High-content imaging workflow for single-cell drug response profiling of acute myeloid leukemia subtypes

Acute myeloid leukemia (AML) exhibits pronounced cellular heterogeneity, which contributes to variable therapeutic responses and limits the predictive power of conventional functional assays. While bulk viability measurements provide aggregated readouts, they fail to resolve phenotypic diversity and dynamic cellular states within heterogeneous populations. Here, we established a high-content fluorescence imaging workflow for quantitative drug response profiling (DRP) in AML at single-cell resolution. The assay integrates three non-toxic fluorescent dyes to capture features of nuclear morphology, mitochondrial function, and apoptosis. Automated high-content imaging combined with computational image analysis enables robust segmentation and extraction of phenotypic features across thousands of individual cells. Using a supervised machine learning approach, cells were classified into viable, apoptotic and dead states, enabling quantitative assessment of drug responses through population-normalized metrics. This approach allows direct integration of image-based data into downstream analysis workflows, facilitating the generation of functional dose-response curves and the determination of IC₅₀ values and drug sensitivity scores (DSS) at single-cell resolution. The workflow was validated across seven AML cell lines, including models of acquired and mutation-driven resistance to BCL-2 inhibition. Image-based DRPs generated for venetoclax (VEN) showed strong concordance with established bulk measurements obtained using the ATP-based cell viability readout (CellTiterGlo®, CTG). Furthermore, screening of a 16-compound panel representing diverse mechanisms of action demonstrated robust agreement between image- and CTG-based drug response profiles, while providing additional phenotypic information at single-cell resolution. Finally, the workflow was successfully transferred to primary AML samples. A pilot 31-compound drug screen identified BCL-2 inhibitors as the most active compounds, consistent with the patient's molecular profile. Together, this workflow establishes a scalable functional phenomics platform for high-resolution drug profiling and phenotypic stratification in AML. The integration of single-cell imaging with AI-based analysis provides a promising foundation for future functional precision oncology approaches, with potential applications in patient-specific DRP and combination therapy optimization.