Analysis of a novel gene, Sdgc, reveals sex chromosome-dependent differences of medaka germ cells prior to gonad formation

In vertebrates, the master sex-determination genes identified so far are expressed in the gonadal somatic cells (supporting cells) (Hattori et al., 2012; Kamiya et al., 2012; Koopman et al., 1991; Matsuda et al., 2002; Smith et al., 2009; Yoshimoto et al., 2010). Therefore, it is generally accepted that sex determination first occurs in the gonadal somatic cells during development. Following this step, under the influence of the somatic cells, the sex of germ cells is determined, and these cells are then fated to develop into either sperm or eggs.

In mice, when germ cells are isolated from XY gonads at 11.5 days post-coitum (dpc) and reaggregated with XX somatic cells from urogenital ridges, the XY germ cells behave like XX female germ cells and enter meiosis. By contrast, germ cells from XY gonads at 12.5 dpc mitotically arrest as prospermatogonia in response to the same treatment (Adams and McLaren, 2002; McLaren, 2000). This observation indicates that by 12.5 dpc, germline sex is largely determined by the action of the sex-determination gene Sry in the gonadal somatic cells.

Somatic cells also exert a strong influence on germ cell sex determination in teleost fish, as demonstrated by the reciprocal transplantation of testicular and ovarian germ cells. In rainbow trout and zebrafish, isolated oogonia and spermatogonia from mature gonads can recolonize the undifferentiated gonads of embryos and produce either sperm or eggs, depending on the sex of the surrounding somatic cells (Okutsu et al., 2006; Wong et al., 2011; Yoshizaki et al., 2010).

In medaka (Oryzias latipes), expression of the sex-determination gene on the Y chromosome, DMY/dmrt1bY (Matsuda et al., 2002; Nanda et al., 2002), has so far been reported to begin in gonadal somatic cells immediately after the gonadal primordium forms at stage 33 (st.33); it suppresses the shift of germ-cell proliferation modes from a self-renewal type of division (type I division) towards a gametogenesis-committed cystic type of division (type II division) at st.35. The shift from type I to type II division is crucial for ovary formation (Nakamura et al., 2012a,,b; Saito et al., 2007; Tanaka et al., 2008).

Taken together, the evidence enumerated above indicates that, in vertebrates, the gonadal somatic cells control the sex of germ cells and determine whether they develop into spermatids or oocytes. In this study, we found that medaka germ cells exhibit sexually different characters even before the formation of the gonadal primordium. We found that the differences at this early stage are controlled in a germ cell-autonomous manner by the number of Y chromosomes, but not by the expression of the sex-determination gene protein DMY/dmrt1bY. Thus, our findings reveal a novel mechanism by which sexually different characters arise at the cellular level.

To study gene expression in germ cells and gonadal somatic cells during gonadal sex differentiation, we used a sox9b-DsRed/olvas-EGFP transgenic medaka (Nakamura et al., 2010), and isolated EGFP-positive germ cells and DsRed-positive supporting cells from XX and XY embryos by fluorescence-activated cell sorting (FACS) at st.33 and st.35. Microarray expression profiling revealed that several transcripts exhibited sexually different expression levels in germ cells by st.33. Among them, a novel transcript, olte54h09 (NCBI accession: FS527139; named Sdgc after its function, sex chromosome-dependent differential expression in germ cells), was highly enriched in XY germ cells relative to XX germ cells (supplementary material Fig. S1A,B). Quantitative PCR (qPCR) analysis of Sdgc and other gonadal markers confirmed that Sdgc was already upregulated in XY embryos by st.33 (supplementary material Fig. S1C). This observation raised the possibility that germ cells exhibit sexually different gene expression even before the expression of DMY/dmrt1bY begins in gonadal somatic cells (supplementary material Fig. S1D).

To confirm this possibility, we performed qPCR and in situ hybridization for Sdgc in st.30 embryos. At st.30, the gonad is not yet formed, and the germ cells are located lateral to the forming hindgut; from this position, they subsequently move towards the prospective gonadal area on the dorsal side (Nakamura et al., 2006). Sdgc expression was detected as clusters on the lateral sides, and was stronger in XY embryos than in XX embryos (Fig. 1A-D). The expression levels of Sdgc were more than twofold higher in XY embryos than in XX embryos (Fig. 1I). When endogenous germ cells were removed by inhibiting germ-cell migration using nanos3/cxcr4-morpholinos (Kurokawa et al., 2006,, 2007), the Sdgc signal totally disappeared (Fig. 1E-I). Together with microarray data showing higher enrichment in isolated germ cells, these observations indicate that Sdgc is expressed in germ cells.

The Sdgc transcript encodes a putative protein of 142 amino acids (supplementary material Fig. S2). Eleven repeats of QGPPAQ(E)GR constitute the majority of the protein, and no known domains were found in the sequence. The 5′UTR was found on three different scaffolds, but none of those scaffolds had yet been mapped to medaka linkage groups (LG). When we performed qPCR analysis of Sdgc in a different strain, Hd-rR-II1, expression was also upregulated in XY embryos (supplementary material Fig. S3). Therefore, upregulation of Sdgc in XY germ cells is not specific to the OKcab strain.

We then examined Sdgc expression in adult testes and ovaries. Sdgc was highly enriched in testes when compared with ovaries (supplementary material Fig. S4A). In testes, the signal was more abundant in type B spermatogonia than in type A spermatogonia, but was absent in spermatocytes (supplementary material Fig. S4B,C). In ovaries, the transcripts were present only at extremely low levels in early stage oocytes (supplementary material Fig. S4D). No expression was detected in oogonia (supplementary material Fig. S4E), consistent with the sexually different expression in germ cells at the early embryonic stages.

Next, we examined the expression levels of Sdgc using medaka with different Y-chromosome constitution and found that Sdgc expression levels increased with the copy number of the Y chromosome (Fig. 1J). We then asked whether the upregulation of Sdgc was due to the presence of a Y chromosome in germ cells or in somatic cells. To address this issue, we generated chimeric medaka by transplanting XY germ cells into XX hosts and vice versa (Nakamura et al., 2012b), using sox9b-DsRed/olvas-EGFP transgenic line as hosts and non-transgenic strains as donors. Prior to manipulation, the host germ cells with EGFP fluorescence were ablated by nanos3/cxcr4-morpholino injection. Sdgc expression levels in XY hosts with XX germ cells resembled that of XX hosts with XX germ cells. By contrast, the expression of Sdgc in both XX and XY hosts with XY germ cells was more than twofold higher than in hosts with XX germ cells (Fig. 1K). The sex of the host did not correlate with any statistically significant difference in the Sdgc expression level in germ cells. These results clearly indicate that upregulation of Sdgc depends on the presence of a Y chromosome in germ cells, but is independent of the genetic sex of the surrounding somatic cells, suggesting that sexually different expression occurs as a germ cell-autonomous event.

The only differences identified to date in the two medaka sex chromosomes reside at the sex-determination locus, leading us to speculate that DMY/dmrt1bY may upregulate the Sdgc gene in XY germ cells in a germ cell-autonomous manner. Interestingly, DMY/dmrt1bY transcripts were detected in XY germ cells at both st.33 and st.35 by microarray (supplementary material Fig. S1D). The germ-cell expression of DMY/dmrt1bY was further confirmed by in situ hybridization in st.30 embryos. The DMY/dmrt1bY signal was also detected, only in XY embryos, as single clusters on both lateral sides of the forming hindgut, where it was colocalized with expression of OLVAS, a marker of germ cells (Fig. 2A-C). When endogenous germ cells were removed by morpholino treatment, the DMY/dmrt1bY signal disappeared (Fig. 2D). To further investigate DMY/dmrt1bY expression in germ cells, we generated a DMY/dmrt1bY-EGFP reporter transgenic medaka, in which DMY/dmrt1bY expression could be visualized by EGFP driven by regulatory elements of DMY/dmrt1bY. DMY/dmrt1bY reporter expression in germ cells was first detected as early as st.28 and persisted until around st.31 (supplementary material Fig. S5A-C). Together, these data indicate that DMY/dmrt1bY is expressed in germ cells before formation of the gonadal primordium.

To determine whether Sdgc is directly regulated by DMY/dmrt1bY protein, we knocked down DMY/dmrt1bY using synthetic antisense oligonucleotides (gripNA). Specifically, we used a dmy-gripNA described previously (Paul-Prasanth et al., 2006), for which the specificity had already been confirmed by in vitro translational suppression and in vivo analysis. We also confirmed that this gripNA effectively decreased green fluorescence in embryos of DMY/dmrt1bY-EGFP reporter transgenic medaka (Fig. 2E,F). Although the reporter fluorescence was dramatically reduced, Sdgc expression levels did not change in gripNA-injected embryos relative to normal embryos at st.30 (Fig. 2G). Overexpression of DMY/dmrt1bY also had no effect on the expression level of Sdgc (supplementary material Fig. S6A,B). These results demonstrate that upregulation of Sdgc is not due to a direct effect of DMY/dmrt1bY protein expression, but rather is crucially dependent on the presence of a Y chromosome in germ cells.

We suspected that upregulation of Sdgc may have been due to a conformational difference or presence of single-nucleotide polymorphisms (SNPs) between the X and Y chromosomes. If this is true, then Sdgc should be located on the sex chromosomes. To test this possibility, we mapped Sdgc using the M-markers (Kimura et al., 2012) (supplementary material Fig. S7 and Table S1). We found that both Sdgc and a sex-linked marker (SL1) flanking DMY/dmrt1bY (Matsuda et al., 2002; Nanda et al., 2002) are located between the MID0122 and MID0123 linkage markers (Fig. 3). Thus, Sdgc maps near the region where DMY/dmrt1bY resides on the sex chromosomes (LG1). Interestingly, in the Kaga strain, 11 SNPs found in the intronic region of Sdgc were heterozygous only in XY specimens (supplementary material Fig. S7), further supporting the idea that the gene is located near DMY/dmrt1bY, where homologous recombination should occur less frequently. Because Sdgc expression was detected in XX germ cells, we reasoned that Sdgc must be present on both the X and Y chromosomes, but that the two alleles have different expression.

Finally, we investigated whether germ cells behaved differently between the two sexes, using an in vitro culture system. First, germ cells dissociated from embryonic tissue at st.30 were plated onto culture dishes together with dissociated somatic cells. Then, the mitotic activity of germ cells was examined in the two different media by monitoring incorporation of EdU (supplementary material Fig. S8A and Table S2). Under both culture conditions, the rate of EdU incorporation in XY germ cells was higher than in XX germ cells (supplementary material Fig. S8C,D). To exclude the possibility that the sex of the dissociated somatic cells might affect the mitotic activity of germ cells, we isolated XX or XY germ cells at st.30 by FACS and cultured them on XX or XY feeder cells, respectively, derived from somatic cells of the tail region at st.30. After 24 h in culture, EdU treatment was performed for an additional 24 h to observe the mitotic activity of the germ cells (supplementary material Fig. S8B). Again, the rate of EdU incorporation in XY germ cells was significantly higher than in XX germ cells, and this difference was not dependent on the sex of the somatic cells (Fig. 4A). This observation suggests that an intrinsic sexual difference within germ cells contributes to the difference in germ-cell mitotic activity.

To examine the relationship between cell-autonomous gene expression and sexually different mitotic activity, we knocked down DMY/dmrt1bY and Sdgc and performed culture experiments on the resultant cells. DMY/dmrt1bY knockdown had no effect on the mitotic activity of XY germ cells (Fig. 4B), whereas knockdown of Sdgc significantly reduced mitotic activity (Fig. 4C), an effect that could be rescued by overexpression of Sdgc (Fig. 4C). Knockdown of Sdgc had no effect on XX germ cells (Fig. 4D), but Sdgc overexpression increased the mitotic activity of XX germ cells (Fig. 4D). Collectively, these results suggest that Sdgc confers higher mitotic activity upon XY germ cells relative to XX germ cells in vitro. This observation further supports the idea that the cell-autonomous acquisition of sexually different characters in germ cells is independent of DMY/dmrt1bY protein expression.

In this study, we have examined the gene expression and behavior of germ cells before gonadal primordium formation in medaka. We found that Sdgc and the sex-determination gene DMY/dmrt1bY were expressed in germ cells at this early stage. However, although Sdgc expression was upregulated only when a Y chromosome is present in germ cells, it was unaffected by DMY/dmrt1bY knockdown or overexpression. This sexually different expression was induced in a germ cell-autonomous manner. In addition, early XX and XY germ cells exhibited different mitotic activities in vitro. Together, these data show that XX and XY germ cells possess different characters even before gonadal primordium formation and the onset of sex determination by gonadal somatic cells.

The master male sex-determining gene evolved relatively recently (5-10 million years ago) from a duplicated copy of dmrt1 (Kondo et al., 2003; Matsuda et al., 2003). This evolutionary process was associated with the insertion of a transposable element called Izanagi that confers a novel expression pattern upon DMY/dmrt1bY (Herpin et al., 2010). Unexpectedly, in our study, sexually different characters of germ cells were not impaired by knockdown and overexpression of DMY/dmrt1bY. Instead, upregulation of Sdgc and increased mitotic activity of germ cells in vitro were dependent on the presence of the Y chromosome in germ cells. In addition, Sdgc mapped close to DMY/dmrt1bY on sex chromosomes and produced higher mitotic activity in XY germ cells than in XX germ cells in vitro. Although the possibility that the transcriptional level of DMY/dmrt1bY affects Sdgc expression cannot be completely ruled out, our findings suggest two possible models for inducing cell-autonomous, sexually different characters of germ cells.

First, the sexually different characters may be caused at the chromosomal level. Previous studies (Kondo et al., 2004,, 2006; Takeda, 2008) have shown that the male sex-determining gene DMY/dmrt1bY was inserted into one homologue of a chromosome pair that became the X and Y. Outside the inserted fragment and where Sdgc is located, X and Y are largely homologous and have the same gene content. Thus, Sdgc was most likely already on this chromosome pair, and both copies might have been expressed at the same level prior to the insertion event. It is likely that insertion of DMY/dmrt1bY and/or subsequent changes are related to restriction of recombination near the Sdgc locus (Kondo et al., 2001; Takeda, 2008), resulting in accumulation of DNA fragments such as repetitive sequences. Recent analyses suggest that these elements affect the structural change of chromatins through a small RNA-related mechanism and/or epigenetic regulation (e.g. Slotkin and Martienssen, 2007). The structural changes caused by these fragments may have facilitated and maintained the expression difference between the X and Y copies, leading to sexually different characters of germ cells. As for development of Sdgc, however, we cannot exclude the possibility that a high-expression version of Sdgc was inserted near DMY/dmrt1bY, and that this linkage has been conserved.

Second, infrequent recombination may result in single-nucleotide polymorphism (SNP) accumulation in the intergenic or non-coding region near Sdgc and DMY/dmrt1bY locus. This effect could be due to the possibility of either genetic hitchhiking around the DMY/dmrt1bY locus or the DMY/dmrt1bY insertion process itself. Both cases can cause sex-specific differences in the X and Y chromosomsal regions around the insertion site and consequently in different transcriptional activity between the X and Y chromosome copies of Sdgc. Consistent with this, we identified sex-specific SNPs around Sdgc. Importantly, in either model, structural changes of the chromosome or different SNPs, Sdgc could potentiate sexually different cell behavior, as evidenced by the reduction of mitotic activity of XY germ cells by Sdgc knockdown in vitro.

Regarding the events that occur differently on the X and Y chromosome prior to gonad formation in vertebrates, X-chromosome reactivation is initiated during the development of primordial germ cells in mice (Sugimoto and Abe, 2007). However, it is still not known whether this reactivation occurs in a germ cell-autonomous manner or instead is affected by surrounding somatic cells. Our chimeric analysis provides good evidence that germ cells in medaka exhibit cell-autonomous, sexually different characters that are independent of somatic cells.

No obvious effect of Sdgc knockdown on germ cells was observed during late developmental stages of sex differentiation in vivo. One possibility is that mitotic activity of germ cells is predominantly regulated by somatic cells at that stage, thus cell-autonomous effects may be overridden. It is also possible that the Sdgc-gripNA, which was injected at the one- or two-cell stage, is not sufficiently active by the sex differentiation stage, which occurs late during development. Further investigation will be necessary to resolve this issue.

Knockdown of DMY/dmrt1bY did not affect expression of Sdgc. DMY/dmrt1bY expression was also detected in the Sdgc-knockdown XY embryos at the onset of sex determination by somatic cells (supplementary material Fig. S9A,B). Thus, given that DMY/dmrt1bY functions as a protein, Sdgc seems to function in a pathway independent of DMY/dmrt1bY.

Our findings provide insight into the relationship between sex chromosome differentiation and sexual plasticity at the cellular level. Many studies indicate that the medaka (Oryzias latipes) sex chromosome arose relatively recently (Herpin et al., 2010; Kondo et al., 2004; Myosho et al., 2012; Takehana et al., 2007) and is not as differentiated as in mammals, in which sexual plasticity is much reduced in germ cells (e.g. Lavery et al., 2011; Koopman et al., 1991). However, our results indicate that the newly arisen sex chromosomes can confer cell-autonomous, sexually different characters upon early stage germ cells. These results may indicate that the medaka sex chromosomes are on the evolutionary path to developing genes that influence sexually different characters of germ cells. These results also indicate that even in medaka, in which sex is determined by a specific gene, additional cell-autonomous mechanisms may affect sexual plasticity.

All the treatments of animals in this research followed the guideline of National Institute for Basic Biology and were approved by the Institutional Animal Care and Use Committee of National Institutes of Natural Sciences. The OKcab strain and sox9b-DsRed/olvas-EGFP transgenic medaka (Nakamura et al., 2010) were used in this study. To produce all XX progeny, XX males, which had been treated with 100 ng/ml 11-ketotestosterone during embryonic development, were crossed with XX females. To produce all XY progeny, first, XY females were generated by treatment with 200 ng/ml estradiol and were crossed with XY males. As a consequence, ∼25% of the progeny were YY males, which were subsequently crossed with XX females to generate all XY progeny. All XX and XY progenies were used for the isolation of germ and somatic cells by FACS, in situ hybridization, cell culture and qPCR experiments as described below. The Hd-rR-II1 strain was also used for qPCR experiment. To ablate germ cells, we injected 2000 ng/µl cxcr4-morpholino (MO) and 1000 ng/µl nanos3-MO into one- or two-cell stage embryos (Kurokawa et al., 2006,, 2007).

A bacterial artificial chromosome (BAC) transgenic method that uses homologous recombination was employed to generate a DMY/dmrt1bY-EGFP-3′UTR reporter construct as previously described (Nakamura et al., 2008). Specifically, the targeting DNA fragment for recombination was prepared to include the DMY/dmrt1bY-3′UTR downstream of the EGFP open reading frame. After homologous recombination, this fragment was inserted immediately downstream of the translation initiation site of the DMY/dmrt1bY gene in BAC clone DMY H1. Microinjection of BAC clone DNA was performed as previously described (Nakamura et al., 2008). For early germ cell tracking, one-cell stage embryos were injected with mCherry:nos3-3′UTR (mCherry ORF fused to the 3′UTR of nanos3) capped RNA as previously described (Herpin et al., 2007).

All XX and all XY progeny of sox9b-DsRed/olvas-EGFP transgenic medaka were used in this study. Approximately 500 gonadal fragments, which included intestines, gonads and body trunks, were dissected from st.33 and st.35 embryos and were digested with 0.2% collagenase (Worthington, CLS4) and 1% trypsin (Worthington, TRL) in Leibovitz's L15 for 2 h at 29°C. The cell suspension was the resuspended with L15 supplemented with 10% FBS to stop the digestion reaction, and was filtered through 35 µm cell strainers.

olvas-EGFP= and sox9b-DsRed-positive cells were isolated by FACS (Beckman Coulter, Coulter Epics Altra) according to fluorescence intensity and size. Total RNA was extracted from 8000-10,000 cells using an RNaqueous-Micro Kit (Ambion, AM1931), followed by cRNA labeling using a Low Input Quick Amp Labeling Kit (Agilent, 5190-2305) according to the manufacturer's instructions. Hybridization of cRNA targets was performed on a 60-mer oligo microarray (Agilent, Design ID: 027381). The microarray experiments were carried out using three independent biological replicates. Scanned data were processed by Feature Extraction (Agilent) and further analyzed by the Subio platform. The microarray data have been deposited at the NCBI Gene Expression Omnibus (GEO) under accession number GSE59521.

Whole-mount in situ hybridization and immunohistochemistry were performed as previously described (Nakamura et al., 2006). A cDNA clone for Sdgc (clone name: olte54h09) was obtained from NBRP medaka (http://www.shigen.nig.ac.jp/medaka/). For immunohistochemistry, anti-OLVAS (medaka Vasa antibody, 1:100, rat) (Aoki et al., 2008) or anti-GFP (1:100, mouse; Clontech, 632381) was used as a primary antibody that was detected with Alexa 488-conjugated secondary antibodies (1:100, goat; Life Technologies, A11029).

sox9b-DsRed/olvas-EGFP transgenic embryos (Nakamura et al., 2010) were used as hosts, and OKcab embryos were used as donors. To eliminate the host's germ cells, 2000 ng/µl cxcr4-MO and 1000 ng/µl nanos3-MO were injected into one- or two-cell stage embryos (Kurokawa et al., 2006,, 2007). The transplantation procedure was performed as previously described (Nakamura et al., 2012b).

Germ cells are located at the lateral sides of the intestine in st.30 embryos. In each experiment, approximately 50 tissue fragments containing germ cells were dissected from XX and XY embryos, and total RNA was extracted using Isogen (Nippon Gene, 317-02503). For the experiment using cxcr4/nanos3-MO-treated embryos and chimeric embryos, 20-30 olvas-EGFP-negative embryos (no endogenous germ cells) were used. For extracting RNA from adult gonads, three testes and one ovary were used in each experiment. After total RNA extraction, cDNA was produced using SuperScript III (Invitrogen, 18080-044) or ReverTra Ace qPCR RT Kit (Toyobo, FSQ-101). Quantification was performed with the ABI StepOnePlus Real-Time PCR system (Applied Biosystems) and SYBR Green (Toyobo, QPS-201). The primers used in this study are shown in supplementary material Table S3. The amount of target gene normalized to the β-actin or olvas (for chimeric analysis) expression and relative to the XX gonads was calculated using the ΔΔC_T method. All qPCR reactions were performed in triplicate. Three independent experiments using independent pools of normal gonads and two independent experiments using independent pools of morpholino-treated gonads and chimera gonads were performed.

To identify SNPs in the genomic region of the Sdgc gene, the genomic fragments were amplified and sequenced using DNA from OKcab (XX: n=4, XY: n=4) and Kaga (XX: n=4, XY: n=6) inbred lines. From OKcab DNA, 535 bp genomic fragments, including an intron, were amplified and sequenced by q-Sdgc-F and -R primers (see supplementary material Table S3). From Kaga DNA, 346 bp genomic fragments were amplified and sequenced using gSdgc-F and q-Sdgc-R primers. As a result, 14 SNPs were observed in the 346 bp amplicon (supplementary material Fig. S7).

For single nucleotide polymorphism (SNP) genotyping, XX Kaga and XY Hd-rR-II1 were crossed to generate the F1 progeny, which were subsequently intercrossed to obtain F2 progeny (Kimura et al., 2012). Genomic DNA from the F2 progeny (n=94) was amplified using gSdgc-F and q-Sdgc-R. The PCR products were then diluted 100-fold and used as templates for nested PCR, the products of which were further genotyped via High resolution melting analysis using a LightScanner (Idaho Technology) (supplementary material Table S1). Locus order was determined together with M-markers (Kimura et al., 2012) by AntMap (http://lbm.ab.a.u-tokyo.ac.jp/~iwata/antmap/).

GripNA (Active Motif) was designed to target the translation initiation site of DMY/dmrt1bY and Sdgc at the sequence CATGTCAGCCCGGGGAGC (Paul-Prasanth et al., 2006) and CTCGGGTTGGTTTTGCAT, respectively. The underlined CAT regions are complementary to the ATG start codon. For qPCR analysis and cell culture experiments, 0.5 mM DMY/dmrt1bY-gripNA and 0.25 mM Sdgc-gripNA were injected into one- or two-cell stage embryos. Injection of 0.1 mM human-CREB gripNA was used as a control.

Sdgc ORF was amplified using FLAG-tagged primers and inserted into a SalI site of the pBLSK+ PTV1-2A-mCherry vector (Nakamura et al., 2012b). olvas-3′UTR was inserted into downstream of the mCherry by InFusion (Clontech, 639649) to generate FLAG:Sdgc:2A:mCherry:olvas-3′UTR. Capped RNA was synthesized with T7 promoter using mMESSAGE mMACHINE kit (Ambion, AM1344). For overexpression, 200 ng/µl of the mRNA was injected into XX embryos. For the rescue experiment, 200 ng/µl of the mRNA and 0.25 mM Sdgc-gripNA were injected into XY embryos. The subsequent culture experiments are described below.

The construction of EGFP:nos3-3′UTR (EGFP ORF fused to the 3′UTR of nanos3) has been described previously (Kurokawa et al., 2006). Based on the EGFP:nos3-3′UTR construct, DMY/dmrt1bY ORF was inserted into downstream of EGFP by In-Fusion (Clontech) to generate EGFP:DMY:nos3-3′UTR. Capped RNA was synthesized with SP6 promoter using mMESSAGE mMACHINE kit (Ambion, AM1340). 50 ng/µl of each construct was injected into one- or two-cell stage embryos.

Approximately 200 tissue fragments containing germ cells at st.30 were dissociated using 0.1% trypsin and 0.1% collagenase in PBS at 29°C for 1 h. The cells were resuspended in L15 containing 10% FBS to stop the enzymatic activity. Next, GFP-positive germ cells were isolated by FACS (Beckman Coulter, Coulter Epics Altra) using the enrich mode. Approximately 5.0×10³ XX and 7.8×10³ XY germ cells were successfully isolated. For the preparation of somatic cells, tails from XX and XY embryos were separately dissociated using trypsin and collagenase under the same conditions described above. Germ cells and somatic cells were divided into two tubes and mixed in the following combinations: germ XX-soma XX, germ XX-soma XY, germ XY-soma XX and germ XY-soma XY. The cell mixtures were divided into three portions and plated on collagen-coated cover slips that had been placed in a six-well dish. After a 2 h incubation, 2 ml of Medium #1 (see supplementary material Table S2) was added to each dish. After a 24 h incubation at 26°C, the culture medium was changed to include 25 µM EdU, and cells were cultured for an additional 24 h, then fixed in 4% paraformaldehyde and analyzed by immunohistochemistry. For knockdown and overexpression experiments, 30-40 tissue fragments containing germ cells at st.30 were dissociated as described above. The cell suspension was then divided into two portions and plated on collagen-coated cover slips. Cell culture and EdU treatment were performed as described above.

We thank Ms C. Noda (National Institute for Basic Biology) for FACS sorting and NBRP medaka for providing cDNA. The DMY H1 BAC clone was kindly provided by Dr Matsuda at Utsunomiya University. We are also grateful to Ms Ichikawa and Mr Suzuki for fish maintenance.