In the zebrafish, no sex-determining gene has been identified, while some sex-related genes, such as cyp19a1a and amh, show sexually dimorphic expression. Interestingly, most of these genes are expressed in the somatic cells. With increasing evidence suggesting roles of germ cells in gonadal differentiation, there is an increasing interest in the factors released by the germ cells for the bidirectional communication between the two compartments. We have reported that Gdf9/gdf9 is an oocyte-specific factor in the zebrafish, similar to that of mammals. Whether and how Gdf9 is involved in gonadal differentiation is unknown. In this study, we compared the expression levels of gdf9, cyp19a1a, and amh among several other sex-related genes in the gonads before, during, and after sex differentiation. The expression of gdf9 started in the gonads before sex differentiation, and its level surged in the differentiated ovary. Its expression pattern was similar to that of cyp19a1a, but reciprocal to amh expression. Using recombinant zebrafish Gdf9 (rzfGdf9), we further showed that Gdf9 significantly suppressed the expression of amh while increased that of activin beta subunits (inhbaa and inhbb) in vitro. Although gdf9 and cyp19a1a showed co-expression during gonadal differentiation, we only observed a slight but not significant response of cyp19a1a to rzfGdf9. Knocking down the expression of gdf9 and cyp19a1a with vivo-morpholinos caused a male-skewed sex ratio. Our data suggested that Gdf9 is likely involved in promoting oocyte/ovary differentiation in the zebrafish and it may act by suppressing amh expression, at least partly, in the somatic cells.
Sex determination and differentiation are long-lasting questions in developmental and reproductive biology. So far, only a few sex-determining genes have been identified in vertebrates, e.g., SRY in mammals , DMY/dmrt1bY in Japanese medaka , amhy in Patagonian pejerrey  and tilapia , DMRT1 in chicken , and DMW in the African clawed frog . In most vertebrate species, the mechanism of sex determination remains largely unknown, especially in teleost fish . In contrast to sex determination, the genetic and physiological mechanisms that govern sex differentiation are better documented in fish .
In teleost fish, gonadal sex differentiation is highly plastic and can be influenced by both genetic and environmental factors such as pH and temperature [9,10]. In the zebrafish, no sex chromosomes or sex-linked genes have been found [11,12], despite being a popular vertebrate model for developmental biology and genetic study. The zebrafish is considered a juvenile hermaphrodite; its gonads first develop as ovary-like organs with oocyte-like germ cells (juvenile ovary) before further differentiation into true ovary and testis [13–16]. During testis differentiation, the oocyte-like germ cells regress through apoptosis , which is accompanied by the appearance of spermatogonia and proliferation of the stromal somatic cells that gradually occupy the gonad (transforming gonad), leading to final differentiation into the testis . Although gonadal differentiation has been well documented in the zebrafish at morphological and histological levels [13,14,17], the molecular mechanism underlying the process remains largely unknown.
It is widely accepted that P450 aromatase (cyp19a1), a key steroidogenic enzyme that converts androgen into estrogen, is actively involved in ovarian differentiation in fish . There are two different cyp19a1 loci in the zebrafish: cyp19ala is mainly expressed in the gonad, while cyp19a1b mainly in the brain [19,20]. Cyp19a1a showed an increased expression during ovarian differentiation in many fish species such as the Japanese medaka , rainbow trout , Nile tilapia [23,24], European seabass, and southern flounder . Similarly, cyp19a1a also increased expression during ovarian differentiation in the zebrafish , but sharply decreased when undergoing transformation into testis . The importance of aromatase in gonadal or sex differentiation is also supported by the evidence that treatment with aromatase inhibitor (AI) during the sex differentiation period led to the formation of testis in the rainbow trout , Japanese flounder , Nile tilapia , Chinook salmon , common carp , and zebrafish [31,32]. Recently, it was also reported that AI treatment could induce sex reversal from female to male in adult zebrafish . Therefore, aromatase is being considered a key factor for ovarian formation, development, and maintenance in fish. As aromatase is expressed in the somatic follicle cells, its importance in sex differentiation suggests a role for the somatic cells in determining or influencing gonadal differentiation. This concept is also supported by other genes such as Sry in mammals , amh in mammals and fish [35,36], amhy in Patagonian pejerrey , and dmy in medaka .
Although the important roles played by somatic cells in gonadal differentiation are well accepted, lines of evidence from recent studies suggest that the germ cells also play an important role in determining gonadal differentiation. In mammals, germ line is required for the initiation of follicle formation in females . Without meiotic germ cells, the ovarian pregranulosa cells transform into Sertoli-like testis cells . Similarly, the germ cells have also been shown to be essential for female sex differentiation in the medaka [39,40], loach , and zebrafish [42,43]. In the zebrafish, specific ablation of primordial germ cells in the embryo by toxin or knocking down of germ cell-specific gene dnd by morpholino led to the formation of sterile males [42,43]. Interestingly, low numbers of germ cells also led to male development in the zebrafish [44,45]. Recently, foxl3, a germ cell-specific gene, was shown to promote germ cell differentiation to oocytes by suppressing spermatogenesis in the medaka . Further evidence in the zebrafish shows that the germ cells are not only critical for gonadal differentiation during the primary sex-determining period, but also important in maintaining female status in adult stage . These results provide strong evidence that germ cells are crucial for ovarian development and maintenance of gene expression in the ovarian somatic cells in fish including the zebrafish. However, the mechanisms by which the germ cells influence the process remain entirely unknown.
Growth differentiation factor 9 (GDF9), a member of transforming growth factor beta (TGFβ) superfamily, is an oocyte-specific growth factor in different vertebrates including mammals, chicken, and zebrafish [48–50], and it is obligatory for early folliculogenesis in mice . Both the granulosa and theca cells are the targets of GDF9 , which regulates a wide range of activities including granulosa cell proliferation , secretion of steroid hormones , and biosynthesis of peptide growth factors . The expression of GDF9 reduced rapidly after fertilization in mice and was no longer detectable by day 1.5 postcoital (pc). The expression showed up again in 4-day neonatal ovaries . In sheep, GDF9 mRNA was detectable in the germ cells before the formation of follicles  and in the oocytes of newly formed primordial follicles around 78–85 days after birth in possums . In the rainbow trout, gdf9 began to express in early gametogenesis period (days 60 to 110 after complete yolk consumption) in the ovary . These lines of evidence suggest that GDF9 is expressed in early developing ovary in different species, but the exact time of its appearance varies depending on species, and its role in gonadal differentiation is largely unknown.
To provide clues to its potential role in gonadal differentiation, we undertook this study to analyze qualitatively and quantitatively the spatiotemporal expression profiles of gdf9 during gonadal differentiation in the zebrafish. In addition, we also analyzed the expression of a variety of genes that are potentially involved in gonadal differentiation for comparison, especially cyp19a1a and amh, which have been reported to be associated with female and male development, respectively [3,18,33]. To understand how Gdf9 acts during gonadal differentiation, we also generated a stable cell line that expresses recombinant zebrafish Gdf9 (rzfGdf9). With recombinant Gdf9, we demonstrated its signaling activity in cultured follicle cells and its effects on the expression of cyp19a1a and amh. In addition, we also examined the regulation of the activin–inhibin–follistatin family by Gdf9.
Material and methods
The wild-type zebrafish (Danio rerio) were kept in flow-through aquaria at 28±1°C on 14L:10D lighting cycle. An albino strain, which has transparent body wall at juvenile stage, was maintained in the laboratory and used for microinjection of vivo-morpholinos (VMO). The fish were fed twice a day with commercial fish food with supplement of live brine shrimp twice a week. All experiments were endorsed by the Research Ethics Committee of University of Macau and the Animal Experimentation Ethics Committee of The Chinese University of Hong Kong.
Culture of fish fry and sampling of gonads from larvae and juveniles
Mature females and males were kept separately for 4–5 days and placed together 1 day before spawning. The fertilized eggs were incubated at 27±1°C. After hatching, the larval fish were cultured under the same photoperiod (14L:10D) and fed with paramecium first followed by commercial baby fish diet (Tetra Baby, Melle, Germany) and Artemia salina nauplii. Samples were collected at different days to examine the expression of gdf9, amh, cyp19a1a, and other related genes. The larvae from 13 to 20 dpf (day post fertilization) were cooled on ice and stored in RNAlater (Ambion, Austin, TX) at 4°C overnight. They were then transferred to –20°C for at least 1 week before isolating the gonads for RNA extraction. For juvenile fish, they were sampled from 27 to 40 dpf when gonads were differentiating. The sampling times were determined according to our previous work on the timeline of zebrafish gonadal development . After quick cooling on ice to sacrifice the fish, each was halved by cross-sectioning around the communicating duct between the two air bladders (Figure 1). The posterior half of the body was kept in RNAlater at 4°C overnight and transferred to –20°C afterward. The anterior part was fixed in Bouin solution for histological examination to identify the stage of gonadal development. The staging of gonadal development was based on the report of Maack and Segner . The presence of perinucleolar oocytes (PO) was used as the marker for the ovary, while the appearance of spermatogonia or spermatocytes marked the differentiation of testis. The gonads with both degenerating oocytes and developing spermatocytes were considered transforming gonad or early testis.
RNA isolation, reverse transcription and polymerase chain reaction
Total RNA was isolated from the gonads of larval or juvenile fish with Tri-Reagent (Molecular Research Center, Cincinnati, OH) according to the protocol of the manufacturer and our previous reports [50,60]. The RNA extracted from the gonads of each fish was reverse transcribed into cDNA at 42°C for 1 h in a total volume of 10 μl reaction solution that contains total RNA, iScript reaction mix with reverse transcriptase (Bio-Rad, Hercules, CA).
Polymerase chain reaction (PCR) amplification was carried out using gene-specific primers flanking an intron (Supplemental Table S1). The reaction was performed in a volume of 30 μl consisting of 10 μl reverse transcription (RT) reaction product (1:10 diluted with water), 1× PCR buffer, 0.2 mM of each dNTP, 2.5 mM MgCl2, 0.2 μM of each primer, 0.6 U Taq polymerase with the profile of 30 s at 94°C, 45 s at 56°C for gapdh, 58°C for cyp19a1a, bactin, dmrt1, 60°C for gdf9, star, sox9b, 63°C for sox9a, and 65°C for amh, and 60 s at 72°C for extension. The cycle numbers of PCR amplification were 30 for gdf9, cyp19a1a, and other genes and 27 for gapdh and bactin.
Real-time qPCR quantification of gdf9, cyp19a1a, and amh expression
The expression levels of gdf9, cyp19a1a, amh, and ef1a in the gonads from each individual larval or juvenile fish at different developmental stages were determined by real-time qRCR. The standard for each gene was prepared by PCR amplification of cDNA fragments with specific primers (Supplemental Table S1). The amplicons were resolved by agarose gel electrophoresis, purified, and quantified by electrophoresis together with the Mass Ruler DNA marker (MBI Fermentas, Hanover, MD). These amplified amplicons were used to construct standard curves in real-time qPCR assays.
Real-time PCR was carried out on the iCycler iQ Real-Time PCR Detection System (Bio-Rad) in a volume of 30 μl that contained 10 μl diluted RT reaction mix, 1× PCR buffer, 0.2 mM of each dNTP, 2.5 mM MgCl2, 0.2 μM of each primer, 0.75 U Taq polymerase, 0.5× EvaGreen (Biotium, Hayward, CA), and 20 nM fluorescein (Bio-Rad). The reaction profile consisted of 38 cycles of 94°C for 30 s, 60°C for 30 s, 72°C for 1 min, and 80°C for 7 s for signal detection. A melt curve analysis was performed at the end of amplification to demonstrate reaction specificity, which was further confirmed by agarose gel electrophoresis.
Fluorescent in situ hybridization
The zebrafish gdf9, cyp19a1a, and amh cDNAs were amplified with specific primers (Supplemental Table S1), and the amplicons were cloned into pBluescript II KS (+). These plasmids were then used as the templates for probe synthesis. The sense and antisense probes were labeled with DIG or fluorescein by RNA labeling kit (Roche Applied Science, Mannheim, Germany). The fish body with gonads was fixed in freshly prepared paraformaldehyde (4%) for 2–4 h at room temperature and stored in PBS buffer at 4°C before processing for paraffin embedding and sectioning at 5 μm thickness. The sections were mounted on slides coated with poly-(L)-lysine, and hybridized with sense (control) or antisense cRNA probes of gdf9, cyp19a1a, and amh, respectively, according to the protocol we reported recently . Briefly, the sections were deparaffinized, washed with PBS, and treated with proteinase K (4 μg/ml in PBS) for 10 min at 37°C. After rinsing with PBS, the sections were postfixed for 15 min with 4% paraformaldehyde in PBS before treatment with glycine (2 mg/ml) for 10 min, followed by 10 min treatment with 0.25% acetic anhydride in 0.1 M triethanolamine buffer. After a 10-min wash with 2 × SSC, the sections were prehybridized at room temperature for 1 h in 66% formamide with 2 × SSC, and hybridized at 58°C overnight with DIG or fluorescein-labeled antisense or sense probes in the hybridization buffer (60% deionized formamide, 7.5% Dextran sulfate, 1 × Denhardt solution, 20 mM Tris-HCl, 2.5 mM EDTA, 350 mM NaCl, and 0.2 mg/ml tRNA). After a series of washes with 5 × SSC, 50% formamide/50% 2 × SSC, 2 × SSC, and 0.2 × SSC at 58°C, the sections were rinsed in maleic acid buffer (0.1 M maleic acid, 0.15 M NaCl, 0.01% Tween 20) and incubated with 100 μl 1% Blocking Reagent (Roche) at room temperature for 60 min. The slides were then incubated with 100 μl diluted anti-DIG or anti-fluorescein horseradish peroxidase (HRP) (1:500 dilution in 1% Blocking Buffer) in a humidified chamber for 2 h at room temperature. After washing slides with maleic acid buffer three times, the sections were treated with the TSA-Plus Fluorescein/cy5/TMR (PerkinElmer, Waltham, MA) in dark for 10 min. The signal was detected by HRP-conjugated anti-fluorescein antibody (Roche) with TSA-fluorescein according to the manufacturer's instruction. The sections were finally washed with PBS and mounted with Prolong Gold antifade reagent (Invitrogen, Carlsbad, CA); DAPI was added at this step. The slides were observed and the images recorded using Olympus Fluoview confocal microscope (Olympus, Tokyo, Japan).
Recombinant production of zebrafish Gdf9
To produce recombinant zebrafish Gdf9, we used Flp-In CHO cells (Invitrogen) as the bioreactor . Briefly, the open reading fragment (ORF) of gdf9 was amplified from the ovarian cDNA with pfu polymerase and a sense primer carrying the Kozak sequence around the start codon ATG to enhance translation efficiency . The ORF was cloned into the pcDNA5/FRT vector (Invitrogen) downstream of the CMV promoter to generate the expression construct, pcDNA5/FRT/zfGDF9. The plasmid was co-transfected into the CHO cells with pOG44 encoding a recombinase by Lipofectamine 2000 and selected by hygromycin B (500 μg/ml) (Invitrogen) for 2 weeks. Individual clones were isolated by serial dilution in 96-well plates, and a clone with integration of gdf9 ORF in the genome was chosen for production of recombinant Gdf9 according to the protocol described by Schatz et al. and modified by our laboratory [61,63]. The expression of gdf9 in the CHO cells was examined by northern blot hybridization , and the preparation of conditioned medium containing recombinant Gdf9 was carried out according to our previous report .
Primary culture of zebrafish follicle cells and incubation of ovarian fragments
The primary culture of zebrafish ovarian follicle cells and incubation of ovarian fragments were performed according to our previous reports [65,66]. For follicle cell culture, the follicles of previtellogenic (PV) to midvitellogenic (MV) stages from about 20 female zebrafish were isolated, washed with medium M199, and incubated in M199 supplemented with 10% FBS at 28°C in 5% CO2 for 6 days for the follicle cells to proliferate. The medium was changed once on the third day of incubation, and the follicle cells were harvested by trypsinization on the sixth day for subculturing in 48-well plates (1 × 105/well). After 18 h of preincubation, the cells were treated with concentrated serum-free conditioned medium from the recombinant CHO cells for 20–180 min. After washing three times with M199, the follicle cells were subject to total RNA extraction with Tri-Reagent for real-time qPCR analysis. For western blot analysis, proteins were extracted from the cultured cells, electrophoresed, and blotted to nitrocellulose membrane. The immunoblot was incubated with p-Smad2 antibody (Cell Signaling, MA) (1:1000) followed by incubation with the HRP-conjugated goat anti-rabbit antibody (1:3000). Signals were detected using the Western Blotting Luminol Reagent (Santa Cruz, CA) and the images were captured on the Lumi-Imager F1 Workstation (Roche).
For ovarian fragment incubation, the ovaries were dissected from six female zebrafish and placed in a petri dish containing 60% Leibovitz 15 medium (L-15, Gibco BRL, Grand Island, NY). The ovaries were carefully separated into small fragments with fine forceps and randomly distributed into 24-well tissue culture plate (Corning, NY). After treatment with recombinant conditioned medium for 12 h at 28°C, the ovarian fragments were collected for total RNA extraction followed by real-time qPCR analysis.
Injection of vivo-morpholinos
The antisense VMOs for gdf9 and cyp19a1a as well as standard control VMO were obtained from Gene Tools (Philomath, OR), and their sequences are as follows: 5΄-GGCGTAATAATAAGCATACCTCTGT-3΄ (gdf9-MO), 5΄-GCCAGCTACAGAACACAAACAGAAC-3΄(cyp19a1a-MO) and 5΄-CCTCTTACCTCAGTTACAATTTATA-3΄(CTL-MO). The VMOs of gdf9 and cyp19a1a were designed to block mRNA splicing so as to knock down gdf9 and cyp19a1a expression. The sterile VMOs were prepared as 0.5 mM (4.2 ng/nl) stocks in distilled water for gdf9-MO, cyp19a1a-MO and CTL-MO. The larvae of 18 dpf with less than 10 mg body weight and undifferentiated gonads were anesthetized with MS222 (0.25 mg/ml), weighted, and injected with VMOs at 12.5 ng/mg body weight. Vivo-morpholinos were injected into the gonad region around the posterior bladder with microinjector every 2 days and seven injections were performed in total. The developmental state of the gonads was determined by histological examination around 40 dpf when the gonads had almost completed differentiation.
The expression levels of gdf9, cyp19a1a, and amh were normalized to that of internal control ef1a. All values were expressed as the mean ± SEM, and the data were analyzed by the Dunnett multiple comparison test with Prism on Macintosh OS X (GraphPad Software, San Diego, CA). The experiments were repeated at least three times.
Expression of gdf9, cyp19a1a, and amh in differentiated gonads at 40 dpf
Juvenile fish around 40 dpf with similar size (1.3–1.5 cm in body length) were sampled, and the body was severed in half around the communicating duct between the anterior and posterior chambers of the swim bladder. The anterior half was fixed for histological examination to identify gonad developmental stage, while the posterior half was stored in RNAlater for gonadal isolation and expression analysis (Figure 1A). The gonads could be easily identified and separated under dissecting microscope after pretreatment with RNAlater (Figure 1B and C).
Histological examination showed that the ovary contained tightly packed PO (<100 μm) (Figure 2A), and the testis was full of spermatogonia, spermatocytes, and even spermatids (Figure 2B). By comparison, the transforming gonads or early testis was small, containing spermatogonia/spermatocytes, remnants of POs, and abundant stromal cells (Figure 2C).
With the gonads categorized into the above three stages of differentiation, we then examined the expression of a variety of genes that are potentially involved in gonadal differentiation with gapdh and bactin being used as the housekeeping genes. The expression of gapdh and bactin showed similar patterns in all three types of gonads despite certain individual variation, indicating that our sampling method involving RNAlater was appropriate for analyzing gene expression (Figure 2D). Among the functional genes analyzed, cyp19a1a, foxl2 (forkhead box protein L2) and sox9b (SRY-box containing gene 9b) are well known to be related to ovarian differentiation , while dax1 (dosage-sensitive sex reversal, adrenal hypoplasia critical region, on chromosome X, gene 1), amh (anti-Müllerian hormone), dmrt1 (doublesex and mab-3 related transcription factor 1), sox9a (SRY-box containing gene 9a), and star (steroidogenic acute regulatory protein) have abundant expression in differentiated testis in fish . Our results in this experiment showed that gdf9 and cyp19a1a exhibited clear dimorphic expression patterns in the differentiating gonads. Their expression could be easily detected in the ovaries, but not testes. In the transforming gonads, about half of individuals showed detectable expression. Interestingly, although foxl2 is well known to be important in controlling cyp19a1a expression, its expression could be detected ubiquitously in all gonads. The same pattern was observed for dax1. By comparison, although the expression of amh could also be detected in all three types of gonads, its expression level was obviously more abundant in the testes than the ovaries. Similar pattern of expression was also observed for dmrt1 and sox9a. For sox9b and star, the expression could be detected in some individuals of all three types of gonads, but the expression pattern was not clear although the positive signal could be more frequently detected in the testes (Figure 2D).
Expression of gdf9, cyp19a1a, and amh in differentiating gonads at 27 dpf
To further illustrate the temporal expression of gdf9, cyp19a1a, and amh during gonadal differentiation, we analyzed their expression levels at an earlier stage (27 dpf, total body length 0.8–1.1 cm) when the differentiation was starting. Again, the developmental stage of the gonads was identified by histological examination of the anterior half of the body. Unlike the stage at 40 dpf when typical ovary and testis were forming or had formed, the gonads at 27 dpf could also be categorized into three types, but of earlier developmental stages: the true ovary full of POs, juvenile ovary with small POs and/or undifferentiated meiotic germ cells, and the gonad with undifferentiated germ cells only (most likely destined to become testis). No testis with typical spermatogenesis was observed at this stage (Figure 3A).
Semiquantitative RT-PCR and real-time qPCR analyses showed that the expression of gdf9 was dimorphic with strong expression in the differentiated ovary, weak expression in the juvenile ovary, but very low expression in the undifferentiated gonad (Figure 3B and C). Similarly, the expression of cyp19a1a could be detected in the ovary and juvenile ovary of some individuals, but not in the undifferentiated gonad (Figure 3B and C). Similar to the observation at 40 dpf, gdf9 and cyp19a1a were co-expressed during gonadal differentiation. As for amh, although its expression could be detected in all three forms of gonads by semiquantitative RT-PCR (Figure 3B), it showed a reciprocal expression pattern as compared with those of gdf9 and cyp19a1a. Real-time qPCR showed that the expression level of amh was very low in the newly formed ovary, increased in the juvenile ovary, and reached the highest in the undifferentiated gonads destined to become the testis (Figure 3C).
Expression of gdf9, cyp19a1a, and amh in undifferentiated gonads (13–20 dpf)
Under our aquarium condition, the zebrafish gonads are either undifferentiated or developed as juvenile ovary with small POs before 20 dpf. The differentiation into true ovary occurs between 25 and 30 dpf, followed by testis formation around 40–45 days . In this experiment, we examined the expression of gdf9, cyp19a1a, and amh between 13 and 20 dpf before differentiation by real-time qPCR. Despite very low levels, the expression of all three genes (gdf9, cyp19a1a, and amh) could be detected in the gonads of 13-dpf fish (Figure 4). From 13 to 20 dpf, the expression of cyp19a1a and amh remained low and stable despite a minor surge at 15 dpf. In contrast, there was an obvious increase in gdf9 expression starting from 16 dpf, and the level remained high and stable to 20 dpf. Despite obvious trends, no statistical significance was detected due to large variations (Figure 4).
Oocyte-specific expression of gdf9 during gonadal differentiation
The expression of gdf9 was known to be oocyte-specific in adult ovaries of both mammals and fish [50,68,69]. However, little information is known about its location in early stage of development during gonadal differentiation. In this experiment, we examined the distribution of gdf9 expression in differentiating gonads (27–30 dpf) by fluorescent in situ hybridization. Strong signal of gdf9 expression was detected in the ooplasm of all POs in the differentiated ovary (Figure 5A). In the juvenile ovary or transforming gonad, degenerating oocytes (around 30 μm in diameter) were sometimes present. These oocytes still showed the expression of gdf9, but the signal was much weaker and sometimes fragmentary (Figure 5B and C). In comparison, the expression of cyp19a1a was weak in the ovary and could be detected sporadically in some follicle cells encompassing the oocytes (Figure 5D). On the contrary, amh was expressed in somatic cells in the juvenile ovary around 30 dpf, but the signal was weak (Figure 5E). In the transforming gonad, the expression of amh was the highest among the three stages and it was expressed in the somatic cells, presumably the Sertoli cells (Figure 5F).
Evidence for roles of gdf9 and cyp19a1a in ovarian differentiation
Aromatase (cyp19a1a) is well known to be important in driving ovarian differentiation in fish . However, as an oocyte-specific growth factor, the role of gdf9 in this process remains entirely unknown. In this experiment, we adopted a special morpholino, VMO, to block the expression of gdf9 in vivo followed by sex ratio analysis. The gdf9-VMO was injected into larval fish at 18 dpf before gonadal differentiation started. Since the effect of aromatase on ovarian development has been reported in the zebrafish [31,33], we also targeted cyp19a1a as the positive control with cyp19a1a-VMO that spans the boundary of exon 8 and intron 7 (Figure 6A). The VMOs for gdf9 and cyp19a1a were injected into the larvae of the same batches at the same time. The VMO for human actin was used as the negative control (CTL-VMO). About 70 larval fish in total were injected with each VMO.
Injection of CTL-VMO did not change the sex ratio at 40 dpf when gonads had nearly finished sex differentiation. The numbers of differentiated females and males were similar with only a few individuals still undergoing gonadal transformation from juvenile ovary to testis. In contrast, the injection of gdf9-VMO significantly decreased the number of female fish and increased that of males, and there seemed to be more fish undergoing transformation. Similar results were obtained for cyp19a1a-VMO (30% ovary, 60% testis, and 10% transforming gonad) (Figure 6B).
Effects of recombinant zebrafish Gdf9 on amh and cyp19a1a expression in vitro
As shown above, cyp19a1a and amh were both expressed in the somatic cells in the gonads and they exhibited reciprocal expression patterns during gonadal differentiation. To demonstrate if the oocyte-derived Gdf9 plays a role in regulating the expression of cyp19a1a and amh, we prepared recombinant zebrafish Gdf9 (rzfGdf9) with the CHO cells. The activity of rzfGdf9 was assessed by its stimulation of Smad2 phosphorylation in cultured follicle cells. As shown in Figure 7A, the conditioned medium could significantly increase the phosphorylation of Smad2, the protein involved in Gdf9 signaling [54,70], in a time-dependent manner. Treatment of the ovarian fragments did not have significant effect on cyp19a1a expression despite a slight increase; however, it significantly reduced the expression level of amh in a dose-dependent manner (Figure 7B).
Effects of recombinant zebrafish Gdf9 on activin–inhibin–follistatin system in vitro
In addition to cyp19a1a and amh, we also examined the effects of rzfGdf9 on the expression of activin–inhibin–follistatin system in cultured ovarian fragments. Our previous work has demonstrated that the subunits of activin and inhibin (inha, inhbaa, and inhbb) are exclusively expressed in the somatic follicle cells [66,71,72] and the expression of beta subunits inhbaa and inhbb increased significantly during follicle activation from primary growth (PG) to PV stage, especially inhbaa, when gdf9 expression decreases [50,73]. Interestingly, while rzfGdf9 reduced amh expression in cultured ovarian fragments, it increased the expression of inhbaa and inhbb, but had no effect on inhibin alpha subunit (inha) and follistatin (fst) (Figure 8).
Adult zebrafish are gonochoristic with ovary and testis present in different individuals; however, the juvenile zebrafish are considered hermaphroditic because they developed ovary-like gonad or juvenile ovary first followed by differentiation into true ovary and testis at later stage [13,17,74]. The juvenile ovary is characterized by meiotic onset in all individuals and appearance of PO in most if not all individuals. The gonadal differentiation starts around 20–25 dpf and completes around 30–45 dpf; however, large variation exists among different individuals [13,16,17,59]. During sex differentiation, three types of gonads could be observed, i.e., ovary, transforming gonad (or early differentiated testis), and testis [16,17]. In this study, transforming gonads started to appear around 25 dpf and the testis with spermatogenic cells such as spermatocytes or even spermatids could be observed in some individuals after 30 dpf; however, large individual variation existed and the gonads of some individuals had not completed their differentiation even at 40 dpf.
Although no single primary sex-determining gene has been identified in the zebrafish and its sex determination is believed to be polygenic , a variety of genes have been implicated in controlling its gonadal differentiation, including cyp19a1a, amh, dmrt1, sox9a/b, ar, figla, and members of Wnt signaling pathway such as rspo1 [26,75–79]. Most of these genes are expressed in the somatic cells, supporting the view that the gonadal somatic cells play an important or determining role in controlling sex differentiation. Interestingly, an increasing body of evidence in recent years also points to the importance of germ cells in the event [43,46]. The mechanisms underlying germ cell signaling toward the somatic cells are unknown; however, this issue has attracted attention to the roles of oocyte-derived secreted factors. Among the best-characterized oocyte-derived factors is GDF9 (Gdf9/gdf9).
As an oocyte-specific growth factor, GDF9 has been well documented to play important roles in influencing the functions of the surrounding somatic follicle cells in mammals [80,81], supporting the notion that the oocyte as the germ cell functions as a signaling center in the follicle to orchestrate folliculogenesis. It increases the production of other growth factors and steroids by the surrounding follicle cells [80,81], and maintains the expression of FSH receptor in the follicle so as to act as an anti-apoptotic factor suppressing granulosa cell apoptosis and follicular atresia . In the zebrafish, apoptosis is considered a major mechanism responsible for degeneration of the oocyte-like germ cells in juvenile ovary, therefore paving a way for testis development. This is supported by the evidence that apoptotic signaling was stronger in the transforming gonads compared to that in the developed ovary . Whether the oocyte-secreted Gdf9 has protective effect to reduce apoptosis during gonadal differentiation will be an interesting issue to explore in the zebrafish.
In fish models, we previously reported that gdf9 was also specifically expressed in the oocytes of the zebrafish, and its expression level was the highest in early follicles. After fertilization, gdf9 transcripts were still detectable in the embryos as maternal mRNA until gastrulation stage . In this study, we showed that the expression of gdf9 started in the gonads before sex differentiation at morphological level and its level increased substantially at 16 dpf. Interestingly, although it was obvious, the increase was not statistically significant due to large data variation. Different from the analysis at later stages, which involved categorization of samples based on histological characterization, the samples from 13 to 20 dpf were pooled without histological classification. The high data variation implies that although gonadal differentiation was not visible at histological level from 15 to 20 dpf, the process might have already started at molecular level. The increase of gdf9 expression, which was not accompanied by cyp19a1a and amh, might be associated with the hermaphroditic phase of gonadal development (juvenile ovary) with oocyte-like germ cells, and different levels of commitment toward femaleness as suggested by Wang et al.  could contribute to the high data variation.
During gonadal differentiation, the expression of gdf9 showed an obvious sexual dimorphic pattern with the level much higher in the differentiated ovary than testis, which was similar to that of cyp19a1a but reciprocal to amh expression. The expression of amh was much higher in the testis than that in the ovary. Our experiment with rzfGdf9 showed that the oocyte-derived Gdf9 suppressed amh expression in vitro, and this effect could be part of the mechanism for the reciprocal patterns of gdf9 and amh expression during gonadal differentiation. The expression of amh in the gonads seems highly diverse among different species. In the medaka, the expression of amh shows no sexual dimorphism with similar levels in males and females . In the Japanese flounder, amh was not detectable in the ovary by RT-PCR or by northern blot analysis . Although gdf9 and cyp19a1a were co-expressed during gonadal differentiation, we did not observe significant response of cyp19a1a to rzfGdf9 despite a slight increase. The importance of Gdf9 in ovarian development was evidenced by VMO-induced male-skewed sex ratio; however, it remains unclear whether the change in sex ratio was due to disrupted signaling pathway towards ovary during gonadal differentiation or sex reversal after differentiation.
Interestingly, activin subunits (inhbaa and inhbb) whose expression increased significantly during follicle activation, especially inhbaa , responded positively to rzfGdf9 treatment. By comparison, no significant responses of inha and fst were observed. This agrees well with the expression profile of inha during folliculogenesis, which increases significantly at late stages before oocyte maturation and ovulation . As for fst, it is abundantly expressed in the oocyte as compared to the somatic follicle cells [71,72], making it less likely the target for Gdf9 in the ovarian fragments. This result suggests that in addition to amh, the oocyte-derived Gdf9 may also target other genes in the somatic follicle cells including the activin system, which represents a major paracrine signaling pathway from the follicle cells toward the oocyte . The stimulation of activin subunits by Gdf9 appears puzzling because they exhibited reciprocal expression patterns during follicle activation or PG-PV transition [50,73]. Our hypothesis for this regulatory relationship is that the decreased mRNA level of gdf9 at this transition may likely represent an increased mRNA turnover and protein synthesis, which stimulates activin output from the somatic follicle cells. The activin in turn acts on the oocyte to stimulate its growth and follicle development [71,72].
In addition to gdf9, cyp19a1a, and amh, we also examined the expression of some other genes that are potentially involved in gonadal differentiation. Interestingly, although these genes have been implicated in gonadal differentiation, favoring the development of either ovary (foxl2 and sox9b) or testis (dmrt1, sox9a, and dax1), none of them exhibited a clear-cut dimorphic expression pattern as shown by gdf9 and cyp19a1a. Despite this, some genes such as dmrt1, sox9a, and amh did show differential expression with higher expression levels in the testis, suggesting that they likely participate in controlling gonadal differentiation quantitatively in a dose-dependent manner.
In summary, we demonstrated sexually dimorphic expression patterns of gdf9, amh, and cyp19a1a during gonadal differentiation in larval and juvenile zebrafish. Gdf9 and cyp19a1a were co-expressed in the differentiating or differentiated gonads but in different cells, and their expression patterns were closely correlated with ovarian development, suggesting important roles for these two molecules in ovarian differentiation and development. Knocking down the expression of gdf9 and cyp19a1a using VMOs both caused a reduction of female ratio. Future studies using genomic knockout of gdf9 and cyp19a1a genes will provide further evidence for their roles and importance in gonadal development.
Supplementary data are available at BIOLRE online.
Supplemental Table S1. Primers used in RT-PCR and real-time qPCR* analysis. All the primers were synthesized by Integrated DNA Technologies, Inc. (Coralville, IA).
The authors have nothing to disclose.