A high-resolution map of nucleosome positioning on a fission yeast centromere

Abstract

A key element for defining the centromere identity is the incorporation of a specific histone H3, CENPA, known as Cnp1p in Schizosaccharomyces pombe. Previous studies have suggested that functional S. pombe centromeres lack regularly positioned nucleosomes and may involve chromatin remodeling as a key step of kinetochore assembly. We used tiling microarrays to show that nucleosomes are, in fact, positioned in regular intervals in the core of centromere 2, providing the first high-resolution map of regional centromere chromatin. Nucleosome locations are not disrupted by mutations in kinetochore protein genes cnp1, mis18, mis12, nuf2, mal2; overexpression of cnp1; or the deletion of ams2, which encodes a GATA-like factor participating in CENPA incorporation. Bioinformatics analysis of the centromere sequence indicates certain enriched motifs in linker regions between nucleosomes and reveals a sequence bias in nucleosome positioning. In addition, sequence analysis of nucleosome-free regions identifies novel binding sites of Ams2p. We conclude that centromeric nucleosome positions are stable and may be derived from the underlying DNA sequence.

Figure 1.

Structure of S. pombe centromeres. (A) Schematic diagram of the centromeres. Each of the three centromeres contains central core cnt and flanking imr and otr repeat regions. imr and otr have repeat sequences so that a probe designed for these regions will hybridize to DNA fragments from multiple loci. Cross-hybridization thus renders nucleosome mapping unfeasible for imr and otr. Similarly, cnt1 and cnt3 on chromosomes 1 and 3 share >70% identical sequences, and thus, only the partial non-overlapping portions can be tiled unambiguously. On the other hand, cnt2 contains unique sequences and can be tiled densely for nucleosome mapping. (B) Gel electrophoresis and Southern. Chromatin or naked genomic DNA was treated with MNase digestion (see Methods for the details of digestion conditions), and the resulting DNA fragments were run on gel. Southern hybridization was performed using a cnt DNA fragment (pKT110) or an otr DNA fragment (pYC148) (Takahashi et al. 1992). Bands containing mono-nucleosomes or DNA of similar size from naked DNA digestion (white boxes) were extracted and hybridized onto the microarray. The otr region shows ladders for both the wild type and the mis12 mutant. In cnt, the mis12 mutant still exhibits distinct ladders, while the wild type shows a smear with only faint mono- and di-nucleosome bands.

Figure 2.

Chromatin structure of (A) ade6 and (B) cnt2. The mono-nucleosome band in Figure 1B was extracted, labeled with Cy5 dye, and hybridized onto the microarray, with MNase-digested naked DNA of similar size labeled with Cy3 as control. (A, black line) The mean of nucleosome data from four biological replicate experiments. (Red dotted lines) The MNase cleavage sites of eight previously known positioned nucleosomes (Bernardi et al. 1991). Positioned nucleosomes detected by our method are shown as histone octamers and agree very well with Bernardi et al. (1991). (B) The median nucleosome signal of three biological replicates, each with (bold black) two technical replicates; (colored lines) individual replicates. The technical replicates come from tiling both strands of cnt2. Positioned nucleosomes based on the median signal are shown as histone octomers. (Red arrows, including the one below pointing upward) The GATA motifs in troughs that are potential Ams2-binding sites.

Figure 3.

Chromatin structure is preserved in mutants. (A) Wild-type versus Cnp1p overexpression. (B) Wild type versus mutants affecting Cnp1p localization. (C) Wild type versus outer kinetochore mutants. The array design in B covered a region longer than that in A and C.

Figure 4.

Nucleosome data correlate well with sequence-based predictions in (A) cnt2 and (B) cnt1 but not in (C) ade6. (Black line) The median nucleosome data for wild type; (blue line) the log nucleosome occupancy probability, computed as in Segal et al. (2006). (B, green line) The nucleosome data in the region repeated in cnt3; (red dotted line) the running window (size 120 bp) log GC content. In cnt1 and cnt2, the correlation between the nucleosome data and GC content is high, and the computational prediction of nucleosome locations works well. In ade6, however, the correlation is only 0.1, and the computational algorithm also fails to predict nucleosome-free regions.

Figure 5.

Enriched motifs in troughs. (A) Poly(T) and (B) histone motifs were found by scanning all trough regions with MEME.