Germ cell determination and the developmental origin of germ cell tumors

The germline is the cell lineage, from zygote to gamete, with the capacity or potential to contribute to the next generation (see Glossary, Table 1). Accordingly, a definitive test of whether a cell is of the germline or soma (see Glossary, Table 1) is genetic: all changes arising in germline cells can be inherited (assuming they do not impede gametogenesis), and no genetic change arising in the soma can be transmitted to offspring, an exclusion known as the ‘Weismann barrier’ (Fig. 1).

Within an animal, gametes trace their lineage back to primordial germ cells (PGCs), which arise during early development (reviewed by Strome and Updike, 2015; see Glossary, Table 1). After mammalian PGCs are induced, they migrate across the developing embryo to the nascent gonad, a journey widely conserved across metazoa (reviewed by Richardson and Lehmann, 2010).

When does the germline irreversibly commit to producing gametes and no other cell type – that is, when are germ cells determined? Here, we argue that this question is closely linked to the origin of germ cell tumors (GCTs; see Glossary, Table 1), which arise from the germline but are composed of diverse cell types. We compare germline development across diverse species to consider when and how determined (committed) germ cells are first distinguished from their germline precursors. We conclude that germ cell determination occurs late in development, after the body plan has been established and organogenesis begun, through a transition widely conserved among vertebrates. Synthesizing molecular and clinical findings, we hypothesize that mammalian GCTs reflect a failure of germ cell determination.

Four lines of evidence support the view that migratory PGCs of mammals are not committed to gametogenesis. First, migratory PGCs express a network of pluripotency factors (‘potency’; see Glossary, Table 1). In mice, this network includes factors characteristic of pre-implantation embryos and embryonic stem cells (ESCs), including the key transcription factors encoded by the genes Nanog, Pou5f1 (encoding Oct4) and Sox2 (Leitch and Smith, 2013; Nicholls et al., 2019). Like rodents, human PGCs express markers of the naïve pluripotent state. Unlike mice, human PGCs do not express ESRRB or SOX2 (Perrett et al., 2008); instead SOX17 functionally replaces SOX2 by binding to the promoters of pluripotency factors and maintaining GCT cell lines in an undifferentiated state (Jostes et al., 2020). Second, PGCs can contribute to somatic lineages of the developing embryo. In mice, nascent PGCs expressing both DPPA3 and PRDM1 contribute to the allantois during normal development (Mikedis and Downs, 2012, 2017). Third, rodent PGCs give rise to pluripotent embryonic germ (EG) cell lines when cultured under conditions similar to those that facilitate derivation of ESCs, with no need for genetic transformation by pluripotency factors (Leitch et al., 2010, 2013; Matsui et al., 1992; Resnick et al., 1992). Human PGCs also produce transient EG-like cells when cultured (Kerr et al., 2008; Shamblott et al., 1998). Finally, PGCs can differentiate to produce teratomas (see Glossary, Table 1), a GCT that displays derivatives of all three (somatic) germ layers. Pioneering work by Leroy Stevens established that mouse PGCs can differentiate to produce spontaneous gonadal teratomas, or to experimentally-produced teratomas when PGCs are transplanted to ectopic locations such as the adult testis and spleen (Stevens, 1964). In humans, the germline can similarly produce GCTs (see below). Taken together, these many lines of evidence indicate that migratory PGCs of mammals are not committed to gametogenesis.

Although mouse PGCs have not been shown to contribute broadly to somatic lineages when transplanted into blastocysts, a study in pigs suggests this is possible in other mammals (Leitch et al., 2014; Mueller et al., 1999). The failure of mouse PGCs to produce chimeras may reflect a developmental incompatibility between PGCs and the blastocyst, as observed elsewhere when donor cells and the intended host environment are not well matched (Cohen et al., 2018).

This developmental potential of migratory PGCs in mammals is mirrored in non-mammalian vertebrates, irrespective of whether those PGCs arose via maternal inheritance or induction (see Glossary, Table 1; Fig. 1B). In frogs, transplantation of PGCs to ectopic sites reveals their capacity for differentiation to a somatic cell fate according to the recipient niche (Wylie et al., 1985). Similarly, frog PGCs deficient in Nanos1 and located in the ectoderm are unable to evade the inductive influences of this niche, and activate ectoderm markers not usually expressed in the germline (Lai et al., 2012). In salamanders, nascent PGCs can be induced to express lineage-defining markers of somatic cells ex vivo (Chatfield et al., 2014). In zebrafish, PGCs deficient in the chemokine receptor Cxcr4 fail to respond to gonad-derived cues and instead migrate to ectopic locations throughout the developing embryo (Gross-Thebing et al., 2017). When these PGCs are also deficient for Dnd1, they differentiate to a somatic fate according to their new niche, indicating that Dnd1 functions to hold the somatic potential of migratory PGCs in check as they encounter ectopic niches (Gross-Thebing et al., 2017). Further, this raises the possibility that wild-type PGCs contribute to somatic lineages in the zebrafish, albeit at low frequency. Collectively, observations from a diverse array of mammalian and non-mammalian vertebrates demonstrate the broad potential of migratory PGCs for non-gametogenic fates – including GCT formation – regardless of whether those PGCs arose via induction or maternal inheritance.

Anne McLaren defined germ cells (see Glossary, Table 1) as ‘those cells, all of whose surviving descendants will become sperm and eggs’ (McLaren, 2003). Thus, germ cells are distinguished from their immediate precursors when they lose the capacity for somatic fate(s). By this definition, vertebrate PGCs are not yet germ cells, as they retain the potential for non-gametogenic fates.

Experimental studies of mammalian PGCs point to a commitment (see Glossary, Table 1) occurring after gonadal colonization. Using transplantation assays, Leroy Stevens discovered that germline potential is restricted within a few days of PGCs completing their migration across the embryo, between embryonic day (E) 12.5 and E14.5 in mice (Stevens, 1964, 1966, 1967). Consistent with this, the capacity for the derivation of pluripotent EG cells is restricted shortly after PGC colonization of the gonads in mice (Labosky et al., 1994; Matsui and Tokitake, 2009), as is the derivation of transient EG-like cells from human embryos (Kerr et al., 2008). Although long-term culture of spermatogonia can yield pluripotent cell lines, these arise exceptionally rarely and, unlike PGCs, transplantation of spermatogonia does not yield teratomas (Kanatsu-Shinohara et al., 2004; Takashima and Shinohara, 2018).

Revisiting the writings of earlier scholars, Jonathan Slack described two levels or states of commitment: specification and determination (Slack, 1991; see Glossary, Table 1). These two commitment states can be distinguished experimentally by grafting tissues and then examining whether cells develop according to the new niche, or according to their original position. By this definition, migratory PGCs of vertebrates may be specified, but their potential is not yet irreversibly restricted. By contrast, if a transplanted cell will only differentiate according to its original position, it is said to be determined, as its developmental potential is now fixed. Accordingly, germ cell determination occurs after PGC colonization in mice.

Analyzing the molecular profile of migratory and gonadal PGCs further reinforces the conclusion that germ cell determination occurs after gonad colonization. We recently examined the transcriptional changes that accompany PGC colonization of the gonads in humans and mice (Nicholls et al., 2019). We determined that genes expressed by migratory PGCs (such as Dnd1, Nanos3, Prdm1 and Tfap2c) are expressed more broadly in somatic lineages in embryos and adults (i.e. their expression is not limited to germ cells or to the germline). In contrast, we identified a set of genes, first activated after PGC colonization, that are exclusively or predominantly expressed by vertebrate germ cells; these include Dazl, Ddx4, Mael and Tdrd12 (Table 2). These genes encode RNA-binding proteins that function in the maternally inherited germ plasm of non-mammalian vertebrates (including fish, frogs and chicken) and some invertebrates (Table 2) (Ewen-Campen et al., 2010). Thus, mammalian germ cell determination coincides with the induction of a set of widely conserved germ-cell factors. One such factor, Dazl, is first activated shortly after PGCs colonize the gonads in mice and humans, and is required for germ cell determination (Nicholls et al., 2019). When the Dazl gene is ablated, mouse PGCs arrive at the nascent gonad but continue to express pluripotency factors, including NANOG, SOX2 and LIN28 (Chen et al., 2014; Gill et al., 2011). The Dazl-deficient mouse germline also retains the capacity to give rise to pluripotent EG cells until at least E15.5, long after wild-type littermates lose this capacity, and almost 10 days after gastrulation (Nicholls et al., 2019). On most inbred genetic backgrounds, the Dazl-deficient germline fails to initiate sexual differentiation or gametogenesis and instead undertakes cell death, with few if any germline cells observed in the gonads of either sex by birth. In the 129-strain of mice (Box 1), some Dazl-deficient cells form testicular or ovarian teratomas, comprising cells from each of the three (somatic) germ layers. In pigs, an outgroup to rodents and primates (Fig. 1B), DAZL deficiency similarly leads to teratomas in ovaries and germline loss in testes (Nicholls et al., 2019). These data demonstrate that Dazl is necessary for germ cell determination, regardless of whether germline development occurs through induction or by maternal inheritance. Although expression of DAZL results in the downregulation of pluripotency factors in ESCs in vitro (Jung et al., 2017; Xu et al., 2013), it remains to be tested whether precocious expression of Dazl is sufficient for germ cell determination in vivo, before gonadal colonization.

In contrast to Dazl, no other widely conserved factors activated on PGC colonization of the embryonic gonads are necessary for germ cell determination. In the genetic absence of Ddx4, Mael or Tdrd12, PGCs initiate gametogenesis following arrival at the gonad, and females are fertile (Malki et al., 2014; Pandey et al., 2013; Tanaka et al., 2000). In the absence of Dazl, each of these other genes is induced after PGC colonization and, despite their expression, the germline retains the capacity for EG cell derivation and teratoma formation (Nicholls et al., 2019). Genetic ablation of Ddx4, a constituent of germ plasm in both vertebrates and invertebrates, does not increase the incidence of teratomas in the 129-strain of mice (Nicholls et al., 2019; Raz, 2000). Similarly, deletion of Gcna, another factor induced after PGCs colonize the gonad and present in the germ plasm of several model vertebrates and invertebrates, does not enhance teratoma incidence in mice (Carmell et al., 2016; Nicholls et al., 2019). Mechanistically, Dazl functions to increase translational efficiency (Li et al., 2019; Mikedis et al., 2020), unlike other conserved germline factors, which typically repress gene expression (Seydoux and Braun, 2006). Thus, Dazl is unique among conserved germ cell factors: it is the first factor shown to be necessary for germ cell determination in vertebrates, regardless of whether the germline is segregated by maternal inheritance or by induction.

Together, this molecular and functional evidence supports a final restriction occurring after PGC colonization of the nascent gonad in mammals. It is likely that germ cells are the last lineage irreversibly committed in mammals, after organogenesis has commenced, almost 1 week after gastrulation in mice (Johnson and Alberio, 2015).

Comparing germ cell determination in mammals with observations in other vertebrates reveals a similar progression occurring around the time that PGCs colonize the gonads. As in eutherian mammals, migratory PGCs of other vertebrates are developmentally uncommitted and express transcription factors found in pluripotent cells, most notably Nanog, and the PouV and SoxB families (Table 2). These transcription factors are crucial in vivo for establishing pluripotency at the maternal-to-zygotic transition, and they are sufficient in vitro to reprogram mammalian somatic cells to pluripotency (Boyer et al., 2005; Lee et al., 2013; Takahashi and Yamanaka, 2006). In many vertebrates, this network of pluripotency factors continues to be expressed throughout PGC migration, being extinguished only after gonadal colonization (Table 2).

A second common feature of this germline transition is the induction of germ cell factors after gonadal arrival (Fig. 2). For example, in the Tammar wallaby, migratory PGCs continue to express POU5F1 until their colonization of the gonad (Frankenberg et al., 2010). After gonadal arrival, these cells first express DDX4, just as in human and mouse embryos (Hickford et al., 2011). A similar pattern of gene expression is observed in birds and turtles. On gonadal colonization, both DAZL and DDX4 are highly expressed in the chicken germline (Jean et al., 2015), as also occurs in turtles, where DDX4 protein is first detected (Bachvarova et al., 2009). In salamanders, where PGCs remain developmentally uncommitted, expression of PouV genes and Nanog is extinguished by the time PGCs are enveloped within the gonad (Bachvarova et al., 2004; Dixon et al., 2010). Like mammals, the salamander germline first expresses orthologs of Dazl and Ddx4 as PGCs arrive at the future gonad (Bachvarova et al., 2004; Johnson et al., 2001). This developmental transition includes the period when salamander PGCs commit irreversibly (Chatfield et al., 2014). The simplest interpretation of these findings is that the transition at PGC colonization of the gonad occurred in the common ancestor of all tetrapods, and that this functioned to commit the germline to gametogenesis.

This transition also occurs in fish, where expression of nanog, pou5f3 and sox3 declines at PGC colonization, and zygotic expression of dazl first occurs after gonadal colonization (Bertho et al., 2021; Zhang et al., 2019). Although dazl is required for PGC formation (Hashimoto et al., 2004; Li et al., 2016), when medaka or zebrafish zygotes genetically deficient in dazl receive maternally supplied RNA, the resulting PGCs complete their migration to the gonad (Bertho et al., 2021; Nishimura et al., 2018). Once in the gonad, genetically dazl-deficient cells are unable to initiate sexual differentiation or meiosis. These observations, together with genetic studies in dnd1-deficient zebrafish (Gross-Thebing et al., 2017), indicate that fish PGCs remain developmentally uncommitted through their migration, and that zygotic expression of dazl and other conserved germ plasm factors – after gonadal colonization – is required for the initiation of gametogenesis. Taken together with analogous findings in Dazl-deficient mice, these observations suggest that germ cell determination in fish occurs late in development, after PGC colonization of the gonad, and after induction of dazl.

In addition to these transcriptional changes, expression of factors required for piRNA biogenesis is also induced shortly after PGC arrival at the gonads. In mice, salamanders and fish, this causes a reorganization of RNA-rich cytoplasmic bodies known as germ granules (Aravin et al., 2009; Eddy, 1975; Ikenishi and Nieuwkoop, 1978; Redl et al., 2021; Voronina et al., 2011).

These comparative findings in diverse vertebrates provide evidence of a widely conserved transition occurring after PGC colonization of the nascent gonad (Table 2; Fig. 2). In many vertebrates, migratory PGCs express a network of pluripotency factors and remain developmentally uncommitted. Upon gonadal colonization, the germline initiates expression of a germ cell-defining program, characterized by the expression of two widely-conserved RNA-binding proteins, Dazl and Ddx4, and the re-organization of germ granules. This sequence of events, broadly conserved among model vertebrates, occurs regardless of whether germline and soma are initially segregated by induction or by maternal inheritance.

Having examined germ cell commitment among diverse model vertebrates, we now explore how these comparative insights contribute to a revised understanding of the developmental origin of GCTs. We present a model whereby GCTs arise from germline cells that, having failed to make this final commitment after arrival at the gonad, retain a broad developmental potential. In humans, we hypothesize that this potential manifests as differentiation to the diverse histologies and cell types that are characteristic of GCTs.

An alternative model states that the first step toward tumorigenesis is the aberrant ‘reprogramming’ of germ cells to pluripotency (see Glossary, Table 1) (Oosterhuis and Looijenga, 2019). This alternative model is grounded in the view that mammalian PGCs are unipotent – built on the observation that mouse PGCs do not produce chimeras when injected into blastocysts (Leitch et al., 2014).

In the following sections, we examine how a failure of germ cell determination accounts for the full range of clinical, genetic and epidemiological observations regarding GCT pathogenesis, without the conceptual complexities of the reprogramming model.

The pathologist Rudolf Virchow first recognized that testicular tumors contain a panoply of cell types and lineages not usually found in close proximity, and he named these tumors ‘teratomas’ (Virchow, 1863). On the basis of this cellular complexity, Virchow postulated that teratomas arise from ‘indifferent elements’ that developed in all possible directions – a concept of cell potential now termed pluripotency. Spontaneous teratomas arise from the germline and are observed in a variety of mammals, most frequently in the gonad (either testis or ovary) or occasionally along the migratory route of PGCs, such as in the adrenal glands or the midline (Anderson and Johnson, 1988; Bishop, 1978; Steiner and Bengston, 1951; Stevens and Little, 1954; Willis, 1938). Spontaneous teratomas also arise in birds, providing further evidence that the tumor-forming potential of the germline is not unique to mammals (Ford et al., 2006; Sheather, 1911). Reflecting an embryonic origin, human GCTs are most prevalent in neonates, pre-pubertal children and young men. Teratomas, together with yolk sac tumors, are commonly referred to as type I germ cell tumors (GCTs), non-seminomatous GCTs (NSGCTs), or pediatric GCTs (Oosterhuis and Looijenga, 2005; Pierce et al., 2018). Type II GCTs are histologically distinct (Box 2), being composed of either a pure seminoma or a mix of cell types comprising both seminoma and non-seminoma, such as choriocarcinoma, embryonal carcinoma, teratoma and yolk sac tumor. Irrespective of their presentation, genetic and epidemiological observations indicate that type I and type II GCTs share a common etiology (see below).

GCTs are among the most heritable of all tumors, underscoring the importance of genetic contributions to GCT susceptibility. Sons or brothers of men with a GCT are at a two- and fourfold increased lifetime risk, respectively, a much greater degree of heritability than for other common cancers (Goldgar et al., 1994; Kharazmi et al., 2015). In men with two first-degree relatives who develop a GCT, the increase in lifetime risk is even greater, being 10- to 26-fold higher than in the general population (Kharazmi et al., 2015). Epidemiological studies find that the incidence of GCTs among US men of European ancestry is five times that among US men of African ancestry (Ghazarian et al., 2015). Despite clear heritability, there is no association between GCT subtype among first-degree relatives (Ghazarian et al., 2015); indicating that GCTs share a common etiology, irrespective of histology. These findings are further reinforced by cases of bilateral GCTs with discordant histology, despite identical genetic and environmental exposures (Thomas et al., 2013).

Given the heritability of GCTs, it was initially anticipated that rare high-penetrance alleles would be found in families with multiple affected members (Leahy et al., 1995). To date, such alleles have not been identified, or shown to account for familial heritability (Litchfield et al., 2018). Instead, genome wide association studies (GWAS) have enabled the discovery of prevalent low-effect alleles that confer heritability (Table 3). Such is the success of GWAS in mapping heritability that meta-analyses have found 49 risk loci that collectively account for 37% of the father-to-son familial risk of GCTs, irrespective of GCT subtype (Litchfield et al., 2017; Wang et al., 2017).

Although GCTs show strong heritability (indicative of risk alleles present in all cells), they simultaneously exhibit a very low rate of de novo mutations occurring in the tumor. Whole-exome studies estimate a mutation rate of ∼0.5 nucleotides per Mb, comparable with other pediatric cancers, but orders of magnitude less than other adult cancers (Shen et al., 2018). Anecdotal evidence further supports a model of GCT pathogenesis that is not built on an accumulation of de novo mutations across the lifespan: unlike many cancers, the risk of developing a GCT decreases substantially after the age of 35 (Znaor et al., 2015).

Taken together, genetic evidence indicates that GCT initiation does not depend on common de novo mutations in the precursor cell. Instead, meta-analyses of GWAS implicate alleles in several biological networks, thereby illuminating the pathways underlying GCT development (Table 3). These analyses implicate pluripotency factors, such as PRDM14, TFCP2L1 and ZFP42, which function in pluripotency (see Glossary, Table 1) of the germline (Loveday et al., 2018; Rezende et al., 2011; Ruark et al., 2013; Wang et al., 2017; Yamaji et al., 2013; Ye et al., 2013). Additional risk alleles are predicted to disrupt genomic binding sites for pluripotency related transcription factors, such as POU5F1 binding to the SALL4 promoter (Litchfield et al., 2017). Robust association with the pluripotency network, which is expressed in PGCs until shortly after colonization of the gonad, points to migratory PGCs, or PGCs that have recently arrived at the gonad, as the cells of origin for human GCTs (Figs 2 and 3).

What role does the pluripotency network play in GCT pathogenesis? Studies in ESCs and induced pluripotent stem cells (iPSCs) find that core pluripotency factors maintain self-renewal by limiting differentiation (Boyer et al., 2005). Mechanistically, this includes the maintenance of a ‘poised’ epigenetic state at the promoters of certain genes, expression of which would otherwise direct lineage commitment to a somatic fate (Bernstein et al., 2006; Lesch et al., 2016). For example, Nanog is required to maintain a repressive H3K27me3 histone mark at many developmental regulators, thereby restraining cells from differentiating (Heurtier et al., 2019). These observations raise the possibility that the pluripotency network functions to shield PGCs from inductive cues encountered along their migratory route through the embryo, thereby preserving the germline. Consistent with this, deletion or knockdown of either Nanog, Pou5f1, Sox2, the miR-290-295 cluster in migratory PGCs, or of Lin28 in embryoid body-derived PGC-like cells, results in a dramatic reduction in the number of cells arriving at the nascent gonad of mice (Campolo et al., 2013; Kehler et al., 2004; Medeiros et al., 2011; West et al., 2009; Yamaguchi et al., 2009; Zhang et al., 2018). The effect of deleting such factors on tumorigenesis has not yet been explored in mice of suitable genetic background (Box 1).

Apoptosis features prominently in germline development and GCT heritability (Table 3). In mice deficient for Bax, a crucial effector of apoptosis, ectopic PGCs are retained along the migratory route, indicating that misplaced PGCs usually undergo programmed cell death (Runyan et al., 2008; Stallock et al., 2003). Although the fate of ectopic PGCs has not been examined, investigators have suggested that a failure of apoptosis may account for the development of GCTs along the midline in young children (Oosterhuis and Looijenga, 2019). Upon their arrival at the nascent gonad, PGCs that do not undertake Dazl-dependent germ cell determination initiate apoptosis in mice (Gill et al., 2011), as is also observed in testes of DAZL-deficient pigs and sheep (McLean et al., 2021; Nicholls et al., 2019). In the 129-strain of mice, some Dazl-deficient germline cells survive, and unilateral testicular teratomas form in one-third of mice. Simultaneously deleting Bax results in bilateral tumors in all 129-strain males (Nicholls et al., 2019). Deletion of Bax enables the survival of cells with a delayed or arrested development, and it increases the incidence of teratomas in mice carrying the Ter mutation (Cook et al., 2009, 2011; Nguyen et al., 2020).

Although BAX is the principal effector of apoptosis in the mouse germline, the human germline also expresses a second effector, BAK1, which is implicated by GWAS in the pathogenesis of GCTs of several subtypes (Poynter et al., 2012; Rapley et al., 2009). Similarly, the receptor/ligand pair KIT and KIT ligand (KITLG) function to protect cells from apoptosis in several developmental contexts, including the mouse germline (Runyan et al., 2006). As with BAK1, polymorphisms in KITLG are implicated by GWAS in the heritability of GCTs, and activating mutations in KIT frequently accompany the transformation of precursor germ cell neoplasia in situ (GCNIS) cells into overt seminomas (Biermann et al., 2007; Kanetsky et al., 2009; Litchfield et al., 2015; Shen et al., 2018; Taylor-Weiner et al., 2016). Downstream of KIT signaling, the AKT and MAPK pathways are similarly implicated in the pathogenesis of GCTs in humans and mice (Kimura et al., 2003; Litchfield et al., 2017; Pierpont et al., 2017; Wang et al., 2017). These broadly consistent findings in mice and humans implicate an escape from apoptosis and the resulting survival of cells that, having failed to restrict their developmental potential, endure in the embryonic gonad (or along the midline) and remain susceptible to tumorigenesis (Fig. 3).

As discussed earlier, germ cell determination restricts germline potential, precluding tumorigenesis. Genetic findings in humans implicate the differentiation of the gonadal soma and the DAZ family of genes in the pathogenesis of GCTs, regardless of histological subtype. As a result of transpositions and duplications, the human XY genome includes six members of the DAZ gene family: four copies of DAZ on the Y chromosome, and the single-copy autosomal genes DAZL and BOLL (Hughes et al., 2012; Saxena et al., 1996). De novo structural rearrangements of the Y chromosome regularly affect DAZ copy number and are a significant cause of spermatogenic failure (Rozen et al., 2012). One common Y-chromosome rearrangement, the gr/gr deletion, removes two copies of DAZ and is associated with an elevated GCT risk, particularly in men with a family history of GCTs (Moreno-Mendoza et al., 2019; Nathanson et al., 2005; Repping et al., 2003). Furthermore, the autosomal gene DAZL is directly implicated in the heritability of GCTs by GWAS (Ruark et al., 2013). Given genetic evidence of the role of Dazl in the shared ancestor of mice and pigs, and of the important role of the DAZ family in humans, the DAZ family likely functions in the restriction of germline potency in humans, precluding tumorigenesis (Fig. 3).

Further support for this model comes from evidence implicating the gonadal soma in inducing germ cell determination. In mice, PGCs arrive at the nascent gonad as it simultaneously differentiates in a Gata4-dependent manner (Hu et al., 2013). The differentiation of the gonad is necessary for the subsequent induction of Dazl in the germline (as also occurs when PGC-like cells are cultured with a gonad), leading to germ cell determination and the initiation of gametogenesis (Hayashi et al., 2012; Hu et al., 2015; Nicholls et al., 2019). Several factors known to function in somatic cells of the nascent gonad, including DMRT1 and GATA4, have been implicated in GCT formation by GWAS (Koster et al., 2014; Litchfield et al., 2017; Turnbull et al., 2010). In mice, these transcription factors function to build and maintain the somatic lineages of the gonad (Hu et al., 2013; Lindeman et al., 2015; Matson et al., 2011). These genetic observations in humans support the model that the embryonic gonad exerts an instructive function in germ cell determination. Epidemiological evidence further implicates the developing gonad, as several syndromes associated with gonadal dysgenesis greatly elevate GCT risk (Pleskacova et al., 2010).

Consistent with an embryonic origin for GCTs, their cellular precursors generally lack de novo mutations. Nonetheless, certain characteristic genetic events drive or accompany the subsequent transformation of benign GCTs to malignancy (Fig. 3). Evidence suggests that a whole genome duplication – likely occurring in benign GCTs – is an early transformation event, followed by several rounds of chromosome arm loss yielding a complex chromosomal complement (Shen et al., 2018; Taylor-Weiner et al., 2016). Although the ploidy of late-stage GCTs is variable, there is often retention (effectively amplification) of chromosome 12p, regardless of tumor histology or subtype (Atkin and Baker, 1983; Shen et al., 2018; Taylor-Weiner et al., 2016). The implicated region of chromosome 12p encompasses the proto-oncogene KRAS (a cluster of factors involved in control of pluripotency, including NANOG, DPPA3 and GDF3; Ezeh et al., 2005) and the cell-cycle regulator CCND2. This characteristic amplification of chromosome 12p is not found in all cells of a GCT (or in all GCTs), suggesting that it is not required for tumorigenesis, but is frequently acquired during tumor progression (Ottesen et al., 2003).

Pure seminomas frequently exhibit mutations in either KRAS or KIT – the only two recurrent sites of mutations (Shen et al., 2018; Taylor-Weiner et al., 2016). Such mutations in KIT are seen in about half of pure seminomas but are rarely observed in non-seminoma or histologically heterogeneous GCTs (Shen et al., 2018). Thus, the initial development of GCTs from the embryonic germline does not require these characteristic genetic changes, which later promote transformation to a malignant phenotype (Fig. 3).

Despite the widespread use of the 129-strain of mice, the genetic basis for spontaneous gonadal teratomas in this strain is unknown (Heaney and Nadeau, 2008) (Box 1). Regardless, the 129-strain has long provided the most tractable model to interrogate the genetics of teratoma development. For example, 129-strain male mice carrying the Steel allele (SlJ, a 650-kb deletion encompassing Kitl) develop teratomas at a rate seven times that of their wild-type littermates (Heaney et al., 2008; Stevens and Mackensen, 1961). In contrast, mice carrying two copies of SlJ, or a deletion of Kit (the receptor of Kitl), are deficient in germ cells and do not develop teratomas (Stevens, 1967). These findings provided the first definitive evidence that teratomas arise from the germline and not from the gonadal soma. Leroy Stevens later isolated a spontaneous mutation in his colony of 129-strain mice, named the Ter allele, where one-third of mice heterozygous for the allele develop testicular teratomas (Stevens, 1973). Subsequent studies mapped Ter to a point-mutation in Dnd1 on chromosome 18 (Asada et al., 1994; Sakurai et al., 1994; Youngren et al., 2005). Although the incidence of teratomas in mice homozygous for Dnd1^Ter is over 90% in a 129-strain background, teratomas do not form when this allele is transferred to the C57BL/6-strain, even when secondary modifiers are included (Cook et al., 2011; Noguchi et al., 1996). Deletion of many genes similarly modifies teratoma incidence in the 129-strain (such as Dazl, Dmrt1, Nanos3 or Tfap2c), but each causes germline apoptosis and sterility on the C57BL/6-strain, and no teratomas arise (Krentz et al., 2009; Nicholls et al., 2019; Noguchi et al., 1996; Schemmer et al., 2013; Suzuki et al., 2008). Although teratomas develop with regularity in the 129-strain, these tumors contain undifferentiated embryonal carcinoma cells (ECCs, characteristic of mixed GCTs) at very low frequency (Schemmer et al., 2013). Conversely, mice deficient in Pten, with a second activating mutation in Kras, develop testicular GCTs with many ECCs (Pierpont et al., 2017). As in humans, these ECCs are sensitive to standard therapy and this model may facilitate the validation of novel agents.

Like human GCTs, spontaneously arising teratomas occur in juvenile mice, reflecting their origin from the embryonic germline (Dawson et al., 2018; Heaney et al., 2012; Pierpont et al., 2017). Further, the strong sex-bias towards males reflects the burden of human disease (Nicholls et al., 2019; Poynter et al., 2010). Despite these similarities, spontaneous GCTs in mice are teratomas or display a mixed GCT histology; no mouse model containing a seminoma component has been described. Although yolk sac and trophoblast elements are rarely seen in mice in vivo, they can differentiate from teratoma-derived cell lines (Stevens, 1960). Given the failure to fully recapitulate a seminoma-like histology, the 129-strain has sometimes been dismissed as irrelevant to human GCTs, particularly those of post-pubertal men (Almstrup et al., 2006; Looijenga et al., 2014; Walt et al., 1993). Despite these differences, the underlying heritability of human GCTs, irrespective of histological subtype, converges with observations in mice (Table 3; Fig. 3). For example, Kitl deletions elevate the incidence of teratoma formation in mice, and GWAS has implicated KITLG in the pathogenesis of GCTs in men (Heaney et al., 2008; Kanetsky et al., 2009; Rapley et al., 2009). Similarly, human GWAS have implicated factors important for germ cell commitment, such as DAZL and DMRT1, the deletion of which dramatically increases teratoma incidence in mice (Krentz et al., 2009; Nicholls et al., 2019; Ruark et al., 2013; Turnbull et al., 2010). Likewise, apoptotic effectors BAK1 and Bax have been implicated in human and mouse GCTs, respectively (Cook et al., 2009, 2011; Nicholls et al., 2019; Poynter et al., 2012; Rapley et al., 2009). Given rapid progress in understanding the heritability of human GCTs, genetic approaches can now be employed to better interrogate GCT development in mouse models. As such, the 129-strain of mice constitutes a valuable resource with which to interrogate the developmental origin and biology of human GCTs.

Complementary observations indicate that GCTs arise from a failure of development: the young age at first diagnosis; high familial heritability; the absence of driver mutations; transcriptional and epigenetic similarity to the embryonic germline; a developmental potential unique to the embryonic germline; and reduced risk with increasing age. Each finding bolsters the view that GCTs arise from the embryonic germline, shortly after gonadal colonization. Which model – a failure of germ cell determination on arrival at the gonad, or the ‘reprogramming’ of unipotent cells to pluripotency – better accounts for the cellular origins of these tumors?

Previous efforts to define the developmental origins of GCTs have pointed to the transcriptional profile or epigenetic state as evidence of an embryonic cell of origin (Oosterhuis and Looijenga, 2019; Sonne et al., 2009). Although GCTs express many core and naïve pluripotency factors (markers normally restricted to migratory PGCs of vertebrates), this expression cannot resolve the precise origin of GCTs; this could indicate that GCTs arise from cells that failed to extinguish such factors following gonadal colonization, or alternatively, from differentiated cells that reacquired their expression. Like gene expression, genomic imprinting status provides clues (Schneider et al., 2001), but interpretation is complicated by ongoing expression of writers and erasers of DNA methylation (DNA methyltransferases and ten-eleven translocation enzymes, respectively) in the developing germline. Thus, it has remained unresolved whether GCTs arise from PGCs that failed to undertake determination on arrival at the gonad, or from more advanced cells that have reprogrammed to pluripotency.

The reprogramming model states that, as the initiating event of tumorigenesis, PGCs must first aberrantly reprogram to pluripotency (Oosterhuis and Looijenga, 2019). This model is grounded in the view that mouse PGCs are unipotent. Evidence to the contrary, such as derivation of pluripotent cell lines and teratomas, has been interpreted as confirming the reprogramming (de-differentiation) of PGCs to a broader potential not normally held by these cells (Leitch and Smith, 2013). Although a large network of factors sufficient to reprogram somatic cells to pluripotency – exemplified by iPSCs, which revert across the ‘Weismann barrier’ – have been interrogated at great depth, most (if not all) are expressed in mammalian PGCs and other model vertebrates. Thus, any reprogramming would be elicited by factors that are normally expressed by PGCs. To account for the inefficiency of reprogramming, it has been argued that reduced apoptosis, such as occurs in Bax-deficient mice, serves to increase the number of PGCs available for reprogramming (Oosterhuis and Looijenga, 2019). Yet deletion of Bax alone does not produce any obvious increase in the incidence of teratomas in otherwise susceptible mice (Nicholls et al., 2019). In mice additionally depleted of Dazl (that fail to undertake determination), the incidence of bilateral teratomas far outweighs any change in the number of PGCs. Conversely, the Ter mutation to Dnd1 causes a large deficiency of PGCs arriving at the gonad, yet the incidence of teratomas in mice carrying this allele is remarkably high (Sakurai et al., 1995; Stevens, 1973). The reprogramming hypothesis would further require that Dazl is necessary to prevent the reprogramming of unipotent cells to pluripotency.

Integrating findings from diverse vertebrates, we advocate a simpler and more parsimonious model: that mammalian PGCs are not committed to gametogenesis. In our model, newly gonadal PGCs that fail to undertake germ cell determination remain pluripotent and are susceptible to tumor formation. The broad array of cell lineages and histologies found in human GCTs (and the absence of driver mutations at tumor initiation) reflect the broad potential of migratory PGCs across Vertebrata. We propose that the survival of such pluripotent cells is a first step of tumorigenesis (Fig. 3), before transformation or differentiation to an overtly tumorigenic phenotype (Dawson et al., 2018).

The development of GCTs, from uncommitted cells of the germline, before sexual differentiation, may also account for the common pathogenesis of ovarian and testicular GCTs (Box 3). We reason that this model better accounts for the broader range of evidence with fewer assumptions, consistent with the demands of Occam's razor. Although the precise origin of human GCNIS cells remains to be definitively established (see Box 2), observations in mice and other vertebrates provide the first genetic evidence for how germ cells are determined – a process that prevents tumorigenesis.

During the past decade, insights and resources from a range of model organisms have enabled a re-evaluation of germline development and the origin of GCTs. A recurring theme emerges: that human GCTs arise from a failure of conserved processes operating in the embryonic germline of mammals and non-mammalian vertebrates. Of importance, several processes that influence germ cell commitment in model organisms are also implicated in the pathogenesis of human GCTs. These processes include pathways that maintain the developmental potency of the germline, apoptotic cell death and the soma-derived induction of germ cell commitment (Table 3).

Current models assume that PGCs are unipotent germ cells, rationalizing observations to the contrary as artifacts of aberrant reprogramming of PGCs to pluripotency. The arguments for the unipotency of PGCs are belied by the demonstrable potential of migratory PGCs in fish, frogs, salamanders, mice and humans. Furthermore, the derivation of EG cell lines – without genetic transformation – indicates that migratory PGCs of mice readily give rise to pluripotent cell lines. No reprogramming of PGCs before tumorigenesis is required if migratory stage-PGCs are developmentally uncommitted cells of the germline. This simplified model is consistent with other childhood cancers, such as pediatric brain tumors, for which an abundance of evidence suggests that they arise from cells blocked in a self-renewing progenitor-like state (Jessa et al., 2019).

Germ cell tumors remain a significant societal burden. Despite the stunning success of current therapy – with a five-year survival rate in excess of 95% (Stang et al., 2013) – 10,000 men will succumb to testis cancer each year (Global Burden of Disease Cancer Collaboration et al., 2017). Given the young age at diagnosis, GCTs are responsible for many years of healthy life lost. Further, the global incidence of GCTs has been rising steadily in recent decades, with the causes likely to be found in embryonic and fetal development (Skakkebæk et al., 2001; Znaor et al., 2015). The formulation of new therapeutic and diagnostic approaches, and an understanding of the dramatic increase in GCT incidence, depend upon a deeper knowledge of the origins of these tumors.

The authors thank D. Winston Bellott, Anthony Mahowald, Maria Mikedis and the Biology Committee of the Malignant Germ Cell International Consortium (MaGIC) for helpful discussions; Emile Wogram for translation of German texts; and Andrew Tubelli for illustrations.