Author contributions: A.G.: collection and assembly of data, manuscript writing; S.P.A.: collection and assembly of data, manuscript writing; M.L.: conception and design, manuscript writing; L.A.: conception and design, manuscript writing and financial support.
First published online in STEM CELLSExpress March 12, 2009
The molecular mechanisms underlying pluripotency and lineage specification from embryonic stem cells (ESCs) are still largely unclear. To address the role of chromatin structure in maintenance of pluripotency in human ESCs (hESCs) and establishment of lineage commitment, we analyzed a panel of histone modifications at promoter sequences of genes involved in maintenance of pluripotency, self-renewal, and in early stages of differentiation. To understand the changes occurring at lineage-specific gene regulatory sequences, we have established an efficient purification system that permits the examination of two distinct populations of lineage committed cells; fluorescence activated cell sorted CD133+ CD45−CD34− neural stem cells and β-III-tubulin+ putative neurons. Here we report the importance of other permissive marks supporting trimethylation of Lysine 4 H3 at the active stem cell promoters as well as poised bivalent and nonbivalent lineage-specific gene promoters in hESCs. Methylation of lysine 9 H3 was found to play a role in repression of pluripotency-associated and lineage-specific genes on differentiation. Moreover, presence of newly formed bivalent domains was observed at the neural progenitor stage. However, they differ significantly from the bivalent domains observed in hESCs, with a possible role of dimethylation of lysine 9 H3 in repressing the poised genes. STEM CELLS 2009;27:1298–1308
The pluripotency and developmental plasticity of embryonic stem cells (ESCs) may be largely controlled by their specific chromatin architecture . The presence of numerous covalent modifications on the N-terminal tails of histones is linked to gene activities. For example, acetylation of histones H3 and H4 (H3ac and H4ac) together with the several methylation states of lysine 4 on histone H3 (H3K4me2/3) present at the gene promoters are linked with transcriptional activation. Conversely, methylation of lysines 9 and 27 on histone H3 (H3K9me2/3 and H3K27me3) are associated with gene silencing [2, 3]. Recently, unique histone modification patterns, named bivalent domains, have been reported which could contribute to the maintenance of pluripotency in both mouse and human ESCs (mESC and hESC) [4, 5]. These domains exhibit both trimethylation of lysines 27 and 4 on histone H3 (H3K27me3 and H3K4me3, respectively) over the promoter sequences of many lineage-specific genes. Bivalent domains keep important lineage-specific genes repressed due to the dominant effect of H3K27me3 over H3K4me3, while preserving their potential to become activated on differentiation. During differentiation into a particular cell lineage, they resolve according to the transcriptional status of the genes required in that specific cell type . However, the bivalent histone modifications were not found at all promoter loci of multiple lineage-control genes in either mouse or human ESCs. A number of key developmental genes were marked only by the permissive H3K4me3 and acetylation marks [7, 8] or did not exhibit either of the marks associated with bivalency at their promoters . How these regions are kept silent in ESCs is still not clear. Given their important role in cell lineage determination, it is likely that these genes become permissive for activation at later developmental stages through as yet uncharacterized epigenetic mechanisms.
This study is an examination of locus-specific patterns of post-translational modifications of histones in hESCs and neural progenitors that differentiate from hESC under specific culture conditions. An efficient purification system was established, in which epigenetic changes were examined in hESC-derived CD133+CD45−CD34− neural stem cells (NSCs) and β-III-tubulin+ putative neurons. Correlation between the gene expression profiles of these cell types and histone modification patterns at the appropriate gene promoters was established. In this study, we show the presence of other permissive histone marks supporting H3K4me3 at the active pluripotency-associated gene promoters, as well as poised bivalent and nonbivalent lineage-specific gene promoters at the undifferentiated hESC stage. H3K9 methylation was found to play a role in repressing various gene promoters on differentiation. Our data support those obtained from similar studies using mESC with the notable exception of the establishment of a novel bivalent domain at the promoters of the myelin basic protein (MBP) and glial fibrillary acidic protein (GFAP) at the neural progenitor cell stage and its resolution into a repressive chromatin structure when these cells differentiate into neurons.
MATERIALS AND METHODS
Wisconsin H9 hESC line (WiCell Research Institute, Madison, WI, www.wicell.org) was cultured as discrete colonies on feeder layers of mitotically inactivated mouse epithelial fibroblasts (MEFs) as previously described . To begin differentiation into the neural lineage, H9 hESCs were plated on pregelatinized plates. Neural differentiation was induced in N2B27 serum-free medium (1 × DMEM/F12 with glutamine, 1 × N2, 1 × B27 [PAA]; 1% MEM-NNAA [Invitrogen, Paisley, UK, http://www.invitrogen.com]; 1% L-glutamine/penicillin-streptomycin [Sigma Aldrich, Dorset, U.K., www.sigmaaldrich.com]). After 4–5 days the medium was changed to Neuronal base medium (DMEM/F12/Neuronal base medium 1:1; 0.5 × N2; 0.5 × B27 [PAA Laboratories Ltd, Pasching, Austria, http://www.paa.at]; 1% MEM-NNAA [Invitrogen]; 1% L-glutamine/penicillin–streptomycin [Sigma]) with bFGF (20 ng/ml; Invitrogen) and cultured for another 20 days. To obtain postmitotic neurons, the cultures were transferred as monolayer fragments to poly-D-lysine/laminin (Sigma/BD Biosciences, Oxford, U.K., www.bdeurope.com) plates. The cells underwent terminal differentiation in Neuronal base medium without bFGF and were collected 12–14 days after plating. Immortalized BJ foreskin human fibroblasts were obtained from LGC Promochem (LGC Promochem, Teddington, U.K., www.lgcstandards.com) and cultured in medium containing 88% Dulbecco's modified Eagle's medium (DMEM) without sodium pyruvate (Invitrogen); 10% FBS (Sigma); 1% MEM-NNAA (Invitrogen); 1% L-glutamine/penicillin–streptomycin (Invitrogen).
Fluorescence Activated Cell Sorting
The H9 hESC-derived monolayers were dissociated into single cell suspensions in accutase solution (Chemicon Temecula, CA, http://www.chemicon.com) and used unfixed for subsequent antibody staining. The single cell suspension recovered from poly-D-lysine/laminin plates were fixed for 10 minutes in 1% formaldehyde (Sigma) and permeabilized in cold 90% methanol for 30 minutes. The cells were washed in the Lyse-Wash Assistant (LWA; BD Biosciences) and filtered using BD falcon filters (12/75 mm). The appropriate fluorochrome conjugated primary antibodies were added (see supporting information Table 1) and the tubes were incubated at room temperature for 30–60 minutes. Compensation for each fluorochrome was performed with use of multifluorochrome beads (BD Biosciences). Fluorescence activated cell sorting (FACS) settings were adapted to a low sheath pressure (20–25 psi), larger nozzle size (100 μm), and reduced sorting speed (1,000–3,000 events per second). Unstained cells were used as a negative control.
Chromatin immunoprecipitation (ChIP) was performed on H9 hESC-derived progeny according to published Q2ChiP method [10, 11] in which 10,000 cells were used for each immunoprecipitation reaction. For the list of the relevant ChIP grade antibodies see supporting information Table 2. An additional no-antibody control ChIP reaction was also included to allow for normalization for any background, if present. Q-PCR of the ChIP products was carried out in 10 μl reaction volumes in a 384-well plate system, using the Applied Biosystems 7900HT Real-time PCR machine. Genomic DNA standards (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com) of know concentration (range of 100–0.01 ng/μl) were included in the Q-PCR run to generate a standard curve from which the ChIP samples could be enumerated. A control Q-PCR with water only was used to control for PCR contamination. The DNA sequence of oligonucleotides used for ChIP analysis is shown in supporting information Table 3.
Total RNA was extracted from the cell populations with TRIzol (Invitrogen) according to manufacturer's protocol and isolated by precipitation with propan-2-ol. After removal of genomic DNA contamination with DNAseI (1 IU/10 μg RNA) (37°C/30 minutes), cDNA was prepared by reverse transcription with MMLV reverse transcriptase (Promega, Southampton, U.K., http://www.promega.com) according to manufacturer's instructions using random 15mers and then stored at −20°C. The DNA sequence of oligonucleotides used for the Q-PCR is given at the supporting information Table 4.
Cell cultures were fixed for 10 minutes at room temperature in 4% PFA, permeabilized in 0.2% Triton X-100 for 45 minutes, and blocked in 5% BSA for 60 minutes at room temperature. After washing in 5% FBS/PBS, the cells were incubated with primary antibodies for 60 minutes at room temperature, followed by incubation with FITC conjugated secondary antibody for another 60 minutes. Finally, the cells were incubated in Hoechst solution for 5 minutes before visualization. The pictures were taken with use of fluorescence microscope Carl Zeiss Axiovert 200M. For the list of antibodies used in the study see supporting information Table 1.
hESCs Differentiate into the Neural Lineage via Progression Through Different Stages Resembling Embryonic Neurogenesis
hESCs are a valuable model of early embryogenesis and lineage-specification which yields sufficient cell numbers for investigations using ChIP. In this study, differentiation of NSCs from H9 hESCs was induced in serum-free adherent monolayer culture on pregelatinized plates (see supporting information Fig. 1). During the first 5 days of differentiation, the H9 hESC colony fragments formed typical neuroepithelial structures which organized into neural tube-like rosettes. After 4–5 days, the medium was supplemented with bFGF to promote proliferation of NSCs for another 20 days (see supporting information Figs. 1 and 2). The NSCs were induced to further differentiation by withdrawal of bFGF and plating on poly-D-lysine/laminin plates. β-III-tubulin as well as postmitotic neuronal markers (neuronal nuclear antigen [NeuN], doublecortin [DCX], microtubule-associated protein-2 (MAP2), neurofilament 200 [NF200]) were detected mainly in the densest areas, originally arising from the monolayer fragments plated on the plates (see supporting information Fig. 3A). Moreover, hESCs gave rise to a high population of cells from glial lineages. Both oligodendrocyte progenitors (A2B5, NG2, and galactocerebroside [GalC] immunopositive) were present together with astrocytic cells (S100β immunopositive) (see supporting information Fig. 3B).
Two Stages of Lineage Specification Can Be Purified During hESC Differentiation into the Neural Lineage
Examination of the histone modification patterns at the promoters of individual lineage specific genes is not possible in a heterogeneous population of differentiated cells; therefore, FACS was used to enrich NSCs and terminally differentiated neurons. H9 hESC-derived monolayers, after 25 days of bFGF induced proliferation, were used as a source of NSCs (Fig. 1A). The CD133 (prominin-1/2) epitope was chosen as a specific NSC marker because expression of this has been reported for hESC-derived NSCs in culture , putative NSCs of the cerebellum , and embryonic neuroepithelial cells . As CD133 can be expressed by immature hematopoietic stem cells , and hESC-derived endothelial progenitors (M. Lako, personal communication) CD133 positive cells were sorted only from the CD34 and CD45 negative population (CD133+) (Fig. 1A) as both of these progenitor cell types express CD34. A distinct population of CD34+ cells was observed within H9 hESC-derived monolayers, but they did not coexpress CD133, whereas no CD45+ cells were detected. Skeletal muscle satellite cells may also express CD133 at some stages of their development  but some evidence suggests that these cells are also CD34 positive  and therefore would be eliminated from the CD133 positive and CD34/CD45 negative population. Our data from other work suggest that differentiation of hESC over similar timescales to those employed in our neuronal differentiation model did not produce significant numbers of PAX7 positive skeletal muscle satellite cells that were also CD133 positive (Fig. 1B). In view of these observations, we are confident that the FACS strategy employed in this work is able to generate highly enriched populations of NSCs. When plated on poly-D-lysine/laminin precoated plates, sorted CD133+ cells underwent further differentiation and gave rise to neurons expressing β-III-tubulin (Fig. 1C). Similarly, of the whole aggregates plated on poly-D-lysine/laminin, not all of the CD133+ progeny expressed β-III-tubulin, which supports the multipotency of CD133+ NSCs. H9 hESC-derived neurons were purified from differentiated populations grown on poly-D-lysine/laminin plates under differentiation conditions for 12–14 days. β-III-tubulin was chosen as a marker for neurons generated in this differentiation protocol due to the lack of specific cell surface markers for postmitotic neurons (Fig. 1D). Other markers have been used for FACS enrichment of neurons such as neural cell adhesion molecule (NCAM, CD56)  but it is difficult to use this system in heterogeneous populations of cells derived from differentiating hESC because CD56 is expressed on other early progenitor cells such as those of skeletal muscle (J. Parris and L. Armstrong, personal communication).
Expression Profiling of Undifferentiated hESC, CD133+ NSCs and Differentiated β-III-Tubulin+ Neurons
H9 hESC exhibited unambiguous expression of a set of important genes associated with pluripotency, including OCT4, NANOG, SOX2, GDF3, LIN28, and TERT (Fig. 2). The expression levels of these genes decreased gradually on hESC differentiation although low levels were still detected in CD133+ cells apart from SOX2, whose expression is highly abundant in neuroepithelial cells and neural progenitors [19, 20]. It has been reported that hESCs express many lineage-specific genes at very low levels  and from all examined lineage-specific genes representing all three embryonic germ layers examined, only GFAP was not detected in H9 hESCs (Fig. 2). However, the expression of such lineage-specific genes was significantly lower in hESCs compared with the differentiated cells in which they are functionally expressed (data not shown). All markers of neural stem/progenitor cells were upregulated in CD133+ cells and further downregulated in the β-III-tubulin+ population. The expression profile observed in CD133+ populations indicates a neuroepithelial status. Unlike adult central nervous system (CNS) NSCs, radial glial (RG) cells do not express GFAP [22, 23]. The expression of OLIG2 and PAX7 together with neurogenic RG markers, such as PAX6, NES (Nestin), and VIM (Vimentin) could indicate that these cells do not correlate to a specific regional identity and CD133+ cells could retain the potential to differentiate into a wide range of neural precursors. An increase in the expression levels of the markers of more mature neural lineages such as MBP and TUBB3 (β-III-Tubulin) in the CD133+ population could indicate that these cells are more permissive for the “leaky” expression of latter neural markers in NSCs compared with hESCs. Importantly, only TUBB3 expression was detected in β-III-tubulin+ populations. No GFAP expression was detected in any of the examined cell populations, indicating a lack of contamination by astrocytic cells from differentiated H9 hESC-derived aggregates. GFAP expression is regulated by DNA methylation present at its promoter in neuroepithelial cells during mouse development and is thus activated only in mature astrocytes . Except for baseline levels of KRT19 (Cytokeratin19), none of examined non-neural markers was detected in H9 hESC-derived progeny. The epithelial marker P63 was not present, which suggests that the expression of KRT19 was not indicative of an epithelial cell fate. VIM was also present in BJ fibroblasts. The baseline expression levels of NES, TUBB3, and KRT19 in BJ fibroblasts could reflect low levels of cellular stress response [25–27].
Changes in Histone Modification Patterns at the Promoters of Pluripotency-Associated Genes Reveals the Repression of Their Promoters During hESC Differentiation
ChiP gives information about the spatio-temporal occupancy of a DNA sequence by a native protein in its in vivo native chromatin environment . The presence of H3ac, H4ac, and H3K4me at gene promoters has been strongly associated with potential transcriptional activation. In particular, H3K4me3 is associated with active transcription . All examined permissive marks were found to be present at the promoters of the examined pluripotency-associated genes (Fig. 3). Low levels of H3K4me2 were detected at the pluripotency markers in BJ fibroblasts, which confirmed a wider abundance of this mark among silent sections of the genome; but importantly, no H3K4me3 was present. Generally, pluripotency-associated gene promoters did not display the H3K9ac mark in BJ fibroblasts, as this mark is associated with transcriptional competence. Because of the difficulty of obtaining high numbers of cells after differentiation and FACS sorting, the presence of only two permissive histone modifications was assessed during hESC differentiation (H3K4me2 and H3K4me3). A gradual decrease in expression levels was associated with the loss of high levels of permissive marks at promoters of the pluripotency-associated genes examined herein (Fig. 3). CD133+ cells did not display H3K4me3 at the pluripotency-associated gene promoters correlating with their much lower expression levels in these cells; however, significant levels of H3K4me2 were still detected at the pluripotency-associated gene promoters in CD133+ and β-III-tubulin+ cells. H3K9me and H3K27me at the gene promoters are mainly associated with transcriptional repression and these marks are generally associated with facultative and constitutive heterochromatin respectively . The promoters of pluripotency markers in H9 hESC did not display significant levels of H3K9me2, H3K9me3, or H3K27me3 (Fig. 3), whereas all three were present in BJ fibroblasts although the level of H3K27 trimethylation at the promoter of LIN28 is very low. The presence of H3K27me3 at pluripotency-associated gene promoters has been reported previously in murine T-cells  and mouse fibroblasts . H3K9me2 was identified first at pluripotency-associated gene promoters in CD133+ NSCs, and its repressive activity was supported additionally by H3K9me3 and H3K27me3 in the β-III-tubulin+ population. The yield and efficiency of chromatin immunoprecipitation varies between antibodies directed against different histone modifications so it is difficult to make direct comparisons between absolute levels of H3K9 dimethylation, H3K9 trimethylation, and H3K27 trimethylation although comparisons of relative levels can be made (Fig. 3B). Controls for the chromatin immunoprecipitations were also performed. Glyceraldehdye 3-phosphate dehydrogenase (GAPDH) is ubiquitously expressed in all human cells and therefore a fragment of its promoter (215 bp before the transcription start site to 20 bp after) is used as a positive control for activating histone modifications (see supporting information Table 3 and associated figure). This section of the GAPDH promoter does not have repressive histone modifications in any of the cell types analyzed.
These data support those of previous groups in which a lack of H3K27me3 has been reported at the Oct4 and Sox2 promoters in mouse neuronal precursor cells , indicating that proteins other than Polycomb complexes play a role in silencing pluripotency genes at the early stages of differentiation.
Bivalent Promoters of Lineage-Specific Genes Resolve at the Progenitor Stage and Are Further Repressed upon Latter Differentiation
The presence of bivalent domains at the promoters of lineage-specific genes has been reported in ESCs, which consist of the permissive H3K4me3 mark together with the repressive H3K27me3 mark present at the same genomic loci [4, 5]. Studies of such domains have shown that lineage-specific transcription factors and genes involved in neurogenesis are particularly abundant among this group [32, 33]. Our data confirm the presence of such bivalent domains in hESC and show a correlation between histone modification patterns and gene expression. All of the examined neural lineage-specific transcription factors within this group, with the exception of GLI3, displayed bivalent domains at their promoters in hESC (Fig. 4). Resolution of these bivalent domains is expected on hESC differentiation and all bivalent domains were indeed resolved in CD133+ NSCs, in which only H3K4me3 remained at the promoters of genes upregulated during neurogenesis (PAX6, PAX7, SOX1, SOX3, and OLIG2). Markers of other lineages, such as NODAL and GATA4, lost H3K4me3 and displayed increased levels of H3K27me3, which remained also at high levels in the β-III-tubulin+ neurons. During further differentiation, the H3K27me3 mark was redisplayed at the bivalent promoters of neural stem/progenitor markers in β-III-tubulin+ cells, together with decrease or even loss of H3K4me3. Surprisingly, despite a lack of expression of OLIG2 and SOX3 in β-III-tubulin+ cells significant levels of H3K4me3 marks were still detected. In other more highly differentiated cells, resolution of bivalency should have taken place, and this is supported by the apparent lack of the H3K4me3 modification and the continued presence of H3K27me3 at the promoters of these same genes in the BJ fibroblasts, which are used as an example of a mostly differentiated cell type in which neuronal specific genes are not expressed. Some bivalent domains seem to persist (such as at the promoters of PAX6 and OLIG2) in BJ fibroblasts which supports the observations of other groups [6, 34]. Our data, however, suggest that this may not be an exact equivalent of the bivalency observed in both mESCs and hESCs because the levels of bivalent histone modifications detected in BJ fibroblasts are different than those of hESCs, with higher levels of the repressive H3K27me3 mark and lower levels of H3K4me3. The reasons for the apparent retention of bivalent chromatin domains in BJ fibroblasts are not clear at present. Interestingly, H3K4me3 was not the only permissive mark observed at the bivalent promoters in H9 hESCs and significant levels of H3K4me2, H3K9ac, H3ac, and H4ac were also present (Fig. 4). This indicates that H3K4me3 may be supported by other permissive marks, despite insignificant levels of transcription in hESCs. H3K4me2 and histone H3 acetylation levels were significantly decreased or totally lost in BJ fibroblasts at the promoters of most of the lineage-specific genes not expressed in this cell line. As obtaining high number of FACS sorted populations is difficult, only the permissive mark examined at bivalent promoters during neural differentiation was H3K4me2. An increase in H3K4me2 levels was detected at most of the bivalent neural promoters activated in CD133+ cells. Significant levels of H3K4me2 were also observed at the promoters of non-neural genes in the H9 hESC-derived progeny, which could reflect the loss of H3K4me3 at the same residue upon resolution of bivalency. H3K4me2 was still detected at non-neural gene promoters also in β-III-tubulin+ cells. Low baseline levels of H3K9me2 and H3K9me3 were detected at the bivalent promoters in H9 hESCs, possibly due to the persistence of spontaneously differentiated cells within the hESC culture (Fig. 4). Both H3K9me2 and H3K9me3 were significantly increased at the silent bivalent promoters in BJ fibroblasts used as a typical differentiated cell reference. In the hESC-derived CD133+ NSCs, H3K27me3 present at the promoters of non-neural markers was supported by H3K9me2 and H3K9me3 appeared additionally in further differentiated β-III-tubulin+ cells. Both H3K9me3 and H3K9me2 were present together with H3K27me3 in β-III-tubulin+ cells at the promoters of silenced early neural stem/progenitor markers. This indicates that Polycomb group (PcG) proteins important for establishing the H3K27me3 mark are not the only regulators that cause repression of bivalent domain promoters at lineage-specific genes during differentiation.
Nonbivalent Promoters of Lineage-Specific Genes Exhibit Changes in Permissive and Repressive Histone Marks Associated with Transcriptional Status of Genes
Several genes upregulated in the CD133+ NSCs do not demonstrate bivalent chromatin domains in the preceding H9 hESC (Fig. 5) supporting published observations of several groups [32, 33]. All examined nonbivalent domain promoters, with the exception of GLI3, belong to proteins important in cytoskeleton formation in differentiated progeny. In case of the NES and VIM promoters, H3K4me3 was detected in pluripotent cells indicating that these genes are also poised for transcription in the early stages of differentiation; however, the mechanisms which keep these genes silent are likely to be Polycomb-independent because they lack H3K27me3. Neither H3K4me3 nor H3K27me3 were displayed at the TUBB3, GFAP, MBP, and GLI3 promoters in H9 hESC. Expression of GFAP and MBP are induced on terminal differentiation of glial lineages and possibly there is little requirement for them to be poised for transcriptional activation in hESCs yet. Other permissive histone marks were found to be present at the lineage-specific genes, which did not display bivalency in H9 hESC (Fig. 5). Interestingly, H3K4me2, H3K9ac, H3ac, and H4ac were present also at the GFAP, MBP, and GLI3 promoters. Only baseline levels of H3K27me3 were displayed at the promoters of NES, GFAP, and MBP in BJ fibroblasts, indicating that other mechanisms independent of Polycomb proteins are required for silencing these genes in non-neural cells. VIM is abundantly expressed in fibroblast cells and a significant increase of H3K4me3 and other permissive marks was observed at its promoter in BJ fibroblasts, with no H3K27me3 detected. Low levels of NES and TUBB3 expression in BJ fibroblasts were reflected in the histone modification pattern at their promoters. Permissive histone marks were at similar levels as in H9 hESCs and no H3K27me3 was displayed at these parts of the genome. As pluripotent cells differentiated to CD133+ NSCs, levels of H3K4me3 and H3K4me2 increased at the promoters of NES, VIM, TUBB3, MBP, and GFAP. Low levels of H3K27me3 also appeared at the promoters of GFAP and MBP, which resembles the establishment of a form of bivalent domain especially because these genes are not yet expressed at significant levels in CD133+ NSCs. Monh et al. reported the appearance of new bivalent domains in mouse neuronal progenitors at the promoters of genes expressed during the latter stages of differentiation (34). Whether the results observed in this study truly report bivalent domains appearing in NSCs is uncertain, as the reported levels of H3K27me3 were much lower than the documented levels of classic bivalent domains present in hESCs. As MBP and GFAP were downregulated in the β-III-tubulin+ neuronal population only the H3K27me3 mark remained at their promoters which is reminiscent of the resolution of bivalent chromatin domains and thus lends support to the possible bivalent status of their promoters in the CD133+ NSCs. As expected, H3K4me3 and H3K4me2 levels were strongly increased at the TUBB3 promoter in β-III-tubulin+ putative neurons, with no H3K27me3 detected. Additionally, significant levels of H3K4me3 were present at the NES and VIM promoters, concomitant with H3K27me3 displayed in β-III-tubulin+ cells, although again not at the levels equivalent to bivalent domains in hESCs. Threshold levels of the H3K9me2 and H3K9me3 marks were observed at several nonbivalent domain promoters in H9 hESCs (Fig. 5). Only the VIM promoter showed no H3K9me in BJ fibroblasts. Interestingly, the levels of H3K9me2 and especially H3K9me3 present at other silent lineage-specific promoters seem to be lower compared with most of the examined bivalent domain promoters, which could again support the prediction that these genes are repressed by other mechanisms than classic repressive histone modifications. A significant increase of the H3K9me2 mark was detected in CD133+ cells at the MBP and GFAP promoters (Fig. 5). It is possible that this mark might play a more important role in keeping the poised genes in progenitor cells in the repressed form, rather than H3K27me3. With the exception of the TUBB3 gene, all other markers were repressed in terminally differentiated cells, with H3K9me2 and H3K9me3 marks detected at their promoters in β-III-tubulin+ cells. Interestingly, the levels of repressive marks were much higher compared with BJ fibroblasts. This could indicate differences between two differentiated cell types, especially with regard to silencing neural stem/progenitor markers during final differentiation. However, further investigation must be performed to exclude the influence of in vitro conditions in regulation of epigenetic changes during hESC differentiation. Only the TUBB3 promoter remained in a fully permissive state during all stages of differentiation.
In this study, we tried to establish the importance of locus-specific post-translational modifications of histones in regulation of transcription during differentiation of hESCs. The presence of a panel of histone modifications at regulatory sequences were correlated with transcriptional status of the examined genes. Given that most studies to date have concentrated on not more than two histone modifications, a full appreciation of the changes occurring at the promoters during differentiation is often not permitted, which may lead to misleading interpretations. One of the main limitations of epigenetic examination of changes occurring during mESC and hESC differentiation lies in the heterogeneity of cell populations produced during differentiation protocols. An efficient purification system was established to examine changes at the promoters of lineage specific genes occurring during differentiation. Two steps of lineage-specification CD133+ NSCs and β-III-tubulin+ putative neurons were found to be an excellent system to examine temporal changes at certain lineage-specific genes during progress of differentiation. Despite genome-wide examination these have not been addressed and the clear trends observed in this study clear elucidate this process in more detail. It may be argued that such trends are specific to the H9 hESC line but because other hESC lines are capable of neuronal differentiation  we believe that the hESC used in this work are representative of pluripotent cell differentiation.
The histone modification pattern present at pluripotency-associated gene promoters was positively correlated with transcriptional status. All pluripotency-associated gene promoters displayed only permissive histone marks in hESCs. We report that H3K9me2 could play a role in initiation of silencing and is followed by H3K27me3 and H3K9me3 during terminal differentiation into neuronal lineages. Additionally, significant levels of permissive histone marks were still observed after in vitro differentiation at silent gene promoters, which on the contrary were not present in human BJ fibroblasts, which suggests that certain levels of permissive marks can still be present at already silent gene promoters, despite the lack of gene expression and that repressive histone modifications exhibit a dominant effect in maintaining the repressed state of the promoters. BJ fibroblasts are a terminally differentiated cell type that we believe represent an opposite state of differentiation to hESC and are a useful comparison for intermediate differentiation states such as the CD133 positive NSCs. It needs further investigation to elucidate whether observed differences arise from the different cell types examined or whether certain mechanisms are not fully functional under in vitro conditions.
Bivalent domains were observed at many early lineage-specific gene promoters, mainly transcription factors, in hESCs. Interestingly, we show that other permissive marks support H3K4me3 in poising the promoters in hESCs. During differentiation bivalent domains were resolved according to the transcriptional status of the gene and repression of the promoters was followed additionally by the elevated levels of H3K9 methylation. Interestingly, the study shows for the first time that in case of bivalent domain gene promoters resolved in the neural progenitor state to favor gene repression upon terminal differentiation, the H3K27me3 mark was redisplayed together with H3K9me2 and H3K9me3 supporting the final repression state. These results are consistent with previous reports, in that expression of only relatively small numbers of genes in human embryonic fibroblasts is controlled by PcG proteins (which establish the H3K27me3 mark) , suggesting that the majority of them are permanently silenced by other mechanisms. It has been reported that bivalent domain promoters together with house-keeping gene promoters and pluripotency-associated gene promoters can be characterized by the presence of high-density CpG islands, which are more likely to be hypermethylated during in vitro differentiation . The association between H3K9 methylation and DNA methylation has been already well documented [38, 39].
A substantial class of genes in hESCs lacks both H3K4me3 and H3K27me3 [33, 34]. As their promoters lack H3K27me3, the relative inactivity of the genes appears to be PcG-independent, yet how they are repressed in ESCs remains unknown. Interestingly, we show that all examined nonbivalent gene promoters displayed significant levels of permissive histone marks, including H3K4me3 in certain cases in hESCs. Given their important role in cell lineage determination, it is likely that these genes become permissive for activation at later developmental stages through as yet uncharacterized epigenetic mechanisms. Data arising from many reports suggest that many genes are regulated mainly by levels of histone acetylation. In many cases, simple inhibition of histone deacetylases is sufficient for activation of expression [40–42]. Moreover, genes encoding cytoskeleton proteins were found to be abundant within repressor element 1-silencing transcription factor (REST) regulated genes during neurogenesis , and correlation between REST and histone deacetylase has been well documented [44, 45]. It is possible that only primary lineage-control genes are required for the very early developmental stages and bivalent domains enable their immediate activation. Importantly, during increases in gene expression, the gene promoters displayed increased levels of permissive marks. Upon repression presence of repressive marks, particularly H3K9me2 and H3K9me3 was detected; however, the increase was much less significant compared with the high levels detected at repressed bivalent domain promoters.
Another question is emerging in recently published data, whether bivalency observed at certain promoters of lineage-specific genes in differentiated cells is equivalent to the same phenomenon in hESCs. In this study, we report that few lineage-specific genes still displayed both H3K4me3 and H3K27me3 in BJ fibroblasts and β-III-tubulin+ neurons; however, the overall relation between two marks was different compared with bivalent domains observed in hESCs. Similarly, the presence of newly formed bivalent domains in CD133+ NSCs is questionable for similar reason. Nevertheless, the possible role of H3K9me2 in repressing certain poised genes in progenitor state should be further approached. The inhibition of gene expression in multipotent state by H3K9 methylation has been reported on many genes expressed in later stages of differentiation such as opioid receptors  or genes encoding synaptic vesicle proteins  that are important in neuronal cells. It should be noted that data arising in the literature is a result of different ChiP protocols with different detection techniques and importantly various threshold levels. Furthermore, only a few bivalent targets have been rigorously confirmed by sequential ChiP to have concomitant H3K4me3 and H3K27me3 marks on the same chromatin segment. The evaluation of the presence of different marks on the same DNA sequence is crucial in making final conclusions. The question whether the bivalency is still a valid phenomenon in lineage-restricted cells still remains open.
The authors are grateful to Marcin Jurga and Colin McGuckin of Newcastle University for the kind gift of antibodies used for immunostaining in this study and to Ian Dimmick and Rebecca Stewart of Newcastle University for their assistance with FACS. The authors also like to thank ONE North East Regional Developmental Agency for funding this work.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.