Endogenous Nodal signaling regulates germ cell potency during mammalian testis development

Germ cells in the mouse fetal gonad are normally exposed to molecular signals from the soma that trigger their differentiation into sperm or oocytes (McLaren, 1983). In an ovary, germ cells enter meiosis and commit to oogenesis around 13.5 days post coitum (dpc) (Monk and McLaren, 1981; Speed, 1982). Recent studies have demonstrated that retinoic acid (RA) present in the ovary triggers germ cell entry into meiosis. In a testicular environment, RA is prevented from inducing meiosis by the action of the degradative p450 enzyme CYP26B1 (Bowles et al., 2006; Koubova et al., 2006). Fibroblast growth factor 9 (FGF9) becomes upregulated following the expression of SRY and SOX9 in the pre-Sertoli cells of the developing testis. In addition to its important functions in regulating somatic cell differentiation in the developing testis (Colvin et al., 2001), FGF9 actively suppresses entry into meiosis and stimulates commitment to spermatogenesis in the germ cell lineage (Barrios et al., 2010; Bowles et al., 2010). During normal development, testicular germ cells maintain expression of pluripotency markers, including Oct4 and Sox2, for about 2 days longer than XX germ cells, then cease dividing and upregulate expression of male fate marker genes, including Nanos2 (Tsuda et al., 2003). Although the transition from pluripotency to differentiation in the period 12.5-14.5 dpc is a crucial stage in the development of male fetal germ cells, it is not known how this transition is regulated at the molecular level.

Germ cells are often considered differentiated, uni-potential cells because they usually generate only gametes. However, during some phases of fetal life they are able to revert to functional pluripotency both in vitro and in vivo. Germ cells isolated between 8.0 and 13.5 days post coitum (dpc) and cultured in a combination of stem cell factor (SCF), leukemia inhibitory factor (LIF) and basic fibroblast growth factor (bFGF, FGF2) can form ‘embryonic germ’ (EG) cell colonies (Matsui et al., 1992; Resnick et al., 1992; Durcova-Hills et al., 2006; Durcova-Hills and Surani, 2008; Matsui and Tokitake, 2009). EG cells closely resemble pluripotent embryonic stem (ES) cells derived from the inner cell mass of blastocysts. Furthermore, the fetal origins of testicular cancer hypothesis postulates that XY germ cells that are not controlled appropriately during fetal life can transform into carcinoma in situ (CIS), the pluripotent precursor cells for testicular germ cell tumor (TGCT) (Skakkebæk et al., 1987; Looijenga and Oosterhuis, 1999). In both of these situations, molecular control mechanisms normally involved in suppression of male germ cell pluripotency during fetal life appear to be dysregulated.

Nodal, a member of the TGFβ morphogen family, is a key regulator of early embryonic development known to establish the anterior-posterior axis, specify mesoderm and endoderm during gastrulation, pattern the nervous system and determine left-right asymmetry (Shen, 2007). Nodal signals by binding to activin receptors in the presence of the obligate co-receptor, Cripto (also known as TDGF-1), leading to phosphorylation (activation) of Smad2/3, which regulates transcription of target genes. Lefty1 and Lefty2, also TGFβ molecules, repress the pathway in a dose-dependent manner to limit the range of Nodal signaling. Nodal upregulates its own expression, as well as that of Lefty1 and Lefty2 (Schier, 2003).

The Activin/Nodal class of TGFβs has been linked to the pluripotency of human and rabbit ES cells and pig, mouse and rat epiblast stem cells (EpiSC) in vitro (James et al., 2005; Vallier et al., 2005; Brons et al., 2007; Tesar et al., 2007; Honda et al., 2009; Alberio et al., 2010). In vivo, loss of Nodal leads to loss of Oct4 expression in the epiblast, as well as premature differentiation of epiblast and trophoblast cells (Brennan et al., 2001; Guzman-Ayala et al., 2004; Camus et al., 2006; Mesnard et al., 2006). These observations prompted us to investigate whether Nodal signaling might play a later role in regulating pluripotency in a differentiated cell lineage, the germ cells, during the organogenesis phase of fetal development.

In this study, we show that Nodal signaling is active in XY but not XX germ cells during mouse gonadal development. Genetic suppression of Nodal signaling led to depressed pluripotency marker expression and early XY germ cell differentiation. Levels of Cripto expression and Nodal signaling correlated with propensity of germ cells to de-differentiate and form EG cell colonies in vitro. Finally, although neither NODAL nor CRIPTO are normally expressed in adult human testes, we found aberrant expression in both patient biopsies positive for CIS and the most malignant TGCT subtypes. Our data indicate a previously unknown role for this pathway in male germ cell development and link dysregulation of this pathway with testis cancer.

Protocols and use of animals in these experiments were approved by the Animal Welfare Unit of the University of Queensland. The conditional Nodal floxed allele (Nodal^fl/fl) and the TNAP^cre line were generated as described (Lowe et al., 2001; Lomelí et al., 2000). Embryos were collected from timed matings of the Nodal^fl/fl, CD1 strain, the W^e/W^e mutant strain (Buehr et al., 1993) and the Oct4ΔPE:eGFP (OG2) strain (Szabó et al., 2002), with noon of the day on which the mating plug was observed designated 0.5 dpc.

Total RNA was extracted and cDNA generated from gonadal tissue, cultured germ cells or isolated cell pools as described previously (Bowles et al., 2010). Endogenous control primers used to normalize gene expression levels were Tbp, Hprt, Sdha, Oct3/4 or Ddx4 (also known as Mvh) for Taqman analyses, and Rn18s or Tbp for SYBR analyses; details listed in supplementary material Table S1. Relative transcript abundance was calculated using the 2^−ΔCT method. Error bars represent s.e.m. calculated from the independent biological replicates and statistical significance assessed using unpaired (two-tailed) Student's t-test.

For pure germ cell isolation, gonadal tissue was collected from Oct4ΔPE:eGFP matings at 10.5 and 11.5 (unsexed) and 12.5, 13.5 and 14.5 dpc (sexed). Tissue was dissociated and germ cell (GFP+) and somatic cell (GFP−) populations were isolated by fluorescence activated cell sorting (FACS) using a FACSAria Cell Sorter (BD Biosciences). For isolation of sub-populations expressing α6-integrin, gonadal tissue was collected from CD1 embryos at 12.5 dpc as described previously (Matsui and Tokitake, 2009) with minor modifications. After dissociation, cells were incubated with FITC-conjugated anti-α6-integrin (CD49F; BD Biosciences, 555735) and/or PE-conjugated anti-SSEA-1 (stage-specific embryonic antigen-1; Bio-Scientific, FAB2155P). Germ cells (anti-SSEA-1+) that were positive and negative for anti-α6-integrin were isolated by FACS using an Influx Cell Sorter (BD Biosciences) and processed for RNA extraction. For analysis of germ cell subpopulations expressing α6-integrin in 12.5 dpc Nodal mutant gonads, single gonad pairs were processed as described above and analyzed using a LSRII Analyzer (BD Biosciences). Statistical significance was determined using a one-way ANOVA. Magnetic isolation of germ cells was performed as described previously (Bowles et al., 2010).

Culture of germ cells and urogenital ridges was performed as described previously (Bowles et al., 2010). Briefly, UGRs were cultured on filters (Millipore, TMTP01300, 5 μm) and germ cells were cultured sparsely (~2000 cells per well) in feeder-free 12-well dishes. Additives to culture were FGF9 (Sigma-Aldrich, 25 ng/ml), basic FGF (FGF2, #1104616, Roche; 25 ng/ml), Nodal (recombinant Mouse, R&D Systems, 100 ng/ml) and FGFR antagonist (SU5402, Calbiochem, 5 mM). Control cultures included dilution vehicle.

In vivo labeling with BrdU labeling reagent (Invitrogen, Cat# 00-0103) was performed by injecting pregnant females intraperitoneally with 1.0 ml/100 g body weight of 1× BrdU. Embryos were harvested 2 hours after BrdU injection, fixed in 4% paraformaldehyde in PBS overnight, dehydrated and paraffin embedded. Immunofluorescence of deparaffinized sections (7 μm) was performed as described previously (Bowles et al., 2010). Primary antibodies used were: anti-DDX4 (AB13840, Abcam; 1:1000), anti-E-Cadherin (#610182, BD; 1:200), anti-Cripto (AB19917, Abcam;1:100), anti-phospho-SMAD2 (#3101, Cell Signaling; 1:300), anti-SMAD2 (AB63576, Abcam; 1:100) and anti-BrdU (#555627, BD; 1:100). Secondary antibodies (all from Molecular Probes; 1:200) were: goat anti-rabbit Alexa Fluor 488, goat anti-rat Alexa Fluor 555 and goat anti-mouse Alexa Fluor 594, and DAPI (D8417, Sigma-Aldrich; 1:2000). Sections were examined by confocal microscopy using a Zeiss LSM-510 META confocal microscope. The number of cells DDX4⁺/BrdU⁺ and DDX4⁺/BrdU⁻ (Fig. 3F) and DDX4⁺ cells (supplementary material Fig. S3D) were counted manually in a minimum of 10 sections per gonad of each genotype (Nodal^+/+ and Nodal^fl/fl; n=4 and 4). Pair-wise comparisons were analyzed using two-tailed Student's t-test.

Western blot was performed by standard protocol on whole gonad-only tissue. Two whole gonads (each Nodal genotype; Fig. 3) or four whole gonads (wild-type analysis; Fig. 1 and supplementary material Fig. S1) were used. Antibodies used were: anti-phospho-SMAD2 (#3101, Cell Signaling, 1:1000), anti-TUBA1A (B-5-1-2, Sigma; 1:1000), anti-DDX4 (AB13840, Abcam; 1:1000), anti-NODAL (AB39242, Abcam; 1:200), goat anti-mouse conjugated to horseradish peroxidase (HRP) (Sigma; 1:2000) and goat anti-rabbit conjugated to HRP (Vector Laboratories; 1:2000). Intensity of PSMAD2 bands was normalized against DDX4 and TUBA1A (Fig. 3C) using Image Lab software (Bio-Rad Laboratories; Version 4.0.1).

EG derivation was performed as described (Durcova-Hills and Surani, 2008) with minor modifications. Briefly, 12.5 dpc Nodal^+/+, Nodal^fl/+ or Nodal^fl/fl single gonad-only pairs were dissociated and seeded onto one well of a six-well culture plate (NUNC) containing 4×10⁵ cells/ml of Mitomycin C-treated STO cells. Cells were incubated in PGC growth media (Durcova-Hills and Surani, 2008), changed daily for 10 days. Wells were stained for AP activity using the AP staining kit (Sigma, 86-R) as per manufacturer's instructions. The total number of AP-positive, densely packed colonies was counted for each well. Statistical significance was assessed by one-way ANOVA followed by a post-test for linear trend.

Use of the patient tissues is approved by an institutional review board (MEC 02.981). Samples were used according to the ‘Code for Proper Secondary Use of Human Tissue in The Netherlands’, from the Dutch Federation of Medical Scientific Societies (FMWV, version 2002). Tissues were collected as described (Looijenga et al., 2003), and tumors were diagnosed according to WHO standards by an experienced pathologist (J. Wolter Oosterhuis, Erasmus MC, Rotterdam, The Netherlands). The testicular parenchyma samples investigated have been described previously (Mosselman et al., 1996). qRT-PCR was performed on a 7900HT Fast Real-Time PCR System (Applied Biosystems).

We have previously conducted a genome-wide microarray analysis of gene expression in mouse fetal gonads (Holt et al., 2006). Subsequent analysis showed that a number of genes associated with the Nodal signaling pathway were expressed more highly in XY compared with XX gonads at 12.0 dpc. Using quantitative reverse transcriptase-PCR (qRT-PCR), we confirmed that Nodal, Cripto, Lefty1 and Lefty2 are predominantly expressed in XY but not XX mouse fetal gonads at 12.5-15.5 dpc (Fig. 1A); in all cases, expression was negligible at 11.5 dpc, peaked at 12.5 (Cripto) or 13.5 dpc (Nodal, Lefty1, Lefty2), and had mostly subsided by 14.5 dpc. We also investigated expression of Nodal and Cripto in purified germ cell populations at 11.5-14.5 dpc (to account more accurately for germ cell numbers) and detected peak expression at 13.5 and 12.5 dpc, respectively (supplementary material Fig. S1A,B). Lack of expression of all four Nodal pathway genes in W^e mutant gonads, which are devoid of germ cells, suggested that these genes are either expressed by the germ cells of the fetal testis or depend on the presence of germ cells (Fig. 1B). Expression of all four genes began at least a day earlier than that of Nanos2, which was previously the earliest known male-specific germ cell marker and encodes a protein essential for male germ cell fate commitment (Tsuda et al., 2003; Suzuki and Saga, 2008).

The obligate Nodal co-receptor Cripto was localized to the surface of XY germ cells at 13.5 dpc, consistent with a role in regulating Nodal signaling in the germline (Fig. 1C). Interstitial and other somatic cells of the testis were negative for Cripto expression. In addition to Cripto, the activin receptors ActRIIB (Acvr2b – Mouse Genome Informatics), Alk4 or Alk7 are also required for Nodal signaling (Shen, 2007). Acvr2b and Alk4 were detected in both purified germ and somatic cells in XY and XX gonads at 13.5 dpc (Fig. 1D). Expression of Alk7 or the alternative Nodal co-receptor Cryptic were not detected at appreciable levels at this time (Fig. 1D). Quantitative RT-PCR analysis on purified germ cell and somatic cell populations (supplementary material Fig. S1B) confirmed male-specific Cripto expression restricted to germ cells. Although some activin receptors are broadly expressed, the highly cell-specific Cripto and Nodal expression (Fig. 1B-D; supplementary material Fig. S1B) suggests that Nodal action is limited to germ cells within the gonad.

Using western blot analysis, we could detect Nodal protein in whole XY gonads from 13.5-15.5 dpc (supplementary material Fig. S1C). To investigate whether Nodal signaling is actively engaged in male germ cells, we tested for the presence of phospho-Smad2 (P-Smad2), an indicator of active activin/Nodal signaling. By western blot, P-Smad2 was detected within fetal testes but not ovaries at 12.5-15.5 dpc (Fig. 1E). Using immunofluorescence, we found that P-Smad2 was largely restricted to the germ cells of 13.5 dpc testes (Fig. 1F). Given the restricted expression of Cripto, the low level of P-Smad2 detected in somatic cells is likely to reflect non-Nodal TGFβ signaling in these cells (Mendis et al., 2010). Together, these results indicate that activin/Nodal signaling is active in fetal XY germ cells, but not XX germ cells or gonadal somatic cells, in the crucial 12.5-14.5 dpc period when their proliferation is arrested and commitment to spermatogenic differentiation is stimulated.

Recently, we showed that the signaling molecule Fgf9, which is highly expressed in the developing testis at 11.5 dpc, promotes male germ cell fate (Bowles et al., 2010). In addition, XY, but not XX, germ cells become dependent on Fgf9 by about 12.5 dpc (DiNapoli et al., 2006). Given the upregulation of Cripto in XY germ cells between 11.5 and 12.5 dpc (Fig. 1A) we reasoned that its expression might be triggered by Fgf9. To test this hypothesis, we carried out urogenital ridge (UGR) organ culture experiments: Cripto expression was increased by Fgf9 (Fig. 2A) and decreased in the presence of an FGF receptor antagonist (Fig. 2B). Nanos2, a marker of male germ cell fate (Suzuki and Saga, 2008), was used as a control in these experiments. In XY Fgf9-knockout mice, expression of both Nodal and Cripto was low (data not shown), as would be expected, as XY Fgf9-null gonads (including resident germ cells) undergo complete male-to-female sex reversal (DiNapoli et al., 2006; Bowles et al., 2010).

We next confirmed that Fgf9 acts directly on isolated germ cells to induce Cripto expression (Fig. 2C). Both XX and XY germ cells behaved similarly under these conditions, suggesting that the presence of Fgf9, rather than the chromosomal constitution of the germ cells, ensures that Cripto is specifically induced in male germ cells in vivo.

Although Fgf9 triggered Cripto expression, there was no concomitant upregulation of Nodal expression in isolated 11.5 dpc germ cells cultured for 18 hours (Fig. 2D). However, in intact UGR cultures, which were started at 11.5 dpc and continued for 18 hours, Fgf9 induced expression of both Cripto and Nodal (supplementary material Fig. S2). We hypothesized that the differing observations with respect to Nodal upregulation reflect the close clustering of germ cells in the UGR system versus the sparse distribution of germ cells in purified germ cell culture. We propose a mechanism whereby FGF9 triggers Cripto expression in germ cells and that this allows Nodal, which is initially present at low levels, to feedback and upregulate its own expression in germ cells. Such a mechanism is consistent with our observation that Cripto expression peaks earlier than Nodal expression (Fig. 1A). In support of this model, we found that when dispersed germ cells were cultured in Fgf9 (to trigger Cripto expression) and exogenous Nodal, expression of Nodal was induced (Fig. 2E,F).

In vitro, activin/Nodal signaling has been implicated in the maintenance of the undifferentiated state in mammalian EpiSC and ES cells (James et al., 2005; Vallier et al., 2005; Brons et al., 2007; Tesar et al., 2007; Honda et al., 2009; Alberio et al., 2010). These findings prompted us to investigate whether Nodal signaling might play an analogous role during male germ cell development. To address this question, we attempted conditional deletion of Nodal by crossing Nodal^fl/fl mice, in which Nodal exons 2 and 3 are flanked by loxP recombination sites (Lowe et al., 2001), with the TNAP^cre strain of mice that expresses Cre recombinase in germ cells (Lomelí et al., 2000). We were unable to obtain conditional mutant embryos, perhaps owing to ectopic Cre recombinase expression in the TNAP^cre strain (Kehler et al., 2004; reviewed by Hammond and Matin, 2009) and/or the reduced expression of Nodal (hypomorphism) in the Nodal^fl/fl strain prior to recombination (Lowe et al., 2001). For this reason, we based all further studies on hypomorphic Nodal^fl/fl embryos, which displayed ~30% of wild-type testicular Nodal expression (Fig. 3A) and survive to adulthood.

Immunofluorescence staining for the germ cell marker DDX4 was indistinguishable between Nodal^fl/fl and wild-type gonads at 11.5 dpc (supplementary material Fig. S3A); this timepoint is prior to germ cell expression of Nodal at 12.5 dpc (Fig. 1A), but after the requirement for Nodal during gastrulation. The equivalent DDX4 expression indicates that germ cell specification and migration occur normally in Nodal^fl/fl mutant embryos. DDX4 expression and germ cell numbers were also comparable between genotypes at later stages (13.5, 14.5 and 17.5 dpc) of gonad development (supplementary material Fig. S3A,B). Together, these data indicate that there is no substantial loss of germ cells due to decreased Nodal expression in Nodal^fl/fl embryonic gonads. We therefore questioned whether XY germ cells deficient in Nodal were qualitatively abnormal.

To determine the effect of greatly reduced Nodal signaling in XY germ cells, we first examined P-Smad2 expression in gonads of the hypomorphic Nodal^fl/fl mouse strain. Diminished levels of P-Smad2 (but not SMAD2) in 14.5 dpc XY Nodal^fl/fl gonads relative to wild type (Fig. 3B,C; and data not shown) revealed that the majority of Smad2 activation in the gonad was specifically dependent on Nodal signaling, and was not part of a more generalized response to other TGFβ signaling.

We next assayed markers of pluripotency and male germ cell differentiation in Nodal^fl/fl gonads. We detected reduced expression of pluripotency markers Nanog, Sox2 and Oct3/4, and, conversely, increased expression of male germ cell differentiation markers Nanos2, p15 and Dnmt3l in Nodal^fl/fl gonads, relative to wild type at 14.5 dpc (Fig. 3D). Because this finding suggested early differentiation of germ cells, we examined the total cohort of germ cells 1 day earlier, when arrest was not yet complete and differences in proliferation might be detected. BrdU incorporation revealed that 27.8% of wild-type germ cells were proliferating at 13.5 dpc, in contrast to only 22.3% of Nodal^fl/fl germ cells (Fig. 3E,F). This 5% difference in proliferating germ cell number, though significant, was not manifest in total germ cell numbers at 13.5, 14.5 and 17.5 dpc (supplementary material Fig. S3B), possibly owing to the sensitivity of our counting methods.

Last, because Fgf9 induces Nodal signaling via Cripto expression (Fig. 2), and because it is also important for suppression of XY germ cell meiosis (Bowles et al., 2010), we investigated the meiosis marker Stra8 in our in vivo model of suppressed Nodal signaling. However, in Nodal^fl/fl mutant gonads, Stra8 expression was not induced in XY germ cells at 14.5 dpc (Fig. 3D). Together, our data suggest that when Nodal signaling is diminished, fetal XY germ cells express markers of male fate commitment and cease proliferation earlier than normal.

Previous studies have shown that mouse fetal germ cells are heterogeneous with respect to cell-surface properties and ability to generate pluripotent EG colonies in vitro. Matsui and Tokitake (Matsui and Tokitake, 2009) flow-sorted mouse fetal germ cells based on α6-integrin expression, and identified an α6-integrin-low population of germ cells with 2.5-fold higher proclivity for forming EG colonies. Because our in vivo findings linked Nodal signaling with maintenance of pluripotency marker expression in germ cells, we investigated whether expression of Nodal signaling components characterized this sub-population of germ cells with the greatest potential to undergo reprogramming to stem cells.

We flow-sorted 12.5 dpc germ cells (SSEA-1-positive) into α6-integrin-negative and -positive pools (Fig. 4A; supplementary material Fig. S4A,B), and assessed expression of Cripto, Nodal, Lefty1, Lefty2 and the pluripotency markers Sox2, Oct3/4 and Nanog in each by qRT-PCR. The α6-integrin-negative population displayed higher expression of all markers than the α6-integrin-positive population, with Cripto, Nodal and Lefty2 exhibiting the greatest difference (2.08, 1.93 and 2.04 fold, respectively) of all genes tested (Fig. 4B). These data indicate that, although Nodal signaling genes appear to be expressed by all XY fetal germ cells (Fig. 1C), highest expression of these characterizes the same subpopulation of fetal testicular germ cells previously identified as particularly pluripotent (Matsui and Tokitake, 2009).

We next investigated the effects of reduced Nodal signaling on EG colony derivation. We cultured 12.5 dpc explanted testes from wild-type, Nodal^+/fl and Nodal^fl/fl embryos under EG colony-forming conditions; at this timepoint in Nodal^fl/fl gonads, Nodal and Cripto expression levels are ~20% and ~70% of wild-type levels respectively. Total germ cell numbers, indicated by Ddx4 expression, remained comparable with wild-type gonads (supplementary material Fig. S4C), as did the ratio of α6-integrin-negative to α6-integrin-positive germ cells assessed by flow cytometry (supplementary material Fig. S4D). In the EG colony derivation assay, each wild-type gonad pair generated 3.7±0.35 alkaline phosphatase-positive EG cell colonies (n=7), whereas 1.0±0.33 EG cell colony arose from each Nodal^fl/fl gonad pair (n=6; Fig. 4C). Thus, reduced Nodal signaling leads to reduced stem cell potential of the total XY fetal germ cell population.

EG cell derivation requires the application of three factors: FGF2, LIF and SCF (Matsui et al., 1992; Resnick et al., 1992; Durcova-Hills et al., 2006; Durcova-Hills and Surani, 2008). We found that FGF2, like FGF9, can induce Cripto expression in 10.5 and 12.5 dpc XX and XY germ cells (supplementary material Fig. S4E), suggesting that its role during in vitro transformation of germ cells to EG colonies may be to stimulate Cripto expression, thereby enhancing pluripotency in those cells.

CRIPTO was first identified from a human embryonal carcinoma cell line, NTERA2 (Ciccodicola et al., 1989), and a later study indicated that CRIPTO is expressed in some non-seminoma germ cell tumors (Baldassarre et al., 1997). Most recently, heterogeneous expression of CRIPTO within the human teratoma-derived NTERA2 cell line identified CRIPTO^High-expressing cells as those having the greatest tumorigenic potential and most undifferentiated phenotype (Watanabe et al., 2010). Our data identifying a function for Cripto during normal XY fetal germ cell development led us to speculate that CRIPTO expression may be maintained or re-activated in carcinoma in situ (CIS), a tumor stem cell thought to arise from fetal germ cells that fail to fully differentiate and are retained in a dormant, pluripotent state in the adult testis (Skakkebæk et al., 1987; Looijenga et al., 2003; de Jong et al., 2008; Gillis et al., 2011). Currently, no CIS model exists in mice (Oosterhuis and Looijenga, 2005; Looijenga et al., 2011), precluding a test of this hypothesis in an experimental model system. For this reason, we examined NODAL, CRIPTO and LEFTY expression by qRT-PCR in a range of human testis biopsies containing CIS cells. In support of our hypothesis, we found highest expression of CRIPTO, NODAL and LEFTY1 in those biopsies in which over 90% of tubules contained CIS (Fig. 5A-C). The low levels of normalized gene expression reflect the fact that CIS cells in these biopsies make up a very small proportion of total cells.

All CIS ultimately progresses to either seminoma, a tumor type composed of cells expressing fetal germ cell markers, or non-seminoma, a diverse class of tumors including embryonal carcinoma (EC), yolk sac tumor (YST), teratoma and choriocarcinoma, that express a range of pluripotency markers (Skakkebæk et al., 1987; Looijenga, 2009). Therefore, we next tested whether Nodal pathway genes are upregulated in these various testis cancers. We detected strong expression of CRIPTO and NODAL in the invasive tumor types EC and YST, and varied expression in mixed non-seminomas (mNS; Fig. 5D,E). Little or no expression was associated with the benign tumors teratoma and choriocarcinoma. LEFTY1 was overexpressed only in YST and mNS (Fig. 5F); the absence of the NODAL inhibitor LEFTY1 in EC may underlie its highly malignant phenotype (Postovit et al., 2008). Within the mNS samples, which contain some differentiated and some non-differentiated tissue, levels of CRIPTO, NODAL and LEFTY1 expression were in direct proportion to the pluripotent EC and YST cell content of each tumor (Fig. 5G-I). None of these genes was overexpressed in seminomas or other types of germ cell tumor (supplementary material Fig. S5A-C); the lack of NODAL/CRIPTO expression may indicate that CIS cells reduce their malignant behavior as they progress to seminoma. Together, these results suggest that the Nodal pathway is activated in TGCTs and is specifically associated with germ cell tumor pluripotency and invasiveness, rather than being a cancer marker per se.

The Nodal signaling pathway is known to play important roles in pattern formation during embryogenesis in a number of vertebrate species. The present study expands the known developmental roles of Nodal signaling by demonstrating that this pathway is transiently active in male mouse fetal germ cells just prior to and during the period of their sexual fate determination. We find that the germ cells with the highest propensity to form pluripotent EG cell lines in vitro are those that express the highest levels of Nodal signaling genes. Hence, our studies suggest that the role of Nodal during germ cell development is to regulate the timing of the transition from proliferation to differentiation in the male germline during fetal life. Finally, the observation that Nodal signaling is reactivated ectopically in adult human TGCT provides a potential link between fetal germ cell regulation and adult disease.

Investigating the regulation of Nodal signaling, we found that Fgf9, which is expressed by the somatic cells of the testis, triggers male germ cells to upregulate Cripto expression around 12.0-12.5 dpc, shortly after germ cell colonization of the developing testis (supplementary material Fig. S6). This upregulation enables Nodal, presumably present at low endogenous levels, to upregulate its own expression and that of downstream targets Lefty1 and Lefty2, peaking at 13.5 dpc. We have previously shown that Fgf9 acts to make XY germ cells more resistant to entering meiosis, through an unknown mechanism (Bowles et al., 2010). Hence, an obvious hypothesis was that FGF9 acts on XY germ cells to trigger Nodal signaling and thereby make those cells less likely to enter meiosis. However, this hypothesis is not supported by our data: knockdown of Nodal to 30% of normal levels had no effect on expression of pre-meiotic gene Stra8 (Fig. 3D), although complete deletion of Nodal is required to confirm this. Instead, our data suggest that the role of Nodal signaling in XY germ cells is to extend the period of pluripotency and delay the onset of differentiation. Following this transient expression of Nodal pathway genes, TGFβ and activin signaling evidently act to ensure correct G1/G0 arrest (Mendis et al., 2010; Moreno et al., 2010). The expression of Cripto by XY germ cells may make such cells temporarily resistant to the effect of TGFβ and activin ligands, as has been suggested in other systems (Gray et al., 2003; Gray et al., 2006). Diminished expression of Cripto at these later time points, presumably owing to a decrease in FGF9 signaling, should allow testicular germ cells to become gradually responsive to other TGFβ factors. An interesting possibility raised by our study is that FGF signaling might play a similar role in other situations where Nodal signaling has been implicated.

Our data indicate a role for Nodal signaling in regulating germ cell potency in vivo. First, fetal germ cells from a genetic model of reduced Nodal signaling are compromised in their ability to generate pluripotent EG cell lines in culture. Second, the gonadal germ cell fraction most competent to form EG cell lines in vitro (Matsui and Tokitake, 2009) is that which expresses the highest levels of Nodal signaling genes. These observations are consistent with several in vitro studies linking TGFβ/activin/Nodal signaling to cell pluripotency. James et al. (James et al., 2005) showed that Smad2/3 is active in undifferentiated human ES cells, and that chemical inhibition of activin receptor Alk4/5/7, using SB431542, suppresses expression of the pluripotency marker Oct4 in mouse blastocyst outgrowths during ES cell derivation in culture. Similarly, SB431542 was found to reduce Smad2/3 phosphorylation and Oct4 expression in cultured rabbit ES cells (Honda et al., 2009), and to promote differentiation/suppress pluripotency markers in pig (Alberio et al., 2010) and mouse (Tesar et al., 2007) epiblast stem cells. However, it is important to note that these effects cannot be attributed specifically to Nodal signaling, as SB431542 blocks signaling mediated by the entire TGFβ/Activin/Nodal branch of the TGFβ superfamily.

One factor that is crucial for in vitro derivation of EG colonies from fetal germ cells is FGF2. We found that FGF2, like FGF9, was able to induce Cripto expression in fetal gonadal germ cells. We propose that a role (if not the role) of FGF2 during in vitro transformation of germ cells to EG colonies is to upregulate or induce Cripto expression, thereby triggering Nodal signaling and enhancing pluripotency potential in those cells (Fig. 6). Interestingly, unlike LIF and SCF, FGF2 is required only for the first 24 hours of EG derivation (Durcova-Hills et al., 2006); a merely transient requirement for FGF2 could be explained if its role is simply to trigger the upregulation of Cripto that facilitates amplification of Nodal signaling.

It has recently been reported that Nodal expression in XY germ cells functions to make them resistant to meiotic induction (Souquet et al., 2012). By adding a chemical inhibitor (SB431542) of activin receptors Alk4, Alk5 and Alk7 to cultured mouse fetal testes explanted at 11.5 dpc, Souquet et al. (Souquet et al., 2012) found that some XY germ cells enter meiosis. Conversely, when exogenous Nodal was added to cultured female gonads, a reduction in the number of meiotic germ cells was observed. In our in vivo model of suppressed Nodal signaling, Stra8 expression was not induced in XY germ cells of Nodal^fl/fl gonads at 14.5 dpc (Fig. 3D), as might be expected if the role of Nodal was to suppress meiosis. However, because Nodal expression is reduced, not ablated, in our model, a role in suppressing meiosis can not be excluded. Differences between endogenous and exogenous systems may account for the different conclusions, therefore analysis of a complete conditional null mutant will be required to determine whether Nodal suppresses meiosis in vivo. A further discrepancy between the two studies relates to the temporal profile of Nodal signaling gene expression. Souquet et al. (Souquet et al., 2012) reported that expression of Nodal, Cripto, Lefty1 and Lefty2 is highest at 11.5 dpc and declines thereafter. Our experiments indicated peak expression at 12.5 (Cripto) and 13.5 dpc (Nodal, Lefty1 and Lefty2; Fig. 1A), consistent with human Nodal expression and phospho-SMAD2 activation predominant from 13.5 dpc onwards. The disparity between our gene expression studies can most likely be attributed to differences in normalizer genes used. When normalizing against Ddx4 expression we observed the same expression profile reported by Souquet et al. (Souquet et al., 2012) (data not shown); however, because expression of Ddx4 increases in germ cells over time, it is considered an unsuitable normalizer for dynamic studies of germ cell gene expression (van den Bergen et al., 2009). Using a more appropriate normalizer, Oct3/4, which is stable in XY germ cells over this timecourse (van den Bergen et al., 2009), we observed peak expression in purified germ cells at 12.5 and 13.5 dpc for Cripto and Nodal, respectively (supplementary material Fig. S1B), as we also found when analyzing gene expression in whole gonads (Fig. 1A).

Finally, we found a strong correlation between Nodal pathway gene expression and malignant TGCTs. In previous studies, expression of CRIPTO has been observed in a variety of carcinomas, including mammary, pancreas, intestine and skin cancer (Strizzi et al., 2005). In these situations, CRIPTO has emerged as an oncogenic growth factor, controlling proliferation, migration and survival (Bianco et al., 2005; Wechselberger et al., 2005), but it remains unclear whether this involvement reflects the role of Cripto in Wnt signaling, Src activation through glypican binding or Nodal signaling (Bianco et al., 2010). Our data strongly correlate CRIPTO expression with TGCT, and specifically implicate dysregulation of the Nodal pathway in carcinogenesis, at least in the context of TGCT.

TGCT is an unusual cancer in that it appears to be seeded during fetal life, yet only manifests after puberty and during young adulthood (Skakkebæk et al., 1987; Looijenga et al., 2011). It is widely believed, based largely on chromatin modification studies, that XY germ cells that do not undergo correct differentiation during fetal development are able to survive as dormant pre-CIS cells in the testis until, at puberty, they are reactivated to form CIS cells that then progress to either seminoma or non-seminoma. Our studies indicate that CRIPTO/NODAL expression is activated ectopically in CIS cells and in the most pernicious forms of TGCT in humans (Fig. 6). It is not yet clear whether CRIPTO expression fails to be downregulated in some germ cells during fetal life, or whether CRIPTO expression is re-initiated in some germ cells during adult life. Our results strongly support a model whereby TGCT is triggered by ectopic action of a program that normally operates only during fetal germ cell development.

Our results have important implications for the etiology and management of testicular cancer. Mechanistic studies of testis cancer biology have been hampered by the lack of a mouse model of CIS, and our results raise the possibility that sustained expression of Cripto in transgenic mouse germ cells may provide such a model. Second, being a cell-surface protein, CRIPTO may prove a tractable target for testicular cancer treatment, as it has proven for various CRIPTO-positive solid tumors (Kelly et al., 2011). Finally, detection of cleaved CRIPTO in human semen may offer a non-invasive method of clinical diagnosis of a variety of testis cancers, a possibility now under investigation.

We thank M. Kuehn for providing Nodal^fl/fl mice, A. Yap, R. Krumlauf and P. Tam for comments on the manuscript, and T. Davidson for mouse colony maintenance. Confocal microscopy was performed at the Australian Cancer Research Foundation Dynamic Imaging Centre of Cancer Biology. Flow cytometry was performed with the assistance of the Queensland Brain Institute Flow Cytometry facility.