Abstract

Open and closed chromatin structures are linked to transcription, H2A.Z, and core promoter elements

Embryos were treated with formaldehyde, and H2A.Z nucleosome core particles were immunopurified (Figs. S2–S3). 652,738 H2A.Z-containing nucleosomes were sequenced (Fig. S4), and mapped to 207,025 consensus locations in the Drosophila r5.2 reference genome (Figs. 1a and S2b, see browser at http://atlas.bx.psu.edu/), thereby providing >3-fold depth of coverage (Fig. S5). Correction for micrococcal nuclease (MNase) digestion bias was imposed (Fig. S6). Those 112,750 nucleosomes detected three or more times were further analyzed, although patterns were identical when all nucleosomes were analyzed. The internal median error of the data was 4 bp (Fig. S7).

Open in a separate window

Figure 1

H2A.Z nucleosomal organization around the 5 ′ end of Drosophila genes

a, Browser shot of an arbitrary locus (nbs-defl). Bar graph represents the number of “W” or “C” (top and bottom traces, respectively) strand reads mapped to each coordinate. b, Composite distribution of H2A.Z nucleosomes relative to the TSS. Nearby genes were either included (gray) or eliminated (black) from the analysis, and normalized accordingly. The equivalent Saccharomyces profile is shown in green¹. c, Correlation of the number of H2A.Z nucleosomal sequencing reads (per gene) to mRNA levels in 8–12 hr embryos.

Fig. 1b displays the predominant embryonic distribution of H2A.Z nucleosomes relative to the transcription start site (TSS) of all coding genes, and is compared to the pattern previously derived from Saccharomyces¹. Patterns around noncoding genes are shown in Fig. S8. 11,994 of the 14,143 Drosophila coding genes (85%) contained at least one H2A.Z nucleosome (detected three or more times) within 1 kb of the TSS. H2A.Z levels correlated with gene expression (Figs. 1c and S9), as has been seen on individual genes and in Saccharomyces^12,13,17.

H2A.Z nucleosomes were predominantly distributed at 175 bp intervals from the TSS (compared to 165 bp in Saccharomyces¹, Fig. 1b), demonstrating that a predominant organizational pattern exists for H2A.Z nucleosomes in Drosophila embryos that transcends a spatial and temporal context. The H2A.Z pattern was compared to the distribution of bulk nucleosomes (i.e., those containing any combination of H2A.Z and H2A), determined using high density tiling arrays (36 bp probe spacing). Within genic regions the same organizational pattern was found (Fig. S10). For both datasets, a nucleosome-depleted region was evident immediately upstream of the +1 nucleosome, which likely reflects a nucleosome-free core promoter region (NFR), as first detected in Saccharomcyces⁷. Like Saccharomyces, a −1 nucleosome was detected ~180 bp upstream of the TSS. However, in contrast, it lacked H2A.Z.

Surprisingly, the genic array of Drosophila nucleosomes, started ~75 bp further downstream from the equivalent position in Saccharomyces, placing the +1 nucleosome at +135 (Figs. 1b and S10). This shift has important implications in how the TSS is presented to RNA polymerase II (Pol II). In Saccharomyces, the TSS resides within the nucleosome border potentially allowing the nucleosome to regulate start site selection and efficiency¹. In Drosophila, the predominant arrangement of nucleosomes might allow unimpeded access to the TSS with potential blockage occurring downstream after initiation.

Drosophila have well-defined core promoter elements such as TATA, Initiator, DPE, and MTE which bind to the general transcription machinery^18–22, although these elements are not found in most genes. For genes lacking these core promoter elements or having a DPE, the canonical nucleosome organization was observed (black pattern in Fig. S11), which was more robust when only H2A.Z containing nucleosomes were examined (blue pattern). In contrast, genes containing TATA, Inr, or MTE had a diminished canonical nucleosome organization and a diminished NFR, indicating that these classes of genes may have a more compact and gene-specific chromatin architecture, including a positioned nucleosome over the TSS. Consequently, they might be more dependent upon chromatin remodelling for expression. When genes become transcriptionally competent, resident nucleosomes could adopt a more open and canonical organization, which includes replacing H2A with H2A.Z. Three observations support this hypothesis. First, H2A.Z and bulk nucleosomes at highly expressed genes were more uniformly organized than those at lowly expressed genes (Fig. S9). Second, bulk nucleosomes for genes that contained H2A.Z at their 5′ end displayed the canonical pattern, while those lacking H2A.Z did not (Fig. S10, black plot vs red trace). Third, within any class of genes except those having an Initiator, H2A.Z nucleosomes adopted a more canonical organization than the bulk set of nucleosomes (Fig. S11). These results suggest that transcription and the presence of H2A.Z are linked to an open and uniform chromatin architecture at promoter regions.

Conserved DNA motifs and H2A.Z nucleosomes are organized around each other

Recent genome sequencing of 12 Drosophila species of differing evolutionary distance has provided an unprecedented opportunity to identify conserved DNA sequence motifs²³. In comparing the distribution of motifs around the TSS²³, we found four recurring patterns: 27 motifs were classified as “nucleosomal”, 57 as “anti-nucleosomal”, 12 as “fixed”, and 98 as “random” (left panels in Fig. 2a and Fig. S12). “Nucleosomal” and “anti-nucleosomal” patterns matched the general distribution of where nucleosomes were relatively enriched or depleted, respectively, relative to the TSS (see Fig. 1b). “Fixed” elements were at a defined distance from the TSS, and “random” elements lacked patterning. The “nucleosomal” and “anti-nucleosomal” patterns suggest that certain motifs are organized to be downstream of the TSS in the midst of nucleosomal arrays, while others are organized to be upstream of the TSS, where nucleosomes are relatively depleted.

Open in a separate window

Figure 2

Organization of conserved DNA motifs around TSSs (left) and nucleosomes (right)

a, Distribution of representative conserved DNA motifs from four distribution classes. b, Distribution of all motifs plotted with Treeview⁴⁶. Bin counts from all motifs in Fig. S12 were converted to fold deviations from the regional average (−/+ 1 kb), converted to a log₂ scale, and plotted. Red/black/green denotes above, near, or below average deviations, respectively.

We examined the organizational relationship of these DNA motifs to individual H2A.Z nucleosomes genome-wide (right panels of Fig. 2a and Fig. S12, and all motifs in Fig. 2b). Strikingly, “nucleosomal” motifs were consistently enriched on the H2A.Z nucleosome surface, whereas “anti-nucleosomal” motifs were consistently depleted. Individual “fixed” motifs were mostly depleted of H2A.Z nucleosomes. These findings along with several controls (Fig. S13) suggest that motifs and nucleosomes adopt a preferred organization around each other, regardless of their genomic location. This organization could be linked to co-evolution of base sequence composition bias in and around nucleosomes. The functional importance of such context remains to be determined.

Drosophila use a CC/GG patterning rather than AA/TT for demarcating nucleosome positions

We examined whether the positions of Drosophila H2A.Z nucleosomes are at least partly defined by the underlying DNA sequence pattern, and whether such pattern might be evolutionarily conserved. We determined the frequency of dinucleotides across Drosophila H2A.Z nucleosomal DNA since 10 bp periodic patterns of certain dinucleotides enhance the wrapping and positioning of DNA around the histone core (Figs. 3a and S14). As seen in Saccharomyces, 10 bp periodic patterns of A/T dinucleotides running counter-phase to G/C dinucleotide was observed. The modest amplitudes of the pattern suggest that such periodicities are infrequent, and thus used selectively (i.e., most nucleosomes lack underlying positioning signals).

Open in a separate window

Figure 3

Positioning properties of Drosophila nucleosomes and DNA

a, Composite distribution of WW and SS dinucleotides (as indicated) along the 147 bp axis of nucleosomal DNA (p-value = 0). The equivalent yeast profile¹ is shown in light shading in the background. b, Average correlation of all Drosophila promoter regions to nucleosome positioning sequence (NPS) patterns, comparing an AA/TT and a CC/GG pattern. The distribution of H2A.Z nucleosomes from Fig. 1b is shown as a gray backdrop. The span of the +1 nucleosome is indicated by the horizontal bar.

We further investigated the rules of nucleosome positioning by scanning promoter regions for correlations to nucleosome positioning sequences previously identified for a relatively small number of yeast or human nucleosomes²⁴, in which AA/TT (yeast²⁵ and worms²⁶) or CC/GG (human)²⁷ dinucleotides occur in a biased and/or periodic arrangement across nucleosomal DNA. Unlike in yeast, the AA/TT positioning pattern failed to identify nucleosome locations (Fig. 3b, black trace). However, the CC/GG pattern (Fig. S15) reproduced the exact position of the +1 nucleosomes (Fig. 3b, red trace), indicating that the Drosophila +1 nucleosome may be positioned in part by CC/GG-based positioning sequences that are utilized preferentially in metazoans. Consistent with this, +1 nucleosomes are highly positioned around the 5′ end of genes (Fig. S16).

Nucleosome-free regions reside at the end of active genes

Despite H2A.Z being enriched at the 5′ end of genes, substantial levels were detected throughout the genome, which allowed us to examine nucleosome organization at the 3′ end of genes (Figs. 4a and S17a). Strikingly, H2A.Z nucleosome levels spiked near the ORF end points then dropped precipitously further downstream into the intergenic regions, where transcripts terminate. The spike occurred ~30 bp upstream from the stop codon and ~160 bp upstream of the transcript polyA site. A similar nucleosome drop-off was seen when bulk nucleosomes were examined (Fig. S17b), but was not evident at genes that lacked H2A.Z. Thus, like the 5′ end, the presence of H2A.Z may be linked to a more open chromatin architecture at the 3′ end of genes. The change in nucleosome density coincided with alterations in nucleosome positioning sequences (Fig. 4b). Thus, such “3′-NFRs” might be defined in part by the underlying DNA sequence. Conceivably, 3′ NFRs might function in transcription termination.

Open in a separate window

Figure 4

H2A.Z nucleosomal organization around the 3′ end of Drosophila genes

a, Composite distribution of nucleosomes relative to ORF end points. Also shown is the distribution of transcript termination sites (polyA sites) in red. Nearby genes were either included (gray) or eliminated (black) from the analysis. b, Average correlation of all Drosophila gene terminal regions to AA/TT or CC/GG NPS patterns. The nucleosome profile from panel a is shown as a gray backdrop.

RNA polymerase II contacts the +1 nucleosome and pauses

The location of the +1 nucleosome at the 5′ end of genes is striking because its upstream border resides at approximately +62 (relative to the TSS), which is near where Pol II pauses during the transcription cycle^3,28–32. To examine the potential linkage between Pol II pausing and nucleosome positions, we first determined the genome-wide location of Pol II in embryos at 1,956 putatively paused genes (Fig. 5a). Pol II was concentrated in a ~300 bp region that peaked around +90, which overlaps the region bound by the +1 nucleosome, and is consistent with other recent placements^30–32. Indeed, the distribution of paused Pol II, as directly measured by permanganate reactivity of thymines on a statistically robust subset of ~50 genes (yellow trace in Figs. 5a and S18a), indicates that pausing occurs between +20 and +50 with the center at +35³⁰. This high resolution permanganate footprinting data, which represents the most definitive means of assessing Pol II pausing, places the front edge of Pol II (~16 bp downstream of the bubble³³) within ~10 bp of the +1 nucleosome border.

Open in a separate window

Figure 5

Distribution of Pol II and Pol II-engaged nucleosomes around the 5′ end of genes

a, Genome-wide location of “Paused” Pol II relative to the TSS. ChIP-chip genomic profiling of Pol II was conducted on Drosophila embryos. The black filled plot shows the distribution of Pol II at 1,956 “Paused” Pol II genes (see Methods). The yellow trace shows the distribution of permanganate-reactive thymines (an indicator of pausing) in 50 genes that undergo pausing³⁰. The distribution of Pol II at these 50 genes is similar to the bulk profile (Fig. S18a). b, H2A.Z nucleosomal distribution at 1,956 genes that contained “Paused” Pol II. The “Not paused” class represents those where Pol II is either absent or not paused. c, Distribution of Pol II-bound nucleosomes. Pol II ChIP was performed on bulk MNase-digested mono-nucleosomal DNA and hybridized to genome-wide tiling arrays. The three traces include distributions at 1,956 Pol II “Paused” genes, and control distributions either at 788 genes that have “No H2A.Z” within 1 kb of the TSS, or at 8,736 genes of the “Not paused” class.

The location of the +1 H2A.Z nucleosome was similar (but not identical) whether or not paused Pol II was present (Fig. 5b), indicating that Pol II was not likely to be the cause of the nucleosome shift compared to Saccharomyces. Rather, the positioned +1 nucleosome might be contributing to pol II pausing, which is consistent with other studies^34–37. Other factors including NELF are likely to make significant contributions to pausing as well^30,38,39.

Intriguingly, genes that contained a paused Pol II showed a ~10 bp downstream shift of H2A.Z nucleosomes (P-value = 10⁻⁹; Fig. 5b). The same shift was observed if H2A.Z sequencing reads (rather than nucleosomes) or bulk nucleosomes are plotted (Fig. S19a,b). The shift suggests that as part of the pausing process, Pol II collides with the +1 nucleosome, possibly displacing it downstream by one turn of the DNA helix. If the downstream nucleosomes are positioned in large part by the principles of statistical positioning^40,41, rather than the underlying DNA sequence, then a shift of the +1 nucleosome is expected to have a ripple effect on downstream nucleosomes.

To test the prediction that Pol II is engaging the +1 nucleosome, bulk mononucleosomes were prepared from formaldehyde crosslinked embryos and immunoprecipitated with antibodies directed against Pol II. DNA corresponding to mononucleosomes (~150 bp) was gel-purified and mapped to the entire Drosophila genome with high resolution tiling arrays. Fig. 5c (black trace) shows that the distribution of nucleosome-Pol II crosslinking at Pol II-paused genes peaked at the +1 nucleosome. This was not seen at genes lacking a paused Pol II or H2A.Z. The selective enrichment at +1 demonstrates that Pol II is predominantly engaged with the +1 nucleosome, and thus the +1 nucleosome may be instrumental in establishing the paused state.

Conclusions

The high resolution map of Drosophila nucleosomes reveals evolutionarily conserved and divergent principles of nucleosome organization. Genes that possess H2A.Z nucleosomes are likely to have experienced a transcription event. They tend to have nucleosome-free promoter and termination regions and intervening arrays of uniformly positioned nucleosomes that become less uniform towards the 3′ end of the gene. H2A.Z nucleosomes in general might not block assembly of the transcription machinery at transcriptionally “experienced” promoters. However, repressed promoters or those containing Initiator elements do appear to have an H2A nucleosome over the TSS.

Conserved DNA sequence motifs (and thus any proteins that bind to them) tend to have an organizational relationship with nucleosomes. “Anti-nucleosomal” motifs including those for proteins such as engrailed, even skipped, fushi tarzu, giant, hunchback, and knirps tend to be located upstream of the TSS and might contribute to the exclusion of nucleosomes over the core promoter. Indeed some have anti-nucleosomal activity^42,43. “Nucleosomal” motifs include sites for achaete, antennapedia, dorsal, tramtrack, and others. Their preference for locations downstream of the TSS where nucleosomes are well organized raises the possibility that they contribute to nucleosome organization.

In Saccharomyces, the location of the TSS just inside the +1 nucleosome border, allows the nucleosome to potentially exert control over initiation, whereas in Drosophila, the general case may be to position the +1 nucleosome to interact with a transcriptionally engaged paused polymerase. Whether the +1 nucleosome is causative or just participatory in the pausing is not known. It is now becoming clear that metazoans regulate transcription in large part through Pol II pausing rather than solely through transcription complex assembly^3,31,32,44. The nucleosome map and its context to DNA regulatory elements, presented here, provides a framework for designing experiments and analyzing existing data to understand how metazoans regulate transcription.

METHODS SUMMARY

D. melanogaster embryos (0–12 hr) were collected and crosslinked with formaldehyde. H2A.Z was immunoprecipitated from chromatin digested with MNase. Mononucleosomal DNA was gel-purified and sequenced using Roche GS20/FLX pyrosequencing technology^1,45. Chromatin from crosslinked embryos was also solubilized by sonication and/or MNase digestion, where indicated, and Pol II immunoprecipitated. Bulk nucleosomes were not immunoprecipitated. MNase-treated samples were gel-purified in the 75–200 bp range. DNA samples were then hybridized to Affymetrix Drosophila tiling microarrays (36 bp average probe spacing).

Supplementary Material

Acknowledgments

Footnotes

References