Introduction

The current definition of “epigenetics” is “the inheritance of variation (-genetics) above and beyond (epi-) changes in the DNA sequence” 1. In eukaryotic cells, the basic unit of chromatin is the nucleosome, composed of 147 bp of DNA wrapped around the core histone octamer 2. Histones contain extensive post-transcriptional modifications for various functions (reviewed in this issue, 3). Histone modifications are known to carry important epigenetic information 3, 4, especially after several histone modifications were shown to be critical to many well-known epigenetic phenomena, including position effect variegation 5, 6, 7, Polycomb silencing 8, 9, 10, 11 and dosage compensation 12, 13. However, how newly deposited histones acquire these modifications during/after DNA replication remains enigmatic. In fact, it is not even clear whether any of these modifications are reestablished on newly deposited histones in a manner precise enough to assure mitotic inheritance of histone modification-based epigenetic information. In this review, we focus on insights obtained from recent progress that lead us to the viewpoint that histone modification-based epigenetic information may not necessarily be precisely inherited during the mitotic divisions. Instead, histone modifications affecting epigenetics may be reestablished in a much more flexible way to maintain distinct transcription states.

Semi-conservative partition, followed by templated modification copying events: an imperfect “ideal” model

In addition to histone modification-based epigenetic inheritance, there are two well-studied biological phenomena in which information is transmitted across cell division: the inheritance of genetic information through the DNA sequence and the inheritance of epigenetic information encoded by CpG methylation. Semi-conservative DNA replication and sister chromatid segregation ensure the faithful duplication and partition of the genetic information. Similarly, symmetric CpG methylations are segregated in a semi-conservative way and faithfully reestablished on the newly synthesized DNA strand by a templated copying event (Figure 1A) 14, 15, 16. Therefore, semi-conservative partitioning of the histone H3-H4 tetramers followed by templated modification copying events provided a compelling model for the inheritance of histone modifications during mitotic divisions (Figure 1B). This model relies on two critical assumptions: (1) inheritable histone modifications are obligated to exist in a symmetric manner within mono-nucleosomes; (2) histone H3-H4 tetramers undergo semi-conservative segregation during DNA replication-dependent chromatin assembly.

Figure 1
figure 1

Semi-conservative partition, followed by templated modification copying events: an imperfect “ideal” model. (A) Model illustration for the inheritance mechanism of DNA methylation. (B) The “ideal” but recently proven incorrect model for the inheritance of histone modifications.

Every nucleosome contains two copies of each core histone. However, it is not known whether histone modifications exist in a symmetric manner within mono-nucleosomes, mainly because of technical difficulties in discriminating symmetrically and asymmetrically modified nucleosomes.

Another key pillar for this precise inheritance model depicted in Figure 1B is the semi-conservative partition of the H3-H4 tetramers during DNA replication-dependent chromatin assembly. Historically, this model about “two symmetrically paired half-nucleosomes” was proposed in 1976 to explain “potential direct inheritance of information in the form of histone structure” 17. However, newly synthesized and parental chromatins exhibit comparable micrococal nuclease digestion kinetics, indicating that newly synthesized chromatin contains full nucleosomes at regular spacing intervals 18. Sedimentation studies with heavy isotope-labeled histones indicated that histone octamers segregate in a conserved manner 19. These findings led to the nucleosome conservative segregation model. Although “hybrid nucleosomes” containing old H3-H4 tetramers and new H2A-H2B dimers, or vice versa, were discovered later 20, 21, 22, 23, H3-H4 tetramers, the core particles of the nucleosomes, which carry the modifications most likely to be involved in epigenetic phenomena, do not dissociate during replication-dependent nucleosome assembly 24, 25, 26, 27. However, expectations for the H3-H4 tetramer split model revived in the past few years, after the discoveries that H3-H4 histones deposit onto chromatin as dimers rather than tetramers 28, 29, 30, and that the histone chaperone Asf1 blocks H3-H4 tetramer formation 31, 32 and even disrupts pre-assembled H3-H4 tetramers to form H3-H4/Asf1 heterotrimers 32. Although without direct evidence, these studies certainly offered a possibility for the H3-H4 tetramer splitting events. Hence, debates and discussions about the H3-H4 tetramer semi-conservative partition model arose in many recent reviews 33, 34, 35, 36, 37, 38, 39. Recently, with selective purification of “old” or “new” H3 histone-containing mono-nucleosomes and stable isotope labeling-based quantitative mass spectrometry analysis, the H3-H4 tetramer partition pattern was reinvestigated in a high-resolution, variant-specific way 40. This study provided strong evidence that the majority of H3-H4 tetramers segregate in a conservative manner. Although splitting events of a fraction of H3.3-H4 tetramers were detected 40, it is unlikely that these events are responsible for epigenetic inheritance mechanisms involving templated modification copying within the same mono-nucleosomes, especially for repressive modifications given that H3.3 is associated with actively transcribed genomic regions.

These observations failed to support the seemingly simple and ideal model for epigenetic inheritance of histone modifications (Figure 1B), leading future investigations into alternative models. These studies also suggest that if there are any templated copying events for histone modifications, the most likely templates would be neighboring pre-modified parental nucleosomes, and, if this were to be the case, then the mitotic inheritance of histone modifications may not be achieved by a highly precise mechanism (see section Templated modification copying events versus modification reinforcement).

Replication-coupled histone modification “copying” versus replication-independent histone modification “maturation”

Replication-coupled modification “copying” on newly deposited histones behind the replication fork?

CpG methylation is faithfully maintained by recruitment of DNMT1 to the replication forks 41 by PCNA 42 and UHRF1 43, 44. Similarly, many histone modifiers have been reported to localize to the replication forks, suggesting potential replication-coupled reestablishment of histone modifications on the nascent chromatin. These include PCNA-mediated replication-coupled recruitment of HDACs 45; CAF-1-mediated MBD1 and SETDB1 recruitment during heterochromatin DNA replication 46, 47, 48; and DNMT1-mediated replication-coupled recruitment of G9a 49. Finally, the PRC2 complex has also been reported to localize to the replication foci throughout S phase 50.

That certain chromatin-modifying enzymes localize to replication foci raises the attractive hypothesis that factors such as PCNA and CAF-1 that interact with the replication machinery may serve as a landing pad for the chromatin-modifying enzymes to facilitate rapid restoration of the histone modifications on nascent chromatin.

Replication-independent modification “maturation” on newly deposited histones?

Histone modifications are often designated as being “epigenetic modifications”. However, certainly not all modifications are “epigenetic”, because many of them may not be inheritable as required under a widely accepted definition of “epigenetics” 1, 4. In fact, even for methylated versions of H3K9 and H3K27 that play critical roles in well-studied epigenetic phenomena, including position effect variegation, Polycomb silencing and X chromosome inactivation in mammals 5, 6, 7, 8, 9, 10, 11, 12, 13, it is not clear whether these modifications are faithfully reestablished on the newly deposited histones. Therefore, the development of a new technology to monitor the establishment of histone modifications on newly deposited histones will help resolve two important questions: (1) which kinds of histone modifications are faithfully reestablished on newly deposited histones; and (2) what is the kinetics for the reestablishment of histone modifications on newly deposited histones?

Stable isotope labeling-based quantitative mass spectrometry analysis allowed simultaneous detection of modifications on pre-existing and newly deposited histones, as well as relative quantification of their global abundance, thus providing a resolution for the above-mentioned technical challenges. This approach was first utilized for studying the reestablishment of H4K20 methylation on newly deposited histones, with the results indicating that H4K20 was progressively methylated throughout the cell cycle 51. Interestingly, little H4K20me2, the main state of mammalian H4K20 methylation, was detected on newly synthesized H4 histones during the S phase, suggesting that H4K20me2 was not reestablished on newly deposited histones in a replication-dependent manner. This was confirmed by an independent study 52. These results are consistent with the findings that H4K20me2-3, established by the Suv4-20 enzymes, requires H4K20me1 53, which is catalyzed by PR-Set7, whose expression is cell cycle-regulated and present from G2 to early G1 54, 55, 56, 57, 58.

A similar strategy was applied to H3K79 methylation, with the improvement that both the H3 histone backbone as well as the methyl group can be simultaneously monitored 59. Surprisingly, old H3 histones continue to be methylated at K79 at a rate comparable to that of newly deposited H3 histones, a clear indication that H3K79 methylation cannot be reestablished in a replication-coupled manner. Given that H3K79me2 shares a similar half-life with the histone modifications most likely to carry epigenetic information (H3K27me3 and H3K9me3) 60, 61, it was proposed that there might be a certain degree of positional “scrambling” of K79 methylation through the cell cycle 59.

Taken together, studies with stable isotope labeling-based quantitative mass spectrometry analysis provided new technology and novel insights for understanding the modification-reestablishment process on newly deposited histones. Current results indicated that histone lysine methylations in general have a much longer half-life than acetylation or phosphorylation 51, 60, 61, and that certain methylation states might be established on newly deposited histones in a chromatin assembly-coupled way. But clearly, modification states like H4K20me2 are established by a mechanism independent of DNA replication 51, 52. These modification states might gradually “mature” through the cell cycle.

Templated modification copying events versus modification reinforcement

Templated modification copying events?

In order to be heritable, pre-existing modifications should, at least to some extent, be able to serve as templates to guide the reestablishment of the same modifications on newly deposited histones 62. Considering that the vast majority of histone H3-H4 tetramers segregate in a conservative manner 26, 27, 40, the most likely templates might be adjacent pre-existing nucleosomes in cis (Figure 2).

Figure 2
figure 2

Modification spreading model that can also be involved in the epigenetic inheritance of histone modifications.

The proposed modification copying events are highly similar to another analogous phenomenon, chromatin modification spreading. Certain histone modification events, including H4K16 deacetylation, H3K9 methylation and H3K27 methylation, can spread along the chromatin to maintain a regional heterochromatic environment 63, 64, 65. These events are typically mediated by the association of a chromatin-modifying enzyme with a partner effector protein that specifically recognizes the modified product, such as Sir2 (H4K16 deacetylase) 66 with Sir3 (H4K16 effector) 67, Suvar39/Clr4 (H3K9 methyltransferase) 5, 68 with HP1/Swi6 (H3K9me3 effector) 6, 7, 68, and PRC2-Ezh2 (H3K27 methyltransferase) 8, 9, 10, 11 with Eed, a subunit of PRC2 as the H3K27me3 effector 50, 65. Therefore, inheritance of histone modifications can be achieved through similar biochemical mechanisms.

A recent study describing allosteric activation of PRC2 by H3K27me3 peptide suggests that PRC2 can “sense” the neighboring repressive environment and intrinsically mediate chromatin modification propagation 65. A similar strategy might be used for other histone-modifying enzymes, although stimulation of histone methyltransferase activities by neighboring methylated nucleosomes in cis awaits biochemical demonstration. Importantly, this mechanism is compatible with both the replication-coupled modification “copying” model and the replication-independent modification “maturation” model.

Modification reinforcement?

Although the above-mentioned templated modification copying events appear practical, at least for some modifications, accumulating evidence suggests many epigenetic states may not be inherited solely by this simple mechanism.

Several chromatin modifications may act coordinately to maintain the epigenetic states reflected by these modifications. One good example is the trio among UHRF1, DNMT1 and H3K9 methylation (Figure 3). UHRF1 recruits DNMT1 to hemimethylated CpG sites during DNA replication to facilitate the maintenance of DNA methylation 43, 44; DNMT1 interacts with G9a and recruits it to replication foci, potentially for replication-coupled H3K9me2 reestablishment on the nascent chromatin 49. Interestingly, H3K9me2 catalyzed by G9a or H3K9me3 catalyzed by Su(var)39-h can also recruit UHRF1 through direct recognition 69 and potentially facilitate DNA methylation. This is a clear feedback loop with players in this system reinforcing each other for eventual maintenance of the epigenetic states that they present. We expect that similar cases for other chromatin modifications will be discovered in the coming years.

Figure 3
figure 3

A reinforcement model for epigenetic inheritance. See text for details.

Finally, we would like to point out that the transcription status of a gene locus itself might play a role in maintaining its epigenetic status. Numerous chromatin modifiers are known to associate with the transcription machinery (for a review, see 70); therefore, it is easily conceivable that active transcription maintained by these modifications may in turn facilitate the reestablishment of these modifications in the next cell cycle. Although less apparent, it is also likely that transcription repression mediated by repressive modifications may also in turn facilitate the reestablishment of these modifications by certain favorable characteristics of the repressed chromatin, which awaits further investigations.

Concluding remarks

In comparison to genetic information which is strictly inherited via “code-copying” (DNA replication), recent progress suggests that epigenetic inheritance mechanisms are much more complicated. Although the mechanisms underlying epigenetic inheritance appear to be far less faithful or rigid in comparison to DNA replication, they do in fact offer an opportunity for plasticity, which is critical to the most fundamental role of epigenetic information, which is to translate one genome into hundreds of epigenomes for the generation of distinct cellular functions.