Tissue and cell interactions in mammalian PGC development

The germ cells of an embryo are the keepers of the germline, producing the gametes (sperm or eggs) that are responsible for the next generation. The germ cells themselves are determined from a population of precursors called primordial germ cells (PGCs). In the mouse, these are first specified (i.e. reversibly associated to a PGC fate in an environment-dependent manner) in the pre-gastrulation embryo at about embryonic day (E) 6.5 (McLaren and Lawson, 2005) (Fig. 1). Once gastrulation has begun and the supporting tissues have started to form, these cells undertake a long migratory journey (starting around E7.5) via the hindgut and the dorsal mesentery to reach the nascent gonadal ridge at approximately E10.5 (Richardson and Lehmann, 2010). Upon colonising the gonadal ridge, the PGCs form clusters and undergo a process of maturation, through which they develop and gain characteristics of later-stage cells, including epigenetic remodelling (Gomperts et al., 1994; Hajkova et al., 2002). During this process, the PGCs become fully committed – or fate restricted – germ cells (Nicholls et al., 2019).

Recent technical advances in mammalian embryogenesis, including improved imaging of the developing mouse embryo (McDole et al., 2018) and ex utero culturing techniques (Aguilera-Castrejon et al., 2021), provide unparalleled access to the post-implantation embryo when PGCs first arise. However, research into PGC development is likely to remain hampered by the rarity of the cell type at early stages, which precludes many assays that require substantial material, such as chromatin immunoprecipitation (ChIP) sequencing. To circumvent this, researchers often use in vitro-derived PGC-like cells (PGCLCs) generated using an embryoid body (EB)-based aggregate of pluripotent stem cells that are exposed to a defined signalling cocktail that promotes PGCLC derivation (Hayashi et al., 2011, 2012; Ohinata et al., 2005). However, these PGCLCs are limited in their similarity to embryonic PGCs, with gene expression profiles of PGCLCs resembling PGCs at about E9.5 (i.e. the ‘migratory’ stage; Hayashi et al., 2011). Furthermore, the CpG methylation status of PGCLCs at the height of their in vitro differentiation (without any co-culture) is equivalent to ∼E10.5 PGCs (von Meyenn et al., 2016; Miyoshi et al., 2016).

Perhaps the reason that the EB-derived PGCLCs are stalled at this stage of maturity is because, although they are generated alongside somatic differentiated cell types, they do so in a fairly disorganised manner without supporting gonadal cell types, deviating from the known spatial organisation and tissue proportions exhibited by the embryo (Brickman and Serup, 2017). Perhaps because of this limitation, EB-derived PGCLCs still require subsequent co-culture with gonadal cell populations and/or in vivo transplantation to progress towards functional gamete maturity (Hayashi et al., 2012, 2011; Ge et al., 2015). Thus, EB-derived PGCLCs are somewhat limited in their development compared with embryo-derived PGCs because they do not progress to gametogenesis without co-culture, which might, perhaps, be a constraining feature of their derivation method.

Furthermore, several lines of research have observed the dependence of PGCs on neighbouring tissues for their development, even before their commitment to germ cells; whether that be extra-embryonic tissues involved in specification (Ohinata et al., 2009), gut endoderm involved in PGC migration and early development (Gu et al., 2009; Laird et al., 2011; Hara et al., 2009), or the genital ridge and early gonad primordium for PGC niche formation and subsequent maturation (Bullejos and Koopman, 2004; Menke et al., 2003; Mayère et al., 2021). Much of this evidence has come from careful dissection or transplantation assays, examining cellular plasticity when separated from their original environment.

Here, we highlight several important examples of the intercellular- and tissue-level dependencies of PGCs during the time course of their development. In particular, we argue that there is evidence for consistent dependency of PGCs on interactions with their environment, including surrounding cells and tissues, which we suggest reflects a general ‘interaction-dependency’ of PGCs. We primarily discuss PGC development in mice, although we bring in other species where relevant. In addition, rather than focussing on the genetic or signalling requirements of PGCs (for which we refer the reader to other recent reviews), we argue that a cell- and tissue-level perspective is important to gain new insights into the mechanisms of PGC development in vivo. These insights might also open new avenues for establishing culture techniques for derivation of mouse PGCLCs in vitro that recapitulate mouse PGC maturation in a developmentally faithful manner. Perhaps by doing so, we might be able to get closer to the efficiency, maturation and reproducibility that the embryo achieves.

Because mammalian PGCs emerge in early development, there has been some discussion about whether PGCs are actively specified in a deterministic fashion (the ‘first cell lineage’ hypothesis) or whether they arise stochastically as a result of delayed somatic differentiation (the ‘last cell standing’ hypothesis) (Johnson and Alberio, 2015). In particular, there is a conceptual distinction between active, early specification in the first cell lineage hypothesis (in which PGC-specific genes are preferentially upregulated resulting in lineage restriction in a few cells) and passive, late specification in the last cell standing hypothesis (in which a proportion of cells stochastically retain their heightened potential whereas neighbouring cells differentiate). In the case of mouse development, the general consensus is that the former (i.e. first cell lineage hypothesis) is the dominant mechanism, whereby Blimp1 (also known as Prdm1) expression in selected cells of the epiblast represses somatic gene expression and commits cells to a PGC fate (Ohinata et al., 2005; Kurimoto et al., 2008). However, others have suggested that lineage segregation occurs much later, and these Blimp1-positive cells are in fact multipotent progenitor populations capable of forming both germ cells and extra-embryonic mesoderm (ExM) tissues (Mikedis and Downs, 2014). In either case, it is becoming increasingly evident that PGC specification requires the interplay of both a permissive environment (often spatially confined and determined by neighbouring tissues) and internal cell state, and that both factors play a role in specifying the location, timing and proportion of PGCs in the embryo.

The first PGCs in the mouse embryo appear at ∼E6.25 in the posterior proximal epiblast (PPE) (Ohinata et al., 2005; Vincent et al., 2005), which is also the source of ExM and the primitive streak-derived embryonic mesoderm (Tam and Behringer, 1997), which begs the questions: why and how are PGCs specified in this region, and what mechanisms enable the distinction between these very different fates?

It has been elegantly shown that the restriction of PGC fate to the PPE, and the loss of germline potential outside of this region, is due primarily to signalling from the extra-embryonic tissues that establishes the early embryonic axes (Fig. 2A). Dissection experiments that carefully separated the epiblast from the extra-embryonic tissues, including the visceral endoderm (VE) and extra-embryonic ectoderm (ExE), of E6.0 mouse epiblasts have shown that, in the absence of either extra-embryonic tissue, the whole epiblast is competent to give rise to PGCs (Ohinata et al., 2009) (Fig. 2B). The effect of these extra-embryonic tissues is therefore to achieve robust embryonic symmetry breaking and axial specification, leading to regionalised identities within the embryo. As such, the extra-embryonic tissues create a permissive signalling environment specifically in the PPE (and restrict this fate in other parts of the epiblast), which allows the regionalisation necessary for PGC specification.

Several studies have subsequently identified the molecular mechanisms that underlie these tissue interactions to be largely the result of BMP and Wnt signalling (Lawson et al., 1999; Ying et al., 2000, 2001; Ying and Zhao, 2001; Ohinata et al., 2009; Aramaki et al., 2013; reviewed by Saitou and Yamaji, 2012). However, it is still unclear how cells within this permissive signalling environment go on to produce a variety of different embryonic (mesoderm, endoderm and PGCs) and extra-embryonic cell types in this region, and how this robust, multi-lineage differentiation is balanced.

PGCs first become evident in a sparse, ‘salt and pepper’ pattern within the PPE, rather than in a spatially defined location based on their proximity to the PPE and ExE interface (Ohinata et al., 2009; Aramaki et al., 2013; Senft et al., 2019). Therefore, regulation of PGC specification within the PPE is likely to occur at the cellular scale, through interactions between cells or as a result of underlying stochasticity in gene expression and fate decisions. However, the exact mechanisms for this specification still remain unclear.

Once the epiblast has been regionalised to specify the PPE, the correct number of PGCs must be specified. As Samuel Butler famously retorted, ‘a hen is only an egg's way of making another egg’ (Butler, 1878), so the developing embryo must carefully balance the proportion of somatic and germline cells to enable both viable and fertile offspring. Much of the research teasing apart this proportional PGC specification has been obtained by careful dissections of the early embryo.

In 1996, Tam and Zhou transplanted distal epiblast cells into the proximal region of host embryos and found that these transplants contributed to extra-embryonic mesoderm, mesoderm and, crucially, PGCs (Tam and Zhou, 1996) (Fig. 2C). Quantitatively, they found that ∼3.7% of transplanted cells were allocated to a PGC fate, in comparison with ∼3.9% of cells in orthotopic transplants. These observations have demonstrated not only the plasticity of the epiblast at this stage, but also that the permissive signalling environment created in the PPE is sufficient to promote the specification of PGCs in the correct proportions, restricting a defined number of cells to the PGC fate. This is particularly striking in light of the competence of the entire PPE to assume a PGC fate in the correct environment (Ohinata et al., 2009) (Fig. 2B), meaning that additional factors must be at play in the embryo that mediate this proportional balance between germline and soma.

The VE has also been implicated in regulation of the appropriate number of PGCs forming in the PPE. In experiments using explants cultured with and without extra-embryonic tissues, researchers have demonstrated that between E6.25 and E7.25 the VE is necessary for the correct number of PGCs to form, with those explants lacking the VE containing approximately half the number of PGC precursors (de Sousa Lopes et al., 2007). However, whether this is due solely to signalling requirements or additional interactions is not yet clear.

Taken together, these data suggest that the permissive environment of the PPE created by surrounding extra-embryonic tissues is responsible not only for the specification and localisation of PGCs, but also for the correct proportions of PGCs to be specified. However, the exact mechanism underlying this phenomenon is not yet fully understood. Cellular heterogeneity within the population may explain variability in differentiation competence in a stochastic manner (Hayashi et al., 2002) or instructive signals could be involved in biasing populations, or some combination of the two. Further experiments examining the specification of PGCs at cellular resolution will be required to elucidate this molecular mechanism fully, with far-reaching implications with respect to how the embryo controls cell fate decisions and cell proportions throughout development.

Beyond the tissue-level permissive signalling environments that enable localised PGC specification, cell-to-cell contacts are also known to be essential for the specification of PGCs (Yoshimizu et al., 2001; Okamura et al., 2003). The importance of cell-cell contacts has been observed following dissociation of cells from E6.5 PPE, which prevents PGC derivation in culture. Conversely, after the E6.5 stage, dissociated cells are able to form PGCs, and their ability to do so increases at later points of development, suggesting a stage-dependent requirement for cell-to-cell contacts (Yoshimizu et al., 2001).

Several suggestions have been postulated to account for this state-specific requirement for cell contacts. For instance, E-cadherin (also known as cadherin 1, encoded by Cdh1) is expressed in PGCs at various periods of their development (Bendel-Stenzel et al., 2000) and can distinguish PGCs from their neighbouring cells. Using an antibody that blocks E-cadherin function, researchers have found that PGC specification is inhibited, suggesting that E-cadherin might be mediating the cell-to-cell contacts or other cellular properties required for PGC specification (Okamura et al., 2003). Interestingly, it has been shown that E-cadherin can signal via mechanotransduction (Desprat et al., 2008; Benham-Pyle et al., 2015), which could play a role in mediating the induction of PGCs in a cell contact-dependent manner. Furthermore, such mechanotransduction is thought to rely on β-catenin activity, which is known to play a key role in PGC specification (Aramaki et al., 2013). These findings point to the possibility that connections between cells are required during the initial stages of PGC development for correct specification.

Whether this potential requirement for cell contact is based on germ-germ contacts or germ-soma contacts, is much less understood. However, an ex vivo study examining the role of E-cadherin using PGCs at a later stage of development has shown that that the ability of PGCs to adhere to a cell monolayer and produce motile morphologies is dependent on the presence of E-cadherin on the feeder cells, and is also inhibited by an E-cadherin-blocking antibody (Di Carlo and De Felici, 2000). Additionally, the researchers observed E-cadherin-mediated aggregation of PGCs in culture where, similarly, E-cadherin inhibition disrupted PGC colony-forming ability and reduced PGC numbers (Di Carlo and De Felici, 2000). These results suggest that potential germ-somatic cell contacts may exist during certain stages of PGC development in the mouse embryo. Several other cell-adhesion proteins, such as β1 integrins, EpCAM and PECAM1, are expressed by PGCs at various stages during migration and in the gonadal ridge (Wakayama et al., 2003; Mork et al., 2012; Anderson et al., 1999a,b; De Felici et al., 2005). Such factors might also play a role in cell-to-cell contact regulation of PGC specification, migration, survival and cluster formation.

Even once PGC-associated gene expression becomes apparent (i.e. Blimp1, the first of a tripartite network of transcription factors responsible for PGC specification, is expressed; Magnúsdóttir et al., 2013), the cells are still relatively plastic in their fate assignment and rely on cues from their environment and neighbouring tissues for fate commitment. Lineage tracing has shown that the precursors of PGCs (from pre-/early-streak- or pre-gastrulation-stage embryos) are not fully lineage restricted, and instead are able to give rise to ExM, as well as PGCs (Lawson and Hage, 1994). Furthermore, when PPE cells from pre-/early-streak (E6.5) embryos are heterotopically transplanted to distal regions, over 90% of grafted cells form neural and surface ectoderm rather than PGCs or ExM (Tam and Zhou, 1996). This is despite the fact that Blimp1 expression in prospective mouse PGCs is evident before E6.5 (Ohinata et al., 2005; Vincent et al., 2005). Indeed, when Blimp1-positive cells from E6.5 embryos are transplanted to distal epiblasts they no longer give rise to PGCs. Conversely, if distal epiblast cells are transplanted to the PPE, they form PGC clusters (de Sousa Lopes et al., 2007). It is worth noting here, however, that work by Mikedis and Downs has demonstrated that Blimp1 expression is not limited to PGCs in the posterior region of the developing embryo (Mikedis and Downs, 2017). Therefore, the notion of ‘PGC-specific’ or ‘PGC marker genes’ is likely to be misleading as, at this early stage of development, there is still plenty of plasticity in the cell fate, and PGCs are likely to take additional specification/commitment cues from the microenvironment in which they find themselves.

This plasticity can be utilised in vitro to culture PGCs into a pluripotent cell line referred to as embryonic germ cells (EGCs) (Resnick et al., 1992). However, the ability of PGCs to convert into EGCs reduces progressively until they reach the gonads at E14.5 (Durcova-Hills et al., 2008; Matsui and Mochizuki, 2014). This aligns with epigenetic reprogramming associated with germ cell fate (Hajkova et al., 2002), loss of teratoma-forming capacity (Stevens, 1964) and pluripotency factor downregulation (Chen et al., 2014; Nicholls et al., 2019) at the point of gonad colonisation. This implies that commitment is progressively attained, either gradually or proportionally at the individual cell level. These dynamics are thought to be due to the direct influence of the gonads, as shown using Gata4 (required to form both the male and female genital ridge) conditional knockout embryos that do not possess genital ridges, in which PGCs migrated to the site where the genital ridge would normally form but did not become competent to undergo gametogenesis and retained their immature PGC identity (Hu et al., 2015). However, once exposed to the genital ridge or its influence, the PGCs develop towards meiosis in a cell-autonomous fashion and in a synchronised manner to those found in the genital ridge (Upadhyay and Zamboni, 1982; Richards et al., 1999; Tedesco et al., 2013). The precise mechanism of this influence from the genital ridge still remains somewhat unclear, but it does highlight that PGCs possess stage- and location-dependent potency, which are mediated through interactions with neighbouring tissues.

Taken together, these results suggest that not only do neighbouring tissues act to create a permissive signalling environment for the specification of PGCs, but they also control the proportion of cell differentiation to somatic and germ fates. There is also some evidence to suggest that local cell-cell contacts, perhaps between somatic cells and PGCs, might be required for early specification events. Both phenomena combine to create microenvironments that enable the commitment to a PGC fate.

Perhaps non-intuitively, even across evolutionarily diverse species, PGCs are not specified where they will eventually develop into gametes. This means they must migrate through neighbouring tissues, from their point of specification (the PPE) to their point of final maturation: the gonad (Richardson and Lehmann, 2010; Molyneaux and Wylie, 2004). The exact cause of this circumlocutory requirement is not yet well understood, although it has been suggested that this added hurdle might reflect a means for the embryo to enforce a selective pressure on cells, with an adaptive migratory advantage as a form of quality control (Cantú and Laird, 2017) imposed by the relationship between PGCs and their surrounding tissues and organs. In some species, this PGC migration may in fact aid the development of somatic cells with which they interact along their route, as is the case in Drosophila (Stepanik et al., 2016).

In terms of cell biology, active cellular motility is facilitated by coordinated cell biological processes involving the protrusion and attachment of the leading edge ahead of the cell, and detachment of the trailing edge behind the cell, leading to directional migration. Yet passive movement or displacement of cells is also possible in development, with cells being carried along by the active motion and fluid-like flows of neighbouring tissues (Zamir et al., 2006). In avian species, PGCs migrate along the bloodstream, for example (Swift, 1914). In other species, it has been argued that PGCs scattered through the endoderm are passively carried along by the axial elongation of the embryo (Freeman, 2003). This passive movement is likely the case in species in which PGCs remain transcriptionally quiescent, such as Caenorhabditis elegans where the PGCs are passively internalised with the endoderm by virtue of their E-cadherin expression (Chihara and Nance, 2012). Similar associations between PGCs and the endoderm exist in Drosophila, in which endodermal invagination similarly leads to PGC internalisation (Warrior, 1994). The evolutionary driver for PGC association with the endoderm might therefore be to achieve internalisation in passive cell populations.

However, it is now well understood that in the mouse, PGCs are inherently motile during periods of their development, based on both cell morphology and dynamic movements seen in vivo and in vitro (Stott and Wylie, 1986; Anderson et al., 2000; Molyneaux et al., 2001). The directional nature of this movement, or ‘migration’, is therefore dependent on the location in which the PGCs find themselves, and the neighbouring somatic cells that cohabit their environment.

There appears to be a conserved association between the developing endoderm and PGCs in several species, including in mammals (reviewed by Richardson and Lehmann, 2010). It is not immediately apparent why this relationship is conserved, and what advantage might be gained from such an association. However, mounting evidence is starting to suggest that the relationship might not be just a remnant of evolution, but an active interaction between cell types that work together to achieve robust development.

Beyond providing signals to promote migration of PGCs once they are within the endoderm (Gu et al., 2009; Laird et al., 2011; Cantú et al., 2016), the endoderm plays a vital, physical role in PGC migration. At approximately E7.5, the PGCs begin to migrate out of the proximal-posterior part of the epiblast and enter the developing endoderm. However, further migration from the posterior hindgut towards more anterior endoderm seems to be dependent on the tissue movement of the endoderm (Anderson, 2000).

In Sox17-null embryos, which have defective hindgut expansion but no discernible defect in surrounding tissues, PGCs are unable to migrate into the gut endoderm (Hara et al., 2009). When examined closely, it has been observed that the PGCs can migrate out of the PPE but stall at the hindgut entrance, leading to reduced PGC numbers at later stages. By creating chimeras using Sox17-null cells implanted into wild-type blastocysts, the researchers have shown that Sox17-null PGCs interact with the hindgut and rescue migration (Hara et al., 2009). Therefore, the observed defect is not due to direct aberrations in the PGCs themselves, but is rather due to malformation of the endoderm, which, in turn, affects PGC development. It has been suggested that this initial cell movement of PGCs into the hindgut is due to passive migration of PGCs, requiring only the proper expansion and morphogenic movement of the hindgut to propel them into the gut. However, within the endoderm PGCs display motile behaviour even prior to exiting this tissue (Molyneaux et al., 2001).

Similarly, further studies that have generated epiblast-specific gene knockouts (e.g. Smad2 or Smad4, which lead to a lack of endoderm formation) result in a loss of migration and associated morphological changes (e.g. protrusions) in PGCs (Gomperts et al., 1994; Senft et al., 2019; Anderson et al., 2000). These results support the notion that the PGCs are dependent on the presence of correctly formed endoderm and that morphological endodermal defects can affect PGC motility and morphology.

Having migrated through the endoderm, the PGCs then must find their way to the developing genital ridges. It was originally believed that PGCs migrate through the dorsal mesentery individually, each independently starting and finishing their migration like runners in a race. However, seminal work from Gomperts and colleagues revealed that PGCs actually use a variety of physical connections between cells during their migration (Gomperts et al., 1994) (Fig. 3). Using SSEA1 (FUT4) as a cell-surface marker for PGCs, they identified long projections between cells at E10.5, including both lamellipodia and filopodia, that form what they termed ‘networks’ of cells within the surrounding tissue. These projections are first visible at E9.5 as so-called ‘pioneer cells’ extended elongations out of the endoderm (Fig. 3A), and by this stage they estimate that 35% of PGCs are networked (Gomperts et al., 1994) (Fig. 3B). By E11, when cells are in the genital ridge, the cells assume a completely different morphology and instead form small clusters of large cells that maximise the inter-PGC cell-surface area (Fig. 3C).

At first, it was unclear whether PGCs were being manipulated by the surrounding influences of morphological embryonic development, such as the pulling and stretching of the axially elongating tissue, which might have contributed to the long, interconnected extensions observed in vivo. To test this, cellular morphologies have been observed in vitro following dissection and plating of E10.5 PGCs on irradiated STO fibroblast feeder cells (Fig. 3D). Even in static culture, by 24 h in vitro, 86% of cells were networked, and by 48 h the cells started to form small clusters of large cells, maximising their contact area (Gomperts et al., 1994). This result suggests that the phenomenon of PGCs extending cellular elongations between one another is an inherent property of the cells, rather than a result of inter-tissue forces.

This network-forming property of PGCs also changes temporally, in that the ability of PGCs to form inter-cell connections increases from E11.5 until E14.5-E15.5, before decreasing again (Di Carlo and De Felici, 2000). This temporal property of the cells correlates well with expression of E-cadherin, which begins to be expressed in PGCs around E10.5, concomitant with their migration out of the gut endoderm (Bendel-Stenzel et al., 2000; Di Carlo and De Felici, 2000). This correlation seems to be functional, because disrupting E-cadherin function with a blocking antibody leads to impaired cluster morphology in mouse gonads (Di Carlo and De Felici, 2000) and an increase in mislocalised, extragonadal PGCs throughout the embryo (Bendel-Stenzel et al., 2000). Similar experiments performed in vitro have also shown this clustering defect and demonstrated that it is associated with reduction in the expression of PGC markers, such as Stella (Dppa3) (Okamura et al., 2003). Other cell-surface molecules have also been implicated in the migrational competence of PGCs, including β1 integrin (Anderson et al., 1999a,b), suggesting that widespread surface markers might play a role in regulating the physical properties of PGCs, in turn affecting their migration ability and their interaction with neighbouring tissues. This suggests a tantalising possibility that cellular morphology and tissue mechanics might play a role in coordinating PGC development.

The migration of PGCs along the endoderm and into the genital ridge of the embryo is a highly reproducible phenomenon of movement, relying on other tissues to be formed in the correct spatiotemporal arrangement to ensure PGCs reach their final destination. It is particularly important that PGCs reach the genital ridge, because mislocalisation can result in teratoma tumour formation (Pierce et al., 2018) with devastating consequences.

How mouse PGCs are guided to the correct destination in a robust manner has been extensively investigated, to try to understand what drives and directs this migration. In the 1990s, experiments culturing PGCs in vitro in small chambers with tissue explants in neighbouring (but not touching) chambers at either end, revealed that the genital ridge, but not the limb bud or mesentery, is able to attract cells directionally (Godin et al., 1990). Subsequent studies have revealed that the chemokine SDF-1 (also known as CXCL12) is expressed by the developing genital ridge, and that PGCs at E11.5 express the chemokine co-receptor CXCR4 (Ara et al., 2003). Mice lacking SDF-1 have roughly the same number of PGCs and they have the same clustering ability, but they are often retained in the gut or mesentery rather than homing to the genital ridges (Ara et al., 2003). Likewise, mouse embryos exposed to exogeneous SDF-1 have impaired directional migration and PGCs are similarly mislocalised, as are PGCs with Cxcr4 mutations (Molyneaux et al., 2003). Corresponding experiments in zebrafish have also resulted in mis-migration, suggesting that this is a conserved mechanism for directing PGC migration (Doitsidou et al., 2002). Together, these results suggest that an inter-tissue chemotactic gradient and receptor expression in PGCs is responsible for the correct directional migration of PGCs in vivo. Importantly, the timing of the genital ridge development must coincide with PGC migration, so the PGCs are dependent – even from a distance – on somatic tissue development.

During migration, the PGCs travel along a pathway laid down by local tissues in the form of extracellular matrix (ECM) proteins. This pathway is a mix of fibronectin, collagen IV and laminin (Fujimoto and Yoshinaga, 1986; García-Castro et al., 1997). PGCs respond differently to each one and change their strength of adhesion in response. In fact, it has been suggested that fibronectin plays a direct role in stimulating and modulating the migration of PGCs (Ffrench-Constant et al., 1991). At the end of their migration, they interact directly with a ‘ribbon’ or ‘tract’ of laminin and upon reaching the gonadal ridge PGCs form clusters within secreted ECM proteins, which may act to protect their niche by reinforcing PGC-to-PGC signals (García-Castro et al., 1997). This highlights a key role for the extracellular environment in the guidance of PGC migration.

If PGCs correctly migrate through the endoderm and dorsal mesentery, they arrive in the developing gonad, termed the genital ridge. The sex-specific developmental fate of germ cells (either oocyte or spermatogonia) is determined not by their genetic composition of sex chromosomes, but rather, by the surrounding cells (Mayère et al., 2021). In fact, if germ cells do not correctly settle in the genital ridge, they will become functional oocytes even if they possess XY chromosomes, demonstrating the dependence of these cells on their somatic surroundings (Evans et al., 1977). However, even this ability to form mislocated oocytes has been demonstrated to require long-range exposure to the developing genital ridge, as Gata4 knockout studies have shown (Hu et al., 2015), in order to become committed to a germ cell fate (Nicholls and Page, 2021).

Likewise, there is also a reliance on germ cells for correct somatic development. In the testis, somatic cells can develop without germ cells; however, this is not true in the developing ovaries, which require the presence of the oocytes to form correctly (McLaren, 1994). Within the ovaries, the position of the germ cells dictates the timing of their genetic and epigenetic changes, with those more anteriorly located progressing first and the somatic cells controlling their coordinated entry into meiosis as a rostro-caudal wave (Bullejos and Koopman, 2004; Menke et al., 2003). This polarised maturation corresponds to the development of the genital ridge itself, starting anteriorly and progressing towards the posterior prior to PGC arrival (Hu et al., 2015). This stage of PGC development shows some of the strongest evidence for a role of co-dependence between the soma and germ cell lineage, although there is evidence for interactions throughout the course of PGC development.

Throughout the animal kingdom, there are plentiful examples of interactions between somatic and germ cells, many of which have been experimentally shown to be necessary for the correct specification, migration and maturation of PGCs. Indeed, it could be argued that this reliance on neighbouring tissues in the developing embryo is a fundamental feature of coordination in development, whereby the co-development of tissues enables reproducible generation of viable and fertile offspring.

In abstract terms, and from the point of view of the developing PGCs, this dynamic environment and multi-tissue interaction means that the context they experience is constantly changing. Termed the ‘travelling niche’ by Gu and colleagues (Gu et al., 2009), PGCs receive input from the surrounding somatic cells from early in their specification, during their time in the endoderm, and when they reach the gonadal ridge. These inputs influence their survival, proliferation and migratory ability. As noted by Cantú et al. (2016), PGC proliferation is often more dependent on location rather than the age of the embryo, and this is dictated by the distinct niche in which a PGC finds itself. Indeed, this reliance on co-development might ensure that cells adhere to a global, embryo-wide timekeeping ‘reference’, whereby cellular states are linked to one another by virtue of their interdependence. So, from first specification, through their migration and as they progress and mature into germ cells, the PGCs are reliant on the surrounding soma for direction and support relevant to their developmental stage (Fig. 4).

In recent years, in vitro methods of generating PGCLCs have been shown to be useful in studying some of the processes associated with PGC development. Such methods have been devised using a range of starting cell types and treatments, but the current state-of-the-art procedure involves a protocol that requires an initial induction of mouse embryonic stem cells (mESCs) into Epi-like cells before aggregation into EBs, and treatment with a cocktail of signalling factors to induce the formation of PGCLCs (Hayashi et al., 2011, 2018). Using this system, research has sought to improve our understanding of the genetic and epigenetic changes that occur in mouse PGCs (Kurimoto et al., 2015; Shirane et al., 2016; Nakaki et al., 2013; Murakami et al., 2016). Similar methods have also been applied to generate human PGCLCs (Irie et al., 2015; Sasaki et al., 2015).

Such methods, however, are not without their limitations. Current methods of PGCLC induction bias cells towards the derivation of PGCLCs to the detriment of other cell types. This makes studying co-development of PGCs with somatic cells and their putative interactions difficult, if not impossible. For instance, one of the outstanding questions in the field is how robust proportions of PGCs are specified and how this might be regulated to generate separate populations from within the same environment. In the current PGCLC system, it would not be possible to address this problem because the directed differentiation imposed by the exogeneous signalling environment obscures the internal systems of control.

Indeed, EBs generally lack multi-tissue organisation and have tended to be heterogeneous in both size and shape (Brickman and Serup, 2017), limiting their utility as a model system in which to study PGCLC interactions with somatic cells or phenomena such as migration. Some degree of movement of PGCLCs in EBs has been observed, whereby human (h)PGCLCs exhibit upregulation of genes associated with migration and shift their relative positions (Mitsunaga et al., 2017). However, this most likely represents the intrinsic mobility of PGCs at this stage of development and is not reflective of a globally coordinated migration (directional movement) towards a gonadal tissue. Part of the reason for this could lie in the fact that although the hPGCLCs expressed CXCR4, the EBs lacked expression of the CXCL12 ligand (which, in the embryo, is a chemoattractant secreted from the genital ridge), belying a general lack of global organisation in EB systems that might limit PGCLC development and their subsequent ability to undergo maturation to germ cells (Nicholls and Page, 2021).

However, the recent identification of putative PGCLCs in self-organised, embryo-like systems, such as mouse gastruloids (van den Brink et al., 2020; Veenvliet et al., 2020), mouse embryo-like structures (Harrison et al., 2017; Sozen et al., 2018), human micropatterns (Minn et al., 2020, 2021) and human epiblast-like models (Zheng et al., 2019) bring with them a novel opportunity to shed light on these types of questions (Hancock et al., 2021). Such systems might be well positioned to offer better recapitulation of the various conditions PGCs encounter over developmental time, by allowing their co-development alongside somatic cell types. However, such studies on PGCLCs generated in embryo-like model systems have yet to be functionally validated (i.e. converted to functional gametes), which is necessary to prove their validity conclusively.

And yet, it is conceivable that an in vitro system containing multi-tissue spatiotemporal patterning akin to the embryo, could potentially allow the timings of signals, physical cues and interactions for the PGCLCs to develop in a manner that closely resembles that of PGCs in vivo. For instance, Zheng and colleagues have used 3D aggregates of human pluripotent stem cells in microfluidic devices to generate human embryo-like models with both the epiblast and amnion components (Zheng et al., 2019). Importantly, the ability to control experimentally the presence or absence of tissues, such as the amnion-like cells, in these models has provided the opportunity to explore a previous enigma in the field – the origin of human PGCs – and to decipher whether they arise from epiblast or amnion progenitors (Kobayashi and Surani, 2018; Hancock and Chen, 2018; Sasaki et al., 2016; Hancock et al., 2021). By using an embryonic model to explore the inter-tissue and inter-cellular requirements and coordination between PGCs and soma, researchers are able to bring new insights to an old problem. We suspect that as these models gain traction, and as embryonic model-derived PGCLCs become better characterised and fully validated, their advent might herald a new perspective in PGC research, focussing on permissive environments, cellular behaviours and tissue interactions in a ‘bottom-up’ manner.

It might also mean that by focussing on tissue- and cell-level interactions in vitro, PGCLCs can be generated that are more reflective of their in vivo counterparts, more reproducible and more mature. For instance, along their migratory pathway, PGCs undergo extensive epigenetic reprogramming that is required to prepare them for their arrival at the gonadal ridge and for them to develop fully into germ cells (Seki et al., 2005, 2007). Currently, EB-derived PGCLCs do not fully recapitulate the development of mature PGCs, with differences in both epigenetic changes and an inability to enter meiosis properly (Ge et al., 2015). Indeed, the further development of PGCLCs towards mature gametes necessarily requires co-culture with gonadal cells and/or transplantation (Hayashi et al., 2012, 2011), implicating the requirement for cell mixtures in promoting efficient differentiation conditions (Ge et al., 2015). Both these methods have been able to generate functional gametes that result in viable offspring; however, the oocyte-like cells display some abnormalities and are less efficient at producing pups (3.9% from PGCLCs) compared with E12.5 PGCs (17.3%) (Hayashi et al., 2012). Further progress has since enabled the completion of gametogenesis completely in vitro, for both mouse PGCLCs (Ishikura et al., 2016; Zhou et al., 2016; Hikabe et al., 2016) and hPGCLCs (Yamashiro et al., 2018; Hwang et al., 2020), using co-culture with mouse gonadal cells. Yet the mouse PGCLCs are significantly less efficient at generating healthy offspring compared with their in vivo counterparts (Ishikura et al., 2016; Hikabe et al., 2016), highlighting some, still unknown, missing element separating PGCLCs and PGCs.

To circumvent this reliance on gonadal somatic cells derived from foetal tissue, others have derived gonadal-like cells in vitro from mESCs (Yoshino et al., 2021), which behave with similar efficacy to in vivo gonadal cells when reconstituted with PGCLCs, but still result in PGCLCs with reduced capacity to generate offspring. From these studies, we can conclude that, although gametogenesis can occur in vitro, the gametogenesis is not yet equivalent to the process in vivo. Additionally, and most crucially, this process is wholly dependent on co-culture with gonadal somatic cells, exemplifying a currently irreplaceable requirement for co-development of PGCs and the soma. Perhaps alternative in vitro systems that enable co-culture even during the earliest stages of PGCLC development, such as embryo-like models, might be better suited to mediate ongoing interactions between the soma and the germline, to better recapitulate PGC development, from specification through maturation to full gametogenesis.

We would like to thank Alfonso Martinez Arias, Robin Lovell-Badge and Azim Surani for helpful discussions around this topic. We apologise to colleagues whose work we could not cite owing to space limitations.