Abstract A056: Patient avatar models for ovarian cancer: Predicting patient response to treatment

Abstract Background: High-grade serous ovarian cancer (HGSOC) remains a leading cause of gynecologic cancer-related mortality, with limited treatment options. Although initial remission is often achieved through surgery and chemotherapy, disease recurrence is common. Existing preclinical models fail to fully recapitulate the complexity of HGSOC and its heterogeneous drug responses. Animal models are gold standard but are fundamentally different than human biology. Among all in vitro models, patient-derived tissue explants represent one of the most advanced models, as they are direct derivatives of the actual tumor obtained through surgical resection or biopsy. However, the inherent challenge of maintaining tissue viability outside the human body for a sufficient duration to enable meaningful scientific investigation limit the use of this model despite its advantages. There is a critical need for patient-derived HGSOC models capable of accurately predicting drug response to support personalized treatment strategies. Methods: Surgically resected tumors were collected intraoperatively from properly consented patients, fragmented into sub-millimeter pieces, and cultured in our perfusion platform. Viability was assessed after 7 days using LIVE/DEAD staining, followed by drug treatment. After 5 days, viability was measured with the 3D CellTiter-Glo assay. Microtumors were fixed and stained for common HGSOC markers (PAX8) and tumor microenvironment (TME) components, then imaged by confocal microscopy and multiplex immunofluorescence (IF). Patient clinical data, including CA-125 levels and progression-free survival, were abstracted from electronic medical records and compared to ex vivo results. Results: We developed a 3D ex vivo platform that enables long-term culture, therapeutic screening, and real-time microscopy under pump-free perfusion conditions. The system uses a microgel-based scaffold supporting gravity-driven nutrient and gas transport, enabling culture of patient-derived tissue microexplants (microtumors) for weeks to months. The multiplex IF showed that patient-derived microtumors retained their native microenvironment, including PAX8+ tumor cells, collagen, vimentin, and immune cell components. Ki67 staining confirmed active proliferation in culture. We used this model as a patient-derived avatar to predict responses to standard-of-care therapy, which correlated closely with clinical outcomes, including CA-125 dynamics, disease-free survival, and overall survival. This approach establishes a platform for therapeutic drug screening with the potential to accelerate prediction of clinical responses. Conclusions: This 3D ex vivo platform offers a robust tool for personalized drug screening, mechanistic studies, and prediction of therapeutic efficacy. It addresses a critical gap in existing ovarian cancer models for determining patient-specific drug responses. Citation Format: Yuxi Zhou, Alfonso Pepe, Ryan A. Smolchek, Jack E. Famiglietti, Athena Tsingelis, Angie Rivera, Diego I. Pedro, Katriana Johnson, Dominick Dag, Christian Young, Matthew A. Genser, Lucia Mazzacurati, Said Cifuentes Maury, Jose L. Serrano-Velez, Atousa Ordobazari, Carlos M. Segura, Naele Lopez-Blanco, Duy T. Nguyen, Gregory Sawyer, Erin George. Patient avatar models for ovarian cancer: Predicting patient response to treatment abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Ovarian Cancer Research; 2025 Sep 19-21; Denver, CO. Philadelphia (PA): AACR; Cancer Res 2025;85 (18Suppl): Abstract nr A056.