The DNA damage response in advanced ovarian cancer: functional analysis combined with machine learning identifies signatures that correlate with chemotherapy sensitivity and patient outcome.

BACKGROUND: Ovarian cancers are hallmarked by chromosomal instability. New therapies deliver improved patient outcomes in relevant phenotypes, however therapy resistance and poor long-term survival signal requirements for better patient preselection. An impaired DNA damage response (DDR) is a major chemosensitivity determinant. Comprising five pathways, DDR redundancy is complex and rarely studied alongside chemoresistance influence from mitochondrial dysfunction. We developed functional assays to monitor DDR and mitochondrial states and trialled this suite on patient explants. METHODS: We profiled DDR and mitochondrial signatures in cultures from 16 primary-setting ovarian cancer patients receiving platinum chemotherapy. Explant signature relationships to patient progression-free (PFS) and overall survival (OS) were assessed by multiple statistical and machine-learning methods. RESULTS: DR dysregulation was wide-ranging. Defective HR (HRD) and NHEJ were near-mutually exclusive. HRD patients (44%) had increased SSB abrogation. HR competence was associated with perturbed mitochondria (78% vs 57% HRD) while every relapse patient harboured dysfunctional mitochondria. DDR signatures classified explant platinum cytotoxicity and mitochondrial dysregulation. Importantly, explant signatures classified patient PFS and OS. CONCLUSIONS: Whilst individual pathway scores are mechanistically insufficient to describe resistance, holistic DDR and mitochondrial states accurately predict patient survival. Our assay suite demonstrates promise for translational chemosensitivity prediction.

The use of ovarian cancer cells from patients undergoing surgery to generate primary cultures capable of undergoing functional analysis.

The use of cell lines or animal models has significant disadvantages when dealing with a set of heterogeneous diseases such as epithelial ovarian cancer. This has clinical relevance in that biomarkers developed using cell line or animal models are often not transferable to the clinical setting. In this study, we describe the development of a robust protocol for developing primary cultures of ovarian cancer which will overcome some of these difficulties. Women undergoing surgery for ovarian cancer were recruited and samples of ascites and solid tumour deposits were used to develop primary cultures. Cells were characterised using a panel of immunofluorescent antibodies prior to use in a variety of assays including functional assessment of DNA repair pathways. During the four year study period, viable cultures, confirmed to be epithelial in origin were generated from 156 of 172 (91%) cases recruited. Characterisation was carried out using a panel of antibodies including pancytokeratin, CA125, EpCAM, MOC-31, D2-40 and vimentin. Senescence occurred between the 2nd and 8th passages in all cultures except one in which spontaneous immortalization occurred. Cells could be successfully cultured even after a period of storage at 4°C and cultured cells were capable of being used for a variety of applications including functional assays. Upon functional assessment there was minimal intra-tumour heterogeneity. It is therefore possible to derive viable ovarian cancer cell cultures in the majority of patients undergoing surgery. Cells cultured directly from patient cancers provide an accurate and highly diverse model.