Functional organoid screening uncovers target dependency and bystander killing in TROP2 ADCs

Introduction: TROP2-targeted antibody-drug conjugates (ADCs) represent a promising therapeutic class for multiple tumor types with elevated TROP2 expression. Membrane-permeable payloads can induce bystander killing, making functional, physiologically relevant testing crucial. We developed an advanced organoid-based ADC evaluation workflow combining large-scale patient-derived (xenograft) organoid (PD(X)O) screening, CRISPR-engineered isogenic models, and 3D high-content imaging (HCI) to assess potency, target dependency, and bystander effects. Methods: Two TROP2 ADCs - an experimental exatecan-based ADC linked via a hydrophilic, cleavable linker, and datopotamab-deruxtecan (Dato-Dxd; DS‑1062) - were evaluated in the organoid screening platform, where models were selected irrespective of TROP2 expression. The exatecan-based ADC was tested in 109 PD(X)Os (101 tumor, 8 normal) across ten cancer types using nine-point dose-response. Dato-Dxd was screened using six-point dose-response assays in 142 PD(X)Os (132 tumor; 10 normal) spanning nine tumor types. Drug incubation ran over five days, and efficacy was determined by ATP-based cell viability assay. Free payloads exatecan and Dxd (both naked and linker-conjugated) served as controls. CRISPR/Cas9-generated TROP2 knockout (KO) organoid pairs were labeled with fluorescent/luminescent reporters for target HCI-based dependency and bystander effect assays. Results: The exatecan-based ADC demonstrated high overall efficacy (68.6% showed >50% inhibition at the highest dose of 1 µM), with pancreatic models even showing 100% sensitivity, and 32.4% of organoids achieving IC 50 < 0.1 µM. Payloads alone were strongly cytotoxic in most models (93.3% of models sensitive with IC 50 < 0.03 µM) with similar potency for exatecan and Dxd, whereas linker-conjugated Dxd was significantly less cytotoxic (average IC 50 > 0.1µM). Dato-Dxd responses varied across the PD(X)O panel (54% showed >50% inhibition at 0.1µM, 37% IC 50 < 0.1µM). For both TROP2 ADCs, a correlation between efficacy and TROP2 mRNA abundance was observed. In isogenic KO organoid models, loss of TROP2 abrogated Dato-Dxd responses, confirming target dependency. Mixed-culture assays quantified Dato-Dxd bystander killing, validating membrane-permeable payload action in neighboring cells. Conclusion: The combined organoid screening and HCI workflow provides a robust, translationally relevant platform for ADC evaluation, integrating efficacy profiling, mechanistic validation, and functional bystander effect modeling. CRISPR-generated isogenic organoids enable precise dissection of target dependency and resistance, supporting the preclinical optimization of TROP2-targeted ADCs.