High-quality Gossypium hirsutum and Gossypium barbadense genome assemblies reveal the landscape and evolution of centromeres

Centromere positioning and organization are crucial for genome evolution, however the research on centromere biology is largely influenced by the quality of genome assembly. Here, we combine Oxford Nanopore and PacBio technologies to de novo assemble two high-quality reference genomes for Gossypium hirsutum (TM-1) and Gossypium barbadense (3-79). Compared to previously published reference genomes, our assemblies showed substantial improvements, with the contig N50 improved by 4.6-fold and 5.6-fold respectively, representing the most complete cotton genomes to date. High-quality reference genomes enabled us to characterize 14 and 5 complete centromeric regions for G. hirsutum and G. barbadense, respectively. Our data revealed that the centromeres of allotetraploid cotton were occupied by CRM and Tekay long terminal repeat families, and the CRM family reshaped the structure of the At subgenomic centromeres after polyploidization. These two intertwined families drove the convergent evolution of centromeres between the two subgenomes, ensuring centromere function and genome stability. In addition, the centromere repositioning and high divergence of sequences between G. hirsutum and G. barbadense contribute to speciation and centromere diversity. This study sheds light on centromere evolution in a significant crop and provides an alternative approach to exploring the evolution of polyploid plants.