Report

A mechanism for preventing asymmetric histone segregation onto replicating DNA strands

See allHide authors and affiliations

Science  28 Sep 2018:
Vol. 361, Issue 6409, pp. 1386-1389
DOI: 10.1126/science.aat8849

How cells ensure symmetric inheritance

Parental histones with modifications are recycled to newly replicated DNA strands during genome replication, but do the two sister chromatids inherit modified histones equally? Yu et al. and Petryk et al. found in mouse and yeast, respectively, that modified histones are segregated to both DNA daughter strands in a largely symmetric manner (see the Perspective by Ahmad and Henikoff). However, the mechanisms ensuring this symmetric inheritance in yeast and mouse were different. Yeasts use subunits of DNA polymerase to prevent the lagging-strand bias of parental histones, whereas in mouse cells, the replicative helicase MCM2 counters the leading-strand bias.

Science, this issue p. 1386, p. 1389; see also p. 1311

Abstract

How parental histone (H3-H4)2 tetramers, the primary carriers of epigenetic modifications, are transferred onto leading and lagging strands of DNA replication forks for epigenetic inheritance remains elusive. Here we show that parental (H3-H4)2 tetramers are assembled into nucleosomes onto both leading and lagging strands, with a slight preference for lagging strands. The lagging-strand preference increases markedly in budding yeast cells lacking Dpb3 and Dpb4, two subunits of the leading strand DNA polymerase, Pol ε, owing to the impairment of parental (H3-H4)2 transfer to leading strands. Dpb3-Dpb4 binds H3-H4 in vitro and participates in the inheritance of heterochromatin. These results indicate that different proteins facilitate the transfer of parental (H3-H4)2 onto leading versus lagging strands and that Dbp3-Dpb4 plays an important role in this poorly understood process.

Posttranslational modifications (PTMs) on histones in eukaryotic chromatin have a profound impact on gene expression. Recently, it has been shown that at least some of these PTMs are traits that are inheritable during mitotic cell division and even through meiosis (14). As the “first” step of transmission of these PTMs, it was proposed that parental histone (H3-H4)2 tetramers, the primary carriers of epigenetic modifications, are randomly and equally distributed to leading and lagging strands of DNA replication forks (5, 6) and serve as the “template” for copying epigenetic modifications onto newly synthesized (H3-H4)2 tetramers, which do not mix with parental (H3-H4)2 (7) and have distinct PTMs from parental histones. This dogmatic view has not been tested in vivo owing to challenges in monitoring histone segregation onto replicating DNA strands. Moreover, the molecular mechanisms underlying the transfer of parental histone (H3-H4)2 tetramers onto replicating DNA remain largely unknown (8, 9).

We used the eSPAN (enrichment and sequencing of protein-associated nascent DNA) method, which can detect whether a protein is enriched on leading or lagging strands at a genome-wide scale (10), to monitor the segregation of newly synthesized and parental histone H3, which should represent (H3-H4)2 tetramers, onto replicating DNA strands (Fig. 1, A and B, and fig. S1A). Briefly, we released G1-arrested yeast cells into medium containing bromodeoxyuridine (BrdU) to label newly synthesized DNA and hydroxyurea (HU) to facilitate analysis (10, 11). Chromatin from G1 and early S phase cells was digested with micrococcal nuclease (MNase), which cleaves DNA between nucleosomes, and was used for deep sequencing (MNase-seq). The digested chromatin was also analyzed by chromatin immunoprecipitation (ChIP)—with antibodies (fig. S1B) against acetylation of histone H3 lysine 56 (H3K56ac), a mark of newly synthesized H3 (12), and trimethylation of histone H3 lysine 4 (H3K4me3, a surrogate mark for parental H3, see below)—and subsequent ChIP–strand-specific sequencing (ChIP-ssSeq) and eSPAN analysis.

Fig. 1 Newly synthesized (H3K56ac) and parental histone (H3K4me3) show a slight preference for leading and lagging strands, respectively.

(A and B) An outline of experimental procedures (A) and a diagram for the hypothetical eSPAN outcome assuming that parental and new (H3-H4)2 tetramers are equally distributed to leading and lagging strands (B). Black lines, parental DNA; red and green lines, newly synthesized Watson and Crick strands, respectively; ssDNA, single-stranded DNA; asterisks, BrdU. (C) Heatmap showing the bias ratio of H3K56ac eSPAN peaks at each of the 20 individual nucleosomes surrounding 134 early DNA replication origins. The individual nucleosome positions are numbered from −10 to +10 and are represented by circles. Each row represents the average log2 ratio Watson/Crick (W/C) of H3K56ac eSPAN sequence reads at one origin and is clustered on the basis of hierarchical clustering analysis. The blue dashed outline indicates separation of two groups of origins, with group 1 origins showing a more consistent bias pattern. (D) The average bias ratio of H3K56ac eSPAN peaks at each of the 20 individual nucleosomes of the 134 early replication origins (n = 3 independent repeats). (E and F) H3K4me3 eSPAN peaks at newly replicated chromatin exhibit a slight lagging-strand bias. Error bars in (D) and (F) indicate standard error of three repeats.

Analysis of MNase-seq data from G1 and early S phase chromatin revealed well-positioned nucleosomes surrounding early replication origins (fig. S2, A to C), consistent with published results (13). Moreover, H3K56ac ChIP-ssSeq peaks colocalized with BrdU-IP-ssSeq (BrdU immunoprecipitation and strand-specific sequencing) peaks (fig. S2, A, D, and E). We chose H3K4me3 as the surrogate mark for parental H3 for the following reasons: First, H3K4me3 was not detected on newly synthesized H3 (14). Second, we observed that the H3K4me3 level at newly replicated regions, but not its total level, was reduced compared to that observed during the G1 phase of the cell cycle (fig. S1, C and D), likely reflecting the dilution of parental H3K4me3 during the short time frame of our experiments. Finally, although at reduced occupancy, the positioning of H3K4me3-containing nucleosomes on newly replicated chromatin, as detected by H3K4me3 ChIP-ssSeq, was similar to that of G1 phase (fig. S2, F and G).

We calculated the log2 ratio of H3K56ac and H3K4me3 eSPAN sequence reads of the Watson (5′ → 3′) strand to those of the Crick (3′ → 5′) strand at 20 individual nucleosomes, which, on average, span the replicated region surrounding each of the 134 early replication origins. If H3K56ac and H3K4me3 were equally distributed onto leading and lagging strands, one would expect that the ratio would be close to zero (Fig. 1B). Instead, we observed that H3K56ac eSPAN peaks exhibited a small, but consistent, leading-strand bias at most nucleosomes except the −1 and +1 nucleosomes (Fig. 1, C and D, and fig. S3A), whereas H3K4me3 eSPAN peaks showed a lagging-strand bias (Fig. 1, E and F, and fig. S3B). These bias patterns were more consistent at group 1 origins than at less-efficient group 2 origins (Fig. 1, C and E, and fig. S3, C and D). We estimated that about 8% more H3K56ac-H4 tetramers were deposited onto leading strands than lagging strands, whereas at least 23% more parental H3K4me3-H4 tetramers were transferred to lagging strands than leading strands. The differential enrichment of parental and newly synthesized (H3-H4)2 tetramers at lagging and leading strands (23 versus 8%) observed here and in Figs. 2 and 3 likely suggests that nucleosomes that are formed from parental (H3-H4)2 tetramers are more stable and resistant to MNase digestion than those formed from newly synthesized tetramers. Thus, parental (H3-H4)2 tetramers are transferred onto both leading and lagging strands, with a slight preference for lagging strands.

Fig. 2 Analyzing segregation of newly synthesized and parental histone H3 in dpb3∆ cells with the RITE system.

(A) Schematic outline for marking the parental and newly synthesized H3 with the HA epitope (H3-HA) and T7 (H3-T7), respectively. The HA epitope was flanked by two LoxP sites, and upon induction of Cre recombinase, the HA epitope is replaced by the T7 tag. (B) Heatmap showing the H3-T7 eSPAN bias pattern in dpb3∆ cells at each of the 134 individual origins ranked, from top to bottom, on the basis of the replication efficiency. (C) The average bias pattern of H3-T7 eSPAN peaks at 134 early replication origins. (D and E) H3-HA eSPAN peaks in dpb3∆ cells show a strong lagging-strand bias. Error bars in (C) and (E) indicate standard error of two repeats.

Fig. 3 Deletion of DPB3 results in a marked increase in the bias ratio of H3K56ac and H3K4me3 eSPAN peaks under normal cell-cycle progression.

(A) A snapshot of BrdU-IP-ssSeq, H3K56ac, and H3K4me3 eSPAN peaks at individual nucleosomes surrounding origin ARS1309 (autonomously replicating sequence 1309) in dpb3∆ cells. (B to E) Analysis of the average bias pattern of H3K56ac eSPAN peaks (B) and (C) and H3K4me3 eSPAN peaks (D) and (E) in dpb3∆ cells, using early S phase cells without HU released from the G1 block at 25°C for 30 min (n = 2 independent repeats).

Unexpectedly, we discovered that the deletion of genes DPB3 and DPB4 (dpb3∆ and dpb4∆ cells), which encode two nonessential subunits of the leading-strand DNA polymerase, Pol ε (15), significantly increased the bias ratio of H3K56ac and H3K4me3 eSPAN peaks compared with wild-type cells (figs. S4 and S5) but had no apparent effect either on the levels of H3K56ac and H3K4me3 (fig. S1B) or on the overall nucleosome occupancy and positioning of G1 and early S phase chromatin (fig. S6). We estimated that ∼41% more H3K56ac-H4 was deposited onto leading strands than lagging strands, whereas 120% more parental H3K4me3-H4 was transferred to the lagging strand than to the leading strand in dpb3∆ and dpb4∆ cells.

We used the recombination-induced tag exchange (RITE) system (16), which marks newly synthesized and parental histone H3 with a T7 or hemagglutinin (HA) tag, respectively (Fig. 2A and fig. S7A), as an independent approach to analyze the impact of DPB3 and DPB4 deletion on histone segregation. The H3-T7 and H3K56ac eSPAN peaks showed a strong leading-strand bias (Fig. 2, B and C, and fig. S7, B and C) in dpb3∆ cells, whereas the H3-HA and H3K4me3 eSPAN peaks exhibited a lagging-strand bias (Fig. 2, D and E, and fig. S7, D and E). We also analyzed the distribution of H3K56ac and H3K4me3 at replicating chromatin in dpb3∆ cells during normal S phase without HU and observed the same effect (Fig. 3 and fig. S8). These results demonstrate that deletion of DPB3 and DPB4 alters the distribution pattern of new and parental (H3-H4)2 tetramers at replicating DNA strands.

To determine how depletion of DPB3 and DPB4 affects the distribution pattern of H3K4me3 and H3K56ac at replicating DNA, we calculated the relative amount of these histones at leading and lagging strands in dpb3∆ or dpb4∆ cells normalized to that in wild-type cells (Fig. 4A). Compared with that in wild-type cells, the amount of H3K4me3 at leading strands in dpb3∆ and dpb4∆ cells was reduced significantly (Fig. 4B and fig. S9A). By contrast, the amount of H3K4me3 at lagging strands in these two mutants increased slightly compared with that in wild type (Fig. 4B and fig. S9A). Moreover, deletion of DPB3 and DPB4 had minor effects on H3K56ac deposition onto replicating DNA strands compared with wild type (fig. S9, B and C). These results indicate that the transfer of H3K4me3 to leading strands is impaired the most in dpb3∆ and dpb4∆ cells (fig. S9, D and E). Thus, Dpb3 and Dpb4 facilitate the transfer of parental (H3-H4)2 onto leading strands. Supporting this idea, Dpb3 and Dpb4 were enriched at leading strands (Fig. 4, C and D), similar to the catalytic subunit of Pol ε, which is also enriched at leading strands (10). Moreover, in vitro, Dpb3-Dpb4 dimers, which are structurally similar to H2A-H2B (17) and copurify with H4 in cells (18), interacted with (H3-H4)2 tetramers (Fig. 4E). These results support the model that once nucleosomes ahead of DNA replication forks are disassembled, Dpb3-Dpb4 molecules serve as “receptors” or “chaperones” for parental (H3-H4)2 tetramers and facilitate their assembly onto leading strands (Fig. 4G).

Fig. 4 Deletion of DPB3 and DPB4 compromises the parental (H3-H4)2 transfer to leading strands.

(A) A diagram for calculating the relative amount of histone (H3-H4)2 at leading and lagging strands in dpb3∆ cells normalized to that in wild-type (WT) cells using the eSPAN datasets. Gray circles, parental (H3-H4)2; orange circles, new (H3-H4)2; black lines, parental DNA; red and green lines, newly synthesized Watson and Crick strands, respectively. (B) Top: The relative levels, coded by color, of H3K4me3 in dpb3∆ cells compared with those in wild-type cells at each of the 134 individual origins. Bottom: The log2 ratio of the average amount of H3K4me3 at lagging and leading strands of 134 replication origins in dpb3∆ cells divided by that in wild-type cells. (C) Dpb3 and Dpb4 eSPAN peaks at the ARS1309 origin. (D) The average bias of Dpb3 and Dpb4 eSPAN peaks at 134 early replication origins. The x axis is the distance, in bases, from the origin. (E) GST–Dpb3-Dpb4 pull-down assays show that Dpb3-Dpb4 dimers bind recombinant (H3-H4)2 in vitro. GST, glutathione S-transferase; His, histidine tag; yH3-H4, yeast histone H3-H4. (F) dpb3∆ and dpb4∆ cells exhibit an increased loss of silencing at the HML locus. NS, not significant. (G) A role for Dpb3-Dpb4 in the transfer of parental histone (H3-H4)2 onto leading strands. The “?” indicates an unknown protein complex mediating the transfer of parental H3-H4 to lagging strands.

Cells lacking Dpb3 or Dpb4 are defective in heterochromatin silencing (17, 19). Using an assay that monitors transient losses of heterochromatin silencing at the mating type loci (20), we observed a significant increase in the switching of the heterochromatin state at the HML locus in dpb3∆ and dpb4∆ cells (Fig. 4F), suggesting that a defect in parental (H3-H4)2 transfer in dpb3∆ and dpb4∆ compromises epigenetic inheritance.

Parental (H3-H4)2 tetramers are distributed onto both leading and lagging strands, with a slight preference for lagging strands. Dpb3-Dpb4 facilitates their assembly onto leading strands, thereby providing a mechanism for preventing a large asymmetric distribution of parental (H3-H4)2 at replicating DNA strands. Our results suggest that different proteins likely promote nucleosome assembly of parental (H3-H4)2 tetramers onto leading versus lagging strands. It is known that Mcm2, a subunit of the MCM helicase, contains a histone-binding motif that binds (H3-H4)2. Moreover, it was proposed that Mcm2 is involved in nucleosome assembly of parental (H3-H4)2 tetramers (21). In addition, DNA polymerase α and DNA polymerase clamp (proliferating cell nuclear antigen, or PCNA) also have chromatin-related functions (22, 23). It would be interesting to determine whether any of these DNA replication proteins have a direct role in reassembly of parental histones. The utilization of different proteins for the parental histone transfer also hints at a strategy used by eukaryotic cells to alter histone distribution patterns and possibly chromatin states in response to developmental stimuli, thereby endowing eukaryotic cells with the ability both to maintain epigenetic states and to generate epigenetic plasticity such as the generation of neuronal bilateral asymmetry in Caenorhabditis elegans during development (24).

Supplementary Materials

www.sciencemag.org/content/361/6409/1386/suppl/DC1

Materials and Methods

Figs. S1 to S9

Table S1

References (2539)

References and Notes

Acknowledgments: We thank S. Jia, G. Struhl, and B. Stillman for comments on this manuscript and F. van Leeuwen, H. D. Ulrich, and J. Rine for yeast strains. Funding: This study was supported by grant NIH R35GM118015 (Z.Z.); grants NSFC 31521002, MOST 2017YFA0103304, and CAS XDB08010100 (R.-M.X.); and Cancerfonden and the Swedish Research Council (E.J. and A.C.). Author contributions: C.Y., H.G., and Z.Z. conceived the project. C.Y., A.S.-C., L.Z., S.G., and S.S. performed experiments indicated online. H.G. performed the data analysis. Z.Z., R.-M.X., E.J., and A.C. supervised the corresponding study. C.Y., H.G., and Z.Z. wrote the manuscript with comments from all authors. Competing interests: The authors declare no conflicts of interest. Data and materials availability: The deep sequencing datasets have been deposited in the Gene Expression Omnibus (GEO) database (GSE112522).
View Abstract

Navigate This Article