Genome-wide nucleosome specificity and directionality of chromatin remodelers

Abstract

How chromatin remodelers cooperate to organize nucleosomes around the start and end of genes is not known. We determined the genome-wide binding of remodeler complexes SWI/SNF, RSC, ISW1a, ISW1b, ISW2, and INO80 to individual nucleosomes in Saccharomyces, and determined their functional contributions to nucleosome positioning through deletion analysis. We applied ultra-high-resolution ChIP-exo mapping to Isw2 to determine its subnucleosomal orientation and organization on a genomic scale. Remodelers interacted with selected nucleosome positions relative to the start and end of genes and produced net directionality in moving nucleosomes either away or toward nucleosome-free regions at the 5' and 3' ends of genes. Isw2 possessed a subnucleosomal organization in accord with biochemical and crystallographic-based models that place its linker binding region within promoters and abutted against Reb1-bound locations. Together, these findings reveal a coordinated position-specific approach taken by remodelers to organize genic nucleosomes into arrays.

Figure 1. ATP-dependent chromatin remodelers bind to specific nucleosome positions at the beginning of genes

(A) Remodeler-bound nucleosomal tags were aligned by the TSS of the underlying gene, binned (5 bp) and smoothed (3 bin moving average). Color intensity represents tag counts. Shown are raw tags, with no background subtraction or normalization. At least one remodeler was detected at about half of all 5866 yeast genes having an annotated TSS. (B) Values from panel A were averaged and plotted as in panel A for remodeler-bound nucleosomes (blue) and all (H3-containing) nucleosomes (gray). Red indicates density, where values represented by blue were divided by values represented by gray. See also Figure S1.

Figure 2. ISWI remodelers interact with terminal nucleosomes

(A,B) Same as Figure 1, except genes having remodeler-bound nucleosomes near their 3′ ends were aligned by the terminal nucleosome dyad of the underlying gene. (C) Heat map representing the venn overlap (illustrated to the left) of those genes containing remodeler-bound nucleosomes near the 5′ end versus 3′ end (see illustration). The overlap is presented as a chi-square distribution (middle) and percentage of overlap (right). The size of the blue circles reflects the number of bound genes.

Figure 3. Chromatin remodelers target similar sets of genes

(A) Heat map representing the venn overlap of genes containing each type of remodeler-bound nucleosome near their 5′-ends, either by chi-square distribution (left) or percentage overlapping (right). Circles reflect population sizes. (B) Same as panel A, but for the 3′ ends of genes.

Figure 4. Model of chromatin remodeler targeting and directionality

Each illustration represents an approximate grouping of genes similarly enriched with remodelers, separated into ISWI and RSC classes (upper and lower panels). Spheres represent the predominant locations of remodelers relative to 5′ and 3′ NFRs. Arrows depict the direction to which the indicated chromatin remodeler moves nucleosomes.

Figure 5. Directionality of chromatin remodelers

(A) Composite plot of nucleosome positions at the 5′ ends of genes of either wild type (WT, gray fill) or mutant (green or pink trace). (B) Line graphs of nucleosome dyad shifts from chromatin remodeler null mutants. The shift is reported as the median distance between a mutant and wild type dyad position for those genes either having (upper panel) or lacking (lower panel) remodeler-bound nucleosomes, as defined in the Methods section. Nucleosome positions are relative to the 5′ NFR, or the terminal nucleosome (TN) at the 3′ end of genes. RSC (Sth1_degron) data is from (Hartley and Madhani, 2009). (C) P-value of the nucleosomal shifts observed in the mutants. Log₁₀ p-values are reported as a heat-map table. White block indicate p-values > 0.01. See also Figure S2.

Figure 6. Directional binding of Isw2 to the +1 nucleosome

(A) Matching of a crystallographic-based model of an Isw2/nucleosome complex (Dang and Bartholomew, 2007) to Isw2 ChIP-exo data. The graph plots the distribution of ChIP-exo crosslinking points (peak-pair midpoints corresponding to exonuclease stop sites) relative to the dyad position of the +1 nucleosome, and orientated such that the nearest TSS is directed to the right. The top 1500 occupied Isw2 peaks were selected, and compared to all others. Also shown is the distribution of Reb1-bound locations (Rhee and Pugh, 2011) for the same collection of genes. The distribution of nucleosome dyads is shown as a gray filled plot. (B) Most genes contain detectable levels of Reb1 and Isw2. All 4,967 genes having an annotated TSS were aligned by their TSS, and the intensity level and positions of ChIP-exo Reb1 and Isw2 peak-pairs plotted. Genes were sorted by intensity level. The order of genes in the two panels are not the same.