SWR deposits H2A.Z predominantly at +1 nucleosomes but also at non-canonical sites

To evaluate SWR’s role in global H2A.Z deposition, we compared chromatin-bound H2A.Z levels in WT cells to those in swc2 mutants, where SWR is inactive for H2A.Z deposition [38]. Chromatin-bound H2A.Z levels were measured using the in vivo disulfide crosslinking method VivosX [39]. This technique utilized cysteine-modified HTZ1(T46C) and HTA1(N39C) as the sole sources of H2A.Z and H2A, respectively. Because H2A.Z is the only variant H2A in yeast and no other cysteines are present in core histones, the cysteine probes on H2A.Z and/or H2A, positioned at the interdisk interface of nucleosomes, are close enough for disulfide crosslinking (Fig 1A). Crosslinking was initiated with the cell-permeable, thiol-specific crosslinker 4,-dipyridyl disulfide (4-DPS). To facilitate resolution of crosslinking adducts, the HTZ1(T46C) gene was fused to a C-terminal 2xFLAG tag, and the HTA1(N39C) gene was fused to an N-terminal 2xV5 tag. As such, the crosslinked adducts of H2A-to-H2A.Z and H2A.Z-to-H2A.Z infer AZ and ZZ nucleosomes, respectively, while uncrosslinked H2A.Z infers non-nucleosomal H2A.Z (Fig 1B) [39]. Band intensities were normalized to histone H4, which serves as a loading control (Fig 1B and 1C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. SWR-dependent H2A.Z-containing nucleosomes.

(A) Structural representations of heterotypic H2A/H2A.Z nucleosome (left; PDB: 5B32) and homotypic H2A.Z nucleosome (right; PDB: 5B33) from humans. Pink residues correspond to T46 of yeast Htz1, and the blue residue corresponds to N39 of yeast Hta1. (B) VivosX analysis of WT and swc2 cells expressing HTZ1(T46C)-2FLAG and 2V5-HTA1(N39C) alleles. Total proteins extracts were analyzed by non-reducing SDS-PAGE and immunoblotting with anti-FLAG antibody (top) and anti-H4 antibody (bottom). Biological replicates (Rep) represent extracts from independent transformants. (C) Quantification of the immunoblots in B. Normalized signals represent band intensities of AZ, ZZ, and Z relative to H4. (D) H2A.Z ChIP-seq profiles for WT (pink) and swc2 (grey) cells. The SWR-dependent H2A.Z signal (blue) was derived by subtracting swc2 from WT. Grey blocks: annotated nucleosomes [40, 41]. Dark blue blocks: +1 nucleosomes. Light blue blocks: SWR-dependent H2A.Z nucleosomes (this study). Heatmaps: H3K9Ac and H4K12Ac signals [34]. Magenta lines: H2A.Z islands. (E) Scatter plot showing H2A.Z ChIP-seq read counts at annotated nucleosomes between biological replicates. Histograms depicting datapoint densities are shown above and to the right of the main plot. (F) Heatmaps of SWR-dependent H2A.Z signals (WT minus swc2) aligned at the center of +1 nucleosomes (left) or at the start and end of open reading frames (ORF) (right). (G) Relative number of identified H2A.Z-enriched sites in the indicated categories. The lower number of +1 nucleosomes in Fig 1G compared to Fig 1F is due to the exclusion of +1 sites lacking data or located in repetitive regions. (H) Relative number of NDR-distal H2A.Z sites enriched for H3K9 or H4K12 acetylation. The plot data for Fig 1C–1E are available in S2 Data. The plot data for the heatmaps in Fig 1F are available in S3 Data (left panel) and S4 Data (right panel). The genomic coordinates of SWR-dependent H2A.Z nucleosomes are provided in S1 Data.

https://doi.org/10.1371/journal.pbio.3003059.g001

In WT cells, we estimated that 12% of bulk H2A.Z is in the ZZ configuration, 57% in the AZ configuration, and 31% in non-nucleosomal forms (Fig 1B and 1C). In swc2, ZZ nucleosomes decreased to 2.5%, AZ nucleosomes decreased to 34%, while non-nucleosomal H2A.Z increased to 47% (Fig 1B and 1C). Similarly, deletion of the SWR1 gene, which encodes SWR’s ATPase motor, resulted in a comparable reduction in ZZ and AZ nucleosomes as observed in swc2 (S1 Fig A–B), confirming that Swc2, as well as Swr1, are required for the remodeling activity of the SWR complex [38]. Overall, these results indicate that SWR is primarily responsible for the formation of most ZZ nucleosomes. However, more than half of the AZ nucleosomes persisted in swc2, suggesting that the genomic distribution of H2A.Z in the absence of SWR activity cannot be ignored (Fig 1C).

To determine the genomic locations of H2A.Z nucleosomes in WT and swc2 cells, we isolated H2A.Z-containing nucleosomes and analyzed the associated nucleosomal DNA by sequencing as previously described [16]. Briefly, bulk chromatin was extracted from spheroplasted cells and fragmented under limited micrococcal nuclease (MNase) digestion (S1 Fig C). Chromatin particles were immunoprecipitated using anti-FLAG agarose to target the 2xFLAG epitope tag on the C-terminus of H2A.Z (S1 Fig D). The immunoprecipitated DNA fragments were sequenced on the Illumina platform and mapped to the yeast genome (version R64-1-1). The resulting H2A.Z enrichment signals of WT and swc2 were initially normalized to Reads Per Million (RPM) based on read counts. To account for the reduced levels of H2A.Z-containing nucleosomes in swc2 (55.8% of WT, Fig 1C), a scaling factor of 0.558 was applied to the swc2 H2A.Z profile (Fig 1D).

In swc2, the chromatin-bound H2A.Z was broadly distributed, with virtually no distinct H2A.Z peaks across the genome (Fig 1D and S2 Fig). This finding suggests that SWR is primarily responsible for site-specific H2A.Z deposition in yeast cells. The normalized H2A.Z counts at 59,039 annotated nucleosomes in non-repetitive regions from two biological replicates were plotted for WT (red) and swc2 (black) (Fig 1E) [40, 41]. The replicates showed strong correlations ( for WT; for swc2, and we identified 10,662 nucleosomal positions with significant levels of SWR-dependent H2A.Z (p<0.05), using the swc2 H2A.Z ChIP-seq data to model noise as the null hypothesis (Fig 1E, and S1 Data). Consistent with previous data, SWR deposits H2A.Z predominantly at +1 nucleosomes, with 3,921 of 4,725 +1 positions showing enrichment (Fig 1F and 1G). However, a substantial amount of H2A.Z was also deposited at non-(+1) positions in a SWR-dependent manner, which was represented by subtracting the swc2 profile from the WT profile (Fig 1D and 1F). Other NDR-proximal locations, such as +2 and -1 nucleosomes, accounted for 1,334 and 1,507 H2A.Z-containing nucleosomes, respectively. These findings indicate that 3,900 H2A.Z-enriched sites (i.e., 10,662 minus 3,921, 1,334 and 1,507) distal to NDRs (defined as not +1, -1 or +2) still require SWR for deposition (Fig 1G).

In addition, we identified clusters of SWR-dependent H2A.Z sites, referred to as H2A.Z islands. Using a criterion that defined an island as containing 6 H2A.Z-containing nucleosomes within a 1,500-bp window, we identified 264 H2A.Z islands across the yeast genome (Fig 1D, S3 Fig magenta bars and S1 Data). Many nucleosomes within these H2A.Z islands are not particularly enriched for histone acetylation (S3 Fig B–C). In fact, only one third of the NDR-distal H2A.Z sites across the genome are associated with H3 acetylation at lysine 9 (N = 1,510) or H4 acetylation at lysine 12 (N = 1,536) (Fig 1H). Therefore, SWR deposits H2A.Z at thousands of distinct H2A.Z sites by an unknown targeting mechanism that is not directly linked to NDRs or histone acetylation.

SWR preferentially inserts H2A.Z into a subset of nucleosomes endogenously enriched for H2A.Z

To uncover unknown determinants of SWR targeting, we employed an unbiased biochemical approach to interrogate the substrate specificity of SWR. This approach utilized a library of nucleosomes liberated from H2A.Z-deficient chromatin by MNase digestion. The use of canonical substrate was crucial, as nucleosomes pre-loaded with H2A.Z are poor substrates for SWR [26].

To prepare native canonical nucleosomes, we used an htz1swr1 strain carrying an episomal 2xV5-tagged HHF2 gene as the sole source of histone H4 (S4 Fig A). Logarithmically growing cells were spheroplasted under conditions optimized to minimize proteolytic clipping of histone tails [42]. Nucleosomes and poly-nucleosomes were liberated from the insoluble chromatin pellet by MNase digestion (S4 Fig B–C). The nucleoproteins were then purified by anti-V5 immunoprecipitation followed by V5 peptide elution. Nucleosomes were separated from poly-nucleosomes by sucrose gradient sedimentation before being dialyzed into a buffer compatible with the remodeling reactions (S4 Fig D). The native nucleosomal substrate, along with a recombinant nucleosomal control, were incubated with native SWR, ATP and Z-B dimers that were doubly tagged with biotin and FLAG (Fig 2A–2B). Both the native and recombinant nucleosomes were active substrates of SWR, as evidenced by the slower migration of remodeled nucleosomes in native polyacrylamide gel electrophoresis (PAGE), caused by the tags (primarily FLAG) on the Z-B dimer (Fig 2A and 2C) [28].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. SWR-mediated H2A.Z deposition reaction using native yeast nucleosomes.

(A) The histone exchange assay. Blue flag indicates the 3xFLAG tag on Htb1. The S-S-Biotin label indicates the cleavable biotin tag on Htz1(V126C). Red sectors: A-B dimers. Green sectors: Z-B dimers. Grey sectors: H3-H4 dimers. (B) Native SWR complex purified by the ASAP technique was separated on an 8-16% polyacrylamide gel and analyzed by SYPRO orange staining. (C) Histone exchange reactions analyzed by 6% polyacrylamide / 0.5x TBE electrophoresis. Recombinant nucleosomes were used in lanes 1 and 2, native nucleosomes in lanes 3 and 4. (D) Strategy for isolating SWR-remodeled nucleosomes. Input: nucleosomes partially remodeled by SWR in the presence of Z-B dimers that were biotinylated on Htz1 and FLAG-tagged on Htb1. (E) Recombinant nucleosomes and native nucleosomes partially remodeled by SWR (input) were pulled down with streptavidin-coated beads and eluted with DTT (eluate). FT fractions represent the unbound materials of the streptavidin pulldown. Cell equivalent amounts of input, FT, and eluate were analyzed by 6% PAGE and SYBR green staining. (F) Same as E except that the remodeling reactions were quenched by EDTA at the indicated times before subjected to streptavidin pulldown.

https://doi.org/10.1371/journal.pbio.3003059.g002

Nucleosomes remodeled by SWR, i.e., the AZ and ZZ species, were pulled down by streptavidin-coated paramagnetic beads and eluted with dithiothreitol (DTT), as the biotin moiety was linked to H2A.Z (Htz1) via a cleavable disulfide linkage (Fig 2D–2E). As a proof of concept, a partially remodeled reaction containing a mixture of recombinant AA, AZ and ZZ nucleosomes (input) was subjected to streptavidin pulldown and DTT elution (Fig 2E, lanes 1–3). The eluate fractions were enriched for the AZ and ZZ species, with undetectable levels of AA nucleosomes, indicating the specificity of the pulldown (Fig 2E). Similarly, native nucleosomes containing one or two copies of the dual-tagged Z-B dimers from the SWR-mediated reaction were pulled down by streptavidin beads and released into the eluate fraction after DTT treatment (Fig 2E, lanes 4–6). To isolate the subset of native nucleosomes preferentially remodeled by SWR, we performed time-course reactions. Aliquots from the remodeling reactions, which require as a cofactor, were quenched by the addition of ethylenediaminetetraacetic acid (EDTA) at 15, 30, and 45 min (Fig 2F). The native nucleosomes were prepared in replicates and reacted with SWR in two independent time-course experiments (Fig 2F and S5 Fig A). DNA from the eluate and flow-through (FT) fractions of the replicate reactions was purified and analyzed by paired-end sequencing (S5 Fig B).

The sequencing reads of the eluate and FT fractions were mapped to the yeast genome, and read coverages were normalized to read counts (S6 Fig). H2A.Z (Z)-enrichment was defined as the difference in read coverage between the eluate and FT fractions. Although SWR broadly inserted H2A.Z into nucleosomes across the genome, preferred nucleosomal sites were observed (S6 Fig). To minimize the influence of DNA linkers on SWR’s substrate selectivity, a read size filter (140-154 bp) was applied to the sequencing data. The reproducibility of the site-specific H2A.Z deposition by SWR was assessed by comparing Z-enrichment values between experimental replicates at previously annotated, non-repetitive nucleosomal sites (N = 59,039) (Fig 3A). Positive correlations were observed for Z-enrichment values between replicates at all three time points (: 0.49, 0.45 and 0.45 for 15, 30, and 45 min, respectively). The top 3% of Z-enriched nucleosomes (1,777) were defined as SWR-preferred sites (red) and the bottom 3% as unpreferred sites (blue) (Fig 3B). Importantly, we found the SWR-preferred sites were significantly enriched for H2A.Z in vivo (see Statistics section in Methods). At the 15-, 30-, and 45-min time points, 35%, 41%, and 45% of preferred nucleosomes (out of 1,777) were enriched for H2A.Z in vivo, respectively (Fig 3B and S1 Data). By contrast, randomly selected sites would be expected to overlap with native H2A.Z nucleosomes at a frequency of only 18% (i.e., 10,662/59,039) (Fig 1G).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Identification of SWR-preferred and unpreferred sites.

(A) Z-enrichment values at 59,039 non-repetitive genomic locations for the indicate time points. Red: top 3% of high Z-enrichment values. Blue: bottom 3%. (B) Pie charts showing the fraction of SWR-preferred sites (top 3%) that overlap with endogenous SWR-dependent H2A.Z sites (blue) versus unenriched sites (green). Black circles (not drawn to scale) represent nucleosomes at +1 positions. (C) Genomic distribution of SWR-preferred sites. Z-enrichment values (red heatmaps) plotted against the endogenous SWR-dependent H2A.Z (light blue), H3K9Ac and H4K12Ac (blue heatmaps). Two representative regions were shown. Annotated +1 nucleosomes are indicated. Magenta lines: H2A.Z islands. The plot data for Fig 3A and track information for Fig 3C are available in S5 Data. The genomic coordinates of SWR-preferred nucleosomes are provided in S1 Data.

https://doi.org/10.1371/journal.pbio.3003059.g003

To further understand how SWR selects its nucleosomal substrates in vitro, we plotted Z-enrichment values across the genome (Fig 3C and S7 Fig). Under our reaction conditions, SWR preferentially loaded H2A.Z into many, but not all, regions that are endogenously enriched for H2A.Z, including both NDR-proximal and distal sites, as well as H2A.Z islands. While a significant proportion of +1 nucleosome sites were identified (13.6%, 14.2%, and 16.2%) using the top 3% threshold, many endogenous H2A.Z +1 sites were missed (Fig 3B). To refine our prediction of SWR-preferred sites specifically at +1 nucleosomes, we performed unsupervised k-means clustering analysis (k = 3) using Z-enrichment values around +1 nucleosomes (S8 Fig). This analysis identified 1,718, 1,471, and 1,544 SWR-preferred +1 nucleosomes (out of 4,731) at 15, 30, and 45 minutes of reaction time, respectively (Fig 4A–4B and S1 Data). Of those, 2,401 were preferred by SWR at least once. Consistent with the genomic distribution of native H2A.Z, the +1 nucleosomes preferred by SWR (for all 3 time points) were generally enriched at intergenic regions but depleted across coding regions (Fig 4C–4D). However, SWR exhibited an artifactual preference for the 3’ end of genes not observed in vivo.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. SWR-preferred nucleosomes at +1 positions.

(A) Z-enrichment values of SWR-preferred +1 sites identified by k-means clustering analysis were aligned at +1 nucleosome dyads. Black vertical bar: +1 nucleosomes preferred by SWR at all three reaction time points. Gray bar: +1 nucleosomes preferred at two time points. Light gray bar: preferred at a single time point. (B) Line plots of the heatmaps in (A). Native H2A.Z represents endogenous SWR-dependent H2A.Z enrichment, derived from the data presented in Fig 1F. (C) Heatmaps of Z-enrichment values along genes with SWR-preferred +1 nucleosomes. Genes were scaled to equal length and aligned at their starts and ends. (D) Line plots of the heatmaps in (C) versus endogenous SWR-dependent H2A.Z (black). The plot data for Fig 4 are available in S6 Data. The genomic coordinates of the sorted +1 nucleosomes are provided in S1 Data.

https://doi.org/10.1371/journal.pbio.3003059.g004

SWR prefers Poly(dA:dT) tracts on the top strand of nucleosomal DNA entry/exit sites

To decipher the sequence composition of the nucleosomal substrates preferred by SWR, we performed dinucleotide motif analysis. Previous studies have showed that native nucleosomes in yeast are preferentially positioned on sequences with alternating dinucleotide patterns that are symmetrical around the nucleosomal dyad. This symmetry is associated with how the two halves of the nucleosomal DNA coil around the histone octamer [43]. We repeated this analysis by plotting the 16 dinucleotide frequencies of 67,538 nucleosome positioning sequences annotated by Brogaard et al. [43] (S9 Fig). For comparison, the average genomic frequencies of dinucleotides were also plotted. In general, dAdA and dTdT dinucleotides are enriched at the minor grooves of the DNA facing the histone octamer, whereas dSdS dinucleotides are out of phase. Other dinucleotides, such as dAdT, dTdA, dWdS, and dSdW, exhibited alternating patterns that are symmetrical around the dyad but their patterns are more nuanced (S9 Fig).

Having established the oscillating patterns of dinucleotide frequencies across positioned nucleosomes, we focused on SWR-preferred nucleosomes (present at two or more time points in Fig 3A, N = 1,069) that overlap the Brogaard annotation to uncover specific dinucleotide motifs (Fig 5 and S10 Fig). Our analysis showed that SWR-preferred nucleosomes (dotted lines) were enriched for dAdA dinucleotides on the top strand on the left half of the nucleosomal DNA relative to genome averages (solid color) (Fig 5A). A similar pattern was independently observed on the right half of the preferred nucleosomes, reflected by the enrichment of complementary dTdT dinucleotides on the top strand (Fig 5B). These findings indicate that the dAdA and dTdT enrichments in SWR-preferred nucleosomes exhibit a palindromic pattern relative to the nucleosomal dyad. Additionally, dTdA and dAdT dinucleotides were enriched, albeit to a lesser extent, in SWR-preferred nucleosomes (Fig 5C–5D). The increase in dWdW dinucleotide frequencies in SWR-preferred substrates was offset by a decrease in dCdC, dGdG, and dGdC frequencies (Fig 5E–5G). By contrast, other dinucleotide motifs exhibited more subtle differences in the preferred substrates (Fig 5H and S10 Fig). When the dinucleotide frequencies of the unpreferred substrate (N = 1,236) were compared to those of the averaged nucleosomes, the reverse trend was generally observed: dWdW dinucleotides were depleted, while dSdS were enriched in the unpreferred substrates (Fig 5 and S10 Fig, gray lines).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Dinucleotide motif analysis of SWR-preferred and unpreferred nucleosomes.

(A-H) Dinucleotide frequencies along positioned nucleosomes are shown as dotted lines for SWR-preferred nucleosomes, gray lines for unpreferred nucleosomes, and shaded areas for genome-wide averages. Preferred and unpreferred sites were defined based on the classification in Fig 3. In total, there are 1,069 SWR-preferred nucleosomes and 1,236 unpreferred nucleosomes, while genome averages are represented by 67,538 nucleosomes. See S1 Data for genomic coordinates and sequences. The plot data for Fig 5 are available in S7 Data.

https://doi.org/10.1371/journal.pbio.3003059.g005

To evaluate whether the preferred nucleosomes contain longer dA or dT tracts, we plotted the frequences of five consecutive dA () and dT () residues (S11 Fig). Consistent with the results of the dinucleotide motif analysis, the and frequencies exhibited a symmetrical dA:dT enrichment profile relative to nucleosome averages. To determine how many SWR-preferred nucleosomes contain poly(dA:dT) tracts, we found that 79% of the preferred nucleosomes contain one or more or , compared to 28% of the unpreferred nucleosomes. As a reference, 57% of nucleosomes across the genome contain one or more or . On average, SWR-preferred nucleosomes contain 3.2 or sites, while unpreferred nucleosomes contain 0.46 site, and genome-wide nucleosomes contain 1.5 sites. When uninterrupted or motifs were considered, 9.0% of preferred nucleosomes contain these sequences, compared to 0.16% of unpreferred nucleosomes and 2.3% of genome-wide nucleosomes. These analyses further support the conclusion that SWR preferentially remodels nucleosomes containing dA:dT tracts.

SWR-mediated H2A.Z insertion activity is stimulated by poly(dA:dT) tracts in nucleosomes

Given the enrichment of dAdA and dTdT dinucleotide motifs in SWR-preferred nucleosomes, we asked if the presence of poly(dA:dT) tracts within the nucleosomal sequence is sufficient to stimulate SWR’s H2A.Z insertion activity in vitro. Using the Widom 601 nucleosome as a control substrate, we systematically introduced 10 or 13 consecutive dA:dT residues (designated as or ) into one half of the nucleosome. In the Position (Pos) 1 mutant, a tract was placed at the DNA entry site [i.e., between the entry site and Super Helical Location (SHL) -6], Pos 2, 3, and 4 contained tracts located between SHL-6 and -5, SHL-5 and -4, and SHL-4 and -3, respectively (Fig 6A, red bars). To ensure that SWR acted exclusively on the nucleosome half containing the poly(dA) tracts, a 2-nt gap was introduced at SWR’s ATPase engagement site on the opposite half, thereby preventing histone exchange (Fig 6A and 6B, blue box) [28]. As a control, a reference substrate containing the canonical Widom sequence was included in all reactions (Fig 6C). Remodeling of both substrates were monitored by fluorescence densitometry on PAGE, with the reference and test substrates end-labeled with Alexa555 and Alexa647, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Effect of poly(dA) tracts on SWR activity using reconstituted nucleosomal substrates.

(A) Alignment of the Widom 601 control sequence versus the poly(dA)-containing variants. Red bars highlighted the poly(dA) tracts in positions (Pos) 1–4. Blue bars indicated the position of the 2-nt gap. Note that the poly(dA) tracts were added to the AT-rich side of the Widom 601 sequence. (B) The test substrate has an Alexa647 fluorophore at the 5’ end of the ‘top’ strand and a 2-nt gap at positions +19 to +20 nt (from the dyad) on the ‘bottom’ strand. (C) The reference substrate has the canonical Widom 601 sequence and an Alexa555 fluorophore at the 5’ end. All nucleosomal substrates are flanked by 7-bp linkers DNA on both sides. This design intentionally excludes long linker DNA, which is known to help recruit SWR, to isolate and reveal the contribution of the sequence effect on the H2A.Z deposition activity of SWR. (D) Fluorescent scans of the nucleosomal DNA. The expected positions for the AA, AZ and ZZ nucleosomes were indicated. An asterisk (*) indicates the expected position for hexasomes. The percentage of AA reacted is the difference in AA band intensities in the +ATP lane relative to the no ATP control. (E) Binding assay between SWR and nucleosomes with and without the dA₁₀ tract at Pos 3. Alexa 647-labeled nucleosomes (30 nM) were incubated with SWR at the indicated final concentrations for 30 min on ice before they were resolved on a 4% polyacrylamide gel containing 0.2xTB and 5% glycerol at 45V for 1.5 hr. The gel was scanned directly through the gel plates using a Typhoon 9500 imager. (F) Quantification of the gel shown in (E) and S12 Fig C. The line plot represents average values from the experimental replicates with error bars indicating the range of the data. The plot data for Fig 6F are available in S8 Data.

https://doi.org/10.1371/journal.pbio.3003059.g006

In the control reaction, where the test substrate lacked the poly(dA) tract but contained the 2-nt gap, SWR replaced the untagged nucleosomal A-B dimer on the intact side with a FLAG-tagged Z-B dimer, generating AZ nucleosomes that migrated more slowly than AA nucleosomes (Fig 6D, top). When was introduced at Pos 3, an increased production of AZ species was observed, indicating that the poly(dA) tract can indeed stimulate SWR’s H2A.Z insertion activity (Fig 6D and S12 Fig A–B, top). Notably, the stimulation was position-dependent: at Pos 2 stimulated SWR to a lesser extent than at Pos 3, whereas at Pos 1 and at Pos 4 inhibited SWR’s activity (Fig 6D and S12 Fig A–B, top). The stimulation observed in the Pos 3 substrate was not due to an artifactually increased SWR activity, as the reference substrate in the same reaction was not enhanced but instead diminished relative to the control (Fig 6D and S12 Fig A–B, bottom).

One explanation for the increased H2A.Z deposition activity in the Pos 3 nucleosome is that SWR binds more tightly to this substrate. To test this hypothesis, we performed a binding assay to compare SWR’s affinity for the Pos 3 nucleosomes versus the control, according to a previously described method [44]. Our results showed that SWR exhibited a higher affinity for the Pos 3 nucleosomes (Fig 6E–6F and S12 Fig C), providing a mechanistic explanation for the stimulatory effect of the DNA sequence on SWR activity.

The nucleosomes used in the remodeling assay above had short (7-bp) linkers on both sides. To assess whether the stimulatory effect of the poly(dA) tract is also observed in substrates that mimic the +1 nucleosome adjacent to the NDR, we utilized a canonical nucleosomal substrate with a 50-bp linker on one side and a 7-bp on the other (Fig 7A). Upon incubation with SWR and epitope-tagged Z-B dimers, the control AA nucleosomes were converted into heterotypic AZ and ZA species (Fig 7B–7C), which appeared as a doublet in the native gel before progressing to the ZZ product (Fig 7D–7E). Previously, site-directed hydroxyl radical footprinting showed that the AZ species, with the Z-B dimer positioned distal to the NDR, corresponds to the lower band of the doublet, while the ZA species, with the Z-B dimer proximal to the NDR, corresponds to the upper band [28].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. SWR remodeling assay for monitoring site-specific insertion of H2A.Z into a nucleosome.

(A) Schematic representation of the asymmetric nucleosomal substrates used in the assay. These substrates feature a a 50-bp linker on one side and a 7-bp linker on the other. Where indicated (dA₁₀), a 10-bp poly(dA:dT) tract was inserted between SHL-4 and -5 on the side closer to the long linker. Red: A-B dimers. Green: Z-B dimers, which contain a 3xFLAG tag on H2B (Htb1). Grey: H3-H4 dimers. Yellow: dA₁₀ tract. (B–C) The AZ nucleosome has the Z-B dimer inserted on the side distal to the long linker, whereas the ZA nucleosome has the Z-B dimer positioned proximal to the long linker. (D–E) Histone exchange reactions were allowed to proceed for 2 hours in D and for the indicated times in E. The resulting nucleosomes were resolved on a 6% polyacrylamide gel with 0.5x TBE at 110V for 2 hours, followed by SYBR Gold staining.

https://doi.org/10.1371/journal.pbio.3003059.g007

When a tract was introduced between SHL-4 and -5 (i.e., Pos 3), a more prominent upper ZA band in the AZ/ZA doublet was observed, indicating that SWR exhibited a stronger insertion preference toward the NDR-proximal side of the nucleosome containing the tract (Fig 7D–7E). Notably, this increase was accompanied by a reduced H2A.Z insertion on the NDR-distal site without the tract relative to the control (Fig 7D–7E).

Together, these findings highlight the role of DNA sequences in modulating SWR’s H2A.Z insertion activity and in guiding H2A.Z deposition across the genome.

A revised model of chromatin remodeling at yeast promoters

The formation of the nucleosome-depleted promoter platform and the closely spaced nucleosome arrays within gene bodies ensure the focused assembly of the transcription preinitiation complex and accurate start site selection. The site-specific incorporation of H2A.Z by SWR at the +1 nucleosome of many genes places the variant histone at a critical juncture, where RNA Pol II transitions from initiation to elongation, contributing to accurate transcriptional response. In this study, we uncovered an underappreciated role of DNA sequences in guiding H2A.Z deposition, giving new insights into how SWR targets its histone substrate at promoters and promoter-distal regions.

Yeast promoters are AT-rich, with poly(dT) and poly(dA) tracts predominantly occupying the sense strand at the upstream and downstream boundaries of the NDR, respectively (S13 Fig) [41]. DNA duplexes containing long poly(dA:dT) are poor substrates for nucleosome assembly as these homopolymeric sequences are rigid and intrinsically curled (i.e., bent and twisted) [45–48]. Thus, yeast promoter sequences are naturally depleted for nucleosomes [4]. However, DNA sequence alone is insufficient to keep ‘nascent’ nucleosomes away from promoters. Chromatin remodeling activities are also required to re-establish and maintain the native NDR observed in vivo [37]. The remodeler RSC, whose activity is enhanced by the poly(dA:dT) tracts, actively removes nascent nucleosomes from promoters, thereby widening the NDR (S13 Fig, step 2) [36, 49]. The INO80 remodeler in turn slides the +1 nucleosome into its native position, a process that is influenced by DNA sequences (S13 Fig, step 3) [37, 50]. In this study, we demonstrate that the remodeling activity of the SWR complex is also modulated by genomic sequences, with poly(dA:dT) tracts enhancing its activity. Together with the presence of NDRs and histone acetylation, DNA sequences contribute to the site-specific deposition of H2A.Z in vivo.

We propose that SWR may act upstream of RSC and INO80 to replace nucleosomal A-B dimers in nascent, promoter-proximal nucleosomes with Z-B dimers. The rationale is twofold. First, before a nascent nucleosome is slid into its native +1 position by RSC and INO80, its upstream edge is likely overlapping with the poly(dA:dT) sequence, hence placing the poly(dA) tract on the top strand of the nucleosomal DNA between SHL-4 and -6. This configuration makes the nascent nucleosome an ideal substrate for SWR-mediated histone exchange. Our data supports this model, as nucleosomes originating from the NDRs, albeit in low abundance, were preferentially remodeled by SWR (Fig 4B). Second, RSC preferentially disassembles and INO80 preferentially slides H2A.Z-containing nucleosomes over canonical nucleosomes [51, 52]. Therefore, it is energetically more efficient for SWR to deposit H2A.Z before RSC and INO80 establish the native promoter architecture (S13 Fig, step 1).

How poly(dA:dT) sequences facilitate H2A.Z insertion

At the mechanistic level, specific DNA sequences may introduce structural stress within the nucleosome, facilitating the unwrapping of nucleosomal DNA at the entry/exit sites. This structural perturbation could expand the nucleosome-free region, enhancing SWR binding, as SWR utilizes its Swc2 and Swc3 subunits to engage long stretches of exposed DNA [29]. Our in vitro binding experiment supports this model. Alternatively, but not exclusively, the rigidity of the poly(dA:dT) sequence may destabilize the nucleosomal substrate and facilitate the eviction step of the histone exchange reaction. As the Swr1 ATPase draws in the nucleosomal DNA at SHL-2 to disrupt histone-DNA contacts of the outgoing A-B dimer, the structural stress introduced by the poly(dA:dT) sequence could promote DNA unwrapping and enhance A-B dimer eviction. Why SWR prefers poly(dA:dT) tracts but not poly(dT:dA) tracts is intriguing. This suggests that poly(dA:dT) may curl the DNA in such a way that is more compatible with the transition state of the SWR-mediated reaction, whereas the curl caused by poly(dT:dA) is inverted and is less favorable.

Previously, we investigated SWR-mediated Z-B dimers deposition onto the opposite sides of an asymmetrically position nucleosome, reconstituted with the Widom sequence with a long linker (50 bp) at one end and a short linker (7 bp) at the other. Our findings revealed that Z-B dimer deposition efficiencies on the two sides depend on both reaction temperature and DNA sequences [28]. We identified a 16-bp region, spanning positions 24 to 39 bp relative to the nucleosomal dyad that, when swapped with the corresponding site on the opposite face of the nucleosome, is sufficient to bias Z-B dimer deposition towards one side at low temperatures (e.g., 4 and C). The favorable sequence (5’-TAGGG AGTAA TCCCC T-3’) contains two short poly(dG:dC) tracts, in contrast to the less favored sequence (5’-TCGTA GCAAG CTCTA G-3’), despite both having equivalent GC content. This observation leads to the speculation that SWR prefers poly(dG:dC) tracts. However, under our current experimental conditions at an ambient temperature of C, SWR did not exhibit a preference for native nucleosomes with high dCdC and dGdG contents. The discrepancy could be attributed to the temperature at which the current SWR assays were performed. Alternatively, the preferred 16-bp region, which overlaps the A-B/Z-B dimer binding site, may contain more nuanced motifs beyond poly(dG:dC) tracts that SWR favors at low temperatures. Thus, the sequence motifs favored by SWR likely have multiple permutations that can be influenced by specific reactions conditions, with poly(dA:dT) tracts identified here representing a key element that could combine with other motifs to fine tune SWR-mediated H2A.Z insertion in vivo.

When investigating substrate preference of SWR using the native nucleosome library, SWR failed to target many +1 nucleosomes that are typically enriched for H2A.Z (Fig 3B–3C and S7 Fig). This failure likely stems from the disruption of long linkers by MNase digestion, which are presumably more important for SWR recruitment than sequence motifs at these sites. Notably, these +1 nucleosomes are often associated with higher frequency of poly(dG:dC) tracts, which do not favor H2A.Z insertion under our reaction conditions [28] (Fig 5), further reinforcing the notion that a subset of endogenous H2A.Z sites do not rely on nucleosomal sequence motifs to facilitate H2A.Z insertion. Therefore, across different genomic locations, a complex interplay between nucleosomal linker DNA length, sequence motifs, and histone modifications orchestrates the site-specific deposition of H2A.Z in vivo.

Intrinsic bias of SWR against genic nucleosomes

SWR preferentially deposits H2A.Z into nucleosomes associated with intergenic over genic sequences. One explanation is that the information-rich genic sequences are depleted for the monopolymeric dA:dT sequences preferred by SWR. When genic regions were subjected to dinucleotide motif analysis, 3-bp periodicities were observed for all 16 dinucleotide motifs, reflecting the biased codon usage and codon-pair preferences of yeast genes (S14 Fig). Therefore, poly(dA:dT) sequences are selected against in the coding regions, rendering these regions unpreferred by SWR. An exception is genes encoding lysine-rich and phenylalanine-rich domains, which include consecutive AAA and TTT codons (S15 Fig). This provides a potential explanation for why some genes in the H2A.Z islands, such as PES4 (Fig 3C), exhibit non-canonical H2A.Z deposition in genic regions.

Finally, SWR exhibits a strong preference for nucleosomes at the 3’ end of genes, consistent with the presence of poly(dA) tracts (on the sense strand) associated with the transcriptional termination signal in yeast [53]. However, unlike the preferred H2A.Z insertion at the promoter regions, insertion at 3’ end is non-physiological, as native H2A.Z is generally enriched at the 5’ end but not the 3’ end of genes. How poly(dA:dT) sequences at terminator regions exclude SWR-mediated H2A.Z insertion in cells is unknown and may involve masking by termination factors in the regions [34].

In conclusion, SWR-mediated H2A.Z deposition in cells is influenced by multiple factors, including NDRs, histone acetylation, and, as revealed by this study, specific DNA sequences. This finding places SWR within a growing list of chromatin remodelers, including RSC, INO80, Chd1, whose remodeling activities are tuned by DNA sequences [36, 50, 54]. One emerging theme is that the genome encodes information to direct chromatin remodeling activities or that remodelers have evolved to recognize sequence signatures associated with distinct genomic regions, such as promoters.

This study also highlights the power of native yeast nucleosomes as substrates to interrogate chromatin remodeling events. When coupled with high-throughput sequencing, this biochemical approach provides deeper insights into how substrate diversity influences chromatin remodeling mechanisms.

Yeast strains

Yeast strains used in this study are listed in S1 Table. The WT strain (yEL379) in the VivosX analysis of Fig 1B–1C was previously described [39]. The swc2 strain (yEL575) was constructed by replacing the SWC2 ORF in yEL379 with the kanMX6 cassette using the homologous gene replacement method [55, 56]. The WT strain (yEL378) in the ChIP-seq analysis was previously reported [39]. The corresponding swc2 and swr1 variants (yEL786 and yEL1041, respectively) were similar constructed by homologous gene replacement.

The yeast strain (yEL905), which was used to prepare native nucleosome libraries, carried an episomal HHF2 gene with a 2xV5 tag at the N-terminus to facilitate nucleosome isolation. In addition, the strain lacked HTZ1 and SWR1 to remove endogenous H2A.Z. To construct yEL905, the precursor strain YYY67 was used [57]. YYY67 lacked the endogenous H3 and H4 genes, i.e., (hht1-hhf1)::LEU2 (hht2-hhf2)::HIS3 and was kept alive by an HHT1-HHF1 URA3 CEN ARS (pMS329) plasmid. A CEN ARS TRP1 plasmid (pEL628) containing the divergently oriented HHT1 and 2xV5-HHF2 genes was transformed into YYY67 and was used to replace pMS329 based on the plasmid shuffle technique [58]. The resulting strain (yEL704), which expressed the 2xV5-HHF2 gene as the sole source of H4, was subjected to a second round of transformation to replace HTZ1 with the hphMX6 cassette [55]. Finally, the strain was subjected to a third round of transformation to replace SWR1 with kanMX6. Successful removal of the HTZ1 and SWR1 genes was verified by colony PCR.

Plasmids

Plasmids used in this study are listed in S2 Table. To construct the plasmid pEL628, a synthetic gene fragment of 2xV5-HHF2 (synthesized by Twist Biosciences) was subcloned into a CEN ARS TRP1 HHT2-HHF2 plasmid (gift from Rolf Sternglanz [57]) at the HHF2 location by digestion with BamHI and NcoI followed by recombination using the Gibson Assembly protocol (New England Biolabs).

The plasmid for glucanase production (pT22-6HisLyticase) was constructed as follows. The sequence encoding amino acid residues 37-548 of glucanase (glucan endo-1,3-beta-glucosidase from Cellulosimicrobium cellulans, Uniprot: P22222) was synthesized by IDT and amplified by PCR using primer pT22(NdeI)Gluc-Gibs-F and pT22(XhoI)Gluc-Gibs-R (S3 and S4 Tables). The PCR product was cloned into the pET22b vector (Novagen) pre-digested with the NdeI and XhoI restriction enzymes (Thermo) using the Gibson Assembly protocol. Sequence integrity of all plasmids was confirmed by Sanger Sequencing (Azenta).

Proteins

Biochemical assays

Sequencing and informatics

Sequencing libraries were prepared using the NEBNext Ultra II DNA library prep kit according to the manufacturer protocol (NEB). Sequencing libraries were pooled and analyzed on a NextSeq 550 instrument using paired-end mode with 32 cycles on both ends. Sequencing data (FASTQ files) were aligned to the S. cerevisiae S288C genome (version R64-1-1) using Bowtie2 with the following options: –local, –very-sensitive-local, –no-unal, –no-mixed, –no-discordant, and –phred33. For H2A.Z ChIP-seq in Fig 1D–1F, S2 Fig, and S3 Fig, as well as the initial analysis of SWR preferred substrates in S6 Fig, a sizing filter of –I 120 –X 170 was used. However, SWR exhibited a preference for nucleosomes with longer linkers (S16 Fig), consistent with previous findings [30]. To minimize the contribution of linker length on SWR’s substrate preference in the in vitro assay, a more stringent filter (–I 140 –X 154) was used (S16 Fig, right). Note that the initial analysis in S6 Fig used the original filter (–I 120 –X 170 filter). Mapped reads were filtered using samtools-view with options –q 30 –F 0x4. Read coverage were calculated using bamCoverage in deepTools with options –binSize 1 –extendReads –effectiveGenomeSize 11756462 –scaleFactor X –blackListFileName Mask_v2.bed (which indicated repetitive sequences in the BED format) (S1 Data). Read coverages for H2A.Z ChIP-seq data were first normalized to Read Per Million (RPM) for both WT SWC2 and swc2 samples. To account for the reduced chromatin-bound H2A.Z levels in swc2 relative to WT, as determined by VivosX, the read coverage for swc2 was further scaled by a factor of 0.558. Read coverages for the eluate and FT fractions from the in vitro SWR reactions were normalized to RPM. Scaling factors (X) for all samples are listed in S1 Data. Quality control plots for nucleosome positions around +1 nucleosomes in WT, swc2, and htz1swr1 are shown in S17 Fig A and C, indicating no changes in nucleosome positioning. Similarly, nucleosome positioning in the eluate and FT fractions from the SWR preference assay were unchanged (S17 Fig B), indicating that SWR does not preferentially remodel mispositioned nucleosomes. Custom Python scripts were used for nucleotide motif counting and frequency calculation. The scripts, along with usage instructions, have been deposited in the Zenodo database. See the Data Availability section for details.

Statistics

To identify SWR-dependent H2A.Z nucleosomes, normalized H2A.Z read counts were computed at annotated nucleosomes within non-repetitive regions (N = 59,039) for both WT and swc2 samples. The difference in H2A.Z signals between WT and swc2 represents SWR-dependent H2A.Z. Using a predefined threshold of 10 (i.e., WT-swc2 10 for both replicates), we identified a total of 10,662 SWR-dependent nucleosomal H2A.Z sites. To assess the significance of H2A.Z signals at these sites, a one-sample Z-test was performed, positing the null hypothesis that these signals represent background noise. Utilizing the normally distributed H2A.Z ChIP-seq data from the swc2 strain as a noise model, Z-scores and the associated p-values were computed under the assumption of the null hypothesis. The maximum p-values observed among the 10,662 H2A.Z sites in WT replicate 1 was 0.015, and in WT replicate 2, it was 0.028. These outcomes led to the rejection of the null hypothesis, indicating that the 10,662 SWR-dependent H2A.Z sites exhibit a significant enrichment of H2A.Z.

To determine if our in vitro assay is better at identifying endogenous H2A.Z sites (alternative hypothesis) than randomly picking nucleosomes (null hypothesis), a one-sample proportion Z-test was performed. Using the 15-min time point as an example, the success rate in identifying H2A.Z nucleosomes in our assay is 35% () relative to 18% (p) in the population (N = 59,039). Given a sample size of 1,777, the Standard Error (SE) is 0.0091. The Z-score, calculated as z = ( - p) / SE, for our in vitro assay performance is therefore [(0.35 - 0.18) / 0.0091] 17.6. This results in a p-value of virtually zero, indicating that our in vitro assay (for the 15 min time point) is significantly better at identifying endogenous H2A.Z sites than random selection. Our 30-min and 45-min time points, with identification rates () of 41% and 45%, respectively, are therefore more significant than the 15-min time point at identifying H2A.Z sites.

To assess if the top 3% of SWR-preferred nucleosomes are more frequently associated with +1 nucleosomes than expected by chance, a one-sample proportion Z-test was performed. At the 15-min time point, 243 out of 1,777 SWR-preferred nucleosomes were associated with +1 nucleosomes, corresponding to a success rate of 13.7% (). This is higher than the expected random association rate of 8.02% (p) if +1 nucleosomes (4,731 out of 59,039) were picked at random. Using an SE of 0.0064, the resulting Z-score is 8.66 for the 15 min time point. Similarly, the Z-scores for the 30- and 45-min time points are 11.3 and 12.7, respectively, further supporting that SWR-preferred nucleosomes are significantly associated with +1 sites.