Zebrafish Nedd8 facilitates ovarian development and the maintenance of female secondary sexual characteristics via suppression of androgen receptor activity

Reversible post-translational protein modifications modulate protein activity in diverse cellular processes (Menzies et al., 2016; Oh et al., 2018; Rape, 2018). Neural precursor cell expressed developmentally downregulated protein 8 (Nedd8), as a ubiquitin-like protein (UBL), is covalently conjugated to the lysine residues of target substrates resulting in protein neddylation, a post-translational modification (Enchev et al., 2015). Like ubiquitylation, neddylation involves E1 (activating), E2 (conjugating) and E3 (catalyzing) enzymes. Neddylation is a reversible modification; protein deneddylation is performed by deneddylase, such as DEN1/SENP8 (Gan-Erdene et al., 2003; Wu et al., 2003). The cullin-RING ligases (CRLs), in the E3 ubiquitin ligase family, are the principal substrates of neddylation (Enchev et al., 2015). Recently, neddylation has also been shown on non-cullin targets, regulating substrate protein activity, stability and subcellular localization (Abidi and Xirodimas, 2015; Enchev et al., 2015; Vogl et al., 2015; Watson et al., 2011; Zuo et al., 2013). Functionally, neddylation is crucial for gene regulation, cell survival, organ development and the stress response (Vogl et al., 2015; Xirodimas et al., 2004; Zou et al., 2018; Zuo et al., 2013). Dysregulation of neddylation is associated with disease pathogenesis (Soucy et al., 2009; Xirodimas et al., 2004). For example, an E1 inhibitor of neddylation (MLN4924) was shown to restrict tumor growth and is currently in clinical trials as a cancer treatment (Shah et al., 2016; Soucy et al., 2009).

To investigate the function of the NEDD system in vivo, various animal models have been developed (Chan et al., 2008; Tateishi et al., 2001; Vogl et al., 2015; Zou et al., 2018). Mice deficient in the Uba3 gene that encodes a catalytic subunit of NEDD8-activating enzyme die in utero at the preimplantation stage, suggesting that the NEDD8 system is essential for cell cycle progression and the morphogenetic pathway (Tateishi et al., 2001). The cardiomyocyte-specific knockout of Nae1, a subunit of the E1 activating enzyme, led to myocardial hypoplasia, ventricular noncompaction and heart failure in mice late in gestation, resulting in perinatal lethality (Zou et al., 2018). In addition, studies of Drosophila DEN1-null mutants indicated that DEN1 deneddylates many cellular proteins in addition to cullin proteins (Chan et al., 2008). However, the functions of other components of the NEDD system have not been illustrated by animal models in vivo. In particular, the function of nedd8 in vivo remains largely unknown.

The zebrafish (Danio rerio) is a model organism that has been widely used for studies of gene function in vivo. To further investigate the role of the NEDD system, we used CRISPR/Cas9 to knock out nedd8 in zebrafish. We found that loss of nedd8 caused defects in ovarian development, and led to the growth of the breeding tubercles (BTs) on the pectoral fins of female zebrafish. The BTs are keratinized multicellular epidermal structures that normally present on the dorsal surface of the anterior rays of zebrafish male pectoral fins, not females (Kang et al., 2013; McMillan et al., 2013; Nachtrab et al., 2011). Moreover, we showed that Nedd8 conjugates to Androgen receptor (Ar), inhibiting its transactivity.

The nedd8 gene is evolutionarily conserved across human, mice and zebrafish (Fig. S1A). During embryogenesis, nedd8 was detected at a very early stage (two-cell) and was ubiquitously expressed up to 16 h post fertilization (hpf) (Fig. S1B). After 16 hpf, nedd8 was specifically expressed in the brain, notochord and tail, with particularly high levels of expression detected in the notochord (Fig. S1B). In adults (≥4 months post fertilization, mpf), nedd8 was highly expressed in the brain, testis, ovaries and muscles (Fig. S1C). Immunohistochemical staining showed that the Nedd8 protein was also expressed in the germ cells in testes and ovaries during gonad development (Fig. S1D). These data suggest that nedd8 might play a role in gonadal development.

To investigate the function of nedd8 in vivo, we generated nedd8-null zebrafish (nedd8^{ihb1227/ihb1227}) using CRISPR/Cas9 (Yu et al., 2019). Mutant zebrafish (nedd8^−/−) were phenotypically identical to their wild-type siblings (nedd8^+/+). The heterozygous zebrafish (nedd8^+/−) were indistinguishable from their wild-type siblings (nedd8^+/+). Notably, the BTs were frequently observed on the surface of adult female nedd8^−/− pectoral fins (∼83% of all nedd8^−/− females), but were hardly observed on the fins of wild-type females (∼0%). BTs are male secondary sexual characteristics, which are typically present on the pectoral fins of adult males (Fig. S2) (Kang et al., 2013; McMillan et al., 2013). Notably, ar-knockout male zebrafish (ar^−/−) do not develop BTs, but do exhibit some female secondary sexual characteristics (Yu et al., 2018). Therefore, we hypothesized that nedd8 might impact ar gene function. Next, we aimed to further characterize the phenotype of nedd8^−/− females to clarify the relationship between nedd8 and ar.

When we crossed nedd8^+/− (♀) zebrafish with nedd8^+/− (♂) zebrafish, the nedd8^−/− offspring exhibited a male sex bias and very low fecundity of nedd8^−/− female offspring while nedd8^−/− male offspring exhibited normal fecundity. Subsequently, in order to obtain more nedd8^−/− zebrafish from each mating, we mated nedd8^+/− (♀) zebrafish with nedd8^−/− (♂) zebrafish to get nedd8^+/− and nedd8^−/− offspring. The female:male ratio of the nedd8^−/− offspring was substantially lower than the female:male ratio of nedd8^+/− offspring (Fig. 1A). The ovulation rates of nedd8^−/− females was also lower than those of the nedd8^+/+ females (Fig. 1B). Furthermore, the nedd8^−/− eggs were fertilized at a significantly lower rate than the nedd8^+/+ eggs (Fig. 1C). Compared with nedd8^+/+ adult females, nedd8^−/− adult females had elongated bodies, transparent ovaries and degenerated eggs (Fig. 1D). In addition, the mean gonadosomatic index (GSI) of the nedd8^−/− adult females was lower than that of the nedd8^+/+ adult females (Fig. 1E). Histological analysis of the adult ovaries showed that nedd8^−/− ovaries contained more oocytes at the primary growth stage (PG) and the previtellogenic stage (PV), but fewer oocytes at the early vitellogenic stage (EV), the midvitellogenic stage (MV) and the full-grown stage (FG), compared with nedd8^+/+ ovaries (Fig. 1F,G).

The crucial period of zebrafish sexual differentiation is 17-35 days post fertilization (dpf); during this time, zebrafish develop ‘juvenile ovaries’ containing gonocytes (Sun et al., 2013). After 40 dpf, juvenile ovaries pass the transitional period of sex determination and develop into an immature ovary and testis (Sun et al., 2013). To identify the stage at which the deletion of nedd8 started to affect gonadal development and sexual differentiation, we examined the gonads of zebrafish from juvenile (24 dpf) to 2 mpf. At 24 dpf, the nedd8^−/− gonads contained more gonocytes (GO, indicated by white arrows in Fig. S3A) and degenerated perinucleolar oocytes (indicated by red arrows in Fig. S3A) compared with their wild-type siblings (Fig. S3A,B). At 40 dpf, the nedd8^+/+ ovaries contained vitellogenic stage oocytes, but the nedd8^−/− ovaries only contained early development stage PG and PV oocytes. Interestingly, the nedd8^−/− testes developed even faster than those of their wild-type siblings: when nedd8^+/+ testes contained spermatogonia (SG) and spermatocytes (SC) only, spermatids (ST) were already present in the nedd8^−/− testes (Fig. S3C,E). In the nedd8^−/− ovaries at 2 mpf, most oocytes were arrested at the early development stage, but the nedd8^+/+ ovaries were filled with oocytes at different developmental stages (Fig. S3F,G). However, no obvious difference was observed between the nedd8^+/+ and the nedd8^−/− testes (Fig. S3F,H). These results suggest that nedd8 disruption leads to defects in ovarian development in female zebrafish, but did not substantially affect male zebrafish.

To get deeper insight into the effects of nedd8 in ovarian development, we examined expression of Vasa (also known as Ddx4), Zili (Piwil2) and Ziwi (Piwil1) in nedd8^+/+ and nedd8^−/− ovaries – three key proteins required for zebrafish germ cell differentiation and maintenance (Dranow et al., 2016; Hartung et al., 2014; Houwing et al., 2008, 2007; Zhu et al., 2019). As marked by Vasa, in juvenile ovaries at 24 dpf, more primordial germ cells (PGCs) and more primary oocytes (Zili and Ziwi expressed around nuclei) were detected in nedd8^−/− compared with nedd8^+/+. In nedd8^−/− ovaries at 2 mpf, the granulosa cells and theca cells marked by Zili and Ziwi were fewer than those in nedd8^+/+ ovaries (Fig. S4A,B). These observations suggested that loss of nedd8 resulted in oocytes arrested at early developmental stages, with few developed granulosa cells and theca cells.

Intriguingly, nedd8^−/− males displayed deeper yellow pigmentation on the anal fins compared with the nedd8^+/+ males (Fig. S5A). As nedd8^−/− females exhibited some male secondary sexual characteristics, and as hyperactive behaviors are affected by androgen signaling (Kinch et al., 2015), we compared the behaviors of nedd8^+/+ and nedd8^−/− males. The nedd8^−/− males were not only more sexually aggressive (chasing females) than nedd8^+/+ males, but were also more active when chasing food (Movies 1 and 2). Our movement tracing records showed that the locomotor activity of nedd8^−/− males was also greater than that of nedd8^+/+ males (n=6 per group, over a 10 min period; Fig. S5B,C).

We also measured levels of serum 11-ketotestosterone (11-KT) and estradiol (E2) in nedd8^+/+ and nedd8^−/− zebrafish. At 3 mpf, serum 11-KT and estradiol were similar between nedd8^+/+ and nedd8^−/− males, but serum 11-KT levels were higher and serum estradiol levels were lower in nedd8^−/− females compared with nedd8^+/+ females (Fig. S6A,B). At 6 mpf, serum 11-KT and estradiol levels were higher in both male and female nedd8^−/− zebrafish compared with nedd8^+/+ males and females (Fig. S6C,D). Subsequently, we examined two female determination genes [cyp19a1a and foxl2 (foxl2a)] and two male determination genes (amh and dmrt1) in ovaries and testes. At 3 mpf, the expression of cyp19a1a and foxl2 was downregulated, and that of amh and dmrt1 was upregulated in nedd8^−/− ovaries (Fig. S6E) (Lau et al., 2016; Yang et al., 2017; Yin et al., 2017), but the expression levels of amh and dmrt1 were upregulated and the expression levels of cyp19a1a and foxl2 were not altered significantly in nedd8^−/− testes (Fig. S6F). At 6 mpf, the expression level of cyp19a1a was upregulated in nedd8^−/− ovaries and testes, but the expression level of amh was downregulated in nedd8^−/− ovaries (Fig. S6E,F). No significant change was observed for other genes at this stage (Fig. S6E,F). In fact, the levels of serum estradiol (E2) reflected the cyp19a1a expression level in nedd8^+/+ and nedd8^−/− zebrafish.

We crossed ar^−/− (♀) and nedd8^−/− (♂) zebrafish to generate double knockout offspring. We previously showed that ar^+/−nedd8^+/+ ovaries develop normally, similar to those of wild-type (ar^+/+nedd8^+/+) females, but ar^−/−nedd8^+/+ ovaries exhibit premature ovarian failure during growth (Yu et al., 2018). This was consistent with results seen here (Fig. 2A,B). Surprisingly, although the gross shapes and tissue structure of ar^+/−nedd8^−/− ovaries were similar to those of wild-type ovaries (ar^+/+nedd8^+/+) with fully-grown oocytes, the ar^+/+nedd8^−/− ovaries mainly contained PG, PV and EV oocytes (Fig. 2A,B; Fig. 1D,F; Fig. S7). Owing to complete loss of ar, the ar^−/−nedd8^−/− ovaries exhibited atretic and degenerated follicles (Yu et al., 2018) (Fig. 2A,B; Fig. S7). These results suggest that loss of one copy of ar at least partially rescued the defects of nedd8^−/− ovary maturation. However, the defects of the ar^−/−nedd8^−/− oogenesis were even more serious than those of ar^−/−nedd8^+/+ and ar^+/+nedd8^−/− oogenesis (Fig. 2A,B; Fig. 1D,F; Fig. S7). This suggests that ar functions downstream of nedd8, as mutants lacking both copies of ar showed ovarian development traits characteristic of ar disruption (Yu et al., 2018). Therefore, we hypothesized that nedd8 might affect folliculogenesis by modulating ar function, and that nedd8 acted upstream of ar. This might explain why the loss of one copy of ar (but not two) rescued nedd8^−/− oogenesis.

The lower fertilization rate of ar^+/+nedd8^−/− eggs suggest that loss of nedd8 not only disrupts ovarian development, but also affects egg quality. To determine whether loss of one copy of ar also rescued the low fertilization rate of ar^+/+nedd8^−/− eggs, we conducted further fertilization rate assays. Different groups of wild-type males (ar^+/+nedd8^+/+) were separately mated with female zebrafish having one of three genotypes: ar^+/+nedd8^+/+, ar^+/+nedd8^−/−, or ar^+/−nedd8^−/−. The fertilization rate of the ar^+/+nedd8^−/− (♀)×ar^+/+nedd8^+/+ (♂) cross was significantly lower than that of the ar^+/−nedd8^−/− (♀)×ar^+/+nedd8^+/+ (♂) cross, although this rate was still lower than that of the wild-type cross [ar^+/+nedd8^+/+ (♀)×ar^+/+nedd8^+/+ (♂); Fig. 2C,D]. These results suggest that loss of one copy of ar partially rescued the low fertilization rate of ar^+/+nedd8^−/− eggs. Moreover, by mating ar^+/−nedd8^−/− (♀)×ar^+/−nedd8^−/− (♂) and subsequently examining sex ratio in their offspring, we found that the sex bias towards male phenotype characteristics of nedd8 mutants was rescued in ar^+/+nedd8^−/− zebrafish (Fig. S8).

It is evident that the germ cell number of zebrafish determines sexual differentiation (Dranow et al., 2016). To determine whether rescuing of sex bias in nedd8-null zebrafish by loss of one copy of ar resulted from rescuing germ cell number at early embryogenesis, we examined germ cell numbers by immunofluorescent staining using anti-Vasa antibody. As expected, loss of one copy ar indeed increased germ cell number in nedd8^−/− background embryos (Fig. 2E,F). Interestingly, the disorganization, including lack of seminiferous tubule structure, in ar^−/−nedd8^−/− testes was quite similar to that of the ar^−/−nedd8^+/+ testes (Fig. S9).

To determine whether BTs developed on the pectoral fins of nedd8^−/− female zebrafish because nedd8 modulated ar, we compared BTs between adult female ar^+/+nedd8^−/− and ar^+/−nedd8^−/− zebrafish at 4 mpf. Although ar^+/−nedd8^−/− females retained some BTs, these females had far fewer BTs than ar^+/+nedd8^−/− females, both with respect to the total number of BTs on the pectoral fins (i.e. the dorsal surface from the second to the fourth pectoral fin rays) and with respect to the average number of BTs per pectoral fin segment (Fig. 3A-C) (Kang et al., 2013). When ar^+/+nedd8^−/− and ar^+/−nedd8^−/− female zebrafish were treated with flutamide (an androgen antagonist) (Martinovic-Weigelt et al., 2011), all BTs disappeared in both mutants, indicating that androgen signaling plays an essential role in BT formation in nedd8-null females (Fig. 3A,C) (McMillan et al., 2013). Furthermore, flk (kdrl) expression was increased in nedd8^−/− females compared with nedd8^+/+ females (Fig. 3D); flk may be associated with androgen-induced BTs formation (McMillan et al., 2013). As expected, flutamide treatment reduced flk expression compared with untreated controls (Fig. 3D).

Importantly, BTs did not develop on the pectoral fins of adult male and female ar^−/−nedd8^−/− zebrafish (Fig. S10), further suggesting that androgen signaling is essential for BT formation, and that nedd8 acts upstream of ar.

To determine how nedd8 affects ar function and ovarian maturation, we quantified the expression level of kitlga (Yao and Ge, 2015) and two male determination genes, amh and dmrt1 (Lin et al., 2017) in female zebrafish with different genetic backgrounds. To confirm that kitlga was indeed an androgen responsive gene, we measured its expression in zebrafish after androgen injection. Injection with dihydrotestosterone (DHT) induced kitlga expression in female zebrafish (Fig. S11A); kitlga was more strongly upregulated in nedd8^−/− females than in nedd8^+/+ females (Fig. 4A), but kitlga gene expression in ar^+/−nedd8^−/− females was substantially lower than kitlga gene expression in ar^+/+nedd8^−/− females. When both copies of ar were deleted in female zebrafish (i.e. ar^−/−nedd8^−/− mutants), kitlga expression was barely detectable (Fig. 4A). Notably, flutamide treatment dramatically reduced kitlga expression in ar^+/+nedd8^+/+ (wild type), ar^+/+nedd8^−/−, and ar^+/−nedd8^−/− females (Fig. 4A), indicating that androgen signaling strictly controls kitlga gene expression.

Similarly, the expression levels of amh and dmrt1 were also greater in nedd8^−/− females than in nedd8^+/+ females, but loss of one copy of ar reversed this effect (Fig. 4B,C). Notably, in the nedd8-null background, the expression level of ar was highest in ar^+/+ ovaries, but moderate in ar^+/− ovaries and undetectable in ar^−/− ovaries (Fig. S11B). Our results thus suggest that the loss of nedd8 enhances androgen signaling and disrupts ovarian maturation by upregulating male determination genes and downregulating female determination genes (Fig. S6E,F; Fig. 4B,C). Therefore, nedd8 might modulate androgen signaling.

To investigate whether nedd8 directly regulates ar, we initially used promoter assays to test how nedd8 expression affected ar transcriptional activity. Promoter assays using probasin (Pbsn), a well-described androgen-responsive gene in rats, has been widely used to monitor androgen signaling (Greenberg et al., 1994; Wang et al., 2014; Yan et al., 1997). Therefore, we used a Pbsn promoter luciferase reporter to monitor zebrafish ar activity. In the presence of DHT, the overexpression of the ar gene in epithelioma papulosum cyprinid cells (EPC), a fish cell line established from carp (Cyprinus carpio) (Fijan et al., 1983), activated the Pbsn promoter (Fig. 5A). However, when the nedd8 gene was overexpressed, Pbsn promoter activity decreased (Fig. 5A), suggesting that Nedd8 might conjugate to Ar and inhibit ar transcriptional activity. The overexpression of genes encoding the nedd8 enzyme E1 (uba3) and E2 (ubc12; ube2m) reversed the increase in Pbsn promoter activity induced by ar overexpression in the presence of DHT (Fig. 5B). This suggested that nedd8 might inhibit Ar transcriptional activity through nedd8-conjugation. Furthermore, the overexpression of a mutated form of nedd8 (nedd8ΔGG, a conjugation-defective mutant generated by Gly-75/76 deletion) (Ryu et al., 2011) did not suppress the Pbsn promoter activity by ar overexpression in the presence of DHT – instead, the overexpression of nedd8-ΔGG increased ar transcription (Fig. 5C). On the contrary, the overexpression of the deneddylase senp8, enhanced the Pbsn promoter activity induced by ar overexpression in the presence of DHT (Fig. 5D), further implying that neddylation may be responsible for the inhibition of Ar activity. These data suggested that nedd8 might inhibit ar transcriptional activity through neddylation modification.

To determine whether zebrafish ar was indeed modified by neddylation, we initially performed in vitro neddylation assays using Ni-NTA-agarose beads. Overexpression of wild-type nedd8 caused clear nedd8-conjugated Ar bands, but no band was detected when nedd8 was absent (Fig. 6A). By contrast, overexpression of the nedd8 mutant (nedd8ΔGG) did not cause Ar modification (Fig. S12A) (Vogl et al., 2015). Moreover, treatment with the E1 inhibitor MLN4924 reduced nedd8-conjugation with Ar, but overexpression of E1 (uba3) and E2 (ubc12) enhanced nedd8-conjugation with Ar (Fig. 6B).

It has been shown that nedd8 overexpression triggers unphysiological neddylation pathways (Enchev et al., 2015). To further confirm that Ar was indeed modified by Nedd8, we sought to conduct in vivo neddylation assays. A monoclonal antibody against zebrafish Ar was developed and its specificity was validated using zebrafish testes (Fig. S13). The in vivo neddylation assay indicated that endogenous Ar was modified by Nedd8 in nedd8^+/+ zebrafish testes, but not in nedd8^−/− zebrafish testes (Fig. 6C).

To determine which residue(s) in Ar was (were) modified by neddylation, we performed mutant screening by taking advantage of in vitro neddylation assays. We mutated all lysine residues in Ar to arginine and made a series of mutants (Fig. 6D). Through in vitro neddylation assays, we found that ar-K475R and ar-K862R completely lost neddylation by Nedd8 overexpression (Fig. 6D). Therefore, lysine 475 and lysine 862 might be the target sites of neddylation, which was further confirmed by the double mutant (Ar-2R, in which K475 and K862 were mutated into arginine simultaneously) (Fig. 6E). Of note, these two sites are evolutionarily conserved (Fig. S12B).

By in vitro ubiquitylation assays, we found that Ar-K475R could not be ubiquitylated, but wild type (Ar), Ar-K862R and Ar-2R (K475R/K862R) could still be ubiquitylated, indicating that K475 in Ar is not only one neddylated site, but also one ubiquitylated site (Fig. 6F). Thus, K862 in Ar might be targeted by neddylation specifically. Subsequent Pbsn promoter luciferase reporter assays showed that the activity of Ar-K475R was still suppressed by nedd8 overexpression, but the activity of Ar-K862R and Ar-2R (K475R/K862R) was not suppressed by nedd8 overexpression (Fig. 6G), suggesting that K862 is the key neddylated site accounting for the modulation of Ar activity in response to neddylation modification. Intriguingly, in the presence of DHT, the activity of Ar-K475R, Ar-K862R and Ar-2R (K475R/K862R) was higher than that of wild-type Ar, implying that the modification of these two lysine residues, either ubiquitylation (K475) or neddylation (K862), was crucial for the inhibition of Ar transcriptional activity (Fig. 6G). Taken together, these data suggest that neddylation might occur at K475 and K862 of Ar, and that neddylation on K862 of Ar effectively repressed ar transactivity.

It has been reported that the SPOP-CUL3-RBX1 ubiquitin ligase complex targets mammalian AR for degradation (An et al., 2014). Given that the SPOP protein is highly evolutionarily conserved between zebrafish and mammals [97% amino acid (aa) match] and nedd8 is well known to activate Cullin-based ubiquitin ligases (Petroski and Deshaies, 2005), we speculated whether nedd8 could inhibit Ar through the activation of the SPOP-CUL3-RBX1 ubiquitin ligase. Initially, we examined Ar protein levels between nedd8^+/+ and nedd8^−/− zebrafish testes and we found no difference in Ar protein level between nedd8^+/+ and nedd8^−/− (Fig. S14A). Further in vivo ubiquitylation assays showed that Ar ubiquitylation in nedd8^−/− zebrafish was similar to that in nedd8^+/+ zebrafish (Fig. S14B). Moreover, overexpression of neddylation components (nedd8, uba3 and ubc12) did not enhance the Spop-induced suppressive effect on the activity of the Pbsn reporter (Fig. S14C). Consistently, knockdown of spop in EPC cells also had no effect on the activity of the Pbsn reporter when Ar and Nedd8 were overexpressed with or without DHT treatment (Fig. S14D,E). Therefore, the inhibition of Ar by nedd8 was not mediated by affecting the activity of SPOP-CUL3-RBX1 ubiquitin ligase.

Based on these observations, we proposed a working model of the regulatory effects of nedd8 on ar activity and gonadogenesis (Fig. 7). When nedd8 is intact, the transcriptional activity of ar is strictly controlled by neddylation, and Nedd8 conjugation appears to serve as a suppressor of ar activity. Under these conditions, zebrafish gonads develop normally, and differentiate into testes or ovaries at various time points. If nedd8 is disrupted, however, ar loses control by neddylation and then ar activity increases substantially after binding to DHT. As a result, in female zebrafish, the male determination genes (e.g. amh and dmrt1) are upregulated, and the female determination genes (e.g. cyp19a1a and foxl2) are inhibited. Conversely, in male zebrafish, the male determination genes (e.g. amh and dmrt1) are upregulated, but the female determination genes (e.g. cyp19a1a and foxl2) are not altered at 3 mpf. Consequently, oogenesis is disrupted, resulting in defects in folliculogenesis and the expression of masculinized secondary sexual characteristics. In addition, nedd8-null males exhibit normal fecundity and become super-activated due to the loss of androgen signaling inhibition.

Neddylation is essential in all model organisms except Saccharomyces cerevisiae (Enchev et al., 2015). Disruption of some components of the NEDD system affects early embryogenesis dramatically, leading to embryonic lethality (Tateishi et al., 2001; Zou et al., 2018). In this study, we could generate viable Nedd8-null zebrafish, which can be used for a series of genotype analysis. Thus, our work provides a practical vertebrate model for further revealing the physiological function of neddylation in vivo.

In addition to CRLs, other proteins are modified by neddylation (Enchev et al., 2015; Ryu et al., 2011; Vogl et al., 2015; Xirodimas et al., 2004; Zou et al., 2018; Zuo et al., 2013). The importance of protein neddylation is clear (Enchev et al., 2015). Of note, uba3, the catalytic subunit of the activating enzyme of the Nedd8 conjugation pathway, could inhibit steroid receptor function, linking neddylation to the suppression of steroid receptor function (Fan et al., 2002). Uba3 has also been shown to bind to the ligand binding domain (LBD) of AR (Fan et al., 2002; Nadal et al., 2017), suggesting that neddylation might occur in the LBD of AR. In this study, we identify that K862, which is located in LBD of Ar, is the key target residue of neddylation and its modification influences ar function dramatically, providing further evidence to support the fact that neddylation occurs in the LBD of Ar and has an important role in modulating Ar transactivity. Interestingly, K475 of Ar is not only modified by neddylation, but also modified by ubiquitylation. In addition, although Ar-K475R has higher activity than wild-type Ar, its activity is suppressed by overexpression of Nedd8. These observations suggest that K475 neddylation may not affect Ar function; in contrast, other modifications in K475, such as ubiquitylation, may account for the regulation of Ar function through modifying K475 of Ar. Further investigation of this phenomenon and the underlying mechanisms will expand our knowledge about the regulation of Ar function through post-translational modifications and the crosstalk between different modifications.

Neddylation has been shown to influence germ cell differentiation in Drosophila ovaries (Lu et al., 2015; Pan et al., 2014) and to regulate gene expression in stem cells in Drosophila testes (Qian et al., 2015). In the Drosophila ovary, if Csn4, a deneddylase-like gene, is present, Nedd8 promotes self-renewal. If Csn4 is absent, Nedd8 promotes differentiation (Lu et al., 2015; Pan et al., 2014). Here, we find that loss of nedd8 in zebrafish causes reduced PGCs at early embryogenesis, resulting in more males and immature oocytes versus mature oocytes compared with wild-type zebrafish. This observation appears to be consistent with nedd8 function in the Drosophila ovary.

In this study, we used a zebrafish model to demonstrate that zebrafish nedd8 facilitates ovarian maturation and the maintenance of female secondary sexual characteristics. However, the loss of ar partially rescues ovarian function in nedd8-null females. Thus, not only did our results demonstrate that neddylation is important for gonadogenesis, but our data also indicated that ar might be a novel target of neddylation.

In fish, the plasticity of gonadal sex differentiation in response to treatment with exogenous steroids is well known (Godwin, 2010): sex steroids play a key role in fish sex determination (Nakamura, 2010). Androgen treatment for early embryonic fish transitions ovaries to testes, producing physiological males; this technique is widely used to generate all-female populations in the aquaculture industry (Nakamura, 2010). Thus, it is clear that additional androgen affects sex determination. Of note, in nedd8 mutants, as well as an increase in Ar activity, the production of androgen (11-KT) was also significantly increased, suggesting that nedd8 might also suppress androgen production. Therefore, the defects of the nedd8 mutant could be due to both increased androgen production and androgen receptor signaling. To further figure out the mechanisms underlying this phenomenon will help us to understand the role of nedd8 in gonadal development more completely.

Studies of ar-knockout zebrafish have shown that androgen signaling plays an essential role in the maintenance of ovary function (Crowder et al., 2018; Tang et al., 2018; Yu et al., 2018). Here, the nedd8-knockout-induced overactivation of the androgen receptor also disrupted ovarian oogenesis and led to the production of more fertile males. However, the female:male ratio of the nedd8^−/− offspring suggested that the nedd8-knockout had a less dramatic effect on androgen levels than did direct androgen treatment. Our results indicated that appropriate levels of androgen were necessary for normal ovarian development and function. Intriguingly, the over-masculinized behaviors exhibited by the nedd8^−/− males suggested that androgen signaling also plays an important role in fish behavior.

As it is difficult to ascertain the sex of, and collect blood from, early-stage zebrafish embryos, we were unable to determine whether the loss of nedd8 affected serum hormone levels at the early stages of development. Based on our observations that the defects of oogenesis appeared in nedd8^−/− ovaries at the early stage (24 dpf), the inhibitory role of nedd8 on Ar activity should have a direct effect on oogenesis of nedd8^−/− females. At 3 mpf, the serum levels of 11-KT and estradiol were similar in nedd8^+/+ and nedd8^−/− males. However, we did observe the difference in serum hormone levels between 3 mpf wild-type and nedd8-null females, and between 6 mpf wild-type and nedd8-null males/females. Of note, expression of cyp19a1a, an aromatase which converts androgens to estrogens (Dranow et al., 2016), was downregulated in nedd8^−/− ovaries at 3 mpf, but upregulated in nedd8^−/− ovaries at 6 mpf. In fact, the serum level of estradiol coordinates with expression of cyp19a1a. The changes of serum hormones in nedd8^−/− females might also influence oogenesis, oocyte maturation, BT growth and overactive behaviors secondarily. To further distinguish the direct and indirect effects of nedd8 will help us to fully understand the role of nedd8.

Zebrafish sex determination and gonad differentiation are complicated, particularly with respect to the molecular control of these processes (Lau et al., 2016; Liew and Orban, 2014; Lin et al., 2017; Orban et al., 2009; Uchida et al., 2002; Yang et al., 2017). Recently, the roles of ar in sexual determination, ovarian development, and maintenance of secondary sexual characteristics have been well-characterized by analyzing ar-knockout zebrafish (Crowder et al., 2018; Tang et al., 2018; Yu et al., 2018). Loss of ar in zebrafish increases the proportion of female offspring, causes male infertility, leads to defects in oocyte maturation, reduces fecundity and produces males expressing female secondary sexual characteristics (Crowder et al., 2018; Tang et al., 2018; Yu et al., 2018). Here, our results indicated that zebrafish nedd8 participates in sex determination, ovarian maturation and maintenance of secondary sexual characteristics by modulating ar activity, further demonstrating the importance of ar in zebrafish sex determination and gonad differentiation.

Importantly, the loss of one copy of ar did not completely rescue abnormalities observed in the nedd8^−/− zebrafish (e.g. disrupted ovarian maturation, reduced egg fertilization rate and the development of BTs on the female pectoral fins). It was possible that the loss of one copy of ar was not sufficient to counteract the nedd8-knockout-induced increase in ar activity. Alternatively, nedd8 might affect folliculogenesis via mechanisms other than the modulation of ar activity only. Indeed, our finding that the defects in ovarian maturation were more severe in ar^−/−nedd8^−/− females than in ar^−/−nedd8^+/+ females supported this second possibility.

Similar to ubiquitylation, neddylation also requires E3 ligases (Enchev et al., 2015). However, unlike the E3 ligases identified in ubiquitylation, few E3 neddylation ligases have been identified (Xirodimas et al., 2004; Zuo et al., 2013). The identification of the E3 ligases mediating ar neddylation is necessary to fully understand the role of neddylation in androgen signaling.

The roles of androgen signaling in prostate cancer pathogenesis has received much recent attention (Watson et al., 2015). Indeed, the post-translational modification of AR may be essential for the development of castration-resistant prostate cancer (CRPC) (Koryakina et al., 2014; Coffey and Robson, 2012; Gaughan et al., 2005; Gioeli and Paschal, 2012; Van der Steen et al., 2013). Here, we demonstrated that Ar was modified by neddylation; this represented a previously undescribed post-translational modification of AR. The association of neddylation with cancer initiation and progression has been widely explored (Abidi and Xirodimas, 2015; Zhou et al., 2018), and the E1 inhibitor of neddylation (MLN4924) has been used in clinical trials for various cancers (Soucy et al., 2010, 2009). Thus, a better understanding of the neddylation mechanisms associated with androgen signaling might inform the development of therapeutic treatments for prostate cancer.

Nedd8-null zebrafish and ar-null zebrafish have been described previously (Yu et al., 2018, 2019). Zebrafish were maintained in a re-circulating water system according to standard protocol. Fish were maintained at 28.5°C with a photoperiod of 14 h of light and 10 h of darkness, and fed regularly. All experiments with zebrafish were approved by the Institutional Animal Care and Use Committee of the Institute of Hydrobiology, Chinese Academy of Sciences under protocol number 2016-018.

Flutamide (Sigma, F9397) was dissolved in dimethyl sulfoxide (DMSO) at a concentration of 10 mg/ml. The final concentration of flutamide for zebrafish treatment was 2 mg/l. Zebrafish (4 mpf) were put into a 1 l tank with flutamide (2 mg/l) and water was changed every day.

DHT (Sigma-Aldrich) injection assay was performed as previously described (Yu et al., 2018). MLN-4924 (Merck) was dissolved in DMSO and used at a final concentration of 1 μM.

After anesthesia with MS-222, the testes and ovaries of zebrafish were dissected and the GSI (gonad weight/body weight×100%) was obtained. The testes and ovaries were then fixed in 4% paraformaldehyde (PFA) overnight at 4°C. The samples were dehydrated and embedded in paraffin, and cut into 4 μm sections. Hematoxylin and Eosin (H&E) staining, immunohistochemical staining and immunofluorescent staining were performed as described previously (Yu et al., 2018; Zhu et al., 2019).

Adult female zebrafish with different genetic backgrounds: ar^+/+nedd8^+/+, ar^+/+nedd8^−/−, and ar^+/−nedd8^−/− were transferred to breeding tanks respectively; adult wild-type male zebrafish (ar^+/+nedd8^+/+) were put into the same breeding tank at a female:male ratio of 1:1. The number of eggs ovulated and the ovulation rate (ovulation rate=number of spawned females/total number of females×100%) were assessed. If the female zebrafish did not spawn after mating with male zebrafish, the experiments were repeated 7 days later. All experiments were repeated at least three times. The female zebrafish were considered sterile if three attempts did not produce eggs, or the eggs could not be fertilized.

Whole mount in situ hybridization was performed as previously described (Liu et al., 2016). The probe for zebrafish nedd8 was amplified from cDNA pools using the primers listed in Table S1, and the probe was synthesized using the Transcript Aid T7 High Yield Transcription Kit (Fermentas).

Blood samples were collected from 3- and 6-month-old zebrafish as described previously (Pedroso et al., 2012). For each zebrafish, 3-10 μl of blood could be collected. Blood collected from three individuals was used as one sample for measurement. The blood samples were centrifuged at 5000 g for 20 min at 4°C, and the supernatants were separated and purified according to the manufacturer's extraction protocol (Cayman Chemical). 11-KT and estradiol (E2) were measured by competitive enzyme-linked immunosorbent assay (ELISA) kits (Cayman Chemical) following the manufacturer's instructions. All standards and samples were measured by three independent experiments performed in triplicate.

EPC cells (Fijan et al., 1983; originally obtained from the American Type Culture Collection) were cultured in Medium 199 (Invitrogen) supplemented with 10% fetal bovine serum (HyClone), maintained at 28°C in a humidified incubator containing 5% CO₂. As a fish cell line, EPC cells should be suitable for investigating the regulation of zebrafish androgen receptor activity. HEK293 T cells, a human embryonic kidney cell line (originally obtained from the American Type Culture Collection), were maintained in DMEM (HyClone) supplemented with 10% fetal bovine serum (HyClone), maintained at 37°C in a humidified incubator containing 5% CO₂. The HEK293 T cell line has extremely high efficiency for plasmid transfection, which is widely used for biochemical assays, such as co-immunoprecipitation assay, in vitro ubiquitylation assay, etc.

The Pbsn-luciferase reporter has been described previously (Wang et al., 2014). The zebrafish ar construct was provided by Dr Zhan Yin (Institute of Hydrobiology, Chinese Academy of Sciences) and was subcloned into pCMV-HA and pCMV-Flag vectors (Clontech Laboratories). All zebrafish ar mutants (lysine-to-arginine) were generated using PCR-based mutagenesis and subcloned into the pCMV-HA vector. Full-length cDNAs of zebrafish nedd8 and mutant nedd8 (nedd8-ΔGG) (1-73 aa) were PCR-amplified and sub-cloned into the pCI-his and pCMV-Myc vectors (Clontech Laboratories). Zebrafish uba3, ubc12, senp8 and spop were PCR-amplified and subcloned into pCMV-HA or pCMV-Flag vectors. All constructs were verified by DNA sequencing.

EPC cells were grown in 24-well plates and transfected with the indicated constructs by VigoFect (Vigorous Biotech), together with pRL-CMV as an internal control. After transfection for 12 h, DHT (40 nM; dissolved in ethanol) was added to the cells and incubated for 12 h. Then the cells were harvested for luciferase assays.

The total cDNA of EPC was used as a template and the full length of EPC spop was cloned by assembling exons and RACE products through overlap PCR. The small interfering RNAs (siRNAs) targeting for spop of EPC cells were designed based on the sequence of EPC spop. The siRNA sequences are: si-spop #1: 5′-GCCTGATGACAAATTGACA-3′; and si-spop #2: 5′-GTGGAAAACGCAGCAGAGATT-3′. The siRNAs for spop and the negative control siRNA (si-nc) were obtained from RiboBio. For knocking down spop, EPC cells were seeded in 6-well plates overnight and transfected with 100 nM siRNAs for spop or the negative control (si-nc) using X-treme GENE HP DNA Transfection Reagent (Roche) following the protocol provided by the manufacturer.

HEK293 T cells were transfected with the indicated constructs for 16-22 h, and then the cells were harvested. For MLN4924 treatments, after the cells were transfected for 12 h, MLN-4924 (1 μM) (Merck) was added into the medium and incubated for 12-14 h, then the cells were harvested. The cells were lysed using the lysis buffer [6 M guanidinium-HCl, 10 mM Tris-HCl, 0.1 M Na₂HPO₄/NaH₂PO₄ (pH 8.0), 10 mM β-mercaptoethanol]. The lysates were mixed with Ni-NTA-agarose beads (Qiagen) pre-washed with lysis buffer and rotated at 4°C overnight. The beads were washed five times using washing buffer [8 M urea buffer (pH 8.0); 8 M urea, 0.1 M Na₂HPO₄/NaH₂PO₄, 0.01 M Tris-HCl, 10 mM β-mercaptoethanol and 0.1% Triton X-100], and washed another five times using washing buffer II [8 M urea buffer (pH 6.3)]. Subsequently, the beads were eluted with the sample-loading buffer and analyzed by western blot.

Zebrafish testes dissected from nearly 100 males were lysed with modified lysis buffer [50 mM Tris (pH 7.4), 150 mM NaCl, 10% glycerol, 1 mM EDTA, 1 mM EGTA, 1% SDS, 1 mM Na₃VO₄, 1 mM DTT and 10 mM NaF] supplemented with a protease inhibitor cocktail. After incubation at 100°C for 10 min, the lysate was diluted 10 times with modified lysis buffer without SDS. The lysates were then incubated with the indicated antibody for 3 h at 4°C. Protein A/G-plus agarose beads (Santa Cruz Biotechnology) were added, and the lysates were rotated gently for 8 h at 4°C. The immunoprecipitates were washed at least three times in wash buffer [50 mM Tris (pH 7.4), 150 mM NaCl, 10% glycerol, 1 mM EDTA, 1 mM EGTA, 0.1% SDS, 1 mM DTT and 10 mM NaF]. Proteins were recovered by boiling the beads in 2×SDS sample buffer and analyzed by western blot. Immunoblotting and co-immunoprecipitation (Co-IP) were performed using the indicated antibodies.

The following antibodies were used for western blot analysis: glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Santa Cruz Biotechnology, sc-47724; 1:2000), Flag (M2; Sigma, 1:5000), anti-HA (Covance, MMS-101R, 1:5000), anti-β-actin (Santa Cruz Biotechnology, sc-47778, 1:1000), anti-Nedd8 (Cell Signaling Technology, #2745, 1:1000), anti-Nedd8 (ABclone, #A1163, 1:1000), anti-ubiquitin (Cell Signaling Technology, P4D1, #3936, 1:1000), anti-Spop (Abcam, #137537, 1:1000). A monoclonal anti-zebrafish Ar antibody was raised against the synthesized peptide corresponding to aa 317 to 500 of zebrafish Ar (Dia-An Inc, 1:50). Anti-Vasa (GTX128306, 1:500), anti-Ziwi and anti-Zili antibodies were used as described previously (Zhu et al., 2019). HEK293 T cells were transfected with different combinations for 24 h, then the cells were lysed in RIPA buffer containing 50 mM Tris (pH 7.4), 1% Nonidet P-40, 0.25% sodium deoxycholate, 1 mM EDTA (pH 8.0), 150 mM NaCl, 1 mM NaF, 1 mM PMSF, 1 mM Na3VO4 and a 1:100 dilution of protease inhibitor mixture (Sigma-Aldrich), After incubation on ice for 1 h, lysates were collected and centrifuged at 10,000 g at 4°C for 15 min. The total cell lysates were boiled with 1×SDS sample loading buffer, separated on SDS-PAGE and transferred to a polyvinylidene difluoride membrane (Millipore). Western blot was performed as described previously (Du et al., 2016). The Fujifilm LAS4000 mini-luminescent image analyzer was used to image the blots.

Total RNA was extracted using RNAiso Plus (TaKaRa) following the protocol provided by the manufacturer. cDNAs were synthesized using the Revert Aid First Strand cDNA Synthesis Kit (Fermentas). SYBR Green mix (Roche) was used for quantitative RT-PCR assays. The primers are listed in Table S1. Actb1 (β-actin) was used as an internal control.

Male nedd8^+/+ (n=6 per group) and nedd8^−/− (n=6 per group) zebrafish (4 mpf) were put into a tank filled with 1 l water respectively and their moving trace within 10 min was recorded respectively using View Point Behavior Technology (Zeb-view). The videos of male zebrafish chasing female zebrafish or chasing food were recorded directly (Movies 1 and 2).

Statistical analysis for sex ratio was performed using Microsoft Excel 2007. Other statistical analysis was performed using GraphPad Prism, v5 (GraphPad Software Inc). Significant differences between groups were determined using Student's t-test (paired or unpaired, as appropriate) or one-way ANOVA followed by Tukey's test for multiple group comparisons. Data are mean±s.e.m. of three independent experiments performed in triplicate. The difference was considered to be significant if the P<0.05 (*P<0.05, **P<0.01, ***P<0.001, ****P<0.0001).

We are grateful to Drs William Tansey, Katja Knauth, Boudewijn Burgering, Bo Zhang, Jingwei Xiong, Zhan Yin, Yonghua Sun and XinHua Feng for their generous gifts of reagents.