REVIEWS Particular features that are created at the time of It is possible that the dilution of H3.3 and its marks DNA replication in a particular domain might also be by one-half after one cell cycle might not affect the exploited. NP95 has affinity for both hemimethylated transcriptional readout of a region. Thus, sustained DNA and histones-b64, and specifically interacts with active gene expression, combined with modifications on peptides that are methylated at H3K9 in vitro by poten- parental H3. 3, might recruit factors that modify newly tially reading histone marks. In addition, NP95 was incorporated H3. 1 with the appropriate marks, and found in a complex with HDACs and G9a"12. Therefore, H3. 3 incorporation might be stimulated. Consistent with as well as binding to hemimethylated DNA, NP95 could this hypothesis, arrays of nucleosomes that contain both interpret the histone environment, thereby creating a H3. 3 and H3. 1 nucleosomes have been observed, and feedback mechanism that involves the mutual reinforce- analysis of histone modifications in this context show ment of histone and DNA methylation marks. In this that, when adjacent to H3. 3 nucleosomes, H3. 1 nucleo- situation histone marks would influence the inheritance somes accumulate active marks However, dilution of of DNA methylation. Further chromatin-binding pro- H3. 3 over a number of generations might be reconciled teins or chromatin modifiers with dual affinity for both by the replication-independent incorporation of H3.3 DNA methylation and a particular histone modification that is promoted by chaperones, such as Hir-related are likely to be identified. protein A(HIRA), following transcription(FIG 4a) The examples above show how histone and DNA methylation in a repressive domain could be maintained Inheritance of CENP-A. The histone H3 variant CENP-A at the replication fork. However, how active chromatin marks the site of centromere identity 49. The associa- marks are propagated is less clear. Recently, transmission tion of CENP-A with centromeres is extremely stable, as of an active state through nuclear transfer in Xenopus shown by quantitative fluorescence recovery after photo- laevis has been reported, and it has been proposed that bleaching(FRAP)analysis, and it remains associated the replacement histone variant H3.3 is required for epi- through cell division%. Although the exact mechanism of genetic memory 6. To evaluate this hypothesis, it is nec- CENP-A deposition at centromeres remains enigmatic, it essary to better understand the mechanisms that involve is a replication-independent process, as is the deposition replication-independent histone exchange processes and of H3.3 CENP-A deposition was first proposed the replacement of histone variants. to occur in G2 phase, because CENP-A assembly can take place in the presence of the DNA replication inhibitor Inheritance of histone variants outside S phase aphidicolin, and CENPA mRNA and the CENP-a pro Histone variants can mark a particular chromatin state: tein peak in G2 phase 7. Recent evidence in mammalian H3. 3 is enriched at active regions, whereas the unique cells now suggests that the loading of new CENP-Aonto incorporation of the centromere-specific histone H3 centromeres is restricted to a discrete cell cycle window in An enzyme that removes acetyl variant CenH3(CENP-A in humans) specifies the site of late telophase-early Gl phase, but the mechanism and groups from histones. centromere identity. Together with the replicative vari- the specific chaperone that facilitate CENP-A deposition ants H3. 1 and H3. 2, the replacement variants H3.3 and remain to be deciphered. n enzyme that catalyses the CENP-A constitute the major histone H3 isotypes that Centromeric DNA is replicated during S phase, in addition of a methyl group to are known in mammals. During S phase, H3. I and which parental CENP-A nucleosomes are distributed to pecific Lys residues in histones H3. 2 are exclusively incorporated, whereas the deposi- daughter strands 7. 9. Therefore, chromatin at the centro- and other non-histone tion of replacement variants, such as H3. 3 or CENP-A, meres contains one-half of the complement of CENP-A occurs outside S phasell 88. Thus, the histone variants nucleosomes after the completion of S phase and during H3. 3 and CENP-A have emerged as candidates for key subsequent G2 and M phases. To reconcile the deficit enzyme that catalyses the players of epigenetic information that can be transmitted in CENP-A molecules, current models predict that dur ing replication, either H3 1-containing nucleosomes specific Lys residues in histones are temporarily placed at centromeres, or, alternatively, nd other non-histone proteins. Inheritance of H3. 3. H3.3 is associated with transcrip- nucleosome gaps' are created that are filled later in the tionally active regions and is enriched in active histone cell cycle(FIG 4b) Heterochromatin protein 1 markssss9,90. Furthermore, nucleosomes that contain Recent studies suggest that CENP-A nucleosomes H3. 3 seem to be less stable than those that contain H3.1 are unusual and that these peculiarities might provide containing protein that binds (REF. 91). The extent to which this depends on the dif- a means of marking this region of the chromosome as ferential modification status of the nucleosomes, the unique. For example, in budding yeast, a specialized Cse4 presence of other variants, such as H2A.Z9, or inher- ( Saccharomyces cerevisiae CenH3)-containing nucleosome east(Swi6), mammals (HP1) ent differences in their structural properties remains has been proposed to exist in a form in which histon and Drosophila melanogaster to be established. Regardless, in vivo, these properties H2A and H2B are replaced by the non-histone protein suggest that H3. 3 nucleosomes are more dynamic or suppressor of chromosome missegregation 3( Scm3) Pericentric heterochromatin amenable to displacement during transcription. Given In D. melanogaster, ahemisome' that consists of one A heterochromatic region that replication leads to a concomitant deposition of molecule each of CenH3, H4, H2A and H2B has been H3. 1, the density of H33-containing nucleosomes described tor. Additional evidence suggests that, like H3.3 ontaining the centromere-spe. is reduced. As the mixing of H3. 1 and H3.3 in the nucleosomes, CENP-A nucleosomes are easier to dis- nd which is considered to be same nucleosome has not been observed.s,a semi- assemble in vitro than canonical nucleosomes.One conservative mechanism at the fork is unlikely for H3.3 might speculate that 'unusual CENP-A-containing nucleo- NATURE REVIEWS MOLECULAR CELL BIOLOGY 22009 Macmillan Publishers Limited All rights reservedHistone deacetylase An enzyme that removes acetyl groups from histones. Lys methyltransferase An enzyme that catalyses the addition of a methyl group to specific Lys residues in histones and other non-histone proteins. Lys acetyltransferase An enzyme that catalyses the addition of an acetyl group to specific Lys residues in histones and other non-histone proteins. Heterochromatin protein 1 (HP1). A chromodomain￾containing protein that binds to methylated K9 on histone H3 and is associated with heterochromatin in fission yeast (swi6), mammals (HP1) and Drosophila melanogaster (HP1). Pericentric heterochromatin A heterochromatic region adjacent to chromatin containing the centromere-spe￾cific histone H3 variant CenH3, and which is considered to be typical constitutive heterochromatin. Particular features that are created at the time of DNA replication in a particular domain might also be exploited. NP95 has affinity for both hemimethylated DNA and histones34–36,84, and specifically interacts with peptides that are methylated at H3K9 in vitro85 by poten￾tially reading histone marks. In addition, NP95 was found in a complex with HDAcs and G9a31,32. Therefore, as well as binding to hemimethylated DNA, NP95 could interpret the histone environment, thereby creating a feedback mechanism that involves the mutual reinforce￾ment of histone and DNA methylation marks. In this situation, histone marks would influence the inheritance of DNA methylation. Further chromatin­binding pro￾teins or chromatin modifiers with dual affinity for both DNA methylation and a particular histone modification are likely to be identified. The examples above show how histone and DNA methylation in a repressive domain could be maintained at the replication fork. However, how active chromatin marks are propagated is less clear. Recently, transmission of an active state through nuclear transfer in Xenopus laevis has been reported, and it has been proposed that the replacement histone variant H3.3 is required for epi￾genetic memory86. To evaluate this hypothesis, it is nec￾essary to better understand the mechanisms that involve replication­independent histone exchange processes and the replacement of histone variants. inheritance of histone variants outside s phase Histone variants can mark a particular chromatin state: H3.3 is enriched at active regions, whereas the unique incorporation of the centromere­specific histone H3 variant cenH3 (ceNP­A in humans) specifies the site of centromere identity. Together with the replicative vari￾ants H3.1 and H3.2, the replacement variants H3.3 and ceNP­A constitute the major histone H3 isotypes that are known in mammals87. During S phase, H3.1 and H3.2 are exclusively incorporated, whereas the deposi￾tion of replacement variants, such as H3.3 or ceNP­A, occurs outside S phase11,88. Thus, the histone variants H3.3 and ceNP­A have emerged as candidates for key players of epigenetic information that can be transmitted in a replication­independent manner. Inheritance of H3.3. H3.3 is associated with transcrip￾tionally active regions and is enriched in active histone marks55,89,90. Furthermore, nucleosomes that contain H3.3 seem to be less stable than those that contain H3.1 (ReF. 91). The extent to which this depends on the dif￾ferential modification status of the nucleosomes55, the presence of other variants, such as H2A.Z92, or inher￾ent differences in their structural properties remains to be established. Regardless, in vivo, these properties suggest that H3.3 nucleosomes are more dynamic or amenable to displacement during transcription. Given that replication leads to a concomitant deposition of H3.1, the density of H3.3­containing nucleosomes is reduced. As the mixing of H3.1 and H3.3 in the same nucleosome has not been observed55,59, a semi￾conservative mechanism at the fork is unlikely for H3.3 inheritance (FIG. 3b). It is possible that the dilution of H3.3 and its marks by one­half after one cell cycle might not affect the transcriptional readout of a region. Thus, sustained active gene expression, combined with modifications on parental H3.3, might recruit factors that modify newly incorporated H3.1 with the appropriate marks, and H3.3 incorporation might be stimulated. consistent with this hypothesis, arrays of nucleosomes that contain both H3.3 and H3.1 nucleosomes have been observed, and analysis of histone modifications in this context show that, when adjacent to H3.3 nucleosomes, H3.1 nucleo￾somes accumulate active marks55. However, dilution of H3.3 over a number of generations might be reconciled by the replication­independent incorporation of H3.3 that is promoted by chaperones, such as Hir­related protein A (HIRA), following transcription59,93 (FIG. 4a). Inheritance of CENP‑A. The histone H3 variant ceNP­A marks the site of centromere identity94,95. The associa￾tion of ceNP­A with centromeres is extremely stable, as shown by quantitative fluorescence recovery after photo￾bleaching (FRAP) analysis, and it remains associated through cell division96. Although the exact mechanism of ceNP­A deposition at centromeres remains enigmatic, it is a replication­independent process, as is the deposition of H3.3 (ReF. 97). ceNP­A deposition was first proposed to occur in G2 phase, because ceNP­A assembly can take place in the presence of the DNA replication inhibitor aphidicolin, and CENPA mRNA and the ceNP­A pro￾tein peak in G2 phase97. Recent evidence in mammalian cells now suggests that the loading of new ceNP­A onto centromeres is restricted to a discrete cell cycle window in late telophase–early G1 phase98, but the mechanism and the specific chaperone that facilitate ceNP­A deposition remain to be deciphered. centromeric DNA is replicated during S phase, in which parental ceNP­A nucleosomes are distributed to daughter strands97,98. Therefore, chromatin at the centro￾meres contains one­half of the complement of ceNP­A nucleosomes after the completion of S phase and during subsequent G2 and m phases. To reconcile the deficit in ceNP­A molecules, current models predict that dur￾ing replication, either H3.1­containing nucleosomes are temporarily placed at centromeres, or, alternatively, nucleosome ‘gaps’ are created that are filled later in the cell cycle99 (FIG. 4b). Recent studies suggest that ceNP­A nucleosomes are unusual and that these peculiarities might provide a means of marking this region of the chromosome as unique. For example, in budding yeast, a specialized cse4 (Saccharomyces cerevisiaecenH3)­containing nucleosome has been proposed to exist in a form in which histones H2A and H2B are replaced by the non­histone protein suppressor of chromosome missegregation 3 (Scm3)100. In D. melanogaster, a ‘hemisome’ that consists of one molecule each of cenH3, H4, H2A and H2B has been described101. Additional evidence suggests that, like H3.3 nucleosomes, ceNP­A nucleosomes are easier to dis￾assemble in vitro than canonical nucleosomes102. One might speculate that ‘unusual’ ceNP­A­containing nucleo￾somes represent centromeric chromatin in an intermediate REVIEWS NATuRe ReVIeWS | Molecular cell Biology VOlume 10 | mARcH 2009 | 199 © 2009 Macmillan Publishers Limited. All rights reserved