Epigenetic reprogramming: Enforcer or enabler of developmental fate?

Authors

  • Alexander N. Combes,

    1. Division of Genetics and Population Health, Queensland Institute of Medical Research
    Search for more papers by this author
  • Emma Whitelaw

    Corresponding author
    1. Division of Genetics and Population Health, Queensland Institute of Medical Research
    2. Griffith Medical Research College, Griffith University and the Queensland Institute of Medical Research, Herston, Brisbane, Queensland, Australia
    Search for more papers by this author

Author to whom all correspondence should be addressed.
Email: emma.whitelaw@qimr.edu.au

Abstract

A single fertilized egg is programmed to differentiate into a multitude of distinct cell types that comprise a multicellular organism. Epigenetic mechanisms such as DNA methylation and histone modifications are intricately involved in regulating developmental potential and cellular identity by establishing permissive or repressive chromatin states that are mitotically heritable. Here, we review the dynamics of major epigenetic marks during early mammalian development, and explore the question of whether DNA methylation and chromatin modifications enable or enforce changes that lead to the first cell fate decision.

Introduction

Within an organism, hundreds of different cell types are defined by the unique subset of information they transcribe from a genome held in common. These sets of transcripts encode the physical components and regulatory information required for each cell type to perform its function. Cell types are maintained during development, and within the adult organism, by perpetuating patterns of gene expression through rounds of cell division. How is this cellular memory maintained? And how do cells with identical genetic information come to differ in the first place? The answer to these questions lies in a combination of transcriptional and epigenetic programming of the genome, but the relative importance of these two processes is still unclear. Here, we will discuss this topic in the context of mammalian development.

Master regulators

Developmental biologists have long postulated the existence of “master regulators”, proteins that drive lineage commitment and initiate set patterns of gene expression in less committed precursor cells. This idea became entrenched following the finding by Harold Weintraub and colleagues that a protein, MyoD, could drive differentiated and stem-like cells to turn into muscle cells in vitro (Davis et al. 1987; Weintraub et al. 1991). In the late 1980s it had become clear that cell-type specific patterns of gene expression were achieved, at least in part, through cell-type specific transcription factors (also called trans-activating proteins) that recruit RNA polymerase to particular genes in the genome (Sen & Baltimore 1986; Bodner et al. 1988; Ingraham et al. 1988; Lobe 1992). The defining feature of transcription factors is that they bind to specific DNA sequences adjacent to the genes that they regulate (for an early review see Tjian & Maniatis 1994). The specificity of binding is conferred through the DNA binding domain, which takes various forms including zinc fingers, helix-loop-helix domains, high mobility group boxes, and others. MyoD turned out to be a transcription factor with a helix-loop-helix DNA binding motif that binds to the CANNTG site found in promoters of genes whose expression defines muscle cells (Olson & Klein 1994). There are now numerous examples of transcription factors that act as master regulators in similar ways (see Rothenberg 2007). The Hox transcription factor family, for example, is critical for proper body patterning in eukaryotes (Moens & Selleri 2006). Unravelling the precise rules used by DNA binding domains to bind to specific sequence motifs is likely to aid in the development of therapeutics, because they can provide the specificity required to set and perpetuate expression of a gene, or group of genes, in the genome.

Modulators of epigenetic reprogramming

Recently there has been increasing interest in the proteins that complex with DNA binding proteins to facilitate transcription. Some of these modify the DNA itself (by methylation) or the histones that are packaging the DNA (by methylation, acetylation, phosphorylation etc.). Since these changes alter transcription in the absence of DNA mutations they are referred to as epigenetic modifications and the proteins that carry out the changes are modifiers of epigenetic state. Epigenetic marks such as DNA methylation and histone modifications, change the DNA structure or accessibility and, as such, change the level or probability of transcription (Turker 2002; Li et al. 2006). These changes are known to be involved in silencing of transposons (Yoder et al. 1997b), parental imprinting (Latham et al. 1995; Reik & Walter 2001a,b), and X chromosome inactivation (Heard et al. 1997). Importantly, these marks are mitotically heritable and therefore can facilitate or enforce long-term changes in tissue identity or “cellular memory”.

DNA methylation of promoter regions generally correlates with transcriptional repression. While the mechanisms targeting DNA methylation to specific sequences remain obscure, the major enzymes involved in establishing this mark are known. These enzymes belong to the DNA methyltransferase (Dnmt) family. Dnmt1 is known as the maintenance methyltransferase as it is most efficient at replicating methylation patterns on the newly synthesized strand of DNA (i.e. a hemimethylated CpG site) although it is also able to establish de novo methylation (Bestor & Ingram 1983; Yoder et al. 1997a). Dnmt3a and Dnmt3b are very similar proteins that both catalyse de novo methylation but are expressed independently (Okano et al. 1998, 1999; Watanabe et al. 2002). Mice lacking Dnmt1, Dnmt3a and Dnmt3b die early in development (Li et al. 1992; Okano et al. 1999). The final core member, Dnmt3l is expressed specifically in the germ line and is a non-catalytic protein that acts in concert with Dnmt3 family members to facilitate imprinting and methylation (Bourc’his et al. 2001; Chedin et al. 2002; Hata et al. 2002).

DNA methyltransferases predominantly methylate the cytosine residues in cytosine-guanine (CpG) dinucleotides. Approximately 80% of CpG sites in the mammalian genome are methylated in differentiated tissues (Lister et al. 2009; Popp et al. 2010). Notable exceptions include CpG islands in promoter regions of genes (Suzuki & Bird 2008; Cedar & Bergman 2009). Promoter regions in developmental genes regions contain a balance of permissive and repressive histone modifications that are proposed to prime activity of genes required later in development (Azuara et al. 2006; Bernstein et al. 2006). These histone modifications may prevent establishment of DNA methylation in CpG islands. Until recently, methylation in mammalian genomes was thought to occur almost exclusively at CpG sites; however, a recent genome-wide study has confirmed earlier findings that a significant proportion of non-CpG methylation exists in embryonic stem (ES) cells (Ramsahoye et al. 2000; Lister et al. 2009).

Histone modifications are another major epigenetic mark. Histones octamers make up nucleosomes which, in combination with double stranded DNA, form chromatin. Histone tails extrude from the nucleosome and undergo various protein modifications including methylation, acetylation, phosphorylation, and others (Grant 2001; Kouzarides 2007). These protein modifications confer permissive or repressive states to the histones that affect DNA structure, gene expression, and may be involved in targeting DNA methylation (Kouzarides 2007; Ooi et al. 2007; Cedar & Bergman 2009). For a review of histone modifications during epigenetic reprogramming in mammals, see Morgan et al. 2005; or Hemberger et al. 2009.

During development, a fertilized egg multiplies and the resulting cells progressively differentiate into precursors of adult cell lineages and tissues. At the start of this process, the cells undergo a global clearing of methylation and dynamic changes in histone modifications (Mayer et al. 2000; Oswald et al. 2000; Santos et al. 2002; Beaujean et al. 2004; Fulka et al. 2004; Morgan et al. 2005). These epigenetic marks are subsequently re-established as differentiation proceeds, and correlate with restrictions in developmental potential (Reik et al. 2001; Hemberger et al. 2009). Epigenetic reprogramming is essential for development, as evidenced by embryonic death in mice lacking genes involved in DNA methylation, chromatin remodeling, or histone modification (Li et al. 1992; Okano et al. 1999; Bultman et al. 2000; Peters et al. 2001; Stopka & Skoultchi 2003). However, it is unclear whether epigenetic modifications trigger differentiation events, enable the process, or maintain cell fate after a decision has been made. We will review the fundamental elements of epigenetic reprogramming in development, and examine the relationship between transcription factors, epigenetic marks and differentiation in the specification of the inner cell mass and trophectoderm in the early embryo.

Epigenetic reprogramming during development

At least two phases of genome-wide epigenetic reprogramming occur during murine development. The first phase occurs in the germ line when primordial germ cells enter the gonad and begin to adopt sex-specific fates in mid-gestation. The second phase involves global demethylation in the zygote, which correlates with removal of sperm- and oocyte-specific epigenetic programs, and reactivation of totipotency (Mayer et al. 2000; Oswald et al. 2000; Santos et al. 2002; Beaujean et al. 2004; Fulka et al. 2004). Both reprogramming events involve widespread loss and subsequent resetting of DNA methylation and histone marks.

Reprogramming in the germ line

The germ line is unique in that it is the only cell type in the body to undergo meiosis and it alone transmits genetic, and perhaps epigenetic, information to the next generation. Primordial germ cells are specified in the early embryo and actively migrate to the developing genital ridge (Ewen & Koopman 2009). Before entry into the genital ridge, germ cells carry a distinct epigenetic program that includes select parental imprints and marks reflecting their progenitors (Sasaki & Matsui 2008). This epigenetic program is reset as germ cells complete their migration into the genital ridge (Yamazaki et al. 2003). DNA methylation is rapidly lost, along with the repressive dimethylation of histone 3 lysine 9 (H3K9me2), and a barrage of other histone modifications (Hajkova et al. 2002, 2008; Lee et al. 2002; Yamazaki et al. 2003). Interestingly, nucleus size increases in response to the changes in nuclear chromatin architecture (Hajkova et al. 2008). The first genome-wide study of DNA methylation in germ cells revealed that after deprogramming, male and female germ cells have approximately 60% and 70% less methylation, respectively, than ES cells (Popp et al. 2010). This reduction in methylation is proportional across most classes of genomic elements with the exception of retroviral/retrotransposon long terminal repeat (LTR) sequences, which retain slightly more methylation than other classes (Popp et al. 2010). This retention of methylation at LTR sequences may serve a protective role in preventing retrotransposon reactivation (Yoder et al. 1997b), and may also account for reports of transgenerational epigenetic inheritance at LTR-linked loci in mice (Morgan et al. 1999; Lane et al. 2003; Rakyan et al. 2003).

Rapid loss of methylation patterns in primordial germ cells, despite the presence of Dnmt1, point to a mechanism involving active demethylation (Hajkova et al. 2002). Recent candidates for this mechanism in mammals include Activation induced cytidine deaminase (Aid or Aicda) and Apolipoprotein B mRNA editing enzyme, catalytic polypeptide 1 (Apobec1) (Morgan et al. 2004; Bhutani et al. 2010). These enzymes are able to deaminate 5-methyl cytosine, leading to a T:G base mismatch that can subsequently be repaired to complete cytosine demethylation without mutation (Morgan et al. 2004). Genome-wide bisulfite sequencing in wild-type and Aid-deficient embryos has confirmed a moderate demethylation role for Aid in primordial germ cells (Popp et al. 2010). In Aid-deficient mice CpG methylation levels were increased by 7% in male and 12% in female germ cells, presumably due to a lack of Aid-mediated demethylation. So, while Aid is involved in demethylation, it is clearly not the only factor required. Further, Aid appears to act only within the germ line as global methylation levels in sperm, fetus, and placenta, were not affected by Aid-deficiency (Popp et al. 2010).

Reprogramming in the early embryo

Upon fertilization, the two haploid genomes, the paternal and the maternal, are packaged differently and require considerable processing before transcription and cell division can begin. The maternal genome is packaged around nucleosomes, whereas the paternal genome is stored predominantly in linear arrays on sperm-specific DNA packaging proteins called protamines (Balhorn 1982; Govin et al. 2007). Protamines in the paternal genome are promptly replaced with acetylated histones, and DNA methylation is strikingly lost in the paternal, but not the maternal pronucleus within the first 12–24 h after fertilization (Fig. 1). (Mayer et al. 2000; Oswald et al. 2000; Santos et al. 2002). Again, this demethylation event must be an active process, because it occurs before the first round of DNA replication. The profile of histone modifications also differs between maternal and paternal genomes at this time; for a list see Morgan et al. 2005 or Hemberger et al. 2009.

Figure 1.

 Global DNA methylation in maternal and paternal genomes during early development. Developmental timeline from the zygote to the early blastocyst depicting levels of maternal (M) and paternal (P) genome methylation in gradient bars (dark shading indicates more methylation). ICM, inner cell mass; TE, trophectoderm. Adapted from Santos et al. 2002.

During the first cleavage division1 in the zygote the maternal and paternal genomes combine so each of the daughter blastomeres2 inherits approximately equal amounts of methylated DNA. DNA methylation levels of the maternal genome in blastomeres sequentially decrease during subsequent divisions (Monk et al. 1987; Rougier et al. 1998) to levels that are undetectable by immunohistochemistry in the morula (Santos et al. 2002). This reduction in methylation is thought to be caused by the exclusion of Dnmt1 from the nucleus and resultant failure to maintain DNA methylation patterns during replication (Carlson et al. 1992). Despite the general loss of methylation, some imprinted sequences and retrotransposons maintain their methylation status during this time (Lane et al. 2003; Edwards & Ferguson-Smith 2007; Blewitt et al. 2008). This process of epigenetic clearing from the zygote to the morula sets the stage for development of the first two cell lineages in the blastocyst.

Establishment of trophectoderm and inner cell mass in the early blastocyst

Epigenetic asymmetry has been observed at the blastocyst stage of the early embryo, coincident with differentiation of the trophectoderm (TE), and inner cell mass (ICM) from the morula. This differentiation event involves interplay between epigenetic marks, cell position, and transcription factors (Fig. 2). Lineage bias towards TE or ICM has been observed as early as the four-cell stage in cell tracing experiments (Gardner 2001; Piotrowska et al. 2001; Piotrowska-Nitsche et al. 2005). The earliest marks associated with predisposition to either lineage are specific chromatin modifications; namely, methylation of arginine residues 2, 17, and 26 of histone H3 (e.g. H3R26me) (Torres-Padilla et al. 2007). These residues are thought to be methylated by coactivator-associated arginine methyltransferase 1 (CARM1). Microinjection of Carm1 mRNA into individual blastomeres resulted in increased H3R26me, upregulation of the ICM-enriched pluripotency marker Nanog, and skewed fate towards ICM (Torres-Padilla et al. 2007). CARM1 itself is not necessary for ICM development (Yadav et al. 2003), but this may be due to redundancy. The report by Torres-Padilla and colleagues establishes a histone modification as the earliest known difference associated with change in cell fate in the mammalian embryo.

Figure 2.

 Progression of cell commitment through the first fate decision. Timeline of events that contribute towards establishment of trophectoderm (TE) and inner cell mass (ICM) fates. Cell lineage biases are established as early as the 4-cell stage and are reinforced by cell polarization, positional cues, epigenetic and transcriptional changes. However, cells are able to switch between fates until Elf5 is silenced by methylation in the ICM.

One definitive step towards differentiation of the ICM and TE is the asymmetric division of cells from the 8 to 16 cell stage. During this division, a number of cell polarity proteins become apically localized and are preferentially retained by outer cells (Thomas et al. 2004; Plusa et al. 2005). Other proteins are basally located and are preferentially retained by the inner cells following cell division (Vinot et al. 2005). This differential inheritance of polarity proteins seems to establish the polarized trophectoderm and the apolar inner cell mass (Zernicka-Goetz et al. 2009). Modulation of polarity proteins alone is sufficient to determine whether a cell will be located externally or internally (Plusa et al. 2005).

One report has also suggested that mRNA for a determinant of TE differentiation, caudal type homeobox 2 (Cdx2), is also preferentially retained in outer (TE-biased) cells during the 8–16 cell asymmetric division (Jedrusik et al. 2008).

Transcription factors, Cdx2, Octamer-binding 4 (OCT4/Pou5f1), Sry-box 2 (Sox2), and Nanog homeobox (Nanog) are key lineage markers and determinants of the ICM and TE. Yet up to the 8-cell stage, Oct4 and Sox2 are variably expressed in all cells, and when Cdx2 and Nanog expression is initiated, their expression is also variable and not exclusive to a particular lineage (Dietrich & Hiiragi 2007). However, as differentiation proceeds, higher expression of Oct-4 and Nanog is associated with cells that are located internally and contribute to the ICM, whereas higher Cdx2 levels are associated with cells towards the outside that will contribute to TE. Forced repression of Oct-4, or overexpression of Cdx2, is able to direct ES cells to differentiate into TE (Niwa et al. 2000, 2005). Upstream of Cdx2, TEA domain family member 4 (Tead4) is required for Cdx2 expression and establishment of the TE (Nishioka et al. 2008). Tead4 transcriptionally activates Cdx2 and other TE genes in outer cells but is suppressed by cell-contact and kinase-mediated inhibition (Nishioka et al. 2009). Contact-inhibition of Tead4-mediated Cdx2 expression provides a molecular mechanism by which cell position can contribute to fate decision in the ICM.

Against the background of the early epigenetic and transcriptional changes mentioned above, de novo DNA methylation is initiated in the forming ICM, which heralds further epigenetic asymmetry between the ICM and TE. The onset of global DNA methylation in the ICM correlates with the expression of Dnmt3b, which is conspicuously absent from the external cells that will contribute to the TE. Dnmt3a is not expressed in either cell lineage at this stage (Watanabe et al. 2002). Notably, the levels of H3 lysine 27 methylation (H3k27me, associated with active chromatin) are clearly higher in the ICM than in the TE (Erhardt et al. 2003; Sarmento et al. 2004), whereas other histone modifications show more subtle variation between the two compartments (Hemberger et al. 2009). Little is known about the locus-specificity of the DNA methylation in the ICM but expression of the pluripotency factors Nanog, Oct4, and Sox2 is maintained in this tissue. One rationalization for the differential methylation of ICM and TE is that strict genome defense is required in the ICM because it will give rise to the embryonic and adult tissues, as opposed to the TE, which is only required to form the extra embryonic tissues that are discarded at birth (Hemberger et al. 2009).

During the formation of the blastocyst, differentiation of the TE and ICM is somewhat flexible. Despite location cues, variable expression of lineage determining factors enables cells to adopt either fate when transplanted to different locations within the early blastocyst (Tarkowski & Wroblewska 1967; Dietrich & Hiiragi 2007). This flexibility is consistent with the fact that blastomeres can be removed without disruption to normal embryonic development. However, the ability to change between fates ends in the late blastocyst stage at the point of lineage fixation. E74-like factor 5 (Elf5) is a TE-specific transcription factor that comes on with differentiation and is necessary for formation of trophoblast stem cells and embryonic development (Donnison et al. 2005). Elf5 expression is promoted by Cdx2 and Elf5, in turn, promotes Cdx2 expression to establish a TE-specific positive feedback loop. This feedback loop is shut down in the ICM by DNA methylation of the Elf5 promoter, establishing a formal boundary between the two lineages (Ng et al. 2008). Thus, during differentiation of the ICM and TE, epigenetic marks are the earliest known contributing factor to this fate decision (Torres-Padilla et al. 2007), and also play a role in finalizing developmental fate by establishing lineage restrictions (Ng et al. 2008).

Do epigenetic changes enforce or enable differentiation?

The differential levels of H3R26me that predispose cells towards ICM fate by activating Nanog, provide tantalizing evidence for epigenetic control of this cell fate decision (Torres-Padilla et al. 2007). However, the variable expression of key transcription factors in blastomeres suggests a model where epigenetic factors, the stochastic segregation or initiation of lineage determinants, and cell position, combine to result in strong bias towards one lineage or the other. This lineage bias is then fixed in place by establishment of TE-specific positive feedback loops centered around Cdx2 and Elf-5 and silencing of Elf-5 in the ICM (Ng et al. 2008).

How could this series of events unfold on a mechanistic level? And do epigenetic events in this context enable, or enforce differentiation? One possibility is that high levels of H3R26me increase the probability of activation of Nanog and Sox2, while low levels would favor Cdx2 expression. Nanog protein would then enforce ICM-specific transcriptional circuits and perhaps trigger an active chromatin state at that locus while repressing TE circuits, and vice-versa for the low-H3R26me Cdx2-expressing cells. Where an active chromatin state was established at the Nanog promoter, daughter cells would inherit this active state and have a much higher probability of expressing Nanog again. There is precedent for this type of mechanism in Drosophila (Simon 1995; Ringrose & Paro 2007). If such a mechanism were in play, then although the histone modification precedes the upregulation of Nanog, the latter is actually the causative factor in the differentiation event, and epigenetic mechanisms enforce this decision (Fig. 3a). Alternatively, H3R26me may play a gatekeeper role by allowing (in the case of ICM) or preventing (for TE) transcription factor access to promoters. If this were critical then modulation of H3R26me levels must occur for differentiation to progress down either pathway. This binary mode of action could be described as enabling because it controls transcription factor access to binding sites, but does not initiate the developmental changes (Fig. 3b). However, because H3R26me is present in all blastomeres, it is unlikely that this mechanism applies.

Figure 3.

 Alternate interpretations of the relationship between epigenetic and transcription factors during control of differentiation. (a) (i) A master regulator (transcription factor, TF) binds its target(s) and activates transcription in a low probability event. (ii) The epigenetic state of the target locus changes from neutral to permissive/active. (iii) The epigenetic state is inherited during DNA replication and increases the probability of TF binding to its target gene, thereby enhancing and enforcing differentiation. (b) (i) A repressive epigenetic state prevents TF binding to its target site. (ii) The epigenetic state of the target locus changes from repressive to permissive/active. (iii) TF can bind its target locus and initiate transcription and differentiation. In this case, the epigenetic change enables the transcription factor to perform its role.

Conclusion and perspectives

Because epigenetic mechanisms appear to lack specificity, it is difficult to imagine how they could be a sole determinant for a fate decision. But in the context of a cell mass where multiple determining factors are present at variable levels, an epigenetic factor could assist in lineage determination by establishing cellular memory of a low-probability transcriptional event, and enforcing the changes that ensue, in daughter cells following division. We consider that epigenetic factors do not direct changes in cell fate, but that they enable such changes to occur and serve to entrench them through cellular memory. During differentiation of the inner cell mass and trophectoderm, there is evidence at multiple points that epigenetic mechanisms enforce changes in cell fate that are specified by other factors (Santos et al. 2002; Torres-Padilla et al. 2007; Ng et al. 2008). However, the precise mechanisms underlying these changes are still to be defined.

Clearing, and subsequent re-establishment of epigenetic marks in the germ line and the early embryo correlates with resetting differentiation and initiating new developmental trajectories. This process is of particular relevance to regenerative medicine approaches based on somatic cell nuclear transfer or induced pluripotent cells (Gurdon et al. 1958; Takahashi & Yamanaka 2006). Both of these approaches involve de-differentiation of a mature somatic cell from a patient and reprogramming towards a cell or tissue type that can subsequently be transplanted to fill a medical need. This approach would avoid the issues of immune rejection and limited organ availability that are currently associated with transplants, and enable the correction of genetic defects. However, there are multiple technical issues surrounding regenerative medicine, including the problem of how to achieve effective deprogramming of differentiated cells. This issue of effective deprogramming is best illustrated by the low efficiency of cloning by somatic cell nuclear transfer in animals, and by abnormalities in the animals that are produced by this technique (Kato et al. 1998; Rideout et al. 2001). Several studies have shown that these animals carry multiple epigenetic defects thought to result from incomplete deprogramming (Rideout et al. 2001). Epigenetic reprogramming during development demonstrates that effective clearing of epigenetic marks is likely mediated by an active enzymatic process. Discovery of the factors involved in this process is the focus of intense investigation around the world.

Genome wide changes in DNA methylation levels and histone modifications during development signal a broad role for epigenetic factors in transcriptional regulation and genome architecture. Precisely what those roles are in each particular instance, to a large extent remains unknown. Knockout studies have demonstrated that development cannot proceed without functional methyltransferases or other key factors that influence epigenetic changes (Li et al. 1992; Okano et al. 1999; Grant 2001; Cedar & Bergman 2009). But questions remain about site selection, breadth of function, and central targets that contribute to the associated phenotypes. The recent application of deep-sequencing technologies has yielded broad insight into the genome-wide methylation status of particular cell types and differentiation states (Lister et al. 2009; Popp et al. 2010) and will prove a powerful approach to elucidating global and specific roles for epigenetic factors.

Footnotes

  • 1

    Cleavage division: A cell division between the zygote (1-cell) and morula (16-cell) stage embryo where the number of cells increases with each division, but the size of the embryo remains the same.

  • 2

    Blastomere: Any individual cell between the zygote and morula stage.

Acknowledgments

This work was supported by research grants from the National Health and Medical Research Council of Australia.

Ancillary