Involvement of Transforming Growth Factor Beta Family Genes in Gonadal Differentiation in Japanese Eel, Anguilla japonica, According to Sex-Related Gene Expressions

The gonochoristic feature with environmental sex determination that occurs during the yellow stage in the eel provides an interesting model to investigate the mechanisms of gonadal development. We previously studied various sex-related genes during gonadal sex differentiation in Japanese eels. In the present study, the members of transforming growth factor beta (TGF-β) superfamily were investigated. Transcript levels of anti-Müllerian hormone, its receptor, gonadal soma-derived factor (amh, amhr2, and gsdf, respectively) measured by real-time polymerase chain reaction (qPCR) showed a strong sexual dimorphism. Transcripts were dominantly expressed in the testis, and their levels significantly increased with testicular differentiation. In contrast, the expressions of amh, amhr2, and gsdf transcripts were low in the ovary of E2-feminized female eels. In situ hybridization detected gsdf (but not amh) transcript signals in undifferentiated gonads. amh and gsdf signals were localized to Sertoli cells and had increased significantly with testicular differentiation. Weak gsdf and no amh signals were detected in early ovaries of E2-feminized female eels. Transcript levels of amh and gsdf (not amhr2) decreased during human chorionic gonadotropin (HCG)-induced spermatogenesis in males. This study suggests that amh, amhr2, and especially gsdf might be involved in the gene pathway regulating testicular differentiation of Japanese eels.


Introduction
The eel (Anguilla spp.) is a catadromous basal teleost with a remarkable migratory life cycle. Mature eels spawn in the ocean, and leptocephali larvae drift toward the coast and metamorphose into glass eels, which grow in continental habitats and then develop as yellow eels. After several years of growth phase in continental waters, the yellow eels undergo a prepubertal secondary metamorphosis and transform into silver eels [1]. The silver eels migrate downstream of rivers toward their spawning ground in the ocean to reproduce. Eels are gonochoristic species with gonadal differentiation occurring during the yellow eel phase [2]. Environmental factors, such as population density, temperature, pH, and social interactions, were proposed to prevail over genetic sex determination in eels [3,4]. Gonadal differentiation and development in eels were also suggested to be related to body size rather than age [2,[5][6][7]. The gonochoristic feature with an environmental sex determination pattern and gonadal differentiation occurring during the yellow stage in is expressed in Sertoli cells and is responsible for the regression of the Müllerian ducts and regulation of testicular differentiation of embryonic testes through binding to its receptor, AMHRII [53,54]. Although teleost fish lack a Müllerian duct, the amh orthologous gene was first identified from the testes of Japanese eel (A. japonica) and originally named eel spermatogenesis related substance 21 (esrs21). Amh transcripts were detected in Sertoli cells of immature testes but disappeared during spermatogenesis. Therefore, Amh was proposed as a spermatogenesis-preventing substance and was suggested to be involved in the regulation of germ cell development [55]. The sexually dimorphic with male specific expression pattern of amh was detected in several other teleosts during testicular differentiation (olive flounder, Paralichthys olivaceus [56], zebrafish [57], black porgy, Acanthopagrus schlegelii [58], and orange-spotted grouper, Epinephelus coioides [59]) or during early sexual maturation (Atlantic salmon parr, Salmo salar [60]). In contrast, the transcript level of amhr2 was not sexually dimorphic during gonad development in medaka [61] and tiger pufferfish [19]. Gsdf was only found in teleost fish, implying that it may possess unique functions. The gene gsdf was first identified from the rainbow trout (Oncorhynchus mykiss) embryos with genital ridges. Gsdf protein was exclusively expressed in the genital ridge somatic cells surrounding the primordial germ cells (PGC) during embryogenesis; moreover, Gsdf was particularly located in the Sertoli cells surrounding the type-A spermatogonia of testis and in granulosa cells of ovaries during gonadal development in rainbow trout [62]. Therefore, Gsdf was thought to play an indispensable role in PGC and spermatogonia proliferation in rainbow trout. The importance of gsdf participating in early germ cell development and male sex differentiation was also reported in other teleost fish, for instance, medaka [63,64], Nile tilapia [65], olive flounder [66], the Chinese tongue sole, Cynoglossus semilaevis [67], and spotted scat, Scatophagus argus [68]. Furthermore, gsdf was determined as a downstream gene of dmrt1 in the male pathway of Nile tilapia and spotted scat [68,69]. Gsdf was also suggested to be essential for testicular differentiation by inhibiting estrogen production in Nile tilapia [69].
Our previous studies on the Japanese eel reported the expression of some sex-specific genes (vasa, cyp19a1; female: figla, sox3; male: dmrt1, sox9a, foxl3a, foxl3b) during gonadal differentiation and development [6,70]. However, the molecular regulators involved in sex determination and differentiation in eels need further investigation. In this study, we analyzed the amh, amhr2, and gsdf expression patterns and the cellular distribution of amh and gsdf during testicular and E2-feminized ovarian differentiation and development of Japanese eels. This report elucidates the potential roles of the TGF-β superfamily members on gonadal differentiation and development in eels. We further propose a model of the expression profiles of sex-related genes during the process of gonadal differentiation in Japanese eels.

Animals
Four batches of eels were obtained for the experiments. The first batch (n = 1200, 300 elvers per tank) of elvers, with a mean body length of 9.54 ± 0.96 cm and a mean body weight of 0.88 ± 0.34 g, was used for E2-induced feminization. The second batch (n = 200, 100 elvers per tank) of elvers with a mean body length of 11.73 ± 0.31 cm and a mean body weight of 1.69 ± 0.12 g was used for cellular localization of gsdf in situ hybridization (ISH)/Vasa immunohistochemistry (IHC) studies. The third batch (3 female and 3 male eels) of yellow eels (2-3 years old; male eels, mean body length of 55.3 ± 1.2 cm, mean body weight of 305.0 ± 11.7 g; female eels, mean body length of 64.6 ± 1.5 cm, mean body weight of 409.3 ± 57.3 g) was used for cellular localization of gsdf ISH/Vasa IHC, and also for quantitative polymerase chain reaction (qPCR) of tissue distribution of gsdf, amh, and amhr2 transcripts. The fourth batch (n = 9) of silver eels (3-4-year-old male eels, mean body length of 66.9 ± 3.5 cm, body weight of 461.4 ± 38.0 g) was used for human chorionic gonadotrophin (HCG) treatment to induce spermatogenesis. All the fish were obtained from the commercial eel farms in Pingtung in southern Taiwan. The experimental fish were acclimated for at least one week for further experiments and analysis. The fish were cultured in indoor 2.5-ton tanks with aerated freshwater and then reared with natural light and temperature conditions at the culture station of the National Taiwan Ocean University (NTOU), Taiwan. The animal experiments were performed in an appropriate way in accordance with the animal experimentation procedure guidelines under the supervision of Drs. Chang, Wu, and Jeng under the authorization of the NTOU Institutional Animal Care and Use Committee (no. 104008).

E2-Induced Feminized Eels
Feminization of elvers was performed at the university culture station. Control (n = 600, 300 elvers per tank) and E2-treated fish (n = 600, 300 elvers per tank) were orally administered with the control diet and the E2-ontaining diet of 10 mg E2 (Sigma-Aldrich Corp., St. Louis, MO, USA) per kg of feed for 180 days (8 April 2015 to 13 October 2015). All the control elvers developed into male eels in our experimental conditions [6]. The gonadal samples were collected for qPCR of gsdf, amh, and amhr2 transcripts. The cellular localization of gsdf, amh, and figla ISH/Vasa IHC studies was also conducted.
Based on our previous histological examinations [6], the gonadal status in the control eels during testicular differentiation could be divided into undifferentiated (named S1, Figure 1A: only early germ cells were observed in the gonads), early, and late differentiating (early differentiating, named S2, Figure 1B: the gonads resembled ovaries and contained early germ cells, oogonia (OG), and primary oocytes (PO), late differentiating, named S3, Figure 1C: intersexual gonads exhibited a male-like testicular structure with OG, PO and degenerating oocytes (DO)), and differentiated (testis, named S4, Figure 1D: the testicular structures were formed). Based on our previous histological analysis [6], the E2-treated female eels could be divided into three stages: undifferentiated (named FS1, only early germ cells were observed in the gonads; Figure 1E), differentiating (named FS2, the gonads contained germ cells, OG and PO; Figure 1F), and differentiated eels (named FS3, OG, PO and pre-vitellogenic oocytes (PVO) were found in the gonads; Figure 1G).

HCG-Treated Male Eels
Nine 3-4-year-old male eels were randomly divided into two groups: (1) control (n = 5) and (2) HCG treatment (n = 4). Male eels were injected by HCG (Gona-5000 injection, China Chemical & Pharmaceutical Co. Ltd., Taipei, Taiwan) to induce testicular development. HCG treatment was conducted as described in our previous study [71]. Eels from the treatment group were given HCG (1 unit/g body weight/injection) consisting of one injection per week for six weeks. They were sacrificed four days after the final injection.

Sampling Procedures
Eels were anesthetized with 800 ppm of 2-phenoxyethanol before sacrifice. The biometric parameters of body weight and body length were measured. The gonadal tissue, or the body segment containing the gonadal tissue when the fish were too small to collect the gonadal tissue independently, was fixed overnight with 4% paraformaldehyde in phosphate-buffered saline (PBS, pH 7.4) for histological examinations and ISH/IHC analyses. For analyzing gene expressions, gonadal tissues were collected and stored at −80 • C until the quantitative real-time quantitative polymerase chain reaction (qPCR) analysis was performed.

Cloning of the Partial-Length amh, amhr2 and Full-Length gsdf cDNAs from Japanese Eel
The amh, amhr2, and gsdf cDNAs were cloned from the testis total RNA of Japanese eels by RT-PCR and rapid amplification of cDNA ends (RACE), following the manufacturer's protocol (SMARTer TM RACE cDNA Amplification Kit, Clontech, Tokyo, Japan). The specific primers used for cloning were designed according to the sequences from the NCBI nucleotide database and the Anguilla glass eel transcriptomic databases (http://molas.iis. sinica.edu.tw/4eels/index.html, accessed on 28 March 2017). After that, the resulting PCR products were cloned into the pGEM ® -T Easy vector and sequenced.

Sequence Alignment and Phylogenetic Analysis
The Gsdf sequence cloned for this study was blasted to the NCBI database and identified. This Gsdf sequence from Japanese eel was further aligned with other fishes using the multiple sequence alignment tool (CLUSTALW, http://www.genome.jp/tools-bin/ clustalw, accessed on 5 April 2021), and the identities were calculated. The presumed signal peptide sequences predicted by SignalP-5.0 (http://www.cbs.dtu.dk/services/SignalP-5. 0/, accessed on 25 March 2021) are indicated in Figure 2.

Histology
Eel body segments or gonadal tissues were dehydrated through a graded methanol series and then embedded in paraffin. The transverse sections (4 µm) were stained with hematoxylin and eosin (H&E) staining to observe the gonadal status, which was classified via light microscopic examination. Our previous report was used as a reference to identify the characterization of gonadal histological structure [6].

Dual ISH and IHC
ISH and IHC staining were performed on the same slides to determine cellular localization and distribution of various genes. The methods from our previous study were modified slightly and followed [71]. IHC and specific antibody were utilized to label the Vasa protein, which was used as a germ cell marker based on our knowledge and previous study [6,43]. cDNA fragments of amh (1092 bp (base pairs), GenBank accession number: MZ905497), gsdf (964 bp, GenBank accession number: MZ905499), and figla (524 bp) were used to synthesize the RNA probe for ISH. Both sense and antisense riboprobes were synthesized with DIG RNA Labelling Mix (Roche, Indianapolis, IN) using T7 or SP6 RNA polymerases (Promega) by in vitro transcription. After deparaffinization with xylene, gonadal transverse paraffin sections (4 µm) were rehydrated through a graded series of ethanol (100-30%) and then in distilled water. Sections were then washed with 0.85% NaCl and PBS before post-fixation in 4% paraformaldehyde for 20 min. After washing in PBS, sections were treated for 5 min at 37 • C with proteinase K (2 mg/mL) diluted in PBS and then fixed for 15 min in 4% paraformaldehyde. Sections were treated with DNase I (5 U/mL) for 15 min. After washing in PBS, sections were rinsed twice in 2X standard saline citrate (SSC). Hybridization was performed at 68 • C overnight in a humidified chamber using 100 mL hybridization buffer (50% deionized formamide; 2X SSC; 5xDenhardt's solution; 50 mg/mL of yeast tRNA; 4 mM ethylenediaminetetraacetic acid [EDTA]; 2.5% dextran sulfate) containing the digoxigenin (DIG)-labeled probe (3 mg/mL). After hybridization, slides were washed in 2X SSC at 65 • C, 2X SSC/50% formamide at 65 • C, 0.2X SSC, and 0.1X SSC at room temperature. Slides were washed in 100 mM Tris-HCl (pH 7.5) containing 150 mM NaCl for 10 min and then washed in the same buffer but containing 0.1% Triton and 0.5% skim milk powder, before being incubated overnight at room temperatures with anti-digoxigenin alkaline phosphatase Fab fragments (1:2000; Roche Pharma, Boulogne-Billancourt, France). The next day, slides were incubated with 2-hydroxy-3naphtoic acid-2 -phenylanilide phosphate (HNPP)/Fast Red detection kit (Roche Pharma) according to manufacturer's instructions. IHC steps were applied after ISH, first washing in PBS and sections were incubated with the guinea pig polyclonal antibodies against eel Vasa (1:1000). Sections were then immersed in the Alexa fluor goat anti-guinea pig (1:150; 488-conjugated; Invitrogen, Waltham, MA, USA) antibody. After washing away any excess antibodies, the slides were mounted with Vectashield mounting medium containing 4 ,6-diamidino-2-phenylindole (DAPI) from Vector Laboratories (Burlingame, CA, USA) that visualize the cell nuclei. An epifluorescence microscope (Olympus Provis) equipped with a DP71 digital camera was used to examine and photograph sections. Digital images were processed with the Olympus Analysis Cell software (Olympus, Tokyo, Japan).

Quantification of Gene Transcripts Using Real-Time Quantitative PCR
The qPCR analyses were conducted using the first-strand cDNA and the methods described in our previous study [6]. Primers for QPCR analysis were designed according to Primer Express version 3.0 (Applied Biosystems, Foster City, CA) and followed the design guideline ( Table 1). The gene expression of Japanese eel amh, amhr2, gsdf, and elongation factor 1 alpha (ef1a) transcripts were analyzed. The cDNA fragments of each gene were cloned and used as standards and an internal control gene for qPCR analyses (amh, 1092 bp; amhr2, 1198 bp, GenBank accession number: MZ905498; gsdf, 964 bp; and ef1a, 1311 bp). Gene-specific primers were designed for qPCR and are described in Table 1. Quantification of gene expression in standards (plasmids with cDNA sequence) and samples was conducted concurrently using a 7300 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) with SYBR green I as a dsDNA minor-groove binding dye. Non-template controls were performed in conjunction with every application, and no amplification was found. The melting curve patterns indicated the amplification of a single amplicon for each gene. The slopes of the respective standard and sample curves of the log cDNA concentrations versus Ct (the calculated fractional cycle number at which the PCR-fluorescence product is detectable above a threshold) were −3.3-(−)3.5, indicating an amplification efficiency of 100% to 90%. Gene expression was analyzed using the 2 −∆∆Ct method while calibrating with an internal control ef1a. The highest transcript value in each comparison was defined as 1, and the relative gene expressions are presented. Table 1. Specific primers used for real-time quantitative polymerase chain reaction (PCR) analyses.

Gene
Sequences Amplicon Size

Data Analysis
Transcript levels are presented as means ± standard error of mean (SEM). The values were analyzed using Student's t-test to evaluate significant differences (p < 0.05) between groups. All statistical analyses were performed using SPSS10.0 (SPSS Inc., Chicago, IL, USA) for Windows.

Sequence Analysis of Japanese Eel gsdf
The gsdf sequence with partial 5 UTR (untranslated region) of 292 bp and an open reading frame (ORF) of 669-bp that encoded 223 amino acids was obtained by cloning from testis total RNA ( Figure S1). The signal peptide, TGF-β domain, and seven conserved cysteine residues are indicated in Figure 2. The ORF of gsdf was compared to the Gsdf sequence from GenBank (accession number: BBI76498). The identity of Japanese eel Gsdf between our cloned sequence (accession number: MZ905499) and BBI76498 was 99.10%, indicating 221 identical residues of 223 amino acids and two dissimilar residues in the precursor region. Additionally, this cloned ORF of Japanese eel was compared to European eel Gsdf (GenBank accession number: XP35245734). The identity between the two sequences was 98.65% (220/223) with a three-residue difference in the precursor region.

Phylogenetic Analysis of Japanese Eel Gsdf
Phylogenetic analysis was used to estimate the evolutionary relationships between the Gsdf sequences of Japanese eel and TGF-β superfamily sequences of various vertebrates. Based on an alignment of amino acid sequences ( Figure S2) and taking the NGF domain sequences determined in Bdnf as an outgroup, a phylogenetic tree was constructed using the neighbor-joining method with 1000 bootstrapping replicates in MEGA5.05 software [72].
As shown in Figure 3, the TGF-β domain sequences clustered into five major clades of Amh, Gdf5, Gsdf, Inha, and Tgfb1 sequences, respectively. The NGF domain sequences of Bdnf clustered into one clade as an outgroup. These clades were well supported by 74% to 98% bootstrap values. The clade of Gsdf sequences (96% bootstrap value) encompassed all teleost Gsdf sequences, including the cloned Gsdf sequence from Japanese eel, as illustrated. The phylogenetic analyses indicated that our cloned sequence of Japanese eel, just as the BBI76498 sequence from Japanese eel, is a Gsdf sequence and one member of TGF-β superfamily.

Tissue Distribution of amh, amhr2 and gsdf in Yellow Eels
Various tissues from yellow male eels (n = 3) and ovaries from yellow female eels (n = 3) were used to analyze amh, amhr2, and gsdf transcripts in the Japanese eel. The transcripts of amh and amhr2 ( Figure 4A,B, respectively) were predominantly expressed in the testis, whereas slight expression of these transcripts were observed in the ovaries. The expressions of amh transcripts were also expressed in the pancreas and at low levels in other extragonadal tissues, including gills and different parts of the brain. Low levels of amhr2 transcripts were also detected in extragonadal tissues ( Figure 4A,B). Gsdf transcripts were exclusively expressed in the gonads, in which much higher levels in the testis rather than in the ovaries were found ( Figure 4C).

Expression of amh, amhr2, and gsdf in the Gonads during Sex Differentiation in Eels
The transcript levels of amh, amhr2, and gsdf in the gonads were analyzed using qPCR to further elucidate gene expression profiles during sex differentiation in Japanese eels. The expressions of amh transcripts significantly increased in differentiating testis (S3) and differentiated testis (S4) compared to undifferentiated (S1) and early differentiating testis (S2) ( Figure 5A). Amhr2 transcripts significantly increased in differentiated testis (S4) compared to early differentiating testis (S2) ( Figure 5B). The expressions of gsdf transcripts significantly increased in differentiating testis (S3) compared to undifferentiated gonads ( Figure 5C). Furthermore, amh, amhr2, and gsdf transcripts demonstrated a strong sex-dimorphic expression ( Figure 5A-C). The transcripts of amh, amhr2, and gsdf were maintained at low levels throughout ovarian differentiation in E2-feminized Japanese eels compared to their strong increase in control males during testis differentiation ( Figure 5A-C). Low gsdf transcript levels decreased even further in the ovaries during differentiation (FS2, FS3) as compared to the undifferentiated stage (FS1), as shown in Figure 5C.

Expression of amh, amhr2, and gsdf in the Gonads during Sex Differentiation in Eels
The transcript levels of amh, amhr2, and gsdf in the gonads were analyzed using qPCR to further elucidate gene expression profiles during sex differentiation in Japanese eels. The expressions of amh transcripts significantly increased in differentiating testis (S3) and differentiated testis (S4) compared to undifferentiated (S1) and early differentiating testis (S2) ( Figure 5A). Amhr2 transcripts significantly increased in differentiated testis (S4) compared to early differentiating testis (S2) ( Figure 5B). The expressions of gsdf transcripts significantly increased in differentiating testis (S3) compared to undifferentiated gonads ( Figure 5C). Furthermore, amh, amhr2, and gsdf transcripts demonstrated a strong sex-dimorphic expression ( Figure 5A-C). The transcripts of amh, amhr2, and gsdf were maintained at low levels throughout ovarian differentiation in E2-feminized Japanese eels compared to their strong increase in control males during testis differentiation ( Figure 5A-C). Low gsdf transcript levels decreased even further in the ovaries during differentiation (FS2, FS3) as compared to the undifferentiated stage (FS1), as shown in Figure 5C.

Expression of amh, amhr2 and gsdf Genes in the Gonads during Eel Spermatogenesis
Similar to the previous studies [43,73], HCG was shown to induce testis development and spermatogenesis in eels. The histological analyses showed that the testis sections of control eels only contained spermatogonia A and B. In contrast, the testis of HCG-treated eels contained spermatogonia A and B, spermatocytes, spermatids, and spermatozoa (data not shown). The gonadosomatic index (GSI %) was 0.06 ± 0.02% in the control group and 3.58 ± 1.82% in the HCG group. The amh and gsdf transcript levels in testis significantly decreased in the HCG-treated eels compared to the control eels ( Figure 6A,C), while no significant difference was observed for amhr2 transcript levels ( Figure 6B).

Cellular Localization of the gsdf Transcripts in the Testis of Yellow Eels
Yellow male eels have differentiated testis, and the morphology of the cells in the gonad allows recognition of different cell types. We used yellow eel testis sections to determine the gsdf expressing cells. Our ISH results showed the gsdf transcript signals were specifically expressed in the Sertoli cells surrounding the spermatogonia (Figure 7).

Cellular Localization of the amh and gsdf Transcripts during Testicular Differentiation in Eels
Fuzzy and almost undetectable amh signals were found in the gonad at undifferentiated S1 stage; however, amh transcript signals were obviously observed in the gonads in the S2-S4 stages. The amh signals were detected in the Vasa-negative cells; moreover, the signals were specifically expressed in somatic cells surrounding the spermatogonia in late differentiating (S3 stage) and differentiated testis (S4 stage), and these somatic cells could be identified as Sertoli cells due to their position (Figure 8).  Small and capital letters indicate significant differences (p < 0.05) among various gonadal status for the control groups and E2-induced feminized Japanese eels, respectively. Asterisks indicate significant differences (* p < 0.05) between the control eels and E2-induced feminized Japanese eels. For the control eels, S1, undifferentiated ef1a was used as an internal control gene. Data were expressed as means ± SEM. The number in the figure showed the fish number in each group. Small and capital letters indicate significant differences (p < 0.05) among various gonadal status for the control groups and E2-induced feminized Japanese eels, respectively. Asterisks indicate significant differences (* p < 0.05) between the control eels and E2-induced feminized Japanese eels. For the control eels, S1, undifferentiated gonad; S2, early differentiating gonad; S3, late differentiating gonad; S4, differentiated testis. For the E2-induced eels, FS1, undifferentiated gonad (E2 treated group); FS2, differentiating ovary; FS3, differentiated ovary. Figure 6. Profiles of (A) amh, (B) amhr2, and (C) gsdf expression in the control and HCGmale Japanese eels. Transcripts were measured by qPCR. ef1a was used as an internal contro Data were expressed as means ± SEM. Asterisks indicate significant differences (* p < 0.05) be the control male eels and HCG-treated male Japanese eels.

Cellular Localization of the amh and gsdf Transcripts in Eel Gonads
3.6.1. Cellular Localization of the gsdf Transcripts in the Testis of Yellow Eels

Cellular Localization of the amh and gsdf Transcripts during Testicular Diffe tion in Eels
Fuzzy and almost undetectable amh signals were found in the gonad at und tiated S1 stage; however, amh transcript signals were obviously observed in the in the S2-S4 stages. The amh signals were detected in the Vasa-negative cells; m the signals were specifically expressed in somatic cells surrounding the sperma in late differentiating (S3 stage) and differentiated testis (S4 stage), and these cells could be identified as Sertoli cells due to their position (Figure 8). Signals of gsdf transcript were observed in the Vasa-negative somatic cells in the gonads at undifferentiated stage (S1); furthermore, intensive hybridization signals were detected in somatic cells of the gonads at differentiating stages (S2 and S3), while gsdf transcript signals became weaker in differentiated testis (S4) as compared with differentiating stage (Figure 9), in agreement with qPCR results (Section 3.4).

Cellular Localization of the amh, gsdf and fig1a Transcripts during Ovarian Development in E2-Feminized Japanese Eels
Low levels of amh and gsdf transcripts were measured by qPCR in the ovary of E2feminized Japanese eels (Section 3.4). ISH did not show any amh signal in the gonads of E2-feminized Japanese eels (Figure 10), while gsdf transcript signals could be detected ( Figure 11A-D). Positive gsdf signals were observed in the somatic cells of the gonad at the FS1 stage ( Figure 11A); however, weak, but positive, gsdf signals were detected in the cytoplasmic of the primary oocytes and somatic cells surrounding the primary oocytes at the FS2 and FS3 stages ( Figure 11B,C).
To follow the development of ovarian tissue, fig1a and cyp19a1 were used. The figla transcripts significantly increased during ovarian development in E2-feminized eels as shown by qPCR in our previous study [6]. We further identified the cellular localization of figla in the ovaries by ISH in this study ( Figure 11E-H). No figla transcript signals were detected in the gonad at the FS1 stage ( Figure 11E), while strong figla signals were observed in the cytoplasmic of primary oocytes at the FS2 and FS3 stages ( Figure 11F,G). No Cyp19a1 signals could be detected by IHC at any stage of E2-induced feminized eels and control eels (data not shown) due to its low expression in gonads [6]. Very low cyp19a1 transcript levels were also detected by qPCR in the ovary of Japanese eels [6]. Signals of gsdf transcript were observed in the Vasa-negative somatic cells in the gonads at undifferentiated stage (S1); furthermore, intensive hybridization signals were detected in somatic cells of the gonads at differentiating stages (S2 and S3), while gsdf transcript signals became weaker in differentiated testis (S4) as compared with differe tiating stage (Figure 9), in agreement with qPCR results (Section 3.4).  of E2-feminized Japanese eels (Figure 10), while gsdf transcript signa ( Figure 11A-D). Positive gsdf signals were observed in the somatic c the FS1 stage ( Figure 11A); however, weak, but positive, gsdf signals cytoplasmic of the primary oocytes and somatic cells surrounding t at the FS2 and FS3 stages ( Figure 11B,C). To follow the development of ovarian tissue, fig1a and cyp19a1 transcripts significantly increased during ovarian development in E shown by qPCR in our previous study [6]. We further identified the of figla in the ovaries by ISH in this study ( Figure 11E-H). No fig

Characterization of the gsdf cDNA and Phylogeny
The Japanese eel Gsdf protein contained 223 amino acids, including the typical TGF-β domain with seven conserved cysteine residues, similar to any other Gsdf sequences from fishes that we presented in Figure 2. The cysteine residues created a rigid structure, a unique cysteine-knot motif, which is related to intra-chain disulfide bonds or dimerization [74,75]. Two gsdf genes were found in Luzon ricefish [20] and European sea bass [76]. However, only one gsdf gene was found in the eel genome (Lin et al., unpublished data) and in most published teleost fish genomes. The phylogenetic analysis of the Gsdf conserved TGF-β domain obviously revealed that eel Gsdf with other teleost fish Gsdf formed a clade, and the clade was separated from other members of TGF-β superfamily. To our knowledge, this report is the first on concerning Japanese eel gsdf gene characterization and phylogeny, although gsdf was previously used as a Sertoli cells marker in Japanese eel for mRNA measurement [77], and gsdf expression pattern in gonads was analyzed using qPCR [77,78].

amh, amhr2 and gsdf Are Dominantly Expressed in the Testis of Japanese Eel
The members of TGF-β superfamily (amh/amhr2/gsdf ), such as amhy in Patagonian pejerrey [18] and Nile tilapia [79,80], amhr2 in tiger pufferfish [19], and gsdfY in Luzon ricefish [20], have been implicated in playing crucial roles in gonadal determination. The genes of the TGF-β superfamily were shown to also be involved in gonadal differentiation and development in some teleost fish [68,81]. The amh of fish was first found in Japanese eel [55], and recent studies have shown that the expression levels of amh and gsdf are significantly higher in testes than in ovaries of E2-treated Japanese eels [78]. The amh and gsdf transcripts of European eels were shown to be significantly higher in intersexual and male gonads than in females [82]. However, the roles of amh, amhr2, and gsdf in gonadal differentiation and/or development were still unclear at that point. Similar to the most studied teleost fish, our present study in Japanese eel showed that amh and amhr2 transcripts were dominantly expressed in testis. However, very low expression levels of amh and amhr2 in the ovaries and extragonadal tissues were also detected. This result is agreement with data obtained from some other teleost fish, such as amh and amhr2 in Nile tilapia [83][84][85], amh in pejerrey [86], European sea bass [87], and Atlantic cod [88]. Due to the wide tissue distribution of amh in Nile tilapia, it has been proposed that amh may have multiple functions not restricted only to the gonads [81,85], which is also suggested according to our findings. In contrast, the expression levels of amh and amhr2 in medaka were not sexually dimorphic with no extragonadal expression, and medaka amh and amhr2 are thus suggested to be related to gonad formation and maintenance in both sexes [61]. Our study showed that gsdf expression in the Japanese eel was restricted to the gonads with a much higher expression in the testis than in the ovary. The same results were also found in rainbow trout [62], medaka [64], Nile tilapia [65], Chinese tongue sole [67], and spotted scat [68].

amh/amhr2 Might Play Important Roles in Testicular Differentiation in Japanese Eel
An increase in gonadal amh expression levels during male sex differentiation was reported in rainbow trout, Nile tilapia, pejerrey, black porgy, Atlantic cod, and zebrafish [81]. The expression levels of both dmrt1 and amh showed a rapid increase followed by a decrease in the expression of foxl2 and cyp19a1a during the key stage of the sex differentiation in temperature-masculinized XX females Nile tilapia; therefore, dmrt1 and/or amh have been proposed as the modulator(s) of foxl2 and/or cyp19a1a downregulation in temperatureinduced testis development in Nile tilapia [89]. Furthermore, higher expression of amh was found at male inducing temperatures than at female inducing temperatures during the gonadal differentiation period in pejerrey [90] and Southern flounder (Paralichthys lethostigma) [91]. Male zebrafish have ovary-like gonads before the ovary-to-testis transformation [57]. Low amh transcript signals in undifferentiated gonads of zebrafish larvae were detected by ISH, and the amh signals gradually increased during the ovary-to-testis transformation. Moreover, intensive signals of amh were detected in testis, while no signal was found in the ovary in differentiated juvenile zebrafish [57]. Our present study showed the gonadal morphology of early male differentiating gonad of Japanese eel was similar to that of juvenile zebrafish and exhibited an ovary-like gonad (Figure 1). An increase in amh transcript levels was shown during testicular differentiation in Japanese eel, similarly to zebrafish. Our results suggest an important role of amh in testicular differentiation of eels. In addition, Amh was suggested to regulate germ cell proliferation, a function that might be conserved in fish [81]. For example, amh was suggested to be required for germ cell proliferation during early gonadal differentiation in madaka [92]. Overexpression of Amh by amh-plasmid feeding could stimulate the spermatogonia proliferation and testicular development in undifferentiated, orange-spotted groupers [93]. In contrast, Amh could inhibit spermatogonial proliferation during spermatogenesis in Japanese eels [55].
Amh function via binding to AMHR2 was demonstrated in mammals [94]. Amhr2 and its ligand Amh have also been identified in some teleost fish [81], except in zebrafish, which possess the amh gene, but the amhr2 gene was not found in the genome [95]. Black porgy amhr2 transcripts concurrently increased with its ligand during testicular differentiation; moreover, both amh and amhr2 transcripts were expressed in the somatic cells surrounding the spermatogonia [58]. Double ISH showed that medaka amh and amhr2 were mainly colocalized in the gonadal somatic cells in both sexes during early gonadal sex differentiation, and the Amh/Amhr2 signals might be related to the regulation of germ cell proliferation and follicular development [92,96,97]. Furthermore, co-localization of amh and amhr2 revealed that amh acts through the autocrine/paracrine regulatory system [92,96,97]. In the present study, we found that the expression patterns of amhr2 were similar to its ligand during testicular differentiation; therefore, amhr2 might also play a vital role during testicular differentiation in Japanese eels. The amhr2 expressing cells in eel have not yet been identified. Therefore, further evaluation of the cellular localization of amhr2 in eel gonads will be the next important task.

gsdf Is Expressed in Sertoli Cells and Might Play an Indispensable Role in Testicular Differentiation in Japanese Eel
Gsdf was reported to play a critical role in male sex determination and/or differentiation in teleost fish; moreover, Gsdf might accomplish its function by inhibiting estrogen synthesis [67,69]. Gsdf expression in Sertoli cells has been indicated in many teleost fish, such as rainbow trout [62], Nile tilapia [65], medaka [63], olive flounder [66], Chinese tongue sole [67], spotted scat [68], Atlantic salmon [98], and orange-spotted grouper [99]. The gsdf of olive flounder was expressed in somatic cells prior to the occurrence of gonadal differentiation, and the expression then quickly increased when testis differentiation was initiated. Therefore, gsdf was proposed to play important roles on promoting testis differentiation and early germ cell development [66]. Furthermore, gsdf was distinctly and primarily expressed in XY undifferentiated gonads of Nile tilapia before any other testis-differentiation-related gene was detected; overexpression of gsdf in XX Nile tilapia demonstrated that gsdf was sufficient to induce the male sex differentiation pathway, and consequently, gsdf was suggested to play a critical role in male determination and/or differentiation of Nile tilapia [65]. Gsdf was shown to suppress the proliferation activities of oogonia in orange-spotted grouper [99]. We found gsdf -positive signals in the Sertoli cells surrounding the spermatogonia in yellow eels. This result in the eel agrees with other teleost fish as we described above. Although gsdf was used as a Sertoli cell marker for mRNA measurement using qPCR in Japanese eels [77], our study was the first report indicating the cellular localization of gsdf in the eel gonads. The present study showed that the expression levels of gsdf increased significantly during testicular differentiation of Japanese eels; moreover, high gsdf expression levels and strong transcript signals were present at differentiating stages (S2 and S3). Interestingly, the gsdf expression pattern was different from those of amh and amhr2 with the highest expression levels of gonadal amh and amhr2 being found at differentiated stage. Furthermore, the gsdf transcript signals could be distinctly observed in somatic cells surrounding the germ cells in undifferentiated gonads, and the signals were detected earlier than amh in Japanese eels. Therefore, our finding in the Japanese eel reinforced the importance of gsdf articipation in male determination and/or differentiation in teleost fish.

The Expression of amh and gsdf but Not amhr2 Is Suppressed by Gonadotropin
Amh was first regarded as a spermatogenesis-preventing substance in Japanese eel [55]; furthermore, Amh was reported to have inhibiting effects on germ cell development and steroidogenesis in zebrafish [100] and medaka [97,101]. Therefore, downregulation of Amh might be necessary for the effect of gonadotrophins on maturation [81]. Previous studies showed the expression of amh decreased during spawning season in male teleost fish [60,102,103]. The expression of testicular amh was inhibited by HCG injection in black porgy [58] and European sea bass [104]. Our present data also showed the amh transcripts in testis significantly decreased in HCG-treated male eels in agreement with a previous report in the Japanese eel [55]. Curiously, the expression level of testis amhr2 was not significantly inhibited by HCG in the present study. The asynchronous expressions of amh and amhr2 in male eels indicate that gonadotropins might have different regulatory functions toward amh/amhr2.
A previous study indicated that gsdf expression levels were reduced progressively during spermatogenesis in yellowfin seabream [105]. Our study also showed the testis gsdf transcripts significantly decreased in HCG-treated male Japanese eels. However, the data concerning the regulation and role of gsdf during fish spermatogenesis are still limited and need further investigation.

The Expressions of amh, amhr2, and gsdf in the Ovary Were Inhibited in E2-Induced Feminized Japanese Eels
The robust male-biased expressions of amh, amhr2, and gsdf have been extensively reported, as these genes are considered pivotal factors in male sex determination and/or differentiation in teleost fish. Conversely, our data showed the transcripts of amh and amhr2 were maintained at low levels during ovarian differentiation and development in E2-induced feminized eels. A recent study on the Japanese eel also reported a similar result. The transcript levels of amh were lower in the undifferentiated gonads of E2-treated eels compared with controls (males) [78]. These results suggest that E2 inhibited amh/amhr2 involvement in testicular differentiation and induced the ovarian formation in eels. The administration of E2 to undifferentiated black porgy led to low expression of amh, and the expression of amh remained at low level in E2-induced ovarian tissue [58]. EE2 (17α-ethinylestradiol) concentration levels in the environment range from <0.20 ng/L to 34.00 ng/L in water, as previously reported [106]. Exposure of EE2 at environmentally relevant concentrations (0.5 and 5 ng/L) to zebrafish during early life had a suppressive effect on the expression of amh and caused subsequent disruption of male gonadal development [107]. However, an asynchronous expression of amh and amhr2 was shown in black porgy [58]. The expression levels of amhr2 significantly increased in the differentiated ovary of E2-and aromatase inhibitor-treated black porgy, and the amhr2 expression levels showed no differences between the ovary and control testis [58]. Therefore, amhr2 may be related to gonadal development in both sexes of black porgy [58].
Although gsdf transcripts/Gsdf proteins are dominantly expressed in Sertoli cells, gsdf /Gsdf were also found in the ovaries of teleost fish. A previous study indicated that rainbow trout Gsdf was present in ovarian granulosa cells [62]. In zebrafish, gsdf was located in somatic cells of bipotential gonads and ovarian granulosa cells [108,109]; moreover, the genes related to lipid metabolism, vitellogenesis, and steroid biosynthesis were downregulated in homozygous gsdf knockout mutant female zebrafish. Therefore, a role for gsdf in the regulation of ovarian follicle maturation and gene expressions for steroid biosynthesis and female fertility in zebrafish was proposed [109]. Gsdf was also found in the somatic cells surrounding the oogonia of the ovary in spotted scat, and possibly played a role in oogonia developmental regulation [68]. In addition, gsdf transcript signals were observed in the somatic cells and cytoplasm of oocytes in olive flounder; therefore, gsdf was proposed to participate in early germ cell development, such as proliferation and differentiation of oogonia in olive flounder [66]. Our study showed gsdf transcript signals in somatic cells of undifferentiated gonads of E2-ferminized eels, and faint signals were also detected in follicle cells and cytoplasm of oocytes of differentiating and differentiated ovary in E2-ferminized Japanese eels. Our qPCR data also indicated that gsdf transcripts remained at low levels and decreased significantly during ovarian differentiation/development in E2ferminized Japanese eels. Our results imply that eel gsdf might be involved in the regulation of early germ cell development in both sexes, and E2 could inhibit the expression of gsdf and promote the feminization of Japanese eels. The suppressive effects of estrogen on the gsdf expression were also shown in several teleost fish, and a previous study concerning Japanese eel also showed gsdf transcripts were significantly lower in the undifferentiated gonads of the E2-treated eels than in the controls (males) [78]. Exogenous E2 could reduce the gsdf expression levels in genetic male medaka during embryonic development [64]. The transcripts of Sertoli cell-specific gsdf of mature medaka significantly decreased, and the testis-ova was first found after 14 days of treatment with estradiol benzoate [63]. Conversely, exposure of embryos of Javafish medaka, Oryzias javanicus (ZZ/ZW), to EE2 caused significant expression of an estrogen up-regulated gene, Choriogenin-H; however, the expression of gsdf in ZZ Javafish medaka fry was not suppressed, and most of the XX fry did not sexually reverse [110].

Conclusions
During our preparation of this paper, a certain degree of similarity to part of our results in Japanese eels was published in 2021 [78]. However, our study still provided further valuable information. (1) The expressions of TGF-β superfamily genes in the gonads were conducted from elvers, juvenile eels, to yellow eels including undifferentiated stage, differentiating stage, differentiated stage, and even up to pre-pubertal stage in male and female. (2) Our data, for the first time, demonstrated the cellular localization of TGF-β superfamily transcripts in eel gonads, and the transcript signals were strongly expressed in males and weak in females. (3) This study was a part of our series studies of gonadal sex differentiation in Japanese eels. We provided integrated information (in addition to TGF-β superfamily) for sex related genes profiles during sex differentiation and further development of Japanese eels.
Gonochoristic feature with environmental sex determination and gonadal differentiation occurred quite late during the yellow stage in Japanese eels (body length around 20-40 cm) and provides a unique model to study the mechanism of gonadal differentiation. A diagram of gene expression during gonadal sex development is proposed as a basis to elucidate the pathways of gonadal differentiation of Japanese eels ( Figure 12). Genetic control for sex determination (the master regulator of male development) is highly diverse, while the downstream genetic pathways for sex differentiation is quite well conserved in vertebrates [8,21,111]. Sex-related genes were sexually dimorphic expressed in the testicular and ovarian tissues. Some sex-related genes (vasa, figla, sox3, cyp19a1, dmrt1, sox9a, foxl3a, foxl3b, foxl2a) were previously investigated during gonadal differentiation in Japanese eels [6,43,70,78]. We investigated the expression profiles of other sex-related genes according to the proposed model of gonadal sex development in Japanese eels ( Figure 12). According to gene expression, our present data support that amh, amhr2, and especially gsdf might be involved in the gene pathway regulating testicular differentiation in Japanese eels.