A Game of Thrones at Human Centromeres I. Multifarious structure necessitates a new molecular/evolutionary model

Human centromeres form over arrays of tandemly repeated DNA that are exceptionally complex (repeats of repeats) and long (spanning up to 8 Mbp). They also have an exceptionally rapid rate of evolution. The generally accepted model for the expansion/contraction, homogenization and evolution of human centromeric repeat arrays is a generic model for the evolution of satellite DNA that is based on unequal crossing over between sister chromatids. This selectively neutral model predicts that the sequences of centromeric repeat units will be effectively random and lack functional constraint. Here I used shotgun PacBio SMRT reads from a homozygous human fetal genome (female) to determine and compare the consensus sequences (and levels of intra-array variation) for the active centromeric repeats of all the chromosomes. To include the Y chromosome using the same technology, I used the same type of reads from a diploid male. I found many different forms and levels of conserved structure that are not predicted by -and sometimes contradictory to- the unequal crossing over model. Much of this structure is based on spatial organization of three types of ~170 bp monomeric repeat units that are predicted to influence centromere strength (i.e., the level of outer kinetochore proteins): one with a protein-binding sequence at its 5 end (a 17 bp b-box that binds CENP-B), a second that is identical to the first except that the b-box is mutated so that it no longer binds CENP-B, and a third lacking a b-box but containing a 19 bp conserved "n-box" sequence near its 5 end. The frequency and organization of these monomer types change markedly as the number of monomers per repeat unit increases, and also differs between inactive and active arrays. Active arrays are also much longer than flanking, inactive arrays, and far longer than required for cellular functioning. The diverse forms of structure motivate a new hypothesis for the lifecycle of human centromeric sequences. These multifarious levels of structures, and other lines of evidence, collectively indicate that a new model is needed to explain the form, function, expansion/contraction, homogenization and rapid evolution of centromeric sequences.