Definition

A non-coding RNA (ncRNA) is defined as an RNA molecule that is not translated into a protein. Classes of non-coding RNAs include transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs), small RNAs such as microRNAs (miRNAs), siRNAs, piRNAs, snoRNAs, snRNAs, exRNAs, scaRNAs and the long ncRNAs [88]. The role of these molecules in placental development, physiology and pathology has been recently reviewed in detail [89]. Here we will discuss solely the role of miRNAs and long ncRNAs in the epigenetic control of placental development.

### MiRNA and Normal Human Placental Development

The miRNAs are single stranded RNA molecules of 19–24 nucleotides, which act primarily by degrading mRNA transcripts or inhibiting translation of miRNA in to proteins [90]. To date, more than 2000 human miRNAs have been discovered, which appear to regulate 50% of human RNAs [91]. A large number of miRNAs detected in the placenta are expressed from a gene cluster located on chromosome 19 (C19MC) [92,93]. This cluster includes 46 intronic miRNA genes that express 58 miRNA species. These miRNAs are primate-specific, and they are expressed almost exclusively in the placenta (and are thus termed trophomiRs). In the human placenta, the expression of C19MC miRNAs is detected as early as 5 weeks of pregnancy and the expression gradually increases as pregnancy progresses [94]. An imprinted, paternally expressed, CpG-rich domain has a regulatory role in C19MC expression [95]. This DMR, is hypermethylated in cell lines that do not express C19MCs [96]. The C19MC region contains genomic transposable elements called "Alu repeats", which have been implicated in recombination and gene duplication events. Because of their sequence complementarity it has been proposed that several C19MC miRNAs could be responsible of the targeting and degradation of transcribed Alu elements. Also, the C19MC miRNAs are expressed in embryonic and in stem cells but their expression drops considerably when these cells di fferentiate, which may indicate a role in the maintenance of an undi fferentiated state [97–101]. Several members of the C19MC cluster are expressed at much higher levels in vCT compared with EVTs and overexpression of the C19MC cluster results in reduced migration of the extravillous trophoblast line HTR8/SVneo [45]. The chromosome 14 miRNA cluster (C14MC) is another miRNA cluster that is expressed in the placenta [102]. This cluster includes the miRNAs: miR-127, miR-345, miR-370, miR-431 and miR-665. These miRNAs have been involved in the regulation of the immune suppressive, anti-inflammatory response and also in the regulation of the ischemia/hypoxia response [103]. The expression of the C14MC members generally declines during pregnancy [104].

The miR-675 is expressed from the first exon of the H19 long non-coding RNA. Up-regulation of miR-675, which is controlled by the stress-response RNA-binding protein HuR, restricts murine placental growth. Deficiency of H19, promotes placental growth and miR-675 overexpression decreases cell proliferation, likely through targeting Igf1R [105]. Consistent with these findings, the expression of miR-675 rises toward the end of murine pregnancy, when placental growth decelerates. In addition, miR-675 restricts proliferation in JEG3 cells, likely through binding to the nodal modulator 1 (NOMO1) protein [44].

Several other miRNAs are likely involved in placental development by inhibiting genes associated to regulation of trophoblast fate, invasion and proliferation (Let-7a, miR-377, miR-145, members of the miR-17\_92 cluster, members of the miR-106a\_363 and miR-106b\_25 clusters, miR-155, miR-34, miR-141-3p and miR- 200a-3p) [106,107]. As additional examples of regulation, mir-431inhibits invasion of trophoblast cells by targeting the ZEB1 gene [108], miR-106a~303 inhibits trophoblast di fferentiation by targeting hCYP19A1 and hGCM1 [43], miR-34 targets SERPINA3, a key gene in a variety of biological processes and highly deregulated in placental diseases [41].

These miRNAs regulate diverse processes such as trophoblast physiology, proliferation and invasion (some mentioned in Table 1 and reviewed in Reference [107]).

lncRNA and Normal Human Placental Development

Long non-coding RNAs (lncRNAs) are RNAs greater than 200 nucleotides in length that do not encode a protein product. They are expressed with cellular and temporal specificity and have been involved in many cellular events, including the regulation of gene expression, post-transcriptional modifications and epigenetic modifications, imprinting and X-chromosome inactivation [109]. They act as sca ffolds (binding other RNAs or proteins), signals and antisense decoys and engage in transcriptional interference. Usually a single lncRNA has multiple functions. The function of lncRNAs in placental development is poorly understood, mostly inferred from studies on placental pathologies. Nevertheless, lncRNAs have been involved in a number of critical trophoblast functions, from proliferation, invasion and migration, to cell cycle progression [110]. H19 was one of the first lncRNAs to be discovered [111]. H19 is located within a large imprinted domain on chromosome 11, at ~100 kb downstream of IGF2. H19 and IGF2 are reciprocally imprinted that is, for H19 only the maternal allele is expressed, while for IGF2, only the paternal allele is expressed [112]. H19 expression could be regulated by PLAGL1, a zinc finger transcription factor, in the human placenta [113]. Two major functions have been described for H19, specifically as a modulator for binding small RNAs and proteins [114] and as a source of the miRNA mir-675 (see above). H19 has variable levels of biallelic expression in the placenta (reports sugges<sup>t</sup> between 9% and 25% expression occurs from the imprinted allele) until 10 weeks of gestation by which time H19 expression is mostly restricted to the maternal allele [115]. H19 expression is restricted to intermediate and vCT and is not found within SCTs in the human placenta. H19 down-regulation in trophoblast cells leads to inhibition of proliferation and apoptosis [116]. Many other lncRNAs have been involved in placental development, including lincRNA SPRY4-IT1, MIR503HG, LINC00629, MEG3, MALAT1, RPAIN and TUG1 [31–37]. The study of the expression of these lncRNAs during placental development and the manipulation of their expression in vitro in choriocarcinoma cell made it possible to infer their possible function in the context of placental development (Table 1).

#### 2.3.3. Histone Modifications in the Developing Placenta

Histone modification is the process of modification of histone proteins by enzymes, including post-translational modifications, such as methylation, acetylation, phosphorylation and ubiquitination. Histone modifications participate in gene expression regulation by modulating the degree of chromatin compaction [117].

Our knowledge concerning the role of histones modification in human placentation is scarce and refers mostly to studies in mice. Methylation frequently occurs on histones H3 and H4 on specific lysine (K) and arginine (A) residues. Histone lysine methylation can lead to activation or to inhibition, depending on the position in which it is located. For instance, H3K9, H3K27 and H4K20 are considered as important 'inactivation' markers, that is, repressive marks, because of the relationship between these methylations and heterochromatin formation. However, the methylation of H3K4 and H3K36 are considered to be 'activation' marks [118,119].

The heterochromatin methylation marker H3K27me3 was found to be highly active in vCT. That was explained by rapid and transient repression of genes at the time of SCT formation. SCTs nuclei were also found enriched for H4K20me3 [30]. However, this report contrasted with another study reporting that the CTs were enriched with H3K4me3 and that the SCTs were transcriptionally activated by the chromatin marker H3K4me2, which co-localized with active RNAP II in the majority of SCT nuclei [29]. In mouse and other mammals, H3 arginine methylation predisposes blastomeres to contribute to the pluripotent cells of the ICM, which appears to require higher global levels of H3 arginine methylation than the TE/trophoblast lineage [120]. Nevertheless, these lower modification levels in the trophoblast lineage are indispensable for normal placental development.

Acetylation, which in most cases occurs in the N-terminal conserved lysine residues, is also an important way to modify the histone proteins, for example, acetylations of lysine residues 9 and 14 of histone H3 and of lysines 5, 8, 12 and 16 of histone H4 by Histone Acetylases (HATs). Acetylation is generally associated with the activation or opening of the chromatin. On the contrary, de-acetylation of the lysine residues by histone deacetylases (HDACs) leads to chromatin condensation and inactivation of gene transcription. Oxygen (O2) concentrations strongly influence placental development partially through modifications of the histone methylation codes. Initially, the gestation environment is hypoxic and O2 concentration increases during development. Hypoxia-inducible factor-1 (HIF-1), consisting of HIF-1α and ARNT subunits, activates many genes involved in the cellular response to O2 deprivation [121]. HIF-1 is also known to recruit and regulate HDACs [122,123]. Moreover, HIF-1 has been found to bind specific sites on the promoter of the H3K9 demethylases thereby inducing their expression. In particular, it induces JMJD1A and JMJD2A that remove dimethyl marks on H3K9me2, JMJD2B [124,125] which removes trimethyl marks (H3K9me3) and more weakly JMJD2C which converts H3K9me3 to me2 [126]. Studies in rodents have shown that HIFs have important roles in the regulation of TSCs differentiation by integrating physiological, transcriptional and epigenetic inputs. Thus, the crosstalk between HIF and the HDACs is required for normal trophoblast differentiation [123,127].

Another example of histone modification during placentation, is the acetylation of histones H2A and H2B by the CREB-binding protein (CBP). CBP acts as an acetyltransferase that decreases the EMT and invasiveness of murine TSCs while maintaining the properties of stem cells [28].

Trophoblastic fusion depends on the regulation of GCMa activity by HATs and HDACs. Human GCMa transcription factor regulates expression of syncytin, which in turn mediates trophoblastic fusion. It has been demonstrated that CBP-mediated GCMa acetylation underlies the activated cAMP/PKA signaling pathway that stimulates trophoblastic fusion [27]. Human pregnancy-specific glycoproteins (PSG) are the major secreted placental proteins expressed by the SCTs and represent early markers of cytotrophoblast differentiation. Pharmacological inhibition of HDACs in JEG-3 cells up-regulated PSG protein and mRNA expression levels. This correlated with an increase in the amount of acetylated histone H3 associated with PSG promoter [26]. Combined acetylation at H3K9 and H3K4 methylation also activates Maspin, a tumor suppressor gene which is negatively correlated with human trophoblasts motility and invasion [24,25]. The invasive capacity exhibited by EVTs is attributed in part to the extracellular matrix degradation mediated by matrix metalloproteinases (MMPs) such as MMP-2 and MMP-9. Differential expression of these MMPs and their tissue inhibitors (TIMPs) has been associated to histone H3K9/27me3 [23].

### 2.3.4. Imprinting and Placental Development

### Placentation and the Materno-Fetal Conflict

Pregnancy in Eutherian mammals is an immunological challenge as reviewed recently [128]. To note, an ancestral inflammatory response in pregnancy and parturition also exist in marsupials (metatherians), as recently observed [129,130]. Other mechanisms are equally conserved in the formation of the placenta, in particular the fusion mechanisms of cytotrophoblasts into syncytiotrophoblasts that are mediated by retroviruses, in eutherians as well as in metatherians [131].

Once the placenta is formed, it will allow nutrients to transit from the mother circulation to the fetal circulation. In the context of the maternal-fetal conflict hypothesis, tightly regulating the placentation process and limiting placental growth is crucial for the mother survival. The genes controlling this regulation are expected to be found different between viviparous and non-viviparous species. For this, mammals appear as an excellent model as a group of ~4500 species divided into egg-laying animals (prototherians, Platypus and Echidnaes, 5 species), animals with a short-lived placenta (metatherians, Marsupials ~250 species) and viviparous species with a long-lived placenta (eutherians, i.e., all the other mammals, where gestation length can be up to 22 months in the African elephant). One major difference found between the genome of placental species and non-placental species of mammals is the presence of imprinted genes only in the first group.

Definition of Imprinted Genes and Links with Viviparity

Imprinted genes are genes that are expressed from either the maternal or the paternal allele, mainly through di fferentially methylation mechanisms. Their existence leads to dramatic phenotypic di fferences in animal hybrids according to the sense of the cross. For instance, interbreeding of lions and tigers results in two morphologically di fferent animals, if the male is the lion or the male is the tiger, leading to a liger or a tigon, respectively [132]. While the tigon has a size like that of its parents, the liger is the largest existing felid (up to >400 kg) and several hypotheses have been raised to explain this fact, mostly connected to the existence of imprinted genes. Experimentally, in the 80s, Solter and Surani carried out nuclear transfer experiments that demonstrated in mice the necessity of a paternal and maternal genome to foster healthy development [133]. Androgenetic embryos lead to the production of a hypertrophic placenta while gynogenetic embryos had a very small placenta and a stunted embryo. Similarly, in humans, development from two paternal genomes leads to hydatiform moles, where the placenta is composed of grapelike vesicles, whereas parthenogenic development leads to the apparition of teratomas [134].

As far as we know today, imprinting is closely associated to viviparity. The sequencing of the platypus genome in 2008 [135] revealed syntenic regions that are relatively well conserved with the eutherian and marsupials, albeit no evidence of imprinted gene can be found in Monotremes. This may be since acquisition of imprinting in a species seems to be associated to the progressive acquisition of CpG islands (besides other mechanisms, such as chromosome translocations or retrotransposons insertions), that appear absent from the platypus genome [136,137]. In marsupials (metatherians), where the placenta is short-lived, the number of imprinted genes is more limited than in eutherian mammals. Two imprinted regions are well conserved between metatherians and eutherians such as the PEG10 and the H19-IGF2 regions [135]. Similarly, an exhaustive analysis of the transcriptome of chicken failed to identify imprinted genes, while allele specific expression does exist [138,139]. The evidence collected therefore strongly links these genes with the placenta presence. Besides, imprinted genes may have a strictly paternal or strictly maternal expression. Series of invalidation experiments in mice indicated that paternal genes tend to increase placental growth while maternal genes tend to limit this growth [140].

#### Example of the H19-IGF2 Cluster; Cross Species Conservation of Imprinted Genes

A well-known example of this is the H19-IGF2 cluster localized distally at 11p15.5 in humans and 7qF5 in mice. In both species, the structure of the locus is conserved (about 100 kilobases separating the two genes, with di fferentially methylated regions inside IGF2 and nearby H19). An IMC (Imprinting Control Region), located 3 kb from the starting point of H19 has also been identified, with seven binding sites for the ZNF transcription factor CTCF. H19 is expressed exclusively form the maternal allele, while IGF2 is expressed from the paternal allele. In mice, a placental specific promoter of Igf2 was discovered. The selective invalidation of this promoter [141] leads to a strong decrease of placental development and placental growth. By contrast, the invalidation of H19, leads to placental and fetal overgrowth [142]. Amongst other imprinted genes that a ffect placental and fetal growth besides H19 and IGF2 are paternally expressed genes, generally identified in mice (Peg1, Peg3, Rasgrf1, Dlk1) and maternally expressed genes (Igf2r, Gnas, Cdkn1c, Grb10).

Interestingly, in mice, the decoy receptor of Igf2, Igf2r is imprinted and with a maternal profile of expression. In humans, surprisingly, the imprinting status of IGF2R seem to be erratic, polymorphically imprinted according to the human individual analyzed. This was first published in 1993 [143] that showed that 2 out of 14 fetuses had an exclusive expression from the maternal allele. Recently it was shown that IGF2R is duly imprinted in macaques [144], showing that even in primates, the imprinting status can vary between relatively close species. Overall, it appears that many placental imprinted mouse genes are biallelic in their expression in humans [145]. Reciprocally, in a study aiming at identifying novel imprinted genes in the human placentas, we compared variants of the placental DNA versus those of cDNAs from the same placentas using SNP microarrays [146,147]. In addition to

four known imprinted genes (IPW, GRB10, INPP5F and ZNF597), we could identify 8 novel imprinted genes in the human placentas (ZFAT, ZFAT-AS, GLIS3, NTM, MAGI2, ZC3H12C, LIN28b and DSCAM). Using a mouse cross allowing the following of the allelic origin, we found an astonishing variegation of the imprinting status: only Magi2 was imprinted in the mouse species.

Imprinted genes may have a general impact on the global methylation status of the placenta. For instance, recently a polymorphism located at the IGF2/H19 locus was shown associated to placental DNA methylation and birth weight in association with Assisted Reproductive Technologies usage [148].

Imprinted genes deregulation in the placenta is linked to placental diseases, as reviewed in References [149,150]. In a recent study, Christians and coworkers, analyzed a list of 120 imprinted genes in relation with global expression of 117 placental samples, including PE and Intra Uterine Growth Restriction (IUGR) cases [151]. The authors identified a significant correlation between birth weight and the expression level of imprinted genes but without significant di fferences between paternally versus maternally expressed genes. Imprinted genes were also more heavily deregulated in preeclampsia than other genes and in this case paternally expressed genes were down-regulated, while maternally expressed genes were up-regulated. The trend was similar for IUGR. Interestingly, the two human-specific microRNA clusters (C19MC and C14MC), both appear to be imprinted (paternally and maternally expressed) for C19MC and C14MC, respectively, clusters that have been duly studied by the team of Yoel Sadovsky [45,89,152]. Recently, we identified duplication in the 19q13.42 imprinted region encompassing the C19MC cluster [153], from a male 26 weeks fetus with severe IUGR, suggesting that a double dose of the miRNA could contribute to the disease. This suggests links between miRNA regulation, imprinting status and the putative consequences for fetal health and growth.

### **3. Epigenetic Alterations in Preeclampsia**

### *3.1. DNA Methylation Alterations in Preeclampsia*

Anomalies of DNA methylation in preeclampsia have been analyzed from di fferent cellular sources. Besides the analysis of placental cells, investigators have analyzed circulating maternal blood cells or cell-free DNA, as well as maternal endothelial cells (much less accessible, though) and cord-blood white blood cells (of fetal origin). A list of genes of which methylation was found altered is presented as Table 3.

A summary of epigenetic mechanisms at work in PE is shown in Figure 2.

#### 3.1.1. Methylation Alterations in the Preeclamptic Placenta

Common Alterations of Gene Expression in PE are Associated to Methylation Alterations

Numerous studies revealed altered expression of various genes in the pathological placentas (as synthesized previously [154]). These alterations of gene expression are partly explained by the existence of epigenetic deregulations. In PE, numerous methylation deregulations have been found in the pathological compared to control placentas, some studies (but not all) taking into account the gestational age, a recurrent issue when normal and pathological placentas are compared, for which there is often a more than 6 weeks di fference [155–160]. The di fferent techniques used to analyze methylation globally are presented in a previous review [161]. These epigenetic changes probably originate from the abnormal placental environment in PE (or IUGR), characterized by alternations of low oxygen tension and hyperoxia. As mentioned above, hypoxia *per se* induces the expression of the Hypoxia-Inducible factor (HIF1 α), which binds to Hypoxia Responsive Element activating the transcription of various genes related with angiogenesis and metastasis-associated genes [162]. Overall, abnormal oxygen signaling in the placental context leads to increased concentrations of Oxygen Reactive Species (ROS) [163]. Oxidative stress may drive an accelerated ageing of trophoblast cells, which could be key to understand the origin of placental disorders. Indeed, several studies

emphasized alterations of telomere length (a mark of ageing) in preeclamptic pregnancies, with a drastic augmentation of short telomeres in PE, especially in Early Onset PE (EOPE) [164–166]. This senescence may be induced by alterations of the managemen<sup>t</sup> of oxidative stress [167–169]. The accelerated transformation of vCTs into SCTs will lead to a decrease life expectancy of the placenta and an alteration of its capacity to bring the gestation harmoniously to its normal term.

**Figure 2.** Overview of the molecular mechanisms at play in preeclampsia. Annotations: eNOS = Endothelial Nitric Oxide Synthase; EZH2 = Enhancer of Zeste Homolog 2; FOXP3 = Forkhead box P3; IGF-1 = Insuline-like Growth Factor 1; ISCU = Iron-sulfur cluster; Lnc-DC = Long non-coding RNA DC; miR-30a-5p = microRNA 30a-5p; miR-155 = micro-RNA 155; miR-210 = microRNA 210; MMP-9 = *Matrix Metalloproteinase-9;* NO = Nitric Oxide; RND3 = Rho Family GTPase 3; ROS = Reactive Oxygen Species; sENG = Soluble endoglin; sFLT1 = Soluble fms-like tyrosine kinase *receptor*-1; STAT3 = Signal transducer and activator of transcription 3; STOX1 = Storkhead Box 1; TET2 = Tet methylcytosine dioxygenase 2; TIMP3 = TIMP Metallopeptidase Inhibitor 3; TUG1 = long non-coding RNA taurine-upregulated gene 1; VEGF = Vascular Endothelial Growth Factor.

It is well known that persisting environmental variations induce changes in the epigenetic marks, including DNA methylation. These marks can either be mere biomarkers or participate actively in regulating genes to overcome the changing environmental conditions (although gene expression changes are often disconnected from methylation alterations).

Overall, several of the genome-wide studies showed that the methylation profiles differ between early and late onset of preeclampsia (EOPE and LOPE), suggesting a different etiology between these two types of PE [170–173]. EOPE shows more pronounced genome-wide hypermethylation changes than LOPE, probably since it is caused by earlier alterations allowing the epigenetic reprogramming to install earlier, in reason of the earlier cellular stress [171,174].

Using the Illumina Methylation 450 BeadChip Array, Yeung and coworkers, identified 303 di fferentially methylated regions in PE, 214 hyper and 89 hypomethylated, after adjusting for gestational age. The genes located nearby or encompassing hypermethylated regions were enriched in gene-ontology (GO) terms such as "ATP transport", in KEGG pathways, such as "steroid hormone biosynthesis", "cellular senescence" and Reactome pathways, such as "Vpr-mediated induction of apoptosis by mitochondrial outer membrane (SLC25A6 and SLC25A4)". The annotation of clusters also revealed an alteration of clusters of homeobox genes, (especially HOXD genes), Wnt2 cell signaling; fertilization and implantation genes; reactive oxygen species signaling (NOX5) and cell adhesion (ALCAM) genes [158]. Amongst the most recent studies, Leavey and coworkers used a novel approach based upon bioinformatics to sort 48 human PE samples through their transcriptome profile before subjecting them to methylation analysis, using the Illumina Human methylation450K array. This made it possible to divide the preeclamptic cases into two groups associated to abnormal methylation marks nearby 'immunological' genes or more 'canonical' EOPE cluster, with for instance abnormally methylated CpG in FLNB, COL17A1, INHBA, SH3PXD2A, as well as in the gene body of FLT1 [160].

In 2015, the study of Zhu and coworkers [175] was the first to analyze simultaneously methylation and hydroxymethylation in the PE placentas. Hydroxymethylation results from the hydroxylation of methyl-Cytosine is a first step towards the active demethylation of DNA through the action of Ten+Eleven Translocation enzyme (TET) proteins, and could play an important role in gene expression regulation [176]. The authors showed that the methylation level is higher in gene promoters and gene bodies in PE versus control placenta. Surprisingly most of the clustering of the genes that were altered, either by methylation or by hydroxymethylation were associated with nervous system development, neurotransmitters, neurogenesis, which are presumably not relevant in a non-neural tissue as the placenta. Nevertheless, positive regulation of vasoconstriction was also enriched as a GO term, as well as regulation of nitrogen compounds, two pathways that have a clear biological sense in terms of placental diseases pathophysiology (association with vascularization and with the modulation of oxidative/nitrosative stresses).


**Table 3.** Di fferentially methylated genes in preeclampsia.


**Table 3.** *Cont.*

The major modifications of methylation occurring in preeclampsia are presented as Figure 3.

Limits of the Genome-Wide, Multicellular Approach for Preeclampsia Methylation Profiling

As mentioned earlier, a recurrent criticism of genome-wide comparisons between normal and PE placenta is linked to the fact that in general placental samples of PE patients are collected at earlier terms than controls. However, the existence of methylation profiles for control placentas throughout gestation [51] now allows to make the part between the effect of the placental ageing and the effects of the pathology per se. Other limits of these approaches are the complexity of the cell material, the variation between the degree of severity of the disease or the various statistical tests that are used in the different studies. Also, several studies have brought attention to the lack of reproducibility in high-throughput genomic, transcriptomic and epigenomic studies. This has been recently discussed by Komwar and coworkers in a recent study were they analyze the sources of variation in preeclampsia high-throughput studies an propose a methodology to ensure reproducibility and thus facilitate the integration of data across studies [198].

**Figure 3.** Overview of major methylation alterations in preeclampsia. The main pathways are shown in green boxes. The significant alterations in methylation may be associated either to increased or decreased gene expression (hypermethylated in red and hypomethylated in blue).

Single-Cell Analysis, the Next Frontier to Methylation Epigenomic Approaches

Gene-scale expression studies were recently carried out at the single cell scale [199] and have provided evidence of gene expression shifts during the CT, SCT and EVT differentiation steps, as well as, allowed the reconstructions of differentiation trajectories. This has also been analyzed by genome-scale DNA methylation analysis. Gamage and coworkers have analyzed by RRBS side population trophoblasts, CTs and EVTs from human first trimester placentas [56]. Forty-one genes involved in EMT and metastatic cancer pathways were found methylated between CT and EVTs, possibly contributing to the invasive phenotype of these cells. In the BeWo cell model where fusion can be induced by forskolin, RRBS analysis performed before and after fusion, showed altered methylation of genes involved in cell differentiation and commitment, together with a gain in transcriptionally active histone marks such as H3K4me3 [54]. Such approaches are for the moment difficult to transpose to placenta pathophysiology. Instead, several systems where methylation influences normal placental function have been studied. As an example, we present below the epigenetic regulation of genes involved in placental invasion and PE.

An Example of Specific Gene Alterations of Methylation: Regulation of Invasion

MMPs are well-characterized proteins involved in trophoblast invasion and angiogenesis during pregnancy. They constitute a family of 23 Zn2<sup>+</sup> and Ca2+-dependent proteases that degrade the extracellular matrix. This family of proteins presents abnormal concentration and behavior in placental diseases such as PE [200], placenta accreta and placenta percreta [201,202]. This has been recently reviewed for preeclampsia [203]. A decreased level of MMP-2 and MMP-9 reduces the remodeling of spiral arteries in early gestation. Besides, other MMPs, such as MMP-1 and MMP-14, may also have a role in this disease. Epigenetic mechanisms are at work for controlling MMP gene expression.

Li and coworkers observed that TET2 is involved in the demethylation of the *MMP-9* promoter, this being associated to the downregulation of the protein and contributing to trophoblast shallow invasion [204].

TIMP3, a MMP inhibitor, shows the highest methylation reduction (over 15%) in EOPE compared to control placentas with an inverse correlation between methylation level and gene expression suggesting an increased transcription of TIMP3 in PE placentas [180,189]. Low levels of TIMP3 lead to poor invasion of the trophoblast and placenta hypoperfusion. Moreover, TIMP3 may be able to inhibit angiogenesis by blocking vascular endothelial growth factor binding to its receptor contributing to impaired placenta blood vessels development. Also, genetic variations of the gene have been associated with cardiovascular disorders and hypertension.

#### 3.1.2. Maternal Blood Epigenetic Marks in Preeclampsia

Alterations in the levels of many plasma and serum proteins have been associated with PE. In 2013, White and coworkers showed that PE was favoring hypermethylation in white blood maternal cells using the methylation-27k arrays from Illumina [205]. GRIN2b. GABRA1. PCDHB7 and BEX1 were found di fferentially methylated, with an enrichment of the neuropeptide signaling pathway. The re-analysis of methylation of genes known to be involved in PE revealed that in maternal circulating leukocytes, CpG sited from 4 genes associated with PE, POMC, AGT, CALCA and DDAH1, showed di fferential methylation in PE compared to control, with moderate methylation di fferences (<6%) [206]. These 4 genes are known to alter immunomodulation and inflammatory response, suggesting that at least alterations of the placental physiology in preeclampsia have epigenomic consequences on maternal circulating cells.

During pregnancy, 3 to 6% of cell-free DNA in the maternal blood plasma is derived from the placenta. Oxidative stress in PE leads to increased trophoblast apoptosis and the release of SCT microparticles and a five to ten-fold increase in circulating fetal DNA in the maternal bloodstream compared with control counterparts [207,208]. These free fetal molecules and their methylation status have been proposed as a non-invasive biomarker of fetal and placental pathologies before the onset of symptoms. This has been shown for Maspin, for which the unmethylated version have a median methylation more than 5.7 fold higher in PE than control pregnancies [209,210]. Another epigenetic marker of preeclampsia is the methylation of RASSF1A (Ras Association domain-containing protein 1) promoter [177,211,212].
