Our results provide evidence that the scaling relationship between the total centromere and genome size, initially observed in grasses [6, 11], is universal for all Eukaryotes (Fig. 2). Although the estimates of total centromere sizes using the immunofluorescence method are likely to be overestimated, this overestimation is consistent across the species analyzed (Fig. 2), and more importantly, the data obtained by immunofluorescence show the same relationship between the total centromere size and genome size as the data obtained by ChIP-seq (Fig. 2). Thus, we can conclude that immunofluorescence measurement of the total centromere size is a reliable method for the type of comparative analysis used in this study.
The independence of the relationship between the total centromere size and genome size on phylogenetic relatedness (see Results) means that it remains the same whether we look at closely or distantly related species, implying that Eukaryotes share a mechanism maintaining a stable proportion of functional centromere to genome size. The potential mechanism also appears the same for taxa with monocentric, metapolycentric, or holocentric chromosomes because the relationship between total centromere size and genome size does not change with chromosome type (Fig. 2), and the proportion of genome occupied by centromeres is stable within each chromosome type (Fig. 3). The mechanism responsible for the strong dependence of total centromere size on genome size may stem from intracellular scaling principles [10] that maintain the size ratio of intracellular components to ensure their proper function [14], perhaps via regulation of the amount of available CenH3, directly or indirectly through chaperones or licensing factors. The larger centromere proportions in metapolycentrics and holocentrics (Fig. 3) could imply a higher concentration of available CenH3 in these organisms. Zhang and Dawe [6] hypothesized that a species’ genome size determines its total centromere area required to stabilize the spindle. They also surmised that the total centromere area is equally distributed to individual chromosomes [6]. Individual centromeres of uniform sizes could contribute to proper congression and segregation because chromosomes with too large or too small functional centromeres tend to missegregate and get lost [15–19]. This would also mean that, within a karyotype, functional centromere size does not vary with chromosome size, a notion that has been supported by showing that small chromosomes of maize introduced into oat equalized their centromere sizes with large chromosomes of oat [9]. However, reports from studies on human [16, 20, 21], fescue hybrid [11, 22], Arabidopsis [23], and recently from maize [10] suggest that a moderate within-karyotype correlation between the size of chromosomes and their centromeres may exist.
The reason for such a within-karyotype correlation is unclear, but it appears that for a chromosome of a specific size, there is a lower limit of kinetochore size reflecting the minimum number of kinetochore microtubules required for proper chromosomal segregation [24–26]. Chromosomes whose kinetochore size falls below this limit are more likely to be lost during repeated rounds of cell division [16, 24–26]. Thus, a sufficiently significant increase in chromosome size could require a corresponding increase in kinetochore size, and/or an increase in kinetochore size could allow an increase in chromosome size. Changes in the size of individual kinetochores could occur either by drift or as a result of deterministic processes such as centromeric or holokinetic drive [27, 28]. It is, therefore, possible there are two antagonistic processes affecting centromere size. The first equalizes the size of individual centromeres (and possibly even entire chromosomes) to ensure proper chromosome behavior during cell division on a cellular level. By contrast, the second process operates on the level of individual chromosomes and may cause centromere and chromosome size divergence within a single karyotype. As the correlation between the total centromere size and genome size (Fig. 2 here; [6, 11]) is much stronger than the within-karyotype correlation between sizes of individual chromosomes and their centromeres [10, 16], it seems likely that the mechanism keeping centromeres of similar sizes prevails.