L11.04 - Ovarian Avatar Models Predictive of Platinum Response in Ovarian Cancer Patients

Abstract

Ovarian cancer is the most lethal gynecological malignancy. A major contributing factor in the overall mortality of this disease is the failure to identify patient subgroups sensitive to targeted and individualized therapy. While all advanced ovarian, primary peritoneal and fallopian tube cancer patients typically receive platinum/taxane chemotherapy, many will respond suboptimally. Ovarian cancer tissue +/- ascites was collected from patients undergoing primary debulking surgery. Tumor was minced into a slurry and injected +/- ascites intraperitoneally (IP) in 3 SCID mice/patient and were followed for engraftment for up to 1 year. Upon engraftment (founder), tumor tissue was resected, minced and divided for molecular analyses (Affymetrix Human U133 Plus 2.0) and expansion of tumor tissue in additional mice. Mice bearing the expanded xenografts were then treated with weekly paclitaxel (15 mg/kg) and carboplatin (50 mg/kg) for 4 weeks, sacrificed and tumors were harvested. Tumor volume was assessed biweekly during treatment by ultrasound. From March 2010 to December 2012, tumors were implanted from 246 individual patients. After an initial pilot study, the tumor engraftment rate was 74%. Tumor engraftment occurred in varying histological subtypes of ovarian cancer. Tumor engraftment was associated with poor outcome in ovarian cancer patients from which source tumors originated (OR 5.074, 95% CI 1.1-23.3). DNA microarray experiments demonstrated that Avatars clustered in 1 of 3 subgroups defined by the ovarian TCGA dataset: proliferative, dedifferentiated or immunoreactive. Tumors originating from 3 source patients with platinum-sensitive disease, as defined by no evidence of recurrence > 2 yrs, demonstrated responsiveness to paclitaxel/carboplatin chemotherapy with tumor regression. Conversely, tumor from a patient with platinum-resistant disease, as defined by recurrence within 6 months, failed to respond to chemotherapy in vivo. IP-derived, ovarian tumorgraft models may be generated from primary surgical specimens at a high engraftment rate. Failure to engraft had positive prognostic implications for the source patient. Response to platinum-based chemotherapy in ovarian Avatar models appear to reflect the responsiveness of source patients. The ovarian Avatar models may be a useful in identifying subgroups of ovarian cancer patients that would most likely to benefit from targeted and/or novel therapeutics.