Germ cells in the teleost fish medaka have an inherent feminizing effect

Germ cells give rise to eggs or sperm. However, recent analyses in medaka (Oryzias latipes) showed that germ cells are also important for feminization of gonads, although this novel role of germ cells has not been characterized in detail. Here, we show that the feminizing effect is inherent to germ cells and is not affected by gametogenic stages or the sexual fate of germ cells. Three medaka mutants were generated to demonstrate this effect: figlα mutants, in which follicle formation is disrupted; meioC mutants, in which germ cells are unable to commit to gametogenesis and meiosis; and dazl mutants, in which germ cells do not develop into gonocytes. All these different stages of germ cells in XX mutants have an ability to feminize the gonads, resulting in the formation of gonads with ovarian structures. In addition to normal ovarian development, we also suggest that the increased number of gonocytes is sufficient for male to female sex reversal in XY medaka. These results may genetically demonstrate that the mechanism underlying the feminizing effect of germ cells is activated before the sexual fate decision of germ cells and meiosis, probably by the time of gonocyte formation in medaka. Author summary Germ cells are the only cells that can transfer genetic materials to the next generation via the sperm or egg. However, recent analyses in teleosts revealed another essential role of germ cells: feminizing the gonads. In our study, medaka mutants in which gametogenesis was blocked at specific stages provides the novel view that the feminizing effect of germ cells occurs in parallel with other reproductive elements, such as meiosis, the sexual fate decision of germ cells, and gametogenesis. Germ cells in medaka may have a potential to feminize gonads at the moment they have developed.


Introduction
The sex of organisms is determined either by genetic factors and/or by environmental factors. Many organisms with genetic sex determination systems have sex determination genes, but these genes differ from species to species [1]. The environmental factors that determine sex include temperature, density, pH and social status [2]. Although initial triggers for sex determination vary, there are some common features in the subsequent sex differentiation process in gonads. One feature is observed in the sexual modes of germ cell proliferation. In most vertebrates, after primordial germ cells (PGCs) reach the gonads, PGCs develop to become gametogenesis-competent germ cells [3]. Whereas the germ cells in females begin to differentiate to undergo oogenesis, the germ cells in males are in a mitotically quiescent state for a while before the initiation of spermatogenesis [4,5].
The teleost fish, medaka (Oryzias latipes), is one of the organisms in which sex is determined genetically [6,7]. The sex determination gene on the medaka Y chromosome, DMY/ dmrt1bY, is initially expressed in the gonadal somatic cells when the gonads form at stage 33 [4 day-post-fertilization (dpf)] [8]. At this stage, the sexual difference in the number and differentiation of germ cells is not observed. The first appearance of sexual differentiation in the gonad can be recognized by the time of hatching [0 day-post-hatching (dph)] [9]. Germ cells in XX gonads initiate oogenesis to enter meiosis around the time of hatching, or several days after hatching, while proliferation of germ cells and initiation of gametogenesis is suppressed in XY germ cells during the larval stage. Spermatogenesis is initiated approximately one month after hatching [10].
Previously, we found that the occurrence of more germ cells in female medaka gonads is not just a result of sex differentiation, but is essential for female differentiation of gonads in medaka. When germ cells are ablated in medaka, XX fish show female-to-male sex reversal, while XY fish with overproliferated germ cells, which is usually associated with oogenesis, exhibit male-to-female sex reversal [11][12][13]. In support of this notion, studies have reported that high temperatures also decrease the number of germ cells, which often accompanies female-to-male sex reversal [14,15]. Therefore, although the sex of medaka is genetically determined by DMY/dmrt1bY, the germ cells seem to play a critical role in sex determination, especially in the feminization of gonads. This finding implies that, in addition to the most wellknown role of germ cells to become gametes, germ cells have a unique and surprising potential to change their surrounding environment. However, the molecular basis of, and the stage of gametogenesis critical for feminization remain unknown.
It has been demonstrated that germ cells in the gonads have to acquire competence for gametogenesis before and/or at the time of entering the gonadal area [3,16]. In medaka, gametogenesis-competent cells in the gonad can be categorized into two types, type I and type II, which are distinguished by the modes of cell division [9]. Type I cells are characterized by stem-type self-renewal division and are histologically defined as gonocytes in the developing gonads. They are completely enclosed by supporting cells and are present in isolation. Certain populations of type I germ cells commit to gametogenesis and undergo type II division, in which germ cells connected by the intercellular bridge synchronously divide to form a cyst. After undergoing type II division, germ cells enter meiosis and initiate folliculogenesis or spermatogenesis [10].
In this study, medaka mutants with the disrupted gene, figlα, meioC or dazl, were used to determine a critical gametogenetic stage(s) for the feminization. In addition, gsdf medaka mutant was used to study the requirements of germ cells in the process of male-to-female sex reversal. Collectively, we suggest that germ cells in medaka are predisposed to develop the ability to feminize the gonad independent of somatic sex. The feminizing effect of germ cells seems to be underlain by a mechanism that is distinct from the mechanisms of commitment to gametogenenesis, entering meiosis and sexual fate determination.

Ovarian follicles are not required for ovary fate and female development in medaka
After cystic division and the start of meiosis, each oocyte at diplotene stage is surrounded by somatic cells (granulosa cells and theca cells) to form follicles in female gonads. We initially assumed that follicles are essential for the feminization of gonads in medaka for three reasons. First, the somatic cells of follicles consist of granulosa cells expressing foxl2 and theca cells expressing aromatase [17,18]. Both genes are involved in feminization. Second, in male-tofemale sex reversal due to overproliferation of germ cells in medaka, follicle formation is observed during early gonadal development [12,13,19]. Third, in zebrafish, follicles are considered to be essential for the feminization of gonads [20,21]. Therefore, we first generated mutant medaka in which follicle formation was disrupted.
Figlα (Function in germ line alpha) is a basic helix-loop-helix transcription factor that is essential for follicle development in mammals [22,23] and is preserved in other vertebrates including teleosts [24,25]. In medaka, figlα transcripts were detected in female germ cells (S1A and S1C Fig) but not in male germ cells at 10 days post hatching (dph) and adult stage (S1B and S1D Fig). Detailed analysis of figlα expression was performed using figlα-EGFP reporter transgenic medaka, in which EGFP was expressed under control of figlα regulatory elements (S1E Fig). Figlα-EGFP was initially detected in germ cells at the late zygotene stage, and strong figlα-EGFP signals were detected in diplotene oocytes (S1F Fig), suggesting that figlα functions in germ cells after entering meiosis, possibly for the formation of follicles. To determine whether deletion of the figlα gene disrupts follicle formation in medaka, we generated a medaka figlα mutant using transcription activator-like effector nuclease (TALEN). Two TALENs targeting the figlα gene were designed, which resulted in two different alleles (ex1-Δ16 and ex2-Δ4+18) with frameshift mutations causing premature truncations upstream of and at the basic helix-loop-helix domain, respectively (S1G and S1H Fig). At the hatching stage, the number of germ cells was not significantly different between XX wild-type and XX figlα homozygous (-/-) gonads (S2A, S2B and S2E Fig), the phenotype of which is consistent with the absence of figlα expression in mitotic germ cells at this stage (germ cells in many XX medaka have initiated oogenesis, but do not often reach the pachytene stage). Previously, we identified foxl3 as a gene involved in the sexual fate decision of germ cells by suppressing the initiation of spermatogenesis in medaka [26]. The expression of FOXL3 is initially observed in both female and male mitotic germ cells (female type I and type II germ cells and male type I germ cells). However, in the male mitotic germ cells, the expression of FOXL3 decreases dramatically by the time of hatching. When foxl3 function is lost in germ cells of female gonads, spermatogenesis takes place instead of oogenesis by 30dph (after this stage, oogenesis is also initiated by unknown mechanisms). Hence, FOXL3 can be a good marker representing the female sexual fate of germ cells during early gonadal development (from hatching stage to 30dph). In figlα-/-mutant gonads, the expression of FOXL3 was found to be normal at hatching stage, which suggests that FOXL3 expression is independently regulated by figlα activity. (S2A, S2B and S2F Fig). At 10 dph, XX wild-type (figlα+/+) fish had plenty of oocytes at the diplotene stage and started to form follicles ( Fig 1A). However, XX figlα homozygous (-/-) mutants had cystic germ cells that had arrested between the zygotene and pachytene stages and lacked isolated oocytes at diplotene stage ( Fig 1B). The follicle-marker genes, gdf9 and bmp15, were expressed in the oocytes of follicles found in the control ovaries but were not expressed in figlα-/-mutant ovaries (Fig 1C-1H). These observations indicate that XX figlα-/mutants did not form follicles.
As in wild-type males, the germ cell proliferation in XY figlα-/-mutants was suppressed at hatching (S2C- S2E Fig). XY figlα-/-mutants developed normally as fertile males with testis in  Next, we examined whether XX figlα-/-mutants developed as females. During gonadal development, XX figlα-/-mutants expressed the female marker genes aromatase and foxl2, whereas the male marker, dmrt1, was hardly detected (S4 Fig). In the adult stage, female secondary sex characteristics were present, and XX figlα-/-mutants had typical ovarian-like gonads with an ovarian cavity, but no follicle formation was observed (Fig 1I-1N). The formation of an ovarian cavity depends on the production of estradiol, which is synthesized by aromatase [27]. Accordingly, the adult gonads expressed the aromatase as well as foxl2 (Fig 1O). The expression of these female genes indicates that female somatic cells arise and present in the stromal region of the mutant but cannot contribute to the follicle formation. Taken together, these results demonstrate that normal female development occurs without forming follicles.
In male-to-female sex-reversed medaka, overproliferation of germ cells is associated with follicle formation during gonadal development [12,13,19]. To investigate whether follicles are required for male-to-female sex reversal in XY fish, we generated gsdf (gonadal soma derived factor) mutants (S5 Fig). Gsdf belongs to a TGF-β family that is conserved among teleosts and is enriched in the gonadal somatic cells of males [28,29]. Although DMY/dmrt1bY is expressed, XY gsdf loss-of-function (gsdf-/-) mutants have an increased number of germ cells comparable to that of XX fish at hatching stage, and more than half show male-to-female sex reversal [19,30]. Consistent with previous studies, our newly generated XY gsdf-/-mutants (S5A and S5B Fig) had gonads similar to those of XX fish, and follicle formation occurred during gonadal development (S5C- S5E Fig). XY gsdf-/-gonads expressed the female marker genes foxl2 and aromatase at 10 dph (Fig 2A and 2B), whereas germ cell-deficient XY gsdf-/- The inherent feminizing effect by germ cells in medaka fish gonads did not express these female markers (Fig 2C and 2D). In the adult stage, the male-tofemale sex-reversal was suppressed in XY gsdf-/-medaka mutants that had lacked germ cells during the embryonic stage (Fig 2E-2H). These results indicate that the male-to-female sex reversal in XY gsdf-/-mutants is mediated by germ cells but not by somatic cell-autonomous feminization.
To determine whether feminization of gonads in gsdf mutants requires follicles, we generated gsdf-/-; figlα-/-double mutants. During gonadal development, XY gsdf-/-; figlα-/-double mutants had larger gonads with many germ cells compared with those in XY gsdf+/-; figlα+/medaka (S6A, S6C, S6D and S6F Fig). Interestingly, the XY double mutants did not form any follicles but expressed foxl2 and aromatase, suggesting that follicles are not required for activation of the feminizing genes (S6 Fig). In the adult stage, nine out of ten XY gsdf-/-; figlα-/double mutants exhibited female secondary sex characteristics and had gonads with an ovarian structure and female gene expression (Fig 3). These results may suggest that follicles are not required for male-to-female sex reversal.

Germ cells prior to gametogenesis are sufficient for feminization
We then examined whether germ cells prior to meiosis have a feminizing effect on gonadal somatic cells. Recently, MeioC was identified as a gene essential for progression of meiosis [31,32]. The medaka homologue of meioC was predicted from the genomic database (Ensembl ID: ENSORLT0000005275, NCBI ID: XM_011477965.1). It showed conserved synteny among vertebrates (S7A Fig), and it was expressed in the germ cells of both testes and ovaries (S7B and S7C Fig). A medaka mutant was generated using CRISPR/Cas9, resulting in three different alleles with frameshift mutations causing premature truncations upstream of the C-terminal conserved region of the MEIOC protein (S7D and S7E Fig). In 5 dph gonads, XX meioC +/+ controls had cystic germ cells and meiotic germ cells (Fig 4A), whereas XX meioC-/mutants had only type I stem-like germ cells in isolation and surrounded by somatic cells. Type II, meiotic germ cells and follicles were not observed ( Fig 4B). This suggested that meioC is required for the transition of germ cells from type I to type II in medaka. Interestingly, FOXL3 protein was expressed in the XX meioC-/-mutants. These results indicate that MEIOC is required for the initiation of gametogenesis, but not for the expression of FOXL3, suggesting that the sexual fate decision of germ cells likely occurs in XX meioC-/-gonads. In the adult stage, XX meioC-/-mutants showed female secondary characteristics and had gonads with ovarian structures (Fig 4D, 4G and 4J). Furthermore, four of nine XY meioC-/-; gsdf-/-double mutants also showed female secondary characteristics and had ovary-like gonads (Fig 4E, 4H and 4K) with the expression of foxl2 and aromatase ( Fig 4L). Therefore, type II and meiotic germ cells are not required for normal female development in XX fish and male-to-female sex reversal in XY fish.

PGC-like germ cells have a feminizing effect on somatic cells
Finally, we investigated whether primordial germ cells (PGCs) have the ability to feminize gonads. In mice, Dazl is required for germ cells to become gametogenesis-competent, and germ cells in the Dazl mutant remain in an undifferentiated state similar to that of PGCs [3]. Therefore, we generated medaka dazl mutants using the CRISPR/Cas9 system and obtained two different alleles (ex2-Δ34 and ex3-Δ2+7) that caused premature truncations upstream of or at the RNA recognition motif of DAZL (S8 Fig). Consistent with studies in mice, dazl-/germ cells at hatching stage were morphologically similar to PGCs at st.30, which were characterized by smaller cells and nucleoli than those in dazl+/+ type I germ cells at the same stage (Fig 5A-5C). The foxl3 and meioC transcripts were not detected in dazl-/-germ cells (Fig 5D-   , and the number of germ cells was decreased in both XX and XY mutant gonads during gonadal development (Fig 5H). In the adult stage, vasa transcripts, a typical marker of germ cells, were not detected in dazl-/-gonads (Fig 5K). These results suggest that germ cells in dazl-/-mutants lose both competence for gametogenesis and the response to sexual cues from gonadal somatic cells. These germ cells seemed to be no longer maintained in the mutants. In fact, germ cells were absent in the adult stage. Interestingly, not all XX dazl-/-mutants developed as males. More than half of the XX mutants (56% in ex2-Δ34 and 67% in ex3-Δ2+7) showed female secondary characteristics (Fig 5I and 5J). The gonads from the phenotypically female XX mutants expressed female genes (Fig 5K). Furthermore, detailed histological analysis revealed that the dazl-/-gonads from females had ovarian cavities lined by ciliated epithelial cells, germinal epithelia, and stromal compartments, which are typical structures of ovaries (Fig 6). Therefore, PGC-like germ cells can support female gonadal development. The inherent feminizing effect by germ cells in medaka fish

Discussion
Germ cells are the only cells that develop into eggs or sperm and produce the next generation. This role has long been recognized as the only function of germ cells. However, recent analyses in teleost fish unveiled another essential role, namely sex determination [11-13, 33, 34]. During normal development in medaka, the first manifestation of sexual dimorphism is the difference in the number of germ cells [9]. However, this difference is not simply a consequence of the direction of sex determination, but rather it is essential for sex determination of the gonad: a greater number of germ cells during gonadal development in XX medaka enables to canalize feminization of gonads, while a lower number of germ cells in XY medaka is not sufficient for feminization, leading to masculinization. Interestingly, manipulating the number of germ cells causes sex-reversal. XY medaka with a greater number of germ cells develop into females, whereas a lower number of germ cells cause XX medaka to differentiate into males [11][12][13]35]. A similar relationship between germ cell number during gonadal development and sexual phenotype in the adult stage has also been reported in zebrafish [36]. In this study, we generated four different medaka mutants to demonstrate that the feminizing effect of germ cells is not a result of the progression of gametogenesis or a sexual fate decision of germ cells.
The dazl-/-mutants provided insight into the feminizing effect. Approximately half of the XX mutants showed a female phenotype with a normal ovarian structure devoid of germ cells. This phenotype is in sharp contrast with previously reported germ cell-deficient XX medaka, which displays female to male sex-reversal [11]. The difference between the two phenotypes may be explained by the existence of a limited time window during which the feminizing effect of germ cells acts on somatic cells to induce female differentiation. The dazl-/-mutants retained a certain number of germ cells during the period in which medaka with female-tomale reversal have a fewer number of germ cells. After this window, the gonad can continue to develop into an ovary even if germ cells are gradually lost in the dazl mutants. The inherent feminizing effect by germ cells in medaka fish The dazl mutants also suggest the timing at which germ cells exert a feminizing effect. Our analyses indicate that the germ cells in the dazl mutants are likely gametogenesis-incompetent PGCs because they possess PGC-like morphology in both appearance and size and do not express the meiotic gene (meioC) or the gene involved in the sexual fate decision of germ cells (foxl3). This is consistent with the function of dazl in mammals [3]. However, gametogenesisincompetent germ cells retain the ability to feminize the gonad. This observation suggests that germ cells acquire this property by the time PGCs develop into gonocytes.
The point mentioned above provides insight into the molecular mechanism of gametogenesis. Our results suggest that several mechanisms operate in parallel in developing gametes. Although feminizing effect of germ cells is critical for gonadal sex determination, the mechanism underlying this effect is likely distinct from the sexual fate of germ cells because the feminizing activity is apparent even if foxl3 is not expressed in the dazl mutant. This notion is supported by our previous report that the XX foxl3 mutant, in which germ cells are committed to spermatogenesis, develops sperm-producing ovaries [26].
In the absence of meioC function, germ cells did not undergo cystic division (type II division), demonstrating that meioC is critical for the process of transition from type I stem-type germ cells to gametogenesis-committed type II germ cells. The function of the medaka meioC is in contrast with that of the mammalian MeioC, which is required for the proper progression of meiotic prophase I, but not for the entry of meiosis [31,32]. Medaka meioC may involve more aspects of gametogenesis than mammalian MeioC. Both XX meioC-/-mutants and XY gsdf-/-; meioC-/-double mutants possess gonads with typical ovarian structures, indicating that the feminizing effect is manifested before germ cells exit from the status of stem cells/ gonocytes and enter the gametogenetic program.
The feminizing effects of germ cells have also been reported in zebrafish [20,32,33,36]. Recently, an oocyte-derived factor, BMP15, was suggested to be required for the maintenance of feminization in zebrafish [21]. On the other hand, in mice and some species of teleost fish such as loach, goldfish and salmon, germ cells are not required for the feminization of gonads [37][38][39][40][41]. These studies provide evidence for the fact that somatic cells can establish and/or maintain feminization without germ cells, but do not deny that feminizing effect of germ cells is absent in these species. In support of this, it is interesting to note that meiotic germ cells in females have antagonizing effects against masculinization in mice [42].
One common feature of germ cells during gonadal sex differentiation in vertebrates is that female germ cells proliferate and initiate oogenesis in an early stage, while male germ cells undergo a mitotically quiescent state before the initiation of spermatogenesis [4,5]. Why do male germ cells need to be quiescent? Why do they not initiate spermatogenesis with the same timing of oogenesis? Our analyses using medaka collectively suggest that germ cells acquire the feminizing effect before committing to gametogenesis, entering meiosis and the sexual fate decision of germ cells. Therefore, germ cells in medaka may be developmentally designed to make the gonad female. One possible answer for the biological meaning of the quiescent state of male germ cells may be to prevent the gonad from being feminized until masculinization of somatic cells is established.

Ethics statement
All experiments were done according to Regulations on Animal Experiments in Nagoya University. The experimental plan using the small teleost fish is approved by the Nagoya University official ethics committee (Approval Number 11).

TALEN-induced mutagenesis
The TALEN target sites for figlα and gsdf were searched using the TALEN Targeter program (https://tale-nt.cac.cornell.edu/node/add/talen) with the following parameters: spacer length of 15-18 bp, repeat array of 16-18 bp, and upstream base of T only. TALEN assembly was performed following a modified version of the original protocol [45]. TALEN plasmids were linearized by NotI digestion and used as templates for in vitro RNA synthesis with the mMES-SAGE mMACHINE T7 transcription kit (Thermo Fisher). TALEN mRNAs (250 ng/μl left and right) were injected into one-or two-cell stage embryos. The F0 founders were crossed with sox9b-DsRed/olvas-EGFP transgenic and non-transgenic medaka. Each mutant allele was identified in F1 adult fish using the primer sets shown in S1 Table and sequence analysis.

CRISPR/Cas9-induced mutagenesis
CRISPR/Cas9 target sites in meioC and dazl were searched using CRISPR/Cas9 target online predictor (CCTop, http://crispr.cos.uni-heidelberg.de/) [46]. Selected targets sites are shown in S7 and S8 Figs and S1 Table. A designed pair of oligonucleotides (50 μM each) was annealed in TE buffer by heating to 95˚C and cooling slowly to 4˚C using a thermal cycler (Biometra). The annealed oligos were ligated into BsaI-HF (New England Biolabs) digested DR274 vectors (Addgene plasmid #42250) [47]. The resulting vector was used as a template to amplify fragments containing T7 and gRNA sequences by PCR followed by gel purification. The PCR primer set is shown in S1 Table. Then, the PCR fragments were used as a template for in vitro transcription with the MEGAscript T7 kit (Thermo Fisher) according to the manufacturer's instructions. The synthesized sgRNAs were purified by ammonium acetate precipitation.
For synthesis of Cas9 mRNA, the pCS2+hSpCas9 vector [48], a gift from Dr. Kinoshita in Kyoto University, was linearized by NotI digestion, followed by in vitro transcription using the mMessage mMachine SP6 kit (Thermo Fisher) according to the manufacturer's instructions. The synthesized RNA was purified with the RNeasy Mini kit (Qiagen). Cas9 mRNA (50 ng/μl) and sgRNA (25 ng/μl) were injected into one-cell stage embryos. The F0 founders were crossed with sox9b-DsRed/olvas-EGFP transgenic and non-transgenic medaka. Each mutant allele was identified in F1 adult fish using the primer sets shown in S1 Table and sequence analysis.

Genotyping
For genotyping of each mutant allele, amplicons including the mutation sites were amplified by SYBR-KOD (TOYOBO), followed by melting curve analysis using StepOnePlus (Applied Biosystems). The primer sets for each mutant allele are shown in S1 Table.

Generation of figlα-EGFP transgenic medaka
A bacterial artificial chromosome (BAC) transgenic method that uses homologous recombination was employed to generate a figlα-EGFP reporter construct as previously described [52]. Specifically, the targeting DNA fragment for recombination was prepared to include the figlα-3 0 UTR downstream of the EGFP open reading frame. After homologous recombination, this fragment was inserted immediately downstream of the translation initiation site of the figlα gene in fosmid clone GOLWFno462_j06. Microinjection of fosmid clone containing EGFP was performed as previously described [52].