Principles of Self-Organization of the Mammalian Embryo

SUMMARY

Early embryogenesis is a conserved and self-organized process. In the mammalian embryo, the potential for self-organization is manifested in its extraordinary developmental plasticity, allowing a correctly patterned embryo to arise despite experimental perturbation. The underlying mechanisms enabling such regulative development have long been a topic of study. In this Review, we summarize our current understanding of the self-organizing principles behind the regulative nature of the early mammalian embryo. We argue that geometrical constraints, feedback between mechanical and biochemical factors, and cellular heterogeneity are all required to ensure the developmental plasticity of mammalian embryo development.

Early embryonic development varies between different species in its detailed manifestation, yet its general paradigm is conserved. The fertilized egg gives rise to descendants through multiple rounds of cleavage divisions. Cellular compositions of subsets of cell populations become distinct as the differentiated cell lineages arise. In turn, reciprocal interactions between lineages establish axial asymmetries and specify regional embryonic structures. Despite the species-specific differences in cellular arrangements and molecular signatures presented at each stage, embryogenesis is always a highly self-organized process that can take place without aid from external cues. Deciphering the developmental principles imbuing these self-organizing properties has been a central goal in developmental biology.

The first days of mammalian embryogenesis present an emerging model to unveil the self-organizing principles of embryo development. During the early stages (i.e., pre-implantation), the mammalian embryo is surrounded by a glycoprotein shell, the zona pellucida, within which the embryo develops from a single cell into a hollow sphere of over a hundred cells, named the blastocyst. The blastocyst is specified into just three cell lineages: the trophectoderm (TE), which is the progenitor for the placenta; the primitive endoderm (PE), which forms the yolk sac; and the epiblast (EPI), which gives rise to the embryo proper. Geometrically, the TE cells form an epithelial monolayer localized on the outside of the embryo that encompasses a fluid-filled cavity, the blastocoel, and the inner cell mass (ICM), a clump of cells that includes all the PE and EPI cells localized to one side of the TE sphere (Bedzhov et al., 2014; Rossant, 2016; Schrode et al., 2013; Zhang and Hiiragi, 2018). The asymmetrically positioned ICM corresponds to the “embryonic pole” and lies opposite the “ab-embryonic pole”, describing the polarity of the blastocyst structure (Figure 1A) (Bischoff et al., 2008; Garbutt et al., 1987; Tam, 2004).

Figure 1. Overview of Preimplantation Development and Fate Decision Processes.

(A) Preimplantation development starts with fertilization, after which the embryos undergo several rounds of cleavage divisions. At the 8-cell stage the embryo becomes compacted and establishes apico-basal polarity. The loss of totipotency commences at the 4-cell stage as different sister cells differentiate the contribution to blastocyst lineages. The first cell fate completes at the 32-cell stage; the second cell fate decision completes at the late blastocyst stage.

(B) In the morula stage embryo (between the 16–32 cell stages), the Hippo signaling component protein AMOT becomes tethered to the apical domain, resulting in the inactivation of Hippo signaling and translocation of Yap and Taz proteins to the nucleus, where they activate Cdx2 and Gata3 expression to specify trophectoderm (TE) fate. In the apolar cells, the lack of apical domain activates Hippo signaling, and the resulting lack of Cdx2 and Gata3 expression allows the cells to express pluripotency factors (such as Nanog and Sox2) to specify inner cell mass (ICM) fate.

(C) During the second cell fate decision, ICM cells express Nanog and Gata6 in a salt-and-pepper manner. The expression of Gata6 prompts the expression of FGF receptor Fgfr1/2, whereas Nanog prompts the expression of FGF. The communication between cells expressing different levels of Nanog and Gata6 by FGF signaling allows the separation of Nanog and Gata6 in two cell populations: the former will specify as epiblast (EPI), whereas the latter will specify as primitive endoderm (PE).

The formation of the three blastocyst cell lineages is achieved by two rounds of cell fate decisions. Each round is regulated by a signaling pathway, the activity of which hinges upon cell non-autonomous factors such as cell position and paracrine signaling. This dependency upon internal signaling activities independent of the external environment provides an important basis for the self-organization potential of the blastocyst. It also provides the foundation for bioengineering strategies to assemble “blastocyst-like structures” that morphologically and biochemically recapitulate the natural blastocyst.

Most of what we know about cell fate decisions in the mammalian embryo comes from studies of the mouse embryo. In the mouse, totipotency—the ability of the single cell to give rise to the entire embryo—becomes gradually lost from the 4-cell stage onward, as fate regulators (herein referred to as lineage transcription factors) segregate their expression to different cells (Goolam et al., 2016; Ralston et al., 2010; Strumpf et al., 2005; Tarkowski and Wróblewska, 1967; Torres-Padilla et al., 2007; Wicklow et al., 2014). The first cell fate decision leads to a segregation of TE from ICM progenitors. An important event to trigger this segregation is the establishment of cell polarization that takes place at the 8-cell stage. At the 8-cell stage, the embryo transforms morphologically from an assembly of loosely attached cells into an integrated entity through a process called compaction (Pratt et al., 1982; White et al., 2016). Concurrent with compaction, each individual cell establishes an apical domain (Johnson and Ziomek, 1981) that organizes canonical apical proteins including the Par complex (Par6-aPKC) and ezrin/radaxin/meosin (ERM) proteins (Louvet et al., 1996; Vinot et al., 2005) restricted by an outer actomyosin ring (Fleming et al., 1986; Ziomek and Johnson, 1980). This apical domain serves as an important signaling organizer to instruct TE and ICM segregation. During the 8–16 stage divisions, the apical domain components are differentially inherited by the daughter cells; cells that inherit apical components from their mother cells reestablish the apical pole and hence become polarized, and cells that don’t become apolar (Anani et al., 2014). In this way, the polarity status categorizes embryonic cells into two distinct populations with differential Hippo signaling, leading in turn to the differential expression of TE or ICM transcription factors to specify cell fate. In apolar cells, Hippo signaling remains active, preventing the downstream transcription factors Yap and Taz from translocating into the nucleus and directly or indirectly enhancing the expression of pluripotency transcription factors Nanog and Sox2, which specify ICM fate. By contrast, the apical domain in polarized cells tethers key Hippo activators, resulting in the translocation of Yap and Taz into the nucleus to induce the expression of Cdx2 and Gata3, which specify cells as TE (Hirate et al., 2013; Kono et al., 2014; Ralston et al., 2010; Strumpf et al., 2005; Wicklow et al., 2014). TE specification is completed at the 32-cell stage as cavitation of the blastocyst begins (Posfai et al., 2017) (Figures 1A and 1B).

Once the first cell lineage segregation is complete, the second wave of cell fate decision initiates. In the expanding blastocyst, some ICM cells secrete the fibroblast growth factor (FGF) ligand Fgf4 (Feldman et al., 1995) while other cells express the receptors Fgfr1 and Fgfr2 (Arman et al., 1998; Molotkov et al., 2017). The resulting receptor-ligand interactions trigger the MAPK signaling cascade to induce the expression of the PE transcription factor Gata6 (Chazaud et al., 2006; Cheng et al., 1998). Gata6 is initially heterogeneously expressed between cells and is co-expressed with EPI transcription factors such as Nanog and Sox2 (Schrode et al., 2014). The subset of cells which are elevated in the expression of Gata6 downregulates Nanog and Sox2, and consequently, these cells become specified as PE. In contrast, those cells with lower Gata6 upregulate Sox2 and Nanog and become EPI cells (Schrode et al., 2014). Cells differentially expressing PE and EPI transcription factors are initially intermingled, forming a salt-and-pepper pattern. Subsequently, the specified PE progenitors express apical polarity proteins and sort to the ICM surface, where they form an epithelial layer (Plusa et al., 2008; Saiz et al., 2013). By the time of implantation, all three lineages are fully committed, so the embryo is ready to escape from zona pellucida and implant into the uterus (Figures 1A and 1C).

As indicated above, an outstanding feature of mammalian embryogenesis is its regulative nature. Classical experimental embryology has shown that blastocyst formation can robustly withstand various experimental perturbations. For instance, adding or depleting cells at the cleavage stages, although it changes the size of the embryo, does not significantly alter lineage composition and still produces correctly patterned blastocysts (Saiz et al., 2016; Tarkowski, 1959). Moreover, the embryo can be de-constructed by separating blastomeres, but as long as the blastomeres are placed back together, the blastocyst structure can be reformed (Morris et al., 2012; Posfai et al., 2017). The ability of the cells to reconstruct the blastocyst is maintained for a long period of time—up to the early blastocyst stage (Posfai et al., 2017; Suwińska et al, 2008). During this time, differentiating lineages display high cellular plasticity; the embryo can sense the loss of cells in one lineage and, accordingly, compensate from the alternate lineage through either asymmetric cell division or trans-differentiation mechanisms.

The mechanisms behind the fascinating regulative nature of the preimplantation embryo have evoked a surge of research interest since their early experimental characterization. The recent use of biophysical, imaging, and mathematical modeling tools is allowing us to decipher the regulatory basis of this phenomenon. In this review, we discuss current understanding of the key principles underlying the self-organization properties of preimplantation development. We argue that the plasticity of preimplantation development entails three cell-non-autonomous factors: (1) geometrical confinement, (2) feedback between biophysical and biochemical signaling, and (3) cellular heterogeneity. These features are also present in other developmental contexts. Therefore, we propose that the early mouse embryo provides a reductive system for the study of self-organization, and the knowledge we bring together from this system can be generalized to a broader range of developmental contexts.

GEOMETRICAL CONSTRAINTS INSTRUCT LINEAGE ALLOCATION

Unlike the embryo at its later stages, the preimplantation embryo lacks pathways inducing collective cell movements, such as convergent extension. Accordingly, the embryo is shaped as a spherical ball, held together primarily by cell adhesion and exhibiting apico-basal polarity. This feature provides the basis for the argument that geometrical factors play an important role in regulating spatial cell allocation, such as the inside-outside—in other words ICM-TE—cell number. In this section we review how geometrical confinement acts together with mechanical factors to regulate cell number in these two lineages and how this accomplishes the self-adjustment of lineage ratios that occurs upon variation in the embryo’s size.

At the morula stage, the presence of the apical domain distinguishes the TE progenitors from those of the ICM. Accordingly, transplantation of the apical domain can convert apolar cells into TE (Korotkevich et al., 2017). Polar and apolar cells are first produced by two rounds of symmetric and asymmetric cell divisions at the 16- and 32-cell stages. However, the un-polarized status of the apolar cells is temporary and can only be maintained if the suppressive cell-cell contact occupies the entire membrane—that is, when the apolar cell is positioned on the inside of the embryo. Thus, both division pattern and cell position conditions have to be fulfilled for ICM fate specification. In reality, the two conditions are not completely independent, as cell position is influenced by the initial cell polarity status (Maître et al., 2016). At the 8-cell stage, the actomyosin complex becomes activated around the cellular cortex, where its contractility generates circumferential cortical tension. The cortical tension is higher in apolar cells and lower in polar cells, as the apical domain antagonizes the membrane localization of the actomyosin complex (Anani et al., 2014; Maître et al., 2016). These tension differences cause the polar cells to flatten against the apolar cells when in physical contact, as principally governed by Laplace’s equation (Maître et al., 2016). The apolar cell therefore tends to internalize when surrounded by polar cells (Maître et al., 2016; Samarage et al., 2015) (Figure 2A). The limited inner space in the 16-cell-stage embryo restricts the number of apolar cells that can be positioned on the inside. Computational simulations on spatial organization of the embryo at this stage are in agreement with experimental observations to suggest that approximately 6 cells can be hosted inside, depending on genetic backgrounds (Anani et al., 2014; Morris et al., 2010; Samarage et al., 2015). Thus, embryo geometry defines the lower limits of the polar to apolar cell ratio. The dependency of TE and ICM lineage assignment on embryo geometry allows outer and inner cell number to adjust according to embryo size, as occurs when cells are added or depleted at the cleavage stages. As the inner cells maintain the potential to reestablish cell polarity until the early blastocyst stage, the deprivation of outer polarized cells would break the restriction of cell position on cell polarity to establish a new balanced state with readjusted lineage ratio (Rossant and Lis, 1979).

Figure 2. Crosstalk between Morphogenesis and Lineage Specification in the Preimplantation Embryo.

(A) At the 16-cell stage, asymmetric cell divisions generate polar and apolar cells. Cortical tension is higher in apolar cells and lower in polar cells. As a result, when an apolar cell is surrounded by polar cells, the tension difference drives the apolar cell to internalize to become ICM.

(B) At the blastocyst stage, the cavity expansion stretches the TE cells, which induces the recruitment of tight junction proteins. The tight junction proteins in turn prompt cavity expansion to form a positive feedback loop. The maximum size of the blastocyst is set by the maximum cavity expansion pressure within which TE layer can suffer.

As cavitation begins, the addition of a hollow space increases the embryo size, and it has been proposed that the size of the cavity influences lineage allocation through feedback between mechanical and biochemical signals. After TE cells form a matured epithelial tissue, a Na⁺/K⁺-ATPase pumps fluid into the extracellular space between the TE cell-cell junctions (Madan et al., 2007), generating multiple micrometer-sized cavities (Dumortier et al., 2019; Ryan et al., 2019). These microcavities form a conjugated network connected by cell contacts. If two connecting cavities differ in size, the pressure imbalance will lead to the transfer of liquid from the smaller to the larger cavity. Thus, the larger cavity preferentially grows at the expense of the smaller one, which is akin to the phenomenon of “Ostwald ripening”(Ostwald, 1896). As a result, the network of microcavities reorganizes ultimately into a single united cavity positioned in the interior of the TE sphere (Dumortier et al., 2019). As fluid is continuously pumped into the cavity, cavity expansion creates pressure that deforms the TE and ICM tissues (Chan et al., 2019). TE cells are stretched laterally, creating mechanical pressure that enhances tight junction protein recruitment to cell contacts. In turn, the recruited tight-junction proteins also prompt the fluid pumping process. Hence, tight-junction protein recruitment and fluid pumping processes give positive feedback, allowing the cavity to expand (Figure 2B). A critical threshold exists for luminal pressure beyond which the blastocyst collapses, setting the maximum size of the blastocyst (Chan et al., 2019). Intriguingly, when this cavity expansion process is perturbed, the resulting blastocyst is smaller and, more importantly, has an increased proportion of ICM cells. The luminal pressure therefore affects the ratio between TE and ICM cells (Chan et al., 2019). The precise mechanism accounting for the effects of cavity size regulation on lineage ratio remains undetermined. It is plausible to speculate that this is achieved through feedback between cell shape and the division pattern. In many systems the cell division has been shown to follow Hertwig’s rule, meaning that the division plane bisects the cells along their short axis (Hertwig, 1884). Cells have also been demonstrated to follow such a rule at the morula stage so that the flattened polar cells preferentially divide symmetrically whereas the columnar-shaped cells divide asymmetrically (Niwayama et al., 2019). It is possible that the reduction of cavity tension relaxes the stretched TE cells, leading to an increase in asymmetric cell division at the blastocyst stage.

Overall, the geometrical properties of preimplantation embryos at different stages are coupled with lineage segregation to allocate cell number in the lineages. This also allows the ratio of the TE and ICM lineages to be balanced upon the loss of cells from either lineage.

REGULATORY FEEDBACK SELF-REINFORCES LINEAGE IDENTITY

While geometrical restrictions allow the embryo to adjust the proportion of outer and inner cells with altered embryo size, the flexibility of cell fate specification requires plasticity at the transcriptional level. For both the first and second cell fate decisions, lineage specification takes about a day. During this time, the differentiating lineages can alter their predestined fate upon changes in the external environment (Eckert et al., 2004; Wigger et al., 2017). Common features found between the first and second cell fate decisions imply that conserved regulatory principles, going beyond specific molecular identities, are shared between these two fate decisions and allow for robust fate bifurcation while engaging cellular plasticity. In this section, we discuss the regulatory logic for the first and second cell fate decisions that allows cells to retain cellular plasticity during robust fate specification processes.

Auto-regulatory Circuits and Mutual Inhibition

Both the first and second cell fate decisions generate binary fate outcomes, enabling progenitor cells to become one cell type or the other, driven by differential sets of lineage-specific transcription factors. The expression kinetics of fate regulators for both fate choices share a similar pattern: when the cell fate decision is commencing, cells show conflicting intermingled expression of lineage transcription factors. How does the expression of lineage transcription factors transition from its initial heterogeneous phase to the stably expressed phase? Promoter analyses using stem cells that are transcriptionally equivalent to blastocyst lineages show that many lineage transcription factors—such as Cdx2, Nanog, Oct4, and Sox2—can activate their own expression by binding to their own promoters, thus forming an auto-regulatory circuit that contributes to the rapid upregulation of their expression (Chen et al., 2009; Loh et al., 2006; Niwa et al., 2005). In addition, promoter analyses combined with transcriptomic assays show that lineage transcription factors form intricate gene regulatory networks. The transcription factors prompting concordant lineage identity positively regulate each other’s expression, whereas those dictating conflicting fates mutually inhibit each other. This regulatory logic holds for TE versus ICM fate separation (Cdx2 versus Oct4 and Nanog) (Chen et al., 2009; Niwa et al., 2005) and the EPI versus PE fate separation (Nanog versus Gata6 and Gata4) (Frankenberg et al., 2011; McDonald et al., 2014). These interactions allow opposing lineage transcription factors to form mutual inhibition loops and enable unique fate outcomes despite initially noisy conditions. Similar network topology has been used in many other developmental contexts, where it effectively improves the robustness of the system to signaling perturbations (Briscoe, 2019; Jaeger, 2011; Olson, 2006; Zuniga, 2015) (Figures 3A and 3B).

Figure 3. Gene Regulatory Networks Control Preimplantation Development.

(A) During the first cell fate decision, the robust separation between TE and ICM lineages are contributed to by the feedback mechanisms between cell polarity and its downstream events, as well as the mutual inhibition between TE and ICM transcription factors

(B) During the second cell fate decision, FGF signaling and Nanog and Gata6 expression form a tri-stable network to produce EPI and PE lineages.

Feedback between Biochemical Signaling Modules

Besides regulatory networks operating at the genomic level, there is also feedback between transcriptional and morphogenetic events and between different steps of fate-related biochemical signaling transduction processes to provide additional checks upon cell fate specification.

At the preimplantation stage, many key developmental events are programmed to take place each at a specific time after fertilization, independent of the embryo’s size or progression through the cell cycle. This intrinsic developmental timing mechanism allows embryos that have altered sizes, such as isolated single blastomeres from early cleavage stages, to form a blastocyst structure and implant into the uterus. This property has been used to test the developmental potential of cleavage stage cells. The robust nature of this developmental timing is best characterized for cell polarization. Although the invariant nature of this clock has long been known, the regulatory basis that allows the embryo to polarize at a specific developmental time, regardless of embryo size, has remained unclear. A recent study that combined embryological and biophysical methods to address this question revealed that the intrinsic developmental timer constitutes a link between transcription and cell-autonomous apical protein dynamics. Specifically, it has been shown that apical domain formation requires two critical conditions: zygotic genome activation (ZGA), which in the mouse embryo occurs at the 2-cell stage, and the cortical localization of the actomyosin complex (M.Z. and M.Z.-G., unpublished data). ZGA activates expression of two transcription factors, Tfap2c and Tead4, the expression of which is sufficiently high at the 8-cell stage to induce the expression of downstream targets that remodel the cytoskeleton (Figure 4A). The ensuing cytoskeleton remodeling events allow apical proteins to be recruited to the cell contact free surface. However, the rate of apical protein recruitment is not constant over the whole cell contact free surface, but instead correlates with local apical protein concentration. When a region has low apical protein concentration, the local apical proteins dissipate rather than accumulate, whereas in regions where the apical protein concentration exceeds a threshold concentration, apical proteins accumulate to contribute to the apical domain. This phenomenon is termed “cooperative recruitment.” The threshold concentration is linked to the levels of Tfap2c and Tead4 (M.Z. and M.Z.-G., unpublished data), and ectopic expression of Tfap2c and Tead4 allows the cooperative recruitment to initiate prematurely, resulting in apical protein polarization at the 4-cell stage (M.Z. and M.Z.-G., unpublished data). However, although cooperative recruitment can polarize apical proteins, formation of the final cap shape of the apical domain also requires a “lateral spreading” behavior of the apical proteins (Figure 4B), which allows the apical domain to expand and occupy a major part of the cell contact free surface (M.Z. and M.Z.G., unpublished data). At the molecular level, this apical lateral mobility is regulated by protein kinase C (PKC), Rho guanosine triphosphatases (GTPases), and the actomyosin signaling axis (Zhu et al., 2017). As proof of principle, it has been shown that the premature induction of Tfap2c-Tead4 and RhoA can reset the timing of embryo polarization to allow the cell to establish the apical domain one cell cycle earlier—at the 4-cell stage (M.Z. and M.Z.-G., unpublished data) (Figure 4C). Thus, the timing of apical domain formation provides a case study for how intrinsic transcriptional events can regulate cell-intrinsic cytoskeletal behavior to control the robust timing of a major morphogenetic transition.

Figure 4. The Regulation of Timing of Cell Polarization.

(A) Formation of the apical domain is regulated by two key conditions: (1) zygotic genome activation (ZGA), which activates the expression of two transcription factors—Tfap2c and Tead4; and (2) the Rho GTPases activity. The activities of Tfap2c, Tead4, and Rho GTPases are high at the 8-cell stage when they induce the formation of the apical domain.

(B) Tfap2c and Tead4 regulate the cooperative recruitment of apical proteins, allowing apical proteins to concentrate as a cluster, whereas Rho GTPases signaling allows the apical proteins to spread laterally in order to form the cap shape of the domain.

(C) Overexpression of Tfap2c, Tead4, and RhoA advance the timing of cell polarization to the 4-cell stage. Scale bars, 15um.

Besides crosstalk between transcriptional and morphogenetic events, the feedback between biochemical signaling modules is crucial for cell fate specification. During the first cell fate decision, cell polarity and cell position instruct Hippo activity to determine expression of cell-fate-related transcription factors. On the other hand, Hippo activity also regulates cell polarity and cell position. Cell polarity inhibits Lats kinase activity in the outer cells. This low Lats kinase activity in turn helps maintain the polarized state of the cell, as forced upregulation of Lats kinase can reverse the cell polarity status and relocate cells to the interior (Frum et al., 2018). In this way, the Hippo pathway interacts bi-directionally with cell position to regulate cell fate at the morula stage. Intriguingly, it has been observed that Yap translocates to the nucleus even before cells become polarized, suggesting polarity-independent pathways also regulate Yap localization (Hirate et al., 2015). Our recent study showed that this nuclear translocation of Yap results from the increased zygotic expression of Tead4 (M.Z. and M.Z.-G., unpublished data).

The Tead4-Yap complex at this stage not only contributes to cell polarity establishment but also inhibits the premature expression of the ICM transcription factor Sox2 (Frum et al., 2019). Thus, cell polarity, Hippo signaling, and cell position all feed back upon each other in both spatial and temporal dimensions to establish a robust fate specification program. In addition, the resulting expression of fate-specifying determinants also forms direct feedback with cell polarity, as overexpression of Cdx2 prompts the enrichment of apical proteins, which drive cells to divide symmetrically to produce more TE cells (Jedrusik et al., 2008). Cdx2 transcripts are directly tethered to the apical domain, providing an additional pathway to restrict the Cdx2 expression to polarized cells (Skamagki et al., 2013). Together these reciprocal feedbacks between polarity and Hippo signaling enhance the cell fate decision outcome (Figure 3A).

Feedback between the dominant signaling pathway FGF-mitogen-activated protein kinase (FGF-MAPK) and its downstream transcriptional outcomes becomes essential for bifurcation of the PE and EPI lineages. Here, this feedback is not only important for the unambiguous emergence of the PE and EPI cell populations, but also for the scaling of lineage composition to ICM size (Morris et al., 2013; Saiz et al., 2020). On the one hand, expression of Fgf4 is promoted by Nanog and inhibited by Gata6; on the other hand, expression of FGF receptors Fgfr1 and Fgfr2 is induced by Gata6 but not Nanog. In this way, the levels of lineage transcription factor expression relate to the ratio of the FGF ligand to receptors in each cell. The levels of FGF receptors correlate with intracellular MAPK signaling and consequent Gata6 expression to form a positive feedback loop. The regulatory interactions, including (1) FGF ligand and Nanog, (2) positive feedback between FGF receptors and Gata6, and (3) the mutual inhibition between Nanog and Gata6, form a tri-stable network that, according to a mathematical simulation, is sufficient to produce Nanog and Gata6 positive cells in a ratio similar to the real embryo (Bessonnard et al., 2014). Computational simulation suggests that the cell non-autonomous regulation mediated by the trans-interaction between FGF ligand and receptor can sufficiently account for the “lateral-inhibition-like” phenomenon in the establishment of the salt-and-pepper patterns for Nanog- and Gata6-positive cells (Saiz et al., 2020) (Figure 3B).

Together, these various feedback loops enable the embryo patterning process while allowing patterning to adapt to alternations in embryo size so as to achieve the regulative nature of development.

Apoptosis-Mediated Cell Clearance Guards Patterning Accuracy

For both the first and second rounds of cell fate specification, the resulting two lineages are segregated in space. In some cases, cell sorting is not carried out perfectly, leaving a few cells in the wrong position. Such mis-sorting is relatively rare for the TE and ICM segregation but more common for the second cell fate decision. In both cases, the mis-sorted cells retained in the EPI will eventually be cleared out by apoptosis, around the late blastocyst stage (Frum et al., 2018; McDole and Zheng, 2012; Meilhac et al., 2009; Plusa et al., 2008). Cell death is a consequence of mis-positioning, as “artificially induced mis-sorting”, such as that resulting from enforced Cdx2 expression in inside cells or ectopic upregulation of Sox2 in outside cells also leads to apoptosis (Frum et al., 2018; Skamagki et al., 2013). How the incompatibility between the cellular transcriptional profile and cell position act to induce apoptosis is not clear. It is likely that different lineages are coded by specific cell surface proteins, the homophilic interaction of which is essential for cell survival; alternatively, the presence of the apical polarity complex on the cell membrane is required for survival of TE and PE cells. Overall, this apoptosis-mediated correction mechanism ensures the maintenance of homogeneity of the EPI and the correct patterning of blastocyst lineages. Interestingly, such a “checkpoint” of EPI homogeneity not only detects a mis-expressed lineage signature, but also chromosome abnormalities. It has been shown that when aneuploid cells are present in the EPI of the late blastocyst, they become selectively eliminated by apoptosis (Bolton et al., 2016; Singla et al., 2020), allowing only normal cells to survive in the embryo. Aneuploid cells often express correct lineage markers at the time of their elimination, suggesting a different mechanism may be involved in cell clearance. Overall, the self-correction mechanism allows normal developmental progression despite various challenges.

CELLULAR HETEROGENEITY LEVERAGES LINEAGE PATTERNING

Development of the mammalian embryo is accompanied by cellular heterogeneity from the earliest developmental stages. As discussed in the previous section, the expression patterns of fate regulators such as Nanog, Cdx2, and Gata6 show a significant degree of heterogeneity toward the beginning of the fate decision processes (Dietrich and Hiiragi, 2007). The level of their expression at this stage is to a certain extent indicative of fate outcomes (Posfai et al., 2017; Xenopoulos et al., 2015). The causes of the heterogeneous expression of these fate regulators are not yet completely clear, but it is reasonable to think that they result from the antecedent heterogeneous activity of upstream regulators. Indeed, several studies over recent years have identified a number of molecules that are either heterogeneously expressed or active as early as the 2- or 4-cell stage, and some of these factors have been experimentally validated to have an impact on Cdx2 or Nanog expression at the 8-cell stage. One such essential protein is a histone arginine methyltransferase called Carm1 (Chen et al., 1999). Carm1 is responsible for R26 methylation of histone H3 (H3R26me). In the mouse embryo, Carm1 localizes to a type of nuclear body called the paraspeckle. The number of paraspeckles, and hence Carm1 abundance and the level of H3R26me, is heterogeneous in different blastomeres at the 4-cell stage (Hupalowska et al., 2018). The heterogeneity of Carm1 expression biases TE-ICM fate decision, high Carm1 expression directs fate to ICM, and low Carm1 expression promotes differentiation into TE. Carm1 instructs cell fate decision by promoting the expression of pluripotency factors such as Sox2 and Nanog (Goolam et al., 2016; Torres-Padilla et al., 2007) and a recently characterized fate regulator Sox21 (Goolam et al., 2016; Torres-Padilla et al., 2007). Carm1 also affects cell polarity by activating the expression of a dominant-negative form of atypical protein kinase C (aPKC), which interferes with the activity of Par complex to impair the TE fate specification (Parfitt and Zernicka-Goetz, 2010). A recent study found that Carm1 also regulates the keratin filament assembly (Lim et al., 2020), which becomes heterogeneously expressed at the 8-cell stage. These keratin filaments contribute to the cortical tension in the polar cells and consequently the Yap nucleus localization. The expression of keratin proteins is controlled by Carm1 effector BRG1- or BRM-associated factors (BAF) complex in the mouse embryo (Panamarova et al., 2016). BAF155 up-regulates keratin expression in Carm1-low cells, and in Carm1-high cells, BAF155 becomes methylated and unable to activate keratin proteins’ expression (Lim et al., 2020; Parfitt and Zernicka-Goetz, 2010). These different routes bridge Carm1 expression with first lineage specification (Figure 5).

Figure 5. Heterogeneities in Fate Regulators Control Differential Fate Outcomes.

Heterogeneous levels of fate regulators at the 2- and 4-cell stages contribute to the heterogeneous expression of lineage transcription factors at the 8-cell stage to influence the fate outcomes of sister blastomeres. The non-coding RNA lincGET is expressed unequally between the sister blastomeres at the 2-cell stage and contributes to the heterogeneous activity of Carm1 at the 4-cell stage. The Carm1 activity in turn positively regulates the expression or DNA binding dynamics of pluripotency factors (Nanog and Sox2), whereas it inhibits Cdx2 expression, and as a result, the cells that have high Carm1 activity preferentially contribute to ICM, whereas the cells that have low Carm1 activity contribute to TE.

What leads to Carm1 heterogeneity at the 4-cell stage? A recent study identifies the non-coding RNA LincGET as acting upstream of Carm1 to regulate its localization. LincGET expression starts at the 2-cell stage, but unequally between the two sister blastomeres, where it physically interacts with Carm1 to promote its nuclear localization. Accordingly, overexpression of LincGET also results in a fate-bias toward ICM (Wang et al., 2018). The upstream signaling regulating heterogeneous expression of LincGET is yet to be identified but possibly relates to the heterogeneous timing of ZGA between two sister cells (Figure 5). If it does, this raises the question of what regulates ZGA and whether this process per se might be responsible for the initiation of lineage segregation.

Although it is clear that certain fate-related players can be heterogeneously expressed prior to the cell fate decision processes, how heterogeneity is beneficial for embryo development remains to be discovered. As of yet this question has not been directly tested by experiments, but insights can be found from other systems. As discussed in the previous sections, expression of one lineage regulator must outcompete the other in order to attain each of the first two cell fate decisions. Thereafter, mutual inhibition and positive feedback loops are deployed to allow the stabilized expression of factors characteristic of one lineage. In this regard, the initially heterogeneous condition may expedite the symmetry breaking process to ensure lineage specification in a restricted developmental time-course. Such an effect has been illustrated in tissue patterning processes, where initially heterogeneous transcription factor expression can be required for the formation of discrete lineage domains (Chang et al., 2008). In addition, it is plausible that the heterogeneous expression may allow cells to respond rapidly to damaging conditions such as the depletion of a specific lineage at an early stage of lineage separation.

RECAPITULATING BLASTOCYST FORMATION IN VITRO USING EMBRYONIC STEM CELLS

Stem cell lines that recapitulate the developmental potential of each blastocyst cell lineage have been derived and propagated in vitro (Kunath et al., 2005; Martin, 1981; Tanaka et al., 1998). The simple organization and regulative nature of mouse embryogenesis have led to the idea of using lineage-equivalent stem cells to reconstruct the embryo. As the first proof-of-principle, it was shown that when embedded in the extracellular matrix, embryonic stem cells (ES cells)—which represent the EPI—can mimic development of the blastocyst’s ICM to undertake polarization and lumen opening, as occurs during implantation (Bedzhov et al., 2014). In the same way, it is possible to use human ES cells to mimic development of the human blastocyst ICM behavior (Shahbazi et al., 2017). Co-culture of ES cells with trophoblast stem cells (TS cells, which represent the TE) in mechanically and biochemically appropriate conditions leads to their self-assembly and self-organization into embryo-like structures (Harrison et al., 2017; Rivron et al., 2018). These structures resemble the gross morphology of natural embryos as well as certain aspects of their spatial gene expression pattern. Although the embryo-like structures generated in this way can recapitulate the hollow morphology of the blastocyst (so-called blastoids), they often lack the layer equivalent to the PE, presumably due to the limited ability of ES cells to trans-differentiate into endodermal lineages in the culture conditions of the study. Two subsequent studies addressed this issue by using a new type of stem cell, so-called extended pluripotent stem (EPS) cells, to generate blastoids (Li et al., 2019; Sozen et al., 2019). EPS cells have been shown to be able to contribute to extra-embryonic as well as embryonic lineages in an embryo-chimera assay (Yang et al., 2017). The modified and enhanced blastoid models not only retain a blastocyst-like morphology, but also recapitulate several earlier morphological processes of blastocyst formation, including the establishment of the apico-basal polarity of the outer cells before cavitation and the recruitment of tight junctions (Li et al., 2019). Although the “TE layer” differentiated from EPS cells is topologically similar to the TE of the real blastocyst, transcriptional analysis indicates that it does not recapitulate pre-implantation TE at a transcriptional level (Posfai et al., 2020). It is possible that as a result of this discrepancy, the transferred EPS blastoid does not “develop” properly into post-implantation like embryo structures in vivo (Li et al., 2019). These results thus suggest a requirement for the identification of a better stem-cell type to mimic the preimplantation TE lineage. This will be one of the important future directions to improve blastocyst-like models. Several human extra-embryonic stem cell lines have been created, either by direct derivation from human embryos and tissues or by induction with growth-factors (Linneberg-Agerholm et al., 2019; Okae et al., 2018; Xu et al., 2002), and they hold great potential for creating human synthetic embryo models in the future. These blastoid models reveal that the formation of the hollow blastocyst morphology is extremely robust and highly resistant to transcriptional variations. Such robustness of blastocyst formation helps maximize the potential for implantation and so could help those embryos in which lineage specification has not fully completed to implant in time. This could be of potential benefit to embryos that show less heterogeneity at earlier stages and thus may be delayed in committing to specific lineages. Mechanistically, the feedback loops between mechanical and biochemical signaling that we discuss above can contribute to the robustness of blastocyst morphogenesis in spite of transcriptional variation. Taken together, these emerging synthetic stem-cell models provide solid proof of the self-organization principles of embryogenesis and bring deeper understanding of the self-organizational capacity of blastocyst formation.

CONCLUSIONS AND PERSPECTIVES

Preimplantation development results in the construction of a humble structure—the blastocyst, the formation of which exploits simple principles and processes. An integral part of blastocyst self-assembly is that it is a robust process that can withstand a variety of experimental perturbations. We argue that such developmental plasticity relies principally upon intrinsic feedback interactions between mechanical and biochemical factors. On top of such feedback loops, the cellular heterogeneity in expression of fate regulators may provide additional benefits to ensure correct lineage patterning. Although the application of mathematical and biophysical tools has increased our understanding of the principles of self-organization utilized during early development, many questions remain unanswered: what, for example, allows cells to sense the mismatch between cell position and cell fate and to invoke the intracellular apoptosis program? Is cellular heterogeneity required for developmental progression? The response to cell non-autonomous factors during the activation of lineage transcription programs provides an important basis for developmental plasticity, but what determines the duration of the phase in which cells can be sufficiently flexible to change their fate in response to altered external cues? Addressing these questions will further unveil the self-organizing principles of mammalian embryo development and help to optimize the establishment of in vitro stem-cell models to build artificial embryos that hold great promise in the study and treatment of infertility.