Abstract

Propagation of centromere identity during cell cycle progression in higher eukaryotes depends critically on the faithful incorporation of a centromere-specific histone H3 variant encoded by CENPA in humans and cid in Drosophila. Cenp-A/Cid is required for the recruitment of Cenp-C, another conserved centromere protein. With yeast three-hybrid experiments, we demonstrate that the essential Drosophila centromere protein Cal1 can link Cenp-A/Cid and Cenp-C. Cenp-A/Cid and Cenp-C interact with the N- and C-terminal domains of Cal1, respectively. These Cal1 domains are sufficient for centromere localization and function, but only when linked together. Using quantitative in vivo imaging to determine protein copy numbers at centromeres and kinetochores, we demonstrate that centromeric Cal1 levels are far lower than those of Cenp-A/Cid, Cenp-C and other conserved kinetochore components, which scale well with the number of kinetochore microtubules when comparing Drosophila with budding yeast. Rather than providing a stoichiometric link within the mitotic kinetochore, Cal1 limits centromeric deposition of Cenp-A/Cid and Cenp-C during exit from mitosis. We demonstrate that the low amount of endogenous Cal1 prevents centromere expansion and mitotic kinetochore failure when Cenp-A/Cid and Cenp-C are present in excess.