A close look at the mammalian blastocyst: epiblast and primitive endoderm formation

Abstract

During early development, the mammalian embryo undergoes a series of profound changes that lead to the formation of two extraembryonic tissues—the trophectoderm and the primitive endoderm. These tissues encapsulate the pluripotent epiblast at the time of implantation. The current model proposes that the formation of these lineages results from two consecutive binary cell fate decisions. The first controls the formation of the trophectoderm and the inner cell mass, and the second controls the formation of the primitive endoderm and the epiblast within the inner cell mass. While early mammalian embryos develop with extensive plasticity, the embryonic pattern prior to implantation is remarkably reproducible. Here, we review the molecular mechanisms driving the cell fate decision between primitive endoderm and epiblast in the mouse embryo and integrate data from recent studies into the current model of the molecular network regulating the segregation between these lineages and their subsequent differentiation.

Introduction: an overview of the early mouse development

In eutherians, embryonic development takes place in the maternal reproductive tract. From the infundibulum, where the egg is fertilized, to the uterus, where the embryo implants several days later, major events take place that influence proper development. In the mouse embryo, the preimplantation period lasts approximately 4 days and involves the specification and segregation of extraembryonic tissues, the trophectoderm (TE) and the primitive endoderm (PrE), from the embryonic tissue, the pluripotent epiblast (Epi). Extraembryonic tissues will play essential roles for the development of the embryo proper [1, 2].

After fertilization, the mouse embryo undergoes a series of seven divisions, which are characterized by high dynamic variations in terms of their cell cycle properties. For example, the first two divisions are completed in 18–20 h, while cell cycle length is shortened to 11 h in the subsequent two divisions [3]. By the second division, the blastomeres already divide asynchronously.

During this period, profound changes occur within the zygotic chromatin, which originates from parental dimorphic genomes. The zygotic genome is initially silent at the time of fertilization, and the zygote relies on maternal stores accumulated during oogenesis. The zygotic genome becomes activated by the end of the one-cell stage with a major activation at the two-cell stage [4, 5].

Several morphogenetic events take place during the first 4 days. The first noticeable event occurs at the eight-cell stage when blastomeres acquire the first signs of polarity. Intercellular adhesion increases in blastomeres due to the relocalization of cadherin–catenin complexes, resulting in compaction. The embryo at this stage resembles a mulberry and has been named morula from the Latin word morum. Concomitantly, blastomeres acquire a stereotypical apical-basal polarity with the apical side facing the external environment. Cellular polarity can be visualized by the reorganization of several intracellular components and proteins such as EZRIN or the aPKC/PAR complexes [6–8]. The nucleus localizes basolaterally while endosomes accumulate on the apical side. The cytoskeleton is also modified such that actin filaments are enriched on the apical side. Two distinct populations of microtubules are distributed to the apical and basolateral regions. Microvilli that are originally homogenously distributed begin to accumulate apically. How compaction and cell polarity are linked is not fully understood. One consequence of compaction is the production of inside and outside cells, resulting from symmetric and asymmetric divisions [6] (Fig. 1).

Fig. 1.

Schematic representation of early mouse embryonic development. The scheme depicts embryonic development starting from the compacted eight-cell stage (2.5 days post-fertilization) to the implanting blastocyst (4.5 days post-fertilization). Starting from the eight-cell stage, the orientation of the plane of cleavage will be important in the generation of inside ICM (grey) and outside trophectoderm (green) cells. While symmetric division (purple) will generate two outside cells, asymmetric division (orange) will generate one inside and one outside cell. Subsequently, ICM cells will differentiate into the Epi (red) and PrE (blue)

At approximately the 32-cell stage, cells facing the external environment have been specified to the trophectoderm lineage and form a fully functional epithelium. The presence of NA⁺/K⁺ ATPases and aquaporins ensures channeling of ions and water, thus allowing for the formation of a fluid-filled cavity sealed by tight junctions, known as the blastocoel [9]. Once the cavity has appeared, the embryo is referred to as a blastocyst.

Preimplantation mouse development is the period when extraembryonic lineages are specified from the embryonic pluripotent lineage. Extraembryonic tissues are critical to ensure the proper development of the embryo inside the uterine wall [1]. At the time of implantation, the embryo is composed of three cell lineages that are molecularly distinct and topographically regionalized (Fig. 1). The Epi is enclosed on one side by the PrE facing the blastocoel cavity and on the other side by the polar TE. Fate mapping studies have shown that these lineages have also acquired a restricted developmental potential. The TE will contribute to the fetal portion of the placenta. The PrE will give rise to the endoderm layers of the yolk sac and is incorporated into the fetal gut with a graded anteroposterior contribution [10]. The pluripotent Epi is the tissue from which the embryo will form and is the source of the ES cells.

Embryological studies have revealed that early mouse development is highly regulative and can adapt to experimental modifications such as the addition, removal, or rearrangement of blastomeres up to the blastocyst stage [8]. Our current model posits that these three cell lineages result from two distinct and consecutive cell fate decisions. The first is the formation of the trophectoderm and the inner cell mass (ICM), involving both cell–cell contacts and cell polarity (Fig. 1) [8, 11]. The ICM then gives rise to the PrE and the Epi. These two cell fate decisions result from a complex interaction between the genetic program of a given cell and its position within the embryo. The second cell lineage decision is a multistep process. PrE and Epi differentiation programs are originally initiated in all blastomeres of the eight-cell-stage embryo. PrE and Epi are then further specified within the ICM of the blastocyst embryo. PrE and Epi cells ultimately sort such that PrE cells lie in contact with the blastocoel.

Understanding the molecular mechanisms that specify early cell fate decisions has long been a quest for developmental biologists. Most of our current knowledge of mammalian embryonic development comes from studies on the mouse. The study of mouse models has provided information with potential applications for subjects such as the prenatal diagnosis of human and livestock and the establishment of ex vivo culture conditions for embryo-derived stem cells that could be used in cell therapy and regenerative medicine.

Molecular mechanisms for Epi versus PrE specification

It was initially thought that the cells located at the surface of the ICM in contact to the blastocoel would be specified as PrE. However, the discovery that at E3.5 Gata6 mRNA is present in a subset of ICM cells [12] and that Nanog and Gata6 expression are mutually exclusive, in a “salt-and-pepper” organization, supported an alternative lineage specification model [13]. Subsequently, Nanog and Gata6 expression patterns have been further characterized. The two proteins are coexpressed in nearly all blastomeres from the eight-cell stage [14, 15] while their mRNAs can be detected as early as the two-cell stage [16, 17]. From the late morula stage (around the 20-cell stage), the cells downregulate either Nanog or Gata6 mRNA levels, and become mutually exclusive by E3.75 (~64-cell stage) [13, 15, 16, 18]. However, individual ICM cells develop at different rates [15, 19], where some begin differential expression as early as E3.25, and others can still co-express NANOG and GATA6 until E3.75. It is not known whether the beginning of this differential expression already reflects specification or whether they are merely fluctuations in expression levels, followed by stabilization of gene expression [20]. Indeed, it was shown that Nanog expression fluctuates in ES cells maintained in standard culture conditions [17, 21]. A PrE reporter also shows reversible expression in ES cells, with an inverse correlation to Nanog expression [22]. Thus, in vitro culture has been able to capture a state where cells identity balances between Epi and PrE. These in vitro fluctuations in expression are rather slow and the technologies used only reflect RNA transcription activity, however it is possible that such fluctuations might create the slight differences between ICM cells that would initiate cell specification [23]. The fact that NANOG can repress its own expression in ES cells [24, 25] could contribute to the mechanism of fluctuation. The Nanog auto-repression feedback loop has not been analyzed in the embryo, however, high levels of Nanog mRNA and protein can be detected in the same Epi cells at E3.5 [26]. Thus, the auto-repression of Nanog might occur at later developmental stages, depending on the presence of activating or repressing cofactors [25].

While the FGF signaling pathway has been shown to be required for preimplantation development [27–29] and PrE differentiation in ES cells [30–32], in vivo analyses confirmed its requirement for PrE specification. Indeed, embryos mutant for Grb2—an adaptor of the RTK pathway—cannot induce PrE cells and, as a consequence, all ICM cells become epiblastic [13]. These data revealed that ICM precursors have a binary choice and differentiate either into Epi or PrE. The involvement of the FGF pathway was further examined by embryo culture in the presence of FGF4 or FGFR + MEK inhibitors and genetic analyses of Fgf4 inactivation. These studies showed that modulation of FGF activity can balance cell identity either toward a PrE (high RTK activity) or an Epi (low RTK activity) fate [33–36]. Heparan sulfate proteoglycans are required to activate FGF4/FGFR2 pathway [37].

In Nanog mutant embryos, all ICM cells express GATA6, demonstrating that they are specified toward a PrE fate [38] (Fig. 2). Forced expression of Gata6 in ES cells downregulates the expression of pluripotency markers including Nanog [39, 40]. Thus, NANOG and GATA6 seem to repress each other’s expression, and indeed NANOG binds to Gata6 regulatory sequences and decreases its activity in ES cells [41].

Fig. 2.

Modeling PrE and EPI cell fate decision in wild-type and mutant contexts. The scheme depicts PrE (blue) and Epi (red) formation starting from unspecified ICM cells (grey) to a determined state. In wild-type embryos, unspecified ICM cells have already initiated both PrE and Epi genetic programs (visualized by the expression of NANOG and GATA6). Through positive and negative regulation (that might be direct or indirect), a cell fated to become Epi will secrete FGF4 ligand, which will instruct an adjacent cell to become PrE, through a signaling cascade involving FGFR2, ERK, and OCT4. PrE progenitors will turn off Epi genetic programs and will mature through the expression of SOX17 and GATA4. In the absence of Fgf4, ICM cells cannot be instructed to become PrE and by default mature to Epi cells. In the absence of Nanog, ICM cells cannot mature to Epi fate and consequently do not secrete FGF4 to instruct cells to become PrE. ICM cells can form intermediate PrE progenitors but these progenitors cannot mature. In absence of Oct4, the signal transduction cascade between FGFR and the regulation of PrE transcriptional program is impaired. Epi cells are formed but PrE progenitors do not properly mature and acquire an undetermined identity (brown). Changes in the size of the letters and the arrows indicate changes in the levels of proteins and the regulation loops

During blastocyst stages, Fgf4 expression depends on the presence of NANOG, as no mRNA can be detected in Nanog mutants [38]. Conversely, GATA6 is expressed in these mutants and does not require activation by FGF4. Interestingly, early inhibition of FGFR + MEK prevents GATA6 expression in a Nanog-deficient context [38]. This suggests that other RTK ligands might be required to induce Gata6 expression at the ~eight-cell stage. The observation that Fgf4 ^−/− embryos still express GATA6 until the ~32-cell stage [35, 36] confirms that other RTK ligands are involved in initiating GATA6 expression. In these embryos, GATA6 expression is not maintained after the 32-cell stage, probably due to the inhibition by NANOG, as GATA6 expression persists in Nanog ^−/− embryos [38]. Thus, FGF4 seems to act in Nanog-expressing cells only for the establishment of the salt-and-pepper pattern to induce PrE fate in neighboring cells. As a consequence, Epi cells must insulate themselves from their own secretion of FGF4 and do so by downregulating the expression of Fgfr2 [16, 18, 20]. In neighboring presumptive PrE cells, RTK activation by FGF4 downregulates Nanog, possibly through the activation/maintenance of GATA6 expression. Recent FISH analysis has revealed that Nanog is expressed from a single allele before the late-blastocyst stage [17]. While this result has been recently challenged [42, 43], further analysis using single-cell allelic sequencing supports the monoallelic expression of Nanog [44]. Switching from mono-allelic to bi-allelic expression could increase the amount of NANOG protein. This switch occurs after Epi/PrE specification and Nanog heterozygous embryos correctly specify Epi and PrE cells [17, 38]. While little is known about the mechanisms inducing Nanog expression, one way to limit its expression levels may be through silencing one allele.

After TE specification, OCT4 is expressed throughout the ICM in both Epi and PrE lineages. Its expression becomes restricted to Epi cells only after E4.25 [45]. Given its role in pluripotency in ES cells, it was unexpected that initiation of Epi and PrE specification occurred normally in Oct4 mutant embryos [46]. However, GATA6 expression is progressively lost from E3.5 in the absence of Oct4 and subsequent PrE markers are not expressed (Fig. 2). This loss might be a consequence of higher expression of NANOG in individual cells [47]. Thus, OCT4 is involved in Epi/PrE maintenance rather than specification. As FGF signaling is key for PrE/Epi cell lineage divergence, the authors analyzed the consequences of modulating FGF activity in the absence of OCT4 [46]. When treated with FGFR + MEK inhibitors, NANOG expression is induced in all ICM cells, as expected, indicating that this mechanism is OCT4-independent. However, in contrast to wild-type embryos, GATA6 expression is elevated in treated Oct4 ^−/− embryos, with some cells co-expressing NANOG and GATA6 [46]. Thus, in the absence of RTK signals, NANOG mediates GATA6 repression through Oct4 expression [38]. In wild-type embryos, when FGF activity is increased, all ICM cells adopt a PrE identity. Unexpectedly, in the absence of Oct4, FGF4 administration can no longer prevent NANOG expression and PrE markers are absent [46]. Therefore, OCT4 mediates the inhibition of NANOG expression by the RTK pathway. It remains to be determined whether the absence of GATA6 in these conditions reflects a lack of initiation or a lack of maintenance. Altogether, OCT4 seems to transduce the artificially low and high activities of the RTK pathway, by repressing either GATA6 or NANOG expression respectively. Still, it is not clear how in the absence of Oct4 but in physiological conditions (without artificial perturbation of FGF activity) the salt-and-pepper expression pattern of GATA6 and NANOG is established. Further genetic analyses will be required to decipher the interaction between OCT4 and RTK signals that drives cell fate specification. Interestingly, OCT4 can be phosphorylated by Erk signaling and this phosphorylation has been shown to either inhibit its DNA binding [48] or reduce its transactivation capacities [49]. This post-translational modification could modulate its activity or its nuclear export, which has been shown to be correlated with cell lineage segregation [50, 51]. It is tempting to speculate that such post-translational modifications could add an additional level of regulation of ICM cell lineage specification.

Loss of ICM cell plasticity and PrE versus EPI cell determination

One classic question in developmental biology relates to when and how a cell becomes specified to a particular identity, which usually leads to a state when it becomes irreversible, or determined. The major difference between determination and specification is the notion of reversibility [52]. To assess when a cell is determined (or when cell plasticity is lost), experimental embryologists have designed a set of paradigms where the cell fate potential is examined in situations where the environment is altered (e.g., grafting a cell in a different environment). In normal conditions, lineage tracing and 4D imaging have suggested that Epi and PrE cell lineages do not change their identity [13, 53, 54]. However, experimental perturbation has revealed that Epi/PrE cells become determined relatively late. This is illustrated by several observations. First, Yamanaka et al. [34] demonstrated that when FGF4 or FGFR/MEK inhibitors are applied, the embryo can re-establish a salt-and-pepper pattern when returned to normal culture conditions until ~E4.0. Second, reducing the number of ICM cells does not seem to impair the PrE/Epi ratio [55], meaning that loss of cell plasticity does not depend on cell number. Lastly, Grabarek et al. [45] addressed the question of cell determination by examining the fate of single ICM cells grafted into morulae. Their data are consistent with the idea that ICM cell lineages are determined by E4.0.

Altogether, ICM cell lineages become determined by the time of implantation and until this time ICM cells have great propensity to establish a salt-and-pepper pattern. This ability is probably due to the fact that the cells do not specify all at the same time. The acquisition of Epi or PrE identity starts for some cells around E3.0–E3.25 and the completion of the salt-and-pepper pattern finishes around E3.75–E4.0. The cells that specify late do so in response to their environment, i.e., in function of the ratio of Epi to PrE cells that have already been specified, which is perceived by unspecified cells through the low or high levels of intercellular FGF4.

We currently do not completely understand why it is so important to generate an equilibrated ratio of Epi and PrE cells. However, it seems that the embryo can tolerate a significant loss of Epi [55] or PrE cells as observed in various mutant contexts [35, 36, 56, 57]. Differences in Epi/PrE ratios or in the efficiencies of ES cell derivation during standard culture conditions (where cells are grown on mitotically fibroblast feeders and medium is supplemented with LIF) were observed between mouse strains and probably reflect variabilities in how a FGF signal is generated and interpreted at the cellular level [38, 58].

The mechanisms triggering the loss of plasticity are currently unknown. They do not seem to depend directly on Nanog and Gata6 expression levels. Indeed, despite a long and strong induction of Nanog expression by FGFR + MEK inhibitors (from E1.5 to E3.5), cells can change and adopt a PrE identity when the embryo is placed back in a control medium [34]. Downstream factors depending on either NANOG or GATA6 expression could be involved in the fixation of cell identity. For example, SOX7, SOX17, and/or GATA4 could lock in the PrE identity. Also, epigenetic mechanisms could establish the transcription of Nanog and Gata6. However, none of these mechanisms has been analyzed.

Until now, only NANOG, GATA6, the FGF pathway and, to a certain extent, OCT4 have been found to be implicated in PrE and Epi specification, or in the balance between these two lineages. However, several factors required for the maintenance of ES cell pluripotency such as SOX2 as well as BMP and WNT signaling pathways could be also involved.

Stochasticity versus predetermination of ICM lineages specification?

One remaining open question is how the salt-and-pepper pattern is established. Does it rely on the stochastic initial expression of Gata6, Nanog, and Fgf4, or does the history of the blastomere predetermine its future cell fate identity? ICM cells arise from two to three rounds of asymmetric cell divisions in the morula (8 > 16-cells, 16 > 32-cells, and 32 > 64-cells) (Fig. 1) [54].

Several groups have attempted to answer this question. Single-cell RNA expression analyses show that cell-to-cell variation exists at E3.25 when Epi and PrE lineage markers are still coexpressed [16, 18, 20]. These experiments thus favor a model in which stochastic activation of Epi- or PrE-specific genes creates a heterogeneity that is progressively reinforced to ultimately result in exclusive expression patterns [20]. These experiments also reveal that Fgf4 is the first gene differentially expressed at 32-cell/E3.25 stage. The strong inverse correlation with Fgfr2 expression from cell to cell [16, 18, 20] demonstrates that this is not a noisy activation and probably reflects the initiation of the specification process.

Yamanaka et al. [34] labeled single cells at the eight-cell stage and followed them in culture to determine whether inner cells arise from the first or the second asymmetric division. They analyzed their contribution to PrE and Epi fates at post-implantation stages. Using this approach, they failed to establish a significant correlation between the origin of the ICM cells and Epi or PrE cell fate. In contrast, Morris et al. followed the entire cell population in culture starting from the eight-cell stage and observed that cells from the first asymmetric division mainly contributed to Epi while cells from the second contributed mainly to PrE and cells from the third exclusively to PrE [54, 59]. They suggest that inner cells produced late inherit features of outer cells, possibly FGFR2, biasing them toward a PrE fate [59].

More recently, Krupa et al. [60] re-addressed this question by analyzing the contribution of reconstituted blastomeres from the 16-cell stage embryo to the Epi and PrE lineages. Interestingly, they observed that the composition of inner and outer cells of the reconstituted embryo affected the relationship between the history and the destiny of the cells. Indeed, when starting with a small number of inner cells (three inner and 13 outer cells) they found that inner cells from the first asymmetric division tend to adopt an Epi fate while cells from the second preferentially adopt a PrE identity. This bias was not observed in reconstituted embryos with a higher number of inner cells (five inner and 11 outer cells). Since FGF signaling is the major driver of PrE/Epi specification, they proposed that when a fewer number of inner cells are generated, the source of FGF4 is low and so they would acquire an Epi fate. However, when more inner cells are born, the level of FGF4 is higher and so they can acquire Epi or PrE identities. Primary and secondary inner cells tend to be differentially enriched in Fgf4 or Fgfr2 transcripts respectively [59, 60]. However, further examination with statistically significant differences or the use of the corresponding mutant embryos as controls would be needed to confirm the interpretation. Altogether, these results indicate that it is the number of inner cells and not the round of division that is important to deliver required levels of FGF4 and FGFR2 to allow PrE specification.

Overall, considering the major role of the FGF4 signal in Epi/PrE specification, future studies may refine how the FGF4 signal is produced and interpreted, either by stochastic activation, inner cell number, or the history of asymmetric divisions.

Maturation of PrE and Epi cells

Once specified, each lineage carries on its own developmental program. In the PrE lineage, maturation can be visualized by the sequential activation of a series of lineage-specific markers. The first PrE markers to be expressed are Gata6 (from the eight-cell stage) and Pdgfra (from eight to 16 cells) [15]. They are initially expressed in all blastomeres and become progressively restricted to the PrE at the late blastocyst stage when cells are sorting, with PDGFRα being restricted later than GATA6. Sox17 is first detected at the 32-cell stage [54, 57, 61] while Gata4 is expressed from the 64-cell stage [15]. Using single-cell expression analysis, a recent study proposes a hierarchy in the sequential activation of PrE-specific genes between E3.25 and E3.5 [20]. At the time when PrE cells are sorted and start to epithelialize, they will express additional markers such as Sox7, Lrp2, and Dab2 [19, 57].

The mechanisms regulating the maturation of the PrE lineage, which is visualized by this sequential activation of markers, are partially understood. Chimeric and RNAi experiments showed that a non-cell autonomous factor secreted from Epi cells induces GATA4 and SOX17 expression [38, 62]. Interestingly, while the process of maturation is impaired in Nanog mutants (where cells express Gata6 but not Sox17, Pdgfra, and Gata4), it can be rescued by addition of FGF4 [38]. This major role of FGF signaling was further confirmed by the phenotypical analysis of Fgf4-deficient embryos [35]. Interestingly, the rescue of Gata4 expression in FGF4-treated Nanog mutants required longer cultures conditions, implying that its activation is delayed compared to the one of Sox17 as it is in wild-type embryos. Overall, these experiments also show that the activation of Sox17 and Gata4 as well as the maintenance of PDGFRα expression occurs after cell specification and are implicated in later events of PrE differentiation (Fig. 2). It is possible that their expression could reinforce PrE identity, avoiding a reversion to an Epi identity, and might thus play a role in the specification event. Recently, OCT4 was shown to be required cell-autonomously for PrE development in vitro and in vivo [46, 63] and higher levels of OCT4 promoted ES cell differentiation to a PrE identity in vitro [64, 65] (Fig. 2). FGF4 administration can partially rescue PrE development of Oct4 ^−/− embryos but the exact role of OCT4 remains to be elucidated [47]. In specified PrE cells, OCT4 and SOX17 both bind to a “compressed” response element to induce PrE target genes while in the epiblast OCT4 partners with SOX2 to bind to a “canonical” response element to activate Epi-specific genes [63]. Thus, OCT4 acts to promote both Epi and PrE maturation and Oct4 mutant embryos have defects in both tissues at E4.5 [46]. Interestingly, Stat3 mutants also have defects in the differentiation/maintenance of Epi and PrE cells [66]. While it is not known whether STAT3 acts early during specification or later for the maturation of Epi and PrE, it is involved in the maintenance of Oct4 expression in ES cells [66].

While the maturation of PrE lineage is now relatively well documented, little is known about the maturation of the epiblast. After E4.5, Nanog expression is rapidly shut down. This down-regulation depends on Fgf4 expression, as in Fgf4 mutants NANOG is detected at high levels [35]. This suggests that if Nanog autorepression is effective [24, 25], it is mediated through the FGF pathway. Whether inhibition by FGF4 is autocrine, as suggested by ES cell experiments [67], or is indirect through PrE differentiation, is not known. Indeed, PrE cells could signal to the neighboring epiblast to activate the first steps of maturation, as it does later for the formation of the amniotic cavity [68]. Interestingly, in Pdgfra mutants, the Epi compartment is expanded, suggesting that the PrE sends signals to regulate the Epi [56].

Several in vitro observations suggest that some members of the KLF transcription factor family could be also involved in the maturation of the Epi. Indeed, KLF4 is one of the four factors necessary to reprogram differentiated somatic cells into induced pluripotent stem (iPS) cells [69], but can be replaced by KLF1, 2, or 5 [70]. In vivo, genetic loss of function of Klf1, 2 or 4 is dispensable during early embryonic development while inactivation of Klf5 impairs blastocyst formation [71, 72]. Further analyses have revealed that high levels of KLF5 are required for trophectoderm differentiation and lower levels differentially regulate Epi and PrE lineages where it is expressed homogeneously. The current model proposes that KLF5 promotes the Epi lineage and represses PrE markers, probably through the inhibition of Sox17 [72]. It is not known yet whether KLF5 plays independent, unrelated roles in Epi and PrE.

PrE and Epi cell sorting is achieved through multiple processes

It was previously thought that ICM cells in contact with the blastocoel would acquire a PrE identity, while ICM cells enclosed inside would form the epiblast. However, this positional model was challenged by the finding that the ICM of the mid-blastocyst embryo is composed of cells expressing either Epi or PrE markers in a salt-and-pepper pattern [13]. Subsequently, PrE and Epi cells undergo a series of rearrangements such that PrE cells will be positioned at the interface between the Epi and the blastocoel. A live-imaging approach using a PrE fluorescent reporter has revealed that these cellular rearrangements, referred to as cell sorting, are driven by multiple processes [15].

First, PrE cells that are in contact with the blastocoel tend to retain their position. This observation suggests that some positional information may exist at this stage or that correctly positioned PrE cells differ from the non-sorted cells. Support for the latter idea comes from the observation that sorted PrE cells express Sox7 and start to epithelialize individually, as shown by the apical expression of LRP2 and DAB2 [19, 57]. Once all PrE cells have reached the surface, the distribution of atypical protein kinase C (aPKC) becomes polarized. Inhibition of aPKC activity leads to defects both in PrE maturation and in PrE cell sorting. Live-imaging analysis demonstrates that PrE cells are correctly sorting but fail to maintain a bona fide position close to the blastocoel cavity [73].

This does not exclude the hypothesis that sorting of PrE and Epi cells is achieved through differential adhesive properties as was originally proposed by Townes and Holtfreter [74] using the amphibian embryos. Indeed, in the absence of the adaptor protein DAB2, committed PrE cells either fail to sort or fail to maintain contact with the blastocoel [75]. Other studies suggest that this could also be the case in mutants lacking the integrin β1 [76–78]. However, it is interesting to note that experimental modulation of adhesion properties does not seem to affect sorting of PrE cells in an in vitro cellular assay where null E-cadherin ES cells are mixed with extraembryonic endodermal cells [79]. A recent analysis of Integrin β1 describes a failure in epithelium formation [76], and cell sorting is occurring apparently normally. This is also possibly the case for LamC1, a ligand of integrins [80]. However, redundancy could mask an eventual earlier function. These observations suggest that while possibly involved in the process of cell sorting, differential adhesion is not the only mechanism involved. Similar conclusions were found by computer modeling of the cell sorting process [53, 81, 82].

Analysis of labeled ICM cells during the sorting process revealed that highly dynamic cellular protrusions were present, which were dependent on the filamentous actin network [53]. Whether this reflects oriented migration or differential surface contraction is not clear [83]. In addition, whether PrE cells are sensitive to directional signals provided by the blastocoel remains to be determined. The presence or the retraction of trophectodermal cytoplasmic processes could also play a role in the sorting process [84].

Strikingly, at this developmental period (E3.75–E4.0), the rate of apoptosis is increased [85–87]. We and others have proposed that selective cell death could be involved to refine the segregation of PrE and Epi cells, affecting cells not properly fated and/or mispositioned [15, 53, 88, 89]. Indeed, when the ICM is homogenous as in Nanog mutants, the “wave” of cell death at E3.75 does not occur [38]. PDGF signaling could play a role in this process since, while not involved in PrE cell sorting per se, PDGF signaling plays a critical role in PrE cell survival at this stage [56, 90]. Interestingly, in Nanog-mutant embryos, the PrE progenitors (expressing Gata6 but not Sox17 and Gata4) fail to survive when cultured in the presence of RTK inhibitors [38]. One likely explanation is that low PDGF signaling either due to low expression of the ligand or receptor is not sufficient to maintain PrE cells.

Embryo-derived stem cells; in vitro tools to study lineage maintenance and differentiation

Three stem cell populations can be isolated from the blastocyst and propagated in vitro. Embryonic stem (ES) cells are derived from the epiblast, trophoblast stem (TS) cells from the trophectoderm, and extraembryonic endoderm (XEN) cells from the primitive endoderm. These stem cell lines provide unique in vitro cellular tools to study the molecular mechanisms that control the maintenance of their identity and their differentiation both at the extrinsic (the signaling molecules) and the intrinsic (the transcription network) levels. In addition, since various studies have revealed that complex interactions between embryonic and extraembryonic tissues are involved in embryonic patterning, these stem cells could be used in in vitro assays to model and understand these interactions. This approach has been used to study cardiac induction [91], fetal hematopoiesis [92], and could be extended to study gut morphogenesis, considering the contribution of VE cells to the gut in vivo [10].

ES were originally derived more than 33 years ago from mouse embryos [93, 94]. Since then, much work has been performed to understand how these cells are maintained and propagated in culture. Current protocols to derive ES cells are now combining two pharmacological compounds (known as 2i) that inhibit FGF and GSK3 signaling pathways [95]. Of note, the interleukine molecule LIF is usually added to the medium and is known to promote pluripotency through the regulation of STAT3 and its downstream target MYC. In this situation, ES cell lines are maintained in a “naive” state, refractory to differentiation. They can be isolated very efficiently, even from recalcitrant mouse strains [33, 95, 96] and from other rodent species such as rat [97, 98]. Our current knowledge posits that FGF signaling is necessary to prime ES cells for differentiation [99]. It is not entirely clear how regulation of GSK3 activity participates to maintain pluripotency [100]. Several reports suggest that the main effect is through the regulation of βcatenin, which in turn acts through multiple mechanisms such as regulating transcriptional activity [101, 102], controlling the pool of available OCT4 [103], or by inhibiting the repressive effect of the TCF3 transcription factor [104, 105]. The BMP pathway was also shown to sustain ES cell self-renewal [106] and could play a role as well in cell lineage specification in the embryo, as BMP4 is specifically expressed in Epi cells at E3.5 [16, 107]. ES cells are also commonly used to study PrE differentiation during embryoid body formation [108].

TS cell lines can be established from blastocyst and early post-implantation embryos [109]. TS cells are representative of the trophectoderm lineage and share characteristics with early extraembryonic ectoderm cells. Indeed, TS cells express TE-lineage markers such as Cdx2, Gata3, Eomes, and contribute to TE derivatives when reintroduced into recipient embryos. TS cells have also been generated following misexpression of CDX2, an active form of TEAD4 or RAS in ES cells [110–113]. TS cells can be efficiently maintained in the presence of FGF4 and Activin [109, 114, 115].

XEN cells are derived from the primitive endoderm of the blastocyst-stage embryo [116]. In addition to PrE markers such as Gata4, Gata6, Sox7, and Sox17, they also express some markers of VE and PE that are not expressed in the PrE, suggesting that they may represent a more mature stem cell type [116, 117]. When reintroduced into recipient embryos, they preferentially contribute to PE rather than VE [116, 118]. Whether this observation indicates that they are biased towards PE derivative or just fail to be properly incorporated into the nascent PrE layer still remains to be determined. XEN cell lines can also be established from ES cells in defined culture conditions [119] and following misexpression of Gata4/6 factors [40]. Contrary to other stem cells, little is known about the extrinsic conditions necessary to maintain and expand them in tissue culture and usually they are derived on mitotic-inactivated fibroblast feeders or in the presence of fibroblast-conditioned medium. Their derivation does not require FGF4 (except within the embryo to specify PrE cells) [35] but requires PDGF signaling [56].

Concluding remarks

Mouse models have been powerful in addressing a broad spectrum of questions in biology. Recent studies showing the key role of RTK signaling, especially FGF, in the divergence between Epi and PrE cell identity, has had a profound impact on the stem cell field by improving the cell culture conditions to isolate and efficiently maintain naive embryonic stem cells. We believe that the study of cell lineage decisions using the mouse as a model system has been and will be instrumental in understanding how these lineages are formed in other mammals, especially humans. However, several reports have shown that the mechanisms involved in PrE/Epi formation in the mouse are not completely transposable to other mammals. Indeed, in bovine embryos, PrE/Epi formation is ERK-dependent but FGF-independent, while in human embryos PrE/Epi formation is not sensitive to ERK and FGF inhibition [120, 121]. These observations warrant studies in other mammalian species to extend our knowledge of how these cell lineages are formed.