PML is recruited to heterochromatin during S phase and represses DAXX-mediated histone H3.3 chromatin assembly

Chromatin organization plays an essential role in establishing cellular identity and maintaining genome integrity. Although the molecular functions of many chromatin regulatory proteins have been defined using biochemical approaches, they often cannot provide insight into how protein targeting, intermolecular interactions and cell cycle progression contribute to their functions. Since many chromatin proteins are organized into discrete yet dynamic structures known as nuclear bodies (Dundr and Misteli, 2010), elucidating their assembly mechanisms could provide insight into their roles in genome regulation.

Promyelocytic leukemia nuclear bodies (PML-NBs) are doughnut-shaped structures measuring 0.1–1.0 µm in diameter, which are composed of an outer shell of PML protein that can surround over 100 different proteins, including death-domain associated protein (DAXX) and the chromatin-remodeling factor α-thalassemia/mental retardation syndrome X-linked (ATRX) (Lallemand-Breitenbach and de Thé, 2010). Although linked to transcriptional silencing, DNA damage, senescence, and anti-viral and stress responses, PML-NB function has remained enigmatic owing to the difficulty of identifying the molecular mechanisms that occur within them. However, the discovery that DAXX and ATRX regulate the deposition of the histone H3 variant H3.3 into heterochromatin (Drane et al., 2010; Goldberg et al., 2010) now suggests that one function of PML-NBs is to regulate H3.3 chromatin assembly.

Although the majority of the histone proteins are expressed during S phase to package newly replicated DNA, histone variants, including H3.3, are expressed throughout the cell cycle and incorporated into chromatin through replication-independent (RI) mechanisms (Weber and Henikoff, 2014). In addition to heterochromatin, H3.3 is also deposited into euchromatin; distinct chromatin assembly factors regulate H3.3 deposition into different genomic regions (Goldberg et al., 2010). While DAXX and ATRX regulate H3.3 deposition at heterochromatin, the HUCA complex, composed of HIRA, CABIN1, UBN1 and Asf1a, is required for H3.3 enrichment at genes (Goldberg et al., 2010; Tagami et al., 2004).

H3.3 chromatin assembly is also required for histone post-translational modification (PTM) deposition, which indicates that it regulates epigenetic inheritance. Specifically, DAXX and ATRX-mediated H3.3 chromatin assembly is required for H3K9 trimethylation (H3K9me3) at pericentromeres, endogenous retroviral (ERV) sequences, imprinted genes, intragenic methylated CpG islands (CGIs) and embryonic stem cell (ESC) telomeres (Elsasser et al., 2015; He et al., 2015; Udugama et al., 2015; Voon et al., 2015). HIRA is required for Polycomb repressive complex 2 (PRC2)-mediated deposition of H3K27 trimethylation (H3K27me3) at bi-valent genes in ESCs (Banaszynski et al., 2013).

Although H3.3 is incorporated into chromatin throughout the cell cycle, DAXX and ATRX-mediated H3.3 deposition has also been specifically linked to S phase, suggesting that it may be coordinated with DNA replication. For example, passage through S phase is required for H3.3 to be incorporated into repetitive regions, including pericentromeres (Kraushaar et al., 2013). In zygotes, DAXX and ATRX colocalize with pericentric DNA during S phase when de novo heterochromatin assembly takes place (Fulka and Langerova, 2014). Additionally, ATRX^−/− cells exhibit an increase in stalled replication forks, fragile telomeres and mitotic catastrophe, which suggests that H3.3 loading is needed to resolve the secondary structure of repetitive DNA during synthesis (Bagheri-Fam et al., 2011; Bérubé et al., 2005; Clynes et al., 2015; Clynes et al., 2014; Huh et al., 2016; Huh et al., 2012; Leung et al., 2013). Since PML-NBs are enriched in the S and G2 phases of the cell cycle (Chang et al., 2013; Dellaire et al., 2006; Everett et al., 1999; Ishov et al., 2004; Luciani et al., 2006; Verdun and Karlseder, 2007; Wong et al., 2010), it is possible that they form at this time to facilitate DAXX and ATRX-mediated H3.3 chromatin assembly.

Although there is debate about whether DNA is a component of PML-NBs (Lallemand-Breitenbach and de Thé, 2018), the finding that PML-NB proteins localize to chromatin sites, including: telomeres (Henson et al., 2002; Wong et al., 2010); the major histocompatibility (MHC) class I gene cluster (Shiels et al., 2001); satellite DNA in patients with immunodeficiency, centromeric instability and facial dysmorphy (ICF) syndrome (Luciani et al., 2006); and latent herpes simplex-1 genomic DNA (Cohen et al., 2018) supports the hypothesis that PML-NBs regulate chromatin. A significant limitation to elucidating PML-NB function in chromatin regulation, however, has been the difficulty of tracking PML-NB protein dynamics at specific genomic sites in living cells. Here, we use a system that permits live-cell visualization of a multi-copy transgene array regulated by the DAXX–ATRX–H3.3 pathway (Newhart et al., 2012) to investigate the mechanisms through which PML-NB proteins are recruited to heterochromatin and their functions in H3.3 chromatin assembly. Specifically, we show that DAXX and PML are recruited through distinct SUMOylation-dependent mechanisms. In contrast to DAXX, which associates with the array throughout the cell cycle, PML is specifically recruited during S phase and represses DAXX-mediated H3.3 chromatin deposition. Taken together, our results suggest that PML-NBs coordinate H3.3 chromatin assembly with DNA replication at heterochromatin.

To study mechanisms of mammalian chromatin regulation, we developed a system that permits direct visualization of a repetitive array (∼200 copies) of an inducible transgene in single living cells (Fig. S1A) (Janicki et al., 2004). The transgene includes lac operator repeats, which allow the integration site to be visualized when lac repressor, fused to an auto-fluorescent protein, is expressed. It also includes tetracycline response elements (TREs), which permit transcription to be induced. When this array is integrated into HeLa cells (cell line HI 1-1), DAXX and ATRX are both recruited to the site, and H3.3 is specifically enriched in the chromatin. In HeLa cells, the transgene array it is refractory to activation (Fig. S1B) (Newhart et al., 2012). However, when the array is integrated into U2OS cells, which are ATRX-null and use the alternative lengthening of telomere (ALT) pathway instead of telomerase to maintain telomere length (Lovejoy et al., 2012), H3.3 is not enriched in the chromatin and transcription can be strongly induced (Fig. S1B) (Newhart et al., 2012). Since PML is also enriched at this transgene array (Newhart et al., 2012), we can use this system to study the mechanisms through which PML-NB proteins are recruited to repetitive DNA and their contributions to H3.3 chromatin assembly.

SUMOylation is the covalent attachment of a small ubiquitin-like modifier (SUMO) protein to a target protein (Zhao, 2018). It is a post-translational modification that regulates protein targeting, including the localization of PML-NB proteins. In humans, there are four SUMO isoforms of which SUMO-1, SUMO-2 and SUMO-3 are ubiquitously expressed (Johnson, 2004). SUMO-1 is has 47% similarity to SUMO-2 and SUMO-3 and the latter are 95% identical. SUMOylation is a three-step process requiring an E1-activating enzyme, an E2-conjugating enzyme and an E3 ligase (Johnson, 2004) that catalyzes the formation of an isopeptide bond between the C-terminus of a SUMO protein and the amino group of the acceptor lysine residue (Bernier-Villamor et al., 2002; Lin et al., 2002; Sampson et al., 2001). SUMOylation can change protein interactions, often through the association of the conjugated SUMO with a SUMO interaction motif (SIM), which typically consists of a core of hydrophobic residues preceded or followed by negatively charged amino acids (Hecker et al., 2006; Song et al., 2004). Many PML-NB proteins, including PML and DAXX, are both SUMOylated and contain SIMs (Ishov et al., 1999; Lin et al., 2006; Nacerddine et al., 2005; Sahin et al., 2014a).

To determine whether SUMOylation regulates the transgene array in the HeLa cell line (HI 1-1), we first evaluated SUMO-1, SUMO-2 and SUMO-3 localization through both immunofluorescence (IF) and YFP-tagged constructs. Owing to the high degree of similarity between SUMO-2 and SUMO-3, the antibody cannot distinguish between them. Therefore, in IF and immunoblot assays, we refer to them as SUMO-2/3. Fig. 1A shows examples of cells in which endogenous SUMO-1 (panels a–c) and SUMO-2/3 (panels g–i) accumulate at the transgene array, which suggests that SUMOylation regulates the chromatin. Additionally, both YFP–SUMO-2 and YFP–SUMO-3 accumulate at the array (Fig. S1C), indicating that both isoforms function at the site. Of interest, SUMO-1 and SUMO-2/3 are not enriched at the array in every cell (Fig. 1A, panels d–f and j–l). Since an asynchronously growing population of cells was examined, this suggests that their recruitment is cell cycle regulated.

Co-staining for PML and SUMO-1 indicates that they are simultaneously present at the transgene array (Fig. 1B). Analysis of the PML (red) and SUMO-1 (green) intensities in relation to the array, marked by the lac repressor protein fused to CFP (blue; Fig. 1B, panel d), indicates that these proteins form a ‘ring’ around the DNA, which is consistent with descriptions of the ‘outer shell’ of PML-NBs (Lallemand-Breitenbach and de Thé, 2010; Lang et al., 2010; Sahin et al., 2014a). The finding that the transgene array is at the center of the PML and SUMO rings (Fig. 1A, panels b,c,h and I, and B, panels c and d) supports the hypothesis that DNA is a component of PML-NBs.

To investigate the contribution of SUMOylation to the transgene array-targeting of PML and the SUMO proteins, we evaluated their localization after knocking down UBC9, the only SUMO-conjugating E2 enzyme in the SUMO pathway (Zhao, 2018) (Fig. 2A). The efficiency of UBC9 depletion was validated by quantifying its protein (Fig. 2C) and mRNA (Fig. 2D) levels. Examples of PML, SUMO-1 and SUMO-2/3 localization at the array and in the nucleus in control and knockdown cells are shown in Fig. 2B, Figs S2 and S3. We found that UBC9 depletion eliminated the PML and SUMO rings at the array (Fig. 2B; Fig. S2A–C, panels g–i), indicating that SUMOylation is required for the ring structure to form. Our finding is consistent with a previous report showing that PML is not targeted to PML-NBs in UBC9^−/− cells (Nacerddine et al., 2005). We also found that PML, SUMO-1 and SUMO-2/3 are enriched at the array in ∼35–40% of uninfected (UI) and pLKO (non-silencing shRNA)-infected control cells (Fig. 2B), which supports the hypothesis that their recruitment is cell cycle regulated. Comparison of the levels of UBC9, PML, SUMO-1 and SUMO-2/3 by immunoblotting in UI and pLKO-infected cells indicates that the lentiviral infection does not substantially change their levels, although a slight decrease in UBC9 was detected (Fig. S1D).

We also evaluated the effect of PML depletion (Fig. 2A,C,D) on the transgene array targeting of the SUMOs. SUMO-2/3 failed to accumulate (Fig. 2B; Fig. S2C, panels j–l), which suggests that PML is the primary protein at the site to which SUMO-2 and SUMO-3 are conjugated. This result is consistent with reports that the PML in PML-NBs is conjugated to oligomerized SUMO-2 and SUMO-3 (Fu et al., 2005; Sahin et al., 2014b). Although SUMO-1 is still present at the array after PML knockdown, it does not form a ring. Instead, a ‘dot’ of SUMO-1 colocalizes with YFP–lac repressor (Fig. 2B; Fig. S2B, panels j–l). This suggests that SUMO-1 is conjugated to a protein(s) other than PML before it colocalizes with PML as a ring at the array (Fig. 1B). Since the SUMO-1 ‘dots’ are present at the array in ∼70% of PML-knockdown cells (Fig. 2B), it is likely that, in the absence of PML, SUMO-1 is blocked at a step upstream of ring formation. Since SUMO-1, unlike SUMO-2 and SUMO-3, cannot form polymeric chains (Tatham et al., 2001), it is likely that SUMO-1 requires the attachment sites provided by SUMO-2/3-conjugated PML to form a ring around the array.

To rule out the possibility that the transgene array-targeting of PML and the SUMOs is diminished in UBC9- and PML-knockdown cells (Fig. 2B; Fig. S2A–C, panels g–l) due to the loss of their expression, we measured their protein and mRNA levels (Fig. 2A,C,D). Only UBC9 and PML were eliminated by their respective shRNAs, indicating that the changes in protein targeting are due solely to PML and UBC9 depletion. Of interest, UBC9 depletion increased PML protein levels (Fig. 2A,C) and decreased the level of PML mRNA (Fig. 2D), which suggests that PML is stabilized and its mRNA diminished when SUMOylation is impaired. UBC9 depletion also increased the amount of monomeric SUMO-1 (Fig. 2A,C) and decreased the level of SUMO-1 mRNA (Fig. 2D), suggesting that when SUMO-1 cannot be conjugated, its transcripts are reduced. Additionally, PML depletion decreased UBC9 mRNA (Fig. 2D), which suggests that PML levels affect UBC9 expression. Taken together, these results suggest that PML-NB organization impacts the transcriptional and post-transcriptional regulation of its components.

To investigate the mechanisms that regulate DAXX and ATRX recruitment to the transgene array, we evaluated their localization after each was depleted (Fig. 3A,C,D). Comparison of the levels of DAXX and ATRX by immunoblotting in UI and pLKO-infected control cells indicates that the lentiviral infection does not alter their levels (Fig. S1D). DAXX and ATRX are enriched at ∼90–100% of the arrays in control cells (UI and pLKO) (Fig. 3B), suggesting that they are present at the site throughout the cell cycle. Examples of DAXX and ATRX localization at the array and in the nucleus in control and knockdown cells are shown in Fig. 3B, and Figs S4 and S5A.

Since DAXX and ATRX remain enriched at the array when the other is depleted, it is likely that they are recruited independently of each other (Fig. 3B, Fig. S4A, panels p–r, and S4B, panels m–o). This result is consistent with our previous observation showing that both endogenous and transiently expressed DAXX accumulate at the array in the ATRX-null U2OS cell line (2-6-3) (Fig. S1B) (Newhart et al., 2012). Although it has been reported that ATRX and DAXX require each other to be targeted to repetitive elements (Lewis et al., 2010), the finding that they share a limited number of binding sites throughout the genome (He et al., 2015) supports the hypothesis that they can also be independently recruited to chromatin.

Following UBC9 depletion, DAXX is not recruited to the transgene array (Fig. 3B; Fig. S4A, panels g–i), which indicates that SUMOylation is required for DAXX recruitment. Although UBC9 depletion significantly reduces ATRX accumulation at the array, it does not eliminate it (Fig. 3B; Fig. S4B, panels g–i), suggesting that SUMOylation is not required for ATRX recruitment. Since PML, SUMO-1 and SUMO-2/3 are also not present at the array when UBC9 is depleted (Fig. 2B), it is likely that they are also not required for ATRX recruitment.

DAXX and ATRX also accumulate at the transgene array following PML depletion (Fig. 3B), which indicates that PML is not required for their recruitment. This result is consistent with reports that DAXX localizes to condensed heterochromatin in PML^−/− fibroblasts (Ishov et al., 1999; Lin et al., 2006). However, it is in contrast with the finding that PML depletion prevents ATRX recruitment to mouse ESC telomeres (Chang et al., 2013). Taken together, these results suggest that PML-NB proteins associate with different genomic sites through different combinations of intermolecular interactions.

Although DAXX and ATRX do not require each other for transgene array recruitment (Fig. 3B), their depletions significantly diminish each other's protein levels (Fig. 3A,C). This result is consistent with a report that ATRX protein levels are reduced in DAXX^−/− ESCs (Lewis et al., 2010). PML depletion also diminishes DAXX and ATRX protein and mRNA levels (Fig. 3A,C,D), which suggests that PML stabilizes them and/or promotes their expression. Taken together, these results further support the hypothesis that PML-NB organization impacts the expression and stability of its components.

To understand the contribution of SUMOylation to the transgene array targeting of DAXX and PML, we mutated their SIMs in YFP-tagged constructs (Figs 4 and 5). The DAXX N-terminal and C-terminal SIMs (i.e. SIM1 and SIM2, respectively) were converted from IIVL into IGAG (denoted mSIM1 and mSIM2; Fig. 4A) (Cuchet-Lourenço et al., 2011). Similar to what is seen in wild-type cells, YFP–DAXX-mSIM1 accumulates at both the transgene array and PML-NBs, marked by isoform 4 of PML fused to the Cherry auto-fluorescent protein (Cherry–PML-IV) (Fig. 4B, panels a–h), which indicates that SIM1 is not required for DAXX to be targeted to these sites. Mutation of SIM1 does, however, reduce DAXX intensity at the array (Fig. 4C), suggesting that SIM1 reinforces its association. Although DAXX-mSIM2 still accumulates at the array, it does not colocalize with Cherry–PML-IV in PML-NBs (Fig. 4B, panels i–l), consistent with previous reports (Cuchet-Lourenço et al., 2011; Ishov et al., 1999; Lin et al., 2006; Ryu et al., 2000; Sahin et al., 2014a). Additionally, the intensity of DAXX–mSIM2 at the array is further reduced compared to DAXX–mSIM1 (Fig. 4C), which indicates that SIM2 plays a significant role in targeting DAXX to heterochromatin (see Fig. 8A). Since simultaneous mutation of both SIMs (i.e. YFP–DAXX-mSIM1/2) does not fully abrogate DAXX recruitment (Fig. 4B, panels m–p, C) but UBC9 depletion does (Fig. 3B), it is likely that other DAXX domains and/or its SUMOylation sites (Lin et al., 2006) contribute to heterochromatin targeting.

We also evaluated the impact of mutating the PML-IV SIM (VVVI>VGGG; denoted PML-IV-mSIM) (Cuchet-Lourenço et al., 2011) and SUMOylation sites (K160R, K490R and K616R, denoted PML-IV-3KR) (Fig. 5A) on transgene array targeting (Fig. 5B, panels a–d). Since PML-IV-mSIM (Fig. 5B, panels e–h), PML-IV-3KR (Fig. 5B, panels i–l) and PML-IV-mSIM-3KR (Fig. 5B, panels m–p) are all targeted to both the array and PML-NBs, marked by Cherry–PML-IV, it indicates that these motifs are not required for recruitment. These results are consistent with reports that PML requires tetramerization of its RING domain, not its SIM and SUMOylation sites, for PML-NB localization (Sahin et al., 2014a; Wang et al., 2018).

To further investigate the mechanism through which PML is recruited to the transgene array, we stably expressed YFP–PML-IV in HeLa HI 1-1 cells and determined the timing of its recruitment to the transgene array. By tracking the cells through two divisions (Movies 1–5), we confirmed that they are able to divide under the imaging conditions and determined that they proceed from G1 to G2 in 30.3±4.3 h (mean±s.d.; n=23; Table S1). The HeLa cell cycle has previously been reported to be ∼22 h long and to be divided into the following phases: G1, 12 h; S phase, 7 h; and G2, 3 h (Posakony et al., 1977). Based on this information, we estimate the timing of the cell cycle phases in HeLa HI 1-1/YFP–PML-IV cells to be as shown in Fig. 6A.

To image YFP–PML-IV recruitment to the array, we transiently expressed Cherry–lac repressor in the HI 1-1/YFP–PML-IV cells (Movies 6–11). Under these conditions, we found that the array did not segregate properly to the daughter cells, which suggests that the high affinity lac repressor–operator interaction (Sasmor and Betz, 1990) interferes with chromosome separation during mitosis. Specifically, we found that the array either segregated to only one daughter cell (Table S1, Movies 8–10, Fig. S8) or was extruded from the nucleus (Movies 6 and 7; Figs S6 and S7), and, for 33.3% of cells, one or both daughter cells died after cell division (Movie 7). Therefore, to determine the timing of PML-IV recruitment to the transgene array, we counted back from metaphase (M) (Fig. 6B,C; Movies 6–11, Figs. S6–S8). YFP–PML-IV was initially recruited to the array ∼10 h before metaphase, with its signal intensity increasing in a linear manner and peaking 7.8±0.8 h (n=9) before metaphase (Fig. 6B and C, panel b; Table S1), which is consistent with recruitment beginning in S phase. The increase in Cherry–lac repressor intensity over this period of the cell cycle (Fig. 6B) could be due to an increase in the numbers and/or accessibility of the lac operator sequence as a result of DNA synthesis.

To confirm that the PML rings form at the array during S phase, we synchronized the cells and localized PML and SUMO-1 in relation to YFP-tagged proliferating cell nuclear antigen (YFP–PCNA), which colocalizes with DNA replication foci (Leonhardt et al., 2000). Both the PML (Fig. S5B, panels c,d) and SUMO-1 rings (Fig. 6D, panels c,d) are present at the array when PCNA is both enriched at the site and in a replication pattern throughout the nucleus. However, when PCNA is diffusely distributed in the nucleus, neither PML (Fig. S5B, panels a,b) nor SUMO-1 (Fig. 6D, panels a,b) encircles the array, which supports the hypothesis that the PML and SUMO rings form during S phase.

Since SUMO-1 colocalizes with PML at the transgene array, both spatially (Fig. 1B) and temporally (Fig. 2B), we used it to quantify PML/SUMO ring formation in synchronized cells (i.e. G1/S; S phase, 2 h; S phase 7 h; and G2/M) (Fig. 6E). In ∼90% of cells fixed 7 h after release into S phase, SUMO-1 encircled the array and PCNA was both enriched at the site and in a replication pattern in the nucleus, which further supports the conclusion that the PML/SUMO rings form during S phase. Although SUMO-1 transgene array rings were also detected in cells fixed and stained at the other time points, PCNA was not simultaneously in a replication pattern suggesting that they were not in S phase. As stress has been reported to induce PML-NB formation (Sahin et al., 2014a), it is possible that the stress associated with expression of lac repressor or the failure of the array to segregate properly during mitosis (Table S1) may account for the SUMO-1 that localizes to the array outside of S phase.

To further confirm that the PML transgene array rings form during S phase, we used 5-ethynyl-2′-deoxyuridine (EdU) (30 min pulse) to label newly synthesized DNA in HeLa HI 1-1 cells expressing both CFP–lac repressor and YFP–PML-IV. Fig. 6F shows that YFP–PML-IV encircles the array during DNA synthesis, which supports the conclusion that PML accumulates at the array during S phase. It is also consistent with reports that PML-NBs are enriched during the S phase (Chang et al., 2013; Dellaire et al., 2006; Everett et al., 1999; Ishov et al., 2004; Luciani et al., 2006; Verdun and Karlseder, 2007; Wong et al., 2010).

To investigate the function of PML-NB proteins in H3.3 chromatin assembly, we knocked them down and measured H3.3 levels at the transgene array by performing chromatin immunoprecipitation (ChIP) (Fig. 7A). First, we used a GFP-specific antibody to immunoprecipitate stably expressed H3.3–YFP (from HI 1-1/H3.3-YFP cells) (Fig. 7C) as previously described (Newhart et al., 2016). The metaphase spread confirms that H3.3–YFP is incorporated into the chromatin (Fig. 7B). We also used an H3.3-specific antibody to immunoprecipitate endogenous H3.3 (Fig. 7D). H3.3 is specifically enriched in the lac and TRE repeats (Fig. 7A, primer pairs 1 and 2; Fig. 7C,D), which is consistent with a report showing that DAXX and ATRX regulate H3.3 incorporation into simple repeats (He et al., 2015). DAXX knockdown significantly decreased H3.3 levels at these sites, which indicates that it is required for H3.3 to be deposited (Fig. 7C,D). UBC9 depletion similarly decreased H3.3 incorporation (Fig. 7C,D). Since UBC9 is required for DAXX to be recruited to the array (Fig. 3B), it is likely that its essential function in H3.3 chromatin assembly is DAXX targeting.

In contrast to DAXX and UBC9, ATRX depletion did not affect H3.3 incorporation into the transgene array (Fig. 7C,D). Although this result is consistent with our previous analysis (Newhart et al., 2016), it is in contrast with reports that ATRX is required for H3.3 incorporation into numerous genomic sites, including telomeres in mouse ESCs (Goldberg et al., 2010; Levy et al., 2015; Wong et al., 2010). It is possible that H3.3 levels are maintained at the array in the absence of ATRX because DAXX, which has intrinsic histone deposition activity (Lewis et al., 2010), is still present at the site (Fig. 3B). It is, therefore, also possible that the loss of H3.3 chromatin assembly, at other chromatin sites, following ATRX depletion is due to the loss of DAXX targeting.

In contrast to DAXX and UBC9, which are required for H3.3 incorporation at the lac and TRE repeats, PML depletion increased H3.3 deposition at these sites (Fig. 7C,D), which suggests that it functions as a repressor. To test this hypothesis, we measured H3.3 levels after co-depleting PML and DAXX (Fig. 7E). We found that the increase in H3.3 deposition is abrogated by PML and DAXX co-depletion (Fig. 7F), which supports the conclusion that PML antagonizes DAXX-mediated H3.3 chromatin assembly.

Since DAXX and ATRX-mediated H3.3 chromatin assembly is required for H3K9me3 deposition (Voon and Wong, 2016), we also evaluated the effects of PML-NB protein depletion on H3K9me3 levels at the array (Fig. 7G). In contrast to H3.3, which is specifically enriched at the lac and TRE repeats (Fig. 7C,D, primer pairs 1 and 2), H3K9me3 is distributed across the transgene (Fig. 7G, primer pairs 1–5). This suggests that, if H3K9me3 is initially co-deposited with H3.3 at the repeats, it subsequently spreads, perhaps through the interactions of the H3K9-methyltransferase chromodomains with H3K9me3 (Grewal and Jia, 2007).

Indeed, PML-NB protein depletion significantly impacted H3K9me3 deposition at the transgene array (Fig. 7G). However, in contrast to H3.3, the changes were relatively small and not restricted to the repeat regions; changes in H3K9me3 were also seen at the promoter (Fig. 7G, primer pair 3) and in the gene body (Fig. 7G, primer pairs 4 and 5) of the transcription unit (Fig. 7A). This suggests that dysregulation or loss of DAXX-mediated H3.3 chromatin assembly at simple repeats can alter histone PTM levels in adjacent regions. Similar to their effects on H3.3 incorporation (Fig. 7C,D), DAXX depletion decreased (Fig. 7G, primer pairs 2, 3 and 4) and ATRX depletion did not affect (Fig. 7G) H3K9me3 levels.

In contrast to its effect on H3.3 (Fig. 7C,D), UBC9 depletion increased H3K9me3 (Fig. 7G, primer pairs 2, 3 and 5), which suggests that the array becomes more heterochromatic in the absence of both PML-NB proteins (Figs 2B and 3B) and H3.3 deposition (Fig. 6C,D). This result is consistent with the finding that H3.3 and PML-NB proteins are enriched at ESC telomeres but not at somatic cell telomeres, which have higher levels of H3K9me3 and DNA hypermethylation (Goldberg et al., 2010; Marion et al., 2009; Wong et al., 2009). Also, in contrast to its effect on H3.3 (Fig. 7C–F), PML depletion decreased H3K9me3 levels (Fig. 7G, primer pairs 1–5), which suggests that its role in antagonizing DAXX-mediated H3.3 chromatin assembly is important for maintaining H3K9me3. This result is consistent with reports that H3K9me3 levels are reduced in PML^−/− cells (Delbarre et al., 2017).

Here, we use a transgene array that can be visualized in single living cells to investigate (1) the mechanisms that recruit PML-NB proteins (i.e. PML, SUMO-1, SUMO-2/3, DAXX and ATRX) to heterochromatin and (2) their functions in H3.3 chromatin assembly. We show that PML, SUMO-1 and SUMO-2/3 form a ring around the array that is consistent with descriptions of the ‘outer shell’ of PML-NBs (Fig. 1) (Lallemand-Breitenbach and de Thé, 2010; Lang et al., 2010; Sahin et al., 2014a) supporting the hypotheses that DNA is a component of PML-NBs and that PML-NBs regulate chromatin.

We also demonstrate that the PML and SUMO rings assemble at the array through a series of SUMOylation-regulated steps (Fig. 2B). Specifically, the PML and SUMO rings do not form when UBC9 is depleted, which indicates that SUMOylation is required. The SUMO-2/3 ring also does not form when PML is depleted, which suggests that PML is the primary protein to which SUMO-2 and SUMO-3 are conjugated within these structures. Likewise, SUMO-1 does not form a ring when PML is depleted, indicating that SUMO-2/3-conjugated PML is required. However, a ‘dot’ of SUMO-1 still accumulates at the array in the absence of PML, which indicates that SUMO-1 is conjugated to a protein(s) other than PML before forming a ring. Notably, PML begins to accumulate at the array during S phase, suggesting that its recruitment signal materializes at the onset of DNA replication (Fig. 6A–C).

In contrast to PML and the SUMOs, DAXX and ATRX are enriched at the array throughout the cell cycle. DAXX and ATRX are also each recruited to this site when each other is depleted, as well as when PML is depleted, which indicates that they are targeted independently of each other as well as independently of PML (Fig. 3B). The finding that PML is only present at the array during the S and G2 phases of the cell cycle also supports the conclusion that PML is not required for DAXX and ATRX recruitment, consistent with a report that DAXX and ATRX are enriched at condensed heterochromatin in PML^−/− cells (Ishov et al., 1999). Although ATRX still accumulates at the array following UBC9 depletion, DAXX does not, which indicates that, like PML, DAXX requires SUMOylation to be targeted (Fig. 3B).

Analysis of constructs with SIM mutations indicates, however, that DAXX and PML have distinct SUMOylation-mediated recruitment mechanisms (Fig. 8A). Mutation of the DAXX C-terminal SIM (i.e. SIM2) prevents PML-NB targeting and significantly reduces transgene array accumulation (Fig. 4). This suggests that the interaction between SIM2 and a currently unknown SUMOylated protein(s) is a significant regulator of DAXX nuclear organization and heterochromatin targeting. Since UBC9 depletion (Fig. 3B), but not simultaneous mutation of both its SIM domains (Fig. 4), fully abrogates DAXX accumulation at the transgene array, it is likely that other DAXX domains and/or its SUMOylation sites (Lin et al., 2006) are also required for heterochromatin targeting.

Although UBC9 depletion also fully abrogates PML transgene array targeting (Fig. 2B), the simultaneous mutation of the SIM and SUMOylation sites in PML-IV did not prevent it from accumulating at either PML-NBs or the transgene array (Fig. 5). This indicates that PML is not targeted to these sites via a direct SIM–SUMO interaction. Instead, it is likely that PML requires the presence of proteins that are themselves targeted by SIM–SUMO interactions (Fig. 8A). This finding is also consistent with reports that PML requires tetramerization of its RING domain, and not its SIM or SUMOylation sites, to be targeted to PML-NBs (Sahin et al., 2014a; Wang et al., 2018).

The idea that DAXX and PML have distinct SUMOylation-dependent targeting mechanisms is also supported by the finding that they are recruited to the transgene array at different times. DAXX is enriched at the site in ∼90% of cells, which suggests that its recruitment signal is present throughout the cell cycle. In contrast, PML is recruited during S phase and remains associated through G2. Since PML colocalizes, both temporally and spatially, with the SUMO proteins in the rings, it also suggests that they do not recruit DAXX to the array. Therefore, the function of the PML/SUMO rings may be to concentrate regulatory factors at chromatin through the formation of biomolecular condensates (Ditlev et al., 2018) or to regulate the ubiquitin-mediated degradation of PML (Boutell et al., 2003; Lallemand-Breitenbach et al., 2001).

Consistent with reports that DAXX is essential for H3.3 deposition at heterochromatic repeats (Drane et al., 2010; Lewis et al., 2010), we also found that DAXX depletion dramatically reduces H3.3 levels at the lac and TRE repeats in the transgene (Fig. 7C,D). Based on previous reports that ATRX is essential for H3.3 chromatin assembly (Voon and Wong, 2016), it is surprising that its depletion did not also decrease H3.3 levels at the array (Fig. 7C,D). However, DAXX, which has intrinsic histone deposition activity (Lewis et al., 2010), is still recruited in the absence of ATRX, which suggests that DAXX is able, on its own, to maintain H3.3 levels over the course of the experiment.

Although loss of ATRX is a hallmark of cancer cells that use the ALT pathway to maintain telomere length (Heaphy et al., 2011; Lovejoy et al., 2012), our finding that ATRX depletion did not noticeably alter the chromatin features of the HeLa cell transgene array is consistent with reports that ATRX loss, by itself, is insufficient to induce ALT (Clynes et al., 2014; Episkopou et al., 2014; Hu et al., 2016; Napier et al., 2015; O'Sullivan et al., 2014). However, when the transgene array was integrated into the ATRX-null ALT-positive U2OS cell line (Newhart et al., 2012), H3.3 was not enriched in the chromatin and transcription could be activated (Fig. S1B). Although we do not know all of the changes required for heterochromatin organization to be disrupted in U2OS cells, this result suggests that loss of ATRX is a contributing factor. Since ALT can be induced by co-depleting DAXX and ATRX, in combination with inducing telomere-specific damage and inhibiting telomerase (Hu et al., 2016), or by co-depleting the histone chaperones, Asf1a and Asf1b (O'Sullivan et al., 2014), it is likely that transcriptional silencing is lost as a result of changes that significantly disrupt histone management (O'Sullivan et al., 2014).

We also found that DAXX-mediated H3.3 chromatin assembly is regulated by SUMOylation. Specifically, UBC9 depletion both prevents DAXX recruitment to the array (Fig. 3B) and decreases H3.3 incorporation (Fig. 7C,D), which suggests that its essential function in H3.3 chromatin assembly is DAXX targeting. In contrast to DAXX and UBC9, PML depletion increased H3.3 deposition (Fig. 7C,D). Since this effect is abrogated by PML and DAXX co-depletion (Fig. 7F), it supports the hypothesis that PML antagonizes DAXX-mediated H3.3 chromatin assembly (Fig. 8B). Furthermore, our finding that PML is recruited to the transgene array during S phase suggests that, at heterochromatin, DAXX-mediated H3.3 chromatin assembly occurs during DNA synthesis (Fig. 8B). This idea is also consistent with a report that inhibiting DNA synthesis blocks H3.3 turnover at heterochromatin (Kraushaar et al., 2013).

Following DAXX, PML and UBC9 depletion, we also detected small but significant changes in H3K9me3 levels across the transgene (Fig. 7G), which supports the conclusion that PML-NBs regulate epigenetic inheritance at heterochromatin. DAXX depletion decreased H3K9me3, which is consistent with reports that DAXX-mediated H3.3 chromatin assembly is essential for its deposition (Voon and Wong, 2016). In contrast, UBC9 depletion increased H3K9me3 levels. Since UBC9 is required for DAXX, PML, the SUMO proteins and H3.3 to accumulate at the array site it is likely that a more-heterochromatic state is established in their absence. This result is consistent with the finding that ESC telomeres, which are enriched with H3.3 and PML-NB proteins (Chang et al., 2013; Goldberg et al., 2010; Wong et al., 2009), have lower levels of H3K9me3 compared to somatic cell telomeres, which have DNA hypermethylation and H4K20 tri-methylation (H4K20me3) and no longer accumulate PML-NB proteins and H3.3 (Goldberg et al., 2010; Marion et al., 2009; Wong et al., 2010; Wong et al., 2009). Taken together, these results support the hypothesis that DAXX-mediated H3.3 chromatin assembly is a malleable silencing mechanism that regulates epigenetically dynamic chromatin, including ESC telomeres (Voon and Wong, 2016) and latent viruses (Cohen et al., 2018).

Notably, PML depletion decreased H3K9me3 levels along the transgene (Fig. 7G). Although this is consistent with a report that H3K9me3 is decreased in PML^−/− cells (Delbarre et al., 2017), this result is surprising given that PML depletion increases H3.3 incorporation (Fig. 7C,D), which is required for H3K9me3 deposition (Voon et al., 2015). This finding is also in contrast with a report showing that PML depletion in ESCs increases telomeric H3K9me3 (Chang et al., 2013). However, since the increase in telomeric H3K9me3 was accompanied by a decrease in H3.3 and ATRX (Chang et al., 2013), the latter of which is required for DAXX to be targeted to ESC telomeres (Lewis et al., 2010), it is possible that the increase in H3K9me3 is due to the loss of DAXX-mediated H3.3 chromatin assembly, similar to the changes that occur at telomeres during differentiation (Marion et al., 2009).

Although we do not know how PML both represses H3.3 deposition and maintains H3K9me3 levels at the transgene array, our finding that PML is recruited to the site during S phase suggests that it does so by modulating DAXX-mediated H3.3 chromatin assembly during DNA synthesis (Fig. 8B). During DNA replication, H3–H4 tetramers do not dissociate and the rate of H3.1–H4 tetramer splitting is very low (Xu et al., 2010). Therefore, it is possible that PML represses H3.3 deposition during S phase by preventing DAXX from replacing H3.1 and H3.2 in the preexisting tetramers with newly synthesized H3.3 (Fig. 8C). It is thought that histone PTM patterns in heterochromatic regions are maintained through cell division by copying them from pre-existing H3–H4 tetramers onto newly assembled nucleosomes (Serra-Cardona and Zhang, 2018; Xu et al., 2010). If, in the absence of PML, DAXX is able to replace the H3 in the preexisting tetramers with new H3.3, it could also explain how PML depletion simultaneously increases H3.3 and reduces H3K9me3 deposition (Figs 7C,D,G and 8C).

We also do not know how PML is recruited to the transgene array during S phase. However, we recently reported that H3.3 colocalizes with RNA and RNA-regulatory proteins at the transcriptionally activated array, which suggests that H3.3 is recruited to its incorporation sites by a transcriptional signal (Newhart et al., 2016; Newhart et al., 2013; Shastrula et al., 2018). It is, therefore, possible that PML is also recruited to the transgene array by the H3.3 recruitment signal and that due to the compact structure of heterochromatin, DAXX-mediated H3.3 chromatin assembly is only initiated during DNA synthesis when chromatin decondensation permits the transcription machinery access to the DNA template. This sequence of events could also explain why passage through S phase is required for H3.3 turnover at heterochromatin (Kraushaar et al., 2013). If transcription-mediated H3.3 chromatin assembly is the signal that recruits PML to chromatin, then the PML-NBs that can be visualized throughout the cell cycle may be chromatin sites that can be constitutively transcribed. A transcription-mediated mechanism of recruiting PML to chromatin could also explain why PML is enriched at telomeres in cells that undertake ALT and which express high levels of telomeric-repeat containing RNA (TERRA) throughout the cell cycle (Henson et al., 2002).

Taken together, these results suggest that the function of PML in modulating DAXX-mediated H3.3 chromatin assembly is an important regulatory control point in the epigenetic inheritance of heterochromatin. Our analyses also indicate that the nuclear organization of PML-NB proteins is an important regulator of their expression and stability (Figs 2C,D and 3C,D). Since H3.3, DAXX, ATRX and PML are all mutated and/or dysregulated in numerous diseases, it is important to understand the impact of their regulatory interdependencies on chromatin organization and epigenetic inheritance. Specifically, H3.3 acquires gain-of-function point mutations in pediatric glioblastomas and skeletal cancers (Lindroth and Plass, 2013; Schwartzentruber et al., 2012). ATRX is mutated in the intellectual disability syndrome for which it is named (Gibbons et al., 1995). Loss-of-function mutations in DAXX and ATRX have been identified in brain (Huether et al., 2014; Khuong-Quang et al., 2012; Liu et al., 2012; Molenaar et al., 2012), pancreatic neuroendocrine tumors (Jiao et al., 2011) and the majority of ALT-positive cancer cells (Heaphy et al., 2011; Lovejoy et al., 2012). Fusion of PML to the retinoic acid receptor α (RARA) drives acute promyelocytic leukemia (APL) (de Thé et al., 2017). Viral proteins degrade DAXX, ATRX and PML and/or disrupt their PML-NB targeting, which suggests that their organization prevents infection (Komatsu et al., 2016). Therefore, fully elucidating the function of PML-NBs in H3.3-mediated epigenetic regulation will also provide insight into disease.

The HeLa cell lines, HI 1-1 and HI 1-1/H3.3-YFP, and their growth conditions were previously described (Newhart et al., 2012). The HeLa HI 1-1/YFP-PML-IV cell line was made by transfecting YFP-PML-IV-C3 with Fugene HD transfection reagent (Promega Corporation, Madison, WI), adding 700 µg/ml of G418 (Life Technologies, Grand Island, NY) to the media, and selecting and screening individual drug-resistant colonies for YFP–PML-IV expression as previously described (Rafalska-Metcalf and Janicki, 2013). Fugene HD transfection reagent (Promega Corporation) was also used for all transient transfections. Expression of the lac repressor fusions proteins is used to validate the presence of the integrated transgene array and to confirm that the cell line is not contaminated.

YFP–, CFP– and the Cherry–lac repressor were previously described (Janicki et al., 2004; Newhart et al., 2012). YFP–SUMO-2 and YFP–SUMO-3 were made by cloning SUMO-2 and SUMO-3 into YFP-C1 (BglII/SalI). The pLU-YFP-PCNA construct was made by cloning YFP-PCNA (Leonhardt et al., 2000; a gift from Christina Cardosa, Department of Biology, Technische Universität Darmstadt, Germany) into the pLU lentiviral vector (XbaI/NheI-SalI). YFP–DAXX and YFP–DAXX-mSIM2 (the C-terminal SIM, IIVL, was converted into IGAG) were gifts of Roger Everett (MRC-University of Glasgow Centre for Virus Research, Glasgow, Scotland, UK) (Cuchet-Lourenço et al., 2011). Mutations were introduced using Agilent QuickChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA). YFP–DAXX-mSIM1 and YFP–DAXX-mSIM1/2 were made by converting the N-terminal SIM, IIVL, into IGAG. YFP-PML-IV-C3 and Cherry-PML-IV-C3 were made by cloning the cDNA from pLINGY (Cuchet et al., 2011) into YFP-C3 and Cherry-C3 (XhoI/SalI-EcoRI). YFP–PML-IV-mSIM was made by converting the SIM, VVVI, into VGGG (Cuchet-Lourenço et al., 2011). YFP–PML-IV-mSIM-3KR was made by cloning PML-IV-KK, which includes K160R and K490R (Cuchet-Lourenço et al., 2011) (a gift of Roger Everett), into YFP-C3 (XhoI/SalI-EcoRI) and introducing the SIM and K616R mutations.

Knockdowns were performed using pLKO, and shRNA against UBC9, PML, DAXX and ATRX [shUBC9, shPML, shDAXX2 (Preston and Nicholl, 2006), shATRX90 (Lukashchuk et al., 2008)] (gifts from Roger Everett) (Lukashchuk and Everett, 2010). Lentiviruses were prepared in 293-T cells as previously described (Everett et al., 2008). Briefly, 10⁶ cells were plated in a 10 cm dish and infected the next day with shRNA-expressing lentiviruses. At 24 h post infection, cells were split into 10 cm dishes (1:3). At 48 h, puromycin was added (0.5 µg/ml) and cells were analyzed at 96 h post infection. The effects of shRNAs on protein depletion was evaluated by immunoblotting, RT-PCR and immunofluorescence; those samples that did not have significant depletion were excluded from the immunofluorescence, ChIP and RT-PCR analyses.

Staining was performed as previously described (Newhart et al., 2012; Rafalska-Metcalf et al., 2010). Antibodies against the following proteins were used under pre-extraction conditions: PML (1:150; catalog no. PG-M3; Santa Cruz Biotechnology, Santa Cruz, CA); SUMO-1 (1:50; catalog no. FL-101, sc-9060; Santa Cruz Biotechnology); and SUMO-2/3 (1:500; catalog no. 8A2; Medimabs, Montreal, Quebec, Canada). For DAXX and ATRX staining, cells were first fixed for 15 min in 4% PFA in 1× PBS and permeablized for 6.5 min in 0.5% Triton X-100 in 1× PBS. Antibodies used were against DAXX (1:4000; catalog no. D7810; Sigma-Aldrich, St Louis, MO) and ATRX (1:250; catalog no. H300; Santa Cruz Biotechnology).

Western blotting was performed with antibodies against the following proteins: PML (1:500; catalog no. A301-167A, Bethyl Laboratories, Inc., Montgomery, TX); UBC9/UBE2I (1:1000, catalog no. ab33044, Abcam, Cambridge, MA, USA); SUMO-1 (1:200; FL-101, sc-9060; Santa Cruz Biotechnology); SUMO-2/3 (1:1000; 8A2; Medimabs, Montreal, Quebec, Canada); DAXX (1:1000; EMD Millipore, Burlington, MA); ATRX (1:1000; H300; Santa Cruz Biotechnology); and GAPDH (2118S; 1:5000; Cell Signaling Technology, Danvers, MA). Horseradish peroxidase (HRP)-conjugated anti-rabbit IgG (1:3000; NA934V; GE Healthcare Ltd, UK) and anti-mouse-IgG (1:3000; NA931V; GE Healthcare Ltd, UK) secondary antibodies were used. The immunoblots were incubated with chemiluminescence solution (Amersham Biosciences, Marlborough, MA, USA) for 4 min and exposed for between 30 and 120 s to Carestream Kodak Biomax films and developed. Enhanced chemiluminescence (ECL) images were analyzed using the public domain program ImageJ. Protein data was normalized to GAPDH protein levels.

RNA isolation and reverse transcription were performed as previously described (Newhart et al., 2016; Newhart et al., 2013). Briefly, RNA was purified with Trizol and Direct-zol RNA miniprep kit (Zymo Research, Irvine, CA, USA). 1 µg of purified RNA was treated with DNase I (1 µl) (Promega Corporation), and reverse transcription was performed with 50 µM random hexamer primers (Integrated DNA technologies, Skokie, IL) using an Omniscript kit (Qiagen, Germantown, MD). The following primer pairs were then used for quantitative PCR (qPCR) analysis using 7500 Fast Real-Time PCR system (Applied Biosystems, Foster City, CA): UBC9, 5′-AAAAATCCCGATGGCACGAT-3′ (For), 5′-TTCCCACGGAGTCCCTTTCTT-3′ (Rev); PML, 5′-GCCTGGAGCACACCCTGTACC-3′ (For), 5′-GGAGCTTGTGGCCAACCTGTC-3′ (Rev); SUMO-1, 5′-TCCAATGAATTCACTCAGGTTTCTCT-3′ (For), 5′-TTCCTCCATTCCCAGTTCTTTTG-3′ (Rev); SUMO-2, 5′-AAGCCCAAGGAAGGAGTCAAG-3′ (For), 5′-CCCCGCCACCTTCAAATT-3′ (Rev); SUMO-3, 5′GAATGACCACATCAACCTGAAGGT-3′ (For), 5′CGGCGTGTGCCTCTTGATC-3′ (Rev); DAXX, 5′-AAGCCTCCTTGGATTCTGGT-3′ (For), 5′-ATCATCCTCCTGACCCTCCT-3′ (Rev); ATRX, 5′-GCAACCTTGGTCGAAAGGAGT-3′ (For), 5′-GGCTCTGGGTGACAAATGTAG-3′ (Rev); and human (h)GAPDH, 5′ATGGAAATCCCATCACCATCTT-3′ (For), 5′-CGCCCCACTTGATTTTGG-3′ (Rev).

Cells were arrested at G1/S and synchronized in S phase via a double-thymidine block G2/M arrest achvieing with nocodazole (Millipore Sigma-Aldrich, Burlington, MA) according to Harper (2005).

Edu staining was performed by using the Molecular Probes Click-iT Plus EdU imaging Kit (Molecular Probes, Life Technologies, Carlsbad, CA). CFP–lac repressor and YFP–PML-IV were transiently expressed for 18 h in HeLa HI 1-1 cells growing on coverslips. The next day, 10 µM EdU was added to the media for 30 min and cells were fixed with 4% formaldehyde in 1× PBS for 15 min. Staining for EdU, using Alexa Fluor 594-conjugated picolylazide, was performed according to the manufacturer's instructions.

Confocal images of PML, SUMO-1, SUMO-2/3, DAXX and ATRX immunofluorescence staining were taken using a Leica DMI 6000 B inverted automated microscope equipped with PLAN APO 100× oil-immersion lens, using SimplePCI software (Hamamatsu, Sewickley, PA, USA), a Yokogawa CSU-10 real-time spinning disk confocal attachment with Nipkow and microlens disks, a 442 nm/70 mW diode laser for CFP protein imaging, a 514 nm/50 mW solid-state laser for YFP protein imaging and a 561 nm/25 mW diode laser for Cherry protein imaging with a Hamamatsu ORCA-AG CCD digital camera with a 1344×1024 chip. Confocal images of fixed cells were also acquired at 12 bits using a Leica TCS SP5II scanning laser confocal using a 63×/1.40 oil objective, 8× zoom and a pinhole of 1 AU according to Nyquist parameters. DAPI-labeled cells were excited with a 405 nm laser, while a multi-line argon laser at 458 or 514 nm was used to image CFP and YFP, respectively, and a 594 HeNe laser was used to image mCherry. All samples were acquired using AOBS and HyD detectors in sequence to maximize signal and minimize spectral crosstalk. Two to four line accumulations, combined with bi-directional scanning, were used to allow laser intensities to be kept low to reduce photobleaching effects. LAS software (Leica) was used for acquisition, processing and presentation.

For live-cell imaging, HeLa HI 1-1 cells stably expressing YFP–PML-IV were plated on 35-mm glass-bottom dishes (CELLview; Greiner Bio-One, Monroe, NC). For imaging the recruitment of YFP–PML-IV to the transgene array, cells were infected with Cherry–lac repressor lentivirus overnight and put on a Leica TCS SP8 WLL system with a 8000 Hz resonant scanner (Leica Microsystems, Inc., Buffalo Grove, IL). Cells were maintained in a Tokai-Hit stage-top incubation chamber (Tokai Hit, Shizuoka-ken, Japan) at 37°C and 5% CO₂. Hardware parameters were adjusted to maintain acceptable resolution while minimizing phototoxic effects, in order to preserve cell viability. Settings included use of the resonant scanner, HyD detectors and 8-line accumulations. The white light laser was used at the following setting: for YFP proteins, 514 nm, 0.2% output; for mCherry proteins 587 nm, 0.05% output. Leica's automatic focus control (AFC) maintained z-localization, and image stacks were captured with a 40×/1.30 oil objective and 2× zoom at 10–20 locations in each sample. Each stack was 10 µm thick on average and comprised six sections taken in 2 µm steps, using a pinhole of 3 AU. Time series images were acquired over 48–72 h, with 30 min sampling intervals, totaling up to 144 time points at each location. Leica's LAS software was used for acquisition, processing and presentation of still images. Post-acquisition processing of the time series included deconvolution using Huygens software (Scientific Volume Imaging, Laapersveld, Hilversum, The Netherlands). Maximum projection, brightness and contrast adjustments, time stamps and AVI file export was done using ImageJ. AVI files were converted to MP4 files using Movavi Video Suite 16 (Limassol, Cyprus). Intensity profiles were produced using both Leica's LAS software and NIS Elements (Melville, NY).

HeLa HI 1-1/H3.3-YFP cells were cultured for 24 h in fresh medium. Metaphase arrest was achieved by incubating cells in 0.1 µg/ml of KaryoMAX Colcemid Solution (Thermo Fisher Scientific, Waltham, MA) for 4 h. Cells were trypsinized, washed with PBS, resuspended in 75 mM KCl hypotonic solution and incubated at 37°C for 30 min. Cells were immediately fixed by dropwise addition of 10 ml fixing solution (3:1 methanol to acetic acid). Fixed cells were dropped onto ice-cold microscope slides and washed three times with fixing solution and allowed to dry for 1 h in a humid chamber. The metaphase spreads were then fixed in 4% formaldehyde (in PBS) for 3 min and stained with 0.1 µg/ml DAPI for 5 min. The slides were mounted with mounting media and images were taken using 100× lens on a Nikon E600 Upright microscope (Nikon Instruments, Inc., Melville, NY). Image-Pro Plus software (Media Cybernetics, Rockville, MD) was used for capturing and processing of images.

Native ChIP was performed in HeLa HI 1-1/H3.3-YFP cells as previously described (Newhart et al., 2012) using a 1 µg/reaction of anti-GFP antibody (catalog no. 11814460001, Sigma-Aldrich) and protein A+G agarose/salmon sperm DNA (EMD Millipore, Temecula, CA). Crosslinked ChIP was performed by incubating cells with 1% formaldehyde at room temperature for 20 min followed by 0.125 M glycine for 5 min. Cells were washed with 1× PBS, pelleted, snap-frozen in liquid nitrogen and stored at −80°C. The cell pellets were lysed in Buffer 1 (50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40 and 0.25% Triton-X) for 10 min at 4°C and pelleted at 1350 g for 5 min. Cells were next incubated in Buffer 2 (10 mM Tris-HCl pH 8, 200 mM NaCl, 1 mM EDTA and 0.5 mM EGTA) for 10 min at 4°C. The pellet was resuspended in Buffer 3 (10 mM Tris-HCl pH 8, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% sodium deoxycholate and 0.5% sodium laurylsarcosine), sonicated, checked for fragmentation efficiency, and spun at 20,000 g for 10 min at 4°C. The supernatants were pre-cleared for 2 h in protein A+G agarose (Invitrogen, Carlsbad, CA) and sonicated bacterial DNA (∼2 μg/per IP). Inputs were collected and the remaining supernatant was incubated with the antibodies against H3.3 (5 µg/reaction; catalog no. 09-838, EMD Millipore, Burlington, MA) and H3K9me3 (1 µg/reaction; catalog no. ab8898, Abcam) overnight at 4°C with rotation. We recently validated the specificity of the anti-H3.3 antibody (Shastrula et al., 2018). Next day, the samples were incubated with protein A+G agarose (EMD Millipore) and bacterial DNA for 2 h at 4°C with rotation. The beads were pelleted at 650 g for 2 min at 4°C and washed 4× times with RIPA wash buffer (50 mM HEPES-KOH pH 7.5, 500 mM lithium chloride, 1 mM EDTA, 1% NP-40 and 0.7% sodium deoxycholate). DNA was eluted in 150 µl elution buffer (50 mM Tris-HCl pH 8.0, 1% SDS and 10 mM EDTA), de-crosslinked overnight and purified by using a Qiaquick PCR purification kit (Qiagen, Germantown, MD).

The sample size (n) used for all experiments is three or more biological replicates with two technical replicates for each with total n values adding up to no less than 6. Inclusion/exclusion criteria, randomization and blinded data were not used in these analyses. The statistical analysis used for these experiments was an unpaired t-test as we are comparing one variable between two groups with their mean and standard deviation.

We would like to thank Roger Everett for reagents, Christina Cardosa for the PCNA construct, Ellen Heber-Katz for critical comments on the manuscript and The Wistar Cancer Center Core Facilities for genomics, molecular screening and imaging support.