Heterogeneous patterning behavior emerges from human stem cell-based embryo model

Undergoing complex self-organization, two-dimensional gastruloids derived from human pluripotent stem cells (hPSCs) are a scalable model system to study developmental processes. However, gastruloid heterogeneity has been obscured by technological limitations. A microraft-based assay was developed to screen, sort, and sequence individual living gastruloids expressing germ layer reporters to quantify the emergence of gastruloid variability. In this proof-of-concept study, distinct patterning trajectories were identified from latent representations using time-series imaging combined with self-supervised deep learning. End point gene expression of isolated single gastruloids was measured to further evaluate heterogeneity and link phenotype to gene expression. The computational analysis captured nuanced technical and biological differences among individual gastruloids across and within batches. Additionally, single gastruloid transcriptomic data revealed clusters by batch and subclusters within experimental batches. Given the small sample size, no detectable association could be inferred from the imaging and sequencing data. Orthogonal measurements will provide deeper insights from each gastruloid; therefore, the preliminary results highlight the need for optimized computational methods and adequately sized training datasets. This platform coupled both imaging and transcriptomic data for comprehensive assays of single gastruloids. Moreover, dissecting the intrinsic heterogeneity of gastruloids derived from the same parental hPSC line provides a powerful lens into the complex variability of stem cell-based embryo models. Importantly, these results call for the careful evaluation of reproducibility in human in vitro systems as applied to human embryogenesis, especially given the current push away from animal models.