Multiparametric phenotyping of compound effects on patient derived organoids

Abstract

Patient derived organoids resemble the biology of tissues and tumors, enabling ex vivo modeling of human diseases from primary patient samples. Organoids can be used as models for drug discovery and are being explored to guide clinical decision making. Patient derived organoids can have heterogeneous morphologies with unclear biological causes and relationship to treatment response. Here, we used high-throughput, image-based profiling to quantify phenotypes of over 5 million individual colorectal cancer organoids after treatment with more than 500 small molecules. Integration of data using a joint multi-omics modelling framework identified organoid size and cystic vs. solid organoid architecture as axes of morphological variation across organoids. Mechanistically, we found that organoid size was linked to IGF1 receptor signaling, while a cystic organoid architecture was associated with an LGR5+ stemness program. Treatment-induced organoid morphology reflected organoid viability, drug mechanism of action, and was biologically interpretable using joint modelling. Inhibition of MEK led to cystic reorganization of organoids and increased expression of LGR5, while inhibition of mTOR induced IGF1 receptor signaling. In conclusion, we identified shared axes of variation for colorectal cancer organoid morphology, their underlying biological mechanisms, and pharmacological interventions with the ability to move organoids along them. Image-based profiling of patient derived organoids coupled with multi-omics integration facilitates drug discovery by linking drug responses with underlying biological mechanisms.