Reconstructing human pancreatic gene networks enhances stem cell-derived β cell induction

Generating functional β cells from stem cells remains a major challenge in regenerative medicine due to the incomplete recapitulation of human pancreatic development in vitro. By integrating newly generated human single-cell RNA sequencing (RNA-seq) datasets (Carnegie stages 10-15) with existing data, we mapped gene co-expression networks (GCNs) underlying pancreatic lineage progression in humans and mice. We observed significant species-specific differences in GCN robustness and dorsal-ventral propensity for progenitor development. Benchmarking three common differentiation protocols against the in vivo datasets showed that they fail to reproduce human-like GCNs, thereby limiting stem cell-derived insulin-secreting β cell (SC-β cell) induction efficiency. To address this, we developed a protocol that reconstructs human pancreatic GCN dynamics, shortens the induction period to 19 days, and achieves up to ∼70% β cell content. SC-islets generated with this method significantly alleviated diabetic symptoms and maintained mature β cell function after transplantation in mice. These findings bridge in vivo mechanisms and in vitro differentiation, advancing stem cell-based therapies.