Primitive means first, not worst: Critical roles for primitive endoderm in embryos and embryo models

Abstract

In mammals, extraembryonic tissues, such as the placenta and yolk sac, are the first cell types to be specified during development because they enable the embryo to take residence and thrive in the uterine environment. Among extraembryonic tissue types, primitive endoderm (PrE), which will eventually contribute to the yolk sac, is especially fascinating. The PrE itself is named for functioning like the embryo’s original gut-like tissue. For many years, our understanding of the PrE was limited by the intrinsically challenging nature of accessing and observing this tissue. However, pioneering studies in mouse have gradually revealed that the PrE is more than just nutritive in function. In fact, the PrE lineage gives rise to signaling centers that oversee developmental processes within the fetus – through processes that are very likely conserved between rodents and primates. Thus, understanding the stages between PrE and yolk sac promises clinically relevant models, including stem cell embryo models, which could lead to enhanced success for in vitro fertilization (IVF). Here, we examine the functions of PrE in the context of embryos, stem cells, and embryo models.

1. Primitive endoderm: origin and destiny

During development, cell fates are gradually established though complex regulatory mechanisms. In many species, including fly, fish, and frog, development outside of the maternal environment means body planning and axial patterning happens first. However, in mammals, including humans, development within the maternal environment means that building the extraembryonic tissues is the first priority. These include placenta, yolk sac, amnion, and umbilical cord. Establishment of these lineages therefore originates in early embryo with very little or no input from the mother. The primitive endoderm (PrE) is one of the first extraembryonic lineages to form during mammalian development. In humans, mice, and other mammals, two major cell fate decisions occur during preimplantation development (embryonic days, E0.5-E4.5). Notable similarities exist between mouse and primate extraembryonic tissue development (Fig. 1A), however, the molecular mechanisms guiding their development have been more comprehensively studied in mice.

Fig. 1. Notable morphological similarity between mouse and human extraembryonic tissues.

A) Three lineages are specified in the mouse and human blastocyst, including the EPI, which will develop into fetal lineages, and two extraembryonic tissues. The PrE and hypoblast are thought to be equivalent between the two species. B-C) At later stages, TE differentiates into multiple tissue types (referred to here by the catch-all “trophoblast”). The PrE differentiates in visceral and parietal endoderm, where visceral endoderm remains in contact with the EPI in both species. Ultimately, parietal endoderm lineages degenerate, while visceral endoderm gives rise to yolk sac structures, which differ greatly between the two species. In spite of these ultimate differences, many features of PrE development appear conserved early in development between mouse and human.

In mice, the selection of PrE and EPI fates is an iterative process, driven by initially random signaling differences among ICM cells, that are reinforced through feedback loops (Fig. 2A). That is, a random subset of ICM cells upregulates expression of FGF4 prior to overt specification of PrE/EPI, around E3.25 (Ohnishi et al., 2014). Soon after, ICM cells that transduce high levels of FGF signaling via MAPK become PrE, while ICM cells with lower FGF/MAPK signaling become EPI by E3.75 (Chazaud et al., 2006; Nichols et al., 2009; Yamanaka et al., 2010). Around this time, PrE cells upregulate expression of both FGFR1 and FGFR2, while EPI cells express only FGFR1 (Kang et al., 2017).

Fig. 2. PrE signaling before and after implantation.

A) During preimplantation development, spontaneous differences in FGF4/FGFR1,2/MEK signaling sort PrE from EPI. B) After implantation and prior to gastrulation, the anterior visceral endoderm (AVE) secretes antagonists of BMP/Nodal/WNT signaling to direct anterior/posterior patterning.

Several mechanisms appear to work alongside FGF/MAPK signaling to boost its impact in PrE specification. For instance, chemical inhibition of p38/MAPK prior to E3.75 disrupts expression of PrE genes (Thamodaran and Bruce, 2016). Similarly, preventing formation of the fully expanded blastocoel cavity disrupts PrE formation, which is thought to be important for concentrating FGF4 ligand inside the blastocyst (Ryan et al., 2019). Interestingly, chemical inhibition of p38/MAPK also disrupts blastocoel expansion (Bell and Watson, 2013; Maekawa et al., 2005), suggesting that p38/MAPK indirectly impacts FGF4 signaling via blastocoel expansion. Consistent with this, supplying exogenous recombinant FGF4 is sufficient to rescue PrE gene expression in p38/MAPK-inhibited embryos (Bora et al., 2021).

FGFR/MAPK signaling subsequently causes upregulation of PrE factors GATA6 and SOX17 and repression of EPI transcription factors NANOG and SOX2 (Chazaud et al., 2006; Frankenberg et al., 2011; Niakan et al., 2010; Plusa et al., 2008; Wicklow et al., 2014). Cell fate commitment is further reinforced by the mutual transcriptional repression between GATA6 and NANOG (Allègre et al., 2022; Bessonnard et al., 2014; Frankenberg et al., 2011; Schrode et al., 2014), as well as Leukemia Inhibitory Factor (LIF) signaling via JAK/STAT (Morgani and Brickman, 2015) and Platelet-Derived Growth Factor (PDGF) receptor signaling (Artus et al., 2010) via PI3K (Molotkov and Soriano, 2018). By E3.75, PrE cells are distributed in an unpatterned manner throughout the ICM. PrE sorting into a coherent layer, and continued differentiation, occurs during implantation and thereafter (Fig. 2A) (Bassalert et al., 2018).

2. The visceral endoderm patterns the epiblast

Following uterine implantation, between E4.5-E6.5 in mice, the PrE gives rise to the visceral and parietal endoderm (Fig. 1B). The parietal endoderm, together with trophectoderm-derived trophoblast giant cells, form the parietal yolk sac, which encapsulates the embryo during early postimplantation stages. The parietal yolk sac is a temporary structure that assists in exchanging nutrients from the maternal environment before the placenta is fully formed (Matsuo et al., 2022).

Just inside the parietal yolk sac, the visceral endoderm closely ensheathes the embryo and adjacent, trophoblast. The visceral endoderm, which is in direct contact with the EPI, is thus appropriately positioned to send regionally specialized signals to the overlying EPI. The visceral endoderm then provides patterning information to the EPI, regulating germ cell gene expression (de Sousa Lopes et al., 2004; Ohinata et al., 2009; Senft et al., 2019; Ying and Zhao, 2001), neural patterning (Thomas and Beddington, 1996), umbilical cord development (Downs, 2022), and primitive streak (site of gastrulation) formation, as discussed further below.

In fact, asymmetric patterning of the visceral endoderm is thought to precede the asymmetry in the EPI that will demarcate the primitive streak. This is thought to occur through a stepwise process involving first the specification of distal visceral endoderm (the very bottom of the egg cylinder, as conventionally oriented), which then migrates in the direction of the soon-to-be anterior region, becoming the Anterior Visceral Endoderm (AVE) (Fig. 2B). The AVE is defined by expression of specific signaling molecules (Fig. 2B) (Stower and Srinivas, 2014; Takaoka et al., 2011, 2017). These include the Wnt antagonist Dkk1, Nodal/BMP antagonist Lefty1, and BMP antagonist Cer1, which repress these pathways within the anterior EPI (Belo et al., 1997; Biben et al., 1998; Kimura-Yoshida et al., 2005; Mohammed et al., 2017; Nowotschin et al., 2019; Yamamoto et al., 2004). Concomitantly, the primitive streak forms, and is restricted to, the EPI region of relatively high Wnt, BMP, and Nodal signaling opposite the AVE.

Interestingly, the BMP antagonists Chordin and Noggin – both famous for their roles in frog and fish axial patterning (De Robertis and Sasai, 1996; Zinski et al., 2018), appear to be coexpressed within an anterior region of the visceral endoderm (Bachiller et al., 2000). However, this interpretation is at odds with those of the authors, who concluded that Chordin and Noggin are not expressed in the AVE but rather in the EPI itself (Bachiller et al., 2000). Resolving the identity and function of AVE signaling molecules in mice is a crucial first step toward understanding human developmental patterning processes.

3. Extraembryonic endoderm – the final act

As gastrulation gives way to organogenesis, the visceral endoderm in contact with the epiblast (called emVE, where em is short for embryo) can intermix with definitive endoderm and contribute to the embryo’s gastrointestinal tract (Kwon et al., 2008; Nowotschin et al., 2019), if even transiently (Batki et al., 2024). Additionally, emVE near the streak transitions through a mesoderm-like state to eventually contribute to placenta and umbilical cord (Rodriguez and Downs, 2017). The many paths and possible roles for all of these cell types throughout embryonic stages is still an area of active investigation.

Meanwhile, visceral endoderm that encapsulates trophoblast (exVE, where ex is extraembryonic) is thought to become the yolk sac endoderm. The yolk sac is the first site of hematopoiesis during development, which can be observed as early as E8.5, prior to formation of liver or bones – the later sites for hematopoiesis (Palis and Yoder, 2001). As the mouse embryo undergoes the process of turning, from the outwardly splayed, bellyflop orientation, to an inwardly curled, fetal position, it drags the yolk sac and associated membranes along, ultimately enclosing itself inside its own protective extraembryonic bubble (Fig. 1C). By contrast, human yolk sac develops inside the amnion as a vestigial waste sac.

4. What we know about human PrE

Disorders related to the PrE can have significant medical implications, including early pregnancy loss and congenital abnormalities (Ferrer-Vaquer and Hadjantonakis, 2013). Defective PrE development or function can lead to improper yolk sac formation, often associated with miscarriage, and can also be linked to rare diseases and certain types of tumors, such as yolk sac tumors (Chousal et al., 2024). Understanding the mechanisms underlying PrE-related disorders is essential for developing diagnostic and therapeutic strategies to improve early embryonic health and pregnancy outcomes.

From a comparative perspective, the PrE lineage shows significant variation across species and important similarities (Goh et al., 2023; Niakan et al., 2012; Ross and Boroviak, 2020). One major difference between primates and rodents is the prominence and location of the mature yolk sac structure. Whereas the yolk sac encircles the amnion and fetus in mice, human yolk sacs are positioned outside of and adjacent to the amnion/fetus capsule (Fig. 1C). This has often led to the assumption that earlier PrE functional roles must not be conserved between humans and primates. The difficulty of obtaining healthy human embryo samples creates a significant barrier to investigating the roles of the human PrE lineage.

Among embryo stages, preimplantation stages are relatively accessible because they can be created by IVF and cultivated in vitro. Such studies have shown striking similarities between mouse and human during preimplantation. As in mice, human PrE (commonly referred to as hypoblast) differentiates from the ICM during the blastocyst stage, expressing many of the same critical lineage-specific transcription factors, including GATA6, SOX17, NANOG, and SOX2 (Blakeley et al., 2015; Boroviak et al., 2015; Gerri et al., 2020; Kuijk et al., 2012; Niakan and Eggan, 2013; Petropoulos et al., 2016; Roode et al., 2012). Curiously, the role of FGF/ERK signaling in PrE specification was initially reported to be dispensable in human blastocysts (Kuijk et al., 2012; Roode et al., 2012). However, this conclusion was recently challenged by studies using higher concentrations and different kinds of FGF/ERK signaling inhibitors, which disrupted PrE gene expression (Dattani et al., 2024; Simon et al., 2024). Therefore, PrE-inducing signals appear to be evolutionarily conserved between mouse and human embryos.

Considerably less is known about human development just following implantation owing to a shortage of samples. Moreover, the quality of human embryo samples available for research study is uncertain. Unlike in mice, early development in humans is thought to be highly error-prone (Currie et al., 2022; Kort et al., 2016). Rare samples have been recovered from medical procedures, including hysterectomy (Hertig et al., 1956) or miscarriage (Dittmar and Mitchell, 2016; Withycombe, 2015). Alternatively, in vitro methods that enable human blastocysts to attach and survive to later stages have been established to simulate the implantation process and enable examination of postimplantation stages (Deglincerti et al., 2016; Shahbazi et al., 2016). With the caveat that all human embryo samples, regardless of source, can deviate from the ideal “normal” developmental path, morphological examination of human specimens suggests that the hypoblast lineage remains in close contact with the epiblast during early post-implantation stages (Fig. 1B). This strongly suggests a conserved role for the PrE as a signaling center to organize epiblast patterning.

Consistent with this, cells with AVE-like transcriptional profiles have been observed in human postimplantation embryos, by single-cell RNA-sequencing (Molè et al., 2021; Zhu et al., 2023). One such study, working in cultured human embryos, identified cells that coexpressed CER1, DKK1, LEFTY1/2, and, notably, NOG – the human orthologue of mouse Noggin (Molè et al., 2021). Limited spatial analysis suggested that this group of cells localized in the anterior hypoblast. Another group also reported enrichment of CER1 in the anterior hypoblast (Zhu et al., 2023), using published RNA-seq data obtained from a single, non-cultured human postimplantation embryo (Tyser et al., 2021) and cultured postimplantation human embryos (Zhou et al., 2019). These observations strongly suggest an evolutionarily conserved role for the PrE lineage in organizing axial patterning. However, a complete molecular characterization and functional testing of the human anterior hypoblast is still needed. This calls for robust models of embryogenesis – with stem cell models presenting a captivating avenue forward.

5. Mouse embryo-derived stem cells paved the way for new knowledge and technology

Stem cell lines provide a powerful, renewable tool for research exploration, as well as proof of the principle that similar lines could be established for the study of human development. Stem cells are defined by the abilities to self-renew and differentiate. Stem cell lines have been derived from all three lineages of the mouse blastocysts, including embryonic stem (ES) cells from the EPI (Evans and Kaufman, 1981; Martin, 1981), trophoblast stem (TS) cells from the trophoblast (Tanaka et al., 1998), and extraembryonic endoderm stem (XEN) cells from the PrE (Kunath et al., 2005) (Fig. 3A). Each of these three stem cell lines retains significant transcriptional similarity to its counterpart lineage in the blastocyst, and each can contribute to its respective lineage when injected into blastocysts. Mouse ES cells are truly pluripotent as they can give rise to the entire fetus when complemented by extraembryonic cell types (Nagy et al., 1993; Poueymirou et al., 2007). By contrast, TS and XEN cells are multipotent as they can only contribute to trophoblast and yolk sac lineages, respectively (Kunath et al., 2005; Tanaka et al., 1998). Notably, each type of stem cell can differentiate primarily into derivatives of their respective origins, indicating that mechanisms exist to maintain each cell line’s lineage restriction.

Fig. 3. The blastocyst is a menagerie of stem cell progenitors.

A) Three types of stem cell can be derived from the mouse blastocyst – each bearing lineage-appropriate characteristics of morphology, gene expression, and developmental potential. Thus, the blastocyst contains stem cell progenitors, but no stem cells. B) Within ES cells, a spectrum of pluripotency exists, and can be cultivated in distinct signaling environments.

Mouse ES cells enabled significant discoveries in developmental biology, including mouse genome editing (Papaioannou, 2024), the establishment of human ES cells (Thomson et al., 1998), and somatic cell reprogramming to produce induced pluripotent stem (iPS) cells in mouse and human (Takahashi et al., 2007; Takahashi and Yamanaka, 2006; Yu et al., 2007). Critically, the study of mouse ES led to the notion that pluripotency is not so much a single, discrete state but a spectrum of developmental states (Fig. 3B). Conditions that capture the developmental progression of pluripotency, from naïve to primed, have been described for both mouse and human pluripotent stem cell lines (PSCs) (Smith, 2024). Importantly, naïve pluripotency is more transcriptionally similar to the preimplantation EPI, while primed pluripotency better resembles a slightly later embryonic stage (Boroviak et al., 2015). The ability to capture these diverse pluripotency states has been fundamental in deriving XEN and PrE-like stem cell types from PSCs, as discussed later in this review.

Protocols to relax the embryonic/extraembryonic lineage restriction and expand the developmental potential of PSCs to a state of apparent totipotency have been summarized elsewhere (Azagury and Buganim, 2024; Baker and Pera, 2018; Riveiro and Brickman, 2020). However, no protocol has supported the clonal formation of a blastocyst from a single starting cell – an ability demonstrated by some embryonic cells – and so this ultimate test of totipotency has yet to be passed. Nevertheless, measuring the ability of various stem cell lines to autonomously “self--organize” into embryo-like shapes in vitro presented an obvious opportunity to explore concepts of developmental potential.

6. The first stem cell models of embryogenesis

Studying human development is challenging for both practical and ethical reasons. Nevertheless, a detailed mechanistic understanding of the processes guiding human development is fundamental for reproductive health. One approach to circumvent these challenges is to model human embryogenesis using human stem cell lines. Stem cell models of mammalian development have many advantages, such as being scalable and genetically modifiable. However, a plate of cells growing in a two-dimensional monolayer does not realistically recapitulate in vivo development. This has led to the creation of more sophisticated, three-dimensional Embryo Models (EMs) – first, mouse and then human. Multiple protocols are available to generate EMs from different mammalian stem cells, with varying degrees of developmental success. The continued optimization of EM generation will help to enable the discovery of fundamental developmental processes that may have significant applications in both research and clinical settings.

The first EMs were made by growing mouse ES cells in a suspension culture lacking self-renewal factors to form embryoid bodies (Brickman and Serup, 2017; Doetschman et al., 1985; Martin and Evans, 1975). Remarkably, embryoid bodies formed spontaneously, often exhibiting sorting and differentiation into germ layers and occasionally demonstrating symmetry-breaking along an apparent anteroposterior axis (ten Berge et al., 2008). Nevertheless, the rarity of embryoid bodies resembling in vivo-grown embryos suggests suboptimal conditions preventing pluripotent stem cells from autonomously and robustly achieving the fetal form – possibly the extraembryonic tissues themselves.

A classical study reported the formation of a visceral yolk sac in about half of the embryoid bodies (Doetschman et al., 1985). This would seem to be at odds with the observation that ES cells do not contribute to extraembryonic tissues in chimeras (Beddington and Robertson, 1989). However, the latter study noted that ES cells could sometimes contribute to extraembryonic tissues with low efficiency. This is consistent with more recent discovery that cultured ES cells can, with low efficiency, show hallmarks of PrE (Canham et al., 2010; Niakan et al., 2010). Moreover, the biological context, whether in culture or chimera, could influence ES cell developmental potential. Consistent with this, yolk sacs have also been observed in human embryoid body-derived organoids (Tamaoki et al., 2023). These observations suggest that combining ES cells with extraembryonic stem cell lines would better recapitulate development than ES cells alone. However, the incorporation of TS cells into embryoid bodies did not promote gastrulation (Harrison et al., 2017). Moreover, PrE cells did not form in these aggregates, suggesting that incorporating PrE-like stem cells might improve EM quality.

7. Calling all extraembryonic cell types

A variety of mouse and human PrE models have been reported (Perera and Brickman, 2023). The first were XEN cell lines, derived from mouse blastocysts (Kunath et al., 2005) and post-implantation embryos (Lin et al., 2016). Overexpression of PrE-specific transcription factors (Fujikura et al., 2002; McDonald et al., 2014; Wamaitha et al., 2015), and the addition of specific signaling molecules (Cho et al., 2012; Niakan et al., 2013) increases the efficiency of ES cells giving rise to cells with XEN-like stem cell properties. These approaches have also been successfully applied to produce human XEN-like cells (Séguin et al., 2008; Wamaitha et al., 2015). However, XEN cells have not yet been derived from human embryos, which creates an obstacle in comparing and evaluating human PSC-derived XEN cell-like models.

Nevertheless, we are beginning to understand the molecular underpinnings of PrE-like stem cell lines. Mouse XEN cells bear transcriptional similarities to PrE in the blastocyst (Pham et al., 2023), but are more similar to a more differentiated state (Rothová et al., 2022). Additionally, in terms of developmental potential, XEN cells were observed to contribute only to portions of the parietal endoderm (Fig. 4A), with very rare contribution to the visceral endoderm in chimeric embryos (Kunath et al., 2005). Thus, XEN cells are more biased towards parietal endoderm fate. Consistent with this notion, isolated visceral endoderm differentiates to parietal endoderm as if by default (Hogan and Tilly, 1978; Ninomiya et al., 2005), suggesting that XEN cells, even when isolated from visceral endoderm, might also track towards a parietal endoderm fate. However, XEN cells can be guided toward visceral endoderm fate using an in vitro differentiation assay (Artus et al., 2012; Paca et al., 2012). These observations suggest that XEN cells, with their limited developmental potential, represent a more primed PrE state.

Fig. 4. Multiple stem cell types representing PrE.

XEN, nEnd, and PrESCs have been derived from blastocysts. XEN cells demonstrated more restricted developmental potential in chimeras than do nEnd and PrESCs, suggesting a more mature developmental state.

The notion that XEN cells represent a more mature PrE state is consistent with the derivation of a reportedly more naïve PrE-like stem cell line. Naïve endoderm stem (nEnd) cells are derived from naïve ES cells grown in medium with the addition of Activin A (ActA), the Wnt signaling agonist CHIR99021 (CHIR), and LIF (Anderson et al., 2017) (Fig. 4B). Interestingly, applying the same conditions to primed ES cells resulted in the formation of definitive endoderm, consistent with the notion that the developmental stage of the starting cell type influences differentiation outcomes accordingly. Morphologically, nEnd cells are distinct from XEN cells, as they appear epithelial rather than mesenchymal and resemble PrE typically observed in blastocyst outgrowths. When injected into embryos, nEnd cells contributed moderately to both parietal and visceral endoderm at E6.5. Notably, the conditions used to derive nEnd from ES cells also supported the derivation of nEnd cells from isolated mouse ICMs (Linneberg-Agerholm et al., 2024). Excitingly, a similar approach enabled the derivation of human nEnd cells from human naïve pluripotent stem cells (Linneberg-Agerholm et al., 2019), suggesting this approach could allow for the derivation of nEnd cells from human blastocysts as well.

More recently, another type of a naïve PrE-like cell was derived from mouse blastocysts (Ohinata et al., 2022) (Fig. 4C). These PrE Stem Cells (PrESCs) bear greater transcriptional similarity to E3.5 PrE than do XEN cells, consistent with a naïve to primed PrE spectrum. In addition, PrESCs possess expanded developmental potential in chimeric embryos compared to XEN cells. PrESCs are capable of contributing to both parietal and visceral yolk sacs around E7.5, as well as the mature yolk sac near the end of term at E18.5 (Ohinata et al., 2022). Whether PrESCs represent a unique state to nEnd cells is unclear because these have not yet been directly compared. A comparison of nEnd and PrESC culture conditions, however, reveals few compelling similarities. While nEnd cells rely on ActA, CHIR, and LIF (Linneberg-Agerholm et al., 2019, 2024), PrESCs rely on FGF4, CHIR, and PDGF signaling pathways (Ohinata et al., 2022, 2023). Therefore, nEnd and PrESCs could represent distinct states along the extra-embryonic endoderm maturation spectrum. Human PrESCs have not, to this date, been reported.

Of note, it is not totally clear why or how these specific PrE-supporting cocktails work. On the one hand, FGF4, PDGF and LIF are all known agonists of PrE proliferation, as described above. On the other hand, the physiological relevance of CHIR in PrE development is unknown. Wnt signaling is also not known to be required for PrE specification or visceral endoderm development (Biechele et al., 2013; Mohamed et al., 2004). This raises the possibility that the target of CHIR – glycogen synthase kinase 3 (GSK3) – could have Wnt-independent roles or that CHIR could have other targets in the PrE. Alternatively, CHIR may more generically support the ex vivo survival of stem cell lines, and cellular behavior that is inherently non-physiological – such as unbridled proliferation. Consistent with this, similar concentrations of CHIR are also used in culturing mouse and human PSCs (Collier and Rugg-Gunn, 2018; Wray et al., 2011; Ying et al., 2008), where self-renewal is also impelled.

8. To PrE or not to PrE

Returning to EMs, studies with ES cells alone demonstrated the developmental deficiencies of EMs, including low efficiency, incomplete differentiation, and patterning defects. This, and the advent of extraembryonic stem cell lines, led to the conceptualization of the integrated embryo model (IEM). IEMs incorporate stem cell models for all three foundational lineages: EPI, PE, and TE (Fig. 5). Given the known signaling interactions between extraembryonic tissues and the epiblast, IEMs could, conceivably, produce more consistent developmental outcomes. In fact, by aggregating mouse ES, TS, and XEN or XEN-like cells, IEMs can exhibit remarkable similarities to mouse embryos, including gastrulation, primordial germ cell gene expression, and some visceral endoderm patterning (Amadei et al., 2021; Sozen et al., 2018). However, this outcome requires the manual selection of IEMs that happen to resemble embryos midway through the protocol – around 10–15 % of all IEMs (Amadei et al., 2022). By continually cherry-picking the most morphologically embryo-like IEMs each day for several days, IEMs can develop beyond gastrulation to establish neural tube, beating heart, and gut tube and then reach a state resembling the mouse embryo at E8.5 (Amadei et al., 2022). Notably, the attrition of normal-looking IEMs over each day of the protocol means that the overall rate at which IEMs develop to this stage is actually only 0.7–1 %. Given this, it is difficult to argue for self-organization in this context.

Fig. 5. Integrated stem cell models of embryogenesis.

A) Stem cell lines representing all three blastocyst lineages are co-cultured to produce an Integrated Embryo Model (IEM). Subsequent manual selection of high-quality IEMs can result, with low overall efficiency in normal advanced embryos. B) An alternative strategy involves overexpression of lineage-specific transcription factors in ESC or PSC to simulate use of stem cell lines from the extraembryonic lineages.

Human IEMs have also been studied, using similar strategies. However, the absence of a bona fide human XEN cell line poses a problem here. PSC-derived XEN-like cells have been used as an alternative (Abel and Sozen, 2023; Shahbazi and Pasque, 2024). In spite of many approaches to making mouse and human IEMs, only a few IEMs have reported observing the formation of AVE-like cells within the EM (Hislop et al., 2024; Molè et al., 2021; Oldak et al., 2023; Weatherbee et al., 2024). However, the cherry-picking aspect of standard IEM procedures is certain to introduce bias. Additionally, identification of AVE-like cells in human IEMs is based on transcriptional similarity to mouse AVE – an untested assumption that could introduce further bias, given that defining mouse AVE is still an unresolved topic of research. One possibility is that a superior model of PrE could be introduced into IEMs. To date, only nEnd cells have been tested in IEMs (Linneberg-Agerholm et al., 2024), but their ability to support postimplantation developmental events was not examined.

The reliance on cherry-picking IEMs raises an additional question: what is a normal reference embryo supposed to look like? Human peri-implantation embryo specimens are exceedingly rare in cultivated collections, such as Carnegie (https://embryology.med.unsw.edu.au/embryology/index.php/Carnegie_Collection) or Kyoto (http://bird.cac.med.kyoto-u.ac.jp/index_e.html). Moreover, given estimated high degree of developmental failure during these early stages (Chard, 1991; Wang et al., 2003; Wilcox et al., 1988), one questions which of the available specimens are the “normal” ones. Nevertheless, IEMs clearly do not consistently operate in a self-organizing manner as embryos do. Clearly this is a research area needing greater investment. The current “try it and see” approaches to IEM development may not be the fastest way to the finish line. Instead, the field of embryo modeling would benefit greatly from a more systematic approach to understanding the necessary components of embryo development. The development of standards to guide experimental design and interpretation has been recently proposed (Martinez Arias et al., 2024). Additionally, the uterine environment, commonly regarded as dispensable in the IEM field, may turn out to be a key collaborator in embryo organization. In any case, the mouse – where embryogenesis is robust, readily accessible, and rich in biologically relevant context, will undoubtedly lead the IEM race.

Data availability

No data was used for the research described in the article.