DAXX safeguards heterochromatin formation in embryonic stem cells Free

Eukaryotic nuclei generally contain two major genomic compartments, referred to as A and B, which respectively contain the active and inactive fractions of the genome. At the chromosome level, these compartments are characterized by alternating gene-rich and gene-poor domains. The B compartment is composed primarily of repetitive sequences, with satellite tandem repeats comprising one of the most abundant classes. Pericentromeric regions in mouse are composed of tandem repeats of a 234 bp sequence, named the ‘major satellite’, which constitutes ∼8 Mb on each chromosome (Consortium et al., 2002; Guenatri et al., 2004). Most genes and repeated sequences of the B compartment are transcriptionally repressed by cooperative epigenetic mechanisms leading to the formation of constitutive heterochromatin. The three- dimensional (3D) organization of constitutive heterochromatin is an important layer regulating transcription (Falk et al., 2019; Lu et al., 2021). In some species, including the mouse, the pericentromeric heterochromatin (PCH) of different chromosomes aggregates to form large DAPI-dense heterochromatin clusters, called chromocenters. Disruption of satellite clustering is often associated with increased DNA damage, and defects in chromosome segregation that are seen in pathologies such as Alzheimer's disease and breast cancer (Hahn et al., 2013; Jagannathan et al., 2018; Mansuroglu et al., 2016; Zhu et al., 2011). Gene-poor domains are preferentially found beneath the nuclear edge and are often referred to as lamina-associated domains (LADs) (Guelen et al., 2008). The tethering to the nuclear envelope is stochastic, and LADs are also found associated with other nuclear landmarks such as chromocenters and nucleoli (Koningsbruggen et al., 2010; Wijchers et al., 2015).

At the molecular level, constitutive heterochromatin exhibits covalent modifications including DNA methylation, such as H3K9 and H4K20 trimethylation (H3K9me3 and H4K20me3). Maintenance of H3K9me3 at pericentromeric satellites relies on the SUV39H1 and SUV39H2 methyltransferases, whereas a distinct H3K9me3 methyltransferase, SETDB1, operates at dispersed DNA repeats and telomeres (Fukuda et al., 2018; Gauchier et al., 2019; Martens et al., 2005; Matsui et al., 2010).

Heterochromatin maintenance is intrinsically linked to chromatin replication during S phase, but also requires the incorporation of histone variants, such as macroH2A and H3.3, into chromatin independently of DNA synthesis (Buschbeck and Hake, 2017; Mendiratta et al., 2019). The histone variant H3.3 differs from the canonical histone H3 by four or five residues and is deposited throughout the cell cycle in active and inactive chromatin by different histone chaperones. These histone-binding proteins prevent promiscuous incorporation into chromatin (Szenker et al., 2011). Together with ATRX, DAXX forms a histone chaperone complex responsible for H3.3 incorporation into heterochromatin regions, including PCH, retrotransposons and telomeres (Drané et al., 2010; Elsässer et al., 2015; He et al., 2015; Lewis et al., 2010; Sadic et al., 2015). At telomeres and retrotransposons, DAXX promotes heterochromatin formation by recruiting SUV39H1 and/or SUV39H2, or SETDB1, thereby facilitating the deposition of H3K9me3 (Elsässer et al., 2015; Gauchier et al., 2019; He et al., 2015; Hoelper et al., 2017). In addition to its role in H3.3 deposition, DAXX also prevents promiscuous incorporation of H3.3 by recruiting soluble H3.3-H4 dimers to PML nuclear bodies, membrane-less hubs for post-translational modification of proteins (Delbarre et al., 2013, 2017; Lallemand-Breitenbach and Thé, 2018).

Heterochromatin is de novo established during early embryogenesis, involving an important remodeling of its molecular composition and 3D organization (Burton and Torres-Padilla, 2014). However, the underlying mechanisms and their functional relevance for early development remain elusive. During preimplantation development, the DNA methylation inherited from parental gametes is erased and reaches its minimal level in the blastocyst, defining the ground-state of pluripotency (Leitch et al., 2013; Ying et al., 2008). This wave of DNA demethylation directly leads to the transcriptional upregulation of several repetitive sequences, a process that is essential for the remodeling of constitutive heterochromatin (Jachowicz et al., 2017; Lu et al., 2021; Probst et al., 2010). For instance, the upregulation of major satellite transcripts at the two-cell stage is required for PCH reorganization into chromocenters and normal developmental progression (Casanova et al., 2013; Probst et al., 2010). However, de-repression of DNA repeats correlates with high DNA damage signaling and their prolonged expression results in developmental arrest, suggesting that silencing through a DNA methylation-independent mechanism is required for normal embryogenesis (Jachowicz et al., 2017; Ziegler-Birling et al., 2009). Polycomb group proteins have been previously shown to bind and facilitate silencing of repetitive sequences upon DNA hypomethylation (Saksouk et al., 2014; Tosolini et al., 2018; Walter et al., 2016). However, knockout of most polycomb-related genes does not impact pre-implantation embryogenesis (Aloia et al., 2013), implying that other factors maintain heterochromatin in the absence of DNA methylation.

In mouse, both DAXX and H3.3 deletions cause genomic instability resulting in early embryonic lethality, with most knockout embryos failing to reach the blastocyst stage (Jang et al., 2015; Liu et al., 2020; Michaelson et al., 1999). DAXX binds major satellite regions during the earliest stages of development, just before the PCH reorganizes to form chromocenters (Arakawa et al., 2015; Liu et al., 2020). Although the importance of DAXX for chromocenter formation remains unknown, the viability of Daxx^−/− embryos drops after chromocenter formation (Liu et al., 2020; Probst et al., 2010), suggesting a direct link between DAXX function at chromocenters and developmental progression. Surprisingly, Daxx is not required for in vitro culture of mouse blastocyst-derived embryonic stem cells (ESCs) (Elsässer et al., 2015; He et al., 2015; Hoelper et al., 2017). In ESCs, DAXX is not recruited to PCH, but rather to retrotransposons and telomeres (Elsässer et al., 2015; Gauchier et al., 2019; He et al., 2015; Saksouk et al., 2014). However, the recruitment of DAXX to chromatin is tightly associated with the low level of DNA methylation that characterize preimplantation embryogenesis (Arakawa et al., 2015; He et al., 2015; Liu et al., 2020). ESCs are typically cultured in serum-based media and accumulate high levels of DNA methylation (Leitch et al., 2013), which could explain the discrepancies with in vivo observations. ESCs can maintain low levels of DNA methylation and reach the ground-state of pluripotency when grown in a serum-free medium, named 2i, but the role of DAXX for their survival and the formation of PCH has yet to be addressed.

Here, we decipher the role played by DAXX at PCH in pluripotent ESCs. We find that DAXX is essential for ground-state ESC survival. We provide evidence that DAXX localizes to pericentric heterochromatin and recruits PML and SETDB1, facilitating heterochromatin formation and organization through H3.3K9 modification. The deletion of Daxx or Pml impacts pericentromeric heterochromatin formation altering its biophysical and clustering properties. Thus, our results identify DAXX as an essential regulator preserving heterochromatin integrity and maintaining the viability of ground-state ESCs.

To assess the role of DAXX in pluripotent cells, we generated a Daxx knockout (KO) mouse ESC line. We targeted exon 3 of Daxx using CRISPR/cas9 technology. Although we confirmed the absence of DAXX mRNA and protein (Fig. 1A; Fig. S1A), there were no obvious changes in H3.3 protein levels in the resulting Daxx KO ESC line (Fig. 1A). Consistent with previous reports, loss of DAXX did not impact the growth of ESCs cultured in a serum-based medium (Elsässer et al., 2015). Likewise, neural differentiation induced by leukemia inhibitory factor withdrawal and retinoic acid addition did not impact on cell viability in Daxx KO ESCs (Diff., Fig. 1B). When the Daxx KO and WT ESCs were converted into the ground state, using a medium with two small kinase inhibitors and vitamin C (hereafter denoted 2iV), Daxx KO cells showed no growth defect and a similar decrease in DNA methylation after 4 days of conversion (Fig. 1B; Fig. S1B). However, prolonged culture in 2iV medium induced a drastic decrease in cell viability (Fig. 1B). Compared to the parental WT ESCs, only 12.6% of Daxx KO ESCs remained after 8 days of culture in 2iV (Fig. 1B). No surviving cells could be detected after 9 to 10 days of conversion. Altogether, these results indicate an essential role for DAXX in the maintenance of pluripotent cell survival upon ground-state conversion.

To further investigate the impact of Daxx knockout, we monitored the transcriptional landscapes under different conditions using RNA sequencing (RNA-seq). Upon differentiation with retinoic acid, our analysis reveals a rather minor differential expression of protein-coding genes between Daxx KO and WT cells, with only eight and 125 genes exhibiting up- and down-regulation, respectively (Fig. S1C). We conclude that the absence of DAXX only has a minor impact on transcription upon neural differentiation. This result is consistent with the absence of morphological or cell growth defects observed, and confirms that DAXX is dispensable for neuronal differentiation. After 4 days of 2iV conversion a substantial differential expression of genes in Daxx KO and WT ESCs was observed (Fig. S1C). To ask whether the drop in viability observed in Daxx KO ESCs upon prolonged 2iV culture was resulting from a ground-state conversion defect, we compared the expression of 166 markers previously identified as differentially expressed during this process (Blaschke et al., 2013). For both cell lines most of these markers changed expression upon conversion according to the expected pattern (Fig. S1D) supporting the conclusion that Daxx KO ESCs can reach the ground-state of pluripotency after 4 days of 2iV conversion.

Overall, our data show that Daxx is essential for the survival of ground-state ESCs but their loss of viability is unlikely to be caused by a conversion defect. As we could not identify any significant gene networks that might explain the subsequent loss of viability in Daxx KO ESCs, it is plausible that DAXX might play a role in regulating the repressed regions of the genome.

As DAXX is known to be responsible for H3.3 deposition in heterochromatic regions, we investigated the impact of Daxx deletion on the transcriptional repression of the silenced portion of the genome (Elsässer et al., 2015).

LADs make up a significant portion of the silenced compartment. To investigate the effect of Daxx deletion on the expression of genes located in the 1180 LADs described in ESCs (Peric-Hupkes et al., 2010), we compared their expression to that of genes with a more central positioning, away from the nuclear lamina (CEN). In serum and differentiated conditions, we found that LAD and non-LAD genes shared a very similar expression profile (Fig. 1C,D). However, we observed an increase in the mRNA levels of several LAD genes normally silenced such as Casp1, Casp4 and Casp12 on chromosome 9, and Naip6 and Naip5 on chromosome 13, in Daxx KO ESCs under 2iV condition (Fig. 1C). Genome-wide, the differential expression of genes located in LADs was significantly increased in the 2iV condition, suggesting that the presence of DAXX is necessary for proper silencing of these genes in ground-state ESCs (Fig. 1D).

Histone modification H3K9me2 is a hallmark of LADs and generally accumulates along the nuclear envelope (Kind et al., 2013; Peric-Hupkes et al., 2010). To assess whether the global upregulation of LAD genes observed in Daxx KO ESCs upon 2iV conversion was associated with a defect in peripheral heterochromatin assembly, we examined the accumulation of H3K9me2 by immunofluorescence. In WT and Daxx KO ESCs grown in the serum condition, the H3K9me2 signal was irregularly diffuse in the nucleoplasm with a significant accumulation at the nuclear edge (Fig. 1E). Quantification of H3K9me2 distribution by linescan density did not show any significant difference in the enrichment of signal at the nuclear periphery between WT and Daxx KO ESCs in serum (Fig. 1F; Fig. S1E). However, after 4 days of 2iV conversion, the H3K9me2 signal remained enriched at the nuclear rim in WT converted cells, but only small patches along the nuclear periphery were observed in the absence of DAXX (Fig. 1E). Linescan quantification confirmed a significant decrease in H3K9me2 enrichment at the nuclear periphery in Daxx KO ESCs grown in 2iV (Fig. 1F). In contrast, lamin B1 accumulated normally at the nuclear edge in both cell lines regardless of growing conditions, indicating no global nuclear lamina assembly defects in Daxx KO ESCs (Fig. S1F). We therefore conclude that DAXX is necessary for the maintenance of peripheral heterochromatin.

Overall, our observations show that Daxx deletion impacts the regulation of the nuclear periphery by altering the tethering of peripheral heterochromatin and the transcriptional silencing of genes located in LADs.

We additionally found that many endogenous retrovirus families were upregulated in Daxx KO ESCs in serum and 2iV (Fig. 2A), consistent with previous observations (Elsässer et al., 2015; He et al., 2015). Several LINE1 elements, notably the L1MdT and L1MdA families, not previously reported as DAXX targets, were also upregulated in pluripotent Daxx KO ESCs. The role of DAXX on transcriptional silencing was not limited to interspersed repeats, as we detected a strong upregulation of major satellites RNA in both the serum and 2iV condition in the Daxx KO cells.

To investigate whether DAXX binds to major satellites in ESCs, we used immunofluorescence with H3K9me3 staining to track PCH after confirming the strong correlation between H3K9me3 and major satellites using immuno-fluorescence in situ hybridization (FISH) (Fig. S2A,B). In both serum and 2iV conditions, DAXX signal concentrated in the nucleoplasm as round foci (Fig. 2B; Fig. S2C). The distribution of foci sizes was similar in both growing conditions, but a small fraction of larger foci could be observed under the 2iV condition. We therefore set a threshold defining small and larger DAXX foci (Fig. S2D). The large DAXX foci exhibited higher levels of DAPI and H3K9me3 specifically under the 2iV condition, supporting that DAXX can relocate to PCH in ground-state ESCs (Fig. S2E,F). To confirm this observation, we segmented nuclei and PCH foci using DAPI and H3K9me3 signals, respectively, allowing us to measure the enrichment of DAXX signal at PCH per nuclei (Fig. S2G). The level of DAXX observed at PCH upon 2iV conversion was significantly higher than in the serum condition (Fig. 2C). Notably, the increase in DAXX enrichment in 2iV was not homogenous among the H3K9me3 foci population, but rather the result of a subset of clusters showing strong DAXX enrichment, allowing us to determine a threshold to define DAXX-positive PCH (Fig. 2D). Although only 14% of cells showed one DAXX-positive PCH cluster in the serum condition, 83% of 2iV-converted cells had at least one DAXX-positive H3K9me3 focus (Fig. 2E). A small proportion of cells showed two or three, or more DAXX-enriched PCH foci (16% and 5%, respectively). Interestingly, our analysis revealed that DAXX-enriched PCH foci were bigger in size and had higher H3K9me3 staining intensities, suggesting a potential link between DAXX and H3K9me3 regulation (Fig. 2F,G).

Altogether, our data show that DAXX accumulates in a subset of PCH clusters in ground-state ESCs, suggesting that it might play a role in the maintenance of PCH and the transcriptional silencing of major satellite sequences.

The DAXX–ATRX complex is recruited to the telomeric regions in the context of DNA hypomethylation, a characteristic typically seen in ground-state ESCs (He et al., 2015). We questioned whether the accumulation of DAXX at PCH could be caused by the decreased level of DNA methylation. We first explored the role of the active demethylation pathway, which shows increased activity in 2iV conditions (Blaschke et al., 2013). The presence of vitamin C stimulates the TET enzymes, which convert 5-methyl cytosine (5mC) into 5-hydroxymethyl cytosine (5hmC). We performed a knockout of both Tet1 and Tet2 in mESCs (Tet1/2 DKO) using Crispr-Cas9 to create a cell line devoid of these two proteins (Fig. S3A,B). Liquid chromatography with tandem mass spectrometry (LC-MS/MS) analysis confirmed that no detectable 5hmC was present in the Tet1/2 DKO ESCs in either serum or 2iV conditions, with only slight differences noted in 5mC levels (Fig. 3A,B).

Upon monitoring DAXX and PCH localization in Tet1/2 DKO ESCs grown in serum or after 4 days of 2iV conversion, immunofluorescent detection showed that the absence of 5hmC minimally affected DAXX localization (Fig. 3C). Similar to what was seen in WT ESCs, Tet1/2 DKO ESCs showed an increase in DAXX signal at a subset of PCH areas following 2iV conversion (Fig. 3D,E). Indeed, 83.3% of nuclei showed at least one DAXX-positive PCH foci, with a slightly higher proportion exhibiting 2 DAXX-positive PCH foci than in WT. These findings suggest that the relocation of DAXX to PCH during ESC ground-state conversion does not rely on an active demethylation pathway.

To investigate further the role of DNA methylation, we analyzed the association of DAXX with PCH in triple knockout ESCs for Dnmt1, Dnmt3a and Dnmt3b (Dnmt TKO) (Dubois et al., 2022). Without the presence of DNA methyltransferase, neither 5mC nor 5hmC could be detected by LC-MS/MS (Fig. 3A,B). Similar to the patterns observed in both the WT and Tet1/2 DKO, the level of DAXX at PCH significantly increased in 2iV conditions, with a nearly identical proportion of DAXX-positive foci (Fig. 3D,E).

In conclusion, the recruitment of DAXX to PCH in ground-state ESCs is neither a consequence of 5mC loss nor does it require the active DNA demethylation pathway. Instead, these findings suggest that the presence of DAXX at PCH could be a distinct hallmark of ground-state ESCs.

To investigate the role that DAXX might play at PCH in ground-state ESCs, we next investigated whether DAXX facilitates PCH formation in ESCs. Given that the loss of pericentromeric silencing is generally caused by defective heterochromatin assembly and often correlates with impaired clustering of chromocenters (Hahn et al., 2013; Healton et al., 2020; Pinheiro et al., 2012; Zhu et al., 2011), we performed DNA FISH experiments against the major satellites and segmented the 3D signal to determine the number of clusters (Fig. 4A). In serum or differentiated cells, the deletion of Daxx had no significant impact on the number of clusters (Fig. 4B). However, after 4 days of 2iV conversion, the average number of major satellite foci was increased by 37% in the Daxx KO ESCs. The presence of an increased number of PCH foci indicates that DAXX is required for their clustering in ESCs specifically under 2iV conditions, and correlates with the loss of viability observed in these cells at later stages of conversion (Fig. 1B). To confirm the role of DAXX in PCH clustering, we used the transcription activator-like effector (TALE) epigenome editing tool to ask whether direct DAXX recruitment to major satellites was sufficient to reduce the number of foci. We generated a FLAG-tagged TALE engineered to specifically target major satellite sequences (hereafter called TALE_MajSat), following the design of a previous study (Miyanari et al., 2013). Immunofluorescence against the FLAG tag of the TALE_MajSat confirmed its colocalization with the major satellite FISH signal (Fig. 4C). When bound to pericentromeres in pluripotent Daxx KO cells, TALE_MajSat–DAXX restored chromocenter clustering, leading to a decrease in the number of PCH foci (Fig. 4D). In WT ESCs, artificial tethering of DAXX to chromocenters had a similar impact, increasing endogenous PCH clustering (Fig. S4A). These results indicate that DAXX enhances physical interactions between major satellite sequences and contributes to the 3D organization of PCH.

We hypothesized that the decreased chromocenter clustering seen in Daxx KO ESCs might result from defective heterochromatin formation. Functional heterochromatin forms a self-segregating subcompartment with limited protein exchange with the rest of the nucleoplasm (Hinde et al., 2015; Strom et al., 2017). To test whether DAXX is important for heterochromatin segregation, we followed GFP–HP1α using live-imaging in WT and Daxx KO ESCs (Fig. 4E). We monitored the mobility of GFP–HP1α by measuring its fluorescence recovery rate at chromocenters after photobleaching (Fig. 4E; Fig. S4B). Compared to what was seen in WT, the half-recovery time was significantly shorter in Daxx KO cells both in serum and 2iV conditions, suggesting higher protein exchange between the chromocenter and the nucleoplasm in the absence of DAXX (Fig. 4F). However, the mobile fraction of GFP–HP1α remained constant in both cell lines, suggesting that the same amount of protein was bound to chromatin (Fig. S4B).

The boundary property of PCH is known to cause high variance in HP1α signal over time at the edge of a chromocenter, a property visible in condensates due to an increase in coordinated movement of molecules at the border of the compartment (Strom et al., 2017). We reasoned that this higher HP1α recovery rate might arise from altered heterochromatin acting as a barrier to protein diffusion. We thus quantified temporal changes in GFP–HP1α signal intensity variation at unbleached chromocenters (Fig. 4G). We found that variance levels were low in the nucleoplasm of WT ESCs and only increased at chromocenters, peaking at their borders (Fig. 4G,H), confirming previous observations (Strom et al., 2017). However, in Daxx KO cells, the pattern of the variance from HP1α signal was altered. In both serum and 2iV conditions, the peak of variance at the edge of chromocenters was significantly lower in the absence of DAXX, suggesting a compromised heterochromatin barrier (Fig. 4H). Moreover, molecular analysis using MNase digestion revealed that Daxx deletion in pluripotent ESCs increased the accessibility of pericentric chromatin, indicating that PCH exists in an abnormal state in these cells (Fig. S4C).

Our data collectively suggest that DAXX is important for regulating the compaction state and chromocenter boundary properties of PCH in ESCs. Specifically, in ground-state conditions, DAXX maintains major satellite clustering. In conclusion, DAXX is essential for proper assembly and spatial organization of PCH in pluripotent ESCs.

PML, identified as a partner of DAXX, has recently been shown to contribute to the transcriptional repression of transposable elements in ESCs (Tessier et al., 2022). To investigate the role of PML in the organization of the heterochromatic compartment in ground-state ESCs, we first analyzed its localization in relation to DAXX and H3K9me3 by immunofluorescence (Fig. 5A). Regardless of culture conditions, PML formed small round foci in the nucleoplasm (Fig. 5A). These are the same nuclear bodies where DAXX accumulates as evidenced by strong Pearson correlation values between the two proteins (Fig. 5A; Fig. S5A). Upon 2iV conversion, a fraction of larger PML foci were observed with higher levels of H3K9me3 and DAPI (Fig. S5B–D). These larger PML foci accumulated around PCH in the form of either complete or partial structures, referred to as PML-rim or PML-arc, respectively (Fig. 5A; Fig. S5E). Both the PML-rim and PML-arc structures were consistently associated with the presence of DAXX at PCH, and were observed in similar proportions to DAXX-positive PCH clusters (Fig. 2E, Fig. 5B). This indicates that PML is recruited to PCH alongside DAXX in ground-state ESCs.

To further examine the role played by PML at PCH clusters, we analyzed the distribution of major satellites in Pml KO ESCs using FISH (Tessier et al., 2022). In Pml KO ESCs, no protein was detectable (Fig. S5F). Similar to what was found for the Daxx KO line, Pml KO ESCs showed a significant increase in the number of major satellite foci only after 4 days of 2iV conversion (Fig. 5C,D), suggesting that PML, like DAXX, is required for proper PCH clustering in ground-state ESCs.

Next, we wished to determine whether Pml deficiency influenced the integrity of peripheral heterochromatin. Immunofluorescence revealed that H3K9me2 accumulated at the nuclear periphery to a greater extent in serum conditions in Pml KO cells compared to WT cells (Fig. 5E,F). However, after 4 days of 2iV conversion, the H3K9me2 signal in Pml KO cells was similar to that observed in Daxx KO cells, with only small patches of H3K9me2 present beneath the nuclear envelope and a significant reduction in peripheral enrichment of the signal (Fig. 5F). In contrast, the lamin B1 signal was not altered by Pml deletion supporting the conclusion that it is specific to peripheral heterochromatin (Fig. S5G,H). To determine how this peripheral heterochromatin alteration influenced gene expression, we analyzed the transcriptional repression of 25 LAD genes that were significantly upregulated in our RNA-sequencing following Daxx deletion (Fig. 1C,D). Out of the 25 selected genes, 22 showed higher mRNA levels in 2iV-converted Pml KO ESCs compared to WT cells when analyzed by RT-qPCR (Fig. 5G). Nonetheless, the transcriptional upregulation in Pml KO cells was less pronounced than in Daxx KO cells, with 15 genes exhibiting a log₂ fold change exceeding two, compared to 23 genes in the absence of DAXX. Our findings suggest that PML, like DAXX, plays a role in the transcriptional silencing of LAD genes in ground-state ESCs.

The deletion of Pml and Daxx have very comparable consequences on the heterochromatin of ESCs after 4 days of 2iV conversion. We therefore wondered how Pml deletion would impact the viability of ESCs under prolonged 2iV conversion (Fig. 5H). After 4 days of conversion, the viability of Pml KO was not impacted, but was decreased by nearly 75% after 8 days. As seen upon the deletion of Daxx, this effect was specific to prolonged 2iV conversion, as the absence of PML had no consequence on the viability under serum condition or after retinoic acid differentiation.

Overall, our data show that PML and DAXX play similar roles in ground-state ESCs. Like DAXX, PML is recruited to PCH and its absence impairs the proper organization of the heterochromatic compartment, significantly impacting cell viability.

We have shown that in ground-state ESCs, DAXX and PML are recruited to major satellites thereby facilitating PCH formation. To further investigate the mechanism underlying the function of DAXX and PML, we searched for DAXX- and PML-interacting factors that could be recruited specifically to chromocenters. Among the DAXX interacting partners, SETDB1, an H3K9me3 methyltransferase, colocalizes with DAXX at PML nuclear bodies and is important for repetitive DNA transcriptional silencing (Cho et al., 2011; Karimi et al., 2011). SETDB1 partially colocalized with PML in a DAXX-independent manner in serum conditions (Fig. S6A). Under 2iV treatment, SETDB1 relocalized to PCH with DAXX and PML in 70% of WT ESCs (Fig. 6A,B; Fig. S6B). By contrast, PML and SETDB1 were only associated with PCH in 5 and 17% of Daxx KO cells, respectively (Fig. 6B). We conclude that DAXX is crucial for PML and SETDB1 recruitment to chromocenters. To investigate whether DAXX was sufficient for SETDB1 recruitment to PCH, we expressed TALE_MajSat fused to DAXX in Daxx KO ESCs. TALE-mediated DAXX binding to major satellite repeats drastically increased the amount of SETDB1 signal observed at chromocenters (Fig. 6C,D).

To determine whether SETDB1 was depositing the H3K9me3 modification at PCH when it is recruited by DAXX, we used Suv39H1 and Suv39H2 double knockout (Suv39dKO) ESCs (Lehnertz et al., 2003; Peters et al., 2001). As H3K9me3 in WT ESCs is primarily deposited by SUV39H1 and SUV39H2, these Suv39dKO cells provide a system where any detected H3K9me3 can only be attributed to SETDB1 activity. Suv39dKO cells displayed no growth defects in 2iV medium, underscoring the non-essential role of these enzymes for pluripotency. Unlike in WT ESCs, Suv39dKO cells did not show an accumulation of H3K9me3 at major satellite foci (Fig. 6E,F). However, upon 2iV conversion, we observed a notable accumulation of H3K9me3 in certain major satellite foci surrounded by PML, suggesting that the fraction of SETDB1 recruited to PCH in the ground state is indeed functionally active (Fig. 6E,F).

In summary, our data demonstrate that DAXX recruits PML and SETDB1 to chromocenters, providing an alternative pathway for the deposition of H3K9me3 at PCH in ground-state ESCs.

Considering our observations that DAXX presence at PCH correlated with higher H3K9me3 levels (Fig. 2G), and its function in the recruitment of SETDB1 (Fig. 6E,F), we asked whether Daxx deletion would impact the level of H3K9me3 at PCH. Surprisingly, the level of H3K9me3 at PCH was very similar in Daxx KO and WT ESCs in serum and 2iV condition (Fig. S6C,D). These findings suggest that the recruitment of SETDB1 by DAXX is not essential for maintaining H3K9me3 in the ground-state.

To better understand the specific roles of DAXX, SETDB1 and H3K9me3 in organizing PCH, we used the TALE_MajSat strategy to specifically target these elements to chromocenters in Suv39dKO ESCs. (Fig. 7A). Both histone methyltransferases were able to individually restore H3K9me3 at chromocenters, when targeted by TALE_MajSat (Fig. 7A; Fig. S6E). The specific recruitment of DAXX to major satellite foci also promoted H3K9me3 deposition via SETDB1 at 70% of chromocenters. The impact of the different TALE_MajSat fusions on PCH clustering was further assessed by quantifying the number of TALE-bound foci (Fig. 7B). Consistent with our observations in Daxx KO and WT ESCs (Fig. 4D; Fig. S4A) targeting DAXX to major satellite also enhanced PCH clustering in Suv39dKO ESCs as evidenced by the reduced number of PCH foci (Fig. 7B). However, the recruitment of either SETDB1 or SUV39H1 did not affect the number of chromocenters, supporting the conclusion that H3K9me3 modification does not influence PCH clustering (Fig. 7B).

In conclusion, the alternative pathway for H3K9me3 deposition at PCH, mediated by DAXX and SETDB1, is not essential for preserving this epigenetic mark. Moreover, the influence of DAXX on PCH organization cannot be solely attributed to SETDB1 or H3K9me3, underscoring a pivotal role for DAXX in PCH organization in ground-state ESCs.

Given that DAXX, unlike SETDB1, had a direct impact on the spatial organization of chromocenters, we examined the role of its H3.3 chaperone activity by expressing a mutated form, DAXX^Y222A, which lowers its interaction with H3.3 (Elsässer et al., 2012). We confirmed that co-immunoprecipitation of TALE_MajSat DAXX^Y222A fusion with H3.3 was drastically reduced compared to its WT counterpart (Fig. S7A). Although DAXX targeted to chromocenters could recruit a tagged version of H3.3, neither the control TALE_MajSat-Δ nor the TALE_MajSat DAXX^Y222A fusion were able to mobilize H3.3 to PCH (Fig. S7B). When expressed in Suv39dKO ESCs, the TALE_MajSat DAXX^Y222A fusion did not rescue H3K9me3 levels at chromocenters (Fig. 7A; Fig. S7C). This could be attributed to the absence of SETDB1, given the DAXX^Y222A mutant was unable to recruit SETDB1 in Daxx KO ESCs (Fig. S7D). Most importantly, the TALE_MajSat DAXX^Y222A fusion did not change the number of chromocenters, arguing that H3.3 is crucial for DAXX-mediated PCH clustering (Fig. 7B). To further confirm a function for H3.3 in chromocenter organization, we depleted endogenous H3.3 using a pool of siRNAs targeting the 3′UTR of H3f3a and H3f3b mRNAs and expressed TALE_MajSat–DAXX in Daxx KO ESCs (Fig. 7C; Fig. S7E). The H3.3 knockdown was rescued by the expression of different HA-tagged versions of H3.3, which were assessed for their impact on PCH clustering by immuno-FISH (Fig. 7C; Fig. S7E). The binding of DAXX at major satellites did not enhance PCH clustering when H3.3 was knocked down (Fig. 7D). The expression of a WT H3.3–HA significantly reduced the number of major satellite foci and rescued the PCH hyperclustering phenotype mediated by TALE_MajSat–DAXX. In contrast, expression of H3.3^G90M, a mutant unable to bind DAXX (Elsässer et al., 2012; Hoelper et al., 2017), did not decrease the number of major satellite foci (Fig. 7D). This result is consistent with the absence of a chromocenter clustering phenotype in cells transfected by TALE_MajSat–DAXX^Y222A (Fig. 7B) and confirms the important role of H3.3 in PCH clustering. Given that DAXX can recruit SETDB1 to PCH, we asked whether a H3.3K9 mutant that cannot be methylated impacts PCH clustering and used H3.3^K9A. Like H3.3^G90M, H3.3^K9A failed to rescue the hyper-clustering phenotype in TALE_MajSat–DAXX-expressing cells, suggesting that the modification of this residue is essential for the spatial organization of chromocenters (Fig. 7D).

Altogether, we conclude that the role of DAXX in PCH organization is intrinsically linked to its chaperone activity. Interaction with H3.3 and H3.3K9 modifications are important in facilitating the spatial organization of pericentromeres in pluripotent ESCs.

During early embryogenesis, adapting heterochromatin to compensate for the wave of DNA demethylation is essential for the maintenance of transcriptional repression of repetitive DNA and protect genome integrity. The mechanisms and molecular factors responsible for heterochromatin reorganization were, however, largely unknown. Here, we describe a novel and essential role for the H3.3-chaperone DAXX and its partner PML in the survival of ground-state pluripotent stem cells. Taken together, our results support a model in which DAXX and PML facilitate both pericentric and peripheral heterochromatin formation in ground-state ESCs (Fig. 8). At PCH, DAXX deposits H3.3 and recruits PML and SETDB1, which can methylate H3.3K9. In the absence of either DAXX or PML, PCH is compromised (Fig. 8). In Daxx KO ESCs, heterochromatin at major satellites is less compact and partially loses its boundary properties, leading to defective chromocenter clustering. At the nuclear periphery, PML and DAXX are necessary for the maintenance of the heterochromatin positioned beneath the nuclear envelope, contributing to the transcriptional silencing of genes tethered to the nuclear lamina. The global failure in heterochromatin compartment formation could directly contribute to the inability of Daxx KO and Pml KO ESCs to maintain the ground-state of pluripotency.

Both DAXX and H3.3 are essential for early embryogenesis, but their specific functions have remained elusive (Jang et al., 2015; Lin et al., 2013; Liu et al., 2020; Michaelson et al., 1999). We found that DAXX is necessary for maintaining the spatial organization and chromatin state of pericentric heterochromatin (PCH), thereby protecting the viability of ground-state ESCs (Fig. 4B,F,H). This finding is consistent with in vivo observations, which show that DAXX and SETDB1 are recruited to PCH just before chromocenter formation, and the viability of Daxx KO embryos drops quickly after the transition to the pluripotent stage (Arakawa et al., 2015; Cho et al., 2012; Liu et al., 2020). These observations support our conclusion that the role of DAXX and SETDB1 at PCH might be crucial for the ground-state of pluripotency. Given that defective PCH assembly is often associated with genomic instability, our model might also explain why H3f3a and H3f3b double knockout ESCs exhibit severe chromosome segregation defects (Jang et al., 2015).

Although Daxx deletion had no impact on major satellite transcription and chromocenter formation upon neuronal differentiation (Fig. 2A, Fig. 4B), DAXX and H3.3 are important for chromocenter clustering during myoblast differentiation (Park et al., 2018; Salsman et al., 2017). In myoblasts, DAXX is recruited to chromocenters, where H3.3 deposition stimulates transcription of major satellites, suggesting a mechanism different to the one we observed in ESCs. Different DAXX and partner recruitment pathways could explain the opposite impact of DAXX and H3.3 on major satellite transcription. In ESCs, DAXX recruits SETDB1, thus favoring a repressive chromatin environment. Upon myoblast differentiation, Setdb1 is downregulated and might not be available for interaction with DAXX, which is recruited to major satellite repeats by the muscle-specific long non-coding RNA (lncRNA) ChRO1 (Park et al., 2018; Song et al., 2015).

Recent insights into chromocenter organization have come from the observation that the intrinsically disordered regions of HP1α facilitate the compartmentalization of chromocenters through liquid-liquid phase separation (Larson et al., 2017; Strom et al., 2017). Given that the first 50 amino acids and the C-terminal half of DAXX are intrinsically disordered (Escobar-Cabrera et al., 2010), it is possible that DAXX might also undergo liquid–liquid phase separation. Multivalent interactions between chromodomains recognizing H3K9me3 have also been proposed to contribute to phase separation of heterochromatin (Wang et al., 2019). However, Daxx deletion impacted the boundary properties of pericentromeric heterochromatin without affecting H3K9me3 (Fig. 4H; Fig. S6D). Furthermore, upon 2iV conversion, the fraction of chromocenters bound by DAXX display higher levels of H3K9me3 (Fig. 2D). It has been well documented that HP1 binding to H3K9me3 relies on its chromodomain (Bannister et al., 2001; Lachner et al., 2001; Mattout et al., 2015). Interestingly, in specific cases, it has been observed that tri-methylation of lysine 9 of H3.3 is required to maintain H3K9me3 levels at telomeres (Udugama et al., 2015). Given that DAXX-mediated chromocenter clustering relies on H3.3K9 modification (Fig. 7C,D), future work will compare the interaction of HP1α and additional heterochromatin-related proteins to H3.3K9me3 versus H3K9me3.

In ground-state ESCs, DAXX binds only a subset of chromocenters in ∼80% of cells (Fig. 2E). Given that chromocenters replicate synchronously during mid-S phase (Guenatri et al., 2004), the recruitment of DAXX is unlikely to rely on replicative chromatin remodeling. Polycomb group proteins have been shown to be recruited to PCH in ESCs (Pailles et al., 2022; Saksouk et al., 2014; Tosolini et al., 2018). The fact that DAXX was shown to rely on PRC1 to bind major satellites in zygotes would support a polycomb-mediated recruitment of DAXX in ground-state ESCs (Liu et al., 2020). However, the presence of H3K27me3, deposited by PRC2, is relatively homogenous among the PCH foci within a nucleus, suggesting that the mechanism is different (Saksouk et al., 2014; Tosolini et al., 2018).

ESCs, especially those grown in 2iV, are characterized by reduced levels of 5mC. Within this context, DAXX has been demonstrated to play a significant role in the formation of heterochromatin at DNA repeat regions (Elsässer et al., 2015; He et al., 2015). However, the absence of DNA methylation was not sufficient to recruit DAXX to PCH, corroborating a previous PCH proteomic analysis (Fig. 3D,E) (Saksouk et al., 2014). Nevertheless, the potential requirement for lower DNA methylation in DAXX recruitment to PCH cannot be excluded, given that Dnmt TKO ESCs might have evolved adaptive mechanisms during their establishment, possibly explaining discrepancies in PCH epigenetic regulation between in vitro and in vivo conditions (Pailles et al., 2022).

DNA damage might also be a factor in the recruitment of DAXX to PCH. Several studies have described the incorporation of H3.3 after DNA damage (Adam et al., 2013; Fortuny et al., 2021; Juhász et al., 2018; Li and Tyler, 2016; Luijsterburg et al., 2016). Both the HIRA and DAXX–ATRX complexes have been proposed to deposit H3.3 after DNA damage, but DAXX has been more specifically observed at chromocenters damaged by UVC (Fortuny et al., 2021). In vivo, DAXX binds preferentially paternal chromosomes, which are actively demethylated by the TET enzymes, a pathway that can result ultimately in a double-strand break (DSB) (Nakatani et al., 2015; Wossidlo et al., 2011). Given that TET enzymes are particularly active in ground-state ESCs (Blaschke et al., 2013), it was tempting to hypothesize that these enzymes might drive DAXX recruitment to PCH. However, our data compellingly indicate that TET enzymes are not key players in DAXX recruitment (Fig. 3D), suggesting that if a DNA damage response is involved, it might be triggered by a different source of damage. Pericentromeric regions are particularly susceptible to the formation of four-stranded guanine-rich structures in vivo known as G-quadruplexes (Henderson et al., 2017). Intriguingly, ATRX, a partner of DAXX, is known to safeguard heterochromatin from G-quadruplex formation, primarily by recruiting SETDB1 (Teng et al., 2021). Although these links provide interesting insights into the potential mechanisms of DAXX recruitment and DNA damage control, they highlight the need for further studies to confirm these hypotheses.

SETDB1 has been shown to be responsible for H3K9me3 deposition at transposable elements and telomeres (Elsässer et al., 2015; Gauchier et al., 2019; Karimi et al., 2011). However, knocking down SETB1 in Suv39h1 and Suv39h2 double-knockout cells destabilizes chromocenters, suggesting that SETDB1 could be involved in PCH formation (Pinheiro et al., 2012).

Our work demonstrates that DAXX is essential for the recruitment of SETDB1 to PCH (Fig. 6A,B). This finding is consistent with a recent study that has highlighted a role for DAXX in the histone chaperone network, where it recruits histone methyltransferases, including SETDB1, and promotes H3K9me3 catalysis on new H3.3–H4 histones before DNA deposition (Carraro et al., 2023). However, in ESCs lacking Suv39h1 and Suv39h2, SETDB1 fails to restore H3K9me3 across all PCH clusters, with restoration only apparent at PML-positive foci (Fig. 6E,F) This finding implies that SETDB1 might preferentially modify histones already integrated into DNA. In contrast to SUV39H1 and SUV39H2, SETDB1 contains a triple Tudor domain recognizing the double modification K14 acetylation and K9 methylation, which might facilitate its binding to hyperacetylated, newly incorporated histone H3.3 (Jurkowska et al., 2017).

In mouse embryonic fibroblasts, SETDB1 promotes H3K9me1 at major satellites. However, SETDB1 has been shown to be specifically recruited during S-phase by the H3.1 and H3.2 chaperone CAF1 (Loyola et al., 2009), suggesting that SETDB1 substrate specificity might change depending on its recruitment pathway. Similar to our data in ESCs, CAF1 recruited SETDB1 to chromocenters only in a subset of S-phase cells. As pericentromeric satellites were shown to be particularly sensitive to replicative stress, SETDB1 recruitment could also result from DNA damage (Crosetto et al., 2013).

This study reveals that DAXX is recruited to pericentromeres in the context of DNA hypomethylation. Although PML nuclear bodies are generally devoid of chromatin, PML formed a shell around DAXX-positive chromocenters in hypomethylated ground-state ESCs (Fig. 5A). Similar structures have been observed in patients with immunodeficiency, centromeric instability and facial dysmorphia (ICF) syndrome associated with mutations in DNA methyltransferases (Luciani et al., 2006). In lymphocytes from ICF patients, DAXX and DNA repair proteins accumulate at hypomethylated pericentromeric satellites, suggesting that our proposed model could apply to other pathologies (Fig. 8).

In conclusion, our study reveals DAXX as an important factor for different heterochromatic compartments in ground-state ESCs. At the nuclear periphery, DAXX and its partner PML are necessary for the clustering of peripheral heterochromatin at the nuclear edge. At PCH, DAXX deposits H3.3 and recruits PML and SETDB1 to facilitate chromocenter formation by enhancing their clustering properties. It would be interesting to characterize whether DAXX also contributes to the 3D organization of other heterochromatin domains, such as the clustering of LINE1 elements (Lu et al., 2021). Beyond early development, our work could also provide crucial insights into the molecular pathways that overcome DNA hypomethylation transitions in pathological contexts, such as cancers.

Feeder-free embryonic mouse ESCs (E14) were used for most experiments except where otherwise notified. E14 mouse ESCs were kindly provided by Pablo Navarro (Pasteur Institute, Paris, France). Daxx KO and Tet1/2 DKO cell lines were constructed using CRISPR/cas9 editing in E14 mES cells. Guide-RNA targeting Daxx was designed using the online CRISPOR tool (Sg-mDaxx: 5′-gGACCTCATCCAGCCGGTTCA-3′; http://crispor.tefor.net/). Guide RNA targeting Tet1 and Tet2 were selected based on a previous study (Mulholland et al., 2020). Oligonucleotides were designed with a BbsI site on 5′ to clone them into the pSpCas9(BB)-2A-Puro(pX459) v2.0 vector (Ran et al., 2013). Pml KO and Dnmt TKO ESCs were generated by CRISPR from E14 cells and are described in previous studies (Dubois et al., 2022; Tessier et al., 2022). Feeder-free Suv39HdKO and corresponding WT R1 ES cells were provided by Alice Jouneau (NRAE, Jouy-en Josas, France) and generated in Antoine Peters's laboratory (Lehnertz et al., 2003).

Pluripotent cells were cultured either in a serum condition, defined as follows: DMEM (Gibco), supplemented with 10% of ESC-certified fetal bovine serum (FBS; Gibco), 2-mercaptoethanol (0.05 mM, Gibco), Glutamax (Gibco), MEM non-essential amino acids (0.1 mM, Gibco), penicillin-streptomycin (100 units/ml, Gibco) and LIF (1000 units/ml) for serum condition. Serum-cultivated cells were grown on 0.1% gelatin-coated plates or stem cell plates (Stem Cell Technology) at 37°C with 5% CO₂. Medium was changed every day and cells were passaged every 2 to 3 days. The other culture condition is the chemically defined serum-free 2i condition, defined as follows: Neurobasal:DMEM/F-12 (50:50, Gibco) medium, supplemented with N2 and B-27 supplements (Gibco), BSA fraction V (0.05% Gibco), 1-thioglycerol (1.5×10⁻⁴ M, Sigma) and ESGRO 2 inhibitors [GSK3i (3 µM) and MEKi (1 µM)] and LIF (10³ units/ml; Merck). Vitamin C (L-ascorbic acid 2-phosphate, Sigma) was added at a concentration of 100 µg/ml (Blaschke et al., 2013). Cells were grown on 0.1% gelatin-coated plates. Medium was changed daily. Cells were passed every 2 days at 1:4 ratio for the first passage then at a 1:6 ratio. Differentiation of mouse ESCs was undertaken by LIF removal for the first 24 h. Then, non-LIF medium was supplemented with retinoic acid (10⁻⁶ M) for 4 days.

Cell growth was quantified by plating ESCs at a density of 2×10⁵ cells/ml in serum and 4×10⁵ cells/ml in 2iV every two days. To assess cell numbers, cell counts were performed three times using a hemocytometer. Specifically, the number of cells in Daxx KO and Pml KO ESCs was compared to WT samples, and the relative cell count was expressed as a percentage.

All cell lines are available upon request.

Cells were harvested using trypsin and 10⁶ cells were plated in a 0.1% gelatin-coated plate and transected with 0.2 to 2.5 µg of DNA using the Lipofectamine 2000 reagent (Thermo Fisher Scientific), following the manufacturer's protocol. TALE vectors were constructed using previously described methods (Ding et al., 2013). A TALE-specific DNA-binding domain targeting major satellite repeats was created by the modular assembly of individual TALE repeats inserted into a backbone vector containing TALE-Nrp1-VP64 as previously described (Therizols et al., 2014). The BamHI-NheI fragment containing VP64 was removed to generate the control vector containing only the DNA-binding domain (T_MS-Δ). The VP64 fragment was also replaced by PCR products encoding the DAXX protein, corresponding DAXX mutants or additional proteins such as SUV39H1 and SETDB1. The coding sequences of the different proteins were amplified by PCR from cDNAs obtained after RNA extraction of serum E14 mouse ESCs. DAXX mutants were generated by PCR from WT DAXX cDNA.

For the overexpression H3.3 and derivative mutants, dsDNA blocks (Eurofins Genomics) containing the coding sequences of the murine H3f3b, mH3f3b^G90M mH3f3b^K9A and digested by BsmBI were cloned into a pCAGG vector (Therizols et al., 2014) digested by EcoRI and NheI.

All plasmids are available upon request.

Total RNA was extracted using RNeasy extraction kit (Qiagen) according to the manufacturer's protocol including DNaseI treatment for 15 min at room temperature (Qiagen). cDNAs were generated from 1 µg of RNA using the Maxima first strand cDNA synthesis kit (Thermo Fisher), with a second round of DNaseI treatment from the Maxima kit for 15 min. Real-time qPCR was carried out using a LightCycler 480 instrument (Roche) and the LightCycler 480 SYBR green master mix (Roche). The qRT-PCR primers used in this study are listed in Table S1. Three independent biological repeats were obtained for each sample. For RNA-seq experiment, RNA quality was assessed using the Agilent 2100 bioanalyzer. Libraries were prepared using oligo(dT) beads for mRNA enrichment, then fragmented and reverse transcribed using random hexamers primer. After adaptor ligation, the double-stranded cDNA is completed through size selection of 250–300 bp and PCR amplification, then quality of the library is assessed by the Agilent 2100 bioanalyzer. Sequencing was performed in 150 bp paired-end reads using an Illumina sequencer platform.

FASTQ files generated by paired end sequencing were aligned to the mouse genome using bowtie2 v2.2.6 (parameters: --local --threads 3; mm9 genome build; Langmead and Salzburg, 2012). Mapped RNA-seq data was processed using tools from the HOMER suite (v4.8; Heinz et al., 2010). SAM files were converted into tag directories using ‘makeTagDirectory’ (parameters: -format sam -sspe). Genomic intervals which extended beyond the end of the chromosomes was removed using ‘removeOutOfBoundsReads.pl’. bigWig browser track files were generated using ‘makeUCSCfile’ (parameters: -fsize 1e20 -strand+-norm 1e8). For gene expression analysis, read depths were quantified for all annotated refseq genes using analyzeRepeats.pl (parameters: rna mm9 -strand both -count exons -rpkm -normMatrix 1e7 -condenseGenes). For repeat analysis, read coverage was quantified for each repeat and then condensed to a single value for each named entry (parameters: repeats mm9 -strand both -rpkm -normMatrix 1e7 -condenseL1). Read depths were then corrected for the number of instances of each repeat prior to expression analysis.

Quantified RNA-seq data was processed using the limma package (R/Bioconductor; Gentleman et al., 2004; Ritchie et al., 2015). Following the addition of an offset value [1 reads per kilobase per million mapped reads (RPKM)] to each gene or repeat, data was normalized across all samples using ‘normalizeBetweenArrays’ with method=‘quantile'. Fold-changes and P-values for differential expression of genes and repeats were determined using empirical Bayes statistics. Briefly, data was fitted to a linear model using ‘lmFit' and specified contrasts were applied using ‘makeContrasts’ and ‘contrasts.fit'. Data was processed using the ‘topTable' function with adjust.method=‘BH’ (Benjamini–Hochberg multiple-testing correction). Differential expression was defined as having a log₂ fold change ≤−1 or ≥1 and an adjusted P-value of ≤0.01. Three biological replicates for each condition represent independently cultured pools of cells.

Heatmaps and boxplots were generated using Prism GraphPad (v9). Histograms were drawn using either Prism GraphPad or Excel (Microsoft). Volcano plots were generated using the plot function in R.

Genomic DNA was extracted with RNase A as described above, plus an additional digestion step: 1 µg of DNA was treated with 10 U DNA Degradase Plus (ZymoResearch) at 37°C for 4 h, followed by inactivation of the enzyme at 70°C for 20 min and then Amicon Ultra-0.5 ml 10 K centrifugal filters (Merck Millipore) were used to filter the solution. The reaction mixture retained on the centrifuge filter was processed for LC-MS/MS analysis; analysis of total 5-mdC and 5-hmdC concentrations was performed using a Q exactive mass spectrometer (Thermo Fisher Scientific). The instrument was equipped with an electrospray ionization source (H-ESI II Probe) coupled to an Ultimate 3000 RS HPLC (Thermo Fisher Scientific). A Thermo Fisher Hypersil Scientific Gold aQ chromatography column (100 mm×2.1 mm, 1.9 μm particle size) heated to 30°C was injected with digested DNA. The flow rate was set to 0.3 ml/min and the column was run for 10 min in isocratic eluent consisting of 1% acetonitrile in water containing 0.1% formic acid. Parent ions were fragmented in positive ion mode, parallel reaction monitoring (PRM) mode at 10% normalized collision energy; MS2 resolution was 17,500, AGC target was 2×10⁵, maximum injection time was 50 ms, and separation window was 1.0 m/z. The inclusion list contained the following masses: dC (228.1), 5-mdC (242.1), and 5-hmdC (258.1). Extracted ion chromatograms (±5 ppm) of basic fragments were used for detection and quantification (dC, 112.0506 Da; 5-mdC, 126.0662 Da; 5-hmdC, 142.0609). Calibration curves were previously generated using synthetic standards in the ranges of 0.2 to 10 pmol injected for dC and 0.02 to 10 pmol for 5mdC and 5hmdC. Results were expressed as percentage of total dC.

Murine ESCs were harvested with trypsin (Gibco) and plated for 4–12 h onto either 0.1% gelatin-coated or ECMatrix-coated (Sigma) glass cover slips. Cells were fixed with 4% paraformaldehyde for 10 min at room temperature, then rinsed three times with PBS. Cells were permeabilized with 0.1× Triton X-100 for 12 min at room temperature, then rinsed three times with PBS. Blocking was done in 3% BSA solution for 30 min at room temperature. All incubations with primary antibodies were performed for either 1 h at room temperature or overnight at 4°C with the following antibodies against: H3K9me3 (cat. no. 39161, Active Motif, 1:1000), DAXX (cat. no. sc-8043, Santa Cruz Biotechnology, 1:500), H3K9me2 (cat. no. 39239, Active Motif, 1:1000), lamin B1 (cat. no. ab16048, Abcam, 1:1000), PML (gift from Hugue de Thé, INSERM U944, Paris, France; 1:2000), SETDB1 (cat. no. 11231-1-AP, Proteintech, 1:100), HA tag (cat. no. 11867423001, Sigma-Aldrich; 1:1000) and FLAG tag (cat. no. F3165, Sigma-Aldrich; 1:1000). Incubation with secondary antibodies (fluorescently labeled anti-mouse-IgG or anti-rabbit-IgG, 1:1000) were performed for 1 h at room temperature. Mounting was performed using ProLong Diamond with DAPI mounting medium (Thermo Fisher Scientific).

Murine ESCs were harvested with trypsin (Gibco) and plated for 4–6 h onto 0.1% gelatin-coated glass cover slips. Cells were fixed with 4% paraformaldehyde (PFA) for 10 min at room temperature, then rinsed three times with PBS. Cells were permeabilized with 0.5× Triton X-100 for 12 min at room temperature, then rinsed three times with PBS. Cells were briefly washed in 2× saline sodium citrate buffer (SSC), then treated with RNaseA (100 µg/ml, Sigma) for 1 h at 37°C. Cells were briefly washed in 2× SSC, then denatured by a serial 2 min incubation into 70, 90 and 100% ethanol. Coverslips were air dried for 15 min. Coverslips were then incubated with 200 nM of PNA probe, placed for 10 min at 95°C for denaturation, then placed for 1 h at room temperature in the dark for hybridization. Cover slips were washed twice in 2× SSC with 0.1% Tween-20 for 10 min at 60°C. Coverslips were immerged at room temperature in 2× SSC with 0.1% Tween-20 for 2 min, then in 2× SSC for 2 min and 1× SSC for 2 min. Mounting is performed using ProLong Diamond with DAPI mounting medium (Thermo Fisher Scientific).

Murine ESCs transfected with HP1α–GFP were harvested with trypsin (Gibco) and plated for 4–6 h onto 0.1% gelatin-coated glass live-cell Nunc slides (Thermo Fisher Scientific). The fluorescence recovery after photobleaching (FRAP) experiment was carried with an LSM 800 confocal microscope (Zeiss). The 488 nm laser was used to bleach and acquire GFP signal. One image was taken before a bleach pulse of 5 ms. The bleaching area was set to target a single pericentric domain. Images were acquired every second during 35 s post-bleach. FRAP analysis was performed using a FRAP analysis ImageJ Jython script (available upon request), that generated FRAP curves and the associated half-recovery time and mobile fraction parameters.

GFP–HP1α variance along time was obtained from ImageJ analysis using standard deviation z-projection along the time for the whole duration of the movie. Quantification of heterochromatin barriers were performed using a 1 μm line across individual non-bleached chromocenter borders, for which the variance intensity along time was measured with the ImageJ software.

Images for immunofluorescence and FISH experiments were obtained with an inverted Nikon Ti Eclipse widefield microscope using a 60× immersion objective and LED sources. Z-stacks images were taken and then deconvoluted using a custom ImageJ deconvolution script Quantifications of images were performed using custom Icy scripts in Icy and ImageJ (all custom scripts available upon request). To quantify DAXX enrichment at PCH and the number of PCH foci, chromocenters were segmented using the spot detector function in ICY on either the H3K9me3 or major satellite FISH signal. Nuclei were isolated using the ICY k-means segmentation script. The nuclear periphery enrichment was measured in ImageJ; briefly linescans of LaminB1 and H3K9me2 at the periphery were isolated using the straighten command and subsequently used to generated intensity profiles.

H3K9me3 intensities at chromocenters were measured using a 2 µm line across individual DAPI-dense chromocenters. Two to three chromocenters per nucleus were analyzed.

Total protein extracts were prepared in RIPA buffer with protease inhibitor cocktail (Roche). Samples were sonicated for 3 min alternating 30 s on and 30 s off. Proteins were separated by electrophoresis in 8–15% poly-acrylamide gels then transferred onto nitrocellulose membranes. Membranes were incubated in Ponceau stain then washed in PBS with 0.1% Tween 20. Membranes were blocked in PBS, 0.1% Tween 20 and 5% milk for 30 min at room temperature before incubation overnight at 4°C with the following primary antibodies: anti-DAXX (cat. no. sc-8043, Santa Cruz Biotechnology, 1:500), anti-lamin B1 (cat. no. ab16048, Abcam, 1:1000), anti-HA tag (cat. no. 11867423001, Sigma-Aldrich; 1:1000), anti-FLAG tag (cat. no. F3165, Sigma-Aldrich; 1:1000), anti-H3.3 (cat. no. 09-838, Millipore; 1:1000), anti-TET1 (cat. no. 61741, Active Motif; 1:1000) and anti-TET2 (cat. no. MABE1132, Sigma-Aldrich; 1:1000). After three washes in PBS with 0.1% Tween 20, membranes were incubated with secondary HRP-conjugated antibody for 1 h at room temperature. Membranes were washed 3 times in PBS with 0.1% Tween 20 and visualized by chemoluminescence using ECL Plus.

The number of objects counted and statistical tests performed are indicated in the text, figure or figure legends. P-values are represented as follows: *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. Statistical tests were performed using GraphPad Prism version 9.0.0. and R software.

We would like to thank A. Bonnet-Garnier and A. Jouneau for the gift of Suv39dKO KO ESCs and advice. We are also grateful to image core microscopy facilities of IRSL, Paris, in particular N. Setterblad for his assistance with live imaging. We warmly thank V. Lallemand-Breitenbach and H. de Thé (Collège de France, Paris, France) for the sharing of PML antibody and helpful advice. We are thankful to Y. Khalil for her help in microscopy analysis. We kindly thank P. Lesage and Annia Carré Simon for their critical readings of the manuscript, and other members of the team for helpful advice. We finally thank support services of IRSL. This work was supported by Agence Nationale pour la Recherche, IDEX SLI, La Ligue Nationale Contre le Cancer (LNCC)-Comité Ile de France and the labex Who am I.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261092.reviewer-comments.pdf