Branching development of early post-implantation human embryonic-like tissues in 3D stem cell culture

Human embryonic stem cells (hESCs) have the intrinsic capacity to self-organize and generate patterned tissues. In vitro models that coax hESCs to form embryonic-like structures by modulating physical environments and priming with chemical signals have become a powerful tool for dissecting the regulatory mechanisms underlying early human development. Here we present a 3D suspension culture system of hESCs that can generate post-implantation, pre-gastrulation embryonic-like tissues in an efficient and controllable manner. The efficiency of the development of asymmetric tissues, which mimic the post-implantation, pre-gastrulation amniotic sac, was about 50% in the 3D suspension culture. Quantitative imaging profiling and unsupervised trajectory analysis revealed that hESC aggregates first entered into a transitional stage expressing Brachyury (or T), before their development branched into different paths to develop into asymmetric embryonic-like tissues, amniotic-like tissues, and mesodermal-like tissues, respectively. Moreover, the branching developmental trajectory of embryonic-like structures was affected by the initial cell seeding density or cluster size of hESCs. A higher percentage of amniotic-like tissues was observed under a small initial cell seeding density of hESCs. Conversely, a large initial cell seeding density of hESCs promoted the development of mesodermal-like tissues. Intermediate cell seeding densities of hESCs in the 3D suspension culture promoted the development of asymmetric embryonic-like tissues. Our results suggest that hESCs have the intrinsic capability to sense the initial cell population size, which in turn regulates their differentiation and self-organization into different embryonic-like tissues. Our 3D suspension culture thus provides a promising experimental tool to study the interplay between tissue topology and self-organization and progressive embryonic development using in vitro hESC-based models.