A Genomic-Profile Derived Summary Measure of Chromosomal Breakpoints Predicts Response to Treatment with the DNA-Damaging Agent Cisplatin.

Abstract

Abstract Cisplatin-based chemotherapy promotes DNA inter- and intra-strand cross-links and is effective treatment in multiple cancer types, yet it is rarely used for first-line therapy of breast cancer. Cisplatin is an effective cytotoxic in BRCA1 mutant cancer cell lines and in breast and ovarian cancers with BRCA1 mutations, prompting speculation that defective DNA double-strand break repair is associated with cisplatin sensitivity. Furthermore, tumors with BRCA1 mutation are presumed to be defective in various aspects of DNA repair and also display increased levels of chromosomal instability. Thus, we hypothesized that the total number of chromosomal breakpoints in a specific tumor may reflect defective DNA repair and/or chromosomal instability, and may therefore be used as a predictor of cisplatin sensitivity. Supporting this relationship, we identified a correlation (r = 0.8, P = 0.08) between the total number of chromosomal breakpoints, estimated by the number of genomic regions showing allelic imbalance, and cisplatin sensitivity in five triple negative breast cancer cell lines. To validate this relationship in a clinical setting we compared the total number of chromosomal breakpoints to therapy response in a cisplatin treated breast cancer cohort. A total of 28 women with stage II or III ER-/PR-/HER2- breast cancer were treated with cisplatin in the neo-adjuvant setting followed by surgery and standard adjuvant chemotherapy. A core biopsy was obtained before treatment commenced, tumor cells were enriched by needle microdissection, and DNA was extracted for genotyping. As only limited amounts of DNA were available, we used a Molecular Inversion Probe assay (in collaboration with Affymetrix), which allowed us to genotype 42,000 SNPs with only 40 ng of starting DNA. We then estimated the total number of putative genomic breakpoints in 21 samples containing at least 75% tumor cells. We compared this summary genomic measure to the response rate as quantitated by the Miller-Payne score determined on the pathological specimen obtained at surgery. This genomic measure showed a remarkably accurate separation of patients by degree of response, correctly classifying 20 out of 21 patients (P < 0.001).These results suggest that the total number of DNA breakpoints may be an accurate biomarker of breast cancers which may be sensitive to cisplatin treatment. The molecular explanations for these findings await further work. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 111.