Developing liquid biopsy tests for malignant effusions using artificial intelligence-assisted, morphology-based isolation of tumor cells

PROJECT SUMMARY/ABSTRACT Background. Malignant effusions (ME) are frequent complications of metastatic breast cancer (MBC) associated with severe symptoms and dire prognosis. Treating MEs involves palliation through the serial removal of excess fluids, which are then typically discarded. Instead, these fluids could be used as substrates for liquid

biopsy (LB) to guide the treatment of MEs and advanced MBCs. Problem. Drug targets and predictors of response for MBC are tremendous unmet clinical needs. Procuring solid metastatic tissue can be challenging due to the inaccessibility of disease sites and the risks associated with tissue collection. A major impetus for this proposed research is the opportunity for LB to circumvent these

limitations. ME circulating tumor cells (ME-CTCs) can serve as surrogates for metastatic tissue for molecular characterization. However, the low proportion of METCs relative to immune cells in many MEs complicates profiling efforts. Solution. We have collaborated with Deepcell (DC), a company that developed an artificial intelligence (AI)-

assisted, morphology-based approach to isolate ME-CTCs. DC’s biomarker-agnostic platform provides an advantage over traditional biomarker-based tumor enrichment methods by creating morphological atlases of ME- CTCs for mining novel biomarkers of treatment response and resistance. Our pilot studies demonstrate the

feasibility of molecular characterization of ME-CTCs isolated using the DC platform. Hypothesis. We hypothesize that isolating ME-CTCs using the DC platform and downstream profiling can facilitate the development of LB tools for evaluating known actionable breast cancer (BCa) biomarkers (e.g., ER/PR/HER2, PIK3CA & ESR1 mutations) and discovering new predictive molecular and morphology-based

biomarkers and drug targets. Specific Aims. In Aim 1, we will first validate the DC platform using primary cells from MEs and ME-derived organoids. Next, we will use the validated platform to isolate ME-CTCs, generate copy number and mutation profiles of these cells and matched archival tumors, compare the status of genes frequently mutated in BCa

(e.g., PIK3CA and ESR1), and detect ME-CTC-specific aberrations. In Aim 2, we will perform single-cell RNA sequencing and immunocytochemistry of isolated ME-CTCs and ME-derived organoids to discover expression- based biomarkers and assess the status of known BCa biomarkers (e.g., ER/PR/HER2). In Aim 3, we will perform

correlative analyses between treatment response vs. ME-CTC morphology and molecular signatures (Aims 1 & 2) and use ME-derived organoids for validation studies. Translational impact. Developing a platform for isolating tumor cells from MEs and liquid biopsy tools to discover novel response biomarkers and drug targets can transform the treatment of MEs from a palliative setting

to a therapeutic opportunity to improve the outcomes of patients who develop these devastating complications.