Spatiotemporal Lifecycle Cell Atlas of Rice

Rice (Oryza sativa) is a premier monocot model and global staple crop. Understanding rice's developmental complexity is a central challenge and core objective in plant science. In this study, we constructed a comprehensive spatiotemporal transcriptomic atlas: from seed to seed. By the spatial information-aware deep learning approach, we reconstructed the developmental trajectories of 202 identified cell types and systematically screened lineage-specific and pleiotropic transcriptional regulators. Among these, master gene exhibited remarkable cross-cell-type regulatory functions: it modulates nutrient allocation between the embryo and endosperm in EAS, coordinates carbon/nitrogen (C/N) transport in mesophyll cells, and precisely regulates source-sink dynamics at the multicellular and multi-organ levels, ultimately determining seed size. Based on spatiotemporal atlas, we systematically screened polarity-distributed genes during tissue development, uncovered the dorsoventral polarity differentiation of the endosperm and mapped a global ligand-receptor interaction profile during seed development. Collectively, our study provides a unified, multi-scale research framework for deciphering how multicellular higher plants gradually develop from a single cell into a complete individual through the precise orchestration of gene expression.