Cryo-EM structures of methylated and unmethylated Sat2R-P human H2A.Z nucleosomes.

(A) Diagram of the palindromic Sat2R-P sequence. Methylated cytosines are labeled in green and underlined. Vertical bar denotes halfway point. (B and C) Atomic models of unmethylated (B) and methylated (C) Sat2R-P H2A.Z structures (the v1 model for both). Two face views (DF1 and DF2) and a side view (middle) are shown. (D) Left, overlay of unmethylated (UM) and methylated (Me) Sat2R-P DNA EM densities. Middle and right, overlays of DNA densities on top of the corresponding v1 DNA atomic models for either the unmethylated (middle) or methylated (right) structures. Arrows point to the visualizable end of the DF2 linker DNA for each structure. (E) RMSD analysis comparing differences between unmethylated and methylated v1 Sat2R-P H2A.Z atomic models.

DNA methylation reduces H3-DNA contacts, alters H4 N-terminal tail orientations, and forms open nucleosome structures.

(A) Density maps of unmethylated (UM) and methylated (Me) Sat2R-P H2A.Z nucleosomes filtered and resampled to ∼3Å. The face shown, and all following zoom-ins, is of DF2. Maps are overlaid onto each other. Boxed regions highlight major areas of difference shown in B-D. (B) Zoom-in of the H3 N-terminus on DF2 of both structures. Density maps are shown as a transparent overlay on top of the corresponding v1 atomic models. (C) Zoom-in of the DF2 acidic patch on both structures. Area corresponding to the extra density found in the Me structure labeled in red. (D) Zoom-in of the H4 N-terminal tail. Dotted lines show each structure’s preferred orientation (E) 3D Models used in the in silico mixing 3D classification analysis. Classes are as follows: (1 and 2) closed nucleosome structures with varying linker lengths; (3 and 6) open nucleosomes with highly flexible linkers; (4) open nucleosomes that have shifted position from center; (5) hexasomes. Both distal faces (DF1 and DF2) are shown of each model. (F) Results of classification analysis. Y-axis is the proportion of particles sorted into each class as a percent of the total input particles from each sample (UM or Me). Error bars represent SEM, n = 3 technical replicates.

DNA methylation increases H2A.Z nucleosome accessibility.

(A) Diagram of the 1HinfI_Sat2R DNA sequence used. CpG sites are shown in green and underlined. HinfI recognition site highlighted in magenta. A Cy5 fluorophore is attached to the 5’ end nearest the HinfI site. (B) Native gel showing HinfI digestion time course of UM and Me human H2A and H2A.Z nucleosomes. SYBR Safe was used to stain total DNA. Digestion efficiency was assessed via loss of Cy5 signal and a downward shift in the total DNA band. (C) Quantification of (B). Cy5 signals were normalized to total SYBR Safe signal for each sample. Error bars represent SEM. Each shape represents data from one independent experiment, n = 5 experiments. Statistics comparing unmethylated and methylated samples at indicated timepoints were completed with a two-tailed t-test assuming unequal variance with p-values listed above each comparison.

Presence or absence of DNA methylation influences H2A.Z deposition.

(A) Percentage of methylated CpG sites (≥ 5x coverage) associated with H2A.Z peaks or H3 reads in sperm pronuclei incubated in interphase egg extract or XTC-2 nuclei following CnT-BS library preparation. H3 data is taken from regions outside of H2A.Z enriched peaks. Data points represent biological replicates (n = 2). (B) Schematic of chromatinization assay. Magnetic beads coated with DNA of interest were incubated in Xenopus egg extract with the addition of CaCl2 to induce cycling into interphase. After 60 min, DNA beads were isolated to assess for histone composition. (C and E) Western blot results of chromatinization assay probing for either H2A.Z (C) or H2A.X-F (E). Ku70 and H4 signals are shown as loading controls. Beads coated with unmethylated (UM) or methylated (Me) 19-mer arrays of Widom 601 (19x601) or 16-mer arrays of HSat2 (16xHSat2) were used. A representative blot of technical triplicates for each condition is presented. (D and F) Quantification of (C and E). H2A.Z and H2A.X-F signals were normalized to H4 intensity. Error bars represent SEM. Data points represent technical triplicates across 3 biological replicates with each shape representing data from a single independent experiment (n = 9).

The SRCAP complex mediates H2A.Z’s preferential association with unmethylated DNA.

(A) Western blot to detect SRCAP in control IgG-depleted Xenopus egg extract (ΔIgG) or SRCAP-depleted extract (ΔSRCAP). Bottom panel shows ponceau staining as a loading control. (B) Western blots of chromatinization assay in IgG control vs SRCAP depleted egg extract. Magnetic beads coated with 19-mer arrays of Widom 601 (19x601) or 16-mer arrays of HSat2 (16xHSat2) were incubated with interphase egg extract for 60 min, before their isolation. (C) Quantification of H2A.Z signal normalized to H4 from (B). (D) Quantification of H2A.X-F signal normalized to H4 from (B). Error bars in (C) and (D) represent SEM from n = 3 biological replicates and each shape represents data from one independent experiment. (E) Western blot staining for SRCAP on 16xHSat2 DNA beads incubated in interphase egg extract. (F) Western blot staining for ZNHIT1 on specified DNA beads incubated in interphase egg extract. Representative image shown from two independent experiments.

DNA methylation prevents SRCAP-C mediated H2A.Z deposition and creates more open and accessible H2A.Z nucleosome structures.

Left, SRCAP-C is capable of depositing H2A.Z on unmethylated substrates, but presence of DNA methylation prevents binding and deposition activity. Another unidentified mechanism deposits H2A.Z in a DNA methylation insensitive manner. Right, DNA methylation decreases DNA wrapping stability, an effect felt greatest on the already open H2A.Z nucleosomes compared to the more tightly wrapped H2A nucleosomes, creating more open and accessible H2A.Z nucleosome structures.

DNA sequences/oligos used in the study.

Antibodies used in this study.

Workflow of Sat2R-P cryo-EM analysis.

(A) Native PAGE analysis of H2A.Z nucleosomes used for cryo-EM analysis. Bands visualized through SYBR Safe staining. (B) Diagram of structure analysis pipeline for either unmethylated (left) or methylated (right) H2A.Z SAT2R-P samples. Further details described in methods.

DNA atomic model generation for Sat2R-P H2A.Z nucleosome structures.

(A and C) Overlays of electron density maps for the DNA with both generated atomic models for either the unmethylated (A) or methylated (C) Sat2R-P H2A.Z nucleosome structure. (B and D) Zoom-in of dyad regions showing fitting of both DNA atomic models (v1 and v2) generated for either the unmethylated (B) or methylated (D) structure. Methylated cytosines are highlighted in green. Top panels are overlays of both models for the two structures. Positioning was determined by assessing densities around bases which retain purine/pyrimidine identity in both models vs those that switch. Bases that switched from purine to pyrimidine, and vice versa, showed more ambiguous densities than those that retained identity. Positioning was determined using the unmethylated structure and then used in modeling the methylated structure. (E) Comparison of predicted CpG positioning for the v1 (top)

Workflow of in silico mixing 3D classification analysis.

Particles from a merged batch of unmethylated and methylated Sat2R-P H2A.Z nucleosome micrographs were picked and fed through the standard analysis pipeline. Six 3D models were then generated representing various open, closed, or hexasome-containing states of the nucleosome. Merged particles were then sorted to each of the classes through CryoSPARC’s heterorefinement tool (n = 3 technical replicates).

Local resolution estimates for unmethylated and methylated Sat2R-P H2A.Z structures.

Left, local resolution estimates of DF1 for both unmethylated (top) and methylated (bottom) Sat2R-P density maps. Right, side view showing local resolution of the DNA backbone for both methylated and unmethylated density maps. Local resolution maps were generated in CryoSPARC using the local resolution estimation tool and overlaid on density maps filtered using CryoSPARC’s local filtering tool.

Cryo-EM structures of 601L H2A.Z nucleosomes show no DNA methylation dependent differences.

(A) Diagram of the palindromic 601L sequence. Bar indicates halfway point. Methylated cytosines are in green and underlined. (B and C) Final density maps of unmethylated (B) or methylated (C) H2A.Z nucleosomes. (D and E) Atomic models generated for (B and C). Methylated cytosines shown in green (F) RMSD analysis comparing differences between methylated and unmethylated 601L H2A.Z atomic models.

Workflow of 601L H2A.Z nucleosome cryo-EM structures.

(A) Native PAGE analysis of 601L H2A.Z nucleosome samples used for cryo-EM. Bands visualized through SYBR Safe staining. (B) Diagram showing analysis pipeline for solving 601L nucleosome structures. Further details described in methods.

Native PAGE and mass photometry analysis of nucleosomes used in HinfI digest assays.

(A) Native PAGE analysis of nucleosome samples. Bands visualized through SYBR Safe staining. (B and C) Mass photometry analysis of methylated or unmethylated H2A (B) or H2A.Z (C) nucleosomes. Peaks around 200-220 kDa represent the fully formed nucleosome population. (D) HinfI digest time course of bare unmethylated or methylated DNA.

Replicate gels of data from Main Figure 3.

Reproducibility comparison of CnT-BS samples

(A) Pearson correlation plot comparing fragment counts mapped to 1000 bp genomic bins across samples. Bins containing zero fragments were first filtered out. Fragment counts were then normalized using DESeq2 to account for sequencing depth differences and underwent a variance stabilizing transformation (VST) before the correlation matrix was generated. (C) Pearson correlation plot comparing CpG methylation across samples. Percent methylation at individual CpG sites was calculated for high coverage CpGs (≥ 10 reads) and normalized using methylKit. The correlation matrix was then generated using pairwise observations across all eight samples. Both correlation matrices were ordered, and rectangles drawn (n = 4, user-specified), via hierarchical clustering.

H2A.Z localizes to hypomethylated TSSs.

(A) Genomic annotation plots for the top quantile of 1000 bp genomic bins containing the most mapped reads averaged between replicates for each specified condition. (B and C) Heatmaps of either H3 CpG methylation calls averaged over 500 bp tiles (left-most panel) or H2A.Z reads (right panels) from two biological replicates across a representation of all annotated Xenopus laevis genes from either sperm (B) or XTC-2 (C) samples. Corresponding H3 and H2A.Z data are shown on the same sets of genes sorted in the same order. (D) IGV snapshot of a region on chromosome 1L showcasing preferential H2A.Z deposition at hypomethylated TSSs. Average H3 methylation tracks were generated across 500 bp tiles for visualization purposes on a scale of 0 to 100% CpG methylation. Grey indicates regions with no methylation data availability. Two biological replicates of H2A.Z and H3 tracks are shown.

Replicates of data from Main Figure 4.

Both H2A.Z variants preferentially associate with unmethylated DNA substrates.

(A) Expression check of exogenously added mRNA expression for ALFA-tagged versions of both H2A.Z.1 and H2A.Z.2. (B and C) Blots of DNA beads incubated with interphase extract expressing either ALFA-tagged H2A.Z.1 (B) and H2A.Z.2 (C). Presence of the specified H2A.Z variant on DNA beads was determined by blotting against the ALFA-tag with H4 shown as a loading control. (D) Quantification of (B and C). ALFA signals were normalized to H4 signals. Unique shapes denote results from three independent experiments (n = 3 biological replicates). Error bars represent SEM. (E) Additional replicate blots of data from (D).

Replicates of data from Main Figure 5.

(A) Replicate blots of Main Figure 5B-D. (B) Replicate blots of Main Figure 5E-F with Ku70 as a loading control.

Tip60-C does not display DNA methylation sensitivity.

Western blots of DNA bead pulldowns from interphase egg extract probing for endogenous p400 (ATPase for Tip60-C) along with respective loading controls.

Collection and model statistics for the Cryo-EM structures.

Alignment statistics of CnT-BS sequencing libraries.

XTC reads were downsampled to 30 million reads to match sperm samples, then all raw reads were trimmed and deduplicated using fastp v0.24.0 and aligned to the Xenopus laevis genome (Xenla 10.1) or lambda genome (to assess for bisulfite conversion efficiency) using Bismark v0.24.2. Methylated CpG percentages for H2A.Z samples were determined for the top 2.5 % of H2A.Z peaks by AUC (obtained using SEACR v1.3) while methylated percentages for H3 associated CpGs was conducted on regions outside of determined H2A.Z peaks. All CpGs met a cutoff of at least 5 reads to be counted. Methylated lambda CpG statistics were calculated from all mapped CpGs with no filtering.

Fragment statistics of filtered genomic bins used for sequencing analysis.

A count matrix was generated from processed bam files over 1,000 bp bins of the Xenopus laevis genome. Mitochondrial reads and bins containing fewer than 17 reads across less than 2 samples were removed to filter out low signal areas. Replicates for each sample were averaged and resulting count matrix used for genomic annotation analysis.