Unbiased chromatin accessibility profiling by RED-seq uncovers unique features of nucleosome variants in vivo

Abstract

Background: Differential accessibility of DNA to nuclear proteins underlies the regulation of numerous cellular processes. Although DNA accessibility is primarily determined by the presence or absence of nucleosomes, differences in nucleosome composition or dynamics may also regulate accessibility. Methods for mapping nucleosome positions and occupancies genome-wide (MNase-seq) have uncovered the nucleosome landscapes of many different cell types and organisms. Conversely, methods specialized for the detection of large nucleosome-free regions of chromatin (DNase-seq, FAIRE-seq) have uncovered numerous gene regulatory elements. However, these methods are less successful in measuring the accessibility of DNA sequences within nucelosome arrays.

Results: Here we probe the genome-wide accessibility of multiple cell types in an unbiased manner using restriction endonuclease digestion of chromatin coupled to deep sequencing (RED-seq). Using this method, we identified differences in chromatin accessibility between populations of cells, not only in nucleosome-depleted regions of the genome (e.g., enhancers and promoters), but also within the majority of the genome that is packaged into nucleosome arrays. Furthermore, we identified both large differences in chromatin accessibility in distinct cell lineages and subtle but significant changes during differentiation of mouse embryonic stem cells (ESCs). Most significantly, using RED-seq, we identified differences in accessibility among nucleosomes harboring well-studied histone variants, and show that these differences depend on factors required for their deposition.

Conclusions: Using an unbiased method to probe chromatin accessibility genome-wide, we uncover unique features of chromatin structure that are not observed using more widely-utilized methods. We demonstrate that different types of nucleosomes within mammalian cells exhibit different degrees of accessibility. These findings provide significant insight into the regulation of DNA accessibility.

Figure 1

The RED-seq method for genome-wide measurement of RE accessibility. (A) RED-seq workflow. RSs are shown in red, yellow boxes (Step 3) represent RS-proximal adaptors, dark blue boxes (Step 5) represent RS-distal adaptors, orange circles represent biotin, light blue boxes represent paired-end PCR primers, large blue circles (Step 1) represent nucleosomes, and DNA is shown in black. (B) Ethidium bromide stained agarose gel indicating bulk digestion levels of chromatin and naked DNA. (C) An example FASTQ file is shown to illustrate the near-uniform sequencing of the RS-containing end of each fragment in the library, signified by the large enrichment of G at position 5, and a CC dinucleotide at positions 7 and 8, derived from the cleaved and blunt-ended Sau96I site (GNCC).

Figure 2

Comparison of RED-seq to naked DNA digestion. (A) RE accessibility reads from mouse ESC chromatin (top) and naked DNA (bottom) from a 3 Mb region of chromosome 14 (Chr14). Shown are normalized reads per million (RPM). (B) Scatterplot of RE accessibility [Log₂(RPM)] for Chr14 from chromatin relative to naked DNA. (C) RE accessibility from chromatin and naked DNA of two Hox genes, Hoxa4 and Hoxa11, which are silent in ESCs. Dotted lines highlight the genomic regions with RE accessibility differences apparent between chromatin and naked DNA. (D) RE accessibility from chromatin and naked DNA of two highly expressed genes in ESCs, Oct4 and Eef1a1.

Figure 3

RED-seq captures the enhanced accessibility of open chromatin regions. Average RE accessibility (A, C) and nucleosome occupancy (B, D) [GEO:GSM1400766] of indicated chromatin domains. RED-seq or MNase-seq data are aligned on the centers of all peaks of DHSs (A-B), or CTCF binding sites (C-D), and averaged within a 2 kb region (-1000 to +1000 bp from the peaks). Normalized RE accessibility and RS density are shown. RE accessibility was normalized as in Figure 2. There are 159,331 DHSs [GEO:GSM1014154] (A-B), and 15,657 CTCF binding sites [GEO:GSE11431] (C-D) plotted. (E-F) Chromatin accessibility determined by RED-seq or DNase-seq and nucleosome occupancy are shown surrounding CTCF binding sites (E) or DHSs (F). Arrows indicate the phased peaks of RE accessibility found within linker regions.

Figure 4

Cell type-specific differences in chromatin accessibility. (A) Average RE accessibility of ESCs (blue) or MEFs (red) shown relative to DNase I hypersensitive sites (DHSs) identified in ESCs [GEO:GSE46588]. (B) Nucleosome occupancy of the same regions is shown for ESCs [GEO:GSM1400766] and MEFs [GEO:GSM1004654]. (C) Average RE accessibility and (D) nucleosome occupancy surrounding CTCF binding regions in ESCs [GEO:GSE11431] are shown for ESCs and MEFs. (E) Average accessibilities over DHSs and CTCF binding sites were quantified for biological replicate experiments from –200 to +200 bp with respect to the indicated feature. P-values indicating statistical significance of accessibility between ESCs and MEFs are indicated. (F, G) RE accessibility of ESCs and MEFs surrounding the Oct4 gene (F) and two genes within the Hoxb cluster (G). RNA Polymerase II (RNA PolII) ChIP-seq reads [GEO:GSE29184] from ESCs and MEFs are shown for the same regions. RED-seq and MNase-seq data are plotted as in Figure 3.

Figure 5

Alterations in RE accessibility during ESC differentiation. (A) Brightfield images of control (EGFP) or Oct4 KD ESC colonies indicate colony flattening and elongated cellular morphology upon Oct4 depletion. (B) Western blot of Oct4 in control (EGFP) or Oct4 KDs, indicating KD efficiency. RNA Polymerase II blot (Pol II) is shown as a loading control. (C, E) Average RE accessibility upon EGFP or Oct4 KD is shown relative to DHSs (C), or CTCF binding sites (E). (D, F) MNase-seq data. Nucleosome occupancy over DHSs (D), or CTCF binding sites (F). RED-seq and MNase-seq data are plotted as in Figure 3.

Figure 6

Loss of chromatin accessibility at some CTCF binding sites correlates with reduced CTCF binding upon ESC differentiation. (A) Differences in RE accessibility at specific DHSs were confirmed by qPCR across an RS of interest at each locus. Remaining uncut DNA after RE digestion of each indicated KD is shown for several DHSs that exhibited accessibility differences by RED-seq. Data are normalized to uncut genomic DNA. (B) Confirmation of restriction enzyme accessibility surrounding CTCF binding sites, as in (A). (C) CTCF ChIP-qPCR data are shown for the indicated KDs at several CTCF binding sites. Controls are CTCF binding sites in which accessibility did not change upon Oct4 KD. Data are presented as a percentage of input DNA. Shown are the mean ± SD of three technical replicates from one representative experiment of two biological replicates performed. (D-E) RED-seq data (D) and MNase-seq data (E) over Klf4 binding sites, plotted as in Figure 3. (F) Average accessibilities over DHSs, CTCF binding sites, and Klf4 binding sites were quantified for biological replicate KD experiments from –200 to +200 bp with respect to the indicated feature. P-values indicating statistical significance of accessibility between EGFP KD and Oct4 KD are indicated.

Figure 7

Enhanced accessibility of DNA bound by H2A.Z-containing nucleosomes. Average RE accessibility (A-C) and nucleosome occupancy (D-F) shown relative to 320,135 randomly selected nucleosomes (A, D), 39,437 H2A.Z-containing nucleosomes [GEO:GSE34483] (B, E), or 8,287 H3.3-containing nucleosomes [GEO:GSE16893] (C, F). Data are plotted as in Figure 3. P-values indicating statistical significance of accessibility between H2A.Z and average nucleosome profiles, as well as H3.3 and average nucleosomes are indicated.

Figure 8

Factors required for H2A.Z or H3.3 deposition are required for enhanced accessibility of regions normally bound by these histone variants. (A) Chromatin accessibility determined by RED-seq averaged over regions of the genome bound by H2A.Z, as in Figure 7. Shown are control (EGFP KD) and Ep400 KD ESCs. (B) Western blot of p400 in control (EGFP) or Ep400 KDs, indicating KD efficiency. Actin is shown as a loading control. (C) Average accessibilities over H2A.Z-marked nucleosomes were quantified for biological replicate experiments from –200 to +200 bp with respect to the H2A.Z peak. P-values indicating statistical significance of accessibility between EGFP and Ep400 KDs are indicated. (D) Chromatin accessibility determined by RED-seq averaged over regions of the genome bound by H3.3. Shown are control (EGFP KD) and Hira KD ESCs. (E) Western blot of Hira in control (EGFP) or Hira KDs, indicating KD efficiency. Actin is shown as a loading control. (F) Average accessibilities over H3.3-marked nucleosomes were quantified for biological replicate experiments from –200 to +200 bp with respect to the H3.3 peak. P-values indicating statistical significance of accessibility between EGFP and Hira KDs are indicated. (G) Effects of Ep400 or Hira KD on average nucleosome accessibility shown by plotting RED-seq data over the same 320,135 randomly selected nucleosomes as in Figure 7A. (H) Effects of Ep400 or Hira KD on chromatin accessibility over CTCF binding sites, as in Figure 3C. (I) Average accessibilities over CTCF-binding sites were quantified for biological replicate experiments from –200 to +200 bp with respect to the peak of CTCF-binding. P-values indicating statistical significance of accessibility between EGFP and Hira KDs are indicated.