Disruption of dmrt1 rescues the all-male phenotype of the cyp19a1a mutant in zebrafish – a novel insight into the roles of aromatase/estrogens in gonadal differentiation and early folliculogenesis

Sex determination and differentiation are hot topics in developmental and reproductive biology. In mammals, sex differentiation is triggered by an upstream sex-determining factor Sry, which drives male development (Bachtrog et al., 2014; Cutting et al., 2013). Although the upstream sex determination signals vary in vertebrates, the downstream differentiation pathways tend to be more conserved (Devlin and Nagahama, 2002; Heule et al., 2014). With the development of new knockout technologies (Cong et al., 2013; Hwang et al., 2013), an increasing number of genes and their functions in gonadal differentiation have been assessed in different species, especially in zebrafish (Dranow et al., 2016; Lau et al., 2016; Lu et al., 2017; Nakamoto et al., 2018; Qin et al., 2018; Zhang et al., 2017, 2015). Unlike in mammals and some fish species, sex determination in domesticated zebrafish is considered to be a polygenic mechanism (Liew et al., 2012; Liew and Orban, 2014). The high plasticity of zebrafish gonadal development provides an excellent model for investigating the interactions between various regulatory factors.

Doublesex and mab-3-related transcription factor 1 (dmrt1) is a functionally important transcription factor in male differentiation across vertebrates (Eggers et al., 2014; Guo et al., 2005; Herpin and Schartl, 2011; Matson and Zarkower, 2012; Matsuda et al., 2002; Nanda et al., 2002; Smith et al., 2009). Recent knockout studies in zebrafish also verified that dmrt1 was necessary for zebrafish male differentiation (Lin et al., 2017; Webster et al., 2017). Aromatase, which is encoded by Cyp19 for the production of estrogens (Simpson et al., 1994), is well known to play a crucial role in fish sex differentiation (Nakamura et al., 2003, 1998; Paul-Prasanth et al., 2013; Takatsu et al., 2013). Knockout of the ovarian aromatase cyp19a1a but not the brain aromatase cyp19a1b in zebrafish resulted in an all-male phenotype (Dranow et al., 2016; Lau et al., 2016; Yin et al., 2017; Song et al., 2020).

The important roles of dmrt1 and cyp19a1a in sexual development have been extensively investigated, but the knowledge about their actions and, in particular, their interactions in gonadal differentiation is still limited. In the present study, we first confirmed the essential role of dmrt1 in driving male differentiation. Then we focused our attention on the potential interaction between the male-promoting dmrt1 and female-promoting cyp19a1a. Surprisingly, we found that, in the absence of the dmrt1 gene, the all-male phenotype of the cyp19a1a mutant was rescued with female follicles developing normally up to the previtellogenic (PV) stage (stage II) in double mutant fish (dmrt1^−/−;cyp19a1a^−/−). Our discoveries suggest that Cyp19a1a plays a role in determining ovarian differentiation by suppressing dmrt1 expression; however, the enzyme plays little role afterwards in early follicle development until the vitellogenic stage.

In this study, we selected a dmrt1 null mutant fish line with a 7 bp deletion (dmrt1^−7/−7) (registered with ZFIN as umo15) for phenotype analysis because it is supposed to induce a frameshift mutation that disrupts Dmrt1 protein synthesis (Fig. S1). The dmrt1^−/− mutants showed a significant female-biased phenotype (dmrt1^−/−, 88.8%±1.3; dmrt1^+/−, 42.6%±11.8; dmrt1^+/+, 50.8%±6.8; Fig. 1A) with normal folliculogenesis and fertility in females (Fig. S2). Additionally, all dmrt1^−/− males were infertile because of failure to induce female spawning and dysfunctional spermatogenesis (Fig. S3), which agrees well with recent similar studies in zebrafish (Lin et al., 2017; Webster et al., 2017) and the reported roles of DMRT1 in sex differentiation of other vertebrates (Matson et al., 2010, 2011; Smith et al., 2009). The female-biased sex ratio in dmrt1 mutant, together with the all-male phenotype of cyp19a1a mutants (Lau et al., 2016; Yin et al., 2017), suggest that cyp19a1a and dmrt1 may represent two major factors/pathways that drive female and male development, respectively. What will happen if both cyp19a1a and dmrt1 are lost? To answer this question, we went on to generate a double mutant of these two genes (dmrt1^−/−;cyp19a1a^−/−).

In our study, all the double mutants (dmrt1^−/−;cyp19a1a^−/−) exhibited masculinized secondary sexual characteristics, including the absence of genital papillae, a slimmer body and a brown anal fin (Fig. S4), making it difficult to distinguish their sex via appearance. In dmrt1^−/− mutants, the fish could be sexed based on apparent morphological features (Webster et al., 2017). The lack of female sexual characteristics in the double mutants (dmrt1^−/−;cyp19a1a^−/−) might be due to insufficient estrogens (plasma, 0.2±0.1 ng/µg; serum, 0.4±0.1 ng/ml; Fig. 1B), which are essential for female characteristics (Brion et al., 2004).

To better identify the sex of dmrt1^−/−;cyp19a1a^−/− fish, we introduced the transgenic line Tg(vas:EGFP)^zf45 into the double mutant line to generate Tg(vas:EGFP)dmrt1^−/−;cyp19a1a^−/− fish. According to previous reports from us and others (Krøvel and Olsen, 2004; Lau et al., 2016; Qin et al., 2018; Wang et al., 2007), the GFP signal is much higher in the ovary than in the testis and its intensity increases dramatically when the gonad differentiates into ovary. This feature allows live assessment of zebrafish sex during sex differentiation before sampling for histological confirmation. The GFP signals in control females (dmrt1^+/−;cyp19a1a^+/−) were strong at 100 dpf (days post-fertilization), in contrast to the weak signals in males. Surprisingly, some double mutant (dmrt1^−/−;cyp19a1a^−/−) individuals also showed strong GFP signals fully comparable with those in female controls (Fig. 1Ca′,c′), suggesting the formation of ovaries in the absence of cyp1a91a. Histological analysis confirmed the formation of well-developed ovaries containing many follicles (Fig. 2Bf,f′).

Analysis of sex ratios at 60, 75 and 100 dpf (Fig. 1D) showed that single cyp19a1a^−/− mutants were all males (100%) at all time points examined, whereas majority of the dmrt1^−/−;cyp19a1a^−/− double mutants were females at 60 dpf (12 females/15, 80% females), which was comparable with that of single dmrt1^−/− mutants (10 females/12, 83% females). Interestingly, at later times (75 and 100 dpf), the female ratio of the dmrt1^−/−;cyp19a1a^−/− double mutants decreased progressively (6 females/14, 43% females, 75 dpf; 3 females/23, 13% females, 100 dpf), while the ratio in dmrt1^−/− mutants remained unchanged over time (12 females/15, 80% females, 75 dpf; 11 females/13, 85% females, 100 dpf). These results clearly indicate that disruption of dmrt1 rescued the all-male phenotype of cyp19a1a^−/− mutant zebrafish. However, the females in the dmrt1^−/−;cyp19a1a^−/− double mutants could not be maintained indefinitely and the fish likely underwent sex reversal from females to males. Further evidence is shown in the following histological analysis.

To confirm the identity of female double mutants (dmrt1^−/−;cyp19a1a^−/−) revealed by GFP signal (see above), we further analyzed the same batch of fish at 75 dpf by histology. As we reported previously (Lau et al., 2016), all cyp19a1a single mutants (dmrt1^+/−;cyp19a1a^−/−) were males at 75 dpf (13 males/13, 0% females), in contrast to the control (dmrt1^+/−;cyp19a1a^+/−; 7 females/14, 50% females) and dmrt1 single mutants (dmrt1^−/−;cyp19a1a^+/−; 12 females/15, 80% females). Surprisingly, the dmrt1 and cyp19a1a double mutants (dmrt1^−/−;cyp19a1a^−/−) showed normal sex ratio (6 females/14, 43% females) comparable with that in the control group, and the females had well-developed ovaries (Fig. 2A). However, in contrast to the ovaries of control females that contained follicles of primary growth (PG), PV and early vitellogenic (EV) stages, the ovaries of the double mutants contained PG and PV follicles only, without any follicles entering vitellogenic growth, suggesting a blockade at the PV-EV transition (Fig. 2A,C). This result indicates clearly that all stages of early folliculogenesis occurred normally in double mutant ovaries, including PG follicle growth, PG-PV transition and development of full-sized PV follicles, and all these took place without the ovarian aromatase (cyp19a1a). This is surprising, as the ovarian aromatase Cyp19a1a has long been believed to play a crucial role in folliculogenesis, including early stages, as it is expressed specifically in the ovary at the time of gonadal differentiation and its expression level increases significantly at the PG-PV transition (Chen et al., 2017; Zhu et al., 2018). This traditional view on roles of aromatase in female development is obviously challenged by our finding with cyp19a1a and dmrt1 double knockout.

To confirm whether the arrest of follicle development was due to a delayed PV-EV transition or permanent blockade, we also examined the ovaries of the double mutants at 100 dpf when the control fish were sexually mature and spawning. The results showed that the follicles in the double mutant ovaries (dmrt1^−/−;cyp19a1a^−/−) remained arrested at PV stage at 100 dpf (Fig. 2B,D). Furthermore, the PV follicles in the double mutants were significantly larger than those in the control ovaries (dmrt1^−/−;cyp19a1a^−/−, 270.22±49.73 μm; dmrt1^+/−;cyp19a1a^+/−, 238.29±35.17 μm; Fig. 2E), still without any vitellogenin accumulation. The mechanism underlying the PV-EV blockade in the ovary of the double mutant (dmrt1^−/−;cyp19a1a^−/−) is unknown, and it will be an interesting issue for further studies. One possible reason is that, without estrogens, the liver would not produce vitellogenin, the precursor of the yolk mass in vitellogenic oocytes (Lubzens et al., 2010; Nagahama and Yamashita, 2008). Our data on plasma and serum estradiol levels also support this hypothesis (Fig. 1B). In mice, although vitellogenesis does not occur, follicular maturation is disrupted and ovulation failure is observed in the absence of aromatase (Fisher et al., 1998). Our evidence indicates that, although estrogens play little role in early folliculogenesis in zebrafish, they are crucial for late stage of follicle development and maturation, consistent with the situation in medaka (Nakamoto et al., 2018) and mammals (Fisher et al., 1998).

In medaka fish, although the cyp19a1a mutant could also form an ovary in the same way as the zebrafish double mutant (dmrt1^−/−;cyp19a1a^−/−), the female mutant eventually underwent sex reversal to become male, suggesting an important role for estrogens in maintaining the ovary (Nakamoto et al., 2018). We also observed similar phenomenon. The follicles in dmrt1^−/−;cyp19a1a^−/− mutant zebrafish exhibited morphological features of atresia after 3 months, such as membrane disintegration, follicle cell proliferation and zona radiata breakdown (Fig. S5) (Morais et al., 2012; Ucuncu and Cakici, 2009). The follicles at 100 dpf gradually underwent degeneration with interfollicular spaces becoming slowly occupied by stromal cells (Fig. 2Bf-f‴; Fig. S6a-a‴), a typical feature seen during the transition from juvenile ovary to testis (Hill and Janz, 2003; Uchida et al., 2002) and in the underdeveloped testes of dmrt1^−/− males (Fig. 2Bg,g′). The follicle degeneration occurred slowly, as evidenced by different degrees of follicle degeneration observed at 100 dpf in dmrt1^−/−;cyp19a1a^−/− double mutants, followed by sex reversal to testis (Fig. S6). However, similar to the testis in dmrt1^−/−;cyp19a1a^−/− double mutant males (Fig. 2Bg,g′; Fig. S6d-d‴), the testis development following ovarian degeneration could not proceed further due to the lack of dmrt1 (Fig. S6). This female-to-male conversion explains the progressive decrease in female-to-male ratio over time (Fig. 1D).

The rescue of the all-male phenotype of cyp19a1a single mutants by dmrt1 mutation raised an interesting question about the regulatory relationship between dmrt1 and cyp19a1a. To address this question, we analyzed the expression levels of dmrt1 and cyp19a1a in wild-type juvenile zebrafish before sex differentiation. At 15 dpf, which is 10 days before sex differentiation (25 dpf) (Chen and Ge, 2013), the expression levels of dmrt1 and cyp19a1a showed the opposite relationship. The individuals with high dmrt1 expression showed lower expression of cyp19a1a, and vice versa (Fig. 3A). Further evidence suggested that the reciprocal relationship of dmrt1 and cyp19a1a could be due to an inhibition of dmrt1 expression by estrogens produced by Cyp19a1a. First, we examined dmrt1 expression in the control (cyp19a1a^+/−) and cyp19a1a mutant (cyp19a1a^−/−) at 25 dpf. As shown in Fig. 3B, two groups of fish could be identified in the presence of Cyp19a1a (cyp19a1a^+/−) at 25 dpf when gonadal differentiation occurs, namely, the dmrt1 high expressers and low expressers. In contrast, all cyp19a1a mutants (cyp19a1a^−/−) showed high expression of dmrt1, likely owing to the lack of estrogens that suppress dmrt1 expression. Second, exposure of male adult zebrafish to estradiol at 0.5 and 5 nM significantly reduced the levels of dmrt1 expression in mature testes (0.5 nM, 0.33±0.14; 5 nM, 0.28±0.19; 0 nM, 1.21±0.86; Fig. 3C), although the treatment did not cause any obvious changes in testes at the histological level (Fig. 3D). The reciprocal relationship between dmrt1 and cyp19a1a is also supported by recent reports that DNA methylation of dmrt1 and cyp19a1a were inversely correlated in developing gonads (Piferrer et al., 2019; Todd et al., 2019).

The results described above strongly suggest a role for cyp19a1a, via estrogens, in suppressing dmrt1 expression in female individuals, thereby driving ovarian differentiation. To understand how estrogens work in the process, we performed a comprehensive genetic analysis, making use of zebrafish mutants of both cyp19a1a and nuclear estrogen receptors (nERs; esr1, esr2a and esr2b), which we generated and reported recently (Lau et al., 2016; Lu et al., 2017).

The fish were sampled at 75 dpf after gonadal differentiation (normally 25-30 dpf); therefore, the sex ratios recorded reflected mostly the maintenance of femaleness rather than female differentiation. As shown in Fig. 4A, the sex ratio of control fish was ∼50% for males and females at 75 dpf. Some individuals in esr1^−/−;esr2a^−/− double mutants (esr1 and esr2a genes are closely linked on the same chromosome) or esr2b^−/− single mutants could develop into females with female ratios relatively lower than the control, which is similar to our recent report (Lu et al., 2017). All fish were males in the presence of dmrt1 but absence of aromatase (cyp19a1a^−/−), and most fish examined were also males in the triple knockout of all three nERs (esr1^−/−;esr2a^−/−;esr2b^−/−) (16/18) with two fish showing intersexuality undergoing female-to-male sex reversal. The all-male phenotype in the triple knockout of nERs could be completely reversed to 100% females by deleting dmrt1 gene (dmrt1^−/−;esr1^−/−;esr2a^−/−;esr2b^−/−). Consistently, quadruple knockout of aromatase and all nERs (cyp19a1a^−/−;esr1^−/−;esr2a^−/−;esr2b^−/−) also resulted in all-male phenotype. Interestingly, disruption of dmrt1 reversed the sex ratio from 100% males in the quadruple knockout of aromatase and nERs (cyp19a1a^−/−;esr1^−/−;esr2a^−/−;esr2b^−/−) to 100% females in the quintuple knockout (dmrt1^−/−;cyp19a1a^−/−;esr1^−/−;esr2a^−/−;esr2b^−/−). Histological analysis showed that the ovaries could form in the quintuple knockout, and the follicles could also develop up to PV stage without aromatase and estrogen receptors (Fig. 4B). Our data using multiple mutants ranging from single (dmrt1^−/−, cyp19a1a^−/− and esr2b^−/−) to quintuple mutant (dmrt1^−/−;cyp19a1a^−/−;esr1^−/−;esr2a^−/−;esr2b^−/−) clearly demonstrate the involvement of nERs in mediating estrogen suppression of dmrt1 expression.

In summary, using a series of zebrafish mutant lines, especially dmrt1^−/− and cyp19a1a^−/− mutants, we provide clear evidence that the ovarian aromatase cyp19a1a works to drive ovarian differentiation by suppressing dmrt1, which is initially expressed in all individuals before sex differentiation to direct gonadal development towards the testis. Further experiments using estrogen receptor mutants (esr1^−/−, esr2a^−/− and esr2b^−/−) demonstrated involvement of nERs in producing estrogen-mediated suppression of dmrt1 expression. Furthermore, the present study provides crucial insight into the role of estrogens in controlling early folliculogenesis. In contrast to the traditional view on estrogen involvement in folliculogenesis, our data show that after knocking down the expression of dmrt1 at gonadal differentiation stage in females, the ovarian aromatase no longer plays any significant roles in promoting early follicle development, including follicle activation (Fig. 5). Further studies on the regulatory mechanisms upstream of cyp19a1a will provide insight into the mechanisms underlying zebrafish sex determination. By generating mutants with mutations of up to five genes, this study also demonstrates the power of zebrafish model in genetic analysis.

The following alleles generated by our previous studies (Lau et al., 2016; Lu et al., 2017) were used in the present study: cyp19a1a^umo5, esr1^umo10;esr2a^umo12 and esr2b^umo13 (see Table S1 for allele information). The alleles mentioned above and the dmrt1 mutant line (registered with ZFIN as dmrt1^umo15; Fig. S1) generated and used in the present study were based on the AB zebrafish strain (China Zebrafish Resource Center, Wuhan, China). The transgenic line Tg(vas:EGFP)^zf45 (Krøvel and Olsen, 2002) was generously provided by Dr. Rüdiger Schulz (Utrecht University, The Netherlands). Wild-type (+/+) and/or heterozygotes (+/−) zebrafish were used as controls in the present study. All experiments were performed according to the guidelines and regulations approved by the Research Ethics Committee of the University of Macau (AEC-13-002).

The fish were raised in an environmental chamber (Thermo Fisher Scientific) and flow-through aquarium system (Tecniplast) at 28±1°C with a photoperiod of 14 hours of light and 10 hours of dark. The larvae were fed with paramecia and artemia, and the adults were fed with artemia and commercial dry food.

The CRISPR/Cas9 approach was used to establish the dmrt1 mutant in zebrafish according to the protocols reported previously (Hwang et al., 2013; Jao et al., 2013; Lau et al., 2016; Lu et al., 2017). An sgRNA (oligo sequence shown in Table S2) targeting exon II of dmrt1 was designed via online websites (http://zifit.partners.org/) to avoid potential off-target effects. The gRNA and Cas9 mRNA were generated through in vitro transcription from DraI-digested pDR274 (Addgene Plasmid #42250) and pCS2-nCas9n (Addgene Plasmid #47929) using the mMACHINE T7 and mMACHINE SP6 kit (Invitrogen) according to the manufacturer's instructions. Then, 4.6 nl of a mixture with gRNA (75 ng/μl) and Cas9 (200 ng/μl) mRNA was injected into one- or two-cell-stage embryos by the Drummond Nanoject system (Drummond Scientific). The F0 mutants were screened using a high resolution melting analysis (HRMA) and a heterozygous mobility assay (HMA), followed by sequencing confirmation (Dahlem et al., 2012; Ota et al., 2013; Zhang et al., 2015). The F0 founders carrying mosaic mutations were mated with the wild-type partners to obtain heterozygous F1(+/−). Sibling female and male F1 individuals carrying the same mutation were mated to obtain the homozygous F2 (−/−).

Three different mutant lines with sequence deletions (−3, −6 and −7) were generated. The CRISPR targeting site and the mutated sequences are shown in Fig S1. The mutant line with a seven-base pair deletion [dmrt1 (−7)] was chosen and maintained for phenotype analysis because it is supposed to induce a frameshift mutation that disrupts Dmrt1 protein synthesis.

Genomic DNA from each individual embryo or a small piece of the caudal fin was extracted by alkaline lysis (Zhang et al., 2015). NaOH (30-50 µl, 50 nmol/µl) was used to extract the genomic DNA at 95°C for 10 min. Then, 3-5 µl Tris-HCl (pH 8.0) was added to the solution to neutralize the reaction system. The genomic DNA was used for HRMA with Precision Melt Analysis software (Bio-Rad) to analyze the difference of the melt curves for distinguishing different genotypes (Dahlem et al., 2012). The primers used for HRMA are listed in Table S2.

The fish were sampled at different time points for phenotype analysis. Histological examination was performed on paraffin sections using Hematoxylin and Eosin (H&E) staining. The fish were sacrificed using MS-222 (Sigma-Aldrich). The gross morphology of each fish was photographed with a digital camera (Canon EOS 7D). The anal fin and cloaca were viewed using a Nikon SMZ18 anatomic microscope and photographed with the Digit Sight DS-Fi2 digital camera (Nikon). Fluorescent images were acquired on the Nikon SMZ18 microscope with a Nikon INTRNSLINGHT C-HGFI laser and photographed with the Digit Sight DS-Fi2 digital camera.

For histological analysis, the entire fish was fixed in Bouin's solution for at least 24 h. Dehydration and infiltration were then performed on the ASP6025S Automatic Vacuum Tissue Processor (Leica). The samples were embedded with paraffin, followed by serial sectioning at 5 µm. The sections were stained with H&E and viewed on the Nikon ECLIPSE Ni-U microscope. The photos were taken using the Digit Sight DS-Fi2 digital camera. Sibling wild-type (+/+) and/ or heterozygous (+/−) fish were used as controls for phenotype analysis.

Female fertility was assessed by the number of eggs ovulated and the percentage of survived embryos. The female fish (+/− or −/−) were naturally mated with wild-type (+/+) male fish, separately. The number of ovulated eggs was counted after spawning. The number of survived embryos was counted 24 h after spawning. The number of eggs ovulated and the embryo survival rate for the female mutants (dmrt1^−/−) were similar to wild-type females (dmrt1^+/+) and the female mutants (dmrt1^−/−) were therefore considered fertile.

Male fertility was assessed by mating performance, which is defined according to their ability to stimulate spawning of the wild-type (+/+) females. At least five individual male fish of each genotype (+/+ or −/−) were tested with the wild-type (+/+) female fish separately. The tests were repeated at 5-day intervals. The spawning rate was defined as the ratio of successful spawning pairs over total pair number. The experiment was repeated at least four times.

Follicles were staged according to morphological and size criteria as we reported previously (Chen and Ge, 2013; Wang and Ge, 2004). Follicle development was staged as follows: PG (primary growth, stage I), PV (previtellogenic, stage II), EV (early vitellogenic, early stage III), MV (mid-vitellogenic, mid-stage III), LV (late-vitellogenic, late stage III) and FG (full-grown).

In general, the gender of zebrafish was identified according to dimorphic morphological features, such as body shape, fin color and genital papillae (Brion et al., 2004), and, if necessary, by dissection under a Nikon SMZ18 stereomicroscope (Nikon, Japan) and/or histological examination. To monitor gonadal differentiation real-time without sacrificing the fish, we introduced the transgene Tg(vas:EGFP)^zf45 into the mutants. As reported previously by us and others, the intensity of GFP fluorescence increased dramatically in the gonads when they differentiate towards ovaries (Lau et al., 2016; Qin et al., 2018; Wang et al., 2007). The sex identity was confirmed further by histology at the end of sampling. In counting the sex of double mutant fish (dmrt1^−/−;cyp19a1a^−/−), only the fish with well-formed ovaries in section were taken as females (Fig. 2Bf-f‴; Fig. S6a-a‴) while the fish with severely degenerating ovaries (Fig. S6b-b‴,c-c‴) or underdeveloped testes (Fig. 2Bg,g′; Fig. S6d-d‴) were taken as males.

For different combinations of quintuple mutants (Fig. 4), fish were firstly separated by genotype. For controls and dmrt1^−/− single mutants, the fish were sexed according to obvious dimorphic morphological features (Webster et al., 2017). For cyp19a1a^−/− single mutants, all fish were males according to our previous report. For other genotypes, all fish were sexed by histological sectioning and analysis.

Total RNA was extracted from juvenile fish (whole fish) and adult fish testes (isolated gonads) using TRIzol (Invitrogen) according to the manufacturer's protocol. Then, reverse transcription was performed using M-MLV reverse transcriptase (Invitrogen). Real-time PCR was performed on the CFX384 Real-Time System (Bio-Rad) using primers listed in Table S2. Melting curve analysis was performed to demonstrate primer specificity. A standard curve was included in each real-time PCR assay for quantification. The relative gene expression level was normalized to the level of ef1a and expressed as a fold change compared with the control group.

The fish were raised until 100 dpf when the body size was large enough for blood collection. In this study, both plasma and serum E2 levels were assessed, and two different batches of fish were used for collecting and preparing plasma and serum. Each fish was anesthetized using MS-222 before sacrifice (Sigma-Aldrich). Once completely anesthetized, the fish blood samples were collected from the heart with a glass capillary or a 10 µl tip and transferred into heparin-sodium-rinsed tubes. The supernatants (plasma) were collected after centrifugation [3000 rpm (850 g), 30 min, 4°C] and then the protein concentrations were quantified using a 2-D Quant Kit (GE Healthcare). The level of E2 in the plasma was measured using an ELISA Kit (Neogen Corporation) according to the manufacturer's instructions.

To measure E2 level in the serum, the blood samples were transferred into 1.5 ml tubes without heparin-sodium treatment. Then the samples were kept at room temperature for 1 h to separate the serum. The supernatants (serum) were collected after centrifugation [3000 rpm (850 g), 30 min, 4°C]. The level of E2 in the serum was measured using an ELISA Kit (Neogen Corporation), according to the manufacturer's instructions.

Adult wild-type fish were treated with E2 (0, 0.5 and 5 nM) for 20 days (from 100 dpf to 120 dpf) continuously. Briefly, 30 fish were divided into three groups, and each group with 10 fish was placed in a 6 liter tank containing E2 at different concentrations. The E2 stock was prepared with ethanol and the same volume of ethanol was added to the control group. The water was changed with new E2 every day.

All values are presented as the mean±s.d., and the significance (*P<0.05; **P<0.01; ***P<0.001; ns, not significant) of the data was analyzed by Student's t-test or one-way ANOVA using Prism 8 (GraphPad) on an Apple Mac.

No statistical method was applied to predetermine sample size. The sample size of control and mutant zebrafish used was according to experimental considerations. For each experiment, sample size is indicated in the relevant figure or figure legend. The sample size reflects the number of independent biological replicates. All experiments were performed at least twice.

We thank the Histology Core of the Faculty of Health Sciences for technical support.