Pericentric heterochromatin is the region of chromatin juxtaposed to the centromeres and is composed of major satellite repeats (Guenatri et al., 2004). In mouse nuclei, pericentric heterochromatin forms cytologically visible “chromocenters” with DNA stains such as DAPI (Probst & Almouzni, 2008). Chromocenters are epigenetically distinguished by H3K9me3 marks written by the histone methyltransferases SUV39H1 and H2 (Müller-Ott et al., 2014). H3K9me3 marks recruit HP1 proteins, which might facilitate chromatin compaction by dimerizing and/or oligomerizing to bridge nucleosomes (Larson et al., 2017; Machida et al., 2018; Keenen et al., 2021). Pericentric heterochromatin is critical for genome stability, and when disrupted, chromosomal abnormalities, defects in segregation, and increased tumorigenesis are observed in mouse models (Peters et al., 2001; Taddei et al., 2001). Studying the distribution of transcriptional regulators relative to chromatin density using epigenetic modifications to classify chromatin compartments revealed an inverse correlation between chromatin density and protein size, with only the smallest proteins freely accessing heterochromatic regions associated with repressive marks (Miron et al., 2020). Further, Maeshima et al. (2015) investigated the importance of small transcription factor size (about 5 nm) in accessing the condensed interior of topologically associated domains (TADs). They demonstrate in silico that 5 nm spherical objects have free movement in condensed chromatin, 10 nm objects have attenuated movement, and objects larger than 15 nm are excluded entirely (Maeshima et al., 2015). Hihara et al. (2012) similarly used computer simulation to model the movement of EGFP pentamers modelled as 13 nm spheres through nucleosomes modelled as 10 nm spheres. They found that the modelled EGFP pentamers showed attenuated movement and penetration into 10 nm spheres when modelled at high concentrations expected of compact chromatin. Gorisch et al. (2005) used FITC-labelled dextrans of sizes 42, 77, 148, 282, 464, and 2500 kDa dextrans to compare how molecular weight (MW) affects molecule concentration in chromocenters, which is a predicted size range of approximately 6.5–51 nm. They showed reduced access of the dextrans 282 kDa and above in HeLa cell heterochromatin, and this accessibility was increased by increasing chromatin acetylation. An attractive model due to its intuitive simplicity is that size-based accessibility resulting from chromatin compaction restricts availability to the interior (Bancaud et al., 2009). In this model, diffusion into heterochromatin is limited by the size of pores between chromatin fibres or nucleosomes. Size-based accessibility could compromise genomic stability if the sensor proteins MRE11, Ku70/80, and PARP1 exceed this size limit. For example, a complex of Ku70/80 will be 150 kDa (Walker et al., 2001), three times the mass of a typical transcription factor (Maeshima et al., 2015) and has an approximate height and width of 7 and 12 nm (Rivera-Calzada et al., 2007). Thus, it is important to understand if there is significant size-dependence in chromocenter accessibility and, if there is, how this relates to sizes of DNA damage sensing and repair machinery.

DNA double-strand breaks (DSBs) in pericentric heterochromatin are repaired primarily by the homologous repair (HR) pathway or non-homologous end joining (NHEJ) pathway in a cell cycle-dependent manner (Tsouroula et al., 2016). In S and G2, DSBs relocate to the periphery of chromocenters and undergo HR, but in G1, DSBs remain in the chromocenter core and undergo NHEJ (Tsouroula et al., 2016). This suggests that proteins involved in NHEJ repair do not have difficulty accessing the interior of chromocenters. However, chromatin decompaction has been proposed as necessary for DSB repair in heterochromatin (Ayoub et al., 2008; Goodarzi et al., 2009; Noon et al., 2010). The relaxation of chromatin structure in response to DNA double-strand breaks could be a requirement for this accessibility.

Beyond to the potential of molecular size to restrict accessibility to chromocenters, numerous recent studies suggest that chromocenters behave as phase-separated compartments (Larson et al., 2017; Strom et al., 2017, 2021; Larson & Narlikar, 2018; Strickfaden et al., 2019; Wang et al., 2019; Erdel et al., 2020), which could provide an alternative mechanism for reducing the concentration of a diffusing protein below that of the surrounding nucleoplasm. Phase separation is emerging as a mechanism to generate membraneless compartments contributing to the subcellular organization of biomolecules. Phase separation occurs when molecules in a solution capable of multivalent interactions reach a critical concentration and undergo unmixing from the solvent to form a stable microenvironment termed a condensate (Boeynaems et al., 2018). By this mechanism, molecules that favourably interact with the environment of the condensate may enter freely, but molecules with unfavourable interactions will be depleted. Interestingly, some small inert proteins and molecules, including EGFP (Bancaud et al., 2009) and the YFP trimer construct (89 kDa) used by Strom et al. (2017), are depleted from chromocenters relative to the surrounding nucleoplasm despite EGFP being smaller than most transcription factors with a molecular weight (MW) of 27 kDa, a diameter of 2.4 nm, and a length of 4.2 nm (Hink et al., 2000) This is unlikely to be explained simply by differences in density and increased volume exclusion in the chromocenters since the measured density of chromocenters is 208 mg/mL while the surrounding nucleoplasm measures 136 mg/mL (Imai et al., 2017) in living cells.

In this study, we examined the chromocenter concentrations and diffusion of multiple DNA double-strand break sensor, mediator, and effector proteins in living murine cells without DNA damage using fluorescent protein-tagged transfected proteins. This informs us about the ability of proteins involved in sensing and repairing DNA through both the NHEJ and HR pathways to access the interior of chromocenters. We compared the relative nucleoplasmic and chromocenter concentrations and measured diffusion coefficients of selected proteins to determine if accessibility or diffusion rates within chromocenters are directly correlated with apparent molecular weight. We find that there is no clear relationship between molecular weight and the extent of depletion within chromocenters. Nonetheless, DDR proteins did show substantial differences in concentration within chromocenters, and many showed a large range of concentrations within cell populations. Chromatin density alone did not explain the depletion of proteins from chromocenters. When we compared the accessibility of EGFP in living cells with recombinant GFP perfused into fixed cells, only the living cells showed depletion of EGFP relative to the surrounding nucleoplasm. The sensor proteins PARP1 and Ku70 were typically enriched in chromocenters, while MRE11 distribution varied between cells with individual cells found to be enriched, depleted, or evenly distributed across the cell population. Since MRE11 can bind PAR (Haince et al., 2008), we tested the effect of PARP1/2 inhibition and found no impact on the distribution of MRE11, indicating that this was not due to an increase in DNA damage or PAR accumulation in chromocenters. Importantly, all proteins examined showed some ability to access the interior of chromocenters demonstrating that DNA damage-associated chromatin decondensation is not required for large DDR proteins to have access to nuclear heterochromatin domains.

The condensed state of chromatin in chromocenters has long been thought to contribute to defining the accessibility of molecules. Heterochromatin has increased mutation rates compared to euchromatic regions (Schuster-Böckler & Lehner, 2012). One suggestion to explain this is decreased repair due to lower rates of transcription-coupled nucleotide excision repair (TCR), but another is the attenuated ability of the repair machinery to access heterochromatic regions (Schuster-Böckler & Lehner, 2012; Roberts & Gordenin, 2014). Indeed, H2AX phosphorylation (γH2AX), which is dependent on the large Ataxia Telangiectasia Mutated (ATM) kinase, is reduced in heterochromatic regions in both yeast and mammalian cells (Kim et al., 2007). The concept of chromatin-mediated regulation of accessibility was originally established based on the differential digestion kinetics of active and inactive genes (Weintraub and Groudine, 1976). The association of DNase I sensitivity with the acetylation state of chromatin domains further implicated chromatin folding in regulating genome accessibility (Hebbes et al., 1994) and is consistent with experiments examining the ability of different sized fluorescent dextrans to diffuse into chromatin of differing density (Görisch et al., 2005). Current models propose that chromatin is compacted by interactions with itself (Hansen et al., 2021), and compaction may be further facilitated by proteins and RNA (Thakur et al., 2019; Wang et al., 2019; Fan et al., 2020; Li et al., 2020). Recent studies into the material states of chromatin indicate that chromatin exists in a gel (solid) state (Erdel, 2020; Strickfaden et al., 2020; Hansen et al., 2021). This predicts the existence of pores between crosslinked chromatin fibres. Thus, we might expect that steric hindrance in condensed regions of chromatin could participate in genome regulation. Consistent with this, an inverse relationship between size and localization to constitutive, facultative, and euchromatic regions of mouse nuclei was recently reported, and larger transcriptional complexes were found to be absent in constitutive heterochromatin regions of fixed cells (Miron et al., 2020).

When we examined the distribution of DNA break sensor proteins PARP1, MRE11, and Ku70, we observed variable behaviour for MRE11 and enrichment for both PARP1 and Ku70. In the case of Ku70, the Ku70/80 heterodimer is 11 × 7 × 7 nm and would be expected to have difficulty entering mouse chromocenters if they had a pore size of only 10 nm. PARP1 was significantly enriched in mouse chromocenters. Ku and PARP1 have previously been reported to associate with heterochromatin (el Ramy et al., 2009; Quénet et al., 2008; Song et al., 2001). Accumulation within chromocenters is expected to reflect binding to chromatin within chromocenters, although differential solubility in distinct solvent environments could also explain the enrichment of any of these proteins. In either case, the observed ability of these proteins to diffuse into chromocenters would require that the gel be sufficiently porous to enable these protein complexes to diffuse through their interior. Thus, the pore size must exceed 10 nm diameter.

The case of MRE11 is particularly interesting and, together with the partitioning of NBS1, suggests that the concentrations of diffusing nucleoplasmic proteins are regulated in some manner. MRE11 can be depleted, homogeneous, or enriched within mouse chromocenters. This enrichment was not due to poly(ADP-ribose) as PARP inhibitors did not affect MRE11 distribution. MRE11 forms a complex with NBS1 and Rad50. NBS1 was consistently depleted and to a greater extent than MRE11. This suggests that either the complex is excluded more than MRE11 alone or that these proteins show differential regulation of chromocenter concentration without complex assembly. Importantly, the results indicate that the initial detection of DNA damage within chromocenters is not limited by steric exclusion of sensor proteins since they all show some access to the interior of the mouse chromocenter.

Further supporting the potential physical accessibility barrier to large molecules like ATM, DSB repair in heterochromatin is reported to depend upon chromatin decondensation mediated by the phosphorylation of KAP1 (Goodarzi et al., 2008). Similarly, PARP activation will drive chromatin decondensation, recruitment of chromatin remodelling complexes, and histone displacement (Poirier et al., 1982; Rouleau et al., 2004; Liu & Yu, 2015; Strickfaden et al., 2016). Thus, it may be that some of the mediator and effector proteins involved in DSB repair require chromatin remodelling to increase the porosity of heterochromatin to function there. Work from Tsouroula et al. (2016) establishes that during S/G2, DSBs occurring within chromocenters are relocated to the periphery of chromocenters to undergo repair by HR. In addition, Rad51 assembly occurs on the periphery of chromocenters. In comparison, independent of the cell cycle, DSBs repaired by NHEJ remain within the chromocenter (Tsouroula et al., 2016). Thus, it was of particular interest to assess the ability of ATM and similarly large downstream proteins such as MDC1, 53BP1, and BRCA1 to diffuse into the chromocenter interior. While we did observe that these larger proteins showed greater depletion from mouse chromocenters, in no case did we see evidence for complete exclusion from chromocenters. For example, ATM exists as a dimer in the absence of DNA damage. A monomer of ATM has an approximate height of 20 nm and width of 10 nm (Xiao et al., 2019). The chromocenters are no more than 40% depleted in ATM relative to the surrounding nucleoplasm, with a mean value of 67%. This indicates that significant quantities of even these larger DNA damage response proteins enter chromocenters.

To determine if the movement of the DNA damage response proteins through chromocenters correlated with expected size, we performed fluorescent correlation spectroscopy (FCS) to measure diffusion coefficients across a size range of the proteins. We found that although GFP had an expected small reduction in diffusion within chromocenters. Amongst the other proteins tested, there was not a strong correlation between predicted size and diffusion. For example, 53BP1-EGFP and RNF168-EGFP have the same mean diffusion coefficient in chromocenters. This could reflect the assembly of complexes for RNF168-EGFP. The difference in diffusion coefficient relative to EGFP for 53BP1 is approximately two-fold, while the difference in the predicted size of monomers is approximately 10-fold. Thus, this is close to the expected difference (eight-fold) in mass to account for the reduced diffusion of 53BP1-EGFP relative to EGFP. Nonetheless, 53BP1 diffusion is slightly faster than what we might expect. This is particularly true when considering that the diffusion coefficient measured in the nucleoplasm is slower than in chromocenters. Two possible explanations are that 53BP1 diffuses as a dimer or oligomer in the nucleoplasm but diffuses as monomers through the chromocenters. Purified GFP in permeabilized fixed cells reflected the increased density of the environment with a slower mean diffusion coefficient in the chromocenters compared to the nucleoplasm, as expected because of the higher mass density present in chromocenters (Imai et al., 2017).

One of the most striking features of these results is the wide range in chromocenter concentrations observed for individual proteins. For example, MDC1 and 53BP1 are both depleted from chromocenters. However, we measured concentrations that ranged from 32 to 120% of the nucleoplasmic concentration for MDC1 and from 26 to 67% for 53BP1. For EGFP alone, we observed a range of 53–87% concentration relative to the nucleoplasm. Notably, we found no examples where EGFP was not depleted. However, when we examined fixed and permeabilized cells, we found that purified GFP incubated with permeabilized fixed cells showed near homogeneous distribution between the nucleoplasm and chromocenters. This argues against a volume exclusion effect dictating differences in chromocenter accessibility observed in living cells. That is, the volume occupied by chromatin, and therefore inaccessible to free GFP, is retained during fixation and permeabilization. We expect similar results between living and fixed cells if volume exclusion is responsible for reduced EGFP concentration in the chromocenters of living cells. Another possibility is that the barrier is imposed by a solvent difference arising from the presence of a phase-separated liquid compartment. This barrier would not be expected to be maintained following fixation and detergent extraction. Its removal could explain the failure to maintain a reduced concentration of GFP in the chromocenters of fixed and permeabilized cells. Supporting this interpretation is that the nucleolus is a well-established phase separated compartment in the nucleus (Brangwynne et al., 2011; Feric et al., 2016; Frottin et al., 2019; Lafontaine et al., 2021), and much less depletion was observed for recombinant GFP in the nucleoli of fixed cells relative to the striking depletion in living cells.

Several membraneless compartments in the nucleus are well established as liquid-liquid phase separated (LLPS) condensates, including the nucleolus and nuclear speckles (Boeynaems et al., 2018). Chromocenters are also considered a membraneless compartment, and recent work has established that several key factors in pericentric heterochromatin formation have phase separation capacity in vitro. The primary and most explored example of this is heterochromatin protein 1 α (HP1 α), which has been demonstrated to have phase separation capacity in vitro (Larson et al., 2017). One current model of heterochromatin formation is that HP1α, which binds H3K9 tri- and dimethylation written by SUV39H1, dimerizes and oligomerizes to bridge nucleosomes compacting chromatin and, at a critical concentration, separates into a heterochromatin phase (Larson et al., 2017). A phase separation role is also supported by work with HP1a in Drosophila (Strom et al., 2017). The histone methyltransferase KMT5C that writes H4K20 methylation marks in heterochromatin has liquid-like behaviour in chromocenters in that it exchanges freely within the chromatin but does not exchange freely with the nucleoplasm upon partial bleaching of chromocenters (Strickfaden et al., 2019). Further, another important heterochromatin protein, methyl CpG binding protein 2 (MeCP2), has been suggested to drive heterochromatin condensate formation in association with DNA, and mutations that disrupt the ability of MeCP2 to form condensates in vitro are found in patients with the neurodevelopmental disorder Rett syndrome (Li et al., 2020). Work on transcription factor kinetics with single-molecule tracking has recently demonstrated that rather than the previously bi-exponential behaviour with non-specific binding and specific binding, kinetics are better described with a third classification representing IDR-based constraint (Garcia et al., 2021a; Garcia et al., 2021b). It was found that the accumulation of glucocorticoid receptors at distinct regions in the nucleus could not be attributed solely to direct DNA binding events. Rather, there was a second subpopulation reliant on the presence of the transcription factor’s IDR through multivalent interactions consistent with an association with a liquid phase separated compartment (Garcia et al., 2021b). This result supports the idea of the chromocenter as having a liquid phase separated compartment as it supports the accumulation of proteins within the compartment based on their non-specific interaction with other compartment components.

Rad51 forms filaments when overexpressed, and free nucleoplasmic concentrations appear to be kept low because of this propensity to polymerize. Tip60, RNF168, Rap80, and NBS1 all showed the presence of numerous small domains. Interestingly, 53BP1, characterized as a protein capable of initiating liquid-liquid phase separation, formed fewer of these structures. Further studies are required to determine if these are liquid condensates and their concentration dependence. Past work has demonstrated that condensed chromatin behaves as a solid-like gel and may act as a scaffold around which phase separation can occur (Strickfaden et al., 2020) and is supported by the observation that chromatin transitions to a gel-like state upon heterochromatin domain formation during differentiation (Eshghi et al., 2021). In this model, a phase separated condensate would exist around the solid-like gel and would regulate the movement of molecules through the chromocenter (Strickfaden et al., 2020). Differential solubility in a distinct liquid nuclear microenvironment rather than physical exclusion appears to be a better explanation for the relative partitioning of DNA damage response proteins in the nucleoplasm relative to mouse chromocenters. A phase separated condensate model of chromocenters could explain why size alone does not determine the relative abundance of these proteins. In this model, the partition coefficient, reflecting the relative solubility in the two liquid phases, will define the distribution. While this merely changes how the understanding of this distribution should be pursued through mutational analysis, it puts additional demands on the analysis of all proteins that are enriched in liquid compartments. For example, it will be important to distinguish between partitioning through weak multivalent interactions that lead to preferential accumulation in a distinct solvent environment from accumulation mediated by high or low specificity binding to the chromatin, which will also result in accumulation beyond the concentration found freely diffusing within the nucleoplasm.