SMAD2 and p38 signaling pathways act in concert to determine XY primordial germ cell fate in mice

Primordial germ cells (PGCs) are generated around embryonic day 7 (E7) in mouse embryos (Lawson and Hage, 1994) and retain the potential to become eggs and sperm. Upon migrating to the genital ridges at approximately E10.5, PGCs start to differentiate into sperm or oocytes under the regulation of cues from the somatic environment (Bowles and Koopman, 2007). In ovaries, wingless-related MMTV integration site (Wnt) signaling and retinoic acid (RA) are essential for oogenesis. Wnt signaling plays important roles in promoting female differentiation and suppressing the male pathway as somatic factors (Vainio et al., 1999; Bernard and Harley, 2007; Ottolenghi et al., 2007). RA signaling drives XX PGCs to enter meiosis by activating stimulated by retinoic acid gene 8 (Stra8), the gene responsible for the initiation of premeiotic DNA replication (Baltus et al., 2006; Koubova et al., 2006; Mark et al., 2008). In fetal testes, SRY expression triggers Fgf9 expression in somatic cells (Kim et al., 2006; Sekido and Lovell-Badge, 2008). FGF9 is essential for the sexual fate determination of XY PGCs because Fgf9 deletion leads to male-to-female sexual reversal (Colvin et al., 2001; Schmahl et al., 2004; Bowles et al., 2010). FGF9 is involved in two events that are essential for PGC fate determination. First, FGF9 suppresses female pathways and permits the expression of Cyp26b1, which encodes a P450 enzyme that directly degrades RA, preventing meiotic entry of XY PGCs (Bowles et al., 2006,, 2010; Kim et al., 2006; MacLean et al., 2007; Ottolenghi et al., 2007). Second, FGF9 initiates Nanos2 expression, a hallmark of male germ cells, to promote further differentiation (Barrios et al., 2010; Bowles et al., 2010). NANOS2 expression is sufficient to induce male PGC fate determination because Nanos2 overexpression in XX PGCs prevents meiosis and forces XX PGCs to enter the male-type differentiation pathway, including the expression of DNMT3L (Suzuki and Saga, 2008). DNMT3L is involved in a male-specific process – genomic imprinting at the mouse embryonic stage (Bourc'his et al., 2001; Bourc'his and Bestor, 2004; Sakai et al., 2004).

Because FGF9 is a signaling molecule and its receptors are expressed in both somatic and germ cells (Bowles et al., 2010), it is unclear whether FGF9 directly influences germ cells or triggers a secondary messenger in somatic cells to regulate PGC differentiation. Recent analysis of Fgf9 and Wnt4 double knockout testes revealed that deletion of Wnt4 from the Fgf9-knockout testes rescued male-to-female sex reversal, suggesting that the main function of FGF9 was to repress the female pathway (Jameson et al., 2012). Consistent with this study, we proved previously that exogenous activin-A promotes male differentiation in the absence of FGF9 signaling (Wu et al., 2013). In addition, nodal/activin-A directly influences germ cells, and the suppression of this signaling pathway using specific inhibitors induces a failure of male differentiation (Souquet et al., 2012; Miles et al., 2013; Wu et al., 2013). Taken together, it is likely that nodal/activin, rather than FGF9 signaling, directly promotes male germ cell fate.

Because of the important roles of nodal/activin signaling, we attempted to elucidate how nodal/activin functioned in XY PGCs. In general, nodal/activin binds to a complex of transmembrane receptors (types I and II) that subsequently phosphorylate SMAD2/3. The phosphorylated SMAD2/3 forms a complex with a common SMAD4, which translocates to the nucleus and activates target genes (Schier, 2003). However, the deletion of Smad4 in XY PGCs results in a much milder phenotype than the suppression of nodal receptors ex vivo, suggesting that a SMAD4-independent nodal pathway exists in XY PGCs (Wu et al., 2013).

p38 mitogen-activated protein kinase (MAPK) signaling is triggered by cellular stresses and extrinsic signals such as inflammatory cytokines and growth factors (Zarubin and Han, 2005). Phosphorylated p38 MAPK (pp38) enters the nucleus and activates certain transcriptional factors (Zervos et al., 1995; Yang et al., 1999; Zhao et al., 1999). During the formation of the anterior-posterior axis, nodal is required for activation of the p38 signaling pathway, which in turn strengthens nodal signaling by phosphorylating SMAD2, and thus increases the level of activated SMAD2 (Clements et al., 2011). Recent studies have revealed a role of p38 signaling in the initiation and maintenance of the expression of the sex-determining gene Sry in somatic cells (Bogani et al., 2009; Gierl et al., 2012; Warr et al., 2012). After sex determination, p38 MAPK signaling works as an essential gatekeeper of meiosis (Ewen et al., 2010). However, the roles of p38 signaling in the induction of the male pathway and the relationship between p38 and nodal/activin signaling are unknown. In this study, we uncovered the distinct roles of Smad2 signaling and p38 signaling in the sexual fate determination of XY PGCs.

We have demonstrated previously that germline-specific deletion of Smad4 caused only small populations of germ cells to lose NANOS2 expression and enter meiosis (Wu et al., 2013). We attributed this phenotype to the low recombination efficiency and/or a SMAD4-independent pathway, which might exist to transmit nodal signals. To distinguish between these two possibilities, we used a ubiquitously expressed Cre transgene (RosaCreERT2 line) to effectively delete Smad4. Although Smad4 was deleted, downregulation of Nodal and Lefty1/2 genes, which are regulated by the nodal signaling pathway, was not observed, suggesting that Smad4 is not the major component for the transduction of nodal signaling in testes (supplementary material Fig. S1).

Next, we asked whether other SMAD proteins are involved in transduction of nodal signaling and male differentiation of XY PGCs. To address this question, Smad2 and Smad3 were conditionally deleted to analyze the effect of each protein on germ cell development. To obtain Smad2/3 double mutant mice, either Smad2^flox/+Smad3^flox/flox/Rosa-CreERT2^+/− or Smad2^flox/+Smad3^flox/+/Rosa-CreERT2^+/− females were crossed with Smad2^flox/+Smad3^flox/+ or Smad2^flox/flox/Smad3^flox/+/Rosa-CreERT2^+/− males. By injecting tamoxifen into pregnant female mice on E9.5 with one injection or E10.5 and E11.5 with two injections, Smad2 and/or Smad3 were successfully deleted in fetal testes. Interestingly, Smad2 deletion alone was sufficient to impede nodal signaling, causing the downregulation of Nodal and Nanos2 expression (Fig. 1A; supplementary material Fig. S2A), and this phenotype was not observed when Smad3 was deleted (Fig. 1A; supplementary material Fig. S2B). Additionally, Smad2/3 double conditional knockout had only a marginal effect on the expression of Nodal and Nanos2 compared with Smad2 conditional knockout testes (Fig. 1A). We concluded that SMAD2 is necessary for the transduction of nodal signaling in male germ cells.

Next, to investigate the fate of Smad2-deficient XY PGCs, we cultured E13.5 testes for 3 days, because many mutant mice did not survive beyond E14.5 in the absence of SMAD2. We found NANOS2-positve cells in 70.8% of TRA98-positive germ cells (1077 out of 1521; n=3) and DNMT3L-positive cells in 63.0% of TRA98-positive cells (1210 out of 1920; n=3) in control testes; however, these populations were decreased to 37.0% (701 out of 1896; n=3) and 31.5% (540 out of 1715: n=3) in Smad2-deficient male gonads (Fig. 1B,C). These results indicate that Smad2 signaling is essential for the induction of male-specific genes.

Previous studies have illustrated that testes with compromised nodal signaling (hypomorphism of the Nodal gene) display reduced pluripotency marker expression and premature differentiation of male germ cells, as judged by increased Dnmt3l expression (Spiller et al., 2012). However, in Smad2-deficient testes, we found that the expression of Dnmt3l was reduced, accompanied by downregulation of Nodal expression (Fig. 1). Therefore, we examined the expression of pluripotency marker genes: SRY-box 2 (Sox2), Nanog, octamer-binding transcription factor 4 (Oct4; Pou5f1 – Mouse Genome Informatics) and undifferentiated embryonic cell transcription factor 1 (Utf1) in Smad2-deficient testes. We did not observe downregulation of these genes at E13.5 (supplementary material Fig. S3). Such a difference might be caused by different genetic backgrounds or/and the different dosage of nodal signals. In the Nodal hypomorphic mice (Spiller et al., 2012), germ cells might still receive nodal signals, whereas Smad2-null germ cells no longer transduce nodal signaling. Thus, we concluded that Smad2 signaling is not involved in the maintenance of pluripotency in our mouse model.

Smad2 signaling might involve the activation of male-specific genes by suppressing RA signaling (Fig. 2A, left panel) or in a manner that is independent of RA signaling (Fig. 2A, right panel). To distinguish between these possibilities, we treated Smad2-mutant testes with AGN193109, a specific inhibitor of RA signaling. If Smad2 signaling is only required for suppressing meiosis, the treatment would promote male differentiation of XY PGCs in Smad2-deficient testes. AGN193109 treatment increased the expression of NANOS2 and DNMT3L in both control and mutant testes, because Nanos2 expression is known to be repressed by RA signaling (Bowles et al., 2010; Saba et al., 2014a). However, we did not observe complete recovery of NANOS2 and DNMT3L expression in mutant testes (Fig. 2B), indicating that Smad2 signaling is responsible for the induction of male-specific genes independent of RA signaling.

To achieve efficient Smad2 deletion, we used Rosa-CreERT2 deleter mice. However, it was difficult to judge whether the results we observed were due to defective Smad2 signaling in XY PGCs or in somatic cells. To address this question, we first investigated whether the deletion of Smad2 affects the expression of SOX9 and forkhead box L2 (FOXL2), which are two key factors involved in the sex determination of somatic cells in testes and ovaries, respectively (Kent et al., 1996; Ottolenghi et al., 2005,, 2007). As shown in supplementary material Fig. S4, we did not observe a significant decrease in Sox9 or increase of Foxl2. In addition, mRNA levels of Bmp2, which is also known as a female somatic factor (Kashimada et al., 2011) were unchanged. Immunohistochemical detection of SOX9 protein indicated that testicular cords were generated normally even in Smad2-deficient testes, suggesting that Smad2 signaling has limited function in the sexual fate decisions of somatic cells. Next, we conditionally deleted Smad2 from germ cells. To this end, Smad2^flox/floxStella-MerCreMer or Smad2^flox/+/CAG-floxed-CAT-EGFP females were crossed with Smad2^flox/+/CAG-floxed-CAT-EGFP, Smad2^flox/+Stella-MerCreMer/CAG-floxed-CAT-EGFP or Smad2^flox/floxStella-MerCreMer/CAG-floxed-CAT-EGFP males. We initially confirmed germ cell-specific recombination by detecting GFP signals in E13.5 testes after injection of tamoxifen at E9.5 and E10.5 (supplementary material Fig. S5A). Using quantitative imaging analysis, we confirmed that 77.6% of GFP^high cells [16.74/(4.84+16.74)%] showed a lower expression of pSMAD2 in mutant testes (supplementary material Fig. S5B-D). pSMAD2 was specifically deleted from GFP^high germ cells in Smad2^flox/floxStella-MerCreMer/CAG-floxed-CAT-EGFP testes (supplementary material Fig. S5E). Under this condition, we found that 80.0% (507 out of 634 cells; n=2) of GFP-positive (Smad2-null) cells were negative for Nanos2 mRNA in the Smad2^flox/flox/Stella-MerCreMer/CAG-floxed-CAT-EGFP testes, whereas only 18.1% of GFP-positive cells (67 out of 370; n=1) were negative for Nanos2 mRNA in the Smad2^flox/+/Stella-MerCreMer/CAG-floxed-CAT-EGFP testes (Fig. 3A,B). To evaluate the fate of Smad2-null germ cells, these testes were further cultured for 2days and the expression levels of DNMT3L were investigated. As expected, the percentage of DNMT3L-negative cells in GFP-positive (Smad2-null) cells was as high as 76.9% (360 out of 468; n=2) in the Smad2^flox/fox/Stella-MerCreMer/CAG-floxed-CAT-EGFP testes, and this percentage was only 14.2% (52 out of 365; n=1) in control testes (Fig. 3C,D). These results indicate that SMAD2 in germ cells is essential for Nanos2 induction and further differentiation in a cell-autonomous manner.

Because meiotic markers were ectopically induced in some of the Nanos2-null germ cells from E15.5 (Suzuki and Saga, 2008), we suspected that some germ cells lacking NANOS2 expression in the Smad2-deficient testes also entered meiosis after this stage. To compare the phenotypes of Smad2- and Nanos2-deficient testes, we re-examined the phenotype of the Nanos2 mutant at E17.5 by detection of two meiotic markers – SCP3 (SYCP3 – Mouse Genome Informatics), a synaptonemal complex protein, and DMC1, a protein involved in the repair of meiotic DNA double-strand breaks (DSBs) (Yoshida et al., 1998; Yuan et al., 2000; Mahadevaiah et al., 2001). At E17.5, we observed that 37.3% (119 out of 319; n=2) of Nanos2-null germ cells expressed SCP3, compared with 3.6% in control testes (58 out of 1621; n=2), whereas DMC1 was not detected in either Nanos2^−/− or Nanos2^+/− testes at E17.5 (supplementary material Fig. S6A,B), suggesting that Nanos2-null germ cells initiate meiosis but are removed before the repair of meiotic DBSs. In Smad2^flox/flox/Stella-MerCreMer testes, we also observed that 26.6% (366 out of 1374) and 25.4% of germ cells were positive for SCP3 (660 out of 2594; n=3) at E16.5 and E17.5, respectively, compared with 2.1% (44 out of 2126) and 5.8% (165 out of 2867; n=3) in control testes; however, the percentage of DMC1-positive cells was only 5.9% (86 out of 1463; n=3) and 1.4% (24 out of 1735; n=2) at E16.5 and E17.5, respectively (Fig. 4; supplementary material Fig. S6C,D). Similarly, DMC1-positive cells were only observed in Smad2^flox/flox/RosaCreErt2 testes after 3 days of culture from E13.5 (supplementary material Fig. S7A). Germ cells in Smad2-deficient testes gradually entered apoptosis, as indicated by the detection of cleaved caspase-3 (supplementary material Fig. S7B,C), resembling the phenotype of Nanos2-null testes (Suzuki and Saga, 2008). These results suggest that Smad2-null germ cells initiate meiosis but fail to further progress meiosis, probably because of low levels of RA signaling.

To analyze the genes downstream of Smad2 signaling that trigger Nanos2 expression, we focused on orthodenticle homolog 2 (OTX2) as a candidate factor, because Otx2 mRNA was specifically detected in E13.5 testes (supplementary material Fig. S8A). In addition, OTX2 proteins were detected in germ cells of E12.5-13.5 testes (supplementary material Fig. S8B). The expression level of OTX2 gradually decreased after E14.5 (supplementary material Fig. S8B).

To determine whether the expression of Otx2 is controlled by Smad2 signaling, we investigated OTX2 expression after blocking Smad2 signaling. As expected, the deletion of Smad2 significantly diminished the OTX2 expression level, suggesting that Otx2 expression is controlled by Smad2 signaling (Fig. 5A). OTX2 might trigger the expression of genes involved in the sexual differentiation of XY PGCs, such as Nanos2. Then, we analyzed the sexual fate of XY PGCs lacking Otx2. Otx2-null mice display embryonic lethality before the formation of gonads (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996). Therefore, we conditionally deleted Otx2 at E10.5 by injecting tamoxifen into Otx2^flox/flox/Rosa-CreERT2^+/− mice. Complete deletion of Otx2 mRNA was confirmed by quantitative (q)PCR at E13.5 (Fig. 5B). Notably, we also observed reduction in the levels of both mRNA and protein of Nanos2 in the mutant testes at E13.5 (Fig. 5B,D). However, the expression of Nanos2 recovered to a similar level to that observed in control testes at E14.5 (Fig. 5C), indicating that Otx2 is not essential for Nanos2 expression. Mice have three Drosophila Otx homologs, Otx1, Otx2 and Otx3, sharing a bicoid-like homeodomain, prompting us to consider whether OTX1 and OTX3 induce Nanos2 expression in the absence of OTX2. Interestingly, we observed that Otx3 is upregulated in Otx2 mutants, implying redundant functions of OTX2 and OTX3 in the induction of Nanos2 expression (supplementary material Fig. S8C).

In addition to OTX2, p38 MAPK is another putative factor activated by nodal signaling. We first investigated the localization of pp38 MAPK by using immunostaining. Previous studies indicated that pp38 is essential for the initiation of Sry expression in somatic cells (Bogani et al., 2009; Gierl et al., 2012; Warr et al., 2012). Consistent with these results, we detected pp38 MAPK in both somatic and germ cells at E11.5 (supplementary material Fig. S9). Although pp38 MAPK was observed in most of the germ cells and in a small population of somatic cells at E12.5, pp38 was exclusively observed in germ cells at E13.5 (supplementary material Fig. S9).

To investigate the involvement of p38 signaling in the sexual differentiation of XY PGCs, we suppressed p38 signaling using the p38 inhibitor SB203580. A previous study indicates that the suppression of p38 signaling leads to the meiotic entry of XY PGCs (Ewen et al., 2010). We further found that the expression levels of two male-specific proteins, NANOS2 and DNMT3L, were significantly reduced after the treatment (Fig. 6E; supplementary material Fig. S10). These results indicate that male germ cells lacking p38 signaling failed to enter the male pathway and initiated meiosis.

Next, we assessed how p38 signaling suppresses meiosis in fetal gonads. One possible mechanism is to regulate the expression of Cyp26b1, a gene encoding an enzyme that degrades RA. If this were the case, the inhibition of p38 signaling would reduce Cyp26b1 expression. However, using either qPCR or in situ hybridization, we did not observe any change in Cyp26b1 expression even when the testes were treated with p38 inhibitor at a high concentration (Fig. 6A-C). We concluded that p38 signaling inhibits meiosis independently of Cyp26b1 expression.

The inhibition of p38 signaling caused induction of meiosis in XY PGCs, implying that the major function of p38 signaling is the suppression of RA signaling, which disrupts XY PGC differentiation (Ewen et al., 2010; Saba et al., 2014a). If this were the case, suppression of RA signaling would rescue the phenotype caused by the loss of p38 signaling (Fig. 6D, top panel). However, if p38 signaling is also required for Nanos2 induction, the loss of RA signaling would not rescue Nanos2 expression (Fig. 6D, bottom panel). To distinguish between these possibilities, the RA and p38 signaling pathways were simultaneously suppressed by the addition of an RA receptor antagonist (AGN193109) and SB203580 to E12.5 testes, and we examined Nanos2 expression after 3 days. Notably, the reduction of Nanos2 expression caused by a high concentration of SB203580 was completely recovered when the RA receptor antagonist was present (Fig. 6E; supplementary material Fig. S10). From these results, we conclude that p38 signaling protects XY PGCs from harmful RA signaling but does not trigger Nanos2 expression. It should also be noted that once Nanos2 is induced, NANOS2 is responsible for the suppression of meiosis in germ cells (Fig. 9) (Suzuki and Saga, 2008; Saba et al., 2014b).

As demonstrated previously, p38 and Smad2 signaling exhibited distinct functions in the promotion of XY PGC differentiation, implying that these signaling pathways work independently. Indeed, pp38 persisted in the Smad2-null testes, in which nodal signaling was dramatically repressed (Fig. 7A-C), suggesting that Smad2 signaling is dispensable for the activation of pp38. Moreover, a negligible change in Nodal expression was observed after p38 inhibitor treatment (Fig. 7D). In addition, signals of phosphorylated SMAD2 persisted after inhibitor treatment, indicating that p38 signaling does not affect the initiation or maintenance of Smad2 signaling (Fig. 7E). Therefore, Smad2 and p38 signaling work in distinct cascades.

Finally, to investigate the cooperative roles of p38 and Smad2 signaling in the promotion of the male pathway, we cultured E12.5 Smad2-deficient testes with the p38-specific inhibitor SB203580 for 2 days (Fig. 8A). We found that Stra8 expression was dramatically increased in SB203580-treated control testes but not in Smad2-deficient testes, suggesting that p38 but not SMAD2 is required for suppression of meiosis at the stage around E14.5 (Fig. 8B). Although the suppression of Smad2 but not p38 signaling affected Nodal expression (Fig. 7D, Fig. 8C), either the inhibition of p38 signaling or the deletion of Smad2 caused the reduction of Nanos2 expression, suggesting that the suppressor of RA signaling (p38) and Smad2 signaling that works independently of RA are both required for Nanos2 expression. Moreover, the suppression of p38 signaling in Smad2-mutant testes completely impeded Nanos2 and Nodal expression and significantly induced Stra8 expression (Fig. 8B-D), demonstrating the cooperative roles of those two factors.

For male sex determination of mouse PGCs, two events are essential – the induction of Nanos2 expression and the suppression of meiosis. In this study, we revealed that two intrinsic signals, SMAD2 and p38, act to induce Nanos2 expression and to suppress meiosis, respectively. Although we provided genetic evidence that SMAD2 is required for PGCs to initiate Nanos2 expression, it is still unknown how SMAD2 induces Nanos2 expression. One possibility is that SMAD2 promotes the expression of OTX2, which in turn induces Nanos2 expression. Indeed, deletion of Otx2 caused a transient reduction in Nanos2 expression, implying that OTX2 might directly control Nanos2 transcription (Fig. 5B,D). However, the redundant role of OTX family members limits the ability of the analysis to fully prove our idea. Interestingly, OTX2 alone could not activate the Nanos2 enhancer in 293T cells as shown by using a luciferase assay, implying the presence of factors that cooperate with OTX2 to induce Nanos2 expression (data not shown). Using a luciferase assay in HeLa cells, it was demonstrated that LIM class homeodomain transcription factor Lim1 (LHX1 – Mouse Genome Informatics) can directly bind to OTX2 and enhance OTX2-mediated gene expression (Nakano et al., 2000). LHX1 is involved in the localization of PGCs in the embryonic hindgut before they migrate into the genital ridge (Tanaka et al., 2010). Further work will be performed to uncover whether these factors cooperate to induce Nanos2 expression.

Mechanisms by which meiosis is suppressed in XY PGCs are likely to be stage dependent. p38 signaling and CYP26B1 were essential for meiotic suppression before NANOS2 expression. The inhibition of p38 signaling resulted in meiotic entry and the failure of Nanos2 induction in XY PGCs. RA contributes to this phenotype because the suppression of RA signaling by a retinoic acid receptor (RAR) inhibitor completely recovered Nanos2 expression. Therefore, p38 signaling is required for the suppression of RA signaling. Given that Cyp26b1 is expressed normally in somatic cells (Fig. 6A-C), it is unlikely that RA from the mesonephros is responsible for meiotic entry in the absence of p38 signaling. How is meiosis induced in the absence of p38 signaling if RA is degraded by CYP26B1? One possible explanation for these results is that RA can be synthesized by the PGC itself. Consistent with this hypothesis, a recent study suggested that E11.5 XX gonads without a mesonephros still enter meiosis under culture conditions devoid of any retinoids (Guerquin et al., 2010). However, a RA reporter mouse line expressing a lacZ controlled by a RA response element in the RARβ gene (RARE-LacZ mice) displays no lacZ signals in PGCs (Bowles et al., 2006), indicating that the level of RA synthesized by PGCs is extremely low or RA is processed by receptors other than RARβ. It is also plausible that RA from the mesonephros remains because of incomplete degradation by CYP26B1 in somatic cells, because we could not exclude the possibility that p38 inhibitor affects the enzyme activity of CYP26B1. Moreover, recent studies also provided a controversial model that RA is required for meiotic initiation in male germ cells but not female germ cells by genetic deletion of retinaldehyde dehydrogenase 2 and 3 (Raldh2/3; Aldh1a2/3 – MGI) (Kumar et al., 2011; Raverdeau et al., 2012). However, their data could not exclude the possibility that Raldh1 was activated under those conditions.

After NANOS2 is expressed at E14.5, NANOS2 is responsible for suppression of meiosis, and the deletion of Nanos2 leads to meiotic initiation from E15.5 (Suzuki and Saga, 2008). We proved that in Nanos2- or Smad2-deficient testes, even germ cells initiate meiosis but fail to further progress meiosis. These germ cells might gradually be removed by apoptosis. Such a phenotype has also been observed in the mice without genes encoding two types of prostaglandin D2 synthase. In these mutant testes, reduction in the expression of Nanos2 and Dnmt3l is accompanied by the induction of Stra8 expression, whereas the expression of Dmc1 is not upregulated (Moniot et al., 2014). Germ cells in Cyb26b1 knockout testes also fail to induce Nanos2 and Dnmt3l expression. However, germ cells at the pachytene stage could be detected in these mutant testes even when apoptosis was also induced (MacLean et al., 2007). These results imply that a high enough dosage of RA might not only be required for meiotic initiation but also important for meiotic progression. Notably, a small population of DMC1-positive cells were observed in Smad2- but not Nanos2-deficient testes, implying that SMAD2 might function (even in a limited manner) to suppress the RA signaling pathway independently of Nanos2 induction.

Comparing results from Smad2, Smad3 and Smad4 conditional knockout mice, we concluded that Nodal expression was dependent on SMAD2 but not SMAD3 or SMAD4 (Fig. 1A; supplementary material Figs S1, S2). This was unexpected because SMAD2 is generally known to act in concert with SMAD4. However, there are several reports that contradict this notion. In human embryonic stem cells, TGFβ/activin signaling requires SMAD2 instead of SMAD3/4 for the maintenance of pluripotency (Avery et al., 2010; Sakaki-Yumoto et al., 2013). These results indicate that the function of SMADs in nodal/activin signaling might be highly dependent on context.

Ubiquitous deletion of Smad4 did affect Nodal expression in fetal testes globally (supplementary material Fig. S1). However, when Smad4 is deleted in a germline-specific manner, a small number of germ cells lose Nanos2 expression and enter meiosis (Wu et al., 2013), suggesting that the SMAD4-dependent signaling pathway might also play a role in male germ cell development. In contrast to the limited function of SMAD4 in germ cells, previous studies have proved that SMAD4-dependent activin signaling is essential for Sertoli cell proliferation and testicular cord expansion (Archambeault and Yao, 2010,, 2014; Liu et al., 2010; Archambeault et al., 2011). Our previous study showed that suppression of nodal/activin signaling by inhibitor SB431542 causes upregulation of female-specific genes in somatic cells (Wu et al., 2013), which was not observed in Smad2-deficient testes (supplementary material Fig. S4). Because SB431542 inhibits type I activin receptor-like kinase (ALK) receptors ALK4, ALK5 and ALK7 (ACVR1B, TGFBR1 and ACVR1C, respectively – Mouse Genome Informatics), we speculate that downstream factors of these receptors other than SMAD2 might play roles in the suppression of the female pathway in somatic cells.

Our results also contribute to an understanding of the mechanism by which Smad2/3 expression is regulated. In fetal testes, Smad3 deletion caused downregulation of Smad2 mRNA expression, indicating that SMAD3 might be required for positive regulation of Smad2 expression (Fig. 1A). However, the reduction in the level of Smad2 mRNA did not affect its ability to transduce nodal signaling because of the normal level of Nodal expression in Smad3-null testes (Fig. 1A). By contrast, the Smad2 deletion slightly increased Smad3 expression at E14.5 (supplementary material Fig. S2A), indicating that SMAD2 negatively regulates Smad3 expression. Further investigations are required to uncover the relationships among SMAD2, SMAD3 and SMAD4 in nodal/activin signaling transduction.

It is still unknown how p38 and Smad2 signaling is initiated. Previous research suggested that FGF9 induces the expression of Cripto (TDGF1 – Mouse Genome Informatics), a co-factor of nodal, implying that FGF9 triggers nodal signaling in XY PGCs (Spiller et al., 2012). However, the involvement of Cripto in germ cell development is unclear. By contrast, XY PGCs in testes lacking Fgf9 and Wnt4 enter the male pathway normally, excluding the possibility that FGF9 is required for nodal/activin activation and male differentiation (Jameson et al., 2012). It has been reported that Wnt signaling suppresses Fgf9, Cyp26b1 and activin expression in female somatic cells (Kim et al., 2006; Bernard and Harley, 2007; Ottolenghi et al., 2007). Therefore, we speculate that re-repression of Wnt signaling (or its downstream factors) by FGF signaling is essential for the activation of nodal and p38 signaling in XY PGCs (Fig. 9). Based on the previous findings and our current results, we propose a model for the sexual fate decision of XY PGCs (Fig. 9). At around E10.5-E11.5, FGF9 secreted from Sertoli cells is required for the activation of nodal and activin expression in germ cells by direct or indirect suppression of Wnt4. After the activation of nodal/activin signaling at around E12.5, SMAD2 acts as an inducer to trigger Nanos2 expression and cooperates with p38 signaling to permit Nanos2 expression through the antagonization of RA signaling. SMAD4 might also be involved in this process with limited function. Endogenous RA from germ cells might exist and impede male differentiation through the induction of meiosis and mitosis (Saba et al., 2014a).

Our studies focus on intrinsic factors that are essential for the sex determination of XY PGCs. Because PGC-like cells have been successfully induced from induced pluripotent stem cells (iPSCs) (Hayashi et al., 2011), our results might have practical implications in the induction of gametogenesis in vitro.

ICR strain mice (Clea Japan) were used to analyze gene expression patterns and were also used in all embryonic gonadal culture experiments. The generation of floxed Smad2, Smad3, Smad4 and Otx2 alleles has been described previously (Zhu et al., 1998; Tian et al., 2002; Yang et al., 2002; Liu et al., 2004; Li et al., 2008). These mutant mice are mixed background. The Rosa-CreERT2 mouse line was purchased from Artemis Pharmaceuticals. Stella-MerCreMer mouse (Hirota et al., 2011), CAG-floxed-CAT-EGFP mouse (Kawamoto et al., 2000) and Nanos2 mutant (Tsuda et al., 2003) lines were established previously. Tamoxifen was diluted in sesame oil (Nacalai Tesque) at a concentration of 10 mg/ml, and 0.5 ml of the diluted tamoxifen was injected each time. All mouse experiments were approved by the Animal Experimentation Committee at the National Institute of Genetics.

Gonads were cultured in 24-well culture plates with DMEM containing 10% horse serum at 37°C on 5 μm nucleopore filters (Hiramatsu et al., 2010; Harikae et al., 2013). Some gonads were cultured in medium containing SB203580 (Sigma-Aldrich), 4-hydroxytamoxifen (Sigma-Aldrich) or AGN 193109 (Toronto Research Chemical). For the experiment shown in Fig. 2B and Fig. 8, 4-hydroxytamoxifen (1 µM) was added to the medium to further delete Smad2 during cultivation.

For immunostaining, gonads were fixed in 4% paraformaldehyde, embedded in OCT compound (Tissue-Tek; Sakura), and sectioned (6 µm) using a cryostat. After the treatment with Target Retrieval Solution (Dako) at 105°C for 15 min and preincubation with 3% skim milk powder in PBST for 30 min, the sections were stained with the following primary antibodies: anti-TRA98 (1:10,000; a gift from Y. Nishimune, Osaka University, Osaka, Japan), anti-cleaved caspase 3 (1:200; Cell Signaling Technology, 9664L), anti-SOX9 (1:250; a gift from Y. Kanai, The University of Tokyo, Tokyo, Japan), anti-pSMAD2 (1:200; Cell Signaling Technology, 3101S), anti-GFP (1:400; Aves Labs, GFP-1010), anti-pp38 (1:200; Cell Signaling Technology, 4511S), anti-UTF1 (1:200; Abcam, ab24273), anti-DMC1 (1:200; Santa Cruz Biotechnology, sc-8973), anti-OTX2 (1:600; R&D Systems, AF1979), anti-SOX2 (1:200; Santa Cruz Biotechnology, sc-17320), anti-DNMT3L (1:200; a gift from S. Yamanaka, Kyoto University, Kyoto, Japan), anti-NANOS2 (1:200; Suzuki et al., 2007) and anti-MVH (1:400; Abcam, ab13840). The sections were then incubated with donkey anti-rabbit IgG, anti-rat IgG, anti-goat IgG or anti-mouse-IgG secondary antibodies conjugated with Alexa Fluor 488 or Alexa Fluor 594 (1:400; Invitrogen). Primary antibodies were diluted in 3% skim milk powder in PBST and incubated overnight at 4°C. Secondary antibodies were diluted in PBST and incubated at room temperature for 1.5 h. For double staining of pp38 and MVH, which are both anti-rabbit antibodies, we first detected pp38 signaling by using anti-rabbit pp38 primary antibody (1:200; Cell Signaling Technology), horseradish peroxidase-conjugated anti-rabbit secondary antibody and TSA-Plus Cyanine 3 system (PerkinElmer) signal detection. Then, MVH signaling was detected by anti-rabbit MVH primary antibody (1:400; Abcam) and donkey anti-rabbit IgG secondary antibodies conjugated to Alexa Fluor 488 (1:400; Invitrogen). Slides were mounted for observation under a scanning confocal microscope (Olympus IX83) or a fluorescence microscope (Olympus BX61). For cell counting experiments, signal-positive cells among MVH- or TRA98-positive germ cells on one or two sections were counted for each independent sample.

The primer sets used for synthesizing antisense probes for in situ hybridization were as follows: Cyp26b1 forward, 5′-CGGAGAATGTGC-GCAAGATCCTACT-3′; reverse, 5′-CCGGGTCAAACACATTCACGT-CCTT-3′; Otx2 forward, 5′-TGTCTTATCTAAAGAACCGCCTTAC-3′; reverse, 5′-CAGCATTGAAGTTAAGCTTCCAAGAG-3′. The probe for Nanos2 was as described previously (Tsuda et al., 2006).

Frozen sections were washed with 1× PBS, followed by tissue acetylation for 10 min by incubating slides in the buffer generated by adding 2 ml of acetic anhydride (Wako) into 400 ml of the acetylation buffer (0.1 M triethanolamine pH 8.0). Then, tissue was permeabilized by incubating slides in PBST buffer for 30 min (0.1% Triton X-100). After washing sections with PBS, prehybridization buffer [5% dextran sulfate (Sigma), 20% 20× SSC (Ambion), 50% formamide (Wako), 2% Denhardt's (Wako), 10 mg of salmon sperm DNA (Agilent Technologies) in DEPC water] was dropped on the slides and incubated for 2 h. After DIG-labeled probe was added, slides were incubated with a cover slide for 12-16 h at 68°C. After hybridization, slides were washed and incubated with blocking buffer for 1 h (0.5% blocking reagent in PBS) (Roche) and incubated with anti-DIG-POD (Roche;1:200) for 1 h. Signals were amplified by TSA-Plus Cyanine 3 system (PerkinElmer). Whole-mount in situ hybridization was performed by the InsituPro system (M&S Instruments) according to the manufacturer's instructions.

Total RNAs prepared using RNeasy Mini Kits (Qiagen) were used for cDNA synthesis using PrimeScript RT Reagent Kits with gDNA Erase (Takara). PCR was performed with KAPA SYBR FAST qPCR Kits using a thermal cycle dice real-time system.

The primer pairs shown in supplementary material Table S1 were used for PCR amplification.

Primers for Smad2 and Smad3 mRNA have been described previously (Yang et al., 2002; Li et al., 2008). To evaluate mRNA levels, G3pdh was used as a normalization control for Mvh, Smad2, Smad3, Smad4, Bmp2, Foxl2, Sox9 and Cyp26B1, and Mvh was used for other genes.

Testis samples were collected and more than three pairs were pooled. After digestion with 1 mg/ml collagenase and 0.15% trypsin, the samples were mixed with 2× SDS-sample buffer (20 μl/pair) and boiled at 95°C for 5 min. For each sample, 10 μl was separated by SDS-PAGE and then transferred onto Pure nitrocellulose blotting membranes (Pall Corporation). The membranes were incubated with 5% milk in TBST for 1 h and incubated with anti-pSMAD2 (1:1000; Cell Signaling Technology, 3101S), anti-MVH (1:1000; Abcam, ab13840) or pp38 (1:1000; Cell Signaling Technology, 4511S) antibodies overnight at 4°C. After washing, the membranes were incubated with anti-rabbit IgG antibody conjugated to horseradish peroxidase, followed by detection using the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific). Pictures were obtained by the AE-9300H EZ-Capture MG (ATTO). The intensity of the bands was calculated by ImageJ (1.47t).

For imaging analysis, four picture sets (pSMAD2, GFP and DAPI) from four different regions of Smad2^flox/+/Stella-MerCreMer/CAG-floxed-CAT-EGFP testes and eight picture sets from eight different region of Smad2^flox/flox/Stella-MerCreMer/CAG-floxed-CAT-EGFP testes were taken under the same conditions with a scanning confocal microscope (Olympus IX83). These pictures were analyzed with tissue quest cell analysis software (version 4.0) (Tissue Gnostics). We marked single cells by DAPI signals and detected GFP-positive cells according to the manufacturer's instructions. In these GFP-positive cells, we defined a cell as a pSMAD2^high cell if the pSMAD2 signal value in the cell was higher than the mean value of pSMAD2 signal in total GFP-positive cells; if not, we considered the cell as pSMAD2^low cell. Percentages of pSMAD2^high and pSMAD2^low cells were then calculated in control and mutant testes.

For quantitative analyses between two different samples, statistical significance was assessed by using Student's t-test. For quantitative analyses among multiple samples, statistical significance was assessed using one-way ANOVA followed by Tukey's post-hoc tests for selected pairs of genotypes (Fig. 1A, Fig. 2B, Fig. 6E and Fig. 8). Asterisks in figures indicate the levels of statistical significance as follows: *P<0.05; **P<0.01; ***P<0.001; ns, not significant.

We thank Dr Y. Nishimune, Dr Y. Kanai and Dr S. Yamanaka for generously providing anti-TRA98, anti-SOX9 and anti-DNMT3L antibodies, respectively. We also thank Dr Y. Furuta, Dr M. Matzuk and Dr C. X. Deng for helping to import Smad4-flox and Samd2/3-flox mice. We thank Dr M. Saitou and Dr S. Aizawa for providing Stella-MerCreMer and Otx2-flox mice, respectively.