Principles of stem cell biology and cancer: future applications and therapeutics. Edited by T. Regad, T. J. Sayers and R. C. Rees. John Wiley & Sons (2015)
Part I. Stem Cells
Differentiation cues have tremendous effects on the transcriptional profiles of ESCs. Pluripotency-related genes become repressed as developmental regulators become activated in a germ layer-specific manner. How environmental signals drive these coordinated changes and what mechanism triggers the differentiation of cells into one particular embryonic layer is largely unknown. Most likely, differentiation signals activate the expression of early transcription factors capable of both inducing the expression of tissue-specific genes and repressing the expression of alternative-lineage and pluripotency-related genes. For example, CDX2 plays a key role in silencing Oct4 expression in the trophectoderm lineage (Niwa et al., 2005), as well as in inducing differentiation genes (Nishiyama et al., 2009). The transcription factors that are among the first to be induced and can access tissue-specific loci at early stages of development have been termed ‘pioneer factors’. These factors have been found to be able to enter compact chromatin regions in order to activate gene expression during development (Zaret and Carroll, 2011).
Several studies have shown that the Polycomb complex participates in the silencing of the pluripotency network during differentiation (Pasini et al., 2007; Shen et al., 2008), while the methyltransferase G9a has been suggested to contribute to the silencing of OCT4 expression by mediating the deposition of the heterochromatin mark H3K9me2/3 at its regulatory regions (Feldman et al., 2006). However, genes like OCT4 and NANOG do not display either H3K27me3 or H3K9me3 marks in human differentiated cells, but show DNA methylation at dispersed CpG sites (Figure 2.2). Nonetheless, Polycomb and G9a might contribute to establishing a transitional chromatin state that favours the final silencing of some pluripotency-related genes through DNA methylation (Epsztejn-Litman et al., 2008). Additionally, recruitment of linker histone H1 (Terme et al., 2011) and histone variant macroH2A (Barrero et al., 2013b) to pluripotency-related genes during differentiation contributes to the repression of these genes.
Figure 2.2. Various silencing mechanisms of pluripotency-related genes in differentiated cells according to CpG densitiy, comparing the histoneand DNA-modification profiles of pluripotency genes in ESCs and human fibroblasts (HFs) in two pluripotency-related genes. (A) OCT4 contains dispersed CpGs that become hypermethylated in differentiated cells. (B) LIN28A contains a CpG island that is hypomethylated in both cell types. Repression of this gene in differentiated cells correlates with presence of H3K27me3. Black, grey and white dots represent hypermethylated, partially methylated and nonmethylated CpG sites. Histone modifications from ENCODE publically available data (ENCODE Project Consortium, 2011) generated in the Bernstein laboratory and displayed in the UCSC genome browser are shown.
Developmental specification implies a transition from a permissive chromatin state to a restrictive state with prevalent Polycomb repression. In pluripotent cells, the H3K27me3 mark is confined to peaks at bivalent GCrich promoters, but it becomes more broadly distributed in differentiated cells and tissues, contributing to maintenance of the silencing of genes with functions in alternate lineages (Zhu et al., 2013). Although DNA methylation of certain CpG islands correlates with repression of genes during differentiation, most CpG islands remain hypomethylated in differentiated cells, with H3K27me3 the most common mechanism of repression of CpG island-containing genes.
Promoters of developmental regulators that become active in early developmental stages resolve their bivalency into H3K4me3 only, a process that entails loss of H3K27me3. This can be achieved by both passive and active mechanisms. Passively, changes in PcG subunit composition during differentiation can help to relocate complexes from genes that become active to genes that need to be repressed, such as the pluripotency network (Morey et al., 2012). More actively, the H3K27me3 demethylases UTX and JMJD3 can be recruited to developmental genes to facilitate activation (Agger et al., 2007; Burgold et al., 2008). Importantly, UTX and JMJD3 associate with different MLL complexes, which likely determines their recruitment to distinct target genes and suggests a coordinated mechanism for the removal of H3K27me3 and the deposition of H3K4me3.
Interestingly, not all bivalent domains are resolved during differentiation. However, these domains are likely to be different in differentiated cells than in pluripotent cells. The presence of additional histone modifications, a different balance of H3K4me2/3 versus H3K27me3 (Figure 2.3) or occupancy by the H2A variant macroH2A (Barrero et al., 2013a) may all contribute to making these domains less permissive in differentiated cells. The role of these domains in differentiated cells remains unclear, but it might reflect a compromise between the plasticity required for some adult cells, such as adult stem cells and germ cells, and the irreversibility of these domains in terminally differentiated cells.
Unlike in developmental regulators, the regulatory regions of tissue-specific genes are often CpG-poor and become activated preferentially at later stages of differentiation, likely due to the binding of developmental transcription factors. Interestingly, loss of DNA methylation at these CpG-poor genes seems to be accompanied by a gain of H3K4me1 or H3K27me3 during early stages of differentiation. These transitions occur without significant change in gene expression and likely function to prime these genes for later induction in terminally differentiated cells (Gifford et al., 2013). Those tissue-specific genes that remain silenced in a layer-specific manner usually retain DNA methylation at their sparse CpG sites (Xie et al., 2013b).
Figure 2.3. Bivalent domains contain different ratios of modifications in pluripotent and differentiated cells. (A) Comparison of the levels of H3K4me3, H3K4me2 and H3K27me3 between human ESCs and keratinocytes (HEKs) around the TSSs of genes that are bivalent in both cell types. (B) Box plot of mRNA levels in genes marked with bivalent domains in both ESCs and HEKs. Keratinocytes show lower levels of H3K4me2 and H3k27me3 at bivalent domains than ESCs, which might contribute to the different transcriptional statuses found in different cell lines. In both cell lines, bivalent genes are expressed at low levels, but expression is more permissive in ESCs than in HEKs. ENCODE publically available data (ENCODE Project Consortium, 2011) generated in the Bernstein laboratory are shown. Aggregation plots were drawn using SitePro, from the CEAS package (Shin et al., 2009). Expression data have been previously published (Aasen et al., 2008).
Broad changes in the heterochromatin mark H3K9me2/3 have been proposed to play a role in differentiation. Initial reports suggested that differentiation of ESCs causes the expansion of H3K9me2 to large megabase-sized domains, termed ‘large organized chromatin K9-modifications’ (LOCKs), which are maintained by the histone methyltransferase G9a (Wen et al., 2009). These domains are likely to contribute to the establishment of heterochromatin in differentiated cells. However, others have reported no significant differences in the coverage of these domains in pluripotent versus differentiated cells (Filion and van Steensel, 2010). A recent report suggests that increased H3K9me3 signal is found in primary cell cultures compared to tissues and ESCs, and that this is likely triggered by the culture environment (Zhu et al., 2013).
Recent work has revealed dynamic changes in enhancer – promoter interactions during differentiation. The changing enhancer landscape in mammalian development results from the recruitment of lineage-determining factors, which associate with enhancers anchored not only to tissue-specific promoters but also to constitutively active ones (Kieffer-Kwon et al., 2013). This reorganization of interactions also involves changes in histone modifications. Most cell-specific enhancers are CpG-poor and appear to be depleted of H3K27me3 in the cell types in which they are active. A subset of enhancers becomes weakly marked with H3K27me3 in differentiated cells, suggesting that the expansion of the H3K27me3 domains that arise during differentiation may contribute to repression of enhancers that are active in other lineages. However, DNA methylation seems to be a more common mechanism in the inactivation of enhancers during differentiation (Xie et al., 2013b).
Despite the highly dynamic exchange of promoter – enhancer interactions during differentiation, recent data suggest that such interactions are restricted to large yet defined megabase-sized chromatin domains, which have been termed ‘topologically associated domains’ (TADs) (Dixon et al., 2012; Nora et al., 2012). These regions are separated by narrow segments devoid of chromatin interactions that appear to function as boundaries or insulators which block the interactions between two adjacent domains (Figure 2.4). Interestingly, TADs align with several domain-wide features of the epigenome, such as H3K27me3 or H3K9me2 blocks and lamina-associated domains, suggesting a common epigenetic environment for the genes located within one domain and a potential role for the boundaries in preventing the spreading of certain histone modifications. Accordingly, genes located within one particular TAD are coordinately expressed during development (Nora et al., 2012). Although TADs can switch their association with particular chromatin landmarks during differentiation (Nora et al., 2012), their boundaries are stable across different cell types and are highly conserved across species (Xie et al., 2013b), indicating that TADs are a fundamental property of mammalian genomes, organized into broad regulatory domains.
Figure 2.4. Changes in the association of TADs with the nuclear lamina during differentiation. Two different domains are shown (TAD1 and TAD2). Genes located in TAD1 are highly expressed in ES Cell, interact within the same domain and are separated from TAD2 by a boundary region that restricts interactions between the two. Most genes located in TAD2 are silent in ES Cell and interact with the nuclear lamina. Differentiation causes silencing and association with the nuclear lamina of genes located in TAD1, and expression of genes located in TAD2. Despite the changes in expression and interaction with the nuclear lamina, boundaries are conserved, restricting interactions within one domain.