Embryonic stem cells (ESCs), which are derived from the inner cell mass of the mammalian blastocysts, possess an unlimited potential for self-renewal and the capacity to differentiate into all kinds of somatic cell types in vitro and in vivo [1, 2]. The pluripotency of ESCs has elicited great excitement regarding their potential applications for cell-replacement therapies for degenerative diseases, including type I diabetes and Parkinson disease. Multiple factors have been identified that are required for maintaining the pluripotency of ESCs, including ESC-selective transcription factors such as Oct4 , Nanog [4, 5], and Sox2  and extracellular signaling molecules such as leukemia inhibitory factor (LIF)  and bone morphogenic proteins  (reviewed in ). For example, mouse ESCs lacking Oct4 have been shown to lose self-renewal capacity and to differentiate into extraembryonic trophectoderm [3, 10]. In addition, Oct4-deficient embryos die at peri-implantation stages because of the conversion of the inner cell mass to the trophoblast lineage . Likewise, withdrawal of LIF and inhibition of its signaling transducer JAK/STAT3 (Janus kinase/signal transducer and activator of transcription 3) have been shown to induce differentiation of mouse ESCs . These studies provide compelling evidence indicating that these factors play a critical role in the maintenance of the pluripotency of ESCs, although their precise mechanism of action is not well-understood, and as yet, very little is known regarding the minimal complement of genes that are sufficient for maintenance of ESC pluripotency.
For decades, a major challenge for developmental biologists has been to elucidate mechanisms whereby a structurally and functionally heterogeneous organism is built from genetically homogeneous cells. A particularly perplexing question has been to determine how pluripotent stem cells within close proximity to one another within a developing blastocyst and thus exposed to very similar (if not identical) environmental cues, ultimately form different cell lineages [11, 12]. Indeed, even under carefully controlled cell culture conditions, apparently “identical” ESCs can be induced to form multiple, different cell types despite sharing very similar local environmental cues [11, 12]. Although very focal gradients of morphogenic substances exist within developing organisms that can pattern a developmental field in a concentration-dependent manner , it is unclear how cells interpret the graded signals of morphogens to exhibit precise gene expression patterns that lead to formation of distinct cell lineages. Moreover, whereas it is easy to see how morphogenic gradients are produced and perpetuated once heterogeneous cell types are formed, this is conceptually much more difficult at the earlier stages of development, when cells are relatively “homogeneous.” As such, the morphogen gradient theory alone is not enough to explain the very complicated cell lineage determination and cell differentiation processes that occur during development of highly complex higher organisms, including the human.
Epigenetic mechanisms are defined as a heritable code other than the genomic sequence and include post-translational modifications to the histones , DNA methylation of CpG (cytosine and guanine separated by a phosphate) nucleotides , ATP-dependent chromatin remodeling, exchange of histones and histone variants , and small RNA molecules  that may serve to guide epigenetic machinery to specific genomic loci. These epigenetic mechanisms have been implicated in the regulation of gene activation and silencing at the level of transcription by regulating how genomic DNA is packaged along with histones into chromatin. For example, DNA sequences that are wrapped “tightly” with histones and condensed into highly folded chromatin fibers are resistant to transcriptional activation (i.e., “heterochromatic”), whereas DNA sequences that are “loosely” associated with histones and present within unfolded chromatin structures are permissive to transcriptional activation (i.e., “euchromatic”) . Of major interest, recent studies indicate that ESCs appear to employ several unique histone modification mechanisms for maintaining pluripotency, as well as permissiveness for activation of cell lineage upon appropriate environmental cues [18, 19]. In this review article, we will discuss the mechanisms controlling these unique histone modifications at lineage-control gene loci in ESCs and consider their contribution to the pluripotency of ESCs. In addition, we will propose a novel “histone modification pulsing” model in which key lineage-control gene loci in individual ESCs undergo pulsatile exchange of histone modifications within chromatin and thereby confer differential responsiveness to environmental cues important for cell lineage determination. Finally, we will consider how these pulsatile histone modifications are altered during differentiation and maturation. Before reviewing the epigenetic mechanisms in ESCs, we will briefly summarize the current state of our knowledge of epigenetic mechanisms regulating gene transcription in differentiated somatic cell types. (Please see the excellent comprehensive reviews of this topic by Jenuwein and Allis  and Margueron et al. .)
Histone Modifications Profoundly Influence Gene Transcription
Genomic DNA in eukaryotic nuclei is packaged into a compact structure known as chromatin. The fundamental unit of chromatin is the nucleosome, which is composed of two copies each of four core histones—H2A, H2B, H3, and H4—wrapped by 146 bp of DNA . Nucleosomes, in turn, are connected together by variable lengths of linker DNA and histone H1. The N-terminal tails of histones are subjected to various post-translational modifications such as acetylation, methylation, phosphorylation, ubiquitination, and ADP-ribosylation . These histone modifications have been demonstrated to control structural chromatin compaction and, in turn, regulate gene transcription through organizing higher order interactions between adjacent nucleosomes, altering the positioning/spacing of nucleosomes, and mediating contacts between histones and DNA . For example, acetylation of H3 and H4 as well as methylation of H3 at lysine 4 (H3 Lys-4) are generally associated with active gene transcription, and these modifications have thus far been found exclusively within euchromatin in higher eukaryotes [23, –25]. Acetylation of histones neutralizes their positive charges and diminishes the interaction between adjacent nucleosomes, which accordingly promotes the unfolding of compacted chromatin fiber and exposure of the DNA template to transcription factors . In addition, acetylated histones and H3 Lys-4 methylation can function to directly tether transcriptional activators to chromatin [27, 28]. Indeed, Pray-Grant et al.  showed that the chromatin remodeling protein Chd1 binds to methylated H3 Lys-4 and recruits the SAGA (Spt-Ada-Gcn5 acetyltransferase) complex to activate gene transcription. In contrast, methylation of H3 Lys-9 and methylation of H3 Lys-27 promote the formation of a compact heterochromatin structure by recruiting heterochromatin-associated protein-1 and polycomb repressive complex-1, respectively [29, 30]. Although both modifications have been shown to associate with gene silencing, the repression mediated by H3 Lys-27 methylation is related to facultative (developmentally regulated) heterochromatin formation, whereas H3 Lys-9-mediated repression connects to constitutive (permanent) heterochromatin in several circumstances [31, –33].
Remarkably, histone modifications and the chromatin patterns they encode can spread over kilobase lengths of genomic DNA in a self-propagating fashion after establishment from a central nucleation point. These patterns are then faithfully transmitted from parent to daughter cell afterward, even in the absence of the initial nucleation signal, giving rise to the concept of epigenetic inheritance [30, 34]. These findings raised the possibility that histone modifications might be used during cellular differentiation to stably program different gene subsets for transcriptional activation or silencing, thereby giving rise to different cell lineages and cell lineage “memory.” Indeed, it has been found that the chromatin of genes whose expression is restricted to a specific cell lineage is programmed with certain histone modifications that contribute to transcriptional activation (e.g., H3 Lys-4 methylation and H4 acetylation) in the cell type in which these genes are expressed. In contrast, the chromatin of these same genes is programmed with a different subset of histone modifications that associate with gene silencing (e.g., H3 Lys-9 methylation and H3 Lys-27 methylation) in other cell types [35, –37]. For example, we recently showed that the promoter loci of genes specific to the smooth muscle cell lineage acquired high levels of H3 Lys-4 methylation and H4 acetylation during development of the smooth muscle cell lineage, whereas the same gene loci were instead programmed with H3 Lys-27 methylation in other cell lineages ( and O.G. McDonald and G.K. Owens, unpublished observations). In summary, the results of the preceding studies provide clear evidence illustrating that histone modifications not only are the marks reflecting the status of transcriptional activity of the genes, but also actively participate in the regulation of gene transcription in differentiated somatic cells.
Many Key Developmental Genes Display Histone Modification Patterns Specific to ESCs, Which Make the Genes Permissive for Activation in Response to Appropriate Environmental Cues
There is evidence that histone modifications play an important role in the regulation of gene expression patterns in ESCs. As is the case in the active gene loci in differentiated somatic cell types, it has been shown that the promoter region of active genes in ESCs, such as Oct4 and Nanog, is marked by acetylation of H3 and H4 and that these histone modifications are important for active gene transcription [39, –41]. These results indicate that ESCs employ similar epigenetic mechanisms for active gene transcription as compared with differentiated somatic cells. However, recent evidence suggests that there are some unique histone modification mechanisms in ESCs for silencing the lineage-control genes, which are not actively transcribed in ESCs but which may be activated later during differentiation.
Azuara et al.  recently examined the histone modification patterns at multiple lineage-control gene loci in mouse ESCs by quantitative chromatin immunoprecipitation (ChIP) assays. They showed that a number of critical transcription factors for cell lineage determination, including Sox1, Nkx2-2, Msx1, Irx3, and Pax3, were not expressed in ESCs but associated with both activating (H3 Lys-9 acetylation and H3 Lys-4 methylation) and repressive (H3 Lys-27 methylation) histone modifications within their promoter loci (Fig. 1). They also showed that expression of these lineage-control genes was aberrantly induced in Eed-deficient ESCs, which lack the Eed-dependent methyltransferase enzymatic activity for H3 Lys-27, and therefore concluded that the presence of H3 Lys-27 methylation is functionally important for preventing expression of lineage-control genes in ESCs. Likewise, results of recent elegant studies by Lander and colleagues  showed that the chromatin of key developmental genes, including Sox, Hox, Pax, and Pou gene family members, displayed a unique histone modification patterns in mouse ESCs, in that the genes contained large stretches of repressive H3 Lys-27 methylation, while simultaneously harboring activating H3 Lys-4 methylation around the transcriptional start sites. They termed this unusual combination of modifications as “bivalent” histone modifications. Of key importance, they demonstrated that these apparently conflicting histone modifications were indeed present at the same physical chromosomes by using sequential ChIP assays for H3 Lys-27 methylation and H3 Lys-4 methylation and concluded that bivalent modification patterns were not due to the presence of two distinct subpopulations of ESCs containing solely H3 Lys-4 methylation or H3 Lys-27 methylation. They also showed that bivalent histone modification patterns resolved during the differentiation from ESCs into a neuronal cell lineage, in that only H3 Lys-4 methylation remained within neuron-specific gene promoters in neuronal precursors, whereas H3 Lys-27 methylation disappeared from these neuron-specific promoters. Conversely, promoters of other genes that remained silent in neurons lost H3 Lys-4 methylation, whereas they retained enrichment for H3 Lys-27 methylation. Taken together, these studies provide evidence that key lineage-control genes in ESCs are marked with a unique combination of activating and repressive histone modifications, which are normally present exclusively in euchromatin or heterochromatin, respectively, in differentiated somatic cells.
Azuara et al.  and Bernstein et al.  postulated that this unique ESC-specific histone modification pattern could contribute to the maintenance of pluripotency in ESCs, because the bivalent histone modifications silenced the genes in ESCs due to the dominant effect of H3 Lys-27 methylation over H3 Lys-4 methylation, while preserving their potential to become activated upon initiation of ESC differentiation via the removal of H3 Lys-27 methylation. H3 Lys-27 methylation has been shown to be related to facultative heterochromatin formation [31, 32], whereas H3 Lys-9 methylation can stably silence the genes by the formation of constitutive heterochromatin [33, 42]. Thus, lineage-control genes in ESCs would be positioned in a relatively “permissive” chromatin conformation that is maintained by the balance between H3 Lys-4 methylation and H3 Lys-27 methylation. This is supported by the findings that bivalently modified lineage-control genes did not contain H3 Lys-9 methylation . Such a permissive chromatin conformation at lineage-control gene loci in ESCs would make these genes accessible for chromatin remodeling complexes and transcription factors for subsequent transcriptional initiation in response to appropriate environmental cues. When ESCs are committed to differentiate into a particular lineage, the repressive histone modifications would be removed from the required lineage-control gene loci and the activating modifications would be maintained, thereby allowing initiation of transcription. On the contrary, when ESCs undergo induction to other cell lineages, the activating histone modifications would be removed from the unnecessary lineage-control gene loci while repressive modifications would be maintained, and the genes would be stably silenced. As such, the bivalent histone modifications at the lineage-control gene loci are likely to behave as the initial platform for either later activation or inactivation in ESCs, and such a transcriptional-permissive status of lineage-control genes may play a key role in the maintenance of pluripotency in ESCs.
However, the bivalent histone modifications were not detected at promoter loci of the multiple lineage-control genes, including Myf5 and Mash1 [18, 19, 43]. Myf5 is a member of MyoD transcription factor family and has been shown to be a critical regulator for muscle lineage determination , whereas Mash1 is a critical transcription factor for the production of neural precursor cells . In addition, the studies by Bernstein et al.  also showed that a number of key developmental genes in ESCs were marked only by H3 Lys-4 methylation (e.g., Pbx3) or did not exhibit either H3 Lys-4 methylation or H3 Lys-27 methylation (e.g., Foxp1). Given their important roles in cell lineage determination, it is likely that these genes are also permissive for activation at later developmental stages through as-yet-uncharacterized epigenetic mechanisms. Taken together, results thus far suggest that only a subset of key developmental genes and lineage-control genes in ESCs display the bivalent chromatin architecture in ESCs, while other lineage-control genes display different histone modification patterns. The functional consequences of these differences are not yet clear, although it could be a result of different kinetics of histone modification exchange at these gene loci, as described in a later section of this review.
What Factors and Mechanisms Program the Unique Histone Modification Patterns in ESCs?
At present, it remains unclear why only a subset of lineage-control genes exhibit bivalent histone modification patterns in ESCs, and the precise mechanisms whereby the bivalent histone modifications are programmed into the selective lineage-control gene loci are unknown. The hierarchical activation model used for hematopoietic cell lineage determination  might explain why only a subset of lineage-control genes are modulated with bivalent histone modifications in ESCs. Based on this model, it is possible that only the primary lineage-control genes—such as the Sox, Hox, Pax, and Pou gene family—that are required for the very early developmental stages are modulated by unique histone modification patterns in ESCs and that activation of these primary genes triggers a sequential activation of downstream lineage-control genes. However, results of previous studies by multiple investigators have shown that a subset of cell lineage-specific genes that are not the transcription factors are also associated with activating histone modifications in mouse ESCs [35, 46, 47]. For example, it has been reported that promoter regions of pancreatic β-cell-specific insulin gene, neuron-specific synaptotagmin-4 gene, and pre-B cell-specific λ5-VpreB1 genes are marked by acetylation of H3 and methylation of H3 Lys-4 in ESCs [35, 46, 47]. Moreover, the synaptotagmin-4 promoter and the λ5-VpreB1 intergenic locus have been shown to be associated with several transcription factors, including RNA polymerase II, in ESCs [46, 47]. The reasons why these cell lineage-specific genes contain activating histone modifications within ESCs are not at all clear, but these results clearly suggest that the epigenetic priming mechanisms do not appear solely on the key lineage-control genes.
Because the bivalent modification patterns are specific for ESCs, it is interesting to speculate that ESC-selective transcription factors, such as Oct4, Nanog, and Sox2, may participate in programming these unique modifications at key developmental gene loci. Boyer et al.  and Loh et al.  recently screened the target genes for Oct4, Nanog, and Sox2 by ChIP assays coupled with DNA microarrays in human and mouse ESCs, respectively. Surprisingly, they found that these ESC-selective transcription factors were associated with not only the actively transcribed genes, but also a number of transcriptionally silent lineage-control genes such as Hox genes and Pax genes in ESCs. These findings are consistent with the hypothesis that ESC-selective transcription factors are important mediators for the assembly of bivalent histone modifications in ESCs. Oct4, Nanog, and Sox2 have been reported to function as transcriptional activators for most of their known target genes  but thus far have not been found to possess intrinsic enzymatic activities for histone modifications. Thus, it is interesting to postulate that these ESC-selective transcription factors introduce activating histone modifications into the selective lineage-control gene loci by recruiting histone-modifying enzymes. Conversely, ESC-selective transcription factors might play a role in introducing repressive histone modifications into the lineage-control gene loci by recruiting the enzymes responsible for repressive histone modifications in ESCs, because, during the differentiation, both ESC-selective transcription factors and repressive histone modifications are diminished from the lineage-control gene loci upon activation. It has recently been shown that the Polycomb protein complexes required for methylation of H3 Lys-27 are associated with a number of key lineage-control gene loci in human and mouse ESCs and that their target genes are highly overlapped with those for ESC-selective transcription factors [50, 51]. Thus, it will be of interest to determine whether ESC-selective transcription factors are capable of directly recruiting the Polycomb protein complexes to introduce repressive histone modifications at bivalently modified lineage-control gene loci in ESCs. However, there are several key points to consider. First, more than 100 loci were identified to be associated with bivalent histone modifications in mouse ESCs, but only half of them were associated with Oct4 and Nanog . Second, although the promoter region of the Myf5 gene has been reported to be “unassociated” with both activating and repressive histone modifications, it was occupied by Oct4 in mouse ESCs [18, 49]. These results suggest that the assembly of bivalent histone modifications is not fully explained by these ESC-selective transcription factors. Rather, it seems likely that additional ESC-selective transcription factors that have not yet been identified may contribute to the formation of bivalent histone modifications.
The specific cis-regulatory elements present within promoters of lineage-control genes may also be a determinant for the presence of bivalent histone modification patterns in ESCs. Indeed, Bernstein et al.  found a strong correlation between the localization of CpG islands and the presence of methylated H3 Lys-4 in mouse ESCs. Recently, Lee and Skalnik  showed that CXXC finger protein 1, a binding factor of unmethylated CpG islands, is a component of the mammalian Set1 H3 Lys-4 methyltransferase complex, suggesting that unmethylated CpG islands are the direct target of this methyltransferase. Therefore, it is possible that H3 Lys-4 methylation that appeared in a subset of key lineage-control genes is mediated by the recruitment of Set1 methyltransferase into unmethylated CpG islands at these promoter regions in ESCs. However, it remains to be determined whether CpG islands at key lineage-control gene loci are selectively unmethylated in ESCs. In fact, results of genome-wide analyses showed that the methylation status of CpG islands was not different between mouse ESCs and adult mouse tissues, such as brain and kidney, although these studies did not focus on key lineage-control gene loci . Thus, it would be interesting to determine whether the DNA methylation status of CpG islands at key lineage-control gene loci between ESCs and differentiated cell types is different and whether these CpG islands are responsible for recruiting trans-acting factors that might be important in programming bivalent histone modifications.
Finally, one must also consider the possibility that the failure to detect ESC-specific histone modifications within some lineage-control genes represents a false-negative. Although this is not an example from the lineage-control genes, it has been reported that activating histone modifications appeared only at the intergenic locus, but not the promoter regions of the pre-B cell-specific λ5-VpreB1 genes in mouse ESCs . In this example, activating histone modifications were expanded and augmented from the intergenic locus toward the promoter regions during B-cell lineage differentiation. Thus, it is possible that histone modifications may exist outside of the regions covered by the PCR primers used in ChIP assays.
In summary, there is compelling evidence for the formation of bivalent histone modifications at key lineage-control gene loci in ESCs. However, the specific mechanisms that give rise to these bivalent histone modifications and their precise role in control of embryonic stem pluripotency and subsequent lineage determination remain to be determined.
Lineage-Control Genes May Be Primed by Rapid Exchange of Histone Modifications in ESCs
Our discussion thus far has focused on consideration of different histone modification patterns at distinct lineage-control gene loci in ESCs. However, it is possible that all or most lineage-control genes use a common histone modification-based mechanism for keeping them in a transcription-permissive state, but it has not been observed at multiple lineage-control gene loci due to technical limitations of ChIP assays. In this section, we will propose a histone modification pulsing model in which individual ESCs exhibit stochastic changes in histone modification patterning at specific lineage-control gene loci. That is, at any given instant in time, a particular ESC may exhibit a unique repertoire of histone modifications at critical lineage-control gene loci that in turn alter their response to environmental cues. Indeed, such a model may reconcile experimental evidence to date.
In this model, all or most of lineage-control genes within ESCs are dynamically associated with histone modifications that exhibit different kinetics and stoichiometries (Fig. 2). For example, some lineage-control gene loci such as Sox1, Nkx2-2, Msx1, Irx3, and Pax3, whose promoter was observed by ChIP assays to contain bivalent chromatin patterns, are associated with cyclic histone modification patterning that occurs with a higher frequency and/or a longer duration. In contrast, the exchange rate of histone modifications of other lineage-control gene loci such as Myf5 and Mash1, whose promoter was not observed to contain bivalent chromatin patterns, may exhibit lower frequencies or shorter duration, such that ChIP assays lack sufficient “sensitivity” to detect these modifications. Because individual ESCs would be expected to be asynchronous with respect to these transient locus-selective histone modification exchanges, the probability and sensitivity in detecting histone modifications using mass-averaging methods such as ChIP assays would thus be dependent on the following variables: (a) the fraction of ESCs within the entire population that exhibits a particular histone modification at a given locus and point in time, (b) the amplitude or magnitude of the enrichment of histone modifications, and (c) the sensitivity of methods available to detect that change, including the affinity of antibodies for specific histone modifications. As such, the data that some lineage-control gene loci exhibited unique histone modification patterns in ESCs would suggest that these patterns were present in a larger fraction of individual ESCs at a given time point and/or were of greater amplitude due to tighter packing of modified histones within a given DNA sequence. In contrast, the data that other loci did not have detectable modifications in ESCs would suggest that the modifications appeared very transiently in individual ESCs and thus fell below the threshold for detection in ChIP assays. During the differentiation of ESCs, such dynamic histone modification patterns would become more stable, leading to the higher levels of histone modifications in differentiated somatic cells. This model would be fully consistent with observations from many laboratories [35, 46, 47] that levels of histone modifications are much higher or are much expanded along the gene loci in differentiated somatic cells than those in ESCs, because a far greater fraction of cells would be expected to simultaneously exhibit the histone modifications.
Indeed, a series of highly innovative studies by Misteli and colleagues [54, 55] provide direct evidence in support of this histone modification pulsing model. Using fluorescent recovery after photobleaching (FRAP) technology, they showed that chromatin was highly dynamic in ESCs with regard to the turnover of chromatin proteins. For example, proteins such as H3 on genomic DNA displayed dissociation constants on the order of seconds in mouse ESCs as compared with those on the order of several minutes in differentiated cell types [54, 55]. They also presented evidence that the hyperdynamic binding of histone proteins to DNA was required for the pluripotent potential of ESCs by experimentally modulating the dynamic features of proteins in ESCs. Of importance, they presented compelling evidence that the hyperdynamic state of histone proteins in ESCs was not a function of high rates of cell replication, in that the hyperdynamic feature was observed in ESCs at any time point in the cell cycle. Moreover, an absolutely critical advantage of these studies is that the FRAP technology permits assessments of chromatin protein dynamics at the single-cell level such that authors were able to directly compare the dynamics in both growing and nongrowing cells and observed no differences. Thus, these results suggest that the dynamic exchange of chromatin proteins, including histones within intact chromatin in ESCs, was not a function of cell replication, but, rather suggest that they may be one of the key determinants for the pluripotency of ESCs. As such, this rapid histone exchange rate is a potential mechanism whereby histone modifications might be dynamically deposited on and off the chromatin template at lineage-control genes in ESCs over time.
There are, however, several major unresolved questions about this model. First, virtually nothing is known regarding what factors and mechanisms control the dynamic rate of histone modification exchanges at gene loci within ESCs. Although it is possible that ESC-selective transcription factors and/or specific cis-regulatory elements determine the dynamics of histone modifications, as yet there is no direct evidence that this is the case. Second, it is unknown whether dynamic changes in histone modification patterning of a given gene locus in a given cell type are caused by the exchange of histone proteins containing alternative histone modifications or by in situ modification of existing histones on DNA through dynamic recruitment of histone-modifying enzymes (e.g., exchange of histone acetyltransferase vs. histone deacetylase enzymes). The latter could be tested by real-time measurements of diffusion rates, dissociation rates, and residence times of histone-modifying enzymes on chromatin within single cells and selected gene loci using methods similar to those employed for Polycomb proteins  and glucocorticoid receptor , including combinations of fluorescence-labeling techniques, in situ hybridization, and FRAP microscopy [56, 57]. Alternatively, the possibility of dynamic exchange of histone proteins might be tested by adaptation of fluorescence resonance energy transfer (FRET), assays using fluorescent histone proteins and labeled DNA, because FRET assays have been used to detect protein-protein interactions in real time and at the single-cell level [58, 59]. Moreover, success in detection of histone modification patterns within as few as 50 ESCs was recently achieved using a modified ChIP assay that employs carrier DNA , thus raising the possibility of eventually increasing the sensitivity of CHIP assays to the single-cell level. In summary, although the histone modification pulsing model provides an intriguing mechanism that may contribute to the pluripotency of ESCs, the model also raises multiple questions that require extensive further studies and development of revolutionary new methodologies, including the real-time imaging of the histone modification status at specific genomic loci at the single-cell level.
Asynchronous Histone Modification Pulsing May Play a Role in the Initial Formation of Heterogeneous Cell Populations in the Developing Embryo
One of the most intriguing and poorly understood areas in developmental biology is how morphologically homogeneous pluripotent stem cells within the inner cell mass of a blastocyst respond differentially to virtually identical local environmental cues, leading to development of heterogeneous cell populations. Whereas this is obviously a very complex process involving many interactive mechanisms, including possible partitioning of the egg cytoplasm, maternal imprinting, and many other mechanisms , and it has been recently reported that there is a heterogeneity in pluripotent stem cells within the inner cell mass [61, 62], it is interesting to postulate that the distinct responsiveness of individual pluripotent stem cells within the developing blastocyst is caused at least in part by the asynchronous dynamic nature of histone modification patterning. That is, adjacent pluripotent stem cells in a blastocyst may respond differently to very similar or identical environmental cues because of the difference in the histone modification status at particular developmental gene loci at a given instant in time. As such, it is interesting to hypothesize that the histone modification pulsing model may be important for creating spatial and temporal cell heterogeneity during the embryonic development, as well as for maintaining pluripotency in ESCs.
According to the histone modification pulsing model, all or most of the key lineage-control genes in pluripotent stem cells are predicted to exhibit a variable exchange rate of histone modifications such as H3 Lys-4 methylation and H3 Lys-27 methylation within their chromatin. Although Bernstein et al.  clearly showed that these histone modifications were present simultaneously on the same physical chromosome loci as determined by the sequential ChIP assays in mouse ESCs, it is still possible that the kinetics of each histone modification pattern (e.g., H3 Lys-4 methylation vs. H3 Lys-27 methylation) is asynchronous in individual cells. A simple diagram to illustrate this principle is shown in Figure 3. In this model, the pluripotent stem cell A (Fig. 3) would exhibit a permissive chromatin architecture consisting of the appearance of activating H3 Lys-4 methylation and absence of repressive H3 Lys-27 methylation at a particular key lineage-control gene locus at a particular time point Z. In contrast, other pluripotent stem cells might contain both repressive H3 Lys-27 methylation and activating H3 Lys-4 methylation (Fig. 3, cell B) or contain only repressive H3 Lys-27 methylation (Fig. 3, cell C) at the same gene locus at the same time point Z. At the time point Z, this particular lineage-control gene would be able to exhibit a response to the appropriate developmental stimuli only in pluripotent stem cell A, but not in pluripotent stem cells B and C. The morphogenic signals and ESC-selective transcription factors could “catch” the responsive gene at time Z and then go on to activate the transcription/expression in pluripotent stem cell A, but this would be prevented by the presence of repressive H3 Lys-27 methylation at the same loci in pluripotent stem cells B and C. Although there is no evidence suggesting that genes lacking both H3 Lys-4 methylation and H3 Lys-27 methylation exhibit (or do not exhibit) a response to developmental stimuli, asynchronous dynamics in histone modification patterns between individual pluripotent stem cells may contribute to the formation of heterogeneous cell populations from the apparently identical pluripotent stem cells that share an identical or virtually identical repertoire of local environmental signals and cues.
This initial activation process of the lineage-control gene could then be amplified by subsequent mechanisms, such as the epigenetic inheritance process described earlier, as well as downstream effects of the initial gene activation event. The “pulsing” histone modifications could be further strengthened spatially and temporally by spreading along chromatin and be inherited through cell division, thereby stably programming activation or silencing after the initial nucleation event and effectively extinguishing the ESC-specific histone modification pulsing from these loci in differentiated cell lineages. For example, it has been demonstrated that a single pulse of transcription at silent genes in ESCs can result in replacement of Polycomb proteins and H3 Lys-27 methylation from Fab7 transgene promoters integrated into Hox gene loci, with the Trithorax proteins and H3 Lys-4 methylation . As such, the gene loci with bivalent histone modifications in pluripotent stem cells could be resolved into chromatin containing only activating modifications such as H3 Lys-4 methylation (if the gene is activated) or repressive modifications such as H3 Lys-27 methylation (if the gene remains silent), once differentiation was triggered. Once a heterogeneous cell population is initially established in this manner, there may be little difficulty in maintaining and extending differentiation through establishment of defined gradients of morphogenic factors and the controlled availability of metabolites, hormones, and secreted matrix components that can promote maturation of individual cell populations to their particular destinies.
In this review, we have proposed a novel histone modification pulsing model that may play an important role in the maintenance of pluripotency and cell lineage determination in ESCs. One very intriguing question is whether re-establishment of an asynchronous hyperdynamic state can restore pluripotency in differentiated somatic cells. Recent nuclear transfer studies have provided clear evidence that somatic cell nuclei can be reprogrammed into pluripotent ESCs, embryos, and even live animals . However, as yet, no studies have tested whether the hyperdynamic pattern of histone proteins is also restored after nuclear transfer, although the successful re-establishment of ESC-specific histone modification patterns has been suggested . Indeed, it is interesting to speculate that one might use measurements of the hyperdynamic nature of histone modifications as a possible means to estimate reprogramming efficiency after nuclear transfer and nuclear hybridization or for the evaluation of plasticity of stem cells.
Further studies are also needed to expand these sorts of analyses to a larger repertoire of activating and repressive histone modification patterns in ESCs. For example, studies have shown that methylation of H3 Lys-36 and H3 Lys-79 and phosphorylation of H3 Serine-10 are correlated with actively transcribed genes [65, 66], whereas methylation of H4 Lys-20 plays a role in gene suppression , and a multitude of other modifications are being discovered and linked to gene activation/silencing. Given that the balance between activating and repressive histone modifications is important for the maintenance of pluripotency of ESCs as well as for the cell lineage determination, extensive studies regarding a greater repertoire of histone modifications in ESCs will likely provide novel insights into this field.
Finally, ESCs are expected to be prime candidates for cell transplantation and regenerative therapies for a wide range of human diseases. As such, an understanding of epigenetic mechanisms that control pluripotency and lineage determination in ESCs will undoubtedly contribute to development of improved methods for therapeutic applications of ESCs. Moreover, such studies are likely to provide novel insights regarding the potential contribution of epigenetic reprogramming mechanisms in control of plasticity and pluripotency of somatic stem cell populations and thereby advance applications of these cells for treatment of human disease.