Mutation of the SWI/SNF complex component Smarce1 decreases nucleosome stability in embryonic stem cells and impairs differentiation

Eukaryotic DNA wraps around histone octamers, each of which contains two copies of histone H2A, H2B, H3, and H4, to form the nucleosome, a functional unit of chromatin structure (Wolffe, 1998). Chromatin is composed of chains of nucleosomes and is packed at various densities related to the transcriptional activity in each region (Baldi et al., 2020). Active chromatin regions are loosely packed in the nucleus, whereas repressed chromatin regions are tightly packed (Chambeyron and Bickmore, 2004; Meshorer et al., 2006; Lim and Meshorer, 2020). The formation of chromatin is a benefit for functional storage of nuclear DNA, but it also interferes with the binding of transcription factors to DNA (Wolffe, 1998; Baldi et al., 2020). To overcome this physical interference, eukaryotic cells utilize the energy of ATP hydrolysis to move histones to make chromatin structures suitable for transcriptional regulation (Tsukiyama, 2002; Becker and Workman, 2013; Clapier et al., 2017). These functional changes in chromatin structure are called chromatin remodeling, and these processes are mediated by ATP-dependent chromatin remodeling factor complexes. These complexes have subunits containing a conserved catalytic ATPase domain and are divided into four subfamilies: imitation switch (ISWI), switch/sucrose non-fermentable (SWI/SNF), chromatin helicase DNA binding (CHD) and INO80 subfamily. All these remodeling complexes commonly change the positions of nucleosomes, but each chromatin remodeling complex also has characteristic functions. ISWI and CHD chromatin remodeling complexes assemble histone octamers and form evenly spaced nucleosomes (Ito et al., 1997; Corona et al., 1999; Torigoe et al., 2011; Fei et al., 2015). INO80 subfamily remodelers replace the histone H2A–H2B dimer with the H2A.Z–H2B dimer (Mizuguchi et al., 2004). SWI/SNF slides or evicts histones to make a suitable platform for transcriptional regulation (Boeger et al., 2004). Brm, a catalytic ATPase domain-containing protein of Drosophila SWI/SNF, was originally discovered as a suppressor of Polycomb group proteins. Therefore, SWI/SNF is recognized in a broad sense as a Trithorax protein (Kennison and Tamkun, 1988).

SWI/SNF chromatin remodeling complexes are composed of more than ten subunits that form large, species-specific complexes of >1 MDa (Kadoch and Crabtree, 2015). Mammalian SWI/SNF chromatin remodeling complexes are related to yeast SWI/SNF and RSC chromatin remodeling complexes in terms of subunit composition. Mammalian SWI/SNF chromatin remodeling complexes are also called Brg1/Brahma-associated factor (BAF) complexes (Wang et al., 1996, 1998). Distinct subfamilies of BAF complexes have been reported in mouse cells and are required to maintain the pluripotent state of undifferentiated cells and their proper differentiation (Ho et al., 2009a,b; Hainer et al., 2019; Ho et al., 2019). The components of the mouse BAF complexes change during differentiation. The ESC-specific BAF complex (esBAF) is mainly composed of Smarca4 (Brg1), Arid1a, Smarcb1, a homodimer of Smarcc1, Smarcd1, Smarcd2, Smarce1, Phf10, Dpf2 and actin-like protein 6a (Ho et al., 2009a,b). Differentiation of ESCs into post-mitotic neurons accompanies the replacement of the components of esBAF complex: Arid1a by a heterodimer of Arid1a and Arid1b, the homodimer of Smarcc1 by a heterodimer of Smarcc1 and Smarcc2, Smarcd2 by Smarcd3, Phf10 by Dpf1 and Dpf2 by Dpf3. The mutually exclusive catalytic subunits, Smarca4 and Smarca2, are also exchanged during differentiation. The post-mitotic neuron-specific BAF complex is called neuronal BAF (nBAF) (Lessard et al., 2007; Staahl et al., 2013; Bachmann et al., 2016; Alfert et al., 2019). BAF complexes are recognized as both a tumor suppressor and oncoproteins, and their components are the most frequently (∼20%) mutated chromatin regulatory proteins in human cancers (Kadoch et al., 2013). Somatic mutations in human SMARCB1 have been identified in rhabdoid tumor, and loss of SMARCB1 from the canonical BAF complex results in the formation of a rhabdoid tumor-specific BAF complex (Nakayama et al., 2017). SMARCE1, thought to be a core component of the BAF complex and which is present in all known canonical subfamilies of the BAF complex, is also mutated in meningioma (Smith et al., 2013, 2014; Gerkes et al., 2016), and genetic mutation of SMARCE1 causes Coffin–Siris syndrome (Santen et al., 2013), a multiple congenital anomaly syndrome. A previous study in Drosophila has shown that heterozygosity of BAP111, an ortholog of mammalian Smarce1, enhances the phenotype resulting from partial loss of Brm, a Drosophila homolog of mammalian Smarca2. This indicates that there is a genetic interaction between BAP111 and Brm (Papoulas et al., 2001). Mouse Smarce1 has an HMG domain in its N-terminal domain, which is predicted to direct the BAF complex to bind to appropriate genomic regions (Wang et al., 1998). However, it is largely unknown how Smarce1 affects the localization of the BAF complex within the genome, the integrity of the BAF complex, the maintenance of a pluripotent state or differentiation of ESCs.

In the present study, we conducted biochemical and cell biological analyses of Smarce1 using homozygous mutant mouse ESCs. We previously developed a method to rapidly generate homozygous mutant mouse ESC lines and constructed a homozygous mutant ESC bank consisting of ∼200 mutant cell lines (Horie et al., 2011). During the phenotypic screening of the homozygous mutant ESCs, we noticed that mutant ESCs of Smarce1, a component of the BAF complex, exhibited abnormal morphology. We observed an ectopic genomic distribution of BAF complex and the induction of instability in nucleosomes that was specific to mutant cells. Mutant cells were also impaired in proliferation and showed abnormal differentiation, accompanied by a deficit of heterochromatinization. These results suggest that Smarce1 is required to maintain the integrity of the BAF complex and that it guides the BAF complex to the appropriate genomic regions to form a proper chromatin structure for transcriptional regulation.

The structures of the of wild-type (WT), homozygous mutant (Smarce1^m/m) and revertant (Smarce1^r/r) Smarce1 alleles in the ESCs used in this study are shown in Fig. 1A. Smarce1^r/r ESCs were obtained by removing the FRT-flanked gene trap unit using Flp recombinase as reported previously (Horie et al., 2011) and were used as a control for the Smarce1-knockout phenotype. Disruption and reversion of Smarce1 were confirmed by western blot analysis (Fig. 1B).

WT ESCs formed round, dome-shaped colonies (Fig. 1C), which is a characteristic feature of undifferentiated mouse ESCs (Robinton and Daley, 2012). In contrast, Smarce1^m/m ESCs exhibited flat, irregular shaped colonies (Fig. 1C). Smarce1^r/r ESCs formed round, dome-shaped colonies similar to WT ESCs (Fig. 1C), indicating that excision of the gene trap unit reverted the ESC phenotype. We examined the expression level of pluripotency genes Oct3/4 (also known as Pou5f1), Nanog and Sox2. Although the morphology of Smarce1^m/m ESCs was different from typical undifferentiated ESCs, expression of Oct3/4 was maintained, and expression of Nanog and Sox2 was slightly increased (Fig. 1D). This observation might indicate that the chromatin structure at the pluripotency gene locus is more open in Smarce1^m/m ESCs compared to WT cells. To address this possibility, we analyzed the histone modification status of the promoter regions of Oct3/4, Nanog, and Sox2 by chromatin immunoprecipitation followed by real-time PCR (ChIP-qPCR) (Fig. 1E–G). As expected, acetylation of lysine 9 on histone H3 (H3K9ac), a marker for open chromatin (Chen and Dent, 2014), was increased in the promoter regions of Oct3/4, Nanog and Sox2 (Fig. 1E–G). However, there was no significant difference in lysine 9 trimethylation of histone H3 (H3K9me3), which is a marker for heterochromatin (Probst et al., 2009) (Fig. 1E–G). To investigate whether the alteration of histone modification is a local event or is present genome-wide, we analyzed the retroelements LINE1 and IAP (Fig. 1H–K). LINE1 and IAP are repetitive elements present in the genome at a high copy number and are known to be regulated by histone modifications (Matsui et al., 2010; Karimi et al., 2011; Sachs et al., 2019). The levels of H3K9ac and H3K9me3 in the LINE1 and IAP regions were almost the same in WT, Smarce1^m/m and Smarce1^r/r (Fig. 1H–K) except for a slight difference in IAP U3 (less than 1.3-fold; Fig. 1K). These data suggest that there is no significant broad-scale reorganization of chromatin in Smarce1^m/m ESCs.

Smarce1 contains an HMG domain, which shares homology with the yeast NHP6A protein (Corpet, 1988; Wang et al., 1998) (Fig. S1A,B). Although yeast NHP6A is not a component of the chromatin remodeling complex, physical and genetic interactions of NHP6A with the RSC chromatin remodeling complex have been reported (Szerlong et al., 2003). In addition, NHP6A mutant yeasts have been reported to have loose histone–chromatin binding (Moreira and Holmberg, 2000; Dowell et al., 2010). These observations suggest the histone–chromatin binding might also be loose in Smarce1^m/m ESCs. To address this possibility, we conducted a biochemical salt extraction assay to examine the binding strength of histones to DNA. Buffers containing different concentrations of salt were added to a nuclear solution of WT, Smarce1^m/m and Smarce1^r/r ESCs to make the final salt concentration 75–450 mM (Fig. 2A). Note that a higher-order chromatin structure is preserved at a salt concentration of 75 mM, and this structure becomes progressively destabilized as the salt concentration increases (Thoma et al., 1979; Allan et al., 1981). Histone H3 was extracted from these nuclei without cutting the genomic DNA. From the WT and Smarce1^r/r nuclei, only a small amount of histone H3 was extracted even at the highest salt concentration (450 mM) (Fig. 2B,C; Fig. S2A,B), indicating a tight association of histone H3 with DNA. Histone H3 was extracted from Smarce1^m/m nuclei with an efficiency similar to that of the extraction from WT and Smarce1^r/r nuclei at a 75 mM salt concentration, which preserves a higher-order chromatin structure (Fig. 2B,C; Fig. S2A,B). However, the extraction efficiency was increased at a moderate salt concentration (300 mM), and histone H3 was readily extracted at the highest salt concentration (450 mM) (Fig. 2B,C; Fig. S2A,B). Although a statistical difference in extraction efficiency was observed only between Smarce1^m/m and Smarce1^r/r and not between Smarce1^m/m and WT, Smarce1^m/m exhibited the highest extraction efficiency at 450 mM salt concentration in all three independent experiments (Fig. S2B). This suggested a loose association of histone H3 with DNA in Smarce1^m/m nuclei. At a salt concentration of 75 mM, we observed a modest increase in the amount of free histones present in the cytosol and nucleoplasm of Smarce1^m/m ESCs compared to WT and Smarce1^r/r ESCs (Fig. 2D–F). However, this difference was not as pronounced as that for the extraction of histone H3 from nucleosomes observed at higher salt concentrations (Fig. 2B,C; Fig. S2A,B). This observation suggests that although histone H3 is bound to chromatin in Smarce1^m/m ESCs with seemingly comparable efficiency to WT and Smarce1^r/r ESCs, the binding in Smarce1^m/m ESCs is looser and more unstable. In accordance with this observation, Arid1a, one of the components of the BAF complex, and a transcriptional repressor protein Kap1 (also known as Tif1β or Trim28) consistently showed higher extraction efficiencies in Smarce1^m/m ESCs compared to WT and Smarce1^r/r ESCs at the highest salt concentration (450 mM) in two independent experiments. This occurred even though a statistical difference was observed only between Smarce1^m/m and Smarce1^r/r ESCs for Arid1a and not for Kap1 (Fig. 2B; Fig. S2C). Unexpectedly, the amount of Kap1 extracted from the nuclei decreased with increasing salt concentration in the extraction buffer (Fig. 2B). It is possible that Kap1 or a complex containing Kap1 acquired hydrophobicity under high salt concentration and has hence have been lost from the soluble fraction due to salt precipitation (Fig. 2B). In contrast to these proteins, extraction efficiencies of Smarcc1 and Smarcc2, which were highly and lowly expressed in WT ESCs, respectively, did not show consistent differences between WT, Smarce1^m/m and Smarce1^r/r ESCs (Fig. 2B; Fig. S2C). These results showing the loose association of chromatin proteins with DNA indicate that Smarce1^m/m ESCs have unstable nucleosomes.

We then analyzed global chromatin architecture by performing an micrococcal nuclease (MNase) sensitivity assay (Fig. 2G–K). When nuclei isolation and MNase treatment were carried out in the presence of 75 mM salt, in which higher-order chromatin structure is maintained (Thoma et al., 1979; Allan et al., 1981), no discernible differences were observed in either the relative intensity or the spacing of the bands, indicating no difference in global digestion pattern of chromatin between WT, Smarce1^m/m and Smarce1^r/r cells (Fig. 2H–K). This result is consistent with the findings of the salt extraction assay in which histone H3 was tightly associated with chromatin in Smarce1^m/m nuclei in low salt (75 mM) concentrations, as in WT and Smarce1^r/r (Fig. 2B,C; Fig. S2A,B). Taken together, these results indicate that the genome-wide nucleosome positioning is unaffected in Smarce1^m/m ESCs despite weak interactions between histones and DNA. These results also suggest the possibility that histones move more dynamically along chromatin fibers in Smarce1^m/m cells than in WT and Smarce1^r/r cells, although this difference could not be detected by the MNase sensitivity assay, which primarily detects the steady-state chromatin structure condition.

Recent studies have shown that a mutation of SMARCB1 reduced the amount of ARID1A, ARID1B and DPF2 in the BAF chromatin remodeling complex (Wang et al., 2017, 2019). These studies suggest that a mutation in one component of BAF chromatin remodeling complex might alter the levels of other components. To explore the possibility that a mutation of Smarce1 induces changes in the components of esBAF chromatin remodeling complex (Fig. S1C,D), we analyzed Smarca4-interacting proteins by immunoprecipitation analysis. Nuclear extracts were prepared from MNase-treated WT, Smarce1^m/m and Smarce1^r/r cells in the presence of 150 mM salt, and were immunoprecipitated with anti-Smarca4 antibody at the same salt concentration (Fig. 3A). As a control, a normal rabbit IgG was used for a mock immunoprecipitation. Smarca4-interacting proteins were further investigated by immunoblot analysis. The amount of Arid1a precipitated with the anti-Smarca4 antibody decreased in Smarce1^m/m (Fig. 3A,B; Fig. S3), suggesting that there was a reduction in the amount of Arid1a in the BAF complex. Smarca4 successfully pulled down Arid3b, which has not been reported as a component of the BAF complex, even though no protein was detected in the input lane due to limited detection sensitivity. Brd9, a bromodomain-containing protein, has been reported to interact with Smarca4 but not with Smarce1 and to be contained in a non-canonical BAF complex called the GBAF complex (Michel et al., 2018; Alfert et al., 2019; Wang et al., 2019) (Fig. S1C,D). Therefore, we investigated the interaction between Smarca4 and Brd9 in WT, Smarce1^m/m and Smarce1^r/r cells but found no obvious differences between the three cell lines (Fig. 3A; Fig. S3). Smarca4 has also been reported to interact with repressor proteins such as the PRC2 protein Ezh2, Kap1 (Li et al., 2017) and HDAC1 (Carey et al., 2015). Weak interactions of Ezh2 and HDAC1 with Smarca4 were detected in WT, Smarce1^m/m and Smarce1^r/r (Fig. 3A). However, no interaction was detected between Kap1 and Smarca4 in the three cell lines (Fig. 3A).

Taken together, the interaction of Smarca4 with Arid1a, a component of the esBAF complex, was decreased in Smarce1^m/m cells. However, the interaction of Smarca4 with Arid3b, Ezh2, HDAC1 and Brd9, a component of the GBAF complex, was unaffected in Smarce1^m/m cells. These results suggest that a deficiency of Smarce1 specifically affects the components of the esBAF complex.

To further analyze the properties of the BAF complex in Smarce1^m/m cells, soluble chromatin from MNase-treated WT and Smarce1^m/m cells was subjected to 10–40% (w/v) sucrose gradient sedimentation analysis. Fractionated BAF component proteins and other chromatin-associated proteins prepared from Smarce1^m/m cells were compared to those of WT cells. We performed experiments with two different salt concentrations, 75 mM and 300 mM. Under 75 mM salt, chromatin is expected to maintain a high-order structure (Thoma et al., 1979; Allan et al., 1981); therefore, interactions between various proteins and chromatin will be detected. Under 300 mM salt, many proteins are expected to dissociate from chromatin.

Under the low salt concentration (75 mM), Smarca4 from Smarce1^m/m cells migrated towards both the top and bottom fractions compared to WT cells (Fig. 4A; Fig. S4). Other components of the BAF complex, Arid1a, Smarcc1, and Smarcc2, from Smarce1^m/m cells also migrated towards the top fractions, and Arid1a migrated towards the bottom fractions as well (Fig. 4A; Fig. S4). The molecular mass of Smarce1 is 46.64 kDa. Given the distribution of gel filtration molecular markers centrifuged in parallel (Fig. 4A, top), migration of the BAF complex components to the top fractions cannot be explained by the lack of Smarce1 alone. Arid1a protein was detected in fractions 4 and 6 in Smarce1^m/m, but not in WT (Fig. 4A). From the distribution of the molecular markers, the molecular mass of proteins in fraction 6 would be ∼230 kDa. Because the molecular mass of Arid1a is 242.05 kDa, the Arid1a protein detected in fraction 6 might represent a free protein dissociated from the BAF complex. This observation is consistent with the results from the immunoprecipitation assay (Fig. 3), which suggested that BAF components such as Arid1a dissociate from the complex in Smarce1^m/m ESCs. However, migration to the bottom fractions contradicted the size reduction of the BAF complex. Smarce1 has an HMGB1 domain that has DNA-binding activity (Wang et al., 1998). Therefore, when Smarce1 is disrupted, the BAF complex might incorrectly interact with chromatin. Migration of the BAF complex to the bottom fractions suggests the interaction of the BAF complex with heterochromatin regions. This unexpected migration of the BAF complex toward the bottom fractions was accompanied by the migration of the PRC2 component Suz12 to the top fractions (Fig. 4A; Fig. S4). Misregulation of Smarca4 localization might have exerted chromatin remodeling ectopically in heterochromatin regions, disrupted the chromatin platform suitable for PRC2 binding and shifted Suz12 toward the top fractions. Kap1 displayed an unusual sedimentation pattern characterized by three distinct peaks in both WT and Smarce1^m/m cells (Fig. 4A; Fig. S4). Notably, in Smarce1^m/m cells, the two peaks on the bottom side were diminished, whereas the peak on the top side was heightened in comparison to WT (Fig. S4). This observation aligns with the increased salt extraction efficiency of Kap1 in undifferentiated ESCs (Fig. 2B). In contrast, Brd9, Arid3b and HDAC1 did not shift to the bottom or top fractions (Fig. 4A; Fig. S4). This observation was consistent with the results of the immunoprecipitation assay showing that the effects of the Smarce1 deficiency were limited to components of the esBAF complex (Fig. 3), although we cannot exclude the possibility that the Smarce1-deficient BAF complex might have interacted with proteins that were not examined in this study, thereby forming a large protein complex that migrated toward the bottom fractions. Arid1b migrated toward the top fractions (Fig. 4A; Fig. S4). Arid1b is not included in the esBAF complex and is thought to be incorporated into the BAF complex during differentiation. ESCs cultured in a serum-containing medium (as in this study) are known to be metastable and exhibit considerable heterogeneity (Marks et al., 2012). The effect of Smarce1 loss on Arid1b migration implies the existence of an ESC subpopulation expressing a Arid1b-containing BAF complex.

Next, we performed the sucrose gradient sedimentation assay at the high salt concentration (300 mM) for which no interaction between Smarca4 and histone H3 was observed (Fig. S5A). At this concentration, Smarca4 from Smarce1^m/m cells migrated toward the top fractions, but not to the bottom fractions in contrast to what was seen in the low salt concentration (Fig. 4B; Fig. S5B). Other components of the BAF complex, Arid1a, Arid1b and Smarcc1 prepared from Smarce1^m/m cells, also co-migrated toward the top fractions only. This observation supports the above-mentioned notion that the migration of the BAF complex components to the bottom fractions under low salt concentrations reflects the interaction of the BAF complex with heterochromatin regions. In contrast to esBAF complex component proteins, non-esBAF complex proteins, such as Arid3b, Brd9, and repressor proteins, were not affected (Fig. 4B; Fig. S5B), indicating the specificity of the effect of Smarce1 knockout on the esBAF complex.

Undifferentiated ESCs have an open chromatin structure permissive to differentiation stimuli (Meshorer et al., 2006). Upon differentiation stimuli, appropriate genomic regions become heterochromatin, and a chromatin structure specific to each cell type is established (Meshorer et al., 2006; Probst et al., 2009). Undifferentiated Smarce1^m/m ESCs did not show any obvious differences in the levels of the heterochromatin markers H3K9me3 and H4K20me3 (Fig. S6). However, abnormal protein composition of the esBAF complex (Fig. 3) and ectopic distribution of repressor proteins (Fig. 4A) in undifferentiated Smarce1^m/m ESCs suggested that the reorganization of chromatin structure upon differentiation stimuli might be impaired in Smarce1^m/m ESCs. Therefore, we investigated the phenotypes of Smarce1^m/m ESCs during differentiation, with a particular focus on changes in chromatin structure.

WT, Smarce1^m/m and Smarce1^r/r ESCs were cultured in hanging drops for 3 days to form embryoid bodies (EBs), and microscopic images were taken for measuring the surface area. Although WT and Smarce1^r/r cells developed equally, the surface area of the Smarce1^m/m EBs was smaller than that of WT and Smarce1^r/r (Fig. 5A,B), indicating a delay in the proliferation of mutant cells. To investigate the differentiation potential of the Smarce1^m/m cells, EBs were transferred onto gelatin-coated coverslips, cultured for an additional 7 days, and stained for mesodermal (α-smooth muscle actin) (Fig. 5C,D) (Hinz et al., 2001) and ectodermal (β-III tubulin) (Fig. 5E,F) (Roskams et al., 1998) markers. The size of the EB outgrowths was smaller in Smarce1^m/m cells compared to that of WT and Smarce1^r/r cells (Fig. S7A,B). WT and Smarce1^r/r cells succeeded in differentiating into α-smooth muscle actin-positive cells, and the differentiated cells were square-shaped (Fig. 5D), which is a typical morphology observed in normal differentiation (Tomov et al., 2016), and appeared throughout the colonies. In contrast, α-smooth muscle actin-positive cells were observed in the peripheral area of the Smarce1^m/m colonies, and they were elongated and rectangular in shape (Fig. 5D). Consistent with the abnormal differentiation, more Nanog-positive undifferentiated cells were observed at the center of the Smarce1^m/m colonies than for WT and Smarce1^r/r colonies (Fig. 5C). In addition to Nanog, other pluripotency markers, namely Oct3/4 and Sox2, continued to be highly expressed after the induction of differentiation (Fig. 5G). These results indicate the defective differentiation of Smarce1^m/m into mesodermal lineages and the persistence of undifferentiated cells. Regarding the differentiation into ectodermal lineages, β-III tubulin-positive cells were observed at the periphery of the colonies for WT and Smarce1^r/r, whereas β-III tubulin-positive cells were found within almost the entire region of the colonies for Smarce1^m/m (Fig. 5E,F). Characteristically, neurite-like structures were prominent in Smarce1^m/m colonies and surrounded Nanog-positive cells (Fig. 5E,F), which was not observed in WT and Smarce1^r/r cells. A recent study has shown that there is enhanced neuronal differentiation in human ARID1A mutant ESCs (Liu et al., 2020). The abundant neurite-like structures observed in Smarce1^m/m cells might have been caused by a reduced amount of Arid1a in the BAF complex (Figs 3, 4). We assessed several additional lineage-specific markers through quantitative RT-PCR (qRT-PCR) and noted altered expression levels, either decreased or increased, in Smarce1^m/m cells upon the induction of differentiation (Fig. 5G). This provides further evidence for the abnormal differentiation of Smarce1^m/m cells.

Next, we investigated heterochromatin formation during ESC differentiation. Upon the stimulation of differentiation, pericentromeric heterochromatin foci identified by DAPI-staining increase in number, become smaller and form discrete structures (Novo et al., 2016). These foci show constitutive heterochromatin as evidenced by immunostaining with H3K9me3 (Probst et al., 2009) and H4K20me3 (Jørgensen et al., 2013) (Fig. 6A,B). We compared the morphology of these foci in WT, Smarce1^m/m and Smarce1^r/r cells. WT and Smarce1^r/r cells formed discrete and round foci (Fig. 6C). In contrast, the foci of Smarce1^m/m cells showed a distorted shape (Fig. 6C). To quantitatively assess the shape of the DAPI-staining foci, we determined the circularity of these foci (Fig. 6D). The circularity of the foci in Smarce1^m/m cells was lower compared to that in WT and Smarce1^r/r cells, suggesting the impaired formation of constitutive heterochromatin. We confirmed that the cell cycle is not affected in Smarce1^m/m cells (Fig. 6E). This eliminates the possibility that the distorted pericentromeric foci were caused by the effect of the Smarce1 mutation on the cell cycle.

To further investigate the impaired formation of heterochromatin in differentiated cells, we conducted a biochemical analysis (Fig. 6F,G). Histone H3 and the repressor proteins Kap1, Ezh2 and HDAC1 were extracted from the nuclei of differentiated cells at various salt concentrations. A greater quantity of histone H3 was extracted from Smarce1^m/m cells than from WT and Smarce1^r/r, suggesting a loose chromatin structure in Smarce1^m/m cells (Fig. 6F,G). As observed in undifferentiated ESCs (Fig. 2B), the amount of Kap1 extracted from the nuclei decreased with increasing salt concentration in the extraction buffer (Fig. 6F). Consistent with the morphological abnormality of the constitutive heterochromatin foci (Fig. 6A–D), Kap1 was more readily extracted from Smarce1^m/m cells than WT and Smarce1^r/r cells (Fig. 6F,G). Kap1 localizes throughout the nucleoplasm in undifferentiated cells and translocates to heterochromatic regions after differentiation, co-localizing with DAPI foci (Cammas et al., 2002; Herzog et al., 2011). We found that, around the peripheral region of the EBs, Kap1 accumulated in the nucleus and colocalized with DAPI foci in both Smarce1^m/m and Smarce1^r/r (Fig. S7C). In contrast, Kap1 foci were not detected at the center of Smarce1^m/m EBs (Fig. S7C), where Nanog-positive cells were enriched (Fig. S7A,B). These results support the biochemical analysis showing the decreased affinity of Kap1 to chromatin in differentiated Smarce1^m/m cells (Fig. 6F,G). Ezh2, an integral component of PRC2 that regulates facultative heterochromatin, and HDAC1, which is associated with both constitutive and facultative heterochromatins, were also more readily extracted from Smarce1^m/m cells than from WT and Smarce1^r/r cells (Fig. 6F,G), suggesting that heterochromatin formation was broadly impaired in Smarce1^m/m cells. It should be noted that the Ezh2 and HDAC1 were not affected in undifferentiated ESCs in all assays performed, such as the salt extraction assay (Fig. 2B), immunoprecipitation with Smarca4 (Fig. 3), and in the sucrose gradient sedimentation analysis (Fig. 4). This observation indicates that the abnormality of repressor protein functions in Smarce1^m/m cells becomes more evident in a differentiated state. To confirm the impaired formation of heterochromatin, we analyzed the global chromatin architecture by an MNase sensitivity assay at a salt concentration of 75 mM. Chromatin is expected to maintain a higher-order structure under this condition (Thoma et al., 1979; Allan et al., 1981) (Fig. 6H). We could not detect any significant differences in the nucleosome repeat length in WT, Smarce1^m/m or Smarce1^r/r cells (Fig. 6I). However, the band intensities of digested DNA fragments that are longer than tetra-nucleosomes were decreased in Smarce1^m/m cells compared to those for WT and Smarce1^r/r cells (Fig. 6H,J). These results demonstrate higher MNase sensitivity of Smarce1^m/m cells than WT and Smarce1^r/r cells, and indicate a global impairment in heterochromatin formation in Smarce1^m/m cells (Fig. 6H–J).

Our biochemical analysis revealed the instability of chromatin proteins in Smarce1^m/m cells (Figs 2B,C, 4, 6F,G). However, these findings do not yet provide insight into the stability of chromatin proteins within living cells. Thus, we investigated the stability of chromatin proteins in live cells using fluorescence recovery after photobleaching (FRAP).

Utilizing the histone H2B–mCherry fusion protein, we investigated the kinetics of histone H2B binding to chromatin after differentiation induction. Notably, the histone H2B–mCherry signal displayed an accelerated recovery after photobleaching in Smarce1^m/m cells (Fig. 6K). This observation suggests an intensified entry and exit of histone H2B from chromatin in Smarce1^m/m cells, corroborating the observed heterochromatin disorder in these cells (Fig. 6A–C). A previous report indicates that there is a decrease in the dynamic behavior of the histone H2B during ESC differentiation (Meshorer et al., 2006). Consequently, it is plausible that the heightened dynamic state of histone H2B contributes to the defective differentiation observed in Smarce1^m/m cells. Furthermore, the successful detection of abnormalities in Smarce1^m/m cells through FRAP implies a global occurrence of nucleosome instability across a substantial chromatin region.

Our biochemical analysis also demonstrated the instability of Ezh2 and Suz12, components of the polycomb complex, in Smarce1^m/m cells (Figs 4A, 6F,G). Although we conducted a FRAP analysis of EGFP–EZH2, no discernible enhancement in the dynamic state was observed in Smarce1^m/m cells (Fig. S8A,B). Unexpectedly, forced EGFP–EZH2 expression increased the proliferative potential of Smarce1^m/m cells (Fig. S8C). It is plausible that the overexpression of EZH2 might have masked the Smarce1^m/m phenotype.

We observed more undifferentiated cells within the outgrowths of the EBs in Smarce1^m/m cells (Fig. 5C,E). This result is consistent with the observation that the promoter regions of pluripotency genes were more open in Smarce1^m/m cells than in WT and Smarce1^r/r cells (Fig. 1E–G). To further characterize these undifferentiated cells, we dissociated the outgrowths of EBs and replated them under ESC culture conditions (Fig. 7). Strikingly, we observed a substantially higher number of colonies from Smarce1^m/m cells compared to WT and Smarce1^r/r cells (Fig. 7A–D). Moreover, these cells were positive for pluripotency markers as determined by immunostaining (Fig. 7A,B) and qRT-PCR (Fig. 7E). These results indicate that Smarce1 deficiency stabilizes the undifferentiated state of ESCs through defective heterochromatin formation.

The current study showed that disruption of Smarce1 decreases nucleosome stability in mouse ESCs and impairs heterochromatin formation during differentiation (Fig. 8). Smarce1 contains an HMG domain that has been shown to interact with DNA (Wang et al., 1998). Other components of the esBAF complex, such as Arid1a, Smarcb1, Smarca4 and Dpf2, also contain DNA-binding domains (Nie et al., 2000; Agalioti et al., 2002; Wilsker et al., 2004; Han et al., 2014; Allen et al., 2015; Mathur et al., 2017; Morrison et al., 2017; Nakayama et al., 2017) or histone-binding domains (Agalioti et al., 2002; Bonaldi et al., 2002; Shen et al., 2007; Xiong et al., 2016; Local et al., 2018). The genomic distribution of the BAF complex is thought to be determined by the overall effect of these BAF complex components. Given that the genomic distribution of the BAF complex seemed altered in the absence of Smarce1 as determined by the sucrose gradient sedimentation assay (Fig. 4A), we speculate that Smarce1 serves as a guide for placing the BAF complex in the appropriate genomic regions. We hypothesize that the Smarce1^m/m-specific BAF complex might exert remodeling effects on ectopic genomic regions, slide histones along the DNA, and induce the loosening of chromatin structure (Fig. 8). It has been reported that loosely structured chromatin is not suitable for the nucleosome binding of Polycomb group proteins (Cao and Zhang, 2004; Martin et al., 2006). Recent studies have also shown that ectopic recruitment of the BAF complex to chromatin to which Polycomb group proteins are already bound leads to the release of Polycomb group proteins (Kadoch et al., 2017; Stanton et al., 2017). Therefore, we speculate that the enhanced release of Polycomb group proteins from chromatin observed in Smarce1^m/m nuclei (Figs 4A, 6F) was caused by the ectopically distributed Smarce1^m/m-specific BAF complex.

Mutation of BAF complex components induces tumorigenesis (Roberts and Orkin, 2004; Wilson and Roberts, 2011; Kadoch et al., 2013; Kadoch and Crabtree, 2015; Alfert et al., 2019). For example, SS18, a component of the BAF complex, is reported to fuse to SSX family members through chromosomal translocation and causes synovial sarcoma. The mutant BAF complex containing such SS18–SSX fusion proteins evicts PRC2 from PAX3 and SOX2 loci, decreases H3K27me3 levels and increases the expression of these genes (McBride et al., 2018). A mutation of SMARCE1 has been reported to cause meningiomas (Smith et al., 2013, 2014; Gerkes et al., 2016). Given the similarities to synovial sarcoma formation, the ectopic distribution of the BAF complex and the eviction of PRC2 observed in Smarce1^m/m in the current study (Figs 4A, 6F) might be responsible for meningioma formation.

The HMG domain of mouse Smarce1 has homology with the HMG box of yeast NHP6A and NHP6B (Fig. S1A,B) (Wang et al., 1998). NHP6A and NHP6B physically and genetically interact with the yeast RSC chromatin remodeling complex, which is closely related to the mammalian BAF complex (Wang et al., 1996, 1998). In addition, the synthetic lethality of triple mutations of the yeast catalytic subunit of the Swi/Snf complex, NHP6A and NHP6B indicates that there is a genetic interaction between these factors (Biswas et al., 2004). Furthermore, it has been reported that the association of histone to chromatin is loosened in yeast NHP6A mutant cells (Moreira and Holmberg, 2000; Dowell et al., 2010), which resembled our finding in Smarce1^m/m cells. A yeast ortholog protein of mouse Smarce1 has not been reported thus far. Similarities between yeast NHP6A and mouse Smarce1 suggest that NHP6A might be the functional yeast counterpart of mouse Smarce1.

We observed a reduced association of Arid1a to the BAF complex in Smarce1^m/m cells (Figs 3, 4). Interestingly, a reduction of Smarce1 in the BAF complex has been reported in Arid1a mutant cells (Mathur et al., 2017). These observations suggest a strong interaction between Smarce1 and Arid1a. A recent study has shown that the prior presence of Smarce1 in the core of the BAF complex is required for the efficient recruitment of Arid1a to form the canonical BAF complex (Mashtalir et al., 2018). The impaired association of Arid1a with the BAF complex observed in Smarce1^m/m in the current study supports this concept. As mentioned above, both Smarce1 and Arid1a possess a DNA-binding domain (Wang et al., 1998; Dallas et al., 2000; Nie et al., 2000; Wilsker et al., 2004; Mathur et al., 2017). The combined loss of the two DNA-binding domains in Smarce1^m/m might exacerbate the misregulation of the BAF complex and contribute to various phenotypes or disease, such as the formation of meningioma in humans.

Recently, the effect of SMARCE1 deficiency on the BAF remodeling complex was investigated in a study that used human cell lines and aimed to elucidate the pathogenic role of SMARCE1 deficiency in meningioma (St Pierre et al., 2022). In line with our findings, that study reported the dissociation of ARID1A from the BAF complex in SMARCE1-deficient cells. Furthermore, that study reported an enhanced formation of the BRD9-containing non-canonical BAF complex and detected specific genomic binding sites associated with this complex. Although we did not detect an increased interaction between Smarca4 and Brd9 (Fig. 3), the findings of that study suggest the possibility that the composition of the BAF complex is altered in Smarce1^m/m cells.

Accumulation of histone acetylation was detected in the promoter regions of the pluripotent factors Nanog, Oct3/4 and Sox2 (Fig. 1E–G). In contrast to these regions, the difference in histone acetylation at the loci of the retroelements IAP and LINE1 was minimal, if detected at all, between Smarce1^m/m and WT or Smarce1^r/r (Fig. 1H–K) even though the repressor protein Kap1, which has been reported to repress these retroelements (Rowe et al., 2010), readily dissociated from chromatin in the salt extraction assay (Fig. 2B). These results suggest that although Kap1 was easily extracted from Smarce1^m/m nuclei, an additional, unidentified silencing mechanism exists for IAP and LINE1.

After the differentiation of ESCs, morphologically more mature neuronal cells were observed for Smarce1^m/m compared with WT and Smarce1^r/r (Fig. 5E,F). As described above, we observed a reduction in the amount of Arid1a associated with the BAF complex in Smarce1^m/m cells (Figs 3, 4). A previous report has shown that there is enhanced neuronal differentiation of ARID1A knockout human ESCs due to an impaired interaction between ARID1A and REST, a repressor of neuronal differentiation (Liu et al., 2020). Furthermore, human SMARCE1 has been reported to interact with REST and is required for REST-mediated repression of neuronal genes (Battaglioli et al., 2002). Based on these reports, we speculate that the function of Rest is impaired in Smarce1^m/m cells because of the reduction of Arid1a and complete loss of Smarce1 in the BAF complex, thus leading to the enhanced neuronal differentiation (Fig. 5E,F). Both SMARCE1 and ARID1A are causative genes for Coffin–Siris syndrome (Santen et al., 2013), a multiple congenital anomaly syndrome. The impaired proliferation and abnormal differentiation of Smarce1^m/m ESCs observed in the present study (Fig. 5) might be associated with some of the developmental disorders of Coffin–Siris syndrome. Single-cell RNA-seq would provide more detailed information on the abnormal differentiation of Smarce1^m/m cells and would be useful in considering the relationship between the findings of the present study and human diseases.

Our observations in Smarce1 mutant cells revealed not only the role of Smarce1 for maintaining the BAF complex integrity but also the functions of the BAF complex itself in the formation of a suitable chromatin environment for transcriptional regulation in undifferentiated and differentiated cells. Further studies using cells that are mutant for other components of the BAF complex will help to elucidate new functions of each component in the maintenance of the BAF complex integrity and chromatin structure formation.

A Smarce1 heterozygous mouse ESC clone (Smarce1^m) was obtained using the piggyBac transposon-based gene trap vector containing the same gene trap unit we used previously (Horie et al., 2011). The piggyBac gene trap vector was generated as follows. First, a 0.82-kb BglII–ApaI fragment of pT2F2GFP (Horie et al., 2011) containing the FRT-flanked GFP gene was inserted into the BglII–ApaI site of pPB-MCS-P5 (Yoshida et al., 2017), resulting in pPB-F2GFP. Next, a 4.8-kb XhoI–PmlI fragment of the Tol2 gene trap vector pT2F2-SAhygpA-N22 (Horie et al., 2011) was cloned into the XhoI–PmlI site of the pPB-F2GFP located between the two inverted terminal repeats of the piggyBac transposon, resulting in pPB-SAhygA-NP22. Gene trapping was conducted as described previously (Horie et al., 2011) and the ESC clone containing the vector insertion at the first intron of the Smarce1 gene was identified. The flanking genomic sequence of the vector insertion site is 5′-TTAATCGCCCCGAGACTGTTTTCTTCC-3′.

To construct the expression vector for H2B–mCherry, the NheI–SacI fragment containing the H2B–mCherry coding region was excised from the ROSA26-H2B-mCherry vector (Ueda et al., 2014) blunt-ended, and cloned into the blunt-ended EcoRI site of the piggyBac transposon vector pPB-CAG-IB (unpublished; details for unpublished vectors are available upon request), which contains the CAG promoter (Niwa et al., 1991) for the expression of the gene of interest and a drug selection cassette consisting of the internal ribosome entry site from encephalomyocarditis virus, the blasticidin S-resistance gene and the polyadenylation signal from the mouse phosphoglycerate kinase 1 gene, resulting in the pPB-CAG-H2B-mCherry-IB. To construct an expression vector for EGFP–EZH2, the EGFP–EZH2 coding region was amplified by PCR from pcDNA3-EGFP-EZH2-V5 (unpublished) using primers shown in Table S1. The PCR product was digested with XhoI and NotI, and cloned into the XhoI–NotI site of the piggyBac transposon vector pPB-CAG-L5-EGFP-IB (unpublished) which contains the same promoter and the drug selection cassette as the pPB-CAG-IB described above, resulting in the pPB-CAG-EGFP-EZH2-IB. These expression vectors were stably introduced into ESCs using the piggyBac transposase as described previously (Yoshida et al., 2017).

Smarce1 homozygous mutant ESCs (Smarce1^m/m) were obtained by doxycycline-induced interchromosomal recombination as described previously (Horie et al., 2011). Revertant ESCs (Smarce1^r/r) were obtained by excising the gene trap unit using Flp-mediated recombination as described previously (Horie et al., 2011). EBs were formed in hanging drops containing 1000 cells in 20 µl of differentiation medium (Horie et al., 2011) in the absence of leukemia inhibitory factor (LIF) on the lid of a culture dish and cultured for 3 days. EBs were cultured for a further 7 days on gelatin-treated coverslips. For the salt extraction assay, differentiation was induced by culturing ESCs (4.0×10⁵ cells) on a low attachment cell culture dish (Greiner, CELLSTAR, Cell-Repellent Surface, 628979) to form EBs in the absence of LIF for 3 days. For further induction of differentiation, EBs were cultured on a gelatin-treated dish for another 7 days. After transferring EBs to the gelatin-treated coverslips or culture dishes, we continued to culture in the same differentiation medium, which was changed every 2 days. To examine the persistence of pluripotent cells within EB outgrowths, we dissociated EB outgrowths, obtained a single-cell suspension, and plated 2×10⁴ cells per well in a 12-well plate. After 7 days, we counted the number of colonies, extracted RNA for RT-PCR analysis, and performed immunostaining to evaluate pluripotency markers. The parental ESC line utilized for genome engineering was verified to be free of mycoplasma contamination.

FRAP was conducted according to the previous report with some modifications (Kimura and Cook, 2001). WT and Smarce1^m/m cells stably expressing H2B–mCherry and EGFP–EZH2 were induced for differentiation using the hanging drop method as described above. After three days, EBs were transferred to a gelatin-coated ibidi µ-Dish (Ibidi, 81156). Before observation, the cell culture medium was replaced with a fresh medium. For observing EGFP–EZH2, FluoroBrite DMEM medium (Thermo Fisher Scientific, A1896701) was used as the basal medium to enhance signal detection sensitivity. For cells expressing H2B–mCherry, the FRAP experiment was performed using the Nikon AXR confocal system equipped with an inverted Nikon Ti2 microscope with a 40× Plan Apo lambda D, NA=0.95 lens. Images of nuclei (512×512 pixels) were acquired with the following settings: 561 nm laser power 0.4, scan zoom 6, 4 AU pinhole size with maximum scan speed, and four times line average. Before photobleaching, five nucleus images were acquired for control. Photobleaching of half of the nucleus was performed with eight iterations of 100% of 561 nm laser power during 8.06 s. After photobleaching, images were captured every 1 min for a duration of 1 h. Throughout the observation, the condition of the cell culture was maintained using a stage-top incubator system (STXG-TIZSHX-SET, Tokai Hit). For cells expressing EGFP–EZH2, imaging was carried out using a 40× Apo water immersion lambda S, NA=1.25 lens. Images of nuclei (256×256 pixels) were acquired with the following settings: 488 nm laser power 4.5, scan zoom 6, 4 AU pinhole size with maximum scan speed, and two times line average. Before photobleaching, five images of nuclei were acquired for control. Photobleaching of a portion of the nucleus was achieved with one stimulation of 100% of 488 nm laser power during 625.35 ms. Following photobleaching, images were captured every 3 s for 175 s.

Cells were trypsinized and washed with PBS. The cell pellet was solubilized with 8 M urea containing 0.1 M NaH₂PO₄, 10 mM Tris-HCl pH 8.0, cOmplete^TM EDTA-free protease inhibitor cocktail (Roche, 11873580001), and 0.1 mM PMSF. The amount of protein was measured by the Bradford method (Bio-rad, 500-0001) using BSA as a standard. An equal amount (15 µg) of each protein was subjected to immunoblot analysis as described below.

Undifferentiated and differentiated ESC nuclei were prepared as described elsewhere with some modifications (Kashiwagi et al., 2011). Cells were washed with PBS and treated with trypsin for dissociation. Trypsin treatment was terminated by adding 10% calf serum-containing medium. Cells were harvested by centrifugation at 300 g for 5 min at room temperature and washed with PBS. Cells were collected again as described above, resuspended, and washed with ice-cold nuclei isolation buffer (NIB) containing 10 mM Tris-HCl pH 7.5, 60 mM KCl, 15 mM NaCl, 1.5 mM MgCl₂, 1 mM CaCl₂, 0.25 M sucrose, 10% (v/v) glycerol, 10 mM sodium butyrate, 1 mM dithiothreitol (DTT), 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail. Cells were collected by centrifugation at 300 g for 5 min at 4°C and re-suspended in NIB. An equal volume of NIB containing 0.2% (v/v) NP40 buffer was then added to cell suspensions to bring the final concentration of NP40 to 0.1% (v/v). Cells were incubated on ice for 10 min and centrifuged at 300 g for 5 min at 4°C. Supernatants containing cytoplasmic proteins were discarded. Pelleted nuclei were resuspended in NIB and centrifuged again at 300 g for 5 min at 4°C. Finally, nuclei were resuspended in NIB.

Nuclei were collected as described above, and a small amount of nuclei solution was taken into saturated 5 M NaCl in 8 M urea buffer to measure the DNA concentration by determining the UV absorbance at 260 nm (20 OD₂₆₀ units corresponded to 1 mg/ml DNA) (Ura and Kaneda, 2001). The DNA concentration of the nuclei solution was adjusted to 1.5 mg/ml DNA with NIB. An equal number of nuclei in NIB was divided into four tubes and extracted with an equal volume of nuclei extraction buffer (NEB) containing 10 mM Tris-HCl pH 7.5, 10 mM EDTA, 0.25 M sucrose, 10% (v/v) glycerol, 10 mM sodium butyrate, 1 mM DTT, 0.1 mM PMSF, cOmplete^TM EDTA-free protease inhibitor cocktail, and different concentrations of NaCl (75, 225, 525 or 825 mM NaCl). The resulting salt (KCl with NaCl) concentration of each tube was 75, 150, 300, or 450 mM, respectively. After overnight incubation on ice, nuclei were subjected to centrifugation at 20,000 g for 15 min at 4°C. The supernatant fraction was collected, and the nuclear pellet was dissolved in 8 M urea buffer containing 0.1 M NaH₂PO₄, 10 mM Tris-HCl pH 8.0, 0.1 mM PMSF, and cOmplete^TM EDTA-free protease inhibitor cocktail. Equal samples in terms of initial nuclei number of each fraction were subjected to SDS-PAGE and then analyzed by immunoblot analysis as described below.

An equal number of nuclei in NIB was divided into five tubes and pre-incubated at 30°C for 10 min. The nuclei were treated with 20, 40, 80, 120 or 160 units/mg DNA of MNase (Worthington Biochemical, LS004797) at 30°C for 10 min. The reaction was terminated by adding EDTA to a final concentration of 5 mM. The MNase-treated DNA samples were treated with 20 µg/ml RNase (Nippon Gene, 313-01461) at 37°C for 1 h and then 40 µg/ml proteinase K at 56°C overnight. On the following day, DNA samples were further extracted twice with 25:24:1 phenol-chloroform-isoamyl alcohol and then extracted once with chloroform-isoamyl alcohol. The extracted DNA samples were precipitated with ethanol and analyzed by 1.5% agarose gel electrophoresis in 1× TAE buffer. The DNA was visualized with ethidium bromide using a UV trans-illuminator. Quantitative analyses of the band intensity and band spacing were performed using a Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, USA).

Isolated nuclei in NIB were pre-incubated at 30°C for 10 min and subjected to MNase (20 units/mg DNA) treatment at 30°C for 10 min. After MNase treatment, NEB225 or NEB525, defined as containing 225 mM NaCl or 525 mM NaCl, respectively, was added to the nuclei solution and incubated overnight on ice. The resulting salt (KCl with NaCl) concentration of the nuclei solution was 150 mM or 300 mM, respectively. The nuclear extract was separated by centrifugation at 12,800 g at 4°C for 10 min. An equal volume of NEB150 or NEB300, containing 150 mM NaCl or 300 mM NaCl, respectively, with 0.2% (v/v) NP40 was added to the nuclear extract to bring the final NP40 concentration to 0.1% (IP buffer). Antibodies against mouse Smarca4 (1.5 µg, Abcam 110641) were added to the nuclear extract and then incubated overnight at 4°C with rotation. As a negative control, an equal amount of normal rabbit IgG (MBL, PM035) was added to the nuclear extract. The next day, Dynabeads protein G (Thermo Fisher Scientific, 10003D) pre-equilibrated with IP buffer were added to the nuclear extract and incubated at 4°C for 4 h with rotation. Proteins that did not bind to the Smarca4 antibodies were separated by placing the Smarca4-associated proteins–Dynabead complexes on a magnet. The complexes were washed three times with IP buffer at 4°C with rotation for 10 min. Smarca4-associated proteins were collected by placing the complexes on a magnet and eluted with a Laemmli SDS-PAGE sample buffer (Laemmli, 1970). The eluted samples were subjected to SDS-PAGE and immunoblot analysis as described below.

Isolated nuclei were treated with MNase (20 units/mg DNA) and extracted with NEB75 or NEB525, defined as containing 75 mM NaCl or 525 mM NaCl, respectively, on ice overnight. The resulting salt concentration (KCl with NaCl) in each extract was 75 and 300 mM, respectively. The next day, the extracts were subjected to centrifugation at 12,800 g for 10 min at 4°C. The supernatants were further overlaid onto 10–40% (w/v) sucrose gradient buffer containing NEB75 or NEB300 (NaCl only) and centrifuged at 50,000 rpm for 3 h at 4°C using a TLS-55 rotor (Beckman). After centrifugation, equal volumes of each fraction were collected from the top of the centrifugation tube. The fractionated samples were mixed with Laemmli SDS-PAGE sample buffer and subjected to SDS-PAGE and immunoblot analysis as described below.

Protein samples dissolved in Laemmli buffer were separated on SDS-PAGE gel and transferred to Immobilon-P PVDF membrane (Millipore, IPVH00010). Transferred protein samples were detected using the following primary antibodies: anti-Smarca4 (1:2000; Abcam, 110641), anti-Arid1a (1:2000; Cell Signaling Technology, 12354), anti-Arid1b (1:2000; Cell Signaling Technology, 92964), anti-Arid3b (1:2000; Bethyl Laboratories Inc., A302-565A), anti-Smarcc1 (1:2000; Cell Signaling Technology, 11956), anti-Smarcc2 (1:2000; Cell Signaling Technology, 12760), anti-Smarce1 (1:2000; Cell Signaling Technology, 33360), anti-Brd9 (1:2000; Active Motif, 61537), anti-Ezh2 (1:2000; Cell Signaling Technology, 5246), anti-Suz12 (1:2000; Cell Signaling Technology, 3737), anti-HDAC1 (1:2000; Millipore, 06-720), anti-Kap1 (1:5000; Active Motif, 61173), anti-LaminB1 (1:400; Santa Cruz, Sc-20682), anti-β actin (1:4000; Sigma, A5441 for Fig. 1, 1:4000; Cell Signaling Technology, 3700 for Fig. S7), anti-H3K9me3 (1:500; Abcam, ab8898), anti-H4K20me3 (1:1000; Abcam, ab9053), and rat anti-Histone H3 serum (1:8000; made in-house by H.K.). Membrane-bound primary antibodies were detected using horseradish peroxidase-conjugated anti-rabbit IgG (Cytiva, NA934), anti-mouse IgG (Cytiva, NA931) and anti-rat IgG (Bethyl Laboratories Inc., A110-305P). Immunoreactive signals were detected using Chemi-Lumi One L (Nacalai Tesque, 07880), Chemi-Lumi One Ultra (Nacalai Tesque, 11644), ECL Western Blotting Detection Reagent (Cytiva, RPN2109) or ECL prime Western Blotting Detection Reagent (Cytiva, RPN2232). Original images for blots presented in this paper are shown in the blot transparency section (Fig. S9).

EBs were seeded on 0.1% (w/v) gelatin-coated coverslips. Cells were fixed with 4% paraformaldehyde, 100 mM HEPES-HCl pH 7.4 buffer for 20 min at room temperature and were washed twice with PBS. After fixation, cells were permeabilized with 0.5% (v/v) Triton X-100 in PBS for 20 min at room temperature and were washed with PBS. Cells were further blocked with Blocking One-P (Nacalai Tesque, 05999-84) for 20 min at room temperature and then incubated overnight at 4°C with the following primary antibodies in antibody dilution buffer (PBS containing 0.1× Blocking One-P) as indicated: anti-Kap1 (1:500; Active motif, 61173), anti-H3K9me3 (2F3) (1:1000; Kimura et al., 2008), anti-H4K20me3 (27F10) (1:1000; Kimura et al., 2008), anti-β-III tubulin (1:125; R&D Systems, MAB1195), anti-α-smooth muscle actin (1:250; Sigma-Aldrich, A5228), anti-Sox2 (1:500, Abcam, ab97959) and anti-Nanog (1:250; ReproCell, RCAB002P-F). After six washing steps with PBST [PBS with 0.1% (v/v) Tween-20] for 5 min each, cells were incubated with the following fluorescence-conjugated secondary antibodies in antibody dilution buffer as indicated: goat anti-mouse IgG Highly cross-adsorbed secondary antibody, Alexa Fluor 488 (1:1000; Thermo Fisher Scientific, A-11029), goat anti-rabbit IgG cross-adsorbed secondary antibody, Alexa Fluor 546 (1:1000; Thermo Fisher Scientific, A-11071) and goat anti-rabbit IgG cross-adsorbed secondary antibody, Alexa Fluor 594 (1:1000; Thermo Fisher Scientific, A-11012). Cells were washed six times with PBST for 5 min each, and were counterstained with 300 nM DAPI. Cells were placed on coverslips, washed with PBS and Milli-Q water, and then mounted on glass slides with ProLong Gold mounting medium (Thermo Fisher Scientific, P36934). Cells were analyzed with a Nikon C2 confocal microscopy system (Nikon), Nikon AXR confocal microscopy system (Nikon) and Keyence BZ-X810 fluorescence microscope (Keyence).

To analyze the expression of pluripotency genes in undifferentiated ESCs (Fig. 1), total RNA was extracted with RNeasy Plus Mini Kits (Qiagen, 74134) and reverse-transcribed with SuperScript IV (Thermo Fisher Scientific, 18090010) using random primers (Promega, C1181). Expression levels of mRNAs encoding Oct3/4, Nanog, Sox2 and β-actin (Actb) were analyzed by real-time PCR on a LightCycler (Roche Diagnostics) using the LightCycler FastStart DNA Master SYBR Green I kit (Roche Diagnostics, 03003230001). The amplification condition for Oct3/4 was 10 min at 95°C for one cycle, followed by 40 cycles of 10 s at 95°C, 5 s at 60°C, and 10 s at 72°C. The conditions for Nanog, Sox2 and Actb were similar except that the extension step was 20 s at 72°C for Nanog and Sox2, and the annealing step was 5 s at 55°C for Actb. To analyze gene expressions after the differentiation of ESCs (Figs 5, 7), total RNA was extracted with a TRIzol reagent (Thermo Fisher Scientific, 15596018). The extracted RNA was treated with deoxyribonuclease (Nippon Gene, 313-03161) to remove the co-extracted genomic DNA and subsequently, it was reverse-transcribed with PrimeScript^TM RT Master Mix (Takara, RR036A). Expression levels of mRNAs were analyzed by real-time PCR on a StepOne Plus real-time PCR system (Thermo Fisher Scientific) using the PowerUp SYBR Green Master Mix for qPCR (Thermo Fisher Scientific, A25742). The amplification conditions for Nanog, Sox2, Oct3/4, Gata4, nestin (Nes), α-smooth muscle actin (Acta2) and Hand1 were 2 min at 50°C for one cycle and 2 min at 95°C for one cycle, followed by 50 cycles of 15 s at 95°C and 1 min at 60°C. The condition for Actb was 2 min at 50°C for one cycle and 2 min at 95°C for one cycle, followed by 40 cycles of 15 s at 95°C, 15 s at 55°C, and 1 min at 72°C. Primer sequences are shown in Table S1. The quantity of each transcript was measured by a standard curve and normalized to the Actb transcript.

The chromatin immunoprecipitation (ChIP) assay was carried out as described previously with some modifications (Kimura et al., 2008; Blecher-Gonen et al., 2013). Briefly, cells were fixed by adding a methanol-free 16% formaldehyde to the cell culture medium to a final concentration of 1% with gentle shaking at 25°C for 10 min. After fixation of cells, 2.5 M glycine solution was added to the medium to a final concentration of 0.15 M and incubated at 25°C for 5 min. Cells were washed twice and suspended in PBS and then were collected by scrapping into tubes. Cells were further collected by centrifugation at 300 g at 4°C for 5 min. The collected cells were snap-frozen in liquid nitrogen and were stored in a −80°C deep freezer until use. Before use, cells were defrosted on ice for 10 min. To prepare the nuclear extract, lysis buffer 1 containing 50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 0.25% (v/v) Triton X-100, 0.5% (v/v) NP40, 10% (v/v) glycerol, 10 mM sodium-butyrate, 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail was added to the defrosted cells and incubated on ice for 10 min. Cells were then collected by centrifugation at 800 g at 4°C for 5 min. Cell pellets were resuspended in lysis buffer 2 containing 50 mM HEPES-KOH pH 7.5, 200 mM NaCl, 1 mM EDTA, 10 mM sodium butyrate, 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail and incubated on ice for 10 min. Cells were collected again by centrifugation as described above. Finally, cells were extracted with lysis buffer 3 containing 50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 0.1% (w/v) sodium deoxycholate, 1% (w/v) SDS, 1% (v/v) Triton X-100, 10 mM sodium butyrate, 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail and were incubated on ice for 30 min. A four-times volume of dilution buffer containing 50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 0.1% (w/v) sodium deoxycholate, 1% (v/v) Triton X-100, 10 mM sodium butyrate, 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail was added to the nuclear extract to bring the final concentration of SDS to 0.2% (w/v). To prepare the nuclear extract, DNA was sonicated using Bioruptor (Diagenode) on high power under the following condition: 15 cycles of 30 s of on and 30 s of off, cooling samples on ice every 5 cycles. After the sonication step, the nuclear extract was collected by centrifugation at 20,000 g for 10 min at 4°C. DNA concentration was estimated by assessing the UV absorbance at 260 nm. Nuclear extracts containing an equal amount of DNA were prepared in tubes, and then an equal volume of dilution buffer was added to bring the final SDS concentration to 0.1% (w/v). For immunoprecipitation, 5 µg of anti-Histone H3K9me3 (2F3) and anti-H3K9ac (1qE5) antibodies (Kimura et al., 2008) and an equal amount of normal mouse IgG (Santa Cruz Biotechology, sc-2025) were added to the nuclear extract and incubated overnight at 4°C with gentle rotation. The next day, pre-equilibrated Dynabeads M-280 sheep anti-mouse IgG (Thermo Fisher Scientific, 11201D) was added to the reaction mixture and further incubated for 4 h at 4°C with gentle rotation. Antibody-bound proteins were collected with a magnet and washed for 10 min each at 4°C with gentle rotation in wash buffer as described below. The bound proteins were washed with low salt wash buffer containing 50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 0.1% (w/v) sodium deoxycholate, 1% (v/v) Triton X-100, 0.1% (w/v) SDS, 10 mM sodium butyrate, 0.1 mM PMSF, and cOmplete^TM EDTA-free protease inhibitor cocktail; high salt wash buffer containing 10 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM EDTA, 0.1% (w/v) sodium deoxycholate, 1% (v/v) Triton X-100, 0.1% (w/v) SDS, 10 mM sodium butyrate, 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail; LiCl wash buffer containing 10 mM Tris-HCl pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.1% (w/v) sodium deoxycholate, 1% (v/v) Triton X-100, 0.1% (w/v) SDS, 10 mM sodium butyrate, 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail; and TE wash buffer containing 10 mM Tris-HCl pH 8.0, 1 mM EDTA, 10 mM sodium butyrate, 0.1 mM PMSF and cOmplete^TM EDTA-free protease inhibitor cocktail. After a final wash with TE buffer, antibody-bound protein complexes were reverse cross-linked with elution buffer containing 10 mM Tris-Cl, pH 8.0, 300 mM NaCl, 5 mM EDTA, and 0.5% (w/v) SDS by heating at 65°C overnight. Reverse cross-linked DNA was further treated with RNase A and Proteinase K and extracted using phenol-chloroform-isoamyl alcohol and chloroform-isoamyl alcohol as described above. Finally, the extracted DNA was precipitated with ethanol, dissolved with 10 mM Tris-HCl pH 8.0 buffer, and subjected to real-time PCR analysis as follows. A serial dilution of input DNA and antibody-bound DNA were prepared from three independent ChIP experiments and analyzed two times using a StepOne Plus real-time PCR system (Thermo Fisher Scientific). PCR cycling conditions are described below. For Oct3/4 detection, 2 min at 50°C, 2 min at 95°C, and 50 cycles of 15 s at 95°C, 15 s at 53°C and 1 min at 72°C. For Nanog and Sox2 detection, 2 min at 50°C, 2 min at 95°C, and 40 cycles of 15 s at 95°C, 15 s at 58°C, and 1 min at 72°C. For the U3 region of IAP detection, 2 min at 50°C, 2 min at 95°C, and 40 cycles of 15 s at 95°C, 15 s at 60°C, and 1 min at 72°C. For the 5′ UTR region of IAP detection, 2 min at 50°C, 2 min at 95°C, and 40 cycles of 15 s at 95°C, 15 s at 62°C, and 1 min at 72°C. For Line L1 ORF2 detection, 2 min at 50°C, 2 min at 95°C, and 40 cycles of 15 s at 95°C, 15 s at 58°C, and 1 min at 72°C. For L1MdF detection, 2 min at 50°C, 2 min at 95°C, and 40 cycles of 15 s at 95°C, 15 s at 60°C, and 1 min at 72°C. Primer sequences are shown in Table S1.

ESCs were dissociated to obtain a single-cell suspension and subsequently fixed using ice-cold 70% (v/v) ethanol. The fixed cells were stored at −30°C until use. Following fixation, the cells were washed with PBS containing 2% (v/v) FBS and 4×10⁵ cells were re-suspended in PBS. Subsequently, the cells were treated with 200 μg/ml RNase at 37°C for 1 h and then stained with 500 μg/ml of propidium iodide for 30 min at 4°C. The stained cells were filtered using Falcon^® 5 ml Round Bottom Polystyrene Test Tube with Cell Strainer Snap Cap (352235) and then analyzed using a FACS Canto II flow cytometry system (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). Cell cycle profiles were analyzed using the ModFit LT software (Verity Software House, Topsham, ME, USA).

Captured images of EBs and DAPI foci were binarized using the Fiji-Image J software and Photoshop CS5.1. The surface area of EBs and circularity of DAPI foci were further analyzed using the Fiji-Image J software. Circularity was calculated by 4π(area/perimeter²). This value varies between 0 and 1. A value of 1 indicates a perfect circle. The 3D reconstruction of the nuclei image was performed using the Imaris software (Bitplane, Oxford Instruments).

We thank Dr K. Ura for assisting sucrose gradient sedimentation assay, Drs K. Yamagata and J. Ueda for providing the H2B-mCherry plasmid, and the members of the Division for Medical Research Engineering at Nagoya University Graduate School of Medicine for their support with FRAP and flow cytometry analyses. We also extend our gratitude to the Division for Medical Research Engineering at Nagoya University Graduate School of Medicine for use of the AXR confocal microscope system (Nikon), FACS Canto II flow cytometry system (Becton, Dickinson and Company), MAX-XP ultracentrifuge (Beckman), ChemiDoc Touch MP Imaging System (Bio-Rad) and Imaris imaging analysis system (Bitplane, Oxford Instruments).

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260467.reviewer-comments.pdf