A stepwise route to domesticate rice by controlling seed shattering and panicle shape
Asian rice (Oryza sativa L.) is consumed by more than half of the worlds population. Despite its global importance, the process of early rice domestication remains unclear. During domestication, wild rice (O. rufipogon Griff.) acquired non-seed-shattering behaviour, allowing humans to increase grain yield. Previous studies argued that the sh4 mutation triggered a reduction in seed shattering during rice domestication; but our experiments using wild introgression lines of O. rufipogon show that the domesticated sh4 allele alone is insufficient for shattering loss. Here, we identified the interaction between three key mutations associated with the interruption of abscission layer formation and panicle architecture that were causal in early rice domestication. An interruption of abscission layer formation requires both sh4 and qSH3 mutations, presenting an apparent barrier to the selection of shattering loss. We identified a causal single-nucleotide polymorphism at qSH3 within the seed-shattering gene OsSh1, which is conserved in indica and japonica subspecies but absent in the circum-aus group of rice. Through harvest experiments, we demonstrated that seed shattering alone did not significantly impact yield; rather yield increases were observed with closed panicle formation controlled by SPR3, which is further augmented by the integration of sh4 and qSH3 alleles. Complementary manipulation of seed shattering and panicle shape result in a mechanically stable panicle structure. We propose a stepwise route for the earliest phase of rice domestication, wherein selection of visible SPR3-controlled closed panicle morphology was instrumental in the sequential recruitment of sh4 and qSH3, which led to the loss of shattering. Significance StatementRice is one of the most important crops worldwide. Loss of seed shattering in domesticated rice, previously attributed to single mutations such as those in sh4, is considered the principal genetic change which resulted in yield increases. However, we show that sh4 alone is insufficient and other genes, such as qSH3, are required to cause abscission layer disruption. The evolution of non-seed-shattering therefore required multiple mutations. Furthermore, shattering loss in genetic backgrounds of wild rice does not correspondingly increase yields. We have identified an interaction in which a second trait, closed panicle formation controlled by SPR3, that both increases the yield and facilitates recruitment of sh4 and qSH3, which synergistically augment yield, leading to a stepwise model for rice domestication.