ABSTRACT: The specificity of action of a pharmaceutical compound to a particular target cell type is desirable to mitigate undesired effects in off target cell types. However, drug screening strategies necessarily employ cell lines that aim to represent but are potentially poor surrogates for the target cell phenotype. In this study, we addressed whether transcriptomic similarity between cell phenotpyes is predictive of their drug response. Thus, we differentiated hiPSC from three individuals into twelve different distinct phenotypes (cardiomyocytes, smooth muscle cells, hepatocytes, renal proximal tubular cells, podocytes, cortical neurons, dopaminergic neurons, neural progenitor cells, nociceptors, astrocytes, macrophages and endothelial cells). We exposed these cells and the MCF-7 cancer cell line with 12 compounds (Taxol (Paclitaxel) (0.04 µM), 17-AAG (Tanespimycin) (0.2 µM), Bardoxolone methyl (CDDO-Me) (0.2 µM), Rapamycin (Sirolimus) (0.2 µM), Tunicamycin (0.2 µM), Vorinostat (0.2 µM), A-366 (1 µM), A-971432 (1 µM), Compound 81 (1 µM), Rosiglitazone (1 µM), Thiamet-G (1 µM), Lithium chloride (5000 µM)) for 7 hours. Analysis of the RNAseq profiles revealed that the impact of drug treatment varies significantly among different cell types. Despite some compounds showing consistent effects across various cell types, there was no clear correlation between transcriptomic similarity and the magnitude of response to the same drug. These findings underscore that accurate predictions of drug responses necessitate employing the most representative models of the target cell types involved.
