‐High and a RAD51‐Low tumour from the BCC cohort. Scale bar is 50 μm.

8. CD8⁺ cytotoxic T‐cell (T_cyt) infiltration analysis in the BCC cohort. Absolute tumour cytotoxic T‐cell infiltration numbers and tumour/stroma (T/S) cytotoxic T‐cell number ratio in RAD51‐High and ‐Low cases (left). Bar is median. Mann–Whitney test.