Epigenome and chromatin structure in human embryonic stem cells undergoing differentiation



Epigenetic histone (H3) modification patterns and the nuclear radial arrangement of select genetic elements were compared in human embryonic stem cells (hESCs) before and after differentiation. H3K9 acetylation, H3K9 trimethylation, and H3K79 monomethylation were reduced at the nuclear periphery of differentiated hESCs. Differentiation coincided with centromere redistribution, as evidenced by perinucleolar accumulation of the centromeric markers CENP-A and H3K9me3, central repositioning of centromeres 1, 5, 19, and rearrangement of other centromeres at the nuclear periphery. The radial positions of PML, RARα genes, and human chromosomes 10, 12, 15, 17, and 19 remained relatively stable as hESCs differentiated. However, the female inactive H3K27-trimethylated X chromosome occupied a more peripheral nuclear position in differentiated cells. Thus, pluripotent and differentiated hESCs have distinct nuclear patterns of heterochromatic structures (centromeres and inactive X chromosome) and epigenetic marks (H3K9me3, and H3K27me3), while relatively conserved gene density-related radial chromatin distributions are already largely established in undifferentiated hES cells. Developmental Dynamics 237:3690–3702, 2008. © 2008 Wiley-Liss, Inc.


Human embryonic stem cells (hESCs) are characterized by sustained self-renewal and the ability to differentiate into the specific cell types of all three germ layers (Thomson et al.,1998; Reubinoff et al.,2000). These qualities make hESCs promising therapeutic tools in regenerative medicine. Several studies on hESCs have focused on optimizing culture conditions (Richards et al.,2002,2003; Eisellerova et al.,2008) to minimize accumulation of genetic defects, such as structural and numerical chromosome aberrations, caused by the long-term maintenance of hESCs in culture (summarized by Allegrucci and Young,2007). Chromosomes thought to be associated with carcinogenesis are more prone to acquiring genetic abnormalities in hESCs (Allegrucci et al.,2004; Storchova and Pellman,2004). In several hESC lines, an increased occurrence of aneuploidy has been observed for chromosomes 12, 17, and X (Brimble et al.,2004; Cowan et al.,2004; Mitalipova et al.,2005). However, genetic instability is not a universal feature of all hESCs (Buzzard et al.,2004; Darnfors et al.,2005), and whether genome aberrations are caused simply by cultivation conditions is unclear. For instance, dissection of hESC colonies during cultivation probably leads to greater genetic stability than trypsinization (Brimble et al.,2004; Mitalipova et al.,2005).

High expression levels of certain genes, such as Oct3/4 and Nanog, are a common feature of both hESCs and mouse embryonic stem cells (mESCs; summarized by Allegrucci and Young,2007). These loci, which are probably responsible for ESC pluripotency, occupy distinct nuclear regions in pluripotent and differentiated hESCs (Wiblin et al.,2005, and summarized by Meshorer and Misteli,2006): The transcriptionally active Nanog gene is more centrally localized in hESCs than in B cells, while Oct3/4 is positioned outside its usual chromosome territory (Wiblin et al.,2005). Likewise, we have found that one or both Oct3/4 alleles are repositioned away from their related chromosome territory in pluripotent hESCs, while Oct3/4 is positioned at the chromosome periphery in retinoic acid (RA) -differentiated hESCs (Bártová et al.,2008). Of interest, the nuclear radial distribution of Oct3/4 remains constant in both pluripotent and RA-differentiated hESCs, similar to what has been observed for c-myc (Bártová et al.,2008). Some studies have demonstrated that c-myc is responsible for mESCs pluripotency (Cartwright et al.,2005). However, c-myc is probably not important for self-renewal of hESCs. Rather, it induces apoptosis and differentiation of these cells (Sumi et al.,2007). This idea is supported by our recent finding that, contrary to Oct3/4, which is responsible for hESCs pluripotency, c-myc localizes to the periphery of its related chromosome territory in both pluripotent and differentiated hESCs (Bártová et al.,2008).

ESC gene expression is regulated by specific histone modifications (summarized by Gan et al.,2007). The N-terminal tails of histones undergo various posttranslational modifications, such as acetylation, methylation, phosphorylation, ubiquitination, and ADP-ribosylation (Jenuwein and Allis,2001; Kouzarides,2007). These histone epigenetic marks are responsible for chromatin compaction, mediate contact between histones and DNA, and regulate gene expression through higher order interactions with neighboring nucleosomes (Narlikar et al.,2002). For example, H3K9 and H3K27 methylation are typical of condensed, transcriptionally silent heterochromatin, while acetylation of H3 and H4 and methylation of H3 at lysine (K4) are associated with transcriptionally active, open chromatin configurations (Jenuwein and Allis,2001). Bernstein et al. (2006) reported that the majority of key developmental genes in ESCs are exposed to bivalent chromatin domains containing large H3K4-methylated regions, which harbor small stretches of H3K27-methylated chromatin. These bivalent domains are thought to silence developmentally important genes in ESCs, while keeping them poised for transcriptional activation upon differentiation.

In this study, we analyzed differentiation-associated changes in the nuclear radial arrangement of select genetic elements and the nuclear patterns of epigenetic histone modification. Our data reveal that differentiation of hESCs is associated with distinct perinucleolar H3K9 trimethylation foci, along with centromere rearrangement and peripheral repositioning of the inactive X chromosome. Moreover, epigenetic variation between two hES cell lines was observed not only for inactivation of X chromosome but also for select pluripotency and differentiation markers.


Changes in Cell Morphology and Gene Expression in Differentiating hESCs

The formation of pluripotent hESC (HUES-9) colonies (Fig. 1 Aa) was studied using light microscopy. We observed that RA treatment induced changes in the morphology of the cells positioned at the center of the colony (Fig. 1 Ab). Specifically, these cells were well dispersed compared with the cells at the periphery, which appeared aggregated (Fig. 1 Ab, magnification in Fig. 1 Ac). Oct3/4 and Nanog expression levels were analyzed to confirm hESC differentiation. Reverse transcriptase-polymerase chain reaction (RT-PCR) revealed that, following RA treatment, expression of Oct3/4 and Nanog were reduced to 6 % and 2 % of starting levels, respectively (Fig. 1B). Accordingly, after RA-treatment, the protein product of Oct3/4 gene was barely detectable by Western blot (Fig. 1C, asterisk). We also found that RA reduced levels of c-MYC protein, while p53 protein levels remained relatively constant. RA treatment did not significantly induce apoptosis, as seen by the absence of lamin B and PARP fragmentation (Fig. 1C).

Figure 1.

Differentiation-induced changes in morphology, gene expression, and histone modification in human embryonic stem cells (hESCs). A: Micrographs of pluripotent hESCs (HUES-9 cell line) growing on an mouse embryonic fibroblast (MEF) feeder layer (a) and their RA-differentiated counterparts (b and c). In the magnified view in c, the white arrow indicates separated cells in the interior of the colony, and the black arrow indicates more aggregated cells at the periphery of the colony. B: Real-Time Polymerase Chain Reaction (RT-PCR) analysis of Oct3/4 and Nanog. Gene expression levels were normalized to the HPRT1 housekeeping gene. C: Western blot analysis of OCT3/4, c-MYC, p53, Lamin B, and PARP in control (C) and retinoic acid-treated (RA) cells. Levels of all proteins were normalized to total protein levels. D: Western blot analysis of H3K4me3, H3K9me3, H3K9Ac, H3K27me3, H3K79me1, and the centromeric histone H3 variant, CENP-A. E: Quantification of hESC Western blot data in (D). Protein levels were normalized to levels of corresponding proteins in control and RA-treated MEFs to account for the contribution of MEFs in hESC-MEF mixed cultures. Data are shown as a ratio of band density between hESCs/MEFs, which was determined by Image QuaNT software (Molecular Dynamics). Scale bar = 2 μm in a; 4 μm in c.

Analysis of the total levels of H3K4 trimethylation (H3K4me3), H3K9me3, H3K9 acetylation (H3K9Ac), H3K27me3, H3K79 monomethylation (H3K79me1), and CENP-A (centromeric variant of H3) showed that levels of H3K9Ac and CENP-A were slightly reduced during RA-stimulated hESC differentiation (Fig. 1D,E). Conversely, H3K79me1 level was increased, while H3K4me3, H3K9me3, H3K27me3 levels were stable in RA-treated hESCs (summarized in Fig. 1E). To exclude the contribution of mouse embryonic fibroblasts (MEFs), we normalized Western blot data from MEF and hESC mixed cultures to relevant protein levels in control and RA-treated MEFs (Fig. 1D,E).

Distinct Patterns of Histone Modification in Pluripotent and RA-Differentiated hESCs

Nuclear patterns of H3K4me2, H3K4me3, H3K9me1, H3K9me2, H3K9me3, H3K9Ac, H3K27me2, H3K27me3, and H3K79me1 were studied in pluripotent and RA-differentiated hESCs. We observed significant changes in the patterns of H3K9me3 (Figs. 2A, 3A), H3K9Ac (Fig. 2B), and H3K79me1 (Fig. 2B) at the nuclear periphery of RA-stimulated HUES-9 cells. In contrast, the patterns of the other types of histone modification remained stable during the differentiation process (Fig. 2A,B). The most pronounced differentiation-related changes involved H3K9me3, which went from a dispersed pattern in pluripotent HUES-9 cells to a focal distribution in differentiated hESCs (Fig. 3A). During differentiation, H3K9me3 regions, which are associated with pericentromeric heterochromatin, were primarily observed in close proximity to the nucleoli (Fig. 3B) as a consequence of centromere (CENP-A) rearrangement (Fig. 3C). Prolonged cultivation (20 passages) also increased the percentage of cells with a focal distribution of H3K9me3 (Fig. 3A, color panel and bar chart), most likely due to an increased percentage of cells undergoing spontaneous differentiation. For this reason, we performed experiments only on pluripotent hESCs that had been passaged 10 times or less.

Figure 2.

Differentiation-associated changes in nuclear distributions of epigenetic marks in human embryonic stem cells (hESCs). H3K4me2, H3K4me3, H3K9me1, H3K9me2, H3K9me3, H3K9Ac, H3K27me2, H3K27me3, and H3K79me1 nuclear distributions (shown in green) were determined by Nipkow disc-based confocal microscopy and analyzed by Andor iQ software. Nuclear volumes are shown in blue. Data represent the average distribution in 30 nuclei from pluripotent or differentiated HUES-9 cells. Significant changes in nuclear distribution are indicated (*).

Figure 3.

Effect of prolonged culture and RA-induced differentiation on H3K9me3 distribution in human embryonic stem cells (hESCs). A: Nuclear location of H3K9me3 (green) in mouse embryonic fibroblasts (MEFs), pluripotent hESCs, and RA-differentiated hESCs (HUES-9). The hESCs were cultured for 2 or 20 passages as indicated. Quantification of focal accumulation of H3K9me3 at perinucleolar regions is shown to the lower right. B: Effect of RA-induced differentiation on the association of H3K9me3 (green) with nucleoli (red). Nucleoli were visualized by TO-PRO-3 staining and with antibody against fibrillarin, a nucleolar component (red). C: Association between H3K9me3 (green) and CENP-A (red) in pluripotent and RA-differentiated hESCs, as determined by confocal microscopy. A high magnification view (×15) is shown for pluripotent hESCs (region 1) and RA-differentiated hESCs (region 2). Scale bars = 1 μm.

Analysis of H3K9me3 interphase patterns in the other hESC line, HUES-1, revealed that pluripotent HUES-1 cells had a distinct H3K9me3 nuclear distribution compared with HUES-9 cells. In HUES-1 cells, H3K9me3 was distributed more focally before RA-induced differentiation, which was further characterized by a more pronounced focal arrangement of H3K9me3 (Fig. 4A). Similarly, X chromosome inactivation, which was accompanied by dense H3K27 trimethylation at the inactive X chromosome, was observed in approximately 50% of pluripotent HUES-1 cells (Fig. 4B, right panel). However, a densely stained inactive X chromosome was not observed in pluripotent HUES-9 cells (see Fig. 6A). These observations correspond well with data published by the International Stem Cell Initiative (2007) showing diverse genotypes and levels of X inactivation in distinct female hESC lines. Given the different nuclear pattern of H3K9me3 and H3K27me3 (i.e., hESC pluripotency markers) between HUES-1 and HUES-9 cells, we expected OCT3/4 protein levels in these cell lines to also differ. Indeed, we found that pluripotent HUES-9 cells had higher OCT3/4 levels than pluripotent HUES-1 cells (Fig. 4C). In both cell lines, RA-induced differentiation considerably decreased OCT3/4 levels. Further analysis of differentiation markers revealed that both RA-stimulated HUES-1 and HUES-9 cell lines undergo an endoderm-like differentiation pathway, as evidenced by an increase in the endoderm-specific marker, cytokeratin Endo-A (Fig. 4C).

Figure 4.

Differentiation-associated changes in nuclear distribution of H3K9me3, H3K27me3 in HUES-1 cells and selected protein levels in pluripotent human embryonic stem cells (HUES-1) and HUES-9 cell lines. A,B: Immunofluorescence labeling of H3K9me3 (A) and H3K27me3 (B) in pluripotent and retinoic acid (RA)-differentiated HUES-1 cells. C: Western blot analysis of OCT3/4 and the cytokeratin, Endo-A, in pluripotent and RA-differentiated HUES-1 and HUES-9 cell lines. Scale bars = 1 μm.

Distinct Histone Patterns in MEFs and Pluripotent hESCs

The chromocenters of MEFs were devoid of H3K4me2, H3K9me1, H3K9me2, H3K9Ac, H3K27me2, H3K27me3 (data not shown), and H3K79me1 (example in Fig. 5A). On the other hand, MEF chromocenters contained higher levels of DNA methylation (Fig. 5A), H3K9me3 (Fig. 3A), and H3K4me3 (Fig. 5A) than those of pluripotent hESCs (HUES-9). RA treatment induced a more focal pattern of DNA methylation and H3K79me1 in both MEFs and hESCs (Fig. 5A–D). DNA methylated foci in RA-treated MEFs were more distinct and larger (Fig. 5A,C). However, the number of foci in RA-treated MEFs was not reduced compared with the number of foci in nontreated fibroblasts (Fig. 5B). In hESCs, RA induced a more focal pattern of DNA methylation in approximately 80% of cells (Fig. 5B), with three to four distinct foci being present in RA-treated hESCs (Fig. 5A, arrow). Approximately 3% of nondifferentiated hESCs exhibited a focal distribution of DNA methylation, likely due to spontaneous differentiation. In both MEFs and hESCs, RA induced the formation of more pronounced H3K79me1 foci (Fig. 5A,D).

Figure 5.

DNA methylation, H3K4me3, and H3K79me1 in mouse embryonic fibroblasts (MEFs) and human embryonic stem cells (hESCs). A: Nuclear patterns of DNA methylation (green), H3K4me3 (green), and H3K79me1 (green) in MEFs, pluripotent hESCs, retinoic acid (RA) -treated MEFs, and RA-differentiated hESCs (blue) (HUES-9). DNA methylation and H3K4me3 were abundant at chromocenters of MEFs, contrary to the dispersed patterns of H3K4me3 in hESCs. RA-induced the appearance of distinct foci of H3K4me3 in MEFs, but did not change the pattern of H3K4me3 in hESCs. More distinct foci were also observed for H3K79me1 in MEFs and hESCs after RA treatment, and RA caused the absence of this histone mark at the nuclear periphery in hESCs. B: Quantification of the number of DNA-methylated foci per cell (black) and the number of cells with these foci (gray) in control and RA-treated MEFs, pluripotent and differentiated HUES-9 cells. C: Quantification of the average area of individual DNA-methylated foci. D: Quantification of the number of H3K79 monomethylated foci per cell (black) and the number of cells with these foci (gray). Scale bar = 1 μm.

In MEFs, H3K27me3 accumulated in one of the chromocenters related to the inactive X chromosome (Fig. 6A). This pattern of H3K27me3 accumulation was not observed in female pluripotent HUES-9 cells, even after RA-induced differentiation (Fig. 6A). This result was confirmed using a combination of immunofluorescence and fluorescence in situ hybridization (FISH) techniques (Fig. 6B), which showed that the highly condensed Barr body in differentiated HUES-9 cells did not contain high levels of H3K27me3. This is in contrast to pluripotent and differentiated HUES-1 cells, in which the inactive X chromosome was densely H3K27 trimethylated (Fig. 4B). H3K27me3 was also observed within the nuclear interior in pluripotent, but not in differentiated HUES-1 cells (Fig. 4B).

Figure 6.

The X chromosomes are not enriched in H3K27me3 in pluripotent or differentiated female pluripotent human embryonic stem cells (HUES-9) cells. A: Pluripotent and RA-differentiated HUES-9 cells as well as control and retinoic acid (RA)-treated mouse embryonic fibroblasts (MEFs; female and male) were labeled for H3K27me3 and counterstained with TO-PRO-3 (blue) to visualize chromocenters associated with inactive X chromosome (Xi-green) in MEFs. B: Visualization of X chromosomes through DNA-fluorescent in situ hybridization (FISH) of H3K27me3-labeled nuclei (HUES-9 cells). C: Immuno-FISH analysis showed RNAP II density of chromosomes X in pluripotent and RA-differentiated HUES-9 cells. Scale bars = 2 μm.

Nuclear Radial Distributions of Genetic Elements in Pluripotent and Differentiated hESCs

The nuclear radial distribution of PML, RARα genes, HSA 10, HSA 12, HSA 15, HSA 17, HSA 19, chromosome X, as well as of the centromeres of chromosomes 10, 12, 15, 17, and 1/5/19 were compared in pluripotent and RA-treated hESCs (Table 1). In many cases, the radial distributions (loci position within specific nuclear layers) were relatively constant, as observed with c-myc and Oct3/4 (Bártová et al.,2008). This suggests that the basic gene density-related radial chromatin distribution is established early during human development. An example is the nuclear radial distribution of the gene-rich chromosomes 17 and 19, which are both centrally positioned in highly differentiated human granulocytes (Bártová et al.,2001). The specific nuclear positions of these chromosomes are already established in pluripotent hESCs (Table 1). The nuclear radial arrangement of RARα was expected to change during hESCs differentiation, owing to its increase in expression (in two independent experiments, RARα expression was 9- to 15-fold higher after RA treatment). However, the nuclear and territorial positions of RARα and its frequent translocation partner PML as well as HSA 15 and HSA 17 were identical in pluripotent and differentiated hESCs (Table 1). Similarly, no significant nuclear radial redistribution was observed for HSA 10 and HSA 12 in control and differentiated hESCs (Table 1).

Table 1. Average Center-to-Locus Distances in Pluripotent and Differentiated hESCsa
Genetic elementPluripotent hESCsRA-differentiated hESCs
CL/R (%)b 
  • a

    For all genetic elements, distances were derived from the HUES-9 cell line, except as otherwise noted. HUES-9, pluripotent human embryonic stem cell; hESC, human embryonic stem cell.

  • b

    Center-to-locus distances (CL) were normalized to the local nuclear radius (R) of hESCs and are expressed as the mean ± standard error of three independent experiments. In some cases, areas of chromosome territories were calculated from maximal projections of 70 confocal sections.

  • c

    P ≤ 0.05. Statistical analyses for individual experiments were performed using Mann-Whitney U-test, and median values were calculated. Comparison of means was performed using the student's t-test.

Centromere 1056.3 ± 0.460.3 ± 0.6c
HSA 1069.3 ± 1.366.6 ± 1.0
Centromere 1256.2 ± 1.955.9 ± 2.1
HSA 1265.2 ± 0.965.2 ± 1.1
PML (15q24.1)60.7 ± 1.859.4 ± 1.4
HSA1557.1 ± 1.9 (Area: 6.6 ± 0.2 μm2)59.5 ± 1.5 (Area: 5.8 ± 0.3 μm2)
Centromere 1544.2 ± 1.750.7 ± 1.6c
RARα(17q21.2)55.9 ± 1.755.8 ± 1.4
HSA1754.0 ± 0.8 (Area: 5.6 ± 0.5 μm2)53.7 ± 0.8 (Area: 5.1 ± 0.3 μm2)
Centromere 1747.1 ± 1.257.8 ± 1.6c
Centromere 17 (HUES-1)48.8 ± 2.051.3 ± 1.3
HSA1947.3 ± 0.9 (Area: 6.3 ± 0.2 μm2)48.8 ± 0.9 (Area: 4.2 ± 0.1 μm2c)
Centromere 1/5/1957.0 ± 1.548.2 ± 1.9c
 X Chromosomes 
Both chromosomes X62.4 ± 1.2 (Area: 6.0 ± 0.5 μm2)70.9 ± 1.1c (Area: 2.3 ± 0.5 μm2c)
Peripherally located X74.7 ± 0.7 (Area: 4.5 ± 0.4 μm2)82.3 ± 0.8c (Area: 4.0 ± 0.4 μm2)
Centrally located X54.5 ± 1.1 (Area: 8.1 ± 0.6 μm2)58.4 ± 1.0 (Area: 5.7 ± 0.5 μm2c)
H3K27-trimethylated Chromosome X (HUES-1)-75.6 ± 1.7 (Area: 5.7 ± 0.2 μm2)

In contrast to the stable radial positions of the transcriptionally active and inactive genes described above, changes in nuclear radial distribution occurred in both constitutive and facultative heterochromatin, such as centromeres and the inactive female chromosome X (Table 1). Differentiation-related centromere redistribution was also apparent through visualization of CENP-A regions (Fig. 3C). This approach revealed that, after differentiation, centromeres clustered around nucleoli (Fig. 3B,C) or were repositioned at the very periphery of interphase nuclei (Fig. 3C; Table 1). This was true for centromeres 10, 15, and 17 in the HUES-9 cell line. On the other hand, during HUES-9 differentiation, centromeres 1, 5, and 19 were repositioned more centrally, closer to the nucleoli. In HUES-1 cells, centromere 17 was relocated closer to the nuclear periphery, although this change was not statistically significant (Table 1). Both observations support the conclusion that, after hESC differentiation, centromeres either form clusters around nucleoli or relocate to the outermost regions of the nucleus.

Differentiation-associated changes in heterochromatic structures were also observed in the inactive X chromosome of hESCs. Analysis of chromosome X radial distribution in pluripotent and differentiated HUES-9 showed that both chromosomes X are repositioned closer to the nuclear envelope in differentiated hESCs (Table 1). However, more pronounced relocation was observed for the H3K27-trimethylated, inactive chromosome X (Fig. 4B) that was situated very peripherally in differentiated HUES-1 cells, compared with the radial position of both X chromosomes in pluripotent stages (Table 1). Despite low H3K27me3 level, in RA-differentiated HUES-9 cells, the X chromosome associated with nuclear periphery was less RNAP II-positive (Fig. 6C, arrows). However, in pluripotent cells both chromosomes X were stained by antibody against phosphorylated form of RNAP II (Fig. 6C).


The distinct nuclear architecture of human somatic cells and corresponding patterns of histone modifications have been described by several laboratories (Cremer et al.,2004; Gilchrist et al.,2004; Bártová et al.,2005; Zinner et al.,2006,2007; Skalníková et al.,2007). The current challenge is to extend this experimental approach to hESCs, which represent a unique model for studying cell renewal and differentiation potential (Meshorer and Misteli,2006; Gan et al.,2007). The suggestion that ESC-specific histone modification patterns contribute to the maintenance of ESCs pluripotency (Azuara et al.,2006; Bernstein et al.,2006) has sparked interest in the nuclear architecture of stem cells. Accordingly, all variants of heterochromatin protein 1 (HP1), which binds to H3K9-methylated chromatin (Rice and Allis,2001), have been shown to be responsible for hESC pluripotency (Bártová et al.,2008). In general, hESCs are thought to display a unique nuclear architecture at loci that are responsible for maintaining the pluripotency of these cells (Wiblin et al.,2005; Bártová et al.,2008).

In the experiments described here, several histone modification marks were highly abundant at the nuclear periphery of pluripotent hESCs, while RA-induced differentiation reduced the levels of H3K9Ac, H3K9me3, and H3K79me1 in this nuclear compartment. Changes in epigenetic patterns have also been observed at the nuclear periphery of human cells exposed to histone deacetylase (HDAC) inhibitors (Gilchrist et al.,2004; Bártová et al.,2005). These observations suggest that the nuclear periphery is a dynamic nuclear compartment. Moreover, nuclear epigenetic patterns appear to be more dynamic than related protein levels, which remain relatively stable during differentiation processes (Fig. 1D,E, and Bártová et al.,2007). On the other hand, HDAC inhibitors frequently induce alterations in both epigenetic patterns and levels of chromatin-related proteins, including the ones analyzed here (Taddei et al.,2001; Bártová et al.,2005).

The perinucleolar region is another dynamic nuclear compartment. This is evident by the observation that the nucleolar periphery of RA-differentiated HUES-9 cells undergoes H3K9me3 clustering (Fig. 3B), which correlates with the H3K9me3 pattern in MEFs (Fig. 3A, first line). H3K9me3 is a marker for pericentromeric heterochromatin in somatic cells (summarized by Lachner et al.,2003) as well as pluripotent and differentiated hESCs (magnified region in Fig. 3C). Therefore, the accumulation of H3K9me3 in close proximity to nucleoli during hESCs differentiation likely results from centromere reorganization. Martin et al. (2006) reported that the genomic structure of mESCs is also dramatically reprogrammed after nuclear transfer. This remodeling involves changes in the nucleolar compartment as well as disconnection between chromatin condensation and HP1β accumulation (Martin et al.,2006). Interestingly, the nuclei of mESCs and hESCs have different patterns of centromere and HP1β localization (compare Martin et al.,2006, with Fig. 3C and Bártová et al.,2008). This supports the conclusion that not all aspects of the structural and functional biology of mouse and human ESCs are similar (Constantinescu et al.,2006; Sumi et al.,2007).

Peripheral repositioning of centromeres is the most common structural change associated with differentiation in human and mouse cell types (Chaly and Munro,1996; Bártová et al.,2001; Harničarová et al.,2006). Wiblin et al. (2005) reported that the proportion of centromeres located close to the nuclear periphery in hESCs is smaller than that in highly differentiated B-cells. Here, analysis of individual centromeres showed that centromeres 10, 15, and 17 undergo differentiation-related peripheral repositioning, whereas centromeres 1, 5, and 19 are relocated more centrally (Table 1). Moreover, immunofluorescence studies with an anti–CENP-A antibody revealed that, during hESC differentiation, some centromeres aggregate at perinucleolar regions, while others are repositioned closer to the nuclear periphery (Fig. 3C). Taking all of this into consideration, we propose that the centromeres become specifically distributed either at the nuclear or nucleolar periphery during hESC differentiation.

The association of perinucleolar centromeres with their heterochromatin marks, such as H3K9me3 and HP1, seems to be stable in terminally differentiated hESCs (Fig. 3C) and even tumor cells treated with HDAC inhibitor (Bártová et al.,2005). Moreover, HDAC inhibitors have the potential to induce differentiation (Harničarová et al.,2006; Bártová et al.,2007), suggesting that the level of histone acetylation at pericentromeric heterochromatin influences centromere position (Taddei et al.,2001). For example, in mESCs, increased hypoacetylation at satellite repeats occurs upon induction of the differentiation process (Keohane et al.,1996). Also intriguing is the dissociation between chromatin condensation and HP1β accumulation observed during the early phases of chromatin remodeling caused by nuclear transfer (Martin et al.,2006). Unlike metaphase chromosomes (Minc et al.,1999), HP1β-abundant chromocenters are restored in somatic cells and in mESCs at G1 phase (Hayakawa et al.,2003; Martin et al.,2006). This is not observed during the G1 phase after nuclear transfer into the oocyte. The absence of HP1β at chromocenters might be due to a reduced level of HP1β in the embryo, lower affinity of HP1β for pericentromeric heterochromatin (discussed by Martin et al.,2006), or distinct acetylation levels at pericentromeric regions.

As documented here, differentiation of hESCs is accompanied by specific nuclear changes involving centromeric heterochromatin. Moreover, in our hESC differentiation model, facultative heterochromatin of the inactive X chromosome became highly condensed and was repositioned closer to the nuclear periphery (Table 1). This seems to be typical of terminal differentiation in somatic cells (Barr and Carr,1962; Heard et al.,1997,2001; Bártová et al.,2001) and of maturation in hESCs (Table 1). Furthermore, we expected that high levels of chromosome decondensation in hESCs would also influence their sensitivity to karyotype instability, which has been observed for HSA 12, HSA 17, and chromosome X of hESCs (Brimble et al.,2004; Cowan et al.,2004; Mitalipova et al.,2005; Baker et al.,2007). Bearing in mind the transcription maps of the chromosomes being studied (Caron et al.,2001), we did not detect any significant differences in selected chromosome condensation and radial arrangement (Table 1) between pluripotent and differentiated hESCs. A higher degree of decondensation of chromosome 17 and X was expected relative to HSA15 and HSA19 in pluripotent hESCs. However, this was not the case, and areas of interphase chromosomes were relatively similar irrespective of chromosome size. Therefore, the degree of chromosome condensation may not be the main factor influencing karyotype instability of chromosome X and HSA17 in hESCs. Moreover, aneuploidy does not influence the nuclear location of chromosomes in tumor cells (Koutná et al.,2000) or in chromosomally abnormal human blastomeres (Finch et al.,2008). Thus, aneuploidy in hESCs would not be expected to induce alterations in the nuclear radial distributions.

The aforementioned chromosome abnormalities are not a general aspect of hESC cultures (Buzzard et al.,2004; Mitalipova et al.,2005), making it unclear if they are induced by cultivation conditions, isolation steps, and/or differing susceptibility of hESCs to genetic instability. Variation among hESC cultures occurs not only at the level of karyotypes, but also at the level of pluripotency markers. For example, expression of Oct3/4, Nanog, and GDF3 varied among seven different hESC lines (Skottman et al.,2005; summarized by Allegrucci and Young,2007), corresponding well with the distinct differences in OCT3/4 and Endo-A levels observed between HUES-1 and HUES-9 cells (Fig. 4C). On the other hand, in H1, H7, and H9 hESCs, Oct3/4, Nanog, and STAT3 expression was relatively stable, while X chromosome inactivation was hESC line-specific (Hoffman et al.,2005). Hoffman and colleagues showed that undifferentiated hESCs exhibit different patterns of X chromosome inactivation. That is, pluripotent, Oct3/4-expressing H9 cells harbor an inactive X chromosome, while pluripotent H7 cells contain two active X chromosomes (Hoffman et al.,2005). Similarly, Silva et al. (2008) showed that different hESC cultures variably express XIST, which generally makes noncoding RNAs required for X inactivation. Silva et al. (2008) categorized HUES-1 and HUES-9 cells into the same cell class based on X chromosome inactivation. This differs from our findings that one X chromosome was H3K27me3-rich in HUES-1 cells, while H3K27me3-poor in HUES-9 culture (Figs. 4, 6). Likewise, different levels of X inactivation, but for H7 and H9 hESC lines, were published by Hoffman et al. (2005) and Silva et al. (2008). These distinctions might be due to different cultivation conditions, which probably alter the genome, epigenome, and proteome of pluripotent hESC lines.

Taken together, our data reveal that hESC differentiation is accompanied by specific changes in the nuclear patterns of histone modification and in particular, that heterochromatin is reorganized during hESCs maturation. Moreover, changes in the interphase profiles seem to be more important indicators of epigenetic plasticity than changes in modified histone protein levels, which were often stable during the differentiation processes (Bártová et al.,2007 and Fig. 1D). Similarly, the general radial organization of chromosome territories is already established in hESCs, as documented for HSA18 and 19 (Wiblin et al.,2005), for Oct3/4 and c-myc (Bártová et al.,2008), and as shown here, for PML and RARα as well as chromosomes 10, 12, 15, 17, and 19. Our observations also confirm that genome, epigenome, and proteome differences exist among hESC lines, with these being accompanied by differences in X chromosome inactivation. One explanation of these differences is the status of X inactivation at the blastocyst stage (i.e., the stage of hESC derivation; Enver et al.,2005). In addition, hESC instability may be caused by nonuniform cultivation conditions. Therefore, further investigation of the variability among different hESC lines, specifically how heterochromatin and its epigenetic marks are established in hESCs, will uncover new insights into the structural and functional aspects of hESC pluripotency and stability.


Cultivation and Differentiation of hESCs

Human embryonic stem cells HUES-9 (9p19) (46,XX, inv9) and HUES-1 (1p21) (46,XX) were a generous gift from the laboratory of Prof. Douglas Melton (HHMI/Harvard University). The cells were cultured on a feeder layer of mitotically inactive MEFs (by mitomycin C) and plated on gelatin (0.1%)-coated, six-well cultivation dishes. MEFs were isolated and prepared for hESC cultivation as previously described (Bártová et al.,2008). Human embryonic stem cells were cultivated in KO-DMEM medium (Invitrogen Gibco) supplemented with 500 U/ml penicillin, 5 mg/ml streptomycin (PAN, Germany), 200 mM Glutamax (Invitrogen Gibco), 10 mM nonessential amino acids (PAN), β-mercaptoethanol (Invitrogen Gibco), and KO-serum replacement (Invitrogen Gibco). The medium also contained a final concentration of 10 ng/ml bFGF (Invitrogen Gibco) and 12 ng/ml hLIF (Chemicon International). Culture medium was changed daily, and every 4 or 5 days, the hESCs were split at a 1-to-3 ratio for further cultivation and differentiation. Differentiation was induced with 4 μM all-trans RA (Sigma-Aldrich). After 4 days of cultivation at 37°C in a humidified atmosphere containing 5% CO2, the differentiated cells were fixed with 4% formaldehyde for further analysis. Medium for differentiation experiments did not contain bFGF and hLIF and was changed every day after differentiation. RA was added at an appropriate concentration daily. Cells at the center of RA-treated colonies were well dispersed, while those at the periphery appeared aggregated (Fig. 1 Ab, magnification in Fig. 1 Ac). Treatment of ESCs with RA and serum induces an endoderm-like differentiation pathway (Pacherník et al.,2002,2005; Bártová et al.,2007). Although centrally positioned pluripotent hESCs enter an endoderm differentiation pathway, their peripheral counterparts become ectoderm-like (Dvořák et al.,2005). Bearing in mind possibility that hESC colony can consist of distinct cell types, we restricted our analysis to cells at the center of the colony.

Western Blot Analysis

Western blots were performed as previously described (Bártová et al.,2005) using the following primary antibodies: anti-c-MYC (Santa Cruz, USA, #sc-764), anti-p53 (Novocastra, UK, #NCL-p53-CM1), anti-H3K4me3 (#ab8580-2b, Abcam), anti-H3K9me3 (Upstate, USA, #06-942), anti-acetyl H3K9 (#06-942, Upstate), anti-H3K27me3 (Upstate, #07-212), and anti-H3K79me1 (#ab2886-25, Abcam). Levels of OCT3/4, which is responsible for hESC pluripotency, were determined using a mouse monoclonal antibody against OCT3/4 (C10) (#sc-5279, Santa Cruz). Monoclonal antibody against the cytokeratin, Endo-A (TROMA-I, Developmental Studies Hybridoma Bank, University of Iowa), was used to study the hESC differentiation pathway.

Real-Time PCR

Total RNA was isolated using Trizol reagent (Invitrogen) and purified with RNeasy mini spin columns (Qiagen) according to the manufacturers' protocols. Contaminated DNA was simultaneously digested using on-column DNase treatment (Qiagen). The iScript cDNA synthesis kit (Bio-Rad) was used to synthesize cDNA from 1 μg of total RNA. Primers used for real-time PCR were designed using the Primer3Plus program (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi). Each primer pair spanned an exon–exon boundary, and the size of the final PCR product was within 90–110 bp in length. Primer sequences were as follows: Oct3/4, 5′-GAGGCAACCTGGAGAATTTG-3′ and 5′-CGGTTACAGAACCACACTCG-3′; Nanog, 5′-CCAACATCCTGAACCTCAGC-3′and 5′-TGCTATTCTTCGGCCAGTTG-3′; RARα, 5′-GAAGATTACTGACCTGCGA AGC-3′ and 5′-CCAACATTTCCTGGATGAGAG-3; HPRT1, 5′-ACACTGGCAAAACAATGCAG-3′ and 5′-ACTTCGTGGGGTCCTTTTC-3′. RT-PCR amplification mixtures (20 μl) contained 25 ng of template cDNA, 2x SYBR Green I Master Mix (10 μl, Roche), and 200 nM forward and reverse primer. Reactions were run on a Rotor-Gene 6000 (Corbett). The cycling conditions consisted of 10 min of polymerase activation at 95°C and 40 cycles of 95°C for 15 sec, 53°C for 15 sec (for each primer), and 72°C for 20 sec. Each assay included a negative control and standard curves of four serial dilutions of hESC cDNA (in triplicate, ranging from 50 ng to 0.4 ng). Analysis was carried out using the detection software supplied with Rotor-Gene 6000. The quantity of each experimental sample was extrapolated from a standard curve and expressed relative to HPRT1 levels (housekeeping gene). Samples containing RNA, but not reverse transcriptase, served as an additional internal control. As a negative control, these samples were also subjected to PCR in the absence of cDNA.

Immunostaining of Interphase Nuclei

Interphase nuclei were fixed in 4% formaldehyde, successively permeabilized in 0.1% Triton X-100 for 12 min and 0.1% saponin (Sigma, Germany), washed twice in phosphate buffered saline (PBS) for 15 min, and incubated for 1 hr at room temperature in 1% bovine serum albumin (BSA) in PBS. The slides were then washed for 15 min in PBS, and preparations were incubated with one of the following antibodies: anti-H3K4me2 (#07-030, Upstate), anti-H3K4me3 (#ab8580-2b, Abcam), anti-H3K9me1 (#ab9045-25, Abcam), anti-H3K9me2 (#07-212, Upstate), anti-H3K9me3 (#07-442, Upstate), anti-H3K9Ac (#06-942, Upstate), anti-H3K27 me2 (#07-322, Upstate), anti-H3K27me3 (#07-449, Upstate), anti-H3K79me1 (#ab2886-25, Abcam), and mouse monoclonal anti-CENP-A (#ab13939, Abcam). Nucleoli were detected using mouse monoclonal antibody to fibrillarin (#ab12367, Abcam) and RNA polymerase II (RNAP II) -positive regions were recognized with the aid of mouse monoclonal antibody against phosporylated form of RNAP II (#ab24759, Abcam; Harničarová et al.,2006). After an overnight incubation of preparations with primary antibody at 4°C, appropriate secondary antibody was applied. The following secondary antibodies were used: anti-rabbit immunoglobulin G-fluorescein isothiocyanate (IgG-FITC; #F9887, Sigma, Germany), donkey anti-goat IgG-FITC (#sc-2024, Santa Cruz), and goat anti-mouse IgG Alexa Fluor-594 (#A11005, Molecular Probes). Mouse monoclonal [33D3] antibody to 5-methyl cytidine (#ab10805, Abcam) was used to determine the level of methylated DNA at CpG islands. Immunocytochemistry using this antibody was performed according to Beaujean et al. (2004). Briefly, cells were fixed in 4% paraformaldehyde at room temperature for 20 min. The preparations were then washed with PBS for 10 min and permeabilized in 0.5% Triton X-100 in PBS for 30 min at room temperature. DNA was denatured with 4 N HCl for 1 hr at 37°C. The cells were then washed three times in PBS-Tween 20 (0.05%) for 10 min. Nonspecific sites were saturated by incubating cells with 2% BSA dissolved in PBS for 1 hr at room temperature. Cells were incubated with 5-methyl cytidine antibody (1:200 in 2% BSA-PBS) for 1 hr at 37°C and subjected to four 15-min washes in PBS-Tween 20 (0.05%). Goat anti-mouse IgG-FITC (#F0257, Sigma) was then applied, and cells were washed three times in PBS-Tween 20 (0.05%) at room temperature for 10 min. Propidium iodide was used as a counterstain.

FISH and Image Acquisition

Whole human chromosome probes from the Biotin-Labeled Human Paint Box (#1088-B) were purchased from Cambio (UK). The following DNA probes were used: HSA10 (#1088-10B), HSA12 (#1088-12B), HSA15 (#1088-15B), HSA17 (#1153-17Cy3-02), and HSA19 (1088-19B). The DNA probe for PML and RARα genes (#LPH 004) was purchased from Cytocell (UK), and alpha-satellite probes for centromere 10 (#CP5020-B.5) and centromere 12 (#P5031-DG.5) were purchased from Oncor (USA). Probes (Poseidon FISH DNA probes) for centromere 15 (KBI-20015), centromere 17 (KBI-20017), and centromeres 1/5/19 (KBI-20026) were obtained from Kreatech Diagnostics (Amsterdam, the Netherlands). The cells used in these experiments were fixed in 4% formaldehyde in PBS for 10 min, and the three-dimensional FISH technique was performed as described elsewhere (Harničarová et al.,2006). MEFs and hESCs were distinguished by using the counterstain, TO-PRO (R)-3 iodide (0.04 μg/ml, Molecular Probes), which labeled whole interphase nuclei and centromeric heterochromatin arranged into chromocenters (Alcobia et al.,2003) in MEFs.

Images of fluorescent labeled interphase nuclei were acquired using a Nipkow discs-based confocal microscope, as already described (Bártová et al.,2005; Harničarová et al.,2006). Nuclear radial distributions of genes and chromosomes were calculated based on the distance between the given loci (fluorescence gravity center of chromosome territories) and the fluorescence gravity center of interphase nuclei. The distances were normalized to the local nuclear radius. In addition, territorial distributions of RARα were determined by measuring the distances between the loci and the center of gravity of the given chromosome territory. Data from the FISH 2.0 software analysis were exported to Sigma Plot 8.0 software (Jandel Scientific, California) to perform the final mathematical evaluation. The Mann Whitney U-test was used to analyze nuclear radial distances. Asterisks (*) indicate significant differences of P ≤ 0.05.


We thank the laboratory of Prof. Douglas Melton (HHMI/Harvard University) for providing the hES cell lines. We also thank Dr. Pavla Gajdušková for help with the real-time PCR experiments.