Recycling of modified H2A-H2B provides short-term memory of chromatin states

Abstract

Chromatin landscapes are disrupted during DNA replication and must be restored faithfully to maintain genome regulation and cell identity. The histone H3-H4 modification landscape is restored by parental histone recycling and modification of new histones. How DNA replication impacts on histone H2A-H2B is currently unknown. Here, we measure H2A-H2B modifications and H2A.Z during DNA replication and across the cell cycle using quantitative genomics. We show that H2AK119ub1, H2BK120ub1, and H2A.Z are recycled accurately during DNA replication. Modified H2A-H2B are segregated symmetrically to daughter strands via POLA1 on the lagging strand, but independent of H3-H4 recycling. Post-replication, H2A-H2B modification and variant landscapes are quickly restored, and H2AK119ub1 guides accurate restoration of H3K27me3. This work reveals epigenetic transmission of parental H2A-H2B during DNA replication and identifies cross talk between H3-H4 and H2A-H2B modifications in epigenome propagation. We propose that rapid short-term memory of recycled H2A-H2B modifications facilitates restoration of stable H3-H4 chromatin states.

Keywords: DNA replication; H2A; H2A.Z; H2B; chromatin; histone PTM cross talk; histone recycling; polycomb; post-translational modifications; ubiquitination.

Graphical abstract

Figure 1

H2A-H2B modifications in nascent chromatin mirror parental chromatin states (A) Scheme outlining the question addressed in this study. Do recycling of modified H2A-H2B contribute to transmission of chromatin states during replication? Flags represent post-translational modifications. (B) Overview of the ChOR-seq workflow. (C–F) (Left) Nascent chromatin (ChOR-seq, blue) and total chromatin (ChIP-seq, gray) occupancy of H3K27me3, H2AK119ub1, H2A.Z, and H2BK120ub1 (C–E). (Right) Heatmap and average profile across peaks of H3K27me3, H2AK119ub1, and H2A.Z (F). (Right) Metagene analysis across H2BK120ub1-decorated genes. TSS: transcription start site; TES: transcription end site. Signal is sorted according to total ChIP-seq intensity and quantified with reads per million (RPMs). Data are represented as average of two replicates. See also Figure S1.

Figure S1

H2A-H2B modifications in nascent chromatin mirror parental chromatin states, related to Figure 1 (A–C) Barplot of mapped reads (mouse) relative to spike-in reads (Drosophila) in noEdU or nascent samples for H2AK119ub1 (A, n = 4), H2A.Z (B, n = 2), and H2BK120ub1 (C, n = 2). (D) Western blot analysis of aniPOND signal with indicated antibodies. A representative image is shown. nE: no EdU control. (E) Quantification of aniPOND signal of nascent (T0) relative to no EdU samples for the indicated antibodies. Average of three replicates. (F) Pearson correlation plot of replicates and conditions of the marks examined. RPM-normalized values were used to probe correlation. (G) Heatmap and average profile across peak boundaries (H3K27me3, H2AK119ub1, and H2A.Z, left to right). RPM scale. (H) Heatmap and average profile across TSS for H2BK120ub1-decorated genes. RPM scale. (I and K) Barplot of mapped reads (human) relative to spike-in reads (Drosophila) in noEdU or nascent samples for H2A.Z and H2BK120ub1. (J and L) (Left) Nascent chromatin (ChOR-seq, blue) and total chromatin (ChIP-seq, gray) occupancy of H2BK120ub1 and H2A.Z. (Right) Metagene analysis across H2BK120ub1-decorated genes (J) and heatmap and average profile across peaks of H2A.Z, (L) RPM scale. (G–L) Data are represented as average of two replicates.

Figure 2

Parental H2AK119ub1 and H2BK120ub1 are recycled during DNA replication (A) Strategy to prevent de novo H2AK119ub1 (orange flag) by depletion of RING1A/B and co-depletion of BAP1 to prolong the half-life of H2AK119ub1. (B) Heatmap and average profile of RING1B occupancy across RING1B peaks on mature and nascent chromatin with or without RING1B/BAP1 depletion. Signal is quantified with reference-adjusted reads per million using exogenous spike-in chromatin (RRPM). (C) Heatmap and average profile across H2AK119ub1 peaks of nascent and total H2AK119ub1 occupancy in untreated (left) or dTAG/Auxin-treated cells (right). Signal is sorted according to total ChIP-seq intensity. RPM scale. (D) Signal across H2AK119ub1 peaks in total or nascent chromatin. Log₂(RRPM + 1) scale. Black line, median; dashed lines, 1.5× interquartile range. (E) Strategy to prevent de novo H2BK120ub1 (blue flag) by inhibiting transcription re-start using triptolide (TPL) with simultaneous EdU labeling (10 min). Thick arrows visualize eviction of RNAPII prior to replication and recruitment post-replication. (F) Signal of total H2BK120ub1 in 1 kb windows overlapping the TSS upon DMSO or TPL treatment. Log₂(RRPM + 1) scale. Black line, median; dashed lines, 1.5× interquartile range. (G) Heatmap and average profile of nascent and total H2BK120ub1 occupancy across H2BK120ub1-decorated genes (>50 kB) in untreated (left) or treated cells. Signal is sorted according to total ChIP-seq intensity. RPM scale. (H) Signal of H2BK120ub1 across genic (TSS-prox: 0–10 kb from TSS, TSS-dist: >50 kb from TSS) or intergenic regions in total or nascent chromatin. Log₂(RRPM + 1) scale. Black line, median; dashed lines, 1.5× interquartile range. Statistics by unpaired Wilcoxon test. Data are represented as average of three (C and D), or two (B and F–H) replicates. See also Figure S2.

Figure S2

Parental H2AK119ub1 and H2BK120ub1 are recycled during DNA replication, related to Figure 2 (A) BAP1/RING1B signal over time analyzed by western blot in single or double depletion time courses. Average of three independent replicates. Western blot was quantified using ImageJ. (B) Experimental setup. Depletion of BAP1 and RING1B was initiated by addition of Auxin/dTAG until completion, and thereafter, EdU was added to analyze nascent chromatin occupancy of residual H2AK119ub1. (C) Western blot Analysis with DMSO-treated dilution curve as reference. Note that the ladder (L; lane 4) has been removed from the Image. “D” refers to DMSO-treated controls, while “+” refers to Auxin/dTAG treatment for 80 min. (D) Quantification of (C), average of three independent replicates. (E) Heatmap and average profile of nascent and total H2AK119ub1 occupancy across H2AK119ub1 peak boundaries. RPM scale. (F) Barplot of mapped reads (mouse) relative to spike-in reads (Drosophila) in noEdU or nascent samples for H2AK119ub1 in cells treated with DMSO or Auxin/dTAG. (G) Barplot of mapped reads (mouse) relative to spike-in reads (Drosophila) for H2AK119ub1 in cells treated with DMSO or Auxin/dTAG. Normalized to input. (H) Barplot of ChOR-intensity normalized to ChIP-intensity in Auxin/dTAG-treated cells. Normalized to DMSO condition. (I) Barplot of mapped reads (mouse) relative to spike-in reads (Drosophila) for H2BK120ub1 in mES cells treated with DMSO or TPL (Triptolide). Normalized to input. (J) Barplot of ChOR-intensity normalized to ChIP-intensity in TPL-treated cells. Normalized to DMSO condition. (K) Barplot of mapped reads (human) relative to spike-in reads (Drosophila) for H2BK120ub1 in HCT116 cells treated with DMSO or TPL20 (20 min TPL treatment). Normalized to input. (L) Heatmap and average profile of nascent and total H2BK120ub1 occupancy across H2BK120ub1-decorated genes (>50 kB) in DMSO (left) or TPL20-treated (right) HCT116 cells. Signal is sorted according to total ChIP-seq intensity. RPM scale. (M) Occupancy tracks for H2BK120ub1 in DMSO (top) or TPL20-treated (bottom) cells. RPM scale. Data are represented as average of three (C–H) and two (I–N) replicates. Statistics by Student t test (two-tailed, unpaired).

Figure 3

H2A-H2B is recycled symmetrically during DNA replication independent of H3-H4 (A) Illustration of the question addressed. Is H2A-H2B segregated symmetrically to both daughter strands? (B) Average SCAR-seq profile, showing replication fork directionality (RFD, measured by OK-seq), and asymmetry (measured as partition to leading and lagging strand), for the indicated histone PTMs across all replication initiation zones with a H2AK119ub1 peak. (C and D) Average SCAR-seq profile for H3K27me3 or H2AK119ub1 in WT or POLE4KO cells. (E and F) Average SCAR-seq profile for H3K27me3 or H2AK119ub1 in WT or MCM2-2A cells. (G and H) Average SCAR-seq profile for H3K27me3 or H2AK119ub1 in WT or POLA1-3A cells. (B), (D), (F), and (H): Note the change in scale to visualize smaller biases. Individual replicates shown in Figure S3. Data in (C), (E), and (G) are represented as average of n = 2–4 replicates. N(IZ): number of initiation zones analyzed (STAR Methods). See also Figure S3.

Figure S3

H2A-H2B is recycled symmetrically during DNA replication independent of H3-H4, related to Figure 3 (A) Asymmetry plot for H2AK119ub1, H3K27me3, and H4K20me0 across H2AK119ub1 peaks (n = 6–8). Positive values indicate leading strand bias. Negative values indicate bias toward lagging strand. (B) Asymmetry plot for H2AK119ub1, H3K27me3, and H4K20me0 across H2AK119ub1 peaks (n = 6–8). Individual replicates. (C) Asymmetry plot for H3K27me3 across called H3K27me3 peaks in different cell lines (n = 3–6). (D) Asymmetry plot for H2AK119ub1 across H2AK119ub1 peaks in POLE4KO cells, individual replicates (n = 2 for WT, n = 4 for 2 POLE4KO clones). (E) Asymmetry plot for H2AK119ub1 across H2AK119ub1 peaks in MCM2-2A cells, individual replicates (n = 3 for WT and 2 MCM2-2A clones). (F–H) Average SCAR-seq profile for H4K20me0 in WT and POLE4KO (F), MCM2-2A (G), or POLA1-3A (H) cells. (I) Asymmetry plot for H2AK119ub1 across H2AK119ub1 peaks in POLA1-3A cells, individual replicates (n = 3 for WT, n = 4 for 2 POLA1-3A clones). (J) Asymmetry plot for H4K20me0 in different cell lines (n = 3–4). (K–M) Average xSCAR-seq profile for H2A.Z in WT and POLE4KO (K), MCM2-2A (L), or POLA1-3A (M) cells (n = 2). (N) Asymmetry plot for H2A.Z in different cell lines (n = 2). (D, E, I, and N) Statistics by unpaired Wilcoxon test.

Figure 4

H2A-H2B marks restore accurately and rapidly after DNA replication (A) Experimental outline of the quantitative ChOR-seq time course. (B–E) Average profile of RRPM-normalized occupancy signal for H2BK120ub1, H2A.Z, H2AK119ub1, and H3K27me3 signal across 3 kb centered on the TSS. Only TSSs occupied by the respective mark were included. Log₂ scale. (F) Average profile of RPM-normalized occupancy signal of nascent or total pan-histones (combined pan-H2A and pan-H3) across all TSSs (n = 30,025). Log₂ scale. (G) Restoration curve for relative abundance of H2BK120ub, H2A.Z, H2AK119ub1, and H3K27me3 post-replication. Data points in gray were excluded from regression analysis (STAR Methods). (H) Kinetic parameters for investigated marks. t(90% restored): relative time (in hours) needed to restore 90% of the total signal. %recycled: estimated abundance at nascent chromatin (T0) across peaks. Data are represented as average of two replicates. See also Figure S4.

Figure S4

H2A-H2B marks restore accurately and rapidly after DNA replication, related to Figure 4 (A–D) Average profile of RRPM-normalized occupancy signal across peak centers (top) or boundaries (bottom) for H2BK120ub1, H2A.Z, H2AK119ub1, and H3K27me3. (E) Average profile of RPM-normalized occupancy signal of nascent or total pan-H2A or pan-H3 across all TSS, (n = 30,025). Z score normalized. (F) Kinetic parameters for investigated marks. t(90% restored): relative time (in hours) needed to restore 90% of the total signal. % recycled: estimated abundance at nascent chromatin (T0) across the entire genome. Considering all regions may not represent as accurate parameters as in contrast to Figure 4H. Data are represented as average of two replicates.

Figure 5

H2A.Z and H2BK120ub1 restoration correlate with transcription (A) Outline of restoration categories for H2A.Z peaks. Peaks are classified as restored once their signal show no further increase in all subsequent timepoints (for example, R60: restored at 60 min) or classified as unstable if their signal is decreasing in subsequent time points (STAR Methods). (B) Restoration kinetics of peaks overlapping enhancers or promoters (within 1 kb). (C) Promoter signature (within 1 kb from TSS) for the different restoration categories for H2A.Z. Log₂(RPKM + 1) scale. Statistics by unpaired Wilcoxon test. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. H3K4me3 data from Sethi et al., RNAPII-Ser5p data from Stewart-Morgan et al. (D) Restoration curve for H2A.Z peaks at promoters stratified according to gene expression level quartiles. (E) As in (D) but for H2BK120ub1 peaks. No peaks overlapped promoters at low- or not-expressed genes. (F) Relative restoration categories according to distance of H2BK120ub1 peaks to TSS. Data are represented as average of two replicates. See also Figure S5.

Figure S5

H2A.Z and H2BK120ub1 restoration correlate with transcription, related to Figure 5 (A) Box plot showing distance to TSS for H2A.Z peak restoration categories. Log₁₀ scale. (B) Restoration lineplot for H2A.Z peaks mapping to promoters stratified according to overlap with H2BK120ub1 peaks. (C) Restoration lineplot for H2A.Z peaks mapping to promoters stratified according to overlap with H3K27me3 or H2AK119ub1 peaks. (D) Genomic occupancy of H2BK120ub1 across the time course. RRPM scale. Data are represented as average of two replicates.

Figure S6

Variant PRC1 sites show fastest H2AK119ub1 restoration, related to Figure 6 (A) Restoration curve for H2AK119ub1 peaks (parsed into 1 kb windows) or 1 kb windows not overlapping H2AK119ub1 peaks. All windows considered. (B) Restoration pie chart for H2AK119ub1 peaks (parsed into 1 kb windows) or 1 kb windows not overlapping H2AK119ub1 peaks. Only non-zero and non-decreasing windows considered. (C) Relative abundance of unstable H2AK119ub1 peaks stratified by their co-occupancy profile with PRC1 and PRC2 subunits. Unstable peaks defined as peaks where signal is decreasing in subsequent time points. (D) Restoration pie chart show restoration categories for high, medium, and low-intensity nascent H2AK119ub1 peaks. Nascent signals were stratified into deciles according to intensity (D1, D2…D10) and the high (D1), medium (D5), and low (D10) deciles are shown. Data are represented as average of two replicates.

Figure 6

Variant PRC1 sites show fastest H2AK119ub1 restoration (A) Outline of restoration categories for H2AK119ub1 peaks. Peaks were classified as in Figure 5A (STAR Methods). (B) Restoration line plot across selected genomic features indicating the cumulative fraction restored at given time points. Parsed peaks were stratified according overlap with CpG Islands (CGI), TSSs, peak centers, or peak boundaries. (C) As in (B) but across peak regions stratified according to overlap with H3K27me3 peaks. (D) As in (B) but across peak regions overlapping either vPRC1 (PGCF1/6, not PCGF2) or cPRC1 (PGCF2, not PCGF1/6). (E) RRPM-normalized H2AK119ub1 signal over time for parsed peaks stratified according to the different restoration categories. Log₂(RRPM + 1) scale. Statistics by unpaired Wilcoxon test. Data are represented as average of two replicates. See also Figure S6.

Figure S7

Cross talk with H2AK119ub1 drives H3K27me3 restoration, related to Figure 7 (A) (Top) ChOR-seq restoration categories for H3K27me3 in WT cells, classified identically as in Figure 5A (STAR Methods). (Bottom) Fraction of H3K27me3 peaks (parsed into 1 kB windows) overlapping H2AK119ub1ub peaks. (B) Western blot analysis of RING1B and H2AK119ub1 levels in DMSO or Auxin-treated cells (using H2A as a loading control). (C) Quantification of (B) relative to H2A. Average of two independent replicates. Quantification using ImageJ. (D) Barplot of mapped reads relative to spike-in for H3K27me3 in 2 h DMSO or Auxin-treated cells. (E) Average profile of RRPM-normalized occupancy signal for H3K27me3 across peak centers in DMSO or Auxin-treated cells. (F) Asymmetry plot showing the restoration of H3K27me3 asymmetry in DMSO-treated MCM2-2A cells over time in across H3K27me3 peaks stratified according to overlap with H2AK119ub1. (G) Asymmetry plot showing the restoration of H3K27me3 asymmetry in DMSO-treated MCM2-2A cells over time across all regions with significant signal (RPM > 0.6) but outside Polycomb domains (neither H2AK119ub1 nor H3K27me3 peaks). (H) Heatmap and average profile of total ChIP JARID2 occupancy across peak centers in genome-wide chromatin in 2.5 h DMSO or Auxin-treated cells. RRPM scale. (I) Asymmetry plot showing the restoration of H2AK119ub1 asymmetry in WT or MCM2-2A cells over time across all H2AK119ub1 peaks stratified according to overlap with H3K27me3 peaks. (F, G, and I) Statistics by unpaired Wilcoxon test. Data are represented as average of three replicates. (D) Statistics by Student t test (unpaired, two replicates).

Figure 7

Cross talk with H2AK119ub1 drives H3K27me3 restoration (A) Illustration of the questions addressed. Does H2AK119ub1 stimulate H3K27me3 restoration on the lagging strand? Does H3K27me3 increase H2AK119ub1 deposition on the leading strand? (B) Experimental outline to assess H2AK119ub1-dependence of H3K27me3. AID-RING1B RING1A^−/− MCM2-2A cells were treated for 2 h with Auxin or DMSO prior to 10 min EdU labeling and indicated chase time course. (C) Asymmetry plot showing restoration of H3K27me3 asymmetry across H3K27me3 peaks in DMSO and Auxin-treated cells over time. Positive values indicate leading strand bias. Negative values indicate bias toward lagging strand. (D) Asymmetry plot showing restoration of H3K27me3 asymmetry stratified by H2AK119ub1 co-occupancy in Auxin-treated cells over time. (E) JARID2 occupancy across JARID2 peaks on nascent chromatin in untreated and Auxin-treated cells. RRPM scale. (F) Average SCAR-seq profile of JARID2 in untreated or Auxin-treated cells. (G) Asymmetry plot showing restoration of H2AK119ub1 asymmetry across H2AK119ub1 peaks in WT or MCM2-2A cells over time. (H) As in (E) but focusing on H2AK119ub1 peaks overlapping cPRC1 sites (RING1B, PCGF2). (I) As in (E) but focusing on H2AK119ub1 peaks overlapping vPRC1 sites (RING1B, no PCGF2). (C, D, and G–I) Statistics by unpaired Wilcoxon test. Data are represented as average of three replicates. (J) Model. (Left) Recycling of H2A-H2B at replication forks transmits modifications on parental H2A-H2B to nascent chromatin. Parental H2A-H2B dimers are recycled to both daughter strands independent of H3-H4 tetramers through a pathway involving POLA1 on the lagging strand. Upon reincorporation on daughter strands, parental-modified H2A-H2B can form nucleosomes with new and parental H3-H4 tetramers and new H2A-H2B dimers. Fast restoration kinetics of H2A-H2B modification together with intra- and inter-nucleosomal PTM cross talk then contribute to establishment of modifications on new H3-H4 and H2A-H2B and maintenance of chromatin states across replication. (Right) We propose that recycled parental H2A-H2B PTMs guide the restoration of newly deposited H3-H4 and H2A-H2B in nascent chromatin. Vice versa, stable H3-H4 PTMs reinforce H2A-H2B modifications in mature chromatin responding to dynamic exchange of H2A-H2B. Replisome components involved in histone recycling are colored in blue for H3-H4 and yellow for H2A-H2B. Yellow arrows depict the path of H2A-H2B at the replication fork, while black arrows indicate positive feedback. See also Figure S7.