Report

Partitioning of Histone H3-H4 Tetramers During DNA Replication–Dependent Chromatin Assembly

See allHide authors and affiliations

Science  02 Apr 2010:
Vol. 328, Issue 5974, pp. 94-98
DOI: 10.1126/science.1178994

Abstract

Semiconservative DNA replication ensures the faithful duplication of genetic information during cell divisions. However, how epigenetic information carried by histone modifications propagates through mitotic divisions remains elusive. To address this question, the DNA replication–dependent nucleosome partition pattern must be clarified. Here, we report significant amounts of H3.3-H4 tetramers split in vivo, whereas most H3.1-H4 tetramers remained intact. Inhibiting DNA replication–dependent deposition greatly reduced the level of splitting events, which suggests that (i) the replication-independent H3.3 deposition pathway proceeds largely by cooperatively incorporating two new H3.3-H4 dimers and (ii) the majority of splitting events occurred during replication-dependent deposition. Our results support the idea that “silent” histone modifications within large heterochromatic regions are maintained by copying modifications from neighboring preexisting histones without the need for H3-H4 splitting events.

Histone and DNA modifications provide key epigenetic information (13). A newly synthesized DNA strand acquires its DNA methylation pattern by copying the preexisting DNA methylation signature from the template strand (1, 4, 5). However, the mechanism by which patterns of histone modifications are passed on to daughter cells through mitotic divisions remains enigmatic. To understand this, the DNA replication–dependent nucleosome partition pattern must be unveiled first. Initial studies indicated that the nucleosomes do not dissociate (6, 7), which was amended by the discoveries of “hybrid nucleosomes” that contain old H3-H4 tetramers and new H2A-H2B dimers or vice versa (811). Nevertheless, H3-H4 tetramers—the core particles of nucleosomes—do not dissociate during replication-dependent nucleosome assembly (1215). Because all six major lysine methylation sites are present on either H3 (Lys4/9/27/36/79) or H4 (Lys20), newly deposited nucleosomes may become methylated by “copying” the modification pattern from nearby parental nucleosomes (1). However, evidence that the H3-H4 tetramers may split emerged with the discoveries that H3-H4 histones deposit into chromatin as dimers rather than as tetramers (1618) and that the histone chaperone Asf1 is capable of disrupting H3-H4 tetramers to form H3-H4/Asf1 heterotrimers (19). Thus, the H3-H4 tetramer partitioning pattern needs a definitive reexamination (1). In addition, H3.3 variant histones do not form hybrid nucleosomes with canonical H3 histones in vivo (16), and they differ from canonical H3 histones for their chromatin localization, chaperon choice, deposition timing, posttranslational modifications, and functions (16, 2022), thus the partitioning pattern of H3.3–containing tetramers is also highly interesting.

We first established stable HeLa cell lines with N-terminally Flag-tagged histone H3.1 or H3.3 under the control of a tetracycline-inducible promoter. To differentiate the “new” histones from the “old” ones and to calculate their ratio, lysine-8 ([13C6, 15N2] heavy isotope–labeled L-lysine, abbreviated as K8 for its 8-dalton mass increase from normal lysine) was used in combination with a cell-cycle arrest reagent (nocodazole), thus specifically labeling the “newly synthesized” histones with K8 while leaving the old histones unlabeled (Fig. 1A). By timing the induction with tetracycline, Flag-H3 histones could be designated as old histones or new ones. In addition, we could also study the two major H3 variants, H3.1 and H3.3, individually. Mononucleosomes were prepared from cells with Flag-H3 incorporated into their chromatin (Fig. 1B and fig. S1C) and subjected to affinity purification with antibody to Flag, which selectively purified Flag-H3–containing mononucleosomes (Fig. 1B and fig. S1D). Flag-H3 histones were associated with native H3 and other core histones, as expected (fig. S1D). Flag-H3, copurified native H3, and other core histones were effectively separated by using 13% SDS–polyacrylamide gel electrophoresis (SDS-PAGE) (fig. S1D). Each histone band was excised individually and subjected to SILAC [stable isotope labeling with amino acids in cell culture (23)]–based quantitative mass spectrometry analysis. The percentage of new (K8) and old (K0) histones in each band was subsequently calculated (see the explanatory illustration in fig. S2).

Fig. 1

H3.1-H4 tetramers do not split. (A and B) Experimental schemes. (C) Summary of K8-labeling status of bulk and affinity-purified histones. (D) Representative mass spectra for peptides derived from bulk H3, H4, and affinity-purified Flag-H3.1, H3.1, and H4.

At 36 hours after cell-cycle release, all cells had gone through the first S phase, with a vast majority of the cells at either the first G2/M phase or the second G1 phase; at 72 hours, all cells had gone through two complete cell cycles, with some cells advancing through the third S phase. These observations were supported by the percentage of K8-labeled bulk histones (Fig. 1 and fig. S3) and with flow cytometry analysis (fig. S3).

After 36 hours of K8 labeling, bulk core histones were approximately half light (K0) and half heavy (K8) (Fig. 1, C and D, and fig. S3), which corresponds to one round of histone deposition. Affinity-purified Flag-H3.1 histones were only 1.0% K8 labeled (Fig. 1, C and D, and fig. S3), demonstrating that they indeed served as the old histones according to the experimental design. Copurified native H3.1 and H4 histones were 3.0% and 3.4% K8-labeled, respectively (Fig. 1, C and D, and fig. S3). Thus, we conclude that the vast majority of H3.1-H4 tetramers follow the nonsplitting model. In contrast, copurified H2A and H2B were close to 50% K8-labeled, which resembles the overall pattern in the bulk histone preparation (Fig. 1C and fig. S3), indicating the extensive exchange of H2A-H2B dimers among nucleosomes.

At 72 hours, Flag-H3.1 histones were 3.8% K8-labeled (fig. S3), indicating minor leaky expression. Nonetheless, their associated native H3.1 and H4 histones remained in similar K8-labeling ranges (6.3% for H3.1 and 5.9% for H4), whereas copurified H2A and H2B histones were close to their bulk counterparts (fig. S3). Taken together, our data clearly demonstrate old Flag-H3.1 histones stay with old H3.1 and H4 histones at the mononucleosome level.

In a second set of experiments, we generated newly synthesized Flag-H3.1 histones by altering the timing of induction (fig. S4). After one round of DNA synthesis, mononucleosomes were affinity-purified and subjected to quantitative mass spectrometry analysis. Flag-H3.1 histones were 93% K8-labeled and native H3.1 and H4 histones copurified with Flag-H3.1 were 90% and 92% K8-labeled, whereas bulk H3 and H4 histones were approximately 50% K8-labeled (fig. S4). In contrast, H2A and H2B histones copurified with newly synthesized Flag-H3.1 were approximately 50% K8-labeled, reflecting the level of labeling in bulk histones (fig. S4). These results indicate that newly synthesized Flag-H3.1 histones associate with newly synthesized native H3.1 and H4 histones, further supporting the H3.1-H4 tetramer nonsplitting model.

The experiments were then extended to the histone variant H3.3, which is known for marking active chromatin (20, 21). We repeated the “on” to “off” experiments for H3.1 (Fig. 1) using the Flag-H3.3 stable cell line. At 36 hours, Flag-H3.3 histones were only 0.3% K8-labeled, indicating almost no leaky expression. However, copurified native H3.3 histones were 6.4% K8-labeled, reflecting a significant level of splitting events (Fig. 2, A and B, and fig. S5). Moreover, at 72 hours Flag-H3.3 was 2.8% K8-labeled, but copurified H3.3 histones were 23% K8-labeled (Fig. 2, A and C, fig. S5). Thus, about one fifth of the Flag-3.3/H4 tetramers had split within roughly two cell cycles. Given that two histone H4 molecules exist in each tetramer, one co-deposited with Flag-H3.3 and the other co-deposited with native H3.3, the density of H4 should lie between Flag-H3.3 and the native H3.3, which is indeed the case at both time points (Fig. 2). The above experiments were repeated in a second Flag-H3.3 stable cell clone that expresses at least fivefold less Flag-H3.3 (fig. S1A) without cell synchronization, and similar results were obtained (fig. S6).

Fig. 2

Significant amounts of H3.3-H4 tetramers split. (A) Summary of K8-labeling status of bulk and affinity-purified histones. (B and C) Representative mass spectra for peptides derived from bulk H3, H4, affinity-purified Flag-H3.3, H3.3, and H4.

To further validate our conclusion, equal amounts of cells expressing Flag-H3.3 in regular medium were mixed with wild-type HeLa cells cultured in K8 medium. Mononucleosomes were then purified from the mixed cells and subjected to affinity purification with antibody to Flag and subsequent quantitative mass spectrometry analysis (Fig. 3A). In this control experiment, affinity-purified Flag-H3.3 histones showed no K8 labeling, and the copurified H3.3 and H4 histones also showed absolutely no K8 labeling. Even the relatively dynamic H2A and H2B histones were less than 1% K8-labeled (Fig. 3, B and D, and fig. S7), despite the existence of roughly 50% K8-labeled bulk histones in the starting material (Fig. 3, B and C, and fig. S7). These data clearly demonstrate the robustness of our assay system, thus ruling out the possibility that core histones might exchange among nucleosomes during the purification processes. Therefore, we conclude that the significant H3.3-H4 tetramer splitting events observed earlier (Fig. 2 and figs. S5 and S6) indeed occur in vivo.

Fig. 3

H3.3-H4 tetramer splitting events occurred in vivo. (A) Experimental schemes. (B) Summary of K8-labeling status of bulk and affinity-purified histones. (C and D) Representative mass spectra for peptides derived from bulk core histones and affinity-purified core histones.

Unlike canonical histones, which are deposited by the DNA replication–dependent pathway during S phase, H3.3 can also be deposited by a DNA replication–independent pathway (16, 20). To test whether a replication-independent pathway is fully responsible for the splitting events, we performed the splitting assay using cells treated with hydroxyurea (HU) or aphidicolin, two reagents that arrest cells at S phase. These experiments allowed us to specifically study the replication-independent deposition pathway. The three H3 variants H3.1, H3.2, and H3.3 can be discriminated by a single peptide after trypsin digestion (Fig. 4A). We took advantage of this property, and successfully achieved individual quantification of H3.1, H3.2, and H3.3 in bulk histone preparations. Seventy-two hours of treatment with 2 mM HU almost fully inhibited the incorporation of new H3.1 (Fig. 4B), demonstrating strong inhibition of DNA replication. In contrast, newly deposited H3.3 accounted for ~40% of the total H3.3 (Fig. 4B), indicating that the replication-independent H3.3 deposition pathway remained effective. In addition, the splitting events in HU-treated cells were significantly reduced from the untreated control cells (7.6% versus 20%) (Fig. 4C and fig. S8). In a separate set of experiments, cells treated with 5 μg/ml aphidicolin displayed full inhibition of new H3.1 deposition while allowing incorporation of 53% new H3.3 in the same cells (fig. S9). Aphidicolin-treated cells also displayed a significantly lower level of splitting events (2.5%) in comparison with that of their parallel untreated control cells (11%) (fig. S9). These results collectively suggest that (i) the replication-independent H3.3 deposition pathway proceeds largely by cooperatively incorporating two new H3.3-H4 dimers and (ii) the majority of splitting events occurred during replication-dependent deposition, although detectable amounts of splitting events were observed during replication-independent deposition.

Fig. 4

Inhibiting DNA replication greatly reduces the number of splitting events for H3.3-H4 tetramers. (A) H3 variants can be discriminated by a single tryptic peptide. Variant-specific amino acids are in red. The amino acid positions are indicated. (B) HU treatment strongly inhibits new H3.1 deposition while allowing replication-independent H3.3 deposition to occur. (C) Summary of K8-labeling status of bulk and affinity-purified histones from cells with or without HU treatment.

Our results support the idea that “silent” histone modifications within large heterochromatic regions are maintained by copying modifications from neighboring preexisting histones (1, 24) without the need for H3-H4 splitting events. However, mechanisms underlying the mitotic inheritance of “active” modifications remain debatable. Our observation that significant amounts of H3.3-H4 tetramers split during replication-dependent nucleosome assembly brings up an intriguing question: Do these tetramer splitting events occur at specific regions of chromatin for specific functions, such as mitotic inheritance (25, 26)? Although we observed significant splitting events only for H3.3–containing tetramers, it remains an open question whether such splitting events are variant-specific or rather chromatin region–specific. We did observe ~2% K8-labeling difference between Flag-H3.1 and copurified H3.1 in several experiments (Fig. 1 and figs. S3 and S4), which could be within our detection error but may also suggest splitting events for a small subset of H3.1–containing tetramers. One possible model is that the replication-dependent nucleosome assembly pathway differs at euchromatic and heterochromatic regions, resulting in specific splitting events, predominantly at euchromatic regions. This is particularly tempting because H3.3 is enriched in euchromatin (20, 21, 27), and H3.1 histones display a similar modification pattern in the vicinity of H3.3 histones (22). Detecting the “splitting hot spots” and unveiling their potential role in the mitotic inheritance of active modifications are interesting directions for future investigation.

Supporting Online Material

www.sciencemag.org/cgi/content/full/328/5974/94/DC1

Materials and Methods

Figs. S1 to S9

  • * These authors contributed equally to this work.

References and Notes

  1. We thank X. Wang from the Howard Hughes Medical Institute/University of Texas Southwestern Medical Center for critical comments on the manuscript. We thank N. Yang and J. Ni for color illustration. We thanks P. Mortensen from the University of Southern Denmark for developing and supporting the MSQuant software. This work was supported by the Chinese Ministry of Science and Technology 863 projects 2007AA02Z1A6 (to B. Z.) and 2007AA02Z1A3 (to S.C.).
View Abstract

Navigate This Article