Homologous recombination DNA repair deficiency and PARP inhibition activity in primary triple negative breast cancer

Abstract

Triple negative breast cancer (TNBC) encompasses molecularly different subgroups, with a subgroup harboring evidence of defective homologous recombination (HR) DNA repair. Here, within a phase 2 window clinical trial, RIO trial (EudraCT 2014-003319-12), we investigate the activity of PARP inhibitors in 43 patients with untreated TNBC. The primary end point, decreased Ki67, occured in 12% of TNBC. In secondary end point analyses, HR deficiency was identified in 69% of TNBC with the mutational-signature-based HRDetect assay. Cancers with HRDetect mutational signatures of HR deficiency had a functional defect in HR, assessed by impaired RAD51 foci formation on end of treatment biopsy. Following rucaparib treatment there was no association of Ki67 change with HR deficiency. In contrast, early circulating tumor DNA dynamics identified activity of rucaparib, with end of treatment ctDNA levels suppressed by rucaparib in mutation-signature HR-deficient cancers. In ad hoc analysis, rucaparib induced expression of interferon response genes in HR-deficient cancers. The majority of TNBCs have a defect in DNA repair, identifiable by mutational signature analysis, that may be targetable with PARP inhibitors.

Fig. 1. RIO study CONSORT diagram and HRDetect analysis.

a RIO study CONSORT diagram. b Effect of rucaparib on Ki67 expression assessed by immunohistochemistry (IHC). The change in proportion of tumor cells expressing Ki67 between baseline and EOT, in patients that had assessable pairs of baseline and EOT samples. BRCA mutation cancers had no evidence of decreased Ki67. c Effect of rucaparib on cleaved PARP expression assessed by immunohistochemistry, as a marker of apoptosis. The change in proportion of tumor cells expressing cleaved PARP between baseline and EOT, in patients that had assessable pairs of baseline and EOT samples. BRCA mutation cancers had no evidence of increased cleaved PARP expression. Grey bars, BRCA wild type patients; Blue bars, BRCA germline mutant patients. Orange line, >30% but <50% reduction; Red line, >50% reduction.

Fig. 2. HRDetect analysis.

a HRDetect scores were established with whole-genome sequencing in baseline biopsies of 26 patients with untreated primary triple negative breast cancer (locally assessed) entering the RIO window clinical trial. HRDetect positive cancers (HRDetect score >0.7) were enriched for inactivating mutations and promoter methylation of HR genes compared to HRDetect negative cancers (p = 0.0005, Fisher’s exact test). Star, an additional patient with ER-positive breast cancer and BRCA2 germline mutation is shown. 2 stars, locally assessed as TNBC but centrally assessed as PR-positive breast cancer with BRCA1 methylation. Magenta bars, BRCA1/2 germline mutation; Blue bars, BRCA1 methylation; Yellow bars, PALB2 germline mutation; Turquoise bars, RAD51C methylation; Grey bar, None/unknown. b Examples of genome plots for a sample with low (top) and high (bottom) HRDetect scores. The histograms associated with each circos plot show mutation counts for each mutation class: the topmost histogram shows the number of mutations contributing to each substitution signature; the middle histogram represents indel patterns; and the bottom histogram shows the number of rearrangements contributing to each rearrangement signature.

Fig. 3. Biomarkers of homologous recombination (HR) repair deficiency in TNBC.

a RAD51 focus assessment in paired baseline and on-treatment biopsies, p = 0.0016 Wilcoxon test. Inset right, example immunohistochemistry images of two cancers, one deficient and one proficient in RAD51 focus formation (brown), in nuclei stained for geminin (blue). Scale bars show 40 μm. b RAD51 immunohistochemistry score (>20% RAD51 foci proficient), as a functional assessment of HR proficiency, was assessed in end of treatment biopsies after 2 weeks of rucaparib in 25 patients. RAD51 foci deficient cancers are enriched for inactivating mutations and promoter methylation of HR genes compared to RAD51 proficient cancers (p = 0.037 Fisher’s exact test). *ER-positive breast cancer and BRCA2 germline mutation is shown. **Locally assessed TNBC but centrally assessed as PR-positive breast cancer with BRCA1 methylation. Magenta bars, BRCA1/2 germline mutation; Blue bars, BRCA1 methylation; Yellow bars, PALB2 germline mutation; Turquoise bars, RAD51C methylation; Grey bars, None/unknown. c Association between RAD51 foci proficiency and HRDetect scores in 18 patients, p = 0.0146 Mann–Whitney U Test. Line indicates median level.

Fig. 4. ctDNA dynamics reveals activity of rucaparib in primary triple negative homologous recombination repair deficient cancers.

a Change in circulating tumour DNA (ctDNA) copies/ml between baseline and end of treatment (EOT) after two weeks of rucaparib. The relative change of on-treatment ctDNA levels (Circulating tumor DNA ratio, CDR15) with HR pathway defects indicated. Right, example digital PCR ctDNA analysis plots. Magenta lines, BRCA1/2 germline mutation; Blue lines, BRCA1 methylation; Yellow lines, PALB2 germline mutation; Turquoise lines, RAD51C methylation; Grey lines, None/unknown. b Associations of ctDNA change on rucaparib at day 15 (CDR15) with left, BRCA1/2 germline mutations (n = 19 patients), middle, RAD51 focus proficiency (n = 12 patients) and right, HRDetect score (n = 15 patients). Centre line, mean; error bars, standard deviation. p values Mann–Whitney U test.

Fig. 5. Expression analysis of primary triple negative homologous recombination repair deficient cancers.

a Association between basal-like and non-basal-like triple negative subtypes, assessed by PAM50, and HRDetect score. p value Fisher’s exact test. n = 20 paired tumour samples. Purple, HRDetect score >0.7; Grey, HRDetect score <0.7. b Change in gene expression on paired tumor biopsies between baseline and end of treatment on rucaparib. Number of genes with a significant change in gene expression (Log fold change >0.5 and false discovery corrected q value < 0.1). Left, categorised by ctDNA suppression or not (CDR15 response <0.25 vs ≥1, n = 8 and n = 3 paired tumour samples respectively). Right, by CDR15 response and HRD score (CDR15 < 0.25 and HRD > 0.7 vs CDR15 ≥ 1 and HRD < 0.7, n = 11 and n = 6 paired tumour samples respectively). c Gene expression changes of PARP1, MKI67, CDKN1A and TMEM173 through treatment, from DESeq2 with false discovery rate (FDR) corrected q value for change. n = 20 paired tumour samples. Magenta lines, BRCA1/2 germline mutation; Blue lines, BRCA1 methylation; Yellow lines, PALB2 germline mutation; Turquoise lines, RAD51C methylation; Grey lines, None/unknown. d Left, Gene set enrichment pathway analysis (GSEA) for gene expression changes through treatment in patients with ctDNA suppression (CDR15 < 0.25, n = 8 paired tumour samples.) on rucaparib. Centre, suppression of G2M checkpoint genes on PARP inhibition, q = 0.006, and right, increased expression of interferon pathway genes on PARP inhibition, q = 0.001. False discovery rate corrected q value for change.