Robust gdf9 and bmp15 expression in the oocytes of ovotestes through the Figla-independent pathway in the hermaphroditic black porgy, Acanthopagrus schlegelii

More than 1,500 fish species are hermaphroditic, but no hermaphroditic lineage appears to be evolutionarily ancient in fishes. Thus, whether more than one sex at a time was present during the evolutionary shift from gonochorism to hermaphroditism in fishes is an intriguing question. Ectopic oocytes were created in the ovotestes of protandrous black porgy via the withdrawal of estradiol (E2) administration. These ectopic oocytes reprogrammed the surrounding cells, which changed from Sertoli cells to follicle-like cells. We observed that gdf9 and bmp15 expression was localized in the primary oocytes and gradually decreased after oocytes entered a secondary oocyte stage. Robust expression of gdf9 and bmp15 in ectopic oocytes was associated with the surrounding Sertoli cells. However, blocking Cyp19a1a activity and increasing androgen levels did not stimulate the expression of gdf9 and bmp15. Thus, the robust gdf9 and bmp15 expression was not related to the inappropriate male microenvironment. Furthermore, in vitro data demonstrated that gdf9 and bmp15 were not downstream genes of Figla signaling. Therefore, our results suggest that there are two independent mechanisms, a Figla-dependent pathway and a Figla-independent pathway, by which oocyte-surrounding cells are altered from a male somatic fate to a female somatic fate. This functional switch might clarify how oocytes created an appropriate microenvironment during the transition from the ancient gonochorism to the present hermaphroditism.


Introduction
Most hermaphroditic fishes change sex during their lifetime in response to internal or environmental cues. These sequential sex changes in fishes include 3 primary forms: protogyny (female-to-male sex change), protandry (male-to-female sex change), and bi-directional sex change. However, no hermaphroditic lineage appears to be evolutionarily ancient in fishes [1]. Thus, an intriguing question is how the simultaneous presence of more than one sex occurred during the evolutionary shift from gonochorism to hermaphroditism in fish. Another in E2-induced female fish < 2 yr old, which can switch from a dominant ovary with a regressed testis to a dominant testis with a regressed ovary (under low E2 levels) after the withdrawal of E2 administration (induced femaleness, Fig 1C) [7,10]. Thus, we used this property to generate an ovotestis in the testicular tissue [9,29], allowing us to investigate how ectopic oocytes in an ovotestis could reprogram the surrounding cells from male fate to female fate. Our studies aimed to explain why both sexes can simultaneously exist in the gonad and produce an ovotestis under low plasma E2 levels during the transition from gonochorism to hermaphroditism.
Here, our experimental data indicated that gdf9 and bmp15 were most highly expressed in primary oocytes. Thus, the presence of robust gdf9 and bmp15 expression in the ectopic oocytes in ovotestes may help the oocytes create an appropriate microenvironment through reprogramming the surrounding cells from Sertoli cells to follicle-like cells. In addition, our in vitro data also indicated that the presence of robust gdf9 and bmp15 expression in ectopic oocytes did not correlate with the oocyte-specific gene factors in germline alpha (figla), whose expression is tightly related to the surrounding cell type: high figla expression when The fish are functional males in the first two spawning seasons, and then half of them change sex to female function in the third spawning season. Digonic gonads (testis and ovary in one gonad) are separated by connective tissue. The testis exists at various stages of the reproductive cycle until the fish changes to femaleness. Only primary oocytes are found before the sex change occurs. B) Active femaleness: The fish are functional females after sex change or induction by removal of the testis of the digonic gonad. C) Passive femaleness: Long-term estradiol (E2) administration results in a dominant ovary with the primary oocytes in the ovary and a regressed testis in < 2-yr-old fish, and no vitellogenic oocytes are observed in E2-induced sex change fish. This sex change is reversed after E2 administration is withdrawn. The gonadal development is classified into 10 different gonadal stages: undifferentiated gonad (stage 1), differentiated gonad (stage 2), developed testes with spermatogonia (stage 3), developed testes with spermatocytes (stage 4), developed testes with spermatozoa (stage 5), regressing testes with an active ovary (stage 6), dominant ovary with regressed testes (stage 7), regressing ovaries with active testes (stage 8), dominant ovary and oocytes with vitellogenesis (stage 9), and female function with rudimentary testes (stage 10). St. = gonadal stage, Vtg oocyte = vitellogenic oocyte, TT = testicular tissue, OT = ovarian tissue, undiff. = undifferentiated gonad, diff. = differentiated gonad.
surrounded by Sertoli cells and low figla expression with follicle-like cells. Taken together, follicular formation (reprogramming surrounding cells from Sertoli cells to follicle-like cells) in ectopic oocytes is not regulated solely in a Figla-dependent manner. This study provides a novel hypothesis for the transition from gonochorism to hermaphroditism: a weak male environment results in oocyte differentiation, and then oocytes can survive and reprogram the surrounding cells.

Animals and experimental design
The experimental fish were acclimated in seawater to the pond environment at the National Taiwan Ocean University culture station with a natural lighting system. All procedures and investigations were approved by the National Taiwan Ocean University Institutional Animal Care and Use Committee (approved number 104007) and were performed in accordance with standard guiding principles. The fish were anesthetized in 2-phenoxyethanol (0.5 ml/l water) during sampling. Experiment 1. The expression profiles of gdf9 and bmp15 during gonadal development and sex change. Fish were collected during maleness (< 2 yr old) and femaleness for genetic analysis and in situ hybridization. To examine ovarian growth in femaleness, we surgically removed the testicular part of the digonic gonad at stage 6 (regressing testes with an active ovary). Surgery was performed as described previously [10,29]. All fish survived. Gonad samples were collected at 2-mo intervals (n = 6-8) for histology and genetic analysis. Experiment 2. The characteristics of ectopically located oocytes in E2-induced ovotestes. We created an ovotestis by ectopically generating oocytes in the testicular region as described in our previous studies [9,29]. Fish (n = 500) were fed a diet containing E2 (6 mg/kg feed; Steraloids, Newport, USA), beginning at stage 1 (undifferentiated gonad; 2-mo-old) and continuing for 4-mo (stage 2, differentiating gonad), after which E2 administration was terminated. Gonad samples were collected monthly (n = 8-30) for histology. All fish survived. An antibody against the Sertoli cell marker Doublesex and mab-3-related transcription factor 1 (Dmrt1) was used to identify the sexual type of cells surrounding ectopic oocytes in the testes. Experiment 3. Effects of sex steroids on gdf9 and bmp15 gene expression. To determine the effects of male microenvironment on the expression of gdf9 and bmp15, fish at gonadal stage 7 (dominant ovary with regressed testes) were intraperitoneally (i.p.) injected with aromatase inhibitor (AI, 1,4,6-androstatriene-3,17-dione, 5 mg/kg body weight; Steraloids) and methyltestosterone (MT, 1 mg/kg body weight; Kingyoker, Taipei City, Taiwan) at days 0, 2, and 4 (n = 8). All fish survived. Gonads were collected 1 day after the third injection (day 5), and gene expression was analyzed. Experiment 4. Effects of Figla signaling on gdf9 and bmp15 gene expression. Our previous studies showed that prolonged and high Figla expression was only found in ectopic oocytes surrounded by Sertoli cells but not follicle-like cells in the ovotestes [9]. To examine the correlation between Figla signaling and gdf9 and bmp15 expression in ectopic oocytes, we developed an in vitro oocyte culture system to induce figla expression by using expression vectors. The vectors were delivered on day 1 (one day after the isolation of the suspended cells), and then samples were collected 2 days after transfection, and gene expression was analyzed.
Cloning of black porgy gdf9 and bmp15. Total RNA in black porgy ovary was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). Total RNA was reversely transcribed to the first-strand cDNA using Superscript III (Invitrogen) with oligo (dT)12-18 primers (Promega, Madison, WI, USA). To amplify the partial cDNA fragment of black porgy gdf9 and bmp15, PCR primers were designed based on conserved nucleotide sequence of European sea bass and fishes (Table 1). cDNA sequences encoding region of gdf9 and bmp15 were elongated by using the SMART 3'-rapid amplification of cDNA ends (RACE) kit (BD Biosciences Clonetech, Mountain View, CA, USA). Primers for RACE were listed in Table 1. The cDNA sequence of gdf9 (GenBank accession no. KY427738) and bmp15 (GenBank accession no. KY427737) was used to design the specific primer for RNA analysis and synthesize the RNA probe for in situ hybridization.
Gonadal histology, in situ hybridization, immunohistochemical staining and immunofluorescence staining Hematoxylin-eosin staining, in situ hybridization (ISH), immunohistochemical (IHC) staining and immunofluorescence (IF) staining were performed as described previously [10,30,31]. Fish gonads were fixed with 4% paraformaldehyde in PBS. cDNA fragments of gdf9 and bmp15 were used to synthesize the RNA probe for ISH (Table 1). For ISH, digoxigenin-11-UTP was used to label RNA with digoxigenine (Roche, Penzberg, Germany). Anti-sense digoxigenin-labeled antisense and sense probes were used to detect the localization of gdf9 and bmp15. The probe was developed overnight, using the standard protocol in our laboratory. Immunostaining was performed with preabsorbed alkaline phosphatase-conjugated sheep Table 1. Primers used in this study.

Gene
Orientation Sequence Analysis anti-digoxigenin antibody (Roche, Penzberg, Germany) at room temperature. Finally, the NBT/BCIP Detection System (Sigma-Aldrich, St. Louis, MO, USA) was used to detect gdf9 and bmp15 expression. The gonads (5-μm-thickness sections) were treated with HistoVT One (Nacalai Tesque, Kyoto, Japan) to expose the antigens of the target protein. The polyclonal Doublesex and mab-3-related transcription factor 1 (Dmrt1) antiserum was produced in white rabbits immunized against a peptide fragment of black porgy Dmrt1 (CEASSETPNFTVS-SIID). The antisera were prepared by Yao-Hong Biotechnology Inc. (New Taipei city, Taiwan). The specificity of the polyclonal anti-Dmrt1 was confirmed by Western blotting described in our previous study [31]. For IHC staining, each section was rehydrated in PBS and incubated with 3% H 2 O 2 in PBS. The section was then incubated with 5% nonfat milk powder for 30 min with anti-Dmrt1 (1:1500 dilution) overnight at 4˚C. This was followed by incubation with an appropriate biotinylated antibody (Vector, Burlingame, CA, USA). Color formation was amplified with an ABC kit (avidin-biotin, Vector; Burlingame, CA, USA) and DAB (3,3'-diaminobenzidine, Sigma). For IF staining, the section was then incubated with 5% nonfat milk powder for 30 min with anti-Factor in germline alpha (Figla; 1:400 dilution, [9]) overnight at 4˚C. Alexa Fluor secondary antibodies (Invitrogen) were used. The specificity of anti-Dmrt1 [31] and anti-Figla [9] has been evaluated in our previous studies. All staining was conducted with triplicate sections for each tissue (n = 3~5 fish in each group).

RNA analysis
Gonads were collected and homogenized in TRIzol reagent (Invitrogen). This homogenate was used for both RNA analysis. The extraction of total RNA and first-strand cDNA was performed according to the manufacturer's protocol. Total RNA extracted from the gonad of the representative fish was reversely transcribed to the first-strand cDNA using Superscript III (Invitrogen) with oligo (dT)15 primers (Promega). This first-strand cDNA was used for RT-PCR and quantitative real-time PCR analyses (qPCR). The number of PCR cycles was preliminarily tested and was in the range of the linear curve for the relationship between the number of cycles and the amount of PCR product. As an internal control, glyceraldehyde-3-phosphate dehydrogenase (gapdh; GenBank accession no. DQ399798) was used. Furthermore, Sertoli cells marker (dmrt1 and anti-mullerian hormone, amh) [29,31], follicle cells marker (forkhead box l2, foxl2 and P450 aronatase gonad form, cyp19a1a) [10], and oocytes marker (figla) [9] were used to analyze the gene expression patters in control fish and E2-treated fish. Specific primers for gdf9, bmp15, dmrt1 (GenBank accession no. AY323953), amh (GenBank accession no. GU256046), foxl2 (GenBank accession no. EU496493), cyp19a1a (GenBank accession no. AY273211), figla (GenBank accession no. EU496494), and gapdh are listed in Table 1. Gene quantification of standards, samples, and controls was conducted simultaneously by qPCR (GeneAmp 7500 Sequence Detection System; Applied Biosystems, Foster City, CA, USA) with SYBR green Master Mix (Applied Biosystems, Vilnius, Lithuania). The PCR specificity was confirmed by a single melting curve (at same temperature) in unknown samples and standards. The respective standard curve of log (transcript concentrations) vs CT (the calculated fractional cycle number at which the PCR-fluorescence product is detectable above a threshold) was obtained. The values detected from different amounts of plasmid DNA contained the fragment of the target gene (10 times serial dilution) of the representative samples in parallel with the respective standard curve. The correlations of the standard curve for the gene analyses were at least -0.999. The qPCR assay was conducted with duplicate repeats in each sample. All samples were normalized to gapdh, and the highest value (control value) of each gene was defined as one. The gapdh transcripts were not significantly different between treatments (data not shown).

In vitro transfection of oocytes
Ovarian cell preparation and oocyte selection were performed as described previously [29]. Ovaries were collected from testis-removal fish (femaleness) at stage 6 (regressing testes with an active ovary), stage 7, and stage 9 (dominant ovary with oocytes with vitellogenesis). The ovaries were dissociated using trypsin (0.36% trypsin, pH 8.2, 5% fetal bovine serum [FBS]) with gentle pipetting. Large clumps and large oocytes were eliminated from the dispersed ovarian cells and tissues through mesh with pore sizes of 70 μm and 40 μm. Filtered ovarian cells (< 40 μm) were seeded in culture medium (L-15 with 5% FBS, 50 U/ml penicillin, and 50 μg/ ml streptomycin). Based on the difference in the adhesive ability between somatic cells and oocytes, dispersed cells were incubated in wells overnight and then divided into 2 groups, adhesive cells and suspended cells. The cell types were confirmed based on gene expression of a germ line marker (vasa), oocyte marker (figla), and follicle cell marker (foxl2). The suspension cells were used for vector delivery. The pcDNA3.1(+) vector (Invitrogen) was used for figla expression. The control group was infected with a vector only. The vectors were delivered by Escort tm Transfection Reagent (Sigma-Aldrich) at day 1 (one day after suspended cell isolation) and the samples were then collected at day 3. The vector delivery ability for oocytes was confirmed by PCR (forward primer: 5'-TAGTAATCAATTACGGGGTCATTAG-3'; reverse primer: 5'-ACAAACTCCCATTGACGTCA-3').

Data analysis
The data were expressed as the mean ± standard deviation (SD). The values were subjected to analyze by one-way ANOVA, followed by a Student-Newman-Keuls multiple test, with P < 0.05 indicating a significant difference. Student's t-test was also conducted to determine significant differences (P < 0.05) between treatments.

Results
Ovarian gdf9 and bmp15 were localized in the oocytes  1A). Histological data (H&E staining) showed that the fish had a digonic gonad (Fig 2A-2C). The ovary was the dominant part of the digonic gonad during the non-spawning season (stage 7; Fig 2A). During the subsequent pre-spawning season, the ovary regressed and the testis grew (stage 4; Fig 2B). During the second spawning season, the testis was the dominant part and functioned as male during the second spawning season (stage 5; Fig 2C). ISH with antisense probes revealed the strong gdf9 and bmp15 expression in the ovary compared to the testis ( Fig  2D-2I). The expression of gdf9 and bmp15 was localized in the primary oocytes (Fig 2D-2I).
In the testis, gdf9 and bmp15 signals were low or absent at all gonadal stages (stages 4, 5, and 7; from non-spawning season to spawning season) (Fig 2D-2I). No signal was observed from sense probes for gdf9 (S1A-S1C Fig) and bmp15 (S1D-S1F Fig). Furthermore, qPCR data confirmed that gdf9 and bmp15 were highly expressed in the ovarian tissue compared to the testis tissue in the digonic gonad at stage 7 ( Fig 2J).

Differential expression of gdf9 and bmp15 in primary oocytes and vitellogenic oocytes
In the third reproductive cycle, half the fish were maintained at maleness (only primary oocytes were observed in the ovary), and the others underwent sex change (entering femaleness) with various stages of oocytes observed in the ovary during pre-spawning season ( Fig  1A). ISH staining revealed a slight expression of gdf9 and bmp15 in early-stage primary oocytes (small oocyte size), and then gdf9 and bmp15 expression gradually increased in late primary oocytes (large oocyte size) (Fig 3A and 3B). In contrast, gdf9 and bmp15 expression was low or absent in the vitellogenic oocytes (Fig 3A and 3B). No signal was observed from sense probes for gdf9 (S1G Fig) and bmp15 (S1H Fig).
To examine the gdf9 and bmp15 expression patterns in the ovary, testis-removal-induced active femaleness was used in >1-yr-old fish (Fig 1B). We surgically removed the testicular part of the digonic gonad in April (stage 6, post-spawning season). Histological data (H&E staining) showed that the surgery was successful and the fish only had an ovary. After the surgery, the primary oocytes gradually grew and then entered the vitellogenic oocyte stage in December (stage 9, pre-spawning season). Advanced oocyte diameters were 32.0 ± 1.21 μm in June (stage 6), 44.7 ± 0.08 μm in August (stage 7), 94.4 ± 8.94 μm in October (stage 7), and 132.0 ± 17.21 μm in December (stage 9). Similar to the ISH results (Fig 3A and 3B), qPCR data confirmed that the gdf9 and bmp15 transcripts were highly expressed in late primary oocytes (stage 7; October), but their expression was significantly lower in vitellogenic oocytes (stage 9; December) ( Fig 3C). Taken together, these results showed that both gdf9 and bmp15 were specifically localized in oocytes and were expressed at high levels during primary oocyte growth.  Fig 1A), stage 4 (Fig 1B), and stage 5 (Fig 1C) was performed. gdf9 and bmp15 mRNA expression was detected by in situ hybridization (ISH) (D-I) and qPCR (J). Both gdf9 (D-F) and bmp15 (G-I) were localized in primary oocytes, and no ISH signal was observed in testes in the second reproductive cycle. qPCR data confirmed that gdf9 and bmp15 were expressed at higher levels in the ovary than in the testes (J). An asterisk indicates a significant difference between the testes and ovary (P < 0.05). TT, testicular tissue; OT, ovarian tissue; CT, connective tissue; RTT, regressed testicular tissue.

Oocyte-surrounding cells transformed from Sertoli cells to follicle-like cells in E2-induced ovotestes expression
We created an ovotestis by ectopically generating oocytes in the testicular region through exogenous E2 administration as described in our previous studies [9,29] (Fig 1C). In previous studies, ectopic oocytes were surrounded by two types of somatic cells: Sertoli cells (Dmrt1 positive and Cyp19a1a negative) and follicle-like cells (Dmrt1 negative and Cyp19a1a positive) [9,29]. To identify the cell types surrounding the ectopic oocytes in the ovotestes, we used a specific antibody for Dmrt1 (Sertoli cell marker) to identify the cell types (Dmrt1-positive for Sertoli cells and Dmrt1-negative for follicle-like cells) by IHC (Fig 4A-4D). IHC staining for Dmrt1 in serial sections revealed that small primary oocytes were surrounded by a large proportion of Dmrt1-positive cells (Sertoli cells) (Fig 4A-4D). In contrast, only a small proportion of Dmrt1-positive cells (considered Sertoli cells) were present around the large primary oocytes in the ovotestes (Fig 4A-4D), and the majority of the surrounding somatic cells (Dmrt1-negative) were considered follicle-like cells.
We further examined the reprogramming of oocyte-surrounding cells during oocyte growth. Our results from serial ovarian sections for various sizes of primary oocytes (30 ectopic oocytes in 5 ovotestes) revealed that the number of oocyte-surrounding cells increased during oocyte growth ( Fig 4E). However, there were fewer Dmrt1-positive cells (Sertoli cells) around large primary oocytes than around small primary oocytes (Fig 4E). These results revealed that oocyte-surrounding cells were transformed from Sertoli cells to follicle-like cells during oocyte growth in the ovotestes (Fig 4F). Taken together, factors released by ectopic oocytes may stimulate transformation of the surrounding cells from Sertoli cells to follicle-like cells.

Robust presence of gdf9 and bmp15 in ectopic oocytes in the ovotestes
Both male and female germ cells were observed in the ovotestes (Fig 5A-5F). ISH staining showed that gdf9 and bmp15 were localized in the primary oocytes in the ovary and ovotestes (Fig 5A-5F). No signal was observed from sense probes for gdf9 (S1I Fig) and bmp15 (S1J Fig). Interestingly, ectopic oocytes in the ovotestes showed stronger expression of gdf9 (Fig 5B) than the normal oocytes in the ovary (Fig 5C). Similar to the gdf9 expression pattern, bmp15 also Robust expression of gdf9 and bmp15 in the ectopic oocytes showed more robust expression in ectopic oocytes in the ovotestes (Fig 5E) than in the normal oocytes in the ovary (Fig 5F). These data revealed that the differential expression of gdf9 and bmp15 between the ectopic oocytes (in the ovotestes) and normal oocytes (in the ovarian tissue) may play a role in reprogramming the cells surrounding the oocytes.
According to qPCR data, the relative expression order of oocyte-expressed genes (figla, bmp15 and gdf9) was as follows: ovary (digonic gonad in status 7) > ovotestes > testes (digonic gonad in status 4) (S3 Fig). The presence of ectopic oocytes in the ovotestes caused the increased oocyte-expressed genes expression as compared to the testes. In contrast, the ovarian tissue of ectopic oocytes in the ovotestes was only a small portion in the ovotestes as compared to the testicular tissue; therefore, the ovary (the dignoic gonad in status 7) had a much higher oocyte-expressed genes expression than ovotestes (S3 Fig). To further study the potential candidate genes in testis related to the presence of ectopic oocytes, we analyzed the gene expression patterns in testis of control fish and in regenerated testis (ovotestis; testis including ectopic oocytes) of E2-terminated fish. According to qPCR data, oocyte-expressed genes (figla, gdf9, and bmp15) had higher expression in the ovotestis of E2-terminated fish compared with in testis of control fish (S4 Fig). However, Sertoli cell markers (dmrt1 and amh) and follicle cell markers (foxl2 and cyp19a1a) had similar expression patterns in both groups (S4 Fig). Thus, our results revealed that Sertoli cell markers (dmrt1 and amh) and follicle cell markers (foxl2 and cyp19a1a) were not changed by the presence of ectopic oocytes in the ovotestis (testis-like) environment compared with the ovarian environment. The data further support the specific expression profiles of figla, gdf9 and bmp15 in the ectopic oocytes of ovotestes. Robust expression of gdf9 and bmp15 in the ectopic oocytes Sex steroids did not stimulate gdf9 and bmp15 expression To further study the correlation between high gdf9 and bmp15 expression in ectopic oocytes and the testicular environment, we used AI (to reduce E2 levels) and MT (to elevate androgen levels) to determine the effects of sex steroids on gene expression. Fish at gonadal stage 7 were injected (i.p.) with AI (aromatase inhibitor, 1,4,6-androstatriene-3,17-dione; 5 mg/kg of body weight) or MT (methyltestosterone, 1 mg/kg of body weight). qPCR results revealed that AI treatment had no effect on gdf9 and bmp15 expression in the ovary (Fig 6A). Furthermore, MT treatment significantly decreased the gdf9 and bmp15 expression (Fig 6B). Taken together, our results revealed that gdf9 and bmp15 expression was not increased by inappropriate sex steroid levels (estrogens or androgens) in the ovotestis (testis-like) environment compared with the ovarian environment.

Expression of gdf9 and bmp15 in oocytes was independent of Figla signaling
Our previous study showed prolonged expression of figla/Figla in ectopic oocytes in the ovotestes compared with normal oocytes in the ovary [9]. High Figla expression was correlated with ectopic oocytes being surrounded with Sertoli cells [9]. In the present study, IF staining confirmed that there was a prolonged Figla expression in ectopic oocytes in the ovotestes compared with normal oocytes in the ovarian tissue (Fig 7A).
To investigate the correlation between figla/Figla expression and gdf9 and bmp15 expression in ectopic oocytes in the ovotestes, we developed an in vitro culture system for oocytes to induce figla expression. Oocytes were isolated based on a previously reported difference in adhesive ability between somatic cells and oocytes [32]. qPCR data showed higher vasa (germ cell marker) and figla (oocyte marker) expression and lower foxl2 (follicle cell marker) expression in suspension cells than in the adhesive cells (data not shown). These data revealed that suspension cells are associated with oocytes. Successful vector delivery to oocytes was shown by PCR ( S2A Fig). qPCR data revealed that vector-induced figla expression varied with Fish were injected with AI (aromatase inhibitor, 5 mg/kg body weight) and MT (methyltestosterone, 1 mg/kg body weight) at day 0, 2, and 4 (n = 8). Gonads were collected 1 day after the third injection (day 5). The relative expression of gdf9 and bmp15 was analyzed by qPCR. Relative differences among control and treated groups were normalized to gapdh, and the highest value of the control group for each gene was defined as 100%. An asterisk indicates a significant difference between the control and treated groups (P < 0.05). https://doi.org/10.1371/journal.pone.0186991.g006 Robust expression of gdf9 and bmp15 in the ectopic oocytes gonadal stage (S2B Fig). Furthermore, qPCR data showed that the fold changes in gdf9 and bmp15 expression did not correlate with the changes in the figla expression levels (Fig 7B). Thus, the robust expression of gdf9 and bmp15 in ectopic oocytes in the ovotestes may have different roles than figla expression.

Discussion
This study focused on how an ectopic oocyte in the ovotestes could create a female microenvironment through reprogramming of the surrounding cells from Sertoli cells to follicle-like cells. We suggest that oocytes may regulate somatic fate through Figla, Gdf9 and Bmp15 signaling.
Ovarian gdf9 and bmp15 expression is associated with primary oocyte development in black porgy Using qPCR, our study revealed that gdf9 and bmp15 were highly expressed in the ovary compared to the testes. ISH results further indicated that gdf9 and bmp15 were localized in oocytes. In addition, both gdf9 and bmp15 were most highly expressed at the primary oocyte stage, with expression gradually decreasing following the secondary oocyte growth stage. Similar to zebrafish [25], Gibel carp [23], ricefield eel (Monopterus albus) [33], and European sea bass [20,21], gdf9/Gdf9 expression was high in oocytes at the primary oocyte stage and then gradually decreased during the secondary oocyte stage. In contrast, bmp15/Bmp15 expressions levels are variable among fish species. In zebrafish [24] and European seabass [20], bmp15/Bmp15 expression did not change during oocyte development. However, bmp15 expression is high in primary oocytes stage in Gibel carp [22]. These data suggest that gdf9/Gdf9 has conserved functions at the primary oocyte stage, but bmp15/Bmp15 does not. Taken together, gdf9 and bmp15 were highly expressed in oocytes at the primary oocyte stage, and this may be related to a species-specific function in black porgy.

Oocytes can differentiate under low plasma E2 levels
Previous work had shown that in black porgy, high endogenous E2 levels in plasma correlate with the sex change in the third reproductive cycle [4]. However, both ovary (primary oocyte stage) and testis can grow under the similarly low endogenous E2 levels in plasma [4,10,34]. Our present research and previous studies [7,34] also showed that E2 administration did not maintain a stable female stage and that a reverse sex change (female-to-male) occurred after E2 withdrawal. Furthermore, AI (aromatase inhibitor) administration in 4-mo-old black porgy caused the regression of testicular tissue and further development of ovarian tissue; therefore, the blockage of Cyp19a1a activity by AI administration did not arrest the early ovarian development or early oocyte growth in juvenile black porgy [34]. In medaka, XX females lacking the function of a female germline-determining gene (foxl3) develop ovaries filled with functional sperm and oocytes [35]. In mammals, germ cells can undergo sustained differentiation outside of the gonad, and in ectopic sites, they all differentiate into oocytes, even in males [36]. Taken together, these findings suggest that oocytes could be developed in an E2-independent manner in some fish species.

Oocytes are required for generating an appropriate female microenvironment
The present work and previous studies [9,29] have shown that in black porgy, ectopic oocytes in ovotestes can reprogram the sexual fate of surrounding cells from Sertoli cells to follicle-like cells. These results revealed that ectopic oocytes could create an ovarian environment appropriate for follicle cell differentiation. The present study further showed that ectopic oocyte-surrounding cells were gradually transformed from Sertoli cells to follicle-like cells during oocyte growth. In addition, in protogynous wrasse [14], protogynous grouper [15] and tilapia (Oreochromis niloticus) [37], chemically induced oocyte depletion results in follicle cell alteration and sexual fate change from female to male soma during the female-to-male sex change. Similar data also showed that in zebrafish, adult females can sexually revert to fertile males after the loss of most of the germ cells [11]. These data suggest that the mechanism for reprogramming the cells surrounding oocytes is possibly determined by the oocytes themselves and their paracrine factors. The oocyte itself plays an important role in the development of the surrounding somatic cells.
Both gdf9 and bmp15 may alter the sexual fate of oocyte-surrounding cells Unlike most vertebrate clades, which have a stable sexual fate after the primary sex determination and gonadal differentiation, hermaphroditic fish have secondary sex determination (sex change) during their life [1,28]. Secondary sex determination is precisely regulated by speciesspecific cues, including social factors (protogynous wrasse, protogynous anemone fish, and bidirectional sex change in gobiid fish), body size (protogynous grouper and protandrous ricefield eel), and age (protandrous sea bream) [1,28]. Thus, a hermaphroditic fish requires constant maintenance of its primary sexual fate through sex-appropriate gene expression before sex change [28].
The present study showed that oocyte-surrounding cells (Sertoli cells) were gradually replaced by follicle-like cells during oocyte growth in the ovotestes. Moreover, in the present study, gdf9 and bmp15 expression levels were higher in the ectopic oocytes in the ovotestes than in normal oocytes in the ovary. In mammals, mutation in Gdf9 results in decreased Cyp19a1 expression [38]. Oocyte-derived Gdf9 and Bmp15 expression have been shown to be essential for ovarian follicle growth [16,17]. Conversely, in zebrafish, all-male phenotypes are observed only with bmp15 mutations and not with gdf9 mutations [39]. These bmp15 mutant females had normal ovary development but oocytes arrested at the primary oocyte stage, after which the oocytes degraded, and the sexual fate switched to maleness [39]. Furthermore, these primary oocyte-surrounded granulosa cells did not express cyp19a1a [39]. Thus, bmp15 is suggested to have dual functions in zebrafish: maintenance of the female fate and the prevention of oocyte maturation [26,27]. Taken together, these differences in function between mammals and fishes may be due to fish having two paralogs of aromatase: cyp19a1a in gonad and cyp19a1b in brain [3,5]. Therefore, the robust oocyte-derived gdf9 and bmp15 expression may play important roles in altering the fate of oocyte-surrounding cells from male to female soma in the ovotestes in black porgy.
The expression of gdf9 and bmp15 in oocytes, not the levels of sex steroids, is related to the types of surrounding cells Our previous studies showed that oocyte-specific figla/Figla expression is associated with the types of surrounding cells [9]. Prolonged Figla expression was found in ectopic oocytes surrounded by Sertoli cells but not follicle-like cells in the ovotestes [9]. Furthermore, sex steroid treatment did not influence figla expression in the ovary [9]. The present study showed patterns of gdf9 and bmp15 expression similar that of figla expression. Robust gdf9 and bmp15 expression was observed in ectopic oocytes surrounded by Sertoli cells in the ovotestes. Neither gdf9 expression nor bmp15 expression was induced by AI or MT treatment. Therefore, this robust gdf9 and bmp15 expression in ectopic oocytes may be induced by oocytes surrounded by Sertoli cells but not in the inappropriate microenvironment (i.e., sex steroid levels) of the testis. We suggest that this unique expression pattern of gdf9 and bmp15 in ectopic oocytes is regulated by the types of surrounding cells.

Ectopic oocyte-surrounding cell reprogramming occurs by both Figladependent and Figla-independent means
Our previous study showed that figla/Figla expression was associated with ovarian differentiation and the reprogramming of ectopic oocyte-surrounding cells [9,10]. In mice, Figla is a key factor in coordinating the expression of ovary-specific genes [40,41]. Figla also represses the expression of male-associated genes during oogenesis in mice [40]. As in mammals, figla is important for ovarian differentiation in fish (zebrafish: [42]; medaka: [43]; tilapia: [44]). Thus, figla may play an important role in ovarian differentiation and follicle formation in black porgy. However, Figla is an oocyte-expressed transcription factor. Thus, this Figla-dependent mechanism needs oocyte-releasing paracrine signaling to alter the soma fate from male to female. The present study demonstrated that gdf9 and bmp15 were robustly expressed in ectopic oocytes (Fig 5 and Fig 8A). We also found that the robust figla, gdf9 and bmp15 expressions but not other examined sex-related genes (dmrt1, foxl2, cyp19a1a) were all specifically detected in the ovotestes (the testes containing ectopic oocytes surrounding Sertoli cells). However, the in vitro data in this study showed that the expression of gdf9 and bmp15 was not induced by high figla expression. In zebrafish, gdf9-deficient female had normally ovary [39]. In contrast, bmp15-deficient female alter sex from female to male during the juvenile stage [39]. Furthermore, Sertoli cell-expressed amh were significantly suppressed by recombinant Gdf9 in zebrafish [45]. Taken together, both gdf9 and bmp15 are suggested to involve in different mechanisms (the stimulation of the follicle cell characteristics and/or the suppression of the Sertoli cell characteristics) in black porgy.
Our data seems to support the positive relationship in the expression of gdf9 and bmp15 (Figs 2, 3, 5 and 6; S3 Fig). More studies with special approaches are needed to further understand the specific roles and expression of gdf9 and bmp15 in the reprogramming sexual fate of oocytes surrounding cells during oocyte development in the testes. Thus, our data suggest that the process of reprogramming the surrounding cells includes at least two parallel pathways: a Figla-dependent mechanism and a Figla-independent mechanism (Fig 8B).

Conclusions
We demonstrated that gdf9 and bmp15 were localized in primary oocytes and that their expression gradually decreased after oocytes entered the secondary oocyte stage. The data indicated that robust expression of gdf9 and bmp15 was associated with ectopic oocytes surrounded by Sertoli cells but not with the inappropriately male microenvironment. Our data also indicated that there are two independent pathways for the conversion of oocyte-surrounding cells from the male somatic fate to the female somatic fate. This functional switch might clarify how fish alter their sexual fate from one sex to the other during the transition from gonochorism to hermaphroditism.
Supporting information S1 Checklist. (PDF) https://doi.org/10.1371/journal.pone.0186991.g008 S1 Fig. References for in situ hybridization (ISH) in the different gonad stages. The references for ISH were confirmed by using sense probes for gdf9 and bmp15. No signal was observed with the sense probes for gdf9 (A-C) and bmp15 (D-F) at stage 7, stage 4, and stage 5 for Fig 2. No signal was observed with the sense probes for gdf9 (G) and bmp15 (H) at stage 9 for Fig 3. No signal was observed with the sense probes for gdf9 (I) and bmp15 (J) in testis for We created an ovotestis by ectopically inducing oocytes in the testicular region with estradiol (E2) administration and then E2 withdrawal. Testis (the digonic gonad in status 4, n = 8), E2-induced ovotestis (testis with ectopic oocytes, n = 4) and ovary (the digonic gonad in status 7, n = 6) were used for qPCR analysis. Oocyte-expressed genes (figla, bmp15 and gdf9) were analyxed by PCR. The gene value in the ovary was defined as 100%. Different small letters indicate significant difference (P <0.05). (TIFF)

S4 Fig. The relative genes expression in the testis and E2-induced ovotestis.
We created an ovotestis by ectopically inducing oocytes in the testicular region with estradiol (E2) administration and then E2 withdrawal. Normal testis (n = 8) and E2-induced ovotestis (n = 4) were used for RNA analysis. qPCR data confirmed that oocytes-expressed figla, gdf9, and bmp15 were expressed at higher levels in the ovotestes than in the testes. No difference of Sertoli cells marker (dmrt1 and amh) and follicle cells marker (foxl2 and cyp19a1a) were observed in the ovotestes than in the testes. An asterisk indicates a significant difference between the testes and ovotestis (P < 0.05).