Immunostaining of interphase nuclei with anti-H1.2 and H1.5 demonstrates punctate structures
Previous studies [12, 24] have revealed that punctate chromatin structures (“chromomeres”) can be observed within fixed and permeabilized interphase nuclei and mitotic chromosomes of HL-60/S4 cells by immunostaining with bivalent rabbit anti-histone H1.5. Similar punctate structures were also observed in HL-60/S4 cells employing the monovalent Fab fragment from the mouse mAb PL2-6, an autoimmune antibody directed against the nucleosome “acidic patch” (consisting of acidic amino acid residues from histones H2A and H2B) [12, 24, 35, 36]. Figure 2a–f and Additional file 1: Figure S1 present images of undifferentiated HL-60/S4 interphase nuclei immunostained with rabbit anti-H1.2 and with rabbit anti-H1.5. The chromomeric patterns are readily visible by both confocal and STED microscopy. Employing stimulated emission depletion (STED) microscopy yielded an estimate of the diameters of H1-enriched foci (anti-H1.2, ~ 60 nm; anti-H1.5, ~ 70 nm), approximately threefold smaller than the diameters estimated by confocal imaging (~ 210 nm). As stated earlier [12, 24], we suggest that these chromomeres may represent the fixed and stained equivalent of constrained polynucleosome clusters observed by a variety of biochemical and microscopy methods (e.g., Hi-C, replication foci and TIRF microscopy). The punctate immunostaining pattern of H1 epitope distribution likely reflects an in situ chromatin higher-order organization within fixed interphase nuclei.
Chromatin immunoprecipitation (IP) with anti-H1 demonstrates differences in genomic element enrichment/depletion
Currently, analysis of the genome-wide distribution of DNA-binding proteins (unmodified or modified by post-translational changes) is frequently performed by chromatin immunoprecipitation followed by DNA sequencing (ChIP-seq). Usually the intact cells are fixed once with formaldehyde, followed by sonication to nucleosome-size fragments and subsequent IP (xChIP). However, information about the influence of in situ chromatin higher-order structure on protein (epitope) distribution is largely lost during the sonication step, which precedes the IP step. We developed a modified ChIP-seq protocol to preserve information about the influence of in situ chromatin higher-order structure on protein (epitope) distribution. The method involves two formaldehyde fixation steps: the first, of intact cells; the second, following permeabilization and antibody binding, but before sonication, IP and DNA sequencing (xxChIP-seq) [23]. A scheme comparing xChIP-seq and xxChIP-seq is shown in Fig. 1, for the specific situation of mapping the distribution of histone H1. This figure contrasts H1 epitopes that are “exposed” in situ and H1 epitopes that are “hidden” by bound protein and/or chromatin higher-order structure “masking”. The xxChIP-seq method was originally developed to define the DNA sequences within “epichromatin”, the surface of chromatin beneath the nuclear envelope by employing the bivalent mAb PL2-6 [12, 23, 24, 37]. In this situation, the epichromatin epitope, which is present on all nucleosomes, is largely “hidden” internally and “exposed” at the chromatin surface. Performing both methods (xChIP-seq and xxChIP-seq) on the same cell type can furnish a detailed genome-wide comparison of “exposed” versus “hidden” epitope regions.
Employing xChIP-seq and xxChIP-seq with rabbit anti-histone H1.2 and H1.5 antibodies, we determined genome-wide distributions of these signals and using peak calling software MUSIC [26] determined the regions (peaks) with their enrichments. The average size of such peaks was around 2000 bp (Figure S2). xChIP H1.2 and xxChIP H1.2 and H1.5 were characterized on average by a slight depletion of GC content at about 500 bp from the peak summit, whereas in the case of xChIP H1.5, the peaks did not have any pronounced nucleotide signature (Additional file 1: Figure S3).
Figure 3a presents parallel tracks along human chromosome 7, illustrating the density of peaks enriched for histones H1.2 and H1.5 xChIP-seq and xxChIP-seq, as well as several other epigenetic signals measured in HL-60/S4 cells [23, 31] (see all other chromosomes in Additional file 1: Figure S4). Figure 3a illustrates interesting correlations among several tracks. In the regions denoted by black arrowheads, there are more (compared to surrounding regions) of xChIP-seq of domains enriched with H1.2 and H1.5, coupled with deficiencies for xxChIP-seq domains enriched with H1.2 and H1.5. This type of behavior might signify the presence of “hidden” H1 epitopes within the xxChIP-seq reads of H1.2 and H1.5, possibly due to the presence of higher-order chromatin structure (Fig. 1). Many of these regions also correlate with enrichments of H3K4me1, H3K9ac, H3K27ac and RNA Pol II, all markers of transcription-permissible regions (based on published ChIP-seq, see Methods). Perhaps the formaldehyde-fixed transcriptional apparatus (transcription factories?) generates steric “hiding” of H1 epitopes in the xxChIP-seq assay. A systematic correlation analysis comparing xChIP-seq and xxChIP-seq (with themselves and each other) is presented in Fig. 3b. It is of interest that H1.2 xChIP and H1.5 xChIP reveal a positive correlation at the kb scale, and that H1.2 xxChIP and H1.5 xxChIP demonstrate even better correlation. On the other hand, the xChIP signals reveal negative correlation with the xxChIP signals. This latter observation supports the view that fixation-preserved in situ higher-order chromatin structure results in a significant fraction of “hidden” H1 epitopes.
As it is clear from Fig. 3b, genome-wide H1.2 and H1.5 xChIP signals are positively correlated. However, in a number of important regulatory locations we observed mutually excluding arrangements of H1.2 and H1.5. Such examples, shown in Fig. 3c, d, are characterized by “swings” of 3–5 nucleosomes in H1.2 and H1.5 linker histone enrichment, with some regions exhibiting H1 isotype predominance for longer distances. We then set to define the locations of all such regions using NucTools with a sliding window of 100 bp. Interestingly, many such regions were at gene promoters. In particular, this analysis revealed 1715 regions where xChIP H1.2 dominates over H1.5 (39.8% of them at promoters), and 5214 regions where xChIP H1.5 dominates over H1.2 (44.2% of them at promoters). Thus, thousands of gene promoters are enriched either with H1.2 or H1.5, suggesting that differential binding of H1 variants has functional implications. Gene Ontology analysis revealed that promoters with H1.2 or H1.5 dominance were enriched for genes related to ATP binding and enzymatic activity (Additonal file 2: Table ST1 and ST2).
Next, we analyzed the genome-wide distribution of regions enriched with xChIP and xxChIP H1 signals in relation to different genomic features defined using our previous ChIP-seq of histone modifications in HL-60/S4 cells (25). Figure 4 presents a summary of the relative enrichment (or depletion) of various chromatin features with H1.2 and H1.5 peaks determined by MUSIC peak calling based on xChIP-seq and xxChIP-seq. Some of the conclusions: (1) For most of the studied features, H1.2 xChIP displays more enrichment, than H1.5 xChIP. For example, H1.2 xChIP shows more enrichment of enhancers, promoters, CpG islands, Alu repeats, H3K27ac, H3K36me3, H3K4me1, H3K9ac, H3K9me3 and epichromatin. (2) xxChIP H1.2 and H1.5 signals resemble each other more than xChIP H1.2 and H1.5. (3) xxChIP generally shows more depletion of the studied chromatin features, than observed with xChIP, except for Alu and L1 repeats. To some extent, the differences observed comparing H1.2 and H1.5 xChIP are obliterated when comparing H1.2 and H1.5 xxChIP. (4) Reminiscent of conclusions derived from the chromosome tracks displayed in Fig. 3a, transcription-related regions and “active” histone modifications (i.e., enhancers, promoters, CpG islands, H3K4me1, H3K4me3, H3K9ac, H3K27ac and H3K36me3) are enriched in H1.2 and H1.5 xChIP-seq and show significant depletions in H1.2 and H1.5 xxChIP-seq reads. These observations support that in situ chromatin higher-order structures, “preserved” by formaldehyde fixation, can create “hidden” histone H1.2 and H1.5 epitopes. They also suggest that transcription-related regions may have their own higher-order structure.
Differential enrichments of H1 variants around protein-binding sites
The apparent occlusion of H1 epitopes, due to the preservation of chromatin higher-order structure surrounding various chromatin protein-binding sites, is presented in Fig. 5. In the case of CTCF-binding regions, H1 xChIP-seq profiles show weak oscillations that have been previously reported in a number of nucleosome positioning studies [38,39,40]. Interestingly, H1.2 and H1.5 variants are not distinguishable in this case. All other protein-binding regions, presented in this figure, have distinct xChIP-seq profiles for H1.2 and H1.5. In addition, H1 xxChIP-seq profiles around CTCF-binding sites show strong depletion compared to the H1 xChIP-seq profiles, suggesting that, in these localized fixed in situ chromatin regions, H1 epitopes are “hidden” due to stabilization of higher-order structure. A similar clear depletion of H1 xxChIP-seq signals, compared to H1 xChIP-seq signals, was observed for other chromatin-binding proteins (e.g., EGR1, GABPA, JMJD1C, Pol II and REST). In terms of the differences of the profile shapes between H1.2 and H1.5, two chromatin-binding proteins stand out: the subunits of cohesin SMC3 and STAG1. Their xxChIP-seq profiles show differences between H1.5 and H1.2 close to the center of the SMC3- and STAG1-binding sites, suggesting differential roles of these H1 variants in interactions with cohesin. For these selected chromatin protein-binding regions, formaldehyde fixation appears to make H1 epitopes (whose presence is demonstrated in the xChIP-seq) inaccessible to antibody in the xxChIP-seq assay. Another case of a very pronounced difference is observed around PU.1-binding sites. Our previous analysis showed that in HL-60/S4 cells PU.1 is associated with highly ordered nucleosome arrays with ~ 10 bp smaller nucleosome repeat length than genome-average [31]. A recent publication noted that PU.1 acts as a non-classical pioneer factor (not able to bind DNA in the nucleosome, but recruiting remodellers that redistribute nucleosomes) [41]. It seems that this nucleosomal organization exposes H1 epitopes in such a way that the xxChIP signal goes up.
“Open” chromatin regions have narrow xChIP and wide xxChIP depletion
Figure 6 presents average H1 epitope exposure around transcription start sites (TSS) for active and inactive genes. Active genes (Fig. 6a) show a significant difference between xChIP and xxChIP. The xChIP profiles contain a sharp and deep decline in apparent H1 occupancy (~ 300 bp wide), corresponding to the nucleosome-depleted region adjacent to the TSS. On the other hand, the xxChIP H1 epitope depletion extends to a ~ 10-fold longer region near the TSS, encroaching onto the gene body, suggesting an extended stabilized higher-order structure. In terms of the differences between H1.2 and H1.5, the epitope depletion of H1.5 is stronger, compared to H1.2. For inactive genes (Fig. 6b), H1 xChIP-seq profiles are essentially unchanged across the TSS regions. The “broad depletions” seen with both H1.2 and H1.5 xxChIP-seq suggest that higher-order chromatin structure is a common feature of TSS regions, regardless of transcriptional activity. It could be that the depletion of xxChIP signal at inactive promoters reflects the decrease of their in situ accessibility.
We then analyzed xChIP profiles around “open” chromatin regions in general. Figure 6c shows average xChIP and xxChIP profiles around DNase I-sensitive sites in HL-60 cells, which are consistently with Fig. 6a. The profiles around CpG islands (Fig. 6d) show the largest depletion of xxChIP, consistently with Fig. 4. Interestingly, xChIP H1.2 and H1.5 profiles around CpG islands are significantly different from each other, consistently with our finding that most regions with “swings” of H1.2 or H1.5 xChIP investigated in Fig. 3c and d are located inside promoters. Collectively, this analysis supports the concept that chromatin regions which are traditionally believed to be “open” generally possess chromatin higher-order structure, which when fixed with formaldehyde in vivo, results in decreased histone H1 epitope exposure.
Figure 7 presents average H1 epitope exposure profiles surrounding genomic regions enriched with different histone modifications. For verification, we have plotted profiles for H3K4me1, H3K27ac, H3K9ac and H3K36me3, using both the data that we reported for HL-60/S4 cells (25), as well as recent data for H3K4me1 and H3K27ac in HL-60 cells [32]. For H3K4me1 (mark of active enhancers), K3K27ac and H3K9ac (general activating marks), we observed strong depletions of H1 xxChIP epitope exposure, consistent with our previous analyses above. Interestingly, the profiles around centers of H3K36me3 domains (the mark of gene bodies of active genes) revealed less difference between xChIP and xxChIP epitope exposure, compared to the other shown histone modifications. Perhaps, this is because H3K36me3-enriched domains are wider than promoter/enhancer marks and less focused on their ChIP-seq peak summits. However, H3K36me3-domains did reveal a difference between H1.2 and H1.5 distribution, with enrichment of H1.2 and depletion of H1.5.
Influence of H1.2/H1.5 enrichment on the nucleosome repeat length (NRL)
NRL is defined as the average distance (bp) between the dyad axes of adjacent nucleosomes and is traditionally used as an integrative parameter characterizing local nucleosome packing. Previous publications suggest that NRL is different near binding sites of transcription factors [42] and affected by the presence of linker histones, although the role of different H1 variants is not clear [43]. Figure 8 presents normalized calculations of the NRL in regions enriched for the H1 variants, using the NucTools algorithm [27]. In the case of xxChIP, NRL was similar for H1.2 and H1.5 (191.8 bp and 188.8 bp, respectively). In the case of xChIP, the difference between H1.2 and H1.5 was slightly larger (190.6 bp and 184.5 bp, respectively). These measurements suggest that chromatin regions enriched with either H1.2 or H1.5 may have different arrangements of nucleosomes.
Interplay between linker histones and DNA methylation
To investigate the relationship between linker histones and DNA methylation, we performed whole genome bisulfite sequencing profiling in HL-60/S4 cells [30]. From our previous publications it is known that DNA methylation profiles around nucleosomes have well-defined patterns, which are significantly different depending on whether the nucleosome is located inside a CpG island or outside of CpG islands [38, 44]. Therefore, in the following analysis we take nucleosomes previously mapped using MNase-assisted H3 ChIP-seq in HL-60/S4 cells [31], and split them into two classes depending on their location inside or outside CpG islands. Furthermore, we narrow down this dataset to take into account only those nucleosomes which are located inside genomic locations enriched with one of four H1-related signals determined here (xChIP H1.2 and H.5 and xxChIP H1.2 and H1.5). Figure 9a, b demonstrates the average DNA methylation profiles calculated around the centers (dyads) of nucleosomes split into these 8 classes.
Figure 9a shows the DNA methylation profiles around nucleosomes inside CpG islands. These profiles are consistent with the idea that CpG islands are in general depleted of nucleosomes, but those few nucleosomes that appear in CpG islands are strongly associated with DNA methylation. These profiles are very different between xChIP and xxChIP, consistent with our previous calculations in Figs. 4 and 6d which show the largest differences between xChIP and xxChIP among all genomic features. DNA methylation profiles around nucleosomes outside of CpG islands are not so dramatically different between xChIP and xxChIP. Quantitatively, average DNA methylation profiles around all nucleosomes showed that DNA methylation was in general higher for xxChIP H1-enriched nucleosomes than xChIP H1-enriched nucleosomes. This can be explained by the increased CpG density in/near xxChIP DNA fragments, with both H1.2 and H1.5 xxChIP showing strong enrichment near CpGs (Fig. 9c). On the other hand, when we considered genome domains enriched with linker histones based on MUSIC peak calling, DNA methylation was depleted in the centers of the H1-enriched peaks and increased at a distance about 500 bp from the centers of the peaks (Fig. 9d); the latter effect was consistent with the GC content signatures of these peaks (Additional file 1: Figure S3). Thus, chromatin regions differentially enriched with H1.2/H1.5 in xChIP/xxChIP are characterized by distinct DNA methylation profiles which may reflect differences in nucleosome packing.