Precise genome-wide mapping of single nucleosomes and linkers in vivo

Abstract

We developed a chemical cleavage method that releases single nucleosome dyad-containing fragments, allowing us to precisely map both single nucleosomes and linkers with high accuracy genome-wide in yeast. Our single nucleosome positioning data reveal that nucleosomes occupy preferred positions that differ by integral multiples of the DNA helical repeat. By comparing nucleosome dyad positioning maps to existing genomic and transcriptomic data, we evaluated the contributions of sequence, transcription, and histones H1 and H2A.Z in defining the chromatin landscape. We present a biophysical model that neglects DNA sequence and shows that steric occlusion suffices to explain the salient features of nucleosome positioning.

Keywords: Biophysical modeling; Chemical cleavage mapping; Linker quantization.

Fig. 1

Concurrent chemical cleavages of H3 histones eliminate the background noise in the chemical mapping experiments. a Integrative Genomics Viewer browser snapshot illustrating nucleosome maps obtained by different techniques. Previous chemical maps, using mutated histones H4S47C (tracks 1–5), are affected by increased background of non-specific cleavages produced by free phenanthroline. Tracks 1 and 2 show the nucleosome occupancy inferred from H4S47C cleavages reported by [23] by symmetrically extending each cleavage position to a footprint of 101 bp (to emphasize the linker positions) between neighboring nucleosomes. Tracks 3–5 show similar data from [26]. The background from the promoter of YBR084W is strongly reduced when we consider only concurrent chemical cleavages, both produced by H3Q85C (tracks 6–8). Tracks 6–8 show the occupancy produced by symmetrically extending the ~ 50-bp fragment centers to a footprint of 101 bp. The absence of histones from the promoter is confirmed by two alternative chromatin immunoprecipitation sequencing (ChIP-seq) methods (tracks 9 and 10) (data from [51, 60]). To make the occupancy profiles comparable, the MNase-ChIP-seq nucleosome map was generated by symmetrically extending the position of the dyad to a footprint of 101 bp. For sonication-ChIP-seq profiles, the occupancy was computed by stacking all the fragments with length between 50 and 200 bp, as in this case the centers of the fragments do not necessarily represent nucleosome dyads. b Average dyad distributions obtained from our H3Q85C cleavage experiment show reduced background noise compared to the map obtained from the H4S47C cleavage experiment. c Length distribution of DNA fragments before and after gel filtration. d Left-right V-plot display of chemical cleavages, where the X-axis shows the position of the center of all 16 aligned centromeres and the Y-axis shows the length of each fragment; red pixels represent the left fragment ends and blue pixels represent the right ends. The two edges of the ~ 80-bp centromere DNA element II (CDEII) are indicated by dotted lines. For H4S47C, the red and blue vertical features between the dotted lines imply that cleavages occur at two distinct positions within CDEII over the population. However, for H3Q85C, no cleavages are seen within the centromeric region, indicating that centromeric nucleosomes do not contain H3 histones. e Cleavage density (left) and average occupancy (right) plots of the data shown in d, comparing H4S47C and H3Q85C cleavage data. Mock control data were obtained by phenanthroline treatment and cleavage reactions performed using wild-type cells as described [26]. f Cleavage density plots comparing H4S47C, H3Q85C, and Mock control cleavage data over aligned transfer RNA (tRNA) genes. g The preferred rotational positions are more evident in the H3S47C cleavage data, even when comparing the nucleosome positions called from the H4S47C cleavage experiment. All dyad distributions were normalized such that the average dyad density equals 1 for every chromosome

Fig. 2

H3Q85C cleavages map rotational positions genome-wide. Nucleosome dyad positions based on raw H3Q85C (a) and H4S47C (b) cleavage data were aligned and dinucleotide frequencies were mapped at each base pair position. The genome-wide H4S47C rotational positioning signature is irregular and much weaker than the H3Q85C signal. A strong and more regular positioning signature (c) is obtained for H4S47C data by using the “called” subset H4S47C nucleosome dyads based on stringent cutoffs [23]

Fig. 3

Characterization of yeast promoters. a Heat map of the nucleosome dyad distribution at yeast promoters. Each row represents a 2-kb window focused on the NDR centers of yeast genes, which are sorted according to the NDR width. b Zoomed-in view of a 500-bp region around the NDR center, including only the +1 and –1 nucleosomes of all yeast genes. The high resolution of the nucleosome positioning data allows the identification in the ensemble of cells of alternative rotational positions (parallel sigmoidal stripes corresponding to +1 and –1 nucleosomes), which differ by multiples of the helical twist. c NDRs are flanked by H2A.Z-containing nucleosomes (data from [61]). NDRs are generally bound by TATA-binding protein (d) (TBP, data from [62]), Reb1 (e) (data from [63]), remodelers RSC, Snf2, and ISW1 (f-h) (data from [64]). NDRs are accessible to DNase I (i) (data from [65]) and transposase Tn5 (j) (data from [66]). k In rsc8-depleted cells, nucleosomes +1 and –1 shift by a multiple of the helical twist, maintaining the same rotational setting. Heat maps showing the average G/C content of the preferred dyad locations. MNase-seq data from [58]

Fig. 4

Nucleosome phasing patterns are characteristic of 5’ but not 3’ gene ends. a Nucleosome dyads from combined H4S47C- and H3Q85C-anchored cleavage data. Gene promoters aligned at +1 nucleosomes. Divergent (top) and tandem (bottom) genes are separated and sorted according to the distance between the +1 nucleosome and the NDR of the upstream gene. The vertical red stripes from the gene bodies indicate strong nucleosome phasing relative to the NDR. The nucleosome distribution in the region between neighboring NDRs is characterized by an interference pattern generated by phasing signals originating from both flanking NDRs. b Nucleosome distribution at the 3’ gene ends aligned at the transcription termination site (TTS). Convergent (top) and tandem (bottom) genes are separated and sorted according to the distance between the TTS and the NDR of the downstream gene. The absence of vertical red stripes at the TTS of convergent genes indicates that the TTS is not a nucleosome phasing element, unless it overlaps with the TSS of the downstream gene, as in the case of some of the tandem genes. Even in the case of tandem genes, it is evident that nucleosomes are not phased relative to the TTS but relative to the position of the nearby TSSs, as the red stripes are not vertical but bent according to the positions of the downstream genes. c, d Organization of in vitro reconstituted nucleosomes (data from [13]) near gene ends shows that DNA sequence is not sufficient to dictate nucleosome phasing patterns. In vitro, NDRs are formed on the regions of high A/T content, suggesting that they may be artifacts of MNase-seq, introduced by the MNase sensitivity of the nucleosomes located on A/T-rich sequences [68, 69], which are easily overdigested. e, f A/T content (ratio of nucleotides A or T) near gene ends shows that TTSs are among the most A/T-rich regions in the yeast genome. In vitro, NDRs near TTSs form at A/T-rich regions and not at the positions where they are observed in live cells

Fig. 5

H3Q85C cleavage data prove linker “quantization.” a A two-dimensional occupancy heat map and length histogram show the coverage of DNA fragments of different sizes relative to +1 nucleosomes. There are three major groups of fragment sizes. Small fragments (~ 50 bp long) originate from two cuts inside the same nucleosome; their centers indicate the position of a nucleosome dyad. Medium fragments (~ 70–130 bp long) and long fragments (~ 135–175 bp long) originate from two cuts inside neighboring nucleosomes, produced at consecutive or alternate cleavage sites, respectively. b Scheme of the cleavage sites and the possible resulting fragments, together with the histogram of DNA fragment lengths obtained in the experiment

Fig. 6

Nucleosome spacing and linker lengths are “quantized” even at the gene level. Length distribution of the long DNA fragments (135–175 bp) resulting from H3Q85C cleavages. Examples of yeast genes among the ones containing the highest numbers of reads over their gene bodies show that the distributions of fragment lengths from individual genes (blue curves) resemble the genome-wide distribution (red curve). This indicates that the same gene can have different levels of nucleosome compaction in different cells and at different times

Fig. 7

Genes sorted by average size of the long fragments from H4Q85C cleavage data. a Genes aligned at the +1 nucleosome and sorted according to the average size of the long fragments (135–175 bp) that were produced by H3Q85C cleavage. The transcription levels, as measured by the levels of nascent transcripts obtained from NET-seq data [73], indicate that the more transcribed genes have more compact nucleosome arrays. b Average nucleosome dyad density for the yeast genes separated into five quintiles: quintile 1, the top fifth from a, through quintile 5, the bottom fifth from a. The decrease in nucleosome spacing correlates with a widening of the NDR, which is characteristic for the highly transcribed genes. c The distribution of linker histone H1 (Hho1 data from [80]) is also correlated with nucleosome spacing: the longer the nucleosome spacing, the more H1 is bound to the corresponding nucleosomes. The H2A.Z variant is enriched in the +1 nucleosomes (data from [61]); the less active genes appear to have more H2A.Z histones incorporated into their +1 nucleosomes. The more active genes contain more H3K4me3 marks on their bodies (data from [81]) and more TBP bound to their promoters (data from [62]). d Average properties for the five quintiles

Fig. 8

Nucleosome organization modeled by statistical mechanics predicts the correct phasing pattern and NDR locations. a Fitted parameters for the energy barrier localized at the gene promoters. The barrier consists of two halves of a Gaussian distribution (H = 7.88 k_BT; σ = 27.64 bp) with the centers shifted by x₀ = 68 bp from the dyads of +1/–1 nucleosomes and a flat energy barrier in between. The additional fitting parameter, the binding energy for a nucleosome away from the barrier, u − μ = –2.79 k_BT. b Predicted nucleosome organization in gene promoters. Promoters are aligned at +1 nucleosomes and sorted exactly as in Fig. 4a. c Predicted nucleosome distribution at the 3’ ends of genes, which are aligned at the TTS. Genes are sorted as in Fig. 4b. A discussion of the applications of Tonks gas modeling of nucleosome positions is presented elsewhere [105]