Epigenetic regulations through DNA methylation and hydroxymethylation: clues for early pregnancy in decidualization.

In mammals with hemochorial placenta, including humans and rodents, endometrial stroma cells extensively proliferate and differentiate into decidual cells under the direction of signals released by invading embryos after implantation ( 63 ). Differentiated decidual cells show huge changes in the transcriptome compared with non-differentiated counterparts ( 69 ). Once pregnancy is terminated, the uterus recovers to an unpregnant state and reenters menstrual cycles after shedding and degeneration of decidual residues and regeneration of the epithelium. It is plausible to hypothesize that DNA methylation plays a role in regulating gene transcription during such a highly dynamic process. Although the transcriptional changes have been studied vigorously ( 69 ), the understanding of epigenetic regulation on this just starts. In an in vitro model of decidualization using estrogen, progesterone, cAMP or a different combination of treated primary or cloned human endometrial stroma cells, DNMT s exhibit a varied expression pattern with mostly transient or sustained decrease ( 59 , 60 , 70 , 71 ), except DNMT1 shows temporal upregulation in the early stage ( 61 , 70 ). The discrepancy between studies could derive from different cell sources, treatment recipes, and other experimental variance ( 62 , 70 ). DNA methylation inhibitor 5-aza-dC induces certain cell morphological and molecular changes similar with decidualized cells treated by hormonal recipe; however, these effects are not exactly the same as induced by hormones, which indicates some other effects independent of its demethylation activity ( 72 ). In a similar in vitro model, no obvious change of global DNA methylation after induced decidualization is detected by another group ( 71 ). Consequently, without a gene specific survey, those observations cannot reveal the real change of DNA methylation during decidualization. In order to screen differentially methylated genes during the progression of decidualization, we profiled the change of DNA methylation for the first time through a non-biased approach, using a methylation-sensitive restriction fingerprinting (MSRF) technique, in an experimentally induced decidualization model in mice ( 67 ). Multiple genes are identified showing hyper- or hypo-methylation in regulatory elements and correlate well with their down- or upregulated expression in the decidual horn compared with an undifferentiated counterpart, clearly showing changes of DNA methylation during this biological event. Enhanced expression of DNMT1 and DNMT3A in decidua, together with the fact that 5-aza-dC significantly prohibits the maintenance of decidualization and aberrantly upregulates two hyper-methylated genes ( Bcl3 and Slc16a3 ), strongly supports an essential role of this epigenetic regulation for successful postimplantation decidual development ( Figure 4 ). The negative effects of 5-aza-dC on decidualization is also supported by another study using a mouse model ( 68 ). Recently, appreciable differential methylation on a single chromatin is uncovered comparing fertile and infertile decidua from genetically different mice ( 73 ). There is considerable discrepancy between the aforementioned studies in humans and mice on regulation of DNMTs expression and the effects of DNA methylation inhibitor. Several factors are probably involved in the divergence, such as incomplete reflection of in vitro conditions for in vivo physiology, difference of species [decidualization is only induced during pregnancy in mice, but occurs cyclically in human during the secretory phase without the presence of embryos, implicating distinctly a potential role in acquiring receptivity ( 63 , 74 )] and differential regulation of enzymes [DNMT3B is barely detectable in the mouse uterus ( 67 ), but important in human decidualized stroma cells ( 70 )]. In spite of the discovery of a set of genes with differential methylation in our study, the number of those genes is relatively low, which may be caused by low resolution and other limitations of MSRF strategy. Interestingly, chromobox 4 ( CBX4) is one of the confirmed differentially methylated genes and is highly inducible in decidua ( 67 ). CBX4 is known as a component of polycomb repressive complex 1 (PRC1), which is known as epigenetic regulatory machinery for gene silencing. Through PRC1-dependent and -independent mechanisms, CBX4 participates in the regulation of cell proliferation, differentiation, and senescence ( 75 ), which are also actively involved during decidualization ( 63 ). PRC1 has diverse compositional variations and ability to perform transcriptional suppression dependent or independent of PRC2 catalyzed H3K27me3 ( 76 ). In respect to histone modification, decline of H3K27me3 and increase of acetyl-H3 and H4 at promoters of decidual marker genes PRL and IGFBP1 have been shown to involve chromatin remodeling in the in vitro human endometrial decidualization model ( 77 ). Strong evidence in vivo has also found that H3K27me3 mediates silencing of inflammatory chemokine genes in decidua to adapt the immune tolerance at the feto-maternal interface ( 78 ). CBX4 has also been found to modulate stability and activity of DNMT3A as a SUMO-E3 ligase ( 79 ), which may link it back to the regulation of DNA methylation locally, as both of them are inducible in the decidual bed. It will be very interesting to explore the function of CBX4 and PRCs and to find the relationship between DNA methylation and histone modification in decidual transformation.

Epigenetic regulation in its strictly modern concept is defined as stably heritable changes in a chromosome without alterations to the DNA sequence ( 1 ). Epigenetic processes mainly include DNA methylation, histone modification, nucleosome positioning, and non-coding RNA. Heredity is transmission between dividing cells or between generations of an organism. However, dynamic changes of classical epigenetic markers for controlling differential gene expression is still loosely described as ‘epigenetic’ regulation, although they may not always fulfill the strict ‘heredity’ definition. DNA methylation is one of the best-studied epigenetic phenomena in plants, fungi, and animals, and our understanding of it has been accelerated recently by the rapid development of ‘next generation sequencing’ techniques. In the past several years, the rediscovery of hydroxymethylation and its catalytic enzymes has started to bridge our understanding of methylation and demethylation, and revealed more dynamic changes than previously known, which, in turn, has tremendously expanded our knowledge of DNA methylome. In this review, we will discuss recent progress on studying DNA methylation and hydroxymethylation in mammalian epigenetic regulation and its relevance in the reproductive system, particularly in pregnant uteri that experience highly tempo-spatial activation and silencing of genes at a high rate.

The TET proteins have three members (TET1, TET2, and TET3) and are comprised of 2-oxoglutarate and Fe(II)-dependent dioxygenases family, which are responsible for creating 5-hydroxymethylcytosine (5hmC) ( 24 , 25 ). The TET1 protein was first found capable of catalyzing the conversion of 5mC into 5hmC ( 25 ). Similar to DNMTs, TET1 and TET3 have the DNA-binding domain CXXC to recognize CpG sites. TET2 lacks this domain; however, as an ancestral TET2 protein, the CXXC domain, IDAX helps to recruit TET2 to target genes and regulate its stability ( 25 ). Pre-existing 5mC is necessary for 5hmC production, as 5hmC is eradicated in Dnmt1/3a/3b triple mutant ESCs, which is absent of 5mC ( 26 ). Whether 5hmC is faithfully maintained during DNA replication by TET proteins as DNMT1 does for 5mC still awaits further exploration. Expression levels of TET1 and TET2 are high in ESCs, but decrease dramatically once differentiation occurs. Meanwhile, TET3 shows an inverse pattern that suggests the distinct functions of TET1, TET2, and TET3 ( 25 ). TET1 and TET2 involve regulating pluripotency in stem cells through a possibly reciprocal feedback process. TET1 and TET2 control promoter hypomethylation of pluripotency-maintaining genes ( Nanog , Esrrb, etc.) to maintain their expression in ESCs ( 24 ). Conversely, OCT4/SOX2 complex directly regulates TET1 and TET2 transcription ( 25 ). Knockdown of TET1 in ESCs causes skewed differentiation into the endoderm-mesoderm lineage and a bias towards trophectoderm differentiation ( 25 , 27 ). In addition to hydroxylase activity, TET1 also controls DNA methylation by binding to CpG-rich regions to prevent unwanted DNA methyltransferase activity, which imparts differential maintenance of unmethylated state at TET1 targets and potentially works as a guardian for epigenetic fidelity ( 28 ). An unexpected role in transcriptional repression of TET1 independent of its catalytic activity is uncovered while being associated with the SIN3A co-repressor complex or being co-recruited with the PRC2 complex at overlapping target genes ( 24 , 26 ). The aforementioned dual roles of TET1 are supported by TET1 being enriched at genes with either H3K4me3 monovalent or H3K4me3/ (trimethylated histone 3 lysine3 27) H3K27me3 bivalent modifications ( 26 , 28 ). However, the importance of TET1 on pluripotency in ESCs is still controversial, as Tet1 mutant mouse can survive for postnatal development ( 27 ). Another in vivo role of TET1 is its specific regulation on demethylation and the activation of meiotic genes in the germ line, indicated by observations of reduced oocytes production and subfertility in Tet1 mutant female mice ( 24 ). Not surprisingly, the three members of TET proteins may have some redundant functions. Tet1 and Tet2 double mutants have more pronounced defects in the female germ line than Tet1 mutants, and upregulated Tet3 may compensate for their roles in the maintenance of ESC pluripotency ( 27 ). Beyond potential roles in ESCs, TET2 controls 5hmC accumulation at regulatory regions of genes in the fetal brain, which will be demethylated and activated later on for memory formation towards adulthood ( 23 ). TET2 is also critical for hematopoiesis and mutation of TET2 relates to decreased 5hmC during myeloid tumorigenesis ( 25 ). TET3 is highly expressed in oocytes and zygotes to control the rapid conversion of 5mC to 5hmC, which may be responsible for immediate demethylation of paternal pronucleus upon fertilization ( 25 ). Techniques based on methylation-sensitive restriction enzymes and sodium bisulfite treatment of DNA are incapable of distinguishing 5mC and 5hmC. Therefore, several chemical/enzymatic modification based methods were developed to selectively convert 5hmC. Conjugation of those techniques with bisulfite sequencing makes the base-resolution mapping of 5hmC possible, which has revealed more accurate global distribution of 5hmC than affinity based enrichment and sequencing methods ( 24 , 29 ). Compared to the relatively constant level of 5mC, 5hmC shows more variability between different cell and tissue types. The brain has a much higher portion of 5hmC than other organs ( 23 ). ESCs has relatively lower ratio of 5hmC. 5hmC (~0.03% of all nucleosides) has far lower abundance compared to 5mC (~0.8%) in the genome of mouse ESC ( 30 ). In contrast with high abundance of 5mC in non-CpG context, 5hmC mostly shows in CpG sites in ESCs ( 29 ) and central neuron system ( 23 ). The high level of asymmetric distribution of hydroxymethylation is another incomprehensible feature of 5hmC, however, this observation needs confirmation by enhanced depth of sequencing due to its low abundance ( 29 ). 5hmC is mostly associated with euchromatin containing actively transcribed genes. The balance between 5hmC and 5mC is different between genomic regions ( 24 ). As indicated in Figure 2 , 5hmC is highly enriched in distal-regulatory elements, which includes enhancers, insulators, p300-binding sites and DNase I hypersensitive sites, in human and mouse ESCs ( 21 , 29 ). The vast majority of repetitive elements are highly enriched with 5mC but not 5hmC ( 29 ). In those regions, distribution of 5hmC shows reverse correlation with 5mC. 5hmC is also distributed in the exons of gene bodies as 5mC but is mostly depleted in CGIs ( 25 , 26 ). In promoters, 5hmC is mostly enriched in those with low to moderate CpG contents ( 29 ), while most affinity based methods found that 5hmC is enriched in CGIs close to TSS and highly related with active transcription ( 26 , 28 ), which, indicating technical bias, may affect the interpretation of data. 5hmC is especially enriched at the start sites of genes whose promoters bear H3K4me3/H3K27me3 bivalent marks, suggesting a model in which 5hmC contributes to the ‘poised’ chromatin signature found at developmentally-regulated genes in ESCs (25,329). Enrichment of 5hmC is also correlated with the monomethylated histone H3 lysine 4 (H3K4me1) markers at the poised enhancers, but not with the acetylated histone H3 lysine 27 (H3K27Ac) markers at active enhancers, indicating a poised epigenetic state or demethylation intermediate ( 24 ).

The mechanism of DNA hydroxymethylation remodeling epigenetic signature is still under vigorous investigation and many novel regulatory machineries have been unraveled. 5hmC may change the chromatin structure by precluding the methyl-binding proteins that only have affinity for 5mC, therefore, disrupting binding of the transcription regulatory machinery after conversion ( 31 ). Conversely, 5hmC sequence prefers to be bound by specific coregulatory proteins, such as chromatin modifiers MBD3 and BRG1, to uniquely control gene de-repression event ( 32 ). 5hmC may also serve as an intermediate product for DNA demethylation. As no direct demethylase has been found so far, indirect demethylation through passive or active pathways occurs. As shown in Figure 3 , in the model of passive DNA demethylation, after 5mC conversion to 5hmC in template strands, large amounts of newly synthesized chains become unmethylated during rapid division of cells, for example, the maternal genome loses methylation gradually for the purpose of reprogramming in early embryos from the zygote to blastocyst stages ( 25 ). Meanwhile, in the model of active DNA demethylation, several pathways possibly mediate this process. First, the activation-induced deaminase (AID)/apolipoprotein B mRNA-editing enzyme complex (APOBEC) family of cytidine deaminases converts 5mC and 5hmC into thymidine (T) and 5-hydroxymethyluracil (5hmU), respectively, followed by thymine-DNA glycosylase (TDG)-mediated T or 5hmU base excision, and finally ended with unmethylated cytosines by downstream repair ( 33 ). TDG is necessary for active demethylation of tissue-specific promoters and enhancers that are developmentally and hormonally regulated ( 33 ). AID/APOBEC assisted demethylation has been uncovered in primordial germ cells (PGCs) and slowly dividing brain cells ( 34 ). Second, TET proteins can generate 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) from 5mC or successively from 5hmC ( 35 ). 5fC and 5caC are excised and repaired to regenerate unmodified cytosines by mammalian TDG and the base excision repair (BER) pathway ( 35 , 36 ). Finally, another possible player of demethylation could be DNMT3A/B, the same enzymes for de novo DNA methylation, which are characterized with 5hmC dehydroxymethylases activity in vitro ( 37 ). Related to this, an earlier finding about hormonal controlled cyclic methylation implies that DNMT3A/B may have deaminase activity for rapid demethylation ( 38 ). Global demethylation and remethylation, processes known as epigenetic reprogramming, only occur twice in the life-cycle of mammals and are under tight regulation by TET family members via mediating conversion to 5hmC. The first occurrence of reprogramming arises soon after fertilization, but it is asymmetrical between the parental genomes found in preimplantation embryos. Paternal DNA loses methylation rapidly after fertilization and before the first division of the zygote, whereas maternal DNA is gradually demethylated during embryonic cleavage ( 24 , 39 ). Nonetheless, most of the imprinting control regions (ICRs), and a few other regions, which are differently methylated between male and female gametes, escape from this global erasure to assure proper transmission of allele specific imprints between generations ( 10 ). During this process, developmental pluripotency-associated protein 3 (DPPA3) protects the maternal chromatin and certain paternal imprinted loci from TET3-mediated conversion of 5mC to 5hmC via local binding to harbored H3K9me2 ( 40 ). The second lifetime occurrence of reprogramming happens during PGCs expansion, migration, and entry into the gonads, which involves TET1 and TET2 driven transformation to 5hmC ( 34 ). Similar to the first wave, there is extensive loci correspondent to the intra-cisternal-A-particles (IAPs), but not imprinted loci, showing a resistance to global demethylation theorized to be for protecting genomic integrity in PGCs ( 41 ).

DNA methylation, as a well-established epigenetic marker, has attracted the efforts of biomedical researchers from diverse fields for several decades. Combining our new knowledge of DNA methylome with histone modification and non-coding RNAs will provide important insights into the mechanisms of chromatin remodeling and regulation of gene transcription, which is instrumental to comprehensive understanding of developmental reprogramming and disease formation. The pregnant uterus provides a good model for exploring the relationship between transient epigenetic regulation and transcription, due to its highly dynamic but reversible molecular and cellular changes during pregnancy or the menstrual cycle. Although emerging evidence suggests that epigenetic regulation is critical to finely tune multiple biological events in the uterus during pregnancy, limited progress has been made towards fully understanding these mechanisms, especially DNA methylation (62,373). Heterogeneous composition of cell types, differential regulation of compartments, and difficulty of recapitulation via in vitro models makes the study of pregnant uteri biologically and technically challenging. However, rapidly developing next-generation sequencing techniques and DNA chemical modifying approaches allow for single-base resolution mapping of methylation and other newly found modifications on DNA. Applying those powerful technical tools and the appropriate experimental designs will help us to acquire a high-resolution view of epigenetic landscapes for the dynamically changing uterus. Integration of newly identified epigenetic information with data of transcriptome and proteome will advance our knowledge of the molecular mechanism behind uterine implantation and decidualization during early pregnancy.