Sexual Fate Change of XX Germ Cells Caused by the Deletion of SMAD4 and STRA8 Independent of Somatic Sex Reprogramming

The differential programming of sperm and eggs in gonads is a fundamental topic in reproductive biology. Although the sexual fate of germ cells is believed to be determined by signaling factors from sexually differentiated somatic cells in fetal gonads, the molecular mechanism that determines germ cell fate is poorly understood. Herein, we show that mothers against decapentaplegic homolog 4 (SMAD4) in germ cells is required for female-type differentiation. Germ cells in Smad4-deficient ovaries respond to retinoic acid signaling but fail to undergo meiotic prophase I, which coincides with the weaker expression of genes required for follicular formation, indicating that SMAD4 signaling is essential for oocyte differentiation and meiotic progression. Intriguingly, germline-specific deletion of Smad4 in Stra8-null female germ cells resulted in the up-regulation of genes required for male gonocyte differentiation, including Nanos2 and PLZF, suggesting the initiation of male-type differentiation in ovaries. Moreover, our transcriptome analyses of mutant ovaries revealed that the sex change phenotype is achieved without global gene expression changes in somatic cells. Our results demonstrate that SMAD4 and STRA8 are essential factors that regulate the female fate of germ cells.


Introduction
Primordial germ cells (PGCs) that emerge in the early mouse embryo have the capability to become either spermatocytes or oocytes and are enclosed by somatic cells in embryonic testes or ovaries. The sexual fate of germ cells in gonads is thought to be determined by factors derived from somatic cells because somatic sex determination precedes germ cell sex determination, and sex reversal in somatic cells leads to sex reversal in germ cells [1]. However, when and how germ cells receive such signals, and the determinants of germ cell sex, remain unknown. Understanding the mechanism of sex determination in germ cells requires elucidation of the pathways associated with male and female differentiation events.
The earliest sign of sexual differentiation in male germ cells is the entry into cell cycle arrest that is likely due to the suppression of retinoic acid (RA) signaling caused by upregulation of the RA-metabolizing enzyme CYP26B1 in somatic cells [2][3][4]. Simultaneously, the male-specific factor NANOS2 is induced by mothers against decapentaplegic homolog (Smad) 2 signaling in germ cells, and the expression is maintained during the embryonic stage [5][6][7][8]. The sex of germ cells in testes is determined by NANOS2 function, as the absence of NANOS2 impedes male differentiation and results in the induction of female characteristics, such as meiotic initiation and upregulation of genes associated with oocyte differentiation [9,10]. Inversely, if Nanos2 is ectopically induced in female germ cells, the cells fail to enter meiosis and begin male-specific gene expression, such as DNA methyltransferase 3-like protein (Dnmt3L) and MIWI2, which are required for male-specific de novo DNA methylation in embryonic testes [10][11][12][13]. The somatic factor sex-determining region Y (SRY) and its downstream effectors SRY-box (SOX) 9 and fibroblast growth factor 9 (FGF9) are essential for the male fate decision because their deletions lead to male-to-female sex reversal, germ cells included [14][15][16][17][18][19]. Among these factors, FGF9 is considered to work directly in germ cells to induce Nanos2 [16,20]. However, a recent report showed that the deletion of a female-specific gene, wingless-related MMTV integration site 4 (WNT4), in Fgf9-null testes rescues the switch from male to female, which suggests that FGF9 acts by suppressing WNT4 expression rather than by actively directing male fate [21]. Thus, the signal that induces the male fate decision remains unknown.
The sexual fate of germ cells in ovaries appears to be induced by two distinct processesmeiosis and oocyte differentiation-that proceed simultaneously [22]. Upon receiving an RA signal, XX germ cells express genes such as stimulated by retinoic acid 8 (Stra8) and Rec8 [2,23,24]. After pre-meiotic DNA replication controlled by STRA8, germ cells enter meiotic prophase I, during which homologous chromosome pairing and recombination occurs in a series of stages: leptotene, zygotene, pachytene, and diplotene. Therefore, sexual differentiation in the ovary is associated with meiotic initiation, a process that is never observed in wild-type testes during the embryonic stage. In Stra8-deficient ovaries, XX germ cells enter neither meiosis nor mitosis and retain DNA content as 2N [23]. However, a recent study showed that oocyte-like cells are generated without premeiotic replication in Stra8-deficient ovaries, which indicates that the feminization of germ cells occurs independent of meiosis and that factors other than RA must be involved in the induction of oocyte differentiation [22]. FOXL2 and WNT4 are possible somatic factors required for female fate determination, as evidenced by the sex reversal of XX PGCs caused by the deletion of Foxl2 and Wnt4 [25]. However, it is controversial whether the effector of Wnt signaling, β-catenin, functions in somatic cells or germ cells [26][27][28]. Therefore, the somatic factors downstream of FOXL2 and WNT4 signals that directly induce oocyte differentiation are unclear.
To clarify the signals that lead to the sexual determination of germ cells, we have focused on the transforming growth factor β (TGFβ) signaling pathway. We have shown that nodal/activin signaling is activated in germ cells in embryonic testis but not in ovaries, and is required for the induction of the male determinant NANOS2 [5,8]. In this study, we investigated the possible role of bone morphogenetic protein (BMP) signaling in the sexual differentiation of germ cells in ovaries because a previous study suggested that Bmp2 expression was under the control of Wnt signaling, although the function of BMP2 remains unknown [29]. BMP signaling is mediated by SMAD 1, 5, or 8 and their common mediator Co-SMAD4 [30]. To determine the function of BMP signaling, we deleted Smad4 in a germ-line-specific manner. We here provide evidence that Smad4 signaling plays essential roles in the induction of oocyte-specific regulators as well as in meiotic progression. Notably, the suppression of Smad4 and RA signaling in germ cells is sufficient to direct the cells to take a male pathway even in the absence of male somatic cells.

Smad4 Signaling Is Required for Germ Cell Survival in the Fetal Ovary
To investigate any involvement of BMP signaling in the fetal ovary, we used immunostaining to examine the expression pattern of phosphorylated SMAD (pSMAD) 1/5/8. pSMAD1/5/8 was detected in both germ and somatic cells in ovaries from E11.5 to E13.5, suggesting that BMP signaling is activated in both cell types (S1A Fig). pSMAD1/5/8 requires SMAD4 to activate downstream genes even though these factors can enter the nucleus upon ligand binding [31]. Thus, we used a floxed allele of Smad4 encoding the Co-SMAD protein to analyze the function of BMP signaling in ovaries. Although SMAD4 may mediate other TGFβ-signaling pathways mediated by SMAD2 and 3, previous studies have shown that phosphorylated SMAD2 was not detectable in ovaries at this stage [5,32]. Therefore, we reasoned that the deletion of Smad4 may disrupt pSMAD1/5/8-dependent BMP signaling rather than pSMAD2/ 3-dependent signaling in the ovary, but we do not exclude the possibility that SMAD4 is also involved in other signaling pathways.
Smad4 was deleted in a germ cell-specific manner by using a transgenic Stella-MerCreMer line [33], in which Cre recombinase activity was induced by tamoxifen (TM) administration under the regulation of the element that controls Stella (a gene specifically expressed in germ cells) expression. We injected TM at E9.5 and E10.5 to delete Smad4 in germ cells (Fig 1A). We refer to this mouse model as the Smad4 (Stella) mutant unless otherwise specified. The efficiency of Cre activity was variable, but was as high as 80% based on the reporter gene expression (S1B Fig). We found that the germ cell number decreased in the mutant ovaries at E16.5 and E17.5 (Fig 1B and 1C). As the reduction of female germ cells was correlated with the increase of cleaved-caspase3 positive germ cells at E16.5 (Fig 1D), it is likely that the observed germ cell loss was due to apoptotic cell death. These results indicate that Smad4 is required for female germ cell survival.

Crucial Role of Smad4 Signaling in Female Germ Cell Differentiation Independent of RA Signaling
To investigate the cause of germ cell death and understand the function of SMAD4, we examined the gene expression changes that occurred before apoptosis. As female differentiation consists of meiosis and oocyte differentiation pathways, we first examined the expression of Stra8, a gene involved in pre-meiotic DNA replication [3,23]. In normal ovaries, STRA8 is expressed in germ cells from E13.5, and most of the germ cells lost this expression at E15.5 due to the progression of meiotic prophase I (Fig 2A). Even in Smad4 (Stella) mutant ovaries, STRA8 expression was observed from E13.5 as well, indicating that the deletion of Smad4 does not interfere with RA signaling. However, its expression was retained up to E15.5 (Fig 2A), suggesting that Smad4-deficiency influences the progression of meiosis.
Next, we investigated the expression of genes involved in oocyte differentiation. We found that the expression of Figla, a key transcription factor required for primordial follicle formation, was decreased in Smad4 (Stella)-mutant ovaries; however, two other regulators, Lhx8 and Sohlh1 [41], were either not decreased (Lhx8) or significantly increased (Sohlh1) at E14.5 ( Fig  3A). Although apoptosis occurred in Smad4-null germ cells at E16.5, some germ cells still survived. However, the number of newborn ovary homeobox protein (NOBOX)-and Forkhead box O3a (FOXO3A)-positive germ cells was fewer in Smad4-deficient ovaries than in the control at E17.5 (Fig 3B and 3C). These data suggest that SMAD4 plays a role in oogenic gene expression, although it is unclear whether SMAD4 directly controls these gene expressions or not.

SMAD4 in Somatic Cells Does Not Contribute to Female-Type Differentiation of Germ Cells
Next, we examined whether SMAD4 in somatic cells contributes to the female differentiation of germ cells. To delete Smad4 from somatic cells, we used WT1-CreERT2 mice. In this mouse line, CreERT2 was inserted into the Wt1 locus, and WT1 is expressed in genital ridges beginning at around E9.5 [42,43]. We injected TM into Smad4 flox/flox /WT1-CreERT2 mice at E10.5 and E11.5 to delete Smad4 in somatic cells. Smad4 expression was repressed, but this had no effect on Figla and Lhx8 expression; however, Sohlh1 expression was slightly upregulated at E14.5 (S2A and S2B Fig

Suppression of Smad4 and RA Signaling Induces Ectopic NANOS2 Expression in the Ovary
To further examine the function of SMAD4 in the early germ cell differentiation pathway, we examined pluripotency-related genes. In germ cells of normal ovaries, the expression of the pluripotency-related genes Sox2, octamer-binding transcription factor 4 (Oct4), Nanog, and undifferentiated embryonic cell transcription factor 1 (UTF1) is downregulated at E14.5 accompanied by the upregulation of genes involved in female differentiation, which indicates that the downregulation is associated with female differentiation (Fig 4) [44,45]. In Smad4deficient germ cells, however, these gene expressions were retained at high levels (Fig 4 and  S3D Fig), indicating that SMAD4 is involved in the suppression of pluripotency-related genes upon entering a female differentiation pathway.
Since Smad4-null germ cells showed severe defects in female differentiation, we speculated that the sexual fate of these PGCs may have been switched from female to male. To test this possibility, we investigated whether NANOS2, a male-specific marker, is expressed in Smad4null germ cells. However, we did not observe upregulation of NANOS2 in the Smad4-null germ cells (S4A Fig). We reasoned that the absence of Nanos2 gene expression may be due to RA signaling, which blocks Nanos2 activation in the testes [4,16], and predicted that simultaneous suppression of Smad4 and RA signaling may lead to Nanos2 induction in XX germ cells. To test this hypothesis, we collected ovaries of Smad4 (Stella) carrying a Cre reporter CAGfloxed-CAT-EGFP at E12.5 and cultured them with the RA receptor antagonist AGN193109. In this genetic background, Smad4-deficient cells could be detected through GFP reporter expression ( Fig 5A). We found that 70% (302/431; n = 5) of GFP-positive (Smad4-null) XX germ cells expressed NANOS2 only in the presence of RAR inhibitor (Fig 5B). Importantly, neither single-mutant ovaries (30.6 ± 16.5%, n = 4; Fig 6A). Conversely, SYCP3 expression disappeared in the NANOS2-expressing XX germ cells ( Fig 6A). Intriguingly, the DKO XX germ cells also showed other male-like features: they began DNMT3L expression, a downstream factor of NANOS2 involved in de novo DNA methylation in fetal testes (13.2 ± 6.9%, n = 4; Fig 6B) [11,13], and retained E-cadherin expression (Fig 6C), a germ cell marker whose expression decreases in XX germ cells from E15.5 [46], as seen in XY germ cells (S5D and S5E Fig). These factors were never observed in Stra8-null ovaries (S5A and S5B Fig). These data indicate that SMAD4 and STRA8 are essential factors responsible for female fate determination.
To further trace the fate of germ cells in DKO mice, we cultured E14.5 DKO ovaries for 4 d, because all of the pregnant mice carrying DKO embryos aborted at E16.5 after TM injection. We examined two critical male gonocyte markers-5-methylcytosine (mC), a marker of DNA methylation, and promyelocytic leukemia zinc finger (PLZF) [47]-and found that most DNMT3L-positive germ cells exhibited strong 5-mC signals and some E-cadherin-positive cells had PLZF expression (Fig 6D and 6E), while no PLZF and 5-mC positive signals were detected in control ovaries (S5C Fig). These results suggest that DKO germ cells acquired an ability to enter the male gonocyte differentiation pathway.

Female-to-Male Gene Expression Changes in DKO Germ Cells without Somatic Sex Reversal
To examine to what extent expression pattern changes in the DKO ovary, we performed microarray analyses. Total RNAs were extracted from control (Smad4 flox/+ , Smad4 flox/flox /Stra8 +/-) testes and ovaries and from Smad4 (Stella) and DKO ovaries at E14.5. To examine genes expressed in germ or somatic cells separately, we first extracted 386 and 360 genes, specifically expressed in control E13.5 male and female germ cells, as well as 680 and 636 genes, specifically expressed in control E13.5 male and female somatic cells, respectively, from our previous data (see Materials and Methods section; also see S1 and S2 Tables) [48]. The upregulation of malespecific genes in DKO ovaries showed a 1-day delay compared with normal male differentiation of germ cells in testes (for example, DNMT3L positive germ cells appeared from E14.5 in control testes but from E15.5 in DKO ovaries judging by comparing S6A Fig and Fig 6B); thus, we considered the developmental time point at E14.5 in the DKO ovary equivalent to that at E13.5 in control samples.
First, principal component analysis (PCA) was performed based on all genes, somatic-specific genes, and germ-specific genes, respectively (Fig 7A-7C). In all cases, PCA divided these samples into two major populations according to a putative sex-dependent axis (PC1 value = 38.6%, 72.3%, and 72.8%, respectively). Male and female control samples were located at opposite ends, as expected. When PCA was performed with all genes or somatic-specific genes, the DKO did not show a strong bias (Fig 7A and 7B). In contrast, when PCA was performed with genes specific for germ cells, the DKO samples were shifted toward the male side compared with Smad4-single KO samples on the putative sex-dependent axis (Fig 7C). We also observed the differences between DKO and male samples on the y-axis, which might be ascribed to the presence of somatic cells and the incomplete knockout efficiency. Nevertheless, it should be noted that PC1 value is as high as 72.8% in this case, suggesting that most of the gene expression profile in DKO germ cells shifts from female to male.
Next, we examined how many germ-cell-specific gene expressions were changed in DKO ovaries. Among 386 male-specific genes that showed a 1.5-fold stronger expression in E13.5 testes than in ovaries (Fig 7D, green dots with y-value greater than log 2 [1.5]), more than half of the cognate genes (52.1%, 201/386) were 1.5-fold upregulated in the DKO ovaries than in the control ovaries (Fig 7D, green points with x-value greater than log 2 [1.5]). Similarly, the  Scatter-plot analysis comparing germ-cell-specific gene expression changes between E13.5 Nanos3 +/− male and female gonads (data from [48]) and between E14.5 DKO ovaries and control ovaries. For each point (gene), the y-value represents the expression change comparing E13.5 testes with E13.5 ovaries and the x-value represents the expression change comparing E14.5 DKO ovaries and control ovaries. Thus, green and pink spots indicate 1.5-fold upregulated or downregulated expression in E13.5 testes versus ovaries (y-value greater than log 2 [1.5] for green spots and y-value less than -log 2 [1.5] for pink spots: the cut-off lines are indicated by horizontal lines). These spots are considered as genes that are specifically expressed in either male (green) or female (pink) germ cells, and their expression changes between DKO and control ovaries can be judged by the x-value. Red spots indicate specific gene examples. Genes located right of the rightmost dotted red line show more than a 1.5-fold increase in DKO, and genes located left of the leftmost dotted red line show less than a 1.5-fold decrease in DKO compared with control ovaries. Because genes that are specifically expressed in germ cells and somatic cells were extracted using microarray data of Nanos3 +/ovaries and testes, we used Nanos3 +/ovaries and testes as control. (E,F) Factor loading of indicated gene groups to principle component. PCA analysis was performed using E13.5 Nanos3 +/− male and female gonads (E) and E14.5 double-mutant and control ovaries (F). See text for details. doi:10.1371/journal.pbio.1002553.g007 expression levels of 71.1% (256/360) female-specific genes (Fig 7D, pink points whose x-value is less than -log 2 [1.5]) were 1.5-fold downregulated in the DKO ovaries. The results of microarray analysis were supported by RT-qPCR, in which eight out of eleven male-specific genes were upregulated in DKO ovaries (S6A Fig). Although Dnmt3l mRNA expression was not yet increased at E14.5, immunostaining data showed it was upregulated at E15.5 ( Fig 6B). These data indicate that not only the representative genes but also a substantial number of germ-cellspecific genes were shifted to male in DKO ovaries.
The results of PCA analysis were further supported by factor loading analysis (Fig 7E and  7F). Theoretically, the genes that give a more positive contribution to the classification will have a factor loading value (y-axis in Fig 7E and 7F) closer to 1 or -1. In the control, the factor loading value of most of the female-specific genes (both somatic and germ cell specific) was close to -1, and that of the male-specific genes was close to 1, indicating that somatic and germ cells contribute equally to the female versus male diversification of developing gonads (see Fig  7E [p = 0.168]). Notably, the comparison of the DKO and control ovaries showed that the factor loading value of most of the germ-cell-specific genes, but not the somatic-cell-specific genes, was closer to 1 or -1, indicating that germ-cell-specific genes contribute much more than somatic-cell-specific genes in distinguishing DKO and control ovaries (p < 0.0000001; Fig 7F). Indeed, somatic cells differentiated to pregranulosa cells marked by FOXL2, but did not differentiate to Sertoli cells, as revealed by the negative signal for SOX9, in DKO ovaries (S6A and S6B Fig, also see S4D Fig). These results suggest that the observed sexual fate change of XX PGCs in DKO ovaries occurs independently from global gene expression change in the somatic cells.

Discussion
In this study, we investigated the function of Smad4 signaling in the sexual differentiation of germ cells in ovaries and clarified the roles of Smad4 and RA signaling in germ cell fate determination (Fig 8). Three important outcomes in this study deserve particular attention.
First, we provided an insight into the critical question in reproductive biology: what factors determine female germ cell fate? RA-Stra8 signaling was once considered to determine female germ cell fate by inducing meiosis. However, a recent study showed a case where female germ cells can complete oogenesis even in the absence of Stra8, indicating that these two pathways are independent [22]. Therefore, uncovering the factors that determine female germ cell fate has been long awaited. Given that the genes involved in follicular formation were compromised in Smad4-deficient germ cells without affecting initiation of Stra8 expression and knocking out of both Smad4 and Stra8 resulted in up-regulation of male-specific germ cell genes, we propose that SMAD4 is a strong candidate for a sexual fate determinant in female germ cells. Lhx8 and Figla may be the potential targets of SMAD4, but we do not exclude that other factors may also be regulated by SMAD4. Future studies to identify the target gene(s) of SMAD4 along with functional studies would facilitate our understanding for the regulators of oogenesis. In addition, what signal SMAD4 mediates in female germ cells remains an open question. In this regard, a deduced candidate is BMP2. As BMP2 is expressed in ovarian somatic cells under the control of Wnt4 [29], we speculate that BMP2 initiates Smad4 signaling. However, we cannot exclude the possibility that other TGFβ ligands activate the pathway as well. Further analysis is required to elucidate the signaling pathway.
We conclude that Smad4 signaling in germ cells functions independent of somatic cells during the sexual fate determination of XX germ cells based on the fact that expression levels of oocyte-specific regulators in Smad4 flox/flox /WT1CreERT2 mutant ovaries were not significantly different from those of control ovaries at approximately E17.5. However, in the maturation stages of oocytes after birth, the deletion of Smad4 from somatic cells (granulosa cells) is accompanied by precocious luteinization and cumulus cell defects [49]. Therefore, the function of SMAD4 is highly stage-dependent in ovaries.
Second, we found that SMAD4 or its downstream signaling regulates meiotic progression during oogenesis. Although controversial [50,51], it is widely accepted that RA induces meiosis in germ cells. However, whether RA alone supports meiotic progression or requires other signals remains unknown. RA alone could not induce meiosis in vitro, which implies that factors other than RA are involved in meiotic progression [52]. The results of our study showed that Smad4-null germ cells expressed Stra8 and Rec8, but most showed meiotic defects before the pachytene stage accompanied by a failure to upregulate genes involved in meiosis (Fig 2 and  S3B Fig). Notably, most of these genes were also inactivated in Stra8 mutant ovaries (for example, Sycp1, Sycp3, Dmc1, Spo11, Mei1, and Hormad1) [53], which implies that SMAD4 and STRA8 regulate the same downstream factors to promote meiosis. Moreover, a Stra8-dependent pathway is reportedly responsible for the downregulation of Stra8 and Rec8 expression levels [53]. We speculated that this pathway is inactivated by the deletion of Smad4 because the expression level of Stra8 in mutant germ cells was higher than that in normal ovaries (Fig 2A  and S3B Fig). Overall, the results of the present study indicated that in addition to RA-dependent meiotic initiation, Smad4 signaling is required for meiotic progression, consistent with the study that successful induction of meiosis requires Activin A, BMP, and RA [52]. This finding is a critical clue for understanding the mechanisms involved in meiosis.
Third, we found profound properties of DKO germ cells. These germ cells became positive for male factors NANOS2 and DNML3L, and some of those cells showed 5-mC, indicating DNA methylation, all of which are normally observed only in male gonocytes. The transcriptome analyses of DKO embryos showed that more than 50% of male-type genes were up-regulated and more than 70% of female type genes were downregulated, indicating that male pathway is promoted when female germ cell development is suppressed by the lack of Stra8 and Smad4. However, as seen in PCA analyses, DKO germ cells were not equal to male germ cells and not all DKO germ cells expressed NANOS2. This may be due to the effects from In the ovary, two signaling pathways are involved in germ cell fate determination; one is an RA-Stra8 pathway that initiates meiosis and the other is a SMAD4 pathway that is activated via a currently unknown factor. In the absence of SMAD4, germ cells fail to proceed through normal meiosis and fail to make any oocytes. SMAD4 acts in germ cells cell-autonomously and regulates the fate of female germ cells because loss of SMAD4 and STRA8 leads to the upregulation of Nodal and NANOS2 expression as well as other male-specific genes. In this situation, somatic cell fate was unchanged and germ cells in the DKO may determine their fate in a cell-autonomous manner or another common factor (X) from female somatic cells is also required.
doi:10.1371/journal.pbio.1002553.g008 ovarian somatic cells; some factor(s) may act to suppress or promote the male pathway. Importantly, however, we showed that the female-to-male sexual fate change of XX germ cells occurred independent of the sexual switch of somatic cells as revealed by sustained FOXL2 expression in DKO ovaries. A similar phenotype was also observed in the ubiquitous Smad4-cKO (via Rosa-CreERT) with an RA inhibitor (S4C and S4D Fig). Owing to the lack of a Y chromosome in XX ovaries, the induction of male-like characteristics in DKO germ cells is not linked to any somatic factors, such as FGF9 and SOX9, which are controlled by SRY. Therefore, our results raise the unexpected possibility that male-specific factors in somatic cells are unnecessary for the male fate determination of germ cells even in the testes, although we do not exclude the possibility that other somatic factors (X) existing in both testes and ovaries are involved in the induction of maleness in germ cells (Fig 8). Consistent with this hypothesis, a recent study has shown that disruption of the germ-cell-specific factor FOXL3 in XX medaka (Oryzias latipes) leads to male differentiation even in ovaries [54]. In the case of our mouse model, however, we cannot expect to induce functional sperm in DKO ovaries, because the deletion of Stra8 leads to meiotic defects, and spermatogenesis requires an appropriate microenvironment supported by Sertoli cells that cannot be provided in mouse ovaries.
In summary, we analyzed the functions of SMAD4 in fetal female germ cells and found that SMAD4 is required for sexual differentiation of female germ cells. We further proved that SMAD4 and STRA8 were crucial for female sex determination, because germ cells lacking these two factors became male gonocyte-like cells, even if these germ cells were surrounded with female somatic cells. Our results provide foundational information for understanding the sex determination system in mammals, which may contribute to the goal of direct gamete induction in vitro.

Ethics
All mouse experiments were approved by the Animal Experimentation Committee at the National Institute of Genetics. Permission number for animal experiment is 27-13.

Mice
Generation of floxed Smad4 alleles and Stra8 knockout mice has been described previously [48,55]. Stella-MerCreMer was established previously [33]. Rosa-CreERT2 mice were purchased from Artemis Pharmaceuticals GmbH. WT1-CreERT2 mice were purchased from Jackson Laboratory. Gt(ROSA)26Sor tm1Sor /J and CAG-floxed-CAT-EGFP mouse lines were established previously [56,57]. All of these mice lines were kept in a mixed background. TM was diluted in sesame oil (Nacalai Tesque) at a concentration of 10 mg/ml, and 0.5 ml of the diluted TM was injected each time. To induce recombination in Stella-MerCreMer mice, TM was injected at E9.5 and E10.5. To induce recombination in Rosa-CreERT2 or WT1-CreERT2 mice, TM was injected at E10.5 and E11.5. ICR strain mice (Clea Japan) were used as control in some cases (S1A, S5D and S5E and S6B Figs).

Microarray and Data Analysis
RNA samples were prepared from E14.5 testes, ovaries, Smad4 mutant ovaries, and double mutant embryos (n = 3 for each case). For each hybridization assay, 200 ng of total RNA was labeled with Cy3 and hybridized to a Whole Mouse Genome Oligo Microarray (G4122F; Agilent) in accordance with the manufacturer's protocol using a Quick Amp Labeling Kit (Cat #51900424), One Color (Agilent), and Gene Expression Hybridization Kit (Agilent). Data have been deposited in Gene Expression Omnibus under accession number: GSE68773.
Extraction of genes specifically expressed in female somatic cells, male somatic cells, and male germ cells was performed using microarray data from [48] by R. To extract genes specifically expressed in male or female germ cells at E13.5, genes showing 1.5-fold changes in expression levels between Nanos3 +/testes and Nanos3 +/ovaries were defined as sex-specific genes. Then, genes showing 1.5-fold changes in expression levels between Nanos3 +/testes and Nanos3 -/testes and between Nanos3 +/ovaries and Nanos3 -/ovaries were defined as germ cell-specific genes, because Nanos3 -/gonads (both ovaries and testes) lack germ cells at E13.5. Genes specifically expressed in male or female germ cells were obtained by the intersection of genes in the two lists. Similarly, genes specifically expressed in male or female somatic cells (if genes are not germ-cell-specific) at E13.5 were also isolated. PCA was analyzed by R using genes in the above lists.

Organ Culture
Gonads were cultured in 24-well culture plates with DMEM containing 10% horse serum at 37°C on 5-μm nucleopore filters [58,59]. Some gonads were cultured in medium For the experiments shown in Fig 6D and 6E, tamoxifen was injected at E9.5 and E10.5, and ovaries obtained at E14.5 were further cultured for 4 d.

Reverse Transcription Real-Time Quantitative Polymerase Chain Reaction (RT-qPCR)
Total RNA was prepared from the fetal gonads of mutant embryos and the gonads of ICR strain embryos at each embryonic stage using RNeasy Mini Kits (Qiagen). Total RNA was then 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 cycler dice real time system (Takara). The primer pairs used for PCR amplification are listed in S3 Table. Histological Analysis Histological analysis was performed as previous reported [5,8]. Briefly, gonads were fixed in 4% paraformaldehyde, embedded in OCT compound (Tissue-Tek; Sakura), and sectioned (6 μm) using a cryostat. After preincubation with 3% skim milk powder in PBST for 30 min, the sections were stained with primary antibodies: anti-TRA98 (1:10000; a gift from Y. Nishi (1:200; Cell Signaling Technology). The sections were then incubated with donkey anti-rabbit/ rat/goat/mouse or anti-chick IgG secondary antibodies conjugated with either Alexa 488, Alexa 594, or cy5 (1:400; Invitrogen). Primary antibodies were diluted in 3% skim milk powder in PBST. Secondary antibodies were diluted in PBST. Slides were mounted for observation under a scanning confocal microscope (Olympus FV1200 IX83). The method for LacZ staining (S1B Fig) was reported previously [61].

Statistical Analysis
Statistical analysis was performed as previous reported [5,8]. For quantitative analyses between two different samples, significance was assessed by Student's t test. For quantitative analyses among multiple samples, significance was assessed using one-way ANOVA followed by Tukey's post-hoc tests for selected pairs of genotypes.
Supporting Information S1 Data. Underlying data used for quantitative analysis.  4). The expression levels of the indicated genes were normalized to that of mouse vasa homolog (Mvh). Significance was assessed using Student's t test for one pair of genotypes and one-way ANOVA followed by Tukey's post-hoc tests for selected pairs of genotypes. Error bars indicate SD. Underlying data is available in S1 Data. Tamoxifen was injected at E9.5 and E10.5, and gonads were recovered at E14.5. The expression levels of the indicated genes were normalized to that of mouse vasa homolog (Mvh). Significance was assessed using Student's t test. Error bars indicate SD. Underlying data is available in S1 Data. (B) Representative image of E15.5 DKO ovarian tissue section stained for FOXL2 and SOX9, and wild-type testis section stained for SOX9 and TRA98. Scale bars: 50 μm. (TIF) S1 Table. Fold change of male germ-cell-specific genes in DKO ovaries compared with control ovaries.