A spatiotemporal and machine-learning platform facilitates the manufacturing of hPSC-derived esophageal mucosa

Abstract

Human pluripotent stem cell-derived tissue engineering offers great promise for designer cell-based personalized therapeutics, but harnessing such potential requires a deeper understanding of tissue-level interactions. We previously developed a cell replacement manufacturing method for ectoderm-derived skin epithelium. However, it remains challenging to manufacture the endoderm-derived esophageal epithelium despite possessing a similar stratified epithelial structure. Here, we employ single-cell and spatial technologies to generate a spatiotemporal multi-omics cell census for human esophageal development. We identify the cellular diversity, dynamics, and signal communications for the developing esophageal epithelium and stroma. Using Manatee, a machine-learning algorithm, we prioritize the combinations of candidate human developmental signals for in vitro derivation of esophageal basal cells. Functional validation of Manatee predictions leads to a clinically compatible system for manufacturing human esophageal mucosa.