Maize centromeric chromatin scales with changes in genome size

Abstract

Centromeres are defined by the location of Centromeric Histone H3 (CENP-A/CENH3) which interacts with DNA to define the locations and sizes of functional centromeres. An analysis of 26 maize genomes including 110 fully assembled centromeric regions revealed positive relationships between centromere size and genome size. These effects are independent of variation in the amounts of the major centromeric satellite sequence CentC. We also backcrossed known centromeres into two different lines with larger genomes and observed consistent increases in functional centromere sizes for multiple centromeres. Although changes in centromere size involve changes in bound CENH3, we could not mimic the effect by overexpressing CENH3 by threefold. Literature from other fields demonstrate that changes in genome size affect protein levels, organelle size and cell size. Our data demonstrate that centromere size is among these scalable features, and that multiple limiting factors together contribute to a stable centromere size equilibrium.

Keywords: CENP-A; cellular scaling; centromere; genome-site; kinetochore.

Figure 1

Centromeric repeats and their relationship to centromere size in the 26 NAM inbred lines. (A) The sizes of 110 fully scaffolded centromeres in each inbred line as measured by CENH3 ChIP-seq. (B) Spearman’s rank correlation analysis between centromere size and the abundance of CentC and CRM. Only 73 centromeres without any gaps (of known or unknown size) were used for this analysis. Correlation coefficients (r) and P-values are indicated in the graph. The data show that CentC is not a primary driver of centromere size in maize, as is the case for alpha satellite in human. CRM shows a positive correlation and is commonly observed in all centromeres.

Figure 2

Correlation of centromere size with genome size and chromosome size. The graphs show data from 110 fully assembled centromeres in the 26 NAM genomes. (A) Spearman’s rank correlation analysis between genome size and centromere size. (B) Spearman’s rank correlation analysis between chromosome size and centromere size. When the > 3 Mb outliers were removed, the correlations were still significant (genome size/centromere size, r = 0.19, P < 0.05; chromosome size/centromere size, r = 0.43, P < 0.0001).

Figure 3

Crossing scheme for generating maize lines with different genome sizes. (A) Crossing schemes for transferring B73 centromeres into the Oaxaca and Z. luxurians backgrounds. (B) Genome sizes of B73, Oaxaca, Z. luxurians and their hybrids. Genome sizes are averages from 3 to 11 plants per family. ⊗ is the self-cross symbol. SD indicates standard deviation.

Figure 4

Centromeres are expanded in B73 X Oaxaca F2 and BC1F2 progeny. (A) CENH3-ChIP profiles of B73, Oaxaca X B73 F2 and Oaxaca X B73 BC1F2 progeny for five centromeres. Centromere size is defined as the length of DNA under the ChIP-seq enrichment curve (see solid bars under the read depth plots). (B) ANOVA analysis of centromere sizes across different lines. Bar graphs show mean centromere size comparison among different lines. Letters represent different groups that are statistically different (P < 0.05). Centromere sizes in the F2 and BC1F2 progeny were significantly larger than in B73, with the sole exception of centromere 9 in the F2 generation. Bars represent SD. (C) Spearman’s rank correlation analyses of centromere sizes across different lines. Dots represent different individuals. Blue: B73, orange/red: F2, dark red: BC1F2. Genome sizes are expressed relative to B73, and are averages based on 3–11 individuals.

Figure 5

Centromeres are expanded in the B73 X Z. luxurians F2 and BC1F2 progeny. Each panel shows ChIP-seq, ANOVA and Spearman’s rank correlation analyses as described in Figure 4. (A) Data for B73 X Z. luxurians F1 and F2 progeny from four centromeres. (B) Data for B73 X Z. luxurians F1, F2, and BC1F2 progeny from three centromeres (BC1F2 data were available for these three centromeres only). Letters represent different groups that are statistically different (P < 0.05). In the F1, the sizes of Cen4, Cen5, and Cen8 were significantly larger than in B73 (and the others were not). In the F2, the sizes of Cen2, Cen3, Cen4, Cen5, and Cen8 were significantly larger than in B73 (while Cen9 and Cen10 were not). The sizes of all centromeres in the BC1F2 were significantly larger than in B73. Blue: B73, purple: F1, orange/red: F2, dark red: BC1F2. Genome sizes are expressed relative to B73, and are averages based on 3–10 individuals.

Figure 6

Centromere size is stable in the CENH3-overexpression lines. (A) Protein blot analysis of maize CENH3 expression levels in roots of wild-type (WT) and CENH3-Ox-1 (Ox) lines. Nuclear protein was diluted to 0.25X, 0.5X, and 1X. The same blot was incubated with antibodies to histone H4 as a loading control. (B) Quantification of CENH3 gene copy number. DNA (gene) copy number was estimated by qPCR, mRNA expression was estimated from RNA-seq, and protein levels interpreted as the relative staining intensity of CENH3 and histone H4 on protein blots. Error bars show standard deviation. WT expression was set to one in each experiment. CENH3 is a single copy gene in WT lines. (C and D) CENH3 ChIP-seq profiles for B73, WT siblings of the Ox lines, and Ox lines for centromeres 4 and 10. The window size is 5 Mb. (E and F) ANOVA analyses of centromere sizes across different lines. There were no significant differences in the sizes of Cen4 and Cen10 between WT and Ox (P < 0.05). The unit of centromere size is Mb. Bars in (B, E, and F) represent SD.