Discovery of small molecules that correct gene networks dysregulated in human disease may allow identification of therapies that treat disease at its fundamental basis by leveraging mechanism-based data. Here, we report the first broad gene network-based drug screen, which led to discovery of a drug candidate that effectively treats aortic valve disease in an animal model. We previously reported haploinsufficiency of NOTCH1 (N1) as a genetic cause of human aortic valve thickening and calcification, the third most common form of heart disease, and described the resulting gene network dysregulation in human induced pluripotent stem cell (iPSC)-derived endothelial cells (ECs). We exposed isogenic N1+/+ or N1+/– iPSC-derived ECs to each of 1595 small molecules or control and developed a machine learning approach that accurately distinguished WT or N1-haploinsufficient cells based on expression of 119 genes assayed by targeted RNA-sequencing. 9 small molecules corrected the gene network of N1+/– ECs sufficiently to be classified as WT. Among hits tested in vivo, the estrogen receptor-related alpha inverse agonist XCT790 significantly reduced aortic valve thickening, calcification, and stenosis in N1-haploinsufficient mice with shortened telomeres, which model the range of age-dependent cardiac disease observed in humans. This strategy, made feasible by human iPSC technology, next generation sequencing approaches, and machine learning, may represent a more effective path for drug discovery compared to conventional screening approaches.
Overall design: RNA-seq on WT vs N1-haploinsufficient iPSC-derived ECs exposed to DMSO or one of 9 network-correcting small molecules (Fmoc-leu, RO4929097, Cytochalasin, GSK837149A, CB1954, Biperiden, TG003, Putrescine, XCT790)
Less...