DNA DSBs (double-strand breaks) represent a critical lesion for a cell, with misrepair being potentially as harmful as lack of repair. In mammalian cells, DSBs are predominantly repaired by non-homologous end-joining or homologous recombination. The kinetics of repair of DSBs can differ widely, and recent studies have shown that the higher-order chromatin structure can dramatically affect the pathway utilized, the rate of repair and the genetic factors required for repair. Studies of the repair of DSBs arising within heterochromatic DNA regions have provided insight into the constraints that higher-order chromatin structure poses on repair and the processing that is uniquely required for the repair of such DSBs. In the present paper, we provide an overview of our current understanding of the process of heterochromatic DSB repair in mammalian cells and consider the evolutionary conservation of the processes.
DNA DSBs (double-strand breaks) are a major form of DNA damage for all organisms. Consequently, DSB repair pathways have been evolutionarily conserved. Indeed, the two major DSB repair pathways, HR (homologous recombination) and NHEJ (non-homologous end-joining), are conserved from bacteria to mammalian cells, although their usage differs between organisms. DSBs also elicit a signalling response that can activate cell-cycle checkpoint arrest and/or apoptosis as well as influence aspects of DSB repair. Although the signalling response in bacteria is less well conserved, the response in the two model yeasts commonly studied (Saccharomyces cerevisiae and Schizosaccharomyces pombe) is remarkably well conserved with the process in higher organisms. Increasing evidence has suggested that higher-order chromatin structure markedly affects DSB repair and signalling. The relative amount of DNA in a mammalian compared with a yeast cell is dramatically different, and, to accommodate their high DNA content, 15–25% of mammalian DNA is highly compacted into HC (heterochromatin). Although yeast have heterochromatic DNA such as rDNA (ribosomal DNA), telomeres and centromeric DNA sequences, the relative ratio to transcriptionally active DNA is much reduced. This distinction may underlie some of the mechanistic differences in DSB repair and signalling between yeast and mammalian cells. In the present article, we discuss how higher-order chromatin structure affects DSBs and signalling in mammalian cells and consider how this may underlie some of the differences observed in yeast.
Differences in the repair of DSBs located within regions of euchromatin compared with heterochromatin in mammalian cells
It has long been recognized that DSBs are repaired with widely differing kinetics. A large fraction [~85% of IR (ionizing radiation)-induced DSBs] are repaired rapidly, whereas ~15% are repaired markedly more slowly, although there is probably a range between these two extremes . These kinetics are observed overtly in primary fibroblasts, where apoptosis does not occur after IR and are evident using a range of procedures to monitor DSB repair, including pulsed-field gel electrophoresis and enumeration of γH2AX (phosphorylated histone H2AX) foci in non-replicating cells [2,3]. Although the complexity of the DSB lesion can influence repair kinetics, more recent studies have strongly suggested that chromatin complexity is the major factor influencing the rate of repair after X- or γ-ray irradiation and that the slow component of DSB repair represents DSBs arising within, or in the vicinity of, HC regions (which we hereafter term HC-DSBs) . Indeed, recent studies have shown that, in G0/G1-phase cells (i.e. in non-replicating cells), DSBs with different end complexities induced by a range of DNA-damaging agents are repaired with similar kinetics and with a fast and slow component of similar magnitude to that shown following X-ray exposure. This analysis includes DSBs induced by TBH (t-butyl hydroperoxide), an agent causing oxidative damage . Furthermore, with all agents, the DSBs that remain at later times after treatment show a marked enrichment at HC regions [4,5]. These results suggest that HC presents a barrier to DSB repair and imply that DSBs relocate to the periphery of HC regions before repair. They nicely couple with earlier studies demonstrating that HC represents a barrier to the expansion of γH2AX foci formation in both yeast and mammalian cells [6,7]. Interestingly, a recent study involving the irradiation of NIH 3T3 cells [where regions of HC can be visualized as densely staining DAPI (4′,6-diamidino-2-phenylindole) centres] with ion particle irradiation that induces DSBs in a track, has provided evidence that the γH2AX foci do not form in a linear track, but ‘bend’ around the periphery of HC regions . Collectively, studies strongly suggest that highly compacted HC impedes the rate of DSB repair.
Role of ATM (ataxia telangiectasia mutated) and DSB-damage response signalling in HC-DSB repair
The notion that HC impedes DSB repair gained reinforcement with the finding that the slow component of DSB repair is specifically defective in cells lacking ATM or the nuclease Artemis. In G0/G1-phase, the process still requires DNA ligase IV and DNA-PKcs (DNA-dependent protein kinase catalytic subunit), two essential NHEJ proteins . This suggests that additional factors are required for the HC-DSB process, but that rejoining ensues by NHEJ in G0/G1-phase. Furthermore, the process additionally requires H2AX, the MRN (Mre11/Rad50/Nbs1) complex, MDC1 (mediator of DNA-damage checkpoint protein 1), RNF (RING finger protein) 8, RNF168 and 53BP1 (p53-binding protein 1), all DNA-damage signalling mediator proteins that accumulate at DSBs [9,10]. In contrast, all of these ATM signalling proteins including ATM are dispensable for the fast component of DSB repair, taken to represent EU (euchromatin)-DSBs . Importantly, KAP1 [KRAB (Krüppel-associated box)–zinc-finger protein-associated protein 1], an HC-building component, was identified as an ATM substrate, and mutation of KAP1 at a single phosphorylation site (Ser824) confers radiosensitivity . Furthermore, exposure to neocarzinostatin or radiation results in pan-nuclear chromatin relaxation revealed as enhanced susceptibility to micrococcal nuclease digestion, siRNA (small interfering RNA) of KAP1 causes similar chromatin relaxation in the absence of DNA damage, and expression of a non-phosphorylatable derivative of KAP1 (S824A KAP1) prevents the change in chromatin relaxation after IR or neocarzinostatin, whereas expression of the phosphomimic derivative (S824D KAP1) causes constitutive relaxation without DNA damage [4,11]. The ability to repair HC-DSBs parallels these findings: i.e. siRNA of KAP1 or expression of S824D KAP1 overcomes the requirement for ATM for the slow component of DSB repair, whereas expression of S824A KAP1 confers an ATM-like repair defect even in the presence of ATM. Collectively, these findings suggest that ATM phosphorylates KAP1 in its C-terminus, which affects HC condensation. However, it is important to appreciate that, although KAP1 appears to be less tightly associated with HC after IR, no major loss of KAP1 at HC regions is observed and, even in the presence of ATM, HC-DSBs are repaired with slower kinetics than EU-DSBs. This suggests that pKAP1 (p indicates phospho-) formation confers a subtle change in HC superstructure without marked dismantling of the superstructure (see below).
Pan-nuclear compared with localized KAP1 phosphorylation
KAP1 phosphorylation occurs in two ways that can be distinguished cytologically and genetically. At early times after IR, pKAP1 forms in a highly dose-dependent pan-nuclear manner . The pKAP1 generated at early times after IR is non-extractable, suggesting that it represents chromatin-bound KAP1. Pan-nuclear pKAP1 dissipates quite rapidly, e.g. by 3–6 h after 2–3 Gy of IR, the signal is no longer strongly visible. However, pKAP1 foci also form at a subset of γH2AX foci, which probably represent HC-DSBs. The pKAP1 foci can be detected at early times after IR using specialized imaging software, although they are not readily apparent due to the pan-nuclear pKAP1. They become clearly visible, however, as the pan-nuclear staining dissipates and persist with similar kinetics to those of the slow DSB-repair component. ATM is required for all pKAP1 formation (at least in non-replicating cells). Importantly, however, the damage-response mediator proteins, including 53BP1, are dispensable for pan-nuclear pKAP1, but are specifically required for pKAP1 foci formation . Given the fact that the mediator proteins are required for HC-DSB repair, this strongly suggests that pan-nuclear pKAP1 is dispensable, but pKAP1 foci formation is essential for HC-DSB repair. A rationale for these findings was provided by the realization that 53BP1 interacts with the Rad50 component of MRN, and thus aids the tethering of ATM at DSBs . Indeed, cells lacking 53BP1 fail to form pATM (as well as pKAP1) foci after IR . Collectively, these studies, coupled with other known findings, suggested a model in which ATM is initially released from the DSB and phosphorylates chromatin-bound KAP1 at low density throughout the nucleus . Alternatively, it is possible that non-chromatin-bound KAP1 is initially phosphorylated and rapidly becomes chromatin-bound, and a recent study reported that KAP1 is transiently recruited to laser tracks . The mediator protein-independence of pan-nuclear pKAP1 is consistent with findings that, although the magnitude of activated ATM (p1981-ATM) is slightly reduced when components of the pathway are lost (e.g. MDC1), loss of 53BP1 in particular, has little impact. pKAP1 foci formation, in contrast, which necessitates dense phosphorylation of KAP1 at HC-DSBs (where KAP1 levels are high) requires 53BP1-dependent ATM tethering. Since 53BP1's localization to the DSB requires the upstream mediator proteins, pKAP1 foci formation also requires these components. Furthermore, it appears that this dense phosphorylation of KAP1 is required to generate sufficient chromatin relaxation to promote HC-DSB repair. Such a model would suggest that some degree of HC relaxation may occur in a pan-nuclear manner after IR, as indeed is observed, but there may be more focused and marked relaxation at the DSB site. To date, no studies have shown that HC relaxation occurs specifically at the DSB site.
pKAP1 formation causes dispersion of the NuRD (nucleosome remodelling and histone deacetylation) component CHD (chromodomain helicase DNA-binding protein) 3 from HC
The obvious question emerging from this model is how does pKAP1 affect the HC superstructure. We addressed this in a recent study exploiting the finding that addition or removal of an ATM inhibitor after IR causes rapid and reversible changes to pKAP1 levels (pan-nuclear and focal) as well as to chromatin relaxation . We asked whether the level of other known HC proteins changes in parallel to formation or loss of pKAP1. Strikingly, we observed that the presence of the larger isoform of CHD3, a component of the NuRD remodelling complex reflected such changes, i.e. CHD3 levels were low when pKAP1 levels were high and vice versa. Another NuRD component, CHD4, however, did not change. CHD3 siRNA, but not CHD4 siRNA, causes enhanced micrococcal nuclease digestion (similar to that observed after IR exposure). The changes in CHD3 chromatin binding reflected both pan-nuclear and focal pKAP1 profiles. At early times after IR, when pan-nuclear pKAP1 levels are high, the level of pan-nuclear chromatin-bound CHD3 diminished and at later times after IR, when pKAP1 formation can be monitored at defined foci, the level of CHD3 at these foci also diminished. This suggests that pKAP1 formation affects CHD3 binding both pan-nuclearly and at specific foci.
CHD3 is known to interact with pKAP1 via a SUMO (small ubquitin-related modifier)–SIM (SUMO-interacting motif) interaction, i.e. CHD3 harbours a SIM domain, whereas KAP1 is a SUMO ligase that undergoes autoSUMOylation [16,17]. The SUMOylation sites on KAP1 are located in its C-terminus close to the Ser824 phosphorylation site, and phosphorylation of KAP1 has been reported to impair SUMOylation . However, we failed to observe any changes in KAP1 SUMOylation levels after IR, but found that inhibition of the interaction between a peptide encompassing the CHD3 SIM domain and SUMO by a KAP1 peptide encompassing Ser842 was greatly increased when the peptide was phosphorylated . Collectively, these and additional data suggest that, when phosphorylated, the C-terminus of KAP1 can out-compete the SUMO–SIM interaction and cause dispersion of CHD3 from HC. Significantly, the SUMO–SIM interaction is highly charge-dependent, and phosphorylation of KAP1 at Ser824 strongly enhances its negative charge, providing a mechanistic basis for the model proposed  (Figure 1).
Distinction between the repair of DSBs located in euchromatin and heterochromatic DNA regions
Does chromatin relaxation occur without changes in epigenetic modifications?
An important question is whether the HC relaxation required to allow DSB repair is accompanied by changes to epigenetic marks. Examination of histone H3K9 (histone H3 Lys9) trimethylation, a silencing marker, and histone H3K9 acetylation, which correlates with an open chromatin structure, failed to reveal any marked changes in levels that correlated with the changes observed in pKAP1 or CHD3 dispersal in either a pan-nuclear or focal manner . Interestingly, CHD3 interacts with HDAC (histone deacetylase) 1/2 and SetDB1 (SET domain bifurcated 1), which promotes H3K9 methylation. Indeed, SetDB1's interaction with CHD3 also occurs via a SUMO–SIM interaction . However, both the HDACs and histone methylases have multiple interactions with HC components and thus, even if dispersion of CHD3 from KAP1 did affect HDAC and/or SetDB1 levels, it may not necessarily confer any change in overall acetylation or methylation levels. This suggests that the changes to HC superstructure that confer relaxation may not be accompanied by marked changes in epigenetic modifications. Extensive resection or resynthesis at such DSBs would cause loss of histone modifications similar to those taking place during replication, and replacement of any such changes must carry a risk of imperfect recovery. The limiting of resection by increasing the accessibility of the DSB would minimize this risk. Thus KAP1 phosphorylation provides a means of transiently relaxing HC to allow DNA repair without promoting epigenetic changes.
Homologous recombination and heterochromatin
The studies discussed above have been carried out in G0/G1-phase cells to avoid any difficulties in interpretation arising from replication. By exploiting markers to identify cells in specific cell-cycle phases, it has more recently been possible to examine DSB repair in cells irradiated and subsequently maintained in G2-phase . Strikingly, these findings have shown that a fast and slow component of DSB repair takes place in G2- as in G1-phase and that the fast component of DSB repair occurs via NHEJ. However, somewhat surprisingly, the slow component of DSB repair in G2-phase represents HR, strongly suggesting that the HC-DSBs preferentially undergo repair by HR. A current working model is that NHEJ makes the first attempt to repair DSBs in both G1- and G2-phase; if rapid rejoining does not occur, then resection takes place in G2-phase and repair occurs by HR. We will not discuss this aspect further here, since it has been recently reviewed elsewhere .
Since the assembly of HC is important for the stability of repetitive DNA, it is surprising that HR is emerging as the mechanism of repair in HC, as recombination events using the wrong template can lead to translocations and genome rearrangements. HR must be tightly regulated to maintain repeat stability, and recent work in Drosophila has shown that, although resection of DSBs occurs in HC, Rad51 foci only form outside, leading to the suggestion that breaks move outside a dynamic HP1α (HC-associated protein 1α) domain for recombination to be completed . The relocalization requires both resection and checkpoint proteins. It provides a mechanism to limit access to the break by the recombination machinery by removing the DSBs from the vicinity of homologous sequences at the vulnerable strand invasion stage, helping to ensure that the sister chromatid is used to template the repair event. Such a model could underlie the movement of HC-DSBs in mammalian cells.
Insight into HC-DSB repair from yeasts
The rDNA in the nucleolus has been used to model HC repair in yeasts. The nucleolus forms a highly specialized structure around the actively transcribed rDNA repeats, which from electron microscopy and psoralin-cross-linking studies were thought not to be packaged into nucleosomes, but more recently have been found to have a dynamic structure of unphased nucleosomes . Little is known about repair of these active repeats . In contrast, inactive repeats and the non-transcribed spacer regions are heterochromatic and silenced for RNA polymerase II transcription (thus transcription-coupled repair is similarly excluded). Recombination in the rDNA is important to maintain copy number, and specific mechanisms have evolved to ensure its regulation. Under normal circumstances, recombination is restricted to sequences close to the replication fork barrier in the non-transcribed spacer regions and coupled to replication (reviewed in ). Since chromatin is remodelled in S-phase, this could suggest that HC is inaccessible to recombination factors without remodelling.
Replication-coupled recombination does not lead to Rad52 foci in the nucleolus in wild-type S. cerevisiae cells, either through nucleolar exclusion or because these events are rapid and do not require the levels of Rad52 to be visible as foci. Analysis of an I-Sce1 DSB in one of the heterochromatic spacer regions visualized by TetI–RFP (red fluorescent protein) binding to the tetO-marked I-Sce recognition site showed that the site was highly dynamic, but predominantly inside the nucleolus. When I-Sce1 was induced, TetI foci were seen on the nucleolar periphery . This suggests that the spatial organization of repair could be conserved between yeast and higher eukaryotes.
Spontaneous nucleolar Rad52 foci were seen in Smc5/6 mutants , correlating with a defect in the resolution of recombination in HR-dependent recovery of collapsed replication forks [28–30]. The Smc5/6 complex, which, like cohesin and condensin, is required for higher-order chromosome structure, has also been implicated in the exclusion of Rad51 foci from HC in Drosophila . In fission yeast, smc6 mutants exhibit an rDNA-silencing defect which additively led to increased spontaneous recombination , and it is therefore not yet clear whether the requirement for Smc5/6 to exclude Rad51 and Rad52 from HC is related to a fundamental role in higher-order chromosome dynamics or an indirect consequence of HC or recombination regulation.
It remains to be seen whether checkpoint proteins are required for the relocalization of DSBs within HC regions in yeast. KAP1 is not conserved in either yeast, and HC in budding yeast is assembled in a functionally divergent manner to fission yeast and other eukaryotes , making it unlikely that targets are conserved. However, multiple chromatin components are phosphorylated in a checkpoint-dependent manner in both yeasts [33,34] and, intriguingly, Msc1, a modifier of H2A.Z, was identified as a suppressor of Chk1 (checkpoint kinase 1) function in fission yeast . Msc1 has multiple domains suggestive of a role in modulating chromatin structure and/or function, including three RING fingers and Jumonji domains. Thus checkpoint protein function in the repair of HC-DSBs may also be conserved.
SUMOylation has been implicated in mechanisms that regulate recombination in HC in fission yeast , and, in budding yeast, Torres-Rosell et al.  found that SUMOylation of Rad52 (independently of Smc5/6) was required to exclude Rad52 foci from the nucleolus. SUMOylation of Rad52 has been implicated in the efficiency of HR in budding yeast [37,38]. Thus SUMOylation of HR and chromatin-associated proteins (although not necessarily the same targets) plays a conserved role in the localization of HR for efficient repair in HC.
The relocalization of DSBs may not be restricted to those in HC as Nagai et al.  showed that unrepairable DSBs were very dynamic and become enriched at the nuclear periphery in budding yeast. Relocation required a SUMO-targeted ubiquitin ligase and Mec1/Tel1 [ATR (ATM- and Rad3-related)/ATM] kinases, but not the downstream checkpoint response. Thus relocalization of difficult to repair DSBs may be a universal phenomenon. To date, the term HC has been used broadly, and it is likely that there will be a range of chromatin complexity extending from highly transcriptionally active regions to highly compacted regions encompassing facultative or constitutive HC. How signalling affects this broad range still awaits study.
There is emerging evidence that, in addition to the regulation of repair pathways through the cell cycle, the spatial organization of repair at HC is highly regulated to maintain repeat stability. We speculate that the regulation of repair in a spatial manner could enforce a temporal order of repair pathways and help to regulate outcomes. The choreography of repair in time and space is essential for genome stability and its elucidation will be key to future developments in the field.
Signalling 2011: a Biochemical Society Centenary Celebration: A Biochemical Society Focused Meeting held at the University of Edinburgh, U.K., 8–10 June 2011. Organized and Edited by Nicholas Brindle (Leicester, U.K.), Simon Cook (The Babraham Institute, U.K.), Jeff McIlhinney (Oxford, U.K.), Simon Morley (University of Sussex, U.K.), Sandip Patel (University College London, U.K.), Susan Pyne (University of Strathclyde, U.K.), Colin Taylor (Cambridge, U.K.), Alan Wallace (AstraZeneca, U.K.) and Stephen Yarwood (Glasgow, U.K.).
p53-binding protein 1
ataxia telangiectasia mutated
chromodomain helicase DNA-binding protein
phosphorylated histone H2AX
KRAB (Krüppel-associated box)–zinc-finger protein-associated protein 1
mediator of DNA damage checkpoint protein 1
nucleosome remodelling and histone deacetylation
RING finger protein
SET domain bifurcated 1
small interfering RNA
small ubiquitin-related modifier
The Jeggo laboratory is supported by the Medical Research Council [programme grant number G0500897], the International Association for Cancer Research [grant number R39E], the Wellcome Research Trust [grant number 6642] and the EMF Biological Research Trust. The Murray laboratory is supported by the Medical Research Council [grant number G0901011], Association for Cancer Research [grant number 10-273] and Cancer Research UK [grant number C9601/A9484].