The CentO satellite confers translational and rotational phasing on cenH3 nucleosomes in rice centromeres

Significance

Centromeres are sites on chromosomes that mediate attachment to microtubules for chromosome segregation and often comprise tandemly repeated “satellite” sequences. The function of these repeats is unclear because centromeres can be formed on single-copy DNA by the presence of nucleosomes containing a centromere-specific variant of histone H3 (cenH3). Rice has centromeres composed of both the 155-bp CentO satellite repeat and single-copy non-CentO sequences. This study shows that rice cenH3 nucleosomes are regularly spaced with 155-bp periodicity on CentO repeats, but not on non-CentO sequences. CentO repeats have an ∼10-bp periodicity in dinucleotide pattern and in nuclease cleavage that suggests that CentO has evolved to minimize its bending energy on cenH3 nucleosomes and that centromeric satellites evolve for stabilization of cenH3 nucleosomes.

Abstract

Plant and animal centromeres comprise megabases of highly repeated satellite sequences, yet centromere function can be specified epigenetically on single-copy DNA by the presence of nucleosomes containing a centromere-specific variant of histone H3 (cenH3). We determined the positions of cenH3 nucleosomes in rice (Oryza sativa), which has centromeres composed of both the 155-bp CentO satellite repeat and single-copy non-CentO sequences. We find that cenH3 nucleosomes protect 90–100 bp of DNA from micrococcal nuclease digestion, sufficient for only a single wrap of DNA around the cenH3 nucleosome core. cenH3 nucleosomes are translationally phased with 155-bp periodicity on CentO repeats, but not on non-CentO sequences. CentO repeats have an ∼10-bp periodicity in WW dinucleotides and in micrococcal nuclease cleavage, providing evidence for rotational phasing of cenH3 nucleosomes on CentO and suggesting that satellites evolve for translational and rotational stabilization of centromeric nucleosomes.

Centromeres, the chromosomal domains that attach to spindle microtubules to segregate eukaryotic chromosomes in mitosis and meiosis, are DNA elements bound by special nucleosomes that contain a centromere-specific variant of histone H3 (cenH3). In most plants and animals, cenH3 nucleosomes are found on centromeric DNA that comprises megabases of tandemly repeated “satellite” sequences. Despite this apparent preference for repetitive DNA, a fully functional centromere, called a neocentromere, can occasionally form by assembling cenH3 nucleosomes on a single-copy DNA sequence that was not previously part of a centromere, indicating that centromere specification is epigenetic in plants and animals (for reviews, see refs. 1–4).

The tandem arrays of highly repeated satellite sequences that compose most plant and animal centromeres can differ dramatically between closely related species (5), and even between different chromosomes (6–8), suggesting that satellite arrays undergo rapid evolution through expansions, contractions, gene conversions, and transpositions. Monomers of satellite repeats range in length from 5 bp in Drosophila to 1,419 bp in cattle although more than half of described monomers in 282 species have lengths between 100 and 200 bp, often regarded as approximately the length of nucleosomal DNA (6, 9). The cenH3 nucleosomes typically occupy only a portion of the satellite repeats, often in discontinuous blocks (7, 10–12), and the same or similar repeats often underlie flanking pericentromeric heterochromatin composed of conventional nucleosomes. Some of these repeats, for example African green monkey α-satellite DNA, have long been known to position conventional nucleosomes, resulting in arrays of regularly spaced nucleosomes, said to be translationally phased (13–15). Nucleosomes can occupy multiple alternative translational phases on the same satellite (16, 17). Translationally phased nucleosomal arrays have also been observed on satellites in cucumber and in several cereal species, where phasing varies among repeats and chromosomal regions (18, 19).

Recently deep-sequencing technology has been applied to centromeres treated with micrococcal nuclease (MNase), which preferentially digests linker DNA between nucleosomes, to determine the positioning of cenH3 nucleosomes on satellite repeats. In human cultured cells, substantial translational phasing of CENP-A, the human cenH3, was reported on α-satellite (20). In maize, a similar approach mapped CENH3 (the name used for plant cenH3s) on the 156-bp maize centromeric satellite CentC and on two retrotransposon-derived centromeric sequences, CRM1 and CRM2 (21). Evidence for translational phasing of CENH3 on CentC and CRM1 was lacking, but 190-bp phasing was observed on CRM2. CentC was shown to have a strong periodicity of AA or TT dinucleotides about every 10 bp, which corresponds to one turn of the DNA double helix. This periodicity is thought to favor a particular orientation of the DNA toward the nucleosome core particle, based on DNA bendability, and is known as rotational phasing of nucleosomes (22–24).

Rice has centromeres characterized by the 155-bp satellite sequence CentO, which is related to maize CentC (25, 26). Although some rice centromeres have megabases of CentO satellites, other evolutionarily new centromeres have little CentO, so CENH3 nucleosomes are found on both CentO and non-CentO sequences (12). For example, Cen8 is comprised of mostly non-CentO sequences and has a CentO array (CentO_8) that is spanned by a sequenced BAC (27). Centromeres like Cen8 are thought to represent an intermediate stage in centromere evolution between rare neocentromeres that form on unique sequences and mature centromeres populated by megabase-sized arrays of satellites (7, 12). Cen8 therefore presents an opportunity to compare the organization of CENH3 nucleosomes on CentO and non-CentO sequences. To that end, we used an antibody to rice CENH3 (27) to perform chromatin immunoprecipitation (ChIP) of CENH3 nucleosomes digested with MNase and sequenced the bound DNA (ChIP-Seq) to determine the positions of CENH3 nucleosomes on rice centromeres. We analyzed the sizes and positions of CENH3 nucleosomal DNA fragments on both CentO and non-CentO sequences to address the role of satellites in organizing centromeric chromatin and analyzed the sequence features of these fragments to look for evidence of nucleosome positioning signals.

Results

CENH3 Nucleosomes Are Well-Positioned on Cen8.

To investigate the positions of CENH3 nucleosomes in rice centromeres, we performed anti-CENH3 ChIP using chromatin well-digested by MNase (4.0 Units) to ∼90% mononucleosome size to eliminate most linker DNA and gel-purified a mononucleosome band (∼80–230 bp). The immunoprecipitated DNA was sequenced using the Illumina Genome Analyzer II platform. We generated 39.6 million (M) 36-bp paired-end reads, of which 28.2 M mapped uniquely to the rice reference genome (TIGR7/IRGSP1). The distribution of fragment lengths (Fig. 1A) had a major peak at 127 bp, with four minor peaks ranging from 93 to 148 bp. For comparative analysis, we also generated 35.1 M 36-bp sequence reads from an MNase-digested rice genomic DNA library, and 273 M paired-end reads from an MNase-digested rice bulk mononucleosomal (input) DNA library, in which canonical H3-containing nucleosomes predominate. All sequencing data are available from GEO (accession no. GSE50755). The distribution of fragment lengths in the input mononucleosome library had a major peak at 147 bp and a minor peak at ∼134 bp (Fig. 1A). The relative distribution of fragment lengths for CENH3 nucleosomes and input mononucleosomes resembles that recently seen for native human cenH3 (CENP-A) nucleosomes and bulk nucleosomes (20) and indicates that cenH3 nucleosomes protect less DNA on average than canonical nucleosomes.

Fig. 1.

CENH3 nucleosome positioning in rice Cen8. (A) Distribution of fragment lengths in size-selected 4.0 U MNase-treated mononucleosome libraries. The green line illustrates the distribution of all of the ∼1.4 M anti-CenH3 ChIP-seq fragments mapped to CENH3 subdomains in 12 rice centromeres. The blue line illustrates the distribution of 1.4 M randomly selected fragments from the input mononucleosomal DNA sequencing library. The red line illustrates the distribution of all mappable fragments (∼241 M) from the mononucleosomal DNA sequencing library. The x axis represents the length of fragments in base pairs. The y axis represents counts for each length. (B) Distribution of ChIP-seq fragments in Cen8. Vertical blue lines represent numbers of fragments within each 1-kb window. The horizontal green bars mark the CENH3-enriched subdomains. Two regions, CentO_8 and a Gypsy /DIRS1 transposon (pink boxes), are enlarged to illustrate CENH3 nucleosome positioning. NucleR scores of the two expanded regions are shown at the bottom.

To identify the CENH3-binding domains in rice centromeres, we split each chromosome into 1-kb windows and plotted the log₂ fold change between the anti-CENH3 ChIP-seq fragment count and the input mononucleosomal DNA fragment count within each window. The ChIP-seq fragment distribution patterns in the four best-sequenced rice centromeres (Cen4, Cen5, Cen7, and Cen8) were similar to those obtained previously using an anti-CENH3 ChIP-454 sequencing dataset (12). We observed many distinct peaks of mononucleosome size in several of the most CENH3-enriched subdomains in Cen8 (Fig. 1B). One of these regions (12,964,000–12,967,000) contains the CentO_8 satellite block and is highly enriched with CENH3 whereas another (13,430,000–13,434,000) contains a Gypsy/DIRS1 transposon and is moderately enriched with CENH3. ChIP-seq fragment distribution at 1-bp resolution revealed well-positioned nucleosomal peaks identified by nucleR (28) in both regions, although with somewhat variable average spacing per well-positioned nucleosome (187 bp for region 12,964,000–12,967,000 or 154 bp for subregion 12,964,000–12,966,000; versus 307 bp for region 13,430,000–13,434,000, which has more poorly positioned nucleosomes). Similarly, well-positioned nucleosome peaks were observed throughout Cen4, Cen5, and Cen7 (Fig. S1) using both fragment count and nucleR score, indicating that most CENH3 nucleosomes are well-positioned.

The CENH3 Nucleosome Cores Are Smaller than Canonical Nucleosome Cores.

Fragments from MNase-digested chromatin can be used to map the boundary/cutting sites of individual nucleosomes. The MNase-resistant nucleosomal core can be visualized through counts of MNase cutting sites from the (+) and (−) strand-specific reads (29). Using this procedure, we aligned the nucleosome centers identified by nucleR from all well-positioned canonical nucleosomes from the input mononucleosomal DNA library and from CENH3 nucleosomes. We found that the distance between the (+) and (−) strand peaks was 147 bp for canonical nucleosomes, but the distance was ∼103–129 bp for CENH3 nucleosomes, depending on which of several closely-spaced peaks was measured (Fig. 2A). Because of the preferential digestion of AT-rich DNA by MNase (30), we used MNase-digested rice genomic DNA to calibrate the mapping of the ChIP-seq dataset. The result after calibration gave a single major peak for both (+) and (−) strands and an average distance of ∼112 bp (Fig. 2B).

Fig. 2.

Core sizes of CENH3 and canonical nucleosomes. (A) The nucleosome cores are anchored by mapping MNase cutting sites from (+) and (−) strands, (red and blue, respectively). The x axis represents the distance from the center. The y axis represents the count of MNase cutting sites. (B) CENH3 nucleosomes were mapped by using strand-specific ChIP-seq reads normalized by reads from MNase-digested rice genomic DNA. The x axis represents the distance from the center. The y axis represents the ratio of read count between ChIP-seq and MNase-digested rice genomic DNA. (C) Agarose gel electrophoresis of DNAs isolated from rice chromatin digested with different amounts of MNase. Chromatin samples digested with 0.5 U and 2.5 U MNase were used to make underdigested CENH3 ChIP-seq libraries. (D–F) V-plots of ChIP-seq fragments from libraries prepared using chromatin digested with (D) 0.5 U MNase, (E) 2.5 U MNase, or (F) 4.0 U MNase followed by size selection (∼80–230 bp). A 300-bp region spanning aligned CENH3 nucleosome centers from each library is shown. The heat maps were generated by quantifying the number of fragments of a given length (y axis) and given distance from the midpoint of fragments to centers (x axis). Color represents quantified fragment scores from the low (black) to high (pink).

The MNase cleavage peaks used to measure CENH3 nucleosomes are sensitive to the extent of digestion. To reduce the frequency of nucleosome-internal cleavages, which increases with MNase digestion, we performed two additional CENH3 ChIP-seq experiments using rice chromatin lightly digested with 0.5 and 2.5 Units of MNase, respectively. Application of 0.5 U of MNase resulted in underdigestion relative to typical ChIP-ready chromatin of mononucleosome size (Fig. 2C). Four libraries were constructed and paired-end sequenced, two from ChIPed and two from input DNA samples (Fig. S2A). Nearly identical patterns of positioning and spacing of the CENH3 nucleosomes were observed using the ChIP-seq data from the three independent ChIP experiments (Fig. S2B).We again observed that the average distance between the (+) and (−) strand peaks for CENH3 nucleosomes was shorter than that for canonical nucleosomes (Fig. S2C).

To confirm and better define the size of rice CENH3 nucleosomes, we made V-plots (31, 32) using ChIP-seq fragments associated with all well-positioned CENH3 nucleosomes aligned at their centers (Fig. 2 D–F). V-plots are based on paired-end read datasets generated by a modified Illumina library preparation protocol that permits efficient recovery of DNA fragments as small as ∼25 bp. On the V-plot, the x axis represents the distance from the fragment midpoint to the center of the CENH3 nucleosome, and the y axis represents the fragment length. The value on the y axis corresponding to the vertex of the V-plot represents the size of the CENH3 nucleosome core particle. The two V-plots using 0.5 U and 2.5 U of MNase both indicated that the protected region of CENH3 nucleosomes is ∼100 bp (Fig. 2 D and E) whereas the V-plot using 4.0 U of MNase had a vertex at ∼90 bp (Fig. 2F). This smaller size might indicate that the DNA on the CENH3 nucleosome has been more precisely trimmed back or that internal cleavage of CENH3 nucleosomes has reduced the size of the protected DNA in the 4.0-U MNase digestion. In either case, the CENH3 nucleosome appears to wrap no more than 100 bp of DNA, sufficient for only a single wrap around the CENH3 nucleosome.

Intrinsic DNA Sequence Features Associated with CENH3 Nucleosomes.

DNA sequences are intrinsically important for nucleosome positioning. Dinucleotides SS (G/C) and WW (A/T) were reported to be signals for nucleosome positions (22–24, 33). To examine whether rice CENH3 nucleosomes exhibit similar intrinsic DNA sequence features, we aligned the centers from 4,255 well-positioned CENH3 nucleosomes from the 4.0-U MNase digestion and analyzed the distribution of SS and WW dinucleotides within ±150 bp from the center. The center and its immediately flanking regions were enriched with SS dinucleotides. In contrast, WW dinucleotides were enriched in regions ±75 bp from the center (Fig. 3 A and B).

Fig. 3.

Sequence features of CENH3 nucleosomal DNA. (A) Centers of positioned CENH3 and canonical nucleosomes are aligned, and sequence signals are displayed within ±150 bp from center. Each curve represents WW or SS dinucleotide frequency distributed in either CENH3 or canonical nucleosomes. The x axis represents the distance from the center; the y axis represents the frequency of SS/WW dinucleotides. (B) Heat map of eight different types of SS/WW dinucleotides ±150 bp from the CENH3 nucleosome center. (C) SS/WW dinculceotide frequency associated with CentO-related CENH3 nucleosomes, including 71 centers in CentO_8 (94,000–106,000 bp in BAC a0038J12). The x axis represents the distance from the center; the y axis represents the frequency of SS/WW dinucleotides.

For comparison, we analyzed 265,487 well-positioned canonical nucleosomes from the 4.0-U library and observed a highly similar dinucleotide distribution pattern (Fig. 3A). However, the enriched WW dinucleotides were detected in regions ±91 bp away from the center, 16 bp further in each direction than for the CENH3 nucleosomes, consistent with our observations that CENH3 nucleosomes are smaller than conventional nucleosomes. These results agree with observations in humans that G/C-enriched regions favor nucleosome centering whereas flanking regions enriched for AA or TT dinucleotides possibly act as repelling elements to restrict the translational positioning of nucleosomes (34).

To investigate whether CENH3 nucleosomes on CentO also possess similar intrinsic DNA sequence features, we used ClustalW 2.0 (35) to align all CentO or CentO-like sequences of length ∼155 bp (available at ftp://ftp.plantbiology.msu.edu/pub/data/TIGR_Plant_Repeats/) and generated a 155-bp consensus CentO sequence (SI Text). This sequence was used to identify CentO repeats in the CentO_8 satellite block from Cen8, using BAC a0038J12 (GenBank ID no. AY360388), which spans the CentO_8 block of Cen8 and is fully sequenced (27). We chose to use BAC a0038J12, which differs in sequence assembly from the CentO_8 sequence of the International Rice Genome Sequencing Project, because much of the assembly of BAC a0038J12 has been verified through restriction digests and because the size of CentO_8 on this assembly most closely matches the length of CentO_8 on Chromosome 8 as estimated from fiber-FISH (26, 27, 36). Seventy-one CentO units were identified within the 94,000- to 106,000-bp region of BAC a0038J12 (Fig. 4A).

Fig. 4.

Mapping of ChIP-seq fragments to CentO repeats in CentO_8. (A) Counts of ChIP-seq fragments (y axis) along the sequence of the CentO_8 region (x axis). Locations of CentO repeats are indicated by arrowheads. (B) Enlargement showing one peak of fragment counts per CentO. (C and D) CENH3-nucleosomes on 155-bp and 167-bp CentO variants. (C) Enrichment in proportion of CentO-derived 11-mers between anti-CENH3 ChIPed and input mononucleosome fragments. The x axis represents the center position of 11-mers along the CentO consensus sequence (shown in color: A, green; T, blue; G, red; C, orange). The y axis represents the log₂-fold change between ChIPed and mononucleosome libraries in the proportion of the 11-mers present in fragments. (D) The log₂ fold change in proportion of 11-mers between anti-CENH3 ChIPed and mononucleosome fragments (Upper), or between anti-H3K4me2 ChIPed and mononucleosome fragments (Lower) along the 167-bp CentO sequence (shown together with the 155-bp CentO consensus for comparison).

We mapped all perfectly matched ChIP-seq fragments to this region, whether a fragment was mapped to a single or multiple sites in the BAC clone. The fragment count peaks in this region corresponded well with individual CentO monomers (Fig. 4B) and were similar to those located in other CENH3-enriched subdomains (Fig. 1 and Fig. S1). These results support the view that each CentO monomer is associated with a single CENH3 nucleosome. Nucleosome centers were aligned to examine the distribution of WW and SS dinucleotides within ±150 bp. We observed a similar SS/WW distribution pattern to that derived from single-copy centromeric sequences, superimposed on an ∼10-bp periodicity in the percentage of WW (Fig. 3C). A 10-bp periodicity of WW dinucleotides is thought to stabilize the rotational setting of the DNA double helix as it bends around the histone core (22, 37).

CENH3 Is Depleted from a 167-bp CentO Variant.

BAC a0038J12 contains six copies of a 167-bp CentO variant that has a tandem duplication of 12 bp (CGAACGCACCCA). We determined the percentage of 11-mer sequences from CentO (both 155-bp and 167-bp variants) in fragments from anti-CENH3 ChIP, anti-H3K4me2 ChIP (38) and from input mononucleosome libraries. Then, we plotted the log₂ fold change between the percentage of ChIPed fragments and input fragments for each 11-mer (Fig. 4 C and D). We found that, for the anti-CENH3 ChIP, the log₂ fold change of two 11-mers (CACCCACGAAC and ACCCACGAACG) that are unique to the duplication junction of the 167-bp CentO sequence were strongly reduced compared with other flanking 11-mers present in both the 155-bp and 167-bp variants. Conversely, we found that these 11-mers were enriched in the H3K4me2 library over input. These results suggest that CENH3 nucleosomes prefer the 155-bp CentO variant and are depleted from the 167-bp CentO variant, which is preferentially occupied by H3 nucleosomes dimethylated on lysine 4.

CENH3 Nucleosomes Are Translationally Phased with 155-bp Periodicity on CentO.

We generated phasograms (33, 34) by plotting the frequency of occurrence of each distance between the fragment midpoints of two ChIP-seq fragments within a window of the reference genome (Fig. S3A).

The phasogram of input mononucleosomal DNA fragments indicated that the average distance between two adjacent nucleosomes is 192 bp (R² = 0.9997, P value = 2.4 × 10⁻⁸) in the rice genome, and a virtually identical spacing was found for mononucleosomes on Chromosome 8, exclusive of CENH3-enriched regions (Fig. S3 B and C). In contrast, phasograms of CENH3 nucleosomes from the CENH3-enriched regions of Cen8 that lack CentO sequences showed no clear global periodicity (Fig. 5A) although we cannot exclude the possibility that small sets of nucleosomes might be regularly spaced.

Fig. 5.

CENH3 nucleosome phasing in rice Cen8. Phasograms were generated using ChIP-seq fragments from Cen8 for non-CentO (Upper) and CentO_8 (Lower) sequences. The x axes show the range of recorded phases. The y axes represent frequencies of distance between the midpoints of two fragments using (A) the total fragments, (B) >127-bp fragments, or (C) <127-bp fragments.

Sequence identity among CentO monomers ranges from 76% to 100% (36) so many CentO fragments are unique. To examine the internucleosomal spacing of CENH3 nucleosomes on CentO, unique fragments from all CENH3 nucleosomes were mapped onto CentO_8, and phasograms were generated. This analysis revealed that the average distance from the middle of one CENH3-protected particle to the middle of the adjacent CENH3-protected particle is ∼155 bp (R² > 0.9999, P value = 8.9 × 10⁻¹⁰) (Fig. 5A). To determine whether this 155-bp periodicity of CENH3 peaks seen for CentO_8 extends to CentO genome-wide, the midpoints of all unique fragments were mapped onto four tandem copies of the CentO consensus. A clear peak was seen on each CentO in the midpoint counts (Fig. S4A). We conclude that CentO imposes a preferred translational position with 155-bp periodic phasing on CENH3 nucleosomes.

Subnucleosome-Sized Particles Map Between Phased CENH3 Nucleosomes on CentO_8.

We noticed that smaller subpeaks could be discerned between successive 155-bp peaks in CentO_8 phasograms (Fig. 5A). Arbitrary separation of CentO fragments into >127-bp and <127-bp size classes revealed that these interstitial subpeaks are seen only in the phasogram representing the larger size fragments (Fig. 5B) whereas the phasogram of smaller fragments showed only the major 155-bp periodicity (Fig. 5C). Consistent with the double periodicity of the phasogram of the longer fragments, midpoint-counts of the >127-bp fragments mapped to the CentO consensus showed two peaks per CentO repeat, and both were offset from the position of the dominant peak seen in the <127-bp fragments (Fig. S4 B–D). The absence of the dominant midpoint-count peak in the longer >127-bp fragments suggests that the midpoints of longer fragments have shifted from the CENH3 nucleosome centers because linker DNA flanking these nucleosomes is also being protected in the >127-bp fragments.

We hypothesized that one or more labile particles that can protect linker DNA from MNase are located between the CENH3 nucleosomes and are pulled down by the anti-CENH3 antibody with CENH3 nucleosomes when they are present on the same DNA fragment, as has been observed in budding yeast (32). These longer protected fragments would include DNA protected by particles on the “left” or “right” side of an adjacent CENH3 nucleosome, and the phasogram peaks and subpeaks would then originate from plotting the distances between the midpoints of these two classes of fragments (see the model in Fig. S5).

To attempt to detect such an internucleosomal particle(s), which might represent other kinetochore proteins, we plotted the distributions of lengths of fragments in the 0.5-U and 2.5-U libraries that crossed nucleosome peak centers in CentO_8 (Fig. 6A) and that crossed the point halfway between the nucleosome centers (peak + 78 bp) (Fig. 6B). A peak of ∼35-bp fragments in the counts of peak + 78 bp fragments suggests that a particle pulled down with CENH3, perhaps through protein–protein contacts, protects DNA fragments of this size. The particle was enriched in the input mononucleosome libraries, indicating that the majority of it dissociated from CENH3 nucleosomes in the MNase digestion, consistent with its being a labile particle between CENH3 nucleosomes. When we made a similar fragment-length plot for fragments at the peak + 78 bp position for all CentO, a peak at ∼35 bp was not evident, although fragments in this general size range were strongly enriched in the input libraries relative to anti-CENH3 libraries (Fig. S6). Therefore, the labile particle that protects ∼35 bp between successive CENH3 nucleosomes may be specific to CentO_8.

Fig. 6.

Fragment-length distribution of CentO fragments from CentO_8. (A) Fragments that pass through the nucleosome peak or (B) the halfway point between nucleosome peaks (peak + 78 bp). The proportion of 35-bp fragments in B is decreased in anti-CENH3 libraries (Left) compared with input libraries (Right) and also decreases with increasing MNase.

CentO Nucleosomes Have an ∼10-bp Periodicity of MNase Cleavage.

To further address the size and position of protected particles on CentO, we made V-plots using all CENH3-enriched CentOs aligned on the CentO consensus, using the 0.5-U or 2.5-U libraries, which were not size-selected. Surprisingly, although a 155-bp periodicity was observed in the V-plots, the dominant pattern was a 10-bp periodicity of MNase cleavage, yielding fragments as small as 20 bp, the limit of recovery (Fig. 7). We infer that the small fragments observed are due to internal cleavage within CENH3 nucleosomes. This 10-bp periodicity is also evident in the CentO fragment length distribution plots (Fig. 6 and Fig. S6). Such a strong 10-bp periodicity is remarkable, given that the CentO V-plot cleavage pattern is very different from the patterns seen under the same light MNase-digestion conditions for V-plots of CENH3 protection that also include non-CentO sequences (Fig. 2 D–F). The pattern of cleavage also differs from MNase digestion of naked CentO sequences (Fig. S7). The regular internal cleavage makes it impossible to use V-plots to infer the sizes of protected particles on CentO. Internal nucleosome cleavage by MNase with 10-bp periodicity has been attributed to the cleavage of accessible nucleotides that lie further from the surface of the histone core at each helical turn as DNA wraps around the core (39). The regular cleavage pattern observed therefore most likely reflects rotational phasing of CENH3 nucleosomes on CentO.

Fig. 7.

Ten-base-pair periodicity of MNase cleavage on CentO. (A) The distances from the midpoints of fragments from the 0.5 U and 2.5 U libraries to the center of three copies of the CentO consensus sequence (x axis) were mapped against the lengths of the fragments (y axis) to generate V-plots. (B) Counts of fragment lengths show a 10-bp periodicity of MNase cleavage on CentO.

Discussion

Size of Protected DNA on CENH3 Nucleosomes.

Rice centromeres provide the opportunity to examine CENH3 nucleosomes over both CentO satellite repeats and non-CentO centromeric sequences. Our ChIP-Seq fragments from Cen8 chromatin revealed well-positioned nucleosomes over both CentO and non-CentO sequences, confirming and refining previous mapping of CENH3-enriched regions (12). Aligning these well-positioned nucleosomes allowed us to map the left and right cut sites around CENH3 nucleosomes, which indicated a protected region of about 112 bp. A precise size estimate was made by constructing V-plots, in which the minimum protected size of CENH3 nucleosomes was found to be ∼100 bp even in lightly digested chromatin. This minimum size is similar to that recently reported for ChIPed nucleosomes containing the human cenH3, CENP-A (20), and budding yeast cenH3, Cse4 (32). The protected sizes of DNA on all of these cenH3 nucleosomes are smaller than the measured sizes (125–150 bp) (40–42) of DNA protected by CENP-A octameric nucleosomes, including the 147 bp protected by octameric CENP-A nucleosomes assembled on human centromeric α-satellite (43). The composition of cenH3 nucleosomes has been controversial, with the proposal of tetrameric, hexameric, and octameric models of the arrangement of subunits (44). Our results do not directly address the composition of rice CENH3 nucleosomes but indicate that these nucleosomes protect only enough DNA for a single DNA wrap around a nucleosome core and protect less DNA than the measured sizes of DNA protected by octameric CENP-A nucleosomes.

Translational and Rotational Phasing on CentO Satellite.

In the CentO_8 satellite block of Cen8, we observed that a single peak of CENH3 ChIP fragments was found on each 155-bp copy of CentO. We used phasograms to determine that the bulk of CENH3 nucleosomes were spaced with a periodicity of ∼155 bp on CentO_8, but not on non-CentO sequences, suggesting a fairly precise translational phasing on CentO repeats. This interpretation was supported by mapping fragment midpoints genome-wide onto the CentO consensus, which revealed a single major peak of midpoint-counts on CentO, implying that there is a preferred translational position of CENH3 on CentO throughout the genome.

In contrast, longer fragments showed a double periodicity of phasogram peaks per CentO repeat, and two peaks of midpoint-counts per CentO, both offset from the peak of midpoint-counts in the smaller fragments. This result implies the presence of CENH3 nucleosomes that protect the linker DNA on the left or right sides of the nucleosome DNA and give rise to two sets of fragments that result in the double periodicity per CentO of the midpoint-count peaks and the phasogram peaks. Such extended protection might occur if MNase sometimes cuts on only one side of a DNA-binding protein(s) between CENH3 nucleosomes, which thereby extends protection from MNase to the right or left. In support of this model, we found evidence for a particle that protects ∼35 bp of DNA between CENH3 nucleosomes in CentO_8, although it was found preferentially in the input libraries. This 35-bp particle may have been recovered in anti-CENH3 libraries through protein–protein contacts with CENH3 nucleosomes even though it was not on the same DNA fragment as a CENH3 nucleosome. Future investigation of the locations of other DNA-binding kinetochore proteins may help to test this model.

Our results are paralleled by recent data from human cell culture that indicate that CENP-A nucleosomes are largely translationally phased on centromeric α-satellite and that shorter and medium-length fragments (100–119 bp and 120–139 bp, respectively) showed better phasing than longer fragments (140–160 bp) (20), consistent with our observations on how fragment length relates to translational phasing. On maize CentC, a 156-bp periodicity in MNase cleavage was observed although the same periodicity in the cleavage of naked CentC DNA obscured whether the cleavage pattern reflected translational phasing (21). In the maize study, only fragments 145–175 bp were selected so phasing similar to that seen in shorter fragments in human and rice centromeres might have been missed. In maize, CentC showed an ∼10-bp periodicity in AA/TT dinucleotides (21), similar to that observed for WW dinucleotides in CentO. Such periodicity of WW dinucleotides for each 10.4-bp turn of the DNA double helix is believed to facilitate bending around nucleosomes and favor rotational phasing (22, 23, 45). In rice, we observed not only a 10-bp periodicity in WW dinucleotides in the CentO sequence, but also a 10-bp periodicity in the MNase cleavage pattern of CENH3 nucleosomes on CentO that was absent from the cleavage pattern when non-CentO nucleosomes were included. Although MNase cuts preferentially within the linker DNA between nucleosomes, it also cuts internally in nucleosomes with an ∼10-bp periodicity at sites that are accessible because they lie further from the surface of the histone core (39). We conclude that the cleavage pattern of MNase on CentO (Fig. 7) provides in vivo evidence for rotational phasing of CENH3 on CentO.

Implications for Evolution of Centromeric Satellites.

Satellite sequences are known to undergo rapid expansions or contractions (7, 8, 36), which has been attributed to their “selfish” competition for preferential transmission in female meiosis, or “centromere drive” (2, 46). However, the reasons for the success or failure of any particular satellite in this competition have remained obscure. Nor has it been clear why satellites populate centromeres or how new satellites arise. Our evidence that CentO confers translational and rotational phasing on rice CENH3 nucleosomes suggests that regular positioning of cenH3 nucleosomes is advantageous for centromere formation. The 10-bp periodicity of WW dinucleotides is believed to stabilize nucleosomes through reducing the deformation energy required to wrap DNA around them (37), and the strongly positioning Widom 601 sequence derived from α-satellite stabilizes conventional nucleosomes relative to other DNA sequences, such as 5S rDNA (47). These observations suggest that satellites may evolve to stabilize cenH3 nucleosomes against the pulling forces they undergo during chromosome segregation. Stabilizing cenH3 nucleosomes in a regular structure on satellite repeats may help to prevent the loss of cenH3 nucleosomes that are under tension, and to facilitate formation of the kinetochore.

Stabilization of cenH3 nucleosomes can explain how new satellites arise and populate centromeres. Tandem duplication of a sequence of any length with a selective advantage for cenH3 stabilization can potentially lead to its expansion into a satellite array through unequal crossover or other modes of recombination (48). The proposed transition of evolutionarily new centromeres that form initially as neocentromeres on unique sequences into mature satellite-based centromeres (3, 49) could then be effected either through transposition of existing satellites from another location to the neocentromere, or through duplication of endogenous neocentromere sequences (49, 50) that stabilize cenH3 nucleosomes, followed by expansion and fixation of the stabilizing repeat through recombination and centromere drive. Although a new satellite sequence would usually be at least long enough to wrap a cenH3 nucleosome, even smaller sequences could contribute to the ∼10-bp periodicity of WW dinucleotides that stabilizes rotational positioning of cenH3 nucleosomes, as has been suggested for the 20-bp CentAs satellite repeats of Astragalus sinicus (51). Satellites longer than the length of two or three nucleosomes are rare (9), perhaps because longer sequences are unlikely to contribute to stabilization equally throughout their length, putting them at a disadvantage to derivative expansions of their own most stabilizing subsequences.

The total size of CENH3 domains in grass species correlates with genome size, and chromosomes of different sizes in the same species tend to have CENH3 domains of a similar size, even if they have different sizes of satellite arrays. This relationship may be explained by a limiting component (52), which may be CENH3. If the expansion of a stabilizing satellite directs CENH3 to a particular centromere, it might deplete CENH3 from other centromeres. Redistributing CENH3 through compensatory expansions or contractions of satellites on the same or other chromosomes or destabilizing selfish satellite interactions through adaptation of kinetochore proteins (53–56) may help to ensure that a functional centromere is present on each chromosome. Recurrent cycles of expansions of stabilizing satellites seeking a genetic advantage in female meiosis followed by restoration of epigenetic centromere specification by adaptation of kinetochore proteins may explain the rapid evolution of satellites and kinetochore proteins (57) and may resolve the paradox of the common occurrence of satellites at centromeres, which nevertheless are epigenetically specified.

Materials and Methods

Materials.

Rice cultivar “Nipponbare” seeds were germinated at room temperature for 3 d. Germinated seeds were then sowed in soil to continue to grow in the greenhouse. Two-week-old seedlings were collected for nuclei and DNA isolation.

The antibody to the N terminus of rice CENH3 has been previously described (27). The antibody recognizes an epitope on CENH3 that appears to be accessible in all chromosomal contexts, not only on CentO and other rice sequences, but in CENH3s of all grass species that have been tested (for example, ref. 52).

MNase Digestion, ChIP, and ChIP-seq.

The collected seedlings were ground into fine powder in liquid nitrogen. The ChIP procedure, including nuclei isolation, MNase digestion, immunoprecipitation using anti-cenH3 antibodies, and recovery of ChIPed DNA for development of ChIP-seq libraries, was carried out by following the published protocols (58), except that purified rice chromatins were differentially digested using 0.5 U, 2.5 U, and 4.0 U of MNase (Sigma; N5386-200U), respectively. ChIPed DNA and input DNA corresponding to 0.5 U and 2.5 U MNase digested chromatins were used for customized ChIP-seq library development following the published protocols (31), which can recover insert sizes down to 20 bp.

Collection of mononucleosome-sized DNA for sequencing was performed as described previously (29). Rice chromatin was digested by MNase into ∼90% mononucleosomes plus 10% dinucleosomes. Mononucleosome-sized DNA fragments were gel purified for library preparation. Nipponbare genomic DNA was also digested by MNase to the same relative extent as DNA from the chromatin digestion, and 100–200 bp DNA were cut from 2% agarose gel and purified for library construction. All libraries were sequenced in the paired mode using the Illumina platform. All sequencing data are available from GEO (accession no. GSE50755).

Data Analysis.

DNA fragments were mapped to the rice genome (TIGR7/IRGSP1) (59, 60), with Bowtie (61) reporting nonmismatches and uniquely mapped sites. The centers of CENH3 and canonical nucleosomes were identified by nucleR (28). V-plots were generated by following our previous study (31, 32). Phasograms were based on calculating the frequency of nucleosome distance between midpoints of all fragments in specific windows. The counting result was plotted as a phasogram, a waveform-like histogram, which represents frequency of distance between all nucleosomal DNA fragments in a specific window. The peaks of phasograms reflect the periodicity or phasing distance between the nearest-neighbor nucleosome, next-nearest neighbors, etc. We used the phasogram peaks to calculate the nucleosome spacing by applying the linear-fitting model from R. The K-mer (11-mer) count was calculated by applying JELLYFISH (62). All ChIP-seq fragments derived from the 4.0 U MNase digestion were mapped to four tandem copies of the CentO consensus using Novoalign. We randomly kept one mapping site, if one fragment was mapped to multiple sites. Data processing and analysis were done using Perl and R.