Distinct influences of tandem repeats and retrotransposons on CENH3 nucleosome positioning

Abstract

Background: Unique structural characteristics of centromere chromatin enable it to support assembly of the kinetochore and its associated tensions. The histone H3 variant CENH3 (centromeric histone H3) is viewed as the key element of centromere chromatin and its interaction with centromere DNA is epigenetic in that its localization to centromeres is not sequence-dependent.

Results: In order to investigate what influence the DNA sequence exerts on CENH3 chromatin structure, we examined CENH3 nucleosome footprints on maize centromere DNA. We found a predominant average nucleosome spacing pattern of roughly 190-bp intervals, which was also the dominant arrangement for nucleosomes genome-wide. For CENH3-containing nucleosomes, distinct modes of nucleosome positioning were evident within that general spacing constraint. Over arrays of the major ~156-bp centromeric satellite sequence (tandem repeat) CentC, nucleosomes were not positioned in register with CentC monomers but in conformity with a striking ~10-bp periodicity of AA/TT dimers within the sequence. In contrast, nucleosomes on a class of centromeric retrotransposon (CRM2) lacked a detectable AA/TT periodicity but exhibited tightly phased positioning.

Conclusions: These data support a model in which general chromatin factors independent of both DNA sequence and CENH3 enforce roughly uniform centromeric nucleosome spacing while allowing flexibility in the mode in which nucleosomes are positioned. In the case of tandem repeat DNA, the natural bending effects related to AA/TT periodicity produce an energetically-favourable arrangement consistent with conformationally rigid nucleosomes and stable chromatin at centromeres.

Figure 1

CENH3 nucleosome core lengths and AA/TT contents. (A) Number of reads per read length (bp). Genome-matching reads with unambiguous termini favour a length of around 155 bp. 'total genome' includes everything that aligned to the maize genome, version 1. The three centromere elements are shown separately. See also Additional File 1, part B. (B) The frequency of AA or TT dimers at each position along the read lengths. The reads were aligned at their starts and the number of AA or TT dimers counted at each position for each aggregated set. We arbitrarily cut off the analysis at position 200 in order to avoid statistical noise from the low number of reads with lengths greater than 200 bases. In the total genomic reads, the change in AA/TT content visible near position 155 suggests a subtle distinction in nucleotide content between the nucleosome core DNA and linker DNA. The CentC, CRM1, and CRM2 reads were identified by blastall alignments. See also Additional File 1, part C.

Figure 2

Alignments of CENH3-ChIP reads to CRM1 and CRM2. Number of CENH3-ChIP reads aligning to each position on CRM2 (A) and CRM1 (B). Vertical lines are spaced every 190 bp. In the case of CRM2, but not CRM1, alignments show strong tendency to peak every 190 bp, particularly toward its long terminal repeats (LTRs), indicating nucleosome phasing.

Figure 3

CentC alignments and micrococcal nuclease (MNase) sequence preference. (A) The number of alignment edges at each position on CentC from CENH3-ChIP reads. Due to the short and repetitive character of Cent, just the alignment edges are depicted rather than the entire alignments. An apparent preferred position is defined by right edges positioned over the bracketed section (but see B). Alignments to CentC were split into two categories, forward and reverse orientation; then the 5'-most position of the forward alignments were combined with the 3'-most position of the reverse alignments to count the total number of left edge positions along the length of CentC. Likewise, the 5'-most position of the reverse alignments and 3'-most position of the forward alignments were combined to count the total number of right edge positions. Since this analysis uses only information from alignment edges, the alignments were first filtered so as to include only the ones with approximate nucleosome lengths (145 to 175 bp). (B) Sequence preference in MNase digestion of CentC. DNA fragments produced by MNase digest of naked DNA or chromatin were captured by ligation to adapters then amplified with primers that selected for CentC-adapter junctions (corresponding to the right edges shown in A). The bar chart indicates the number of cloned fragments (per 100) whose right edges were within 3 bp of the enriched site found in the CENH3-ChIP 454 dataset (the bracketed section in A, centred on the last base in GGGTGTCGGGGTG). Errors bars are standard errors of the means, calculated based on the sample sizes (74 sequences for 'Naked DNA No.1', 37 for 'Naked DNA No.2', 46 for 'Naked DNA No.3', 46 for 'Naked DNA No.4', 71 for 'Chromatin No.1', 41 for 'Chromatin No.2' and 47 for 'Chromatin No.3'.

Figure 4

Start-to-start distance analysis of nucleosome spacing on CRM1 and CRM2. (A) A schematic description of the start-to-start distance analysis. Three 155-bp nucleosomes (represented by purple circles) starting 190 bp apart give rise to 155-bp reads with alignments of both orientations--forward in blue, reverse in red. Distances between the start of the first forward alignment and start of the first reverse alignment correspond to the length of one nucleosome core. Distances between start of the first forward alignment and subsequent reverse alignments increase by 190 bp, corresponding to an additional nucleosome core and linker region. (B) A short range start-to-start distance analysis for CRM1 and CRM2 nucleosomes. The number of starts separated by each possible distance was quantified in terms of a coincidence index value. We arbitrarily cut off this plot at 650 bp to emphasize the presence of a peak at ~155 bases (as well as two subsequent peaks separated by ~190 bp). Vertical lines are spaced every 20 bp. (C) A long range start-to-start distance analysis for CRM2 CENH3 nucleosomes. Distances between start positions from alignments of opposite orientations were calculated and the number of starts separated by each possible distance between -7577 and +7577 bp quantified in terms of a coincidence index value. The length of the CRM2 reference sequence is 7577 bp, precluding any longer distances. For a random sampling of alignment pairs, the probability is lowest in order for them to be separated by the longest possible distance (the element's length), as there is only one possible combination for either the positive or negative extreme. In contrast, the number of combinations that allow for a distance of zero is almost as large as the length of the element. Hence, the triangular shape of the plot, with a peak centred near zero and tapering off in both directions. Vertical lines are spaced every 190 bp.

Figure 5

A Southern analysis of nucleosome spacing on CRM1, CRM2 and CentC. Southern blots of micrococcal nuclease (MNase) digested CRM1, CRM2 and CentC nucleosome cores. (A) Total chromatin was lightly digested with MNase and isolated DNA fragments corresponding to nucleosome monomers, dimers and higher order polymers were separated by gel electrophoresis and probed for CentC and for CRM1 and CRM2 LTRs. (B) A rare fragment, ~30-bp shorter than the majority of the nucleosome dimers, was often visible, especially for CentC (blue arrow). In some cases a shorter fragment could also be seen in the trimer range.