Introduction: While most patients with high grade serous ovarian cancer (HGSC) respond to platinum-based chemotherapy, the response is rarely durable and recurrence almost inevitable. A characteristic of HGSC is defective DNA repair. A class of drugs called PARP inhibitors (PARPi) exploit this vulnerability and have proven useful in delaying recurrence. However, eventually tumors respond by making one or more proteins that prevent cancer cells from dying in response to PARPi. One such protein is the anti-apoptosis protein Bcl-XL, for which there is an inhibitor, navitoclax, in clinical trials. If we can identify the patients that become dependent on Bcl-XL, responses can be augmented and resistance to PARPi can be overcome by adding navitoclax. Our hypothesis is that patient derived organoids can be used as a pragmatic way to identify those patients most likely to benefit from this drug combination.

Methods:  To develop an efficient and reliable patient derived organoids models we are adapting conditional reprogramming and combining it with novel hydrogel based synthetic ECM supports. We generate HGSC patient-specific tumour organoids in 384 well plates with greater than 90% success. Organoids are stained with novel non-toxic mix-and-read dyes and fluorescence imaged by confocal microscopy. Chemoresponses to drugs targeting DNA repair and/or anti-apoptotic proteins are inferred from the micrographs using AI algorithms.

Results: PDOs capture the inherent heterogeneity of the disease, albeit local to the sampled site. Methods employing deep learning AI algorithms for automated analyses of 3D confocal image stacks of fixed and cleared immunostained organoids will enable inferring chemoresponses for individual cell types.

Conclusions: Identifying which patients are most likely to respond to the combination of a PARPi with the Bcl-XL inhibitor navitoclax may enable provision of the one-two-punch needed to eliminate their cancer or dramatically prolong response.