The role of DNA methylation in genome-wide gene regulation during development

The regulation of gene expression during development is a complex process that utilizes many molecular control mechanisms, mainly at the level of transcription. The basic genomic unit of transcription comprises a promoter, located 5′ to the beginning of the gene body, together with other enhancer or silencer regulatory elements that are spatially spread out over the surrounding gene domain. Trans-acting factors play a major role in directing the temporal and cell type-specific control of gene expression, but much of the coordination and stability of this process is mediated by highly regulated changes in chromatin structure and DNA methylation, which control the accessibility of trans-acting factors to cis-acting regulatory elements (Bogdanović et al., 2016; Domcke et al., 2015; Reizel et al., 2018). In this Hypothesis article, we present a model for how these changes in DNA methylation and chromatin structure play a crucial role in controlling gene expression during development.

DNA methylation can be looked at as a form of textual annotation that provides an additional layer of regulatory information, but is not an integral part of the DNA sequence itself. However, in order to appreciate the overall role of DNA methylation in the control of gene expression, it is imperative to first understand the underlying molecular logic of how methylation patterns are established during early development.

In the early mammalian embryo, DNA methylation patterns derived from the gametes are almost completely erased and a new profile is then formed according to a pre-determined program (Reik et al., 2001). Establishment of a new pattern in the offspring is initiated by a wave of de novo methylation that indiscriminately modifies almost all CpG residues in the genome while sparing CpG island-like regions by virtue of cis-acting sequences (Brandeis et al., 1994; Straussman et al., 2009), which dictate a histone modification state (H3K4me3; Greenfield et al., 2018; Long et al., 2016) that sterically prevents de novo DNA methylases (Dnmt3a and 3b) from interacting with underlying DNA (Greenfield et al., 2018; Guo et al., 2015; Noh et al., 2015; Ooi et al., 2007; Otani et al., 2009; Zhang et al., 2010). Once this basal pattern is formed, at about the time of embryo implantation, the molecular machinery responsible for this de novo activity is then downregulated, but the resulting methylation pattern itself is scrupulously maintained in all subsequent cell divisions throughout life (Cedar and Bergman, 2012; Hon et al., 2013; Reizel et al., 2021; Stein et al., 1982), mainly through the action of the Dnmt1 methylase located in the replication fork. Thus, the overall genomic methylation profile in each individual cell actually represents an epigenetic replica of the pattern originally set up in the embryo.

With this general pattern of DNA methylation serving as a starting point, further stages of development are accompanied by targeted stage-specific and cell type-specific changes in DNA methylation (Argelaguet et al., 2019), mediated by local recruitment of either TET demethylases or the de novo methylases Dnmt3a and 3b. These programmed alterations then help to define the specific nature and function of each cell, and hence each biological system, in the body (Cedar and Bergman, 2012; Ziller et al., 2013). Thus, the genome in each cell is covered by an expansive ‘lawn’ of DNA methylation but it is not yet clear how this contributes to the control of gene expression.

Although there is ample evidence demonstrating that DNA methylation inhibits transcription (Razin and Cedar, 1991), its biological role in vivo has not yet been adequately elucidated. The effect of DNA methylation on gene repression is most strikingly observed in cell transfection experiments, in which the expression of inserted unmethylated and in vitro methylated foreign templates can be directly compared (Pollack et al., 1980; Wigler et al., 1981). This effect has been verified by more recent experiments that use epigenetic engineering to change the methylation state of endogenous loci and, in this way, alter their expression patterns (Nuñez et al., 2021; Policarpi et al., 2021; Razin and Cedar, 1991). Nonetheless, the general biological strategy of this programmed epigenetic system during development in vivo is still not clear.

Two possibilities have been put forward to explain how genome-wide DNA methylation might influence gene transcription during development. One idea is that covering the entire genome with DNA methylation serves to reduce transcriptional noise (Bird, 1995; Suzuki et al., 2007), perhaps by decreasing fortuitous interactions between the transcription machinery and the many cryptic low-affinity promoter-like sequences spread out over the entire genome. This would have the effect of making developmentally planned transcription much more efficient by specifically focusing most protein-DNA interactions to those regions that have been pre-programmed to be unmethylated, and thus are accessible for binding. In this model, DNA methylation acts as a non-specific shield covering the entire genome. Indeed, it is this concept that probably explains the promoter methylation and consequent long-term repression of the many endogenous viral sequences present in the genome (Gautsch and Wilson, 1983; Walsh et al., 1998).

A second way by which genome-wide DNA methylation may affect gene expression during development is by providing a mechanism to repress tissue-specific gene promoters globally (Cedar, 1988; Miranda and Jones, 2007). According to this idea, the presence of DNA methylation itself would render these genes inactive in all tissues of the organism. Expression is then seen only in the cell type in which the specific promoter undergoes programmed demethylation, thus enabling transcription. Although attractive, it is unlikely that this simple mechanism plays a major role in tissue-specific gene regulation, mainly because we know that it is probably a combination of specific and non-specific transcription factors acting directly on the promoter that constitutes the rate-limiting component of gene activation. Indeed, RNA analysis has shown that the expression levels of these genes can be orders of magnitude higher than the transcription levels seen in non-relevant tissue; by comparison, the artificial removal of DNA methylation from these promoter sites only raises levels of gene expression by, at most, one or two orders of magnitude, thus suggesting that promoter modification plays a very minor role in this regulatory process (Goren et al., 2006). Furthermore, it appears that many cell type-specific or partially specific genes are actually driven by promoters that are constitutively unmethylated (Reizel et al., 2015; Schug et al., 2005) and packaged in an open chromatin configuration (Heintzman et al., 2009; Thurman et al., 2012) and, as a result, are not influenced by global modification.

On the basis of accumulating data supporting the important role of non-promoter regulatory elements in developmentally programmed gene regulation, we propose that the major influence of global DNA methylation is in fact on the many surrounding enhancers and silencers (Fig. 1) (Lee et al., 2015; Zhang et al., 2013). Regulatory elements of this nature are largely non-CpG island sequences distributed throughout the genome. They thus automatically undergo de novo methylation at the time of implantation, remaining in this methylated state in almost all tissues and cell types and only becoming demethylated in a tissue-specific manner concomitant with terminal differentiation (Hon et al., 2013; Meissner et al., 2008; Ziller et al., 2013). This programmed demethylation has the effect of making these sequences accessible to trans-acting factors in the nucleus (Argelaguet et al., 2019), thus enabling them to influence the level of target gene expression in that cell type (Bogdanović et al., 2016; Domcke et al., 2015; Reizel et al., 2018). The fact that these regulatory elements remain methylated in other cell types serves as a way of controlling factor enlistment and gene expression, even though the target-gene promoters themselves may be unmethylated (Aran and Hellman, 2013; Neiman et al., 2017; Reizel et al., 2018).

The idea that global DNA methylation of regulatory elements plays a major role in gene regulation is strongly supported by the observation that this entire system is highly programmed such that, in each cell type, demethylation is preferentially directed to regulatory elements that affect the genes specifically expressed in that particular cell (Hon et al., 2013; Ziller et al., 2013). The removal of methyl groups is evidently initiated and carried out by specialized trans-acting factors that have ‘pioneer’ function, being able to recognize specific motifs within the element even though they are packaged within a closed and presumably inaccessible, chromatin structure (Cirillo et al., 2002; Decker et al., 2009; Martínez de Paz and Ausió, 2016; Zaret, 2020; Zhang et al., 2018). These same factors also bring about recruitment of the molecular machinery (e.g. TET DNA demethylases) needed to carry out local demethylation and it is this event that allows these regulatory regions to be stably accessible (Guilhamon et al., 2013; Mayran et al., 2018; Reizel et al., 2021; Vanzan et al., 2021). Indeed, in cells depleted for TET enzymes, biochemical demethylation does not take place and, as a result, target genes are not properly regulated (Orlanski et al., 2016) even though this treatment does not affect the levels of trans-acting factors that drive this process (Reizel et al., 2018).

In order to define better the role of demethylation in opening up non-promoter regulatory regions, it is worthwhile examining the kinetics of this process. It has already been demonstrated, for example, that the mere binding of pioneer factors is itself sufficient to bring about changes in local chromatin structure. Over-expression of FoxA, for example, has been shown to cause a rapid increase in chromatin accessibility (Donaghey et al., 2018), partly because high concentrations of this transcription factor enable it to overcome the chemical barriers inherent to this reaction. During normal liver development, FoxA is indeed present at very high levels in hepatoblasts where it is required for enhancer activation (Nicetto et al., 2019; Reizel et al., 2020), bringing about changes in chromatin structure as well as inducing demethylation. This state is then autonomously maintained and is no longer dependent to the same degree on the original transient factors involved in its establishment (Mayran et al., 2018; Reizel et al., 2018; Reizel et al., 2021). It thus appears that the main role of tissue-specific demethylation is not necessarily to open up regulatory regions, but rather to help stabilize and thus maintain their accessible chromatin structure and, in this way, ensure long-lasting cell type identity.

Very early experiments in tissue culture have also supported a role for DNA methylation in regulating chromatin accessibility. For example, it was shown that the presence of methyl groups on DNA has a strong effect on chromatin structure, making it resistant to DNaseI digestion in a manner that is completely independent of sequence content (Keshet et al., 1986). It has also been demonstrated that, in vivo, methylation operates by limiting local accessibility to factors in the surrounding nucleoplasm, and this is probably accomplished through a number of different molecular mechanisms, including both histone modification as well as higher order structural changes (Blattler et al., 2014; Hashimshony et al., 2003; Jones and Wolffe, 1999; Siegfried et al., 1999). This important concept was recently confirmed by studies on the regulatory element demethylation that occurs during normal development and tissue adaptation, where it was shown by ATAC-seq that these sites are in an inaccessible conformation in most cell types but become accessible following demethylation. Furthermore, when demethylation is prevented by virally induced TET depletion, these sites remain in their closed structure and protein factors are no longer able to bind to their specific motifs within these elements (Reizel et al., 2018).

It therefore appears that DNA methylation stands at the foundation of a sophisticated system for controlling the use of regulatory sequences in a tissue- or stage-specific manner while keeping them inactive in other tissues. Because many, if not most, of the binding motifs within these elements are actually target sites for transcription factors that are expressed ubiquitously in many different cell types (Sonawane et al., 2017), it is to a large extent the presence of methylation that prevents factor accessibility and, in this way, serves as a global repression mechanism (Reizel et al., 2018). Thus, although gene-promoter activation may be highly dependent on the presence of tissue-specific transcription factors, distal regulatory sequences are opened up by targeted epigenetic drivers that bring about site-specific local demethylation, thus stabilizing the inaccessible chromatin state and enabling a wide range of cis-trans interactions that can modulate the expression of associated genes (Zaret and Mango, 2016). One of the powerful aspects of this regulatory scheme (Fig. 1) is that TET-mediated demethylation takes place in a regional manner, covering about 500 nucleotides with multiple motifs, thus serving as a way to amplify the epigenetic effect and better ensure stability (Reizel et al., 2021; Vanzan et al., 2021).

This general concept (Fig. 1) has actually been validated by bioinformatic analyses of ENCODE data comparing tissue-specific demethylation tiles and transcription factor binding in a large number of different cell types in mice. For any factor, including those that are expressed in many different cell types, binding is observed only within tissue-specific undermethylated sites but not at these same exact loci in other tissues, where they are methylated (Reizel et al., 2018). This picture is also consistent with data suggesting that non-specific factors actually play an important role in tissue-specific gene expression (Sonawane et al., 2017).

It should be noted that a considerable percentage of the non-promoter regulatory sequences that influence local gene expression are actually not enhancers but rather are silencer elements (Doni Jayavelu et al., 2020; Huang et al., 2019; Ngan et al., 2020; Pang and Snyder, 2020). Genomic sequences of this nature can be easily identified by searching for peaks of chromatin accessibility (e.g. via ATAC-seq) that lack the usual markers typical of enhancers, such as H3K4me1. Many of these elements are enriched in repressor motifs such as REST, USF1 or BATF and, when tested in a reporter system, have been shown to inhibit gene expression in cis, both in vitro and in vivo (Doni Jayavelu et al., 2020; Huang et al., 2019). Like enhancers, these elements appear to be utilized in a tissue-specific manner. Indeed, by analyzing ChIP-seq data from many different cell types (ENCODE), it is possible to verify that these presumed silencers actually bind to repression factors in vivo, but only in cell types in which they are demethylated (Edrei et al., 2021 preprint; Reizel et al., 2018); in some cases, it has been shown that methylation of these sites can actually counter gene repression (Edrei et al., 2021 preprint). Given that many of these trans-acting repressors are ubiquitously expressed in all tissues, it appears that, as occurs with enhancers, the cell-type specificity of silencer action is determined by developmentally programmed demethylation. Thus, the general strategy for potentiating non-promoter regulatory sequences, whether they be enhancers or silencers, appears to employ pre-programmed tissue-specific demethylation as a means of opening up these regions, thereby making them accessible to trans-acting factors.

Because protein-DNA interactions are dependent on the local concentrations of components, the ability to recruit multiple regulatory elements to within close 3D topological reach of the promoter greatly enhances the effectiveness of gene regulation (Pennacchio et al., 2013). DNA methylation programming plays an important part in this scheme, primarily because it enables cell type-specific accessibility to a wide network of cis-acting factor-binding sites. Furthermore, once this available chromatin configuration is established, it will then be maintained even after the original pioneer factors are no longer present, with the new methylation pattern itself thus serving as a template of stable cell identity (Mayran et al., 2018; Reizel et al., 2021).

The process of cell fate determination during development is largely directed by local morphogens, which serve as developmental signals to transduce programmed changes in methylation into stable patterns of cell type-specific gene regulation. It now appears that this same built-in molecular concept may also underlie the process of tissue adaptability to environmental changes that take place during life. It has been demonstrated, for example, that the fixed methylation pattern observed in adult differentiated cells is not totally derived from events that occur during differentiation in the embryo, and many demethylated sites are actually generated postnatally as a result of a dynamically changing biochemical environment. For example, many specific demethylation events occur in mouse hepatocyte DNA within the first 6 weeks after birth, and these most likely reflect a biological response resulting from the sudden transition to a fat-rich nursing diet and then, 3 weeks later, from the introduction of a chow diet at the time of weaning. A high percentage of these methylation changes have been shown to require insulin signaling (Reizel et al., 2015, 2018). Sexual maturation in the male also brings about the specific demethylation of regulatory elements within a number of different tissues in a process that is dependent on testosterone (Reizel et al., 2015), and similar epigenetic changes have also been attributed to pregnancy (Dos Santos et al., 2015). It thus appears that the genomic strategy of gene regulation used during embryonic development has been co-opted for the process of tissue adaptation, with hormone signaling, as opposed to morphogens, now serving as the mediator of environmental challenges (Dor and Cedar, 2018).

Unicellular organisms, such as bacteria or yeast, usually transcribe almost their entire genome, with only a few genes undergoing programmed regulation by specific transcription factors, mainly repressors. However, in higher organisms with large genomes, the orchestration of gene expression during the development and maturation of each and every cell type is much more complex and cannot easily be carried out using this same simple regulatory strategy, especially considering that a large fraction of the genome needs to be repressed in each cell. The system that has evolved in higher organisms provides an elegant solution to this problem by employing DNA methylation as a global inhibitor of transcription, and this may represent the crux of epigenetic regulation.

There are two fundamental aspects of this system that make it a unique global regulator: its mode of establishment and its molecular mechanism. Methylation-mediated gene repression is set up in the early embryo in a process that is essentially non-specific and thus includes all regulatory elements that are not marked for protection. Because methyl groups are then autonomously copied through cell division, this pattern is preserved in every cell type throughout the lifetime of the organism. Thus, this global repression template is faithfully maintained without the need for any specific regulatory factors (Cedar and Bergman, 2012; Reizel et al., 2018).

Another important component inherent in this scheme is the mechanism by which DNA methylation affects gene expression. Methyl groups operate over a small surrounding region by affecting local chromatin structure, independent of the underlying sequence, and this thus restricts factor accessibility in a global manner (Reizel et al., 2018). Given that almost all protein-DNA interactions are dependent on accessibility, this usually represents the rate-limiting step in most regulatory processes. It should be noted that chromatin structure per se actually undergoes extensive disruption during each passage of the replication machinery in S phase. It is very likely that it is the rapid copying of underlying methylation patterns following DNA synthesis (Gruenbaum et al., 1983) that plays a dominant role in enabling proper chromatin repackaging and, in this way, serves as a key mechanism for preserving the genome accessibility profiles that direct gene regulation.