Unique features and emerging in vitro models of human placental development

Abstract Background The placenta is an essential organ for the normal development of mammalian fetuses. Most of our knowledge on the molecular mechanisms of placental development has come from the analyses of mice, especially histopathological examination of knockout mice. Choriocarcinoma and immortalized cell lines have also been used for basic research on the human placenta. However, these cells are quite different from normal trophoblast cells. Methods In this review, we first provide an overview of mouse and human placental development with particular focus on the differences in the anatomy, transcription factor networks, and epigenetic characteristics between these species. Next, we discuss pregnancy complications associated with abnormal placentation. Finally, we introduce emerging in vitro models to study the human placenta, including human trophoblast stem (TS) cells, trophoblast and endometrium organoids, and artificial embryos. Main findings The placental structure and development differ greatly between humans and mice. The recent establishment of human TS cells and trophoblast and endometrial organoids enhances our understanding of the mechanisms underlying human placental development. Conclusion These in vitro models will greatly advance our understanding of human placental development and potentially contribute to the elucidation of the causes of infertility and other pregnancy complications.

is not an immunologically isolated organ, and many lymphocytes reside there. Thus, after implantation, embryos are exposed to maternal immune cells but do not trigger immune rejection. Abnormal placentation is associated with pregnancy complications such as infertility, miscarriage, and hypertensive disorders of pregnancy (HDP). These disorders are collectively called the "great obstetrical syndromes". 4 Although some pregnancy complications such as fetal growth restriction and HDP occur during the 2 nd and 3 rd trimester in humans, attention should be focused on the early stages of intrauterine development to understand the underlying mechanisms of these diseases. There is no doubt that the placenta plays crucial roles in the development and health of all offspring, but there is still little information on the human placenta. In this review, we discuss the differences in the anatomy, transcription factor networks, and epigenetic properties between mice and humans during placental development and then describe the molecular pathology of perinatal complications. Finally, we introduce several recent technical breakthroughs in the field of human placental research.

| PL ACENTATI ON DURING E ARLY PREG NAN C Y
Studies on the human placenta have been conducted using samples obtained from artificial abortion or preterm delivery. Animal models, such as the mouse, are also useful in elucidating the molecular mechanisms of placentation. However, it should be noted that although preimplantation development is similar between humans and mice, there are substantial differences in placental structure and trophoblast subtypes after implantation.

| Placental development before and during implantation
The mature oocyte and sperm fertilize to form a totipotent zygote.
After several cell divisions, a zygote gives rise to a blastocyst that is composed of inner cell mass (ICM) and surrounding trophectoderm (TE) (Figure 1). TE is classified into the polar and mural TE: The former is adjacent to ICM, and the latter is not in contact with ICM.
Whereas the polar TE proliferates rapidly, the mural TE stops cell division and differentiates into primary trophoblast giant cells (TGCs) in mice. TGCs are multinucleated cells that replicate their genomic DNA without cell or nuclear division. ICM differentiates into epiblast and primitive endoderm, which are destined to form embryonic and yolk sac tissues, respectively. Mouse blastocysts implant in utero at around embryonic day 4.5 (E4.5). During implantation, endometrial stromal cells undergo a special process called decidualization. This response is driven by the sex steroid hormones, estrogen, and F I G U R E 1 Comparison of periimplantation embryos in human and mouse. Epiblast and trophoblast directly interact in mice, but not in human. In mice, FGF4 from the epiblast is required for trophoblast proliferation. In humans, epiblasts and trophoblasts are separated by mesenchymal cells and cannot interact directly. Therefore, human trophoblast proliferation is likely to be epiblast (FGF4) EVT independent (maintenance of human TS cells requires EGF, not FGF). EPI: epiblast; PE: primitive endoderm; TE: trophectoderm; TSCs: trophoblast stem cells; TGCs: trophoblast giant cells progesterone, in both mice and humans. 7 Decidualization begins after implantation in mice, whereas it is induced during the mid-secretory to late secretory phase of the menstrual cycle irrespective of implantation in humans. 7,8 The interaction between trophoblasts and maternal decidua is particularly important for placental development. 9 Impaired decidualization, such as endometriosis, has been reported to increase the risk of infertility and perinatal complications. 7,10

| Placental development after implantation in mice
In mice, a longitudinally elongated structure, known as the egg cylinder, is formed at about E6.0 ( Figure 1). The extraembryonic ectoderm and the ectoplacental cone protrude into the maternal decidua. At the periphery of the ectoplacental cone, secondary TGCs differentiate. FGF4 provided by ICM and epiblast is required for the proliferation of polar TE and extraembryonic ectoderm. 5 At E7.5, the ectoplacental cavity is formed. The base of the ectoplacental space is called the chorion. In the extraembryonic cavity, the extraembryonic mesoderm-derived allantois extends from the fetal tail toward the chorion. At E9.0, the allantois fuses to the chorion and fetal blood vessels develop. At this stage, the chorionic ectoderm fuses with the roof of the ectoplacental space, and the ectoplacental space disappears. Failure of this process is the most frequent cause of mid-pregnancy embryo lethality in mouse mutants. 11,12 By E10.0, the basic structure of the so-called chorioallantoic placenta completes ( Figure 2). The outermost layer (maternal side) of the placenta consists of secondary TGCs and serves to anchor the placenta to the uterus. A spongiotrophoblast layer is formed inside TGCs. After approximately E13.5, the spongiotrophoblast layer contains glycogen cells. The fetal side of the placenta, the labyrinthine layer, is invaded by fetal blood vessels from the allantois to form a capillary network.

| Placental development after implantation in humans
In humans, implantation occurs at E6.0-8.0. Trophoblasts adhere to endometrial epithelial cells and penetrate decidualized endometrial stroma. In mice, this invasion is triggered by TGCs, but in humans, primitive syncytiotrophoblast cells appear around undifferentiated cytotrophoblast (CT) cells and invade the decidua. 13,14 Primitive syncytiotrophoblast cells secrete enzymes that digest the decidua and enlarge the space for the embryo. In contrast to the mouse epiblast that acquires a columnar morphology, the human epiblast forms a disk-like structure. Although the gross morphology is different, both mouse and human epiblasts form pseudostratified columnar epithelium. In the human embryo, epiblast cells adjacent to the trophoblast form squamous epithelium known as amnion and primitive endoderm-derived yolk sac is formed. By E14.0, primitive streaks appear, gastrulation starts, and primordial germ cells (PGCs) are specified.  Although the mouse and human placentas are morphologically different, they are both classified as the hemochorial placenta. 16,17 In humans, the major sites of maternal nutrient and gas uptake are lined by a single layer of ST cells. In mice, there are two layers of syncytium.
The feto-maternal interface is highly variable among mammalian species and believed to be optimized by species. 18 At this interface, placental development proceeds through complex and sophisticated interactions between the trophoblast and the endometrium. Indeed, the maternal endometrium is known to play important roles in placental development. A study using human placental explant culture showed decidual stroma cell (DSC)-derived Neuregulin-1 (NRG-1) promotes EVT differentiation. 19 Moreover, DSCs secretes a variety of growth factors and cytokines. They alter the expression of MMPs and integrins in trophoblast, which promotes trophoblast invasion. 20 At the same time, DSCs also secretes inhibitors of MMPs which suppress excessive trophoblast invasion. 21 Thus, the trophoblast differentiation and invasion into maternal tissues are thought to be exquisitely controlled in part by the paracrine factors released from DSCs. 22 The endometrial epithelium also plays an important role for placentation. As described above, the maternal arterial circulation to the placenta is not fully established until 10-12 weeks of pregnancy.
Endometrial glands are thought to be necessary to provide nutrients for the conceptus before the maternal blood supply. Indeed, the intervillous space before the entry of the maternal spiral artery is filled with secretions from the endometrium. 23 Endometrial glands secrete amino acids, lipids, proteins, and sugars that are nutrients of the conceptus. 24 Endometrial glands also supply growth factors such as VEGF, LIF, and EGF. 25 Studies using human villus explants showed that EGF stimulates CT cell proliferation at 4-5 weeks of gestation and the secretion of hCG and human placental lactogen (hPL) from ST cells at 6-10 weeks of gestation. 26 Thus, secretion from both of stromal cells and epithelial cells is important at each stage of trophoblast growth and differentiation.

| TR AN SCRIP TI ON FAC TOR NE T WORK S REG UL ATING TROPHOB L A S T DE VELOPMENT
In mouse preimplantation embryos, Oct4 (Pou5f1) and Cdx2 antagonize each other and play an important role in the segregation of the embryonic and extraembryonic lineages. 27,28 Oct4 protein is expressed in all cells of the morula but gradually restricted to ICM.
Cdx2 is expressed in the outer cells of the morula and completely restricted to TE at the blastocyst stage. 29 The Hippo signal transcrip- expression, on the other hand, seems to show well-conserved expression patterns between mice and humans. The role of Hippo signaling during human preimplantation development has not been well studied. 33 Whereas Cdx2 expression persists in the extraembryonic ectoderm after implantation in mice, it is controversial whether CDX2 is expressed in trophoblast lineages after implantation in humans. [34][35][36] During mouse placental development, Tead4 and Cdx2 act in concert with multiple transcription factors, including Eomes, Elf5, Ets2, Esrrb, Gata3, Sox2, and Tfap2c. 37,38 Dysfunction of these genes leads to abnormal placentation and embryonic lethality shortly after implantation. [39][40][41][42][43][44][45][46][47][48][49][50] Mouse TS cells have been used to analyze the interactions and regulation of these transcription factors. For example, Esrrb and Sox2 are found to be targets of the fibroblast growth factor (FGF) signaling mediated by Fgfr2c. Removal of FGF or addition of FGF signaling inhibitors rapidly reduces TS cell-specific gene expression. Interestingly, the expression of ESRRB and SOX2 is negligible in TE and CT cells in humans. This observation is consistent with the F I G U R E 3 Comparison of signaling pathway and transcription factors between human and mouse TSCs. Wnt and TGF-β signals are inversely involved in mTSCs and hTSCs. Some of the transcription factors required to maintain mTSCs are not expressed in hTSCs. In mTSCs, Esrrb and Sox2 are known to be activated downstream of FGF signals. ESRRB and SOX2 are not expressed in hTSCs, which is consistent with the FGF-independent nature of hTSCs. The specific transcription factors for hTSCs are currently unknown absence of FGFR2C in human TE and CT cells. 44,45,51 Therefore, ESRRB and SOX2 might be dispensable for human trophoblast development.
Additionally, human TE and CT cells do not express EOMES. 32,35,52 In contrast, GATA3, TFAP2C, ELF5, and ETS2 are expressed in TE and/ or CT cells in humans. 32,35,36,53,54 These differences and similarities of the expression patterns of transcription factors in human and mouse trophoblast cells are summarized in Figure 3.

| EPI G ENE TI C REG UL ATI ON IN THE PL ACE NTA
Epigenetics is important not only for cell fate determination but also for stabilization of lineage-specific gene expression patterns.
Genetic information must be properly regulated by epigenetic modifications. 57 Epigenetic modifications such as DNA methylation and histone modifications play crucial roles in placental development.

| DNA methylation reprogramming after fertilization
Among epigenetic modifications, methylation of cytosine residues in CpG dinucleotides has been well studied. DNA methylation plays important roles not only in cell lineage specification but also in the control of the genome imprinting and the X chromosome inactivation. DNA methylation goes through two dramatic reprogramming during mammalian development. One occurs during gametogenesis, and the other does after fertilization. 58 These two waves of reprogramming are observed in both humans and mice. Global DNA demethylation during gametogenesis allows expression of genes required for meiosis and also resets the parent-of-origin-dependent DNA methylation at the imprinted loci. Some retrotransposons are resistant to demethylation and mediate the inheritance of epigenetic information across generations. 59 Global DNA demethylation also occurs after fertilization, but the imprinted loci are protected at this stage. In mice, the sperm-derived genome is demethylated by the ten-eleven translocation (Tet) 3 demethylase, which is known as active demethylation. On the other hand, the oocyte-derived genome is protected from demethylation by

| Regulation of placenta-specific gene expression by DNA methylation
Many genes have been identified to show placenta-specific expression, 72 and some of them are regulated by DNA methylation.
For example, the promoter of ELF5 is hypomethylated in trophoblast cells but hypermethylated in most embryonic lineage cells in both humans and mice. 36 The promoters of INSL4 and DSCR4 show similar methylation patterns. 73,74 We also recently identified 55 promoters that exhibit placenta-specific hypomethylation in humans. 52 The DNA-binding capacity of many transcription factors is influenced by DNA methylation. In vitro studies using synthetic meth- has been reported that LTRs act to stimulate the secretion of corticotropin-releasing hormone, a regulator of gestational age. 84 It has also been reported that LTRs upstream of the nitric oxide synthase gene NOS3 drive placenta-specific transcriptional isoforms in the human placenta. Moreover, syncytins, envelope genes derived from retroviruses, play an important role in both mouse and human ST cells. 85,86 Thus, retrotransposons and their DNA methylation patterns may contribute to the evolution of the placenta.

| Development of trophoblast cells and histone modifications
Histone modifications are essential for trophoblast development in mice. Embryos deficient in EHMT2 (also known as G9A), which mediates H3K9 dimethylation, fail to fuse the chorioallantoic membranes, which is lethal in mid-pregnancy. 87

| Placenta-specific genomic imprinting:
In mice, the development of both female-developing embryos (embryos composed of the oocyte-derived nuclei) and androgenetic embryos (embryos composed of sperm-derived nuclei) is lethal. So, the development of these causes the death of the embryo. However, their phenotypes are very different. Parthenogenetic embryos rarely form the placenta, whereas androgenetic embryos show placental hyperplasia. Similarly, in humans, androgenesis is associated with complete hydatidiform mole, a disease characterized by trophoblast overgrowth. Thus, genomic imprinting is important for placental development.
Some imprinted genes show uniparental expression in a tissue-specific manner, and many tissue-specific imprinted genes have been found in mice, especially in the placenta and brain. We performed a comprehensive analysis of the imprinted genes in the mouse placenta, 96 which revealed that many placenta-specific imprinted genes reported by that time were due to contamination of maternal tissues, and only 12 genes actually underwent imprinting.
We also demonstrated that uniparental expression of these placenta-specific imprinted genes was disrupted in the placentas of  [99][100][101] The reason why humans and mice have acquired almost completely different sets of placenta-specific imprinted genes is currently unknown, but it is possible that they contribute to differences in placental structure, gestational period, or litter size between these species.

| Incomplete X chromosome inactivation in human placenta:
One of two X chromosomes is inactivated in female mammalian cells. In marsupials, both embryonic and extraembryonic tissues show imprinted X chromosome inactivation, in which the paternal X chromosome is selectively inactivated. In mice, inactivation of the X chromosome in embryonic tissues is random, whereas the paternal X chromosome is selectively inactivated in the placenta.
It has been reported that X chromosome inactivation is skewed in the human placenta. The paternal X chromosome is preferentially inactivated in the human placenta but the degree varies among individuals. 64,102

| EPIG ENOMIC AB NORMALITIE S AND PL ACENTAL DISORDER S
Large-scale knockout studies in mice indicate that about 25-30% of all genes are essential for survival. However, with few exceptions, such analyses have focused on embryos and little attention has been paid to placentas. 103 A recent study revealed that morphological abnormalities of the placenta are observed in approximately two-thirds of embryonic lethal mutant mice. 12 It is also known that placental abnormalities tend to be associated with abnormalities of the heart, brain, and vascular embryos. In humans, infants with congenital heart disease have also been reported to exhibit a significantly higher frequency of placental abnormalities than healthy controls. [104][105][106] Abnormalities in genomic imprinting may be associated with the development of small for gestational age (SGA), intrauterine growth restriction (IUGR), and HDP in humans. For example, in the placenta of IUGR, there is a trend toward increased expression of the paternally imprinted gene PHLDA2 and decreased expression of the maternally imprinted genes MEST and PLAGL1. Decreased expression of paternally imprinted genes MEG3 and GNAS has also been reported. 121 HDP is also thought to be associated with epigenetic abnormalities. The expression of several miRNAs is reported to be dysregulated in preeclamptic placentas compared to normal placentas. 122,123 The human placenta-specific imprinted gene CUL7 is also reported to be hypomethylated and shows increased expression in human fetal growth restriction placentas. 124 Deficiency of CUL7 also causes fetal growth restriction, severe post-natal growth restriction, and 3M syndrome type I, a congenital anomaly syndrome characterized by characteristic facies. 125 Furthermore, it has been reported that Cul7-deficient mice cause vascular abnormalities in the decidua and  124 CYP2J2, which is a placenta-specific imprinted gene 64 and encodes one of the cytochrome P 450 enzymes known as drug-metabolizing enzymes, 126 is reported to be highly expressed in HDP patients. 126 In addition, the metabolite EET of CYP2J2 is also increased in a rat model of HDP. These results support the hypothesis that aberrant expression of imprinted genes may be associated with IUGR and HDP.

TA B L E 1 Previous in vitro models for human trophoblast cell research
A complete hydatidiform mole (CHM) is caused by androgene- Recently, we established cell lines from CHM samples and demonstrated that silencing of the imprinted gene P57KIP2 confers resistance to cell cycle arrest by contact inhibition in these cell lines. 127

| DE VELOPMENT OF US EFUL TOOL S TO S TUDY HUMAN PL ACENTAL DE VELOPMENT
The establishment of mouse TS cells was first reported in 1998. 5 It was only recently that human TS cells were established. 52 Furthermore, in recent years, various useful tools have been developed to study placental development, such as artificial mouse embryos and human trophoblast and endometrial organoids ( Figure 5).

| Human trophoblast stem (TS) cells:
In 1998, mouse TS cells were first derived from blastocysts or extraembryonic ectoderm of post-implantation embryos using FGF4, heparin, and mouse embryonic fibroblasts (MEF). 5  In summary, human TS cells retain unique characteristics of trophoblast cells and therefore are very useful to analyze human placental development and function.

| Artificial embryos generated using TS and ES cells:
TS cells have also been used for modeling of early embryos. Mouse TS cells, when combined with ES cells, self-organize to form post-implantation embryo-like structures that mimic early embryo development. 128 Blastocyst-like structures were also generated using mouse ES and TS cells. These "blastoids" have the potential to implant into the mouse uterus and induce decidualization. Although ethical issues must be addressed, the development of human artificial embryos using TS and ES cells will open up great potential for the study of human embryogenesis.

| 3D trophoblast organoids
Soon after human TS cell establishment was reported, three-dimensional trophoblast organoids have been generated using CT cells purified from first-trimester placental tissue. 129,130 These organoids contain CT-and ST-like cells and can be maintained for a long time.
They also give rise to EVT-like cells as human TS cells do. Such trophoblast organoids are useful for studying placental development and function under more physiological conditions.

| New models and data resources to study the feto-maternal interface
The generation of organoid models has also been reported for the maternal endometrium. Three-dimensional culture of endometrial glands leads to self-organized cyst-like structures, which respond to sex hormones such as estrogen and progesterone. Endometrial organoids can also replicate the phenotype of endometriosis and endometrial cancer. 131 Endometrial receptivity is essential for successful implantation and pregnancy. Indeed, two-thirds of implantation failure is caused by endometrial receptivity. 132 The combination of organoid technologies that mimic the development of embryos, trophoblast cells, and endometrium has the potential to recapitulate implantation and placentation more precisely than the conventional trophoblast-endometrium coculture assay. 133,134 This would expand our understanding of the human early developmental process, which has been a "black box" due to ethical constraints.
Recently, single-cell RNA-seq analysis has provided a compre-

Our derivation of human TS cells was approved by the Institutional
Review Board (IRB) approval at Tohoku university (2014-1-879).

Conflict of interest:
Authors have no conflict of interest to be declared.