Nuclear DNA-Gated Electrochemiluminescence Microscopy with Deep Learning for Single-Cell Apoptosis Profiling

Accurately assessing how individual cells respond to anticancer agents remains challenging because most assays provide bulk or binary readouts and cannot translate single-cell imaging into scalable, mechanism-relevant quantification. Here, we develop a nuclear DNA-gated electrochemiluminescence (ECL) microscopy assay in which genomic DNA acts as an intrinsic coreactant for Ru(bpy)₃²⁺, converting apoptosis-associated chromatin condensation and fragmentation into spatially resolved decreases in nuclear ECL intensity. The resulting signal provides a dose- and time-dependent measure of the apoptotic progression. For drug-treated cells, a deep learning model trained on ECL images enables high-throughput classification of apoptotic and nonapoptotic cells, allowing quantitative apoptosis profiling across large datasets. Using this framework, we compare the apoptotic potency of three natural anticancer compounds and reveal distinct spatial damage signatures for two photosensitizers, reflecting their different ROS-mediated pathways. This integrated approach offers a quantitative and scalable strategy for single-cell apoptosis profiling.