Inferring patterning states across murine and primate epithelial organoids using cell-surface morphometrics

Tissue morphogenesis is governed to a large extent by latent patterning states, yet it remains unclear whether these states are conserved across species. To address this, we developed a microfluidic bioreactor in which mouse and macaque intestinal organoids experience an identical spatiotemporal gradient of Wnt and BMP signals that drives growth-factor–guided epithelial foraging. Leveraging the richness of high-speed light-sheet movies, we extracted a broad panel of cell-surface morphometrics—junctional tension, apical curvature, membrane ruffling and bioelectric potential maps—and used these features to train a Markov-Switching Linear Regression model. On a single-organoid basis the model isolates a discrete set of patterning states that reliably predicts the latency to initiate crypt-like budding. Although the algorithm is trained solely on budding onset times, it also forecasts ultimate branching complexity, underscoring the developmental relevance of the inferred states. The linkage between state identity and morphogenetic performance is strikingly similar in mouse and macaque organoids. Moreover, each state is characterized by a stereotyped constellation of surface features that partially overlaps between species, emphasizing conserved epithelial “expressions” of internal patterning programs across mammals.