In this article, we report the following important findings regarding the mechanism of assembly of SAHF. First, SAHF results from condensation of individual chromosomes, each chromosome condensing into a single SAHF focus. Second, formation of SAHF by the histone chaperone ASF1a depends on its ability to bind to histone H3, in addition to the requirement for HIRA binding that we showed previously (86
). Third, HP1γ is phosphorylated on serine 93 in senescent cells, and this modification is required for its deposition in SAHF but not for its localization to PML bodies. Fourth, high levels of chromatin-bound HP1 proteins are not required for chromosome condensation, deposition of macroH2A proteins in SAHF, or other hallmarks of the senescence program, such as expression of SA β-gal activity and senescence-associated cell cycle exit.
Based on these findings and others reported previously (57
), we propose the following stepwise model, comprised of dependent and independent steps, for the formation of SAHF (Fig. ). Initially, histone chaperone protein HIRA and HP1 (HP1α, -β, and -γ) are recruited to PML nuclear bodies. Our previous kinetic analysis of SAHF formation showed that HIRA and HP1 proteins enter PML nuclear bodies prior to any detectable chromosome condensation or any other molecular marker of SAHF (86
). Since HP1βΔN does not affect the localization of HIRA to PML bodies but does block recruitment of endogenous HP1 proteins to PML bodies and does not go to PML bodies itself (data not shown), we conclude that HIRA enters PML bodies independently of HP1 proteins.
FIG. 10. A stepwise model for the formation of SAHF in senescent human cells. This model indicates the key steps in SAHF formation, after initiation by senescence triggers. Dashed lines are steps which are currently poorly defined (see text for details). Abbreviations: (more ...)
The role served by the translocation of HIRA and HP1 proteins to PML bodies is not known. However, PML bodies have been suggested as sites of assembly of macromolecular complexes and also as sites of protein modification (24
). Significantly, in cells approaching senescence, HP1γ becomes phosphorylated on serine 93. Interestingly, we have not detected analogous phosphorylation of HP1α and -β. The nonphosphorylatable mutant HP1γ(S93A) efficiently enters PML bodies but does not efficiently localize to SAHF. Thus, HP1γ might be phosphorylated inside PML bodies and phosphorylation might target HP1γ to SAHF (86
). Alternatively, HP1γ might be phosphorylated after the protein exits PML bodies en route to SAHF.
In the earliest discernible change in chromatin structure itself, individual chromosomes condense to form single SAHF. Previously, we showed that chromatin condensation depends on the histone chaperone ASF1a and is driven by a complex of ASF1a and its binding partner HIRA (86
). Here we have extended this to show that chromosome condensation requires an interaction of ASF1a with histone H3 and HIRA. Since the HIRA/ASF1a complex serves as a chaperone for deposition of the histone H3/H4 complex into chromatin (65
), it seems likely that chromosome condensation by HIRA and ASF1a depends on their chaperone activity. Conceivably, chromosome condensation depends, in part, on increased nucleosome density due to HIRA/ASF1a-mediated nucleosome deposition. This is consistent with many previous reports that transcriptionally active chromatin is depleted of nucleosomes. This is true at both a genome-wide and a local chromatin level (2
). Moreover, a previous study reported that the facultative heterochromatin of the inactive X chromosome has a higher nucleosome density than most other regions of the nucleus (62
). Together, this suggests that whole chromosome condensation and gene silencing may result from increased nucleosome density throughout the chromosome.
The HIRA/ASF1a chaperone complex preferentially utilizes histone H3.3 as a deposition substrate (46
). Significantly, histone H3.3 accumulates in fibroblasts approaching senescence and in nondividing differentiated cells, in some cases to about 90% of the total histone H3, presumably with the majority being in inactive chromatin (11
). Unfortunately, because histone H3.3 and canonical H3.1 differ only by five amino acids, they cannot presently be differentiated immunologically and there is no straightforward way to ask whether endogenous histone H3.3 is specifically enriched in SAHF. The idea that SAHF might contain histone H3.3 may initially seem unlikely, because deposition of histone H3.3 is typically linked to transcription activation (5
), whereas SAHF is a form of transcriptionally silent facultative heterochromatin (56
). However, the apparent inconsistency in this idea is merely an extension of an existing paradox. Specifically, HIRA and its orthologs in other species are typically involved in gene silencing and formation of heterochromatin (9
), whereas HIRA's favored deposition substrate, histone H3.3, is linked to transcriptional activation (5
). However, to our knowledge, histone H3.3 per se has not been shown to directly cause or contribute to transcription activation, and a proportion of histone H3.3 does carry posttranslational marks characteristic of transcriptionally silent chromatin (32
). Therefore, histone H3.3 is unlikely to be exclusively linked to transcription activation. Instead, deposition of histone H3.3 may be associated with any major remodeling of chromatin, perhaps as a way to “reset” histone modifications. To express this idea, Ooi and coworkers have suggested that histone H3.3 is a chromatin “repair” variant (58
). Concordant with this proposal, after egg fertilization in flies, dHIRA activity is required for the replacement of protamines by histone H3.3-containing nucleosomes in decondensing sperm chromatin (46
). By this view, the HIRA/ASF1a complex might drive formation of SAHF by deposition of histone H3.3-containing nucleosomes.
In line with the idea that chromosome condensation to form SAHF results primarily from increased nucleosome density, chromosome condensation into SAHF does not require the accumulation of H3K9Me or the deposition of heterochromatic proteins HP1 and macroH2A. Our previous kinetic analysis showed that chromatin condensation occurs prior to the accumulation of H3K9Me and the deposition of HP1 and the histone variant macroH2A in chromatin (86
). Here we have shown that chromosome condensation, triggered by an activated Ras oncogene or ectopic expression of ASF1a, efficiently occurs in the absence of high levels of stably bound HP1 proteins. Together, these results eliminate the possibility that H3K9Me, HP1, or macroH2A drives chromosome condensation. The finding that facultative heterochromatin can form in the absence of stably bound HP1 proteins is consistent with studies of facultative heterochromatin in nucleated vertebrate erythrocytes, which ordinarily forms without HP1 proteins (26
). In sum, the HIRA/ASF1a complex appears to drive chromosome condensation by acting upstream of characteristic heterochromatin modifications and associated proteins, most likely by contributing to nucleosome assembly through the deposition of histone H3/H4 complexes.
The final steps of SAHF formation consist of recruitment of macroH2A and HP1 proteins to chromatin. These two steps are not separable, based on a temporal analysis alone (86
). However, we have shown here that the recruitment of macroH2A occurs in the absence of stably bound HP1 proteins. At this time, we cannot exclude the possibility that recruitment of HP1 proteins to chromatin depends on prior loading of histone macroH2A. However, since we know of no evidence in support of this idea, we propose that HP1 and macroH2A proteins are independently loaded onto chromatin at approximately the same time. We find that phosphorylation of HP1γ on S93 is required for its efficient recruitment to heterochromatin. Interestingly, another study found that HP1γ phosphorylated on this residue is localized to euchromatin in immortal and transformed cells (45
) (it should be noted that these authors numbered the processed form of HP1γ and so referred to the same residue as S83). Thus, phosphorylation of this site might target HP1γ to different chromatin sites depending on the physiological context.
Remarkably, loading of abundant HP1 proteins onto chromatin is not required for two hallmarks of the senescent phenotype: expression of SA β-gal and senescence-associated cell cycle exit. We obviously cannot rule out the possibility that the residual chromatin-bound HP1 proteins are sufficient to mediate HP1 functions that are required for these senescence phenotypes. However, these results raise the possibility that HP1 proteins do not contribute to the acute onset of the senescent phenotype. Instead, HP1 proteins might be required for the long-term maintenance of SAHF and the senescent state. Alternatively, HP1 proteins might secure the senescent state in the face of genetic alterations or cellular perturbations that compromise other aspects of the senescence program. These ideas remain to be tested.
In contrast to our results with HP1 proteins, Narita and coworkers found through shRNA knock-down experiments that the HMGA protein HMGA2 is required for the formation of SAHF (56
). This might suggest that HMGA proteins are incorporated into SAHF quite early during SAHF assembly, perhaps at the time of chromosome condensation. However, until this is directly demonstrated, we have omitted HMGA's point of entry into SAHF from our model.
Although Fig. provides a framework model for the formation of SAHF, many other questions remain. For example, we do not know the triggers responsible for localization of HIRA and HP1 proteins to PML bodies. We do not know the specific reason for HIRA's localization to PML bodies and the spatial and mechanistic relationships between its localization to PML bodies and the formation of SAHF. Finally, we do not know the identity of the kinase responsible for the phosphorylation of HP1γ, the histone methyltransferase that methylates lysine 9 of histone H3 to create H3K9Me, or the factors required for the deposition of macroH2A into SAHF. Studies to answer these questions are ongoing. Meanwhile, the model proposed in Fig. provides a valuable conceptual framework for thinking about these questions, as well as summarizing a large body of existing knowledge.