Sex differentiation is the process of gonadal development from bipotential gonads to testis and ovary, which is often initiated by upstream sex-determining factors1,2,3,4. In the past two decades, evidence has accumulated that both somatic and germ cells play critical roles in sex differentiation and gametogenesis in the differentiated ovary and testis; however, the exact mechanisms and factors involved remain to be elucidated1,5,6. Studies in mammals have led to the view that the somatic cells differentiate first in response to sex-determining signals, which in turn program the differentiation of the germ cells into male or female gametes to form testis and ovary, respectively7,8,9,10. However, this view has been challenged by some studies in fish models. Evidence from both the zebrafish and medaka models suggests important feminizing roles for the germ cells in gonadal differentiation. In zebrafish, germ line deficient fish generated by either mutation of genes specifically expressed in the germ cells such as the dead end gene (dnd)11,12 or expression of a cellular toxin in germ cells12 all developed into males with testis structure, suggesting important roles for germ cells in ovarian differentiation. Similarly, germ cell deficiency induced by morpholino-mediated knockdown or mutation of germ cell-specific genes in medaka fish resulted in sex reversal of genetic females (XX) to males13,14. A recent study showed that spermatogenesis could take place in the female gonadal environment (ovary) in the mutant of germ cell-specific foxl3, indicating again the importance of germ cell-intrinsic cues for sperm-egg fate decision15.

Zebrafish is a well-known model organism in biological, biomedical, and environmental research6,16,17,18. Interestingly, unlike mammals and some other fish species, the domesticated zebrafish strains used in research do not seem to have any master sex-determining genes, such as Sry in mammals19,20, dmy/dmrt1Y in medaka21,22 and amhy in tilapia23. The sex of zebrafish is therefore determined by a polygenic mechanism, involving multiple genes24,25. The high plasticity of zebrafish gonadal differentiation, which can be influenced by various internal and external factors, makes it an attractive model for investigating the actions and interactions of different factors, including transcription factors such as the factor in the germline alpha (FIGLA) and newborn ovary homeobox gene (NOBOX).

FIGLA and NOBOX are both oocyte-specific transcription factors that play important roles in promoting ovarian development and oogenesis in vertebrates26,27,28,29. FIGLA was first identified in mice for regulating the expression of zona pellucida genes29. Despite being an oocyte-specific factor, the loss of FIGLA in mice did not affect gonadal differentiation at the embryonic stage; however, it caused an arrest of oogenesis at the diplotene stage of meiosis, resulting in failed cyst breakdown and formation of primordial follicles30. NOBOX was also discovered in mice, and it played an important role in the formation of primary follicles26,31. Similar to FIGLA, the NOBOX null mice also exhibited female infertility with early loss of primary follicles and defective folliculogenesis, exhibiting signs of premature ovarian failure (POF) or insufficiency (POI)26,32. Although the expression of NOBOX can be detected in primordial and growing follicles31,33, the early defects in folliculogenesis have limited studies on its roles in late stages.

Both FIGLA (Figla/figla) and NOBOX (Nobox/nobox) have also been studied in teleosts. In medaka, mutation of figla caused defects in germ cell cyst breakdown and therefore follicle formation, similar to that of Figla knockout (KO) mice14,27,30. As for Nobox, the follicles in medaka mutant (nobox−/−) were arrested at the stage of 300 μm-diameter oocytes in the ovary, indicating an essential role for nobox in follicle growth27. In zebrafish, the loss of figla or nobox both resulted in all-male offspring34,35. In figla mutant (figla−/−), the oogenesis was blocked early at the stage of follicle assembly or the transition from cystic prefollicular oocytes at chromatin nucleolar stage (CN, stage IA) to follicular perinucleolar oocytes (PN, stage IB), which was followed by sex reversal to males with normal spermatogenesis, resulting in all-male phenotype34. As for Nobox, the ovary was extremely underdeveloped in the mutant (nobox−/−). In contrast to the figla mutant, early PN follicles at the primary growth (PG) stage could occasionally be observed in the nobox mutant; however, these follicles failed to develop further to form functional ovaries and all individuals also underwent sex reversal to become males with normal spermatogenesis35. Since both figla and nobox are members of the female-promoting pathway, their loss would disrupt the equilibrium between the male and female-promoting pathways, leading to testis development. As the sex change occurred quickly in female mutants of figla and nobox before or shortly after ovarian differentiation due to the high plasticity of gonadal differentiation and relatively dominant male-promoting pathway, it is difficult or impossible to characterize their exact functions in controlling folliculogenesis. This has led us to hypothesize that the female-promoting factors such as figla and nobox could be better studied for their roles in folliculogenesis if the dominance of the male-promoting pathway is alleviated. We have recently tested this idea with aromatase mutant (cyp19a1a−/−) by disrupting the male-promoting gene dmrt136.

Aromatase (cyp19a1a) and doublesex and mab-3-related transcription factor 1 (dmrt1) are important for female and male differentiation, respectively. As in other vertebrates, cyp19a1a is essential for the production of estrogens in fish37, which are important for female gonadal differentiation38. On the other hand, dmrt1 is a critical male-promoting transcription factor essential for testis development and spermatogenesis39. As expected, deletion of the cyp19a1a gene in the zebrafish resulted in an all-male phenotype whereas the loss of dmrt1 gene led to a female-biased sex ratio and underdeveloped testis in males36,40,41,42,43. Interestingly, our recent study showed that simultaneous disruption of dmrt1 and cyp19a1a rescued the all-male phenotype of the cyp19a1a mutant in zebrafish and the follicles in the double mutant (cyp19a1a−/−;dmrt1−/−) could develop well up to the previtellogenic (PV) stage (stage II) with normal formation of cortical alveoli but not yolk granules. This observation suggests that the absence of ovaries in the cyp19a1a−/− mutant is likely due to the early and quick diversion of females to males via sex reversal, making it difficult to evaluate the roles of estrogens in follicle development. This discovery reveals that estrogens are not essential for early follicle development including PG-PV transition and subsequent growth of PV follicles, but they are important for late stages of vitellogenic growth36. This observation supports our view that the female-promoting factors can be better characterized for their roles in controlling folliculogenesis when the male-promoting pathway is attenuated or blocked to delay or prevent the female-to-male sex reversal.

In this study, we attenuated the male-promoting pathway by disrupting dmrt1 to study the differential roles of Figla and Nobox in controlling zebrafish folliculogenesis. We created two double mutants of figla and nobox with dmrt1 (figla−/−;dmrt1−/− and nobox−/−;dmrt1−/−) with the aim to prevent early female-to-male sex reversal so as to allow figla−/− and nobox−/− to fully display their developmental defects in ovarian formation and folliculogenesis. Our data provided strong evidence that both Figla and Nobox play important roles in controlling follicle development; however, they work at different time points of ovarian development and folliculogenesis. Figla is mainly involved in controlling cyst breakdown and follicle formation whereas Nobox controls subsequent follicle development including follicle activation (primary growth-secondary growth transition) and vitellogenic growth. Furthermore, the oocyte specific transcription factor Nobox controls vitellogenic growth by regulating aromatase (cyp19a1a) expression in the follicle cells, which might be mediated by oocyte-secreted signaling molecules, e.g., Gdf9 and Bmp15.


Roles of Figla in follicle formation

Cyst breakdown or follicle assembly is a critical stage in folliculogenesis44. We recently demonstrated in zebrafish that the loss of Figla (figla−/−) prevented cyst breakdown as no individual follicles could form in the mutant, suggesting an important role for Figla in controlling the transition from the cystic oocytes at prefollicular chromatin nucleolar (CN) stage (stage IA) to individual follicles with perinucleolar (PN) oocytes (stage IB). The mutant showed an all-male phenotype in the end34. To further explore the roles of Figla in follicle development, we created a double mutant of figla and dmrt1 genes (figla−/−;dmrt1−/−) to prevent early sex reversal.

Histological analysis of gonadal development at 60 dpf showed well-differentiated ovary and testis in the control fish (figla+/−;dmrt1+/−) with normal sex ratio and gametogenesis. In dmrt1 single mutant (figla+/−;dmrt1−/−), the sex ratio was biased towards females with normal follicle development; however, the males showed underdeveloped and dysfunctional testis with spermatogenesis blocked at early meiotic stage. As observed previously, all figla single mutants (figla−/−;dmrt1+/−) were males showing normal spermatogenesis. Interestingly, all double mutant fish (figla−/−;dmrt1−/−) had underdeveloped testis identical with the males of dmrt1 single mutant. Also, the secondary sex characteristics in figla and dmrt1 double mutants (figla−/−;dmrt1−/−) were all male-like (with breeding tubercles and without genital papilla) at 60 and 150 dpf (Fig. 1a; Supplementary Fig. 1). Importantly, no follicles could be seen in the gonads of the double mutant at 60 dpf, suggesting failed follicle formation as 60 dpf provided enough time for early follicle formation (Fig. 1). These results suggest that disruption of dmrt1 could not rescue the female fate of the figla mutant as seen with cyp19a1a mutant36. The gonads could not develop into functional testis or ovary in the absence of dmrt1 and figla, suggesting critical roles for dmrt1 and figla in spermatogenesis and oogenesis, respectively.

Fig. 1: Phenotype analysis of different genotypes of figla and dmrt1 mutations at 60 dpf.
figure 1

a Gonads and secondary sex characteristics of four different genotypes: normal ovary and testis in the control (figla+/−;dmrt1+/−; n = 15 independent fish, 7 males, 8 females); normal ovary and underdeveloped testis in dmrt1 single mutant (figla+/−;dmrt1−/−; n = 16 independent fish, 2 males, 14 females); all-male testis in figla single mutant (figla−/−;dmrt+/−; n = 13 independent fish, 13 males) and underdeveloped testis in figla and dmrt1 double mutant (figla−/−;dmrt1−/−; n = 15 independent fish). Asterisk, breeding tubercles on pectoral fins; arrowhead, genital papilla. b Sex ratio in four different genotypes at 60 dpf. The sample sizes for independent fish are shown at the top of the columns. PG primary growth, PV previtellogenic, EV early vitellogenic, me meiotic cells, sc spermatocytes, sg spermatogonia, sz spermatozoa.

Roles of Nobox in follicle development

Similar to figla, knockout of nobox, another oocyte-specific transcription factor, also led to an all-male phenotype in zebrafish35. However, in contrast to figla whose mutation caused complete failure of follicle formation from the cystic oocytes, the follicles could occasionally form and develop into typical PN follicles at early primary growth (PG) stage in nobox mutants (nobox−/−), more advanced compared to the figla mutant (figla−/−)34,35. However, the mutant ovaries were extremely underdeveloped and the follicles failed to grow further. The mutant female fish soon underwent sex reversal to become males35. To prevent gonadal masculinization, we created a double mutant with dmrt1 mutation to block gonadal masculinization.

At 30 dpf, PN follicles of early PG stage could be observed in the gonads of about 50% nobox single mutant fish (nobox−/−;dmrt1+/−; 5 females/10), similar to the female ratio in the control fish (nobox+/−;dmrt1+/−; 5/8, 62.5%); however, compared to the control, the follicles in the nobox single mutant were obviously underdeveloped with smaller number and size, suggesting a deficiency in ovarian development (Fig. 2a, b). At 50 dpf, none of the nobox single mutant fish contained follicles in the gonads and all fish were males with normal spermatogenesis (11 males/11) as we reported previously35, suggesting a quick and early sex reversal from females to males. Interestingly, all double mutant fish (nobox−/−;dmrt1−/−) had PN follicles at PG stage in their gonads (8/8) at 30 dpf and these follicles seemed better developed than those in the nobox single mutant. At 50 dpf, the sex ratio of the double mutant fish remained to be female-biased (10 females/12, 83.3%) with well-formed PG follicles in the ovary, in contrast to the all-male phenotype in nobox single mutant (Fig. 2c, d); however, the follicles were all arrested at early PG stage without formation of cortical alveoli, in contrast to those in the control and dmrt1 single mutant (nobox+/−;dmrt1−/−) (Fig. 2a, d). These results suggest that the loss of dmrt1 successfully blocked early sex reversal of nobox mutant to males.

Fig. 2: Rescue of the all-male phenotype of nobox mutant by simultaneous mutation of dmrt1.
figure 2

a Gonadal histology of nobox mutant and double mutant with dmrt1 mutation in early gonadal differentiation (30 and 50 dpf). b Sex ratio in four different genotypes of nobox and dmrt1 mutations at 30 dpf (nobox+/−;dmrt1+/−: n = 8 independent fish; nobox+/−;dmrt1−/−: n = 7 independent fish; nobox−/−;dmrt1+/−: n = 10 independent fish; nobox−/−;dmrt1−/−: n = 8 independent fish). c Sex ratio in four different genotypes of nobox and dmrt1 mutations at 50 dpf (nobox+/−;dmrt1+/−: n = 15 independent fish; nobox+/−;dmrt1−/−: n = 14 independent fish; nobox−/−;dmrt1+/−: n = 11 independent fish; nobox−/−;dmrt1−/−: n = 12 independent fish). d Follicle composition of different genotypes at 30 and 50 dpf (n = 3 independent fish). The horizontal black lines represent the mean. PN perinucleolar oocytes, CN chromatin nucleolar oocytes.

To explore the fate of the arrested follicles in the double mutant ovary, we examined the fish at 120 dpf when the control zebrafish were sexually mature with all stages of follicles present in the ovary. Interestingly, we could commonly see follicles that had entered previtellogenic stage (PV, stage II) with abundant cortical alveoli, indicating that the follicles could overcome the arrest and underwent follicle activation (PG-PV transition) in the absence of Nobox (Fig. 3a). This is similar to our observation in the double mutant of cyp19a1a and dmrt1 (cyp19a1a−/−;dmrt1−/−), but with a significant delay36. What should be noted is that while the follicles could develop to the PV stage, the ovaries of the double mutant were smaller than the well-formed ovaries in nobox+/−;dmrt1+/− controls (Fig. 3a, b). The mutant ovaries showed a progressive degeneration with gradual loss of oocytes and a multitude of empty cavities or vacuoles, presumably the remnants of the degenerated or lost oocytes (Fig. 3a, c). The phenotype of oocyte loss was similar to the observation in Nobox null mice26. The ovarian degeneration was followed by the initiation of sex reversal to males (Fig. 3d). For the convenience of studying and understanding this process, we divided the process of ovarian degeneration in the double mutant into four stages. At Stage I, the ovaries were well formed and structured, containing both PG and PV follicles; however, they were smaller than the control ovaries and contained abundant somatic cells in the interfollicular spaces. Stage II ovaries still contained PG and PV follicles, yet they were markedly smaller in size. At Stage III, the ovaries contained PG follicles only, and at Stage IV, the ovarian tissues had been supplanted by testicular tissues with meiotic cells (Fig. 3).

Fig. 3: Long-term degeneration of the double mutant ovaries.
figure 3

a The ovaries in the double mutant (nobox−/−;dmrt1−/−) showed different degrees of degeneration at 120 dpf with loss of oocytes and decreased ovarian size (n = 5 independent fish). In addition, empty cavities or vacuoles left by degenerated oocytes were often observed (asterisks). The degenerated ovaries were gradually replaced by testicular tissues with meiotic cells (me); however, spermatogenesis could not proceed further due to the lack of dmrt1. The ovarian degeneration process was categorized into four stages based on gonadal size and morphological features. In stage I, the ovary was significantly smaller than the control (approximately half the size), containing both PG and PV follicles. Stage II was characterized by a dramatically reduced ovarian size, while still housing PG and PV follicles. In stage III, the ovary contained PG follicles only. In stage IV, the ovary was devoid of all follicles, featuring empty cavities or vacuoles left by degenerated oocytes. Additionally, testicular tissues began to emerge with meiotic germ cells. b Ovarian sizes of the control and double mutant ovaries undergoing degeneration. The area size of the largest section of each ovary was measured with ImageJ and the data are expressed as the ratios relative to the control. c Number of vacuoles resulting from oocyte loss. The vacuoles were counted on the largest section and classified based on their sizes compared to those of PG and PV follicles. d Composition of gonadal tissues in the control and double mutant ovaries. The areas of different gonadal tissues were measured using ImageJ.

Deficiency of estrogen signaling in nobox mutant females

The blockade of follicle development at PV stage with normal formation of cortical alveoli but without yolk accumulation in the double mutant oocytes (nobox−/−;dmrt1−/−) raised an interesting question about the involvement of estrogen signaling as estrogen is essential for vitellogenin production and therefore the transition from PV to early vitellogenic (EV) stage45,46. To test this hypothesis, we performed a series of experiments.

First, we examined the genital papillae at 100 dpf as this female secondary sex characteristic is estrogen-dependent47,48. The genital papillae formed well in control females; however, they were significantly regressed in female double mutants (nobox−/−;dmrt1−/−) with shortened lengths (Fig. 4a, b). Second, we assessed the expression of cyp19a1a in the ovary and determined the estradiol (E2) concentrations in the serum of double mutant females. Both the mRNA expression of cyp19a1a in PV follicles and serum estradiol level were significantly lower in double mutant females (nobox−/−;dmrt1−/−) than the control (nobox+/−;dmrt1+/−) and dmrt1 single mutant (nobox+/−;dmrt1−/−). There was no significant difference in cyp19a1a expression at PG stage, probably because its expression level was very low or barely detectable at this stage49 (Fig. 4c, d). These lines of evidence indicate that estrogen synthesis was significantly reduced in the double mutant (nobox−/−;dmrt1−/−). Third, we created a triple mutant of nobox, cyp19a1a and dmrt1 genes (nobox−/−;cyp19a1a−/−;dmrt1−/−). Compared with cyp19a1a−/−;dmrt1−/−, the ovaries from the double and triple mutants containing nobox−/− (nobox−/−;dmrt1−/− and nobox−/−;cyp19a1a−/−;dmrt1−/−) were generally smaller (Fig. 5a, b), which was also reflected in body weight (Fig. 5c). Histological analysis showed that the triple mutant phenocopied the double mutant (nobox−/−;dmrt1−/−) at 100 dpf with no additional impact and the follicles could enter PV stage but not vitellogenic stage in both mutants. Although the triple mutant phenocopied the double mutant of cyp19a1a and dmrt1 (cyp19a1a−/−;dmrt1−/−) in terms of follicle activation characterized with the formation of cortical alveoli, its ovary was underdeveloped with fewer follicles (Fig. 5b). The mutations of these genes did not seem to have any pleiotropic effects as no significant difference was observed in body length (Fig. 5d).

Fig. 4: Evidence for estrogen deficiency in double mutant females.
figure 4

a Genital papilla in the control (nobox+/−;dmrt1+/−; n = 3 independent fish) and double mutant (nobox−/−;dmrt1−/−; n = 3 independent fish) females (arrow). The dotted line showed the length of the genital papilla. b Length of genital papilla in the controls and double mutant females (n = 3 independent fish). Data shown are mean ± SEM (P = 0.0088 by unpaired Student’s two-tailed t test). c Expression of cyp19a1a in PG and PV follicles (n = 5 independent samples). Total RNA was extracted from the isolated PG and PV follicles and reverse transcribed into cDNA for real-time PCR analysis. Each data point represents PG or PV follicles isolated and pooled from two fish for each genotype. Data shown are mean ± SEM, P values revealed by one-way ANOVA and Tukey’s test. d Serum E2 levels in the control (nobox+/−;dmrt1+/−; n = 5 independent fish), dmrt1 single mutant (nobox+/−;dmrt1−/−; n =5 independent fish) and double mutant (nobox−/−; dmrt1−/−; n = 6 independent fish) females at 100 dpf. Data shown are mean ± SEM (nobox+/−;dmrt1+/− v.s nobox+/−;dmrt1−/−: P = 0.6649; nobox+/−;dmrt1+/− v.s nobox−/−;dmrt1−/−: P = 0.0095; nobox+/−;dmrt1−/− v.s nobox−/−;dmrt1−/−: P = 0.0018; P values revealed by one-way ANOVA and Tukey’s test).

Fig. 5: Ovarian growth and follicle development in the triple mutant of cyp19a1a, nobox and dmrt1.
figure 5

a Gross morphology of the ovaries (arrow) in females of the control (nobox+/−;cyp19a1a+/−;dmrt1+/−, n = 3 independent fish), cyp19a1a and dmrt1 double mutant (nobox+/−;cyp19a1a−/−;dmrt1−/−, n = 3 independent fish), and nobox and dmrt1 double mutant (nobox−/−;cyp19a1a+/−;dmrt1−/−, n = 3 independent fish). b Ovarian histology and secondary sex characteristics in different genotypes: control (nobox+/−;cyp19a1a+/−;dmrt1+/−, n = 3 independent fish), cyp19a1a and dmrt1 double mutant (nobox+/−;cyp19a1a−/−;dmrt1−/−, n = 3), nobox and dmrt1 double mutant (nobox−/−;cyp19a1a+/−; dmrt1−/−, n = 3 independent fish), and nobox, cyp19a1a and dmrt1 triple mutant (nobox−/−;cyp19a1a−/−; dmrt1−/−, n = 3 independent fish). Red asterisk, breeding tubercles; arrowhead, genital papilla; black asterisk, vacuoles left by the lost oocytes. c, d Body weight and body length of the triple mutant of cyp19a1a, nobox and dmrt1 (n = 6 independent fish). Data shown are mean ± SEM, P values revealed by one-way ANOVA and Tukey’s test.

Finally, to provide direct and conclusive evidence for the involvement of estrogen signaling in the follicle blockade at PV-EV transition in the double mutant (nobox−/−;dmrt1−/−), we performed an experiment to treat both double mutants nobox−/−;dmrt1−/− and cyp19a1a−/−;dmrt1−/− with E2, the major endocrine hormone that promotes vitellogenesis50. To avoid the toxic effects of estrogens on follicle development at high dosage51,52, we adopted the treatment scheme optimized in our recent study52. The fish were treated with E2 by oral administration of E2-containing diet (2 μg/g) for 15 days from 70 to 85 dpf (Fig. 6a). Histological analysis showed that E2 treatment could rescue the PV-EV blockade with yolk accumulation resumed in both double mutants (Fig. 6b). The follicles in the double mutant cyp19a1a−/−;dmrt1−/− could develop to the full size of FG stage, similar to the follicles in the control fish. In contrast, although vitellogenic growth also resumed in the double mutant nobox−/−;dmrt1−/−, the follicles could only grow to the range of mid-vitellogenic (MV) stage (Fig. 6c). In agreement with the histological observations, the E2-treated cyp19a1a−/−;dmrt1−/− females could spawn to produce fertilizable eggs (Fig. 6d) and live offspring (Fig. 6e); however, the E2-treated nobox−/−;dmrt1−/− females were infertile (Fig. 6d).

Fig. 6: Rescue of vitellogenic growth in the double mutants (nobox−/−;dmrt1−/− and cyp19a1a−/−;dmrt1−/−) by E2 treatment.
figure 6

a Schematic illustration of E2 treatment. b Ovarian histology of different genotypes in the control group with normal feeding and the E2 treatment group fed with an E2-containing diet (n = 3 independent fish). Vitellogenic growth characterized with yolk granule accumulation resumed in both double mutants (nobox−/−;dmrt1−/− and cyp19a1a−/−;dmrt1−/−), which were blocked at the PV stage with the formation of cortical alveoli but not yolk granules. c Follicle composition of different genotypes in different treatment groups (n = 3 independent fish). The data points shown are diameters of individual follicles and the statistical significance of the means was demonstrated by unpaired Student’s two-tailed t test. d Fecundity of different genotypes and treatments. Each data point represents the number of eggs spawned by each female mated with one wild-type male (Control group: n = 4 independent experiments; Other groups: n = 3 independent experiments). The sexes of examined females were further confirmed by histology after mating. Data shown are mean ± SEM, P values revealed by one-way ANOVA and Tukey’s test. e The offspring from E2-treated females of cyp19a1a−/−;dmrt1−/− double mutant.

Potential roles of Gdf9 and Bmp15 in mediating Nobox actions

Nobox is an oocyte-specific transcriptional factor while estrogens are produced in the surrounding follicle cells by aromatase (cyp19a1a)26,33,53. This suggests that the reduced expression of cyp19a1a in nobox mutant must be mediated by other signaling molecules from the oocyte. As growth differentiation factor 9 (GDF9) and bone morphogenetic protein 15 (BMP15/GDF9B) are the two best characterized oocyte-derived growth factors54 and they both could be regulated by NOBOX in mice26,55,56,57, we hypothesized that these TGF-β family members could be potential factors that mediate Nobox regulation of aromatase expression. In support of this idea were recent studies showing downregulation of cyp19a1a in bmp15 null zebrafish58. To provide further evidence for the possible involvement of Gdf9 and Bmp15 in Nobox regulation of cyp19a1a, we examined the expression of gdf9 and bmp15 in the double mutants (nobox−/−;dmrt1−/−) (no females available in nobox single mutant). The results showed that both genes were almost shut off in PG and PV follicles in the absence of nobox (nobox−/−;dmrt1−/−) compared to the control (nobox+/−;dmrt1+/−) and dmrt1 single mutant (nobox+/−;dmrt1−/−) (Fig. 7).

Fig. 7: Expression of gdf9 and bmp15 genes in PG and PV follicles of nobox−/−;dmrt1−/− double mutant at 100 dpf.
figure 7

Total RNA was extracted from the isolated PG and PV follicles and reverse transcribed into cDNA for real-time PCR analysis. Each data point represents PG or PV follicles isolated and pooled from two fish for each genotype, totaling 10 fish (n = 5 independent samples). The mRNA levels of each target gene were normalized to that of the housekeeping gene ef1a, and expressed as a fold change relative to the control. Data shown are mean ± SEM, P values revealed by one-way ANOVA and Tukey’s test.


Our previous studies showed that knocking out the oocyte-specific transcription factor figla or nobox led to an all-male phenotype in zebrafish34,35, suggesting crucial roles for these germ cell factors in sex differentiation and gametogenesis, especially folliculogenesis. The all-male phenotype in figla and nobox mutants has prevented us from understanding their functions in folliculogenesis due to early differentiation of the juvenile ovary to males. To explore the roles of figla and nobox, especially their differential roles in folliculogenesis, we weakened the dominance of the male-promoting pathway by deleting dmrt1 gene, which is essential for male differentiation and spermatogenesis in vertebrates36,59. We have successfully used this approach to study ovarian aromatase (cyp19a1a), leading to the discovery that cyp19ala and estrogens are essential for ovarian differentiation but not for early folliculogenesis from PG to PV stage36. Two double mutants were created in the present study (figla−/−;dmrt1−/− and nobox−/−;dmrt1−/−) for analysis, which has further explored the differential roles of the two transcription factors in controlling folliculogenesis and their action mechanisms, especially Nobox.

Zebrafish is a juvenile hermaphrodite with all individuals forming ovary-like gonads first before developing further into functional ovaries and testes60. In the juvenile ovary, the germ cells initiate meiosis to become oocytes at chromatin nucleolar (CN) stage (stage Ia) and in some fish perinucleolar (PN) stage (stage Ib) as well34,61. The CN oocytes are clustered in germ cell cysts with synchronous development in each cyst, and some oocytes may develop into PN follicles after the process of cyst breakdown or follicle assembly to become individual follicles. How these events are regulated remains largely unknown. Our recent studies demonstrated that two oocyte-specific transcription factors, Figla and Nobox, played important roles in juvenile ovary to regulate ovarian differentiation and follicle formation.

In figla-null zebrafish, the germ cells of CN stage remained in cystic clusters undergoing meiosis; however, these oocyte-like cells could not form individual follicles, indicating failed cyst breakdown or follicle formation. All mutant fish turned into males quickly, generating an all-male phenotype34. Similar to figla mutant, the nobox-null zebrafish also showed extremely underdeveloped ovaries in juveniles; however, individual follicles of PN stage could occasionally form in the mutant, in sharp contrast to figla-null fish. Unfortunately, further assessment of the functional importance of Nobox in regulating subsequent follicle development was not possible due to the early and quick masculinization into males, similar to the figla mutant35. To evaluate roles especially differential roles of Figla and Nobox in controlling folliculogenesis during and after follicle assembly, we attenuated the male-promoting pathway so as to allow the mutant females of figla and nobox to fully display their defects in follicle development without early conversion to males. We have used this approach to discover that aromatase (cyp19a1a) and therefore estrogens are not essential for the development of follicles from PG to PV stage, in contrast to the traditional views. The loss of cyp19a1a gene also resulted in an all-male phenotype with PN follicles occasionally formed in the juvenile ovaries42, similar to the nobox mutant35. Interestingly, the ovary could form normally and folliculogenesis resumed in the double mutant of cyp19a1a and dmrt1 (cyp19a1a−/−;dmrt1−/−), which prevented early sex reversal from females to males36.

Using the same approach, we created two double mutant fish in the present study, figla−/−;dmrt1−/− and nobox−/−;dmrt1−/−, and analyzed their folliculogenesis. The results showed that figla−/−;dmrt1−/− exhibited the same phenotype as that seen in the figla single mutant (figla−/−), i.e., cystic CN oocytes without formation of PN follicles. In contrast, the PN follicles could form adequately in the females of nobox−/−;dmrt1−/−. Not only could these follicles develop beyond the PG stage, but they could also progress to the PV stage, marked by the formation of cortical alveoli. This indicates a successful PG-PV transition or follicle activation. Despite these observed follicle developments, the ovaries of the double mutant were significantly smaller than those of the control, containing far fewer follicles. This reduction may be partially due to a loss of oocytes, as suggested by the presence of numerous vacuoles in the mutant ovaries. These results indicate clearly that Figla and Nobox are both important for folliculogenesis, albeit at different stages of the process. Figla primarily influences cyst breakdown and follicle formation, while Nobox acts as a promoter and stabilizer for subsequent progression of follicle development, especially the transition from PV to EV stage. This agrees well with the genetic studies in the mouse model62,63, suggesting functional conservation of these oocyte-specific transcription factors in vertebrates. Also, the expression of Figla was normal in Nobox-null ovary in mice and medaka;26,27 however, no expression of Nobox could be detected in the Figla-null gonads27,64. There results suggest that FIGLA might be an upstream regulator of Nobox expression. Alternatively, FIGLA disruption might result in a complete failure of folliculogenesis, therefore preventing Nobox expression. More studies are needed in different models to address this issue, including zebrafish.

Interestingly, nobox−/−;dmrt1−/− displayed a similar phenotype to that of cyp19a1a−/−;dmrt1−/− in that follicles could form and develop to PV stage with normal formation of cortical alveoli. However, the PV follicles in both double mutants could not develop further into the vitellogenic growth phase due to the lack of yolk accumulation. Since vitellogenesis in non-mammalian vertebrates is estrogen-dependent, the similarity between nobox−/−;dmrt1−/− and cyp19a1a−/−;dmrt1−/− raised an interesting question about the possible involvement of estrogens in Nobox regulation of follicle development.

To address this issue, we performed a series of experiments or analyses. First, we determined cyp19a1a expression in the ovary and E2 level in the serum. Both decreased significantly in nobox−/−;dmrt1−/− females compared to the control fish and dmrt1 single mutant. In agreement with this was the regression of the female secondary sex characteristics, genital papilla, in nobox−/−;dmrt1−/− females. The most direct and conclusive evidence for roles of estrogen signaling in Nobox actions was the observation that treatment with E2 could rescue the phenotypes of not only nobox−/−;dmrt1−/− but also cyp19a1a−/−;dmrt1−/−. The PV follicles in both mutants resumed vitellogenesis by accumulating yolk mass in the oocytes in response to E2 supplementation.

Interestingly, although E2 treatment could rescue vitellogenic growth in the double mutant nobox−/−;dmrt1−/−, the follicles could only develop maximally to the MV stage, in contrast to the double mutant cyp19a1a−/−;dmrt1−/−, whose follicles could develop to the final FG stage in response to E2 and undergo oocyte maturation and ovulation. The blockade of follicle development at MV stage or MV-LV transition in nobox−/−;dmrt1−/− after E2 treatment suggests that in addition to regulating cyp19a1a, Nobox may also control other regulatory factors, which are essential for follicle development beyond the MV stage. The identity of these factors remains unknown. Interestingly, similar follicle arrest at MV stage has also been reported in bmpr2b mutant as well as the double mutants gdf9−/−;inha−/− and bmp15−/−;inha−/−52,65,66, suggesting potential roles for BMP family members in promoting follicle growth at the MV stage. This would be an interesting issue to explore in future studies.

Since nobox and cyp19a1a are expressed in two different compartments of the follicle, namely oocyte and follicle cells respectively, it is unlikely that Nobox could directly regulate cyp19a1a expression. Instead, the regulation is likely mediated by factors secreted by the oocyte. Although the identity of such factors remains unknown, the potential candidates include Gdf9 and Bmp15, which are the two best characterized oocyte-specific growth factors in mammals54,67. To provide supportive evidence for this, we examined the expression of gdf9 and bmp15 in the ovary of nobox−/−;dmrt1−/−. The data showed that the expression of these two genes was almost shut off in the mutant ovary. Further studies are needed to verify their roles in mediating Nobox control of cyp19a1a expression.

In mammals, Nobox has been reported to be a potential transcription factor to regulate the expression of GDF9 and BMP15. Knockout of Nobox gene in mice resulted in a complete loss of expression of both Gdf9 and Bmp15 together with other oocyte-specific genes26. NOBOX binding elements (NBEs) have been identified in the promoter region of mouse Gdf9 gene and their functionality has been confirmed by luciferase reporter and ChIP assays56. As for BMP15, although its expression was shut off in Nobox-null mice26, direct evidence is needed to support NOBOX regulation of Bmp15 expression at the transcription level. Our data in the present study also implicate Gdf9 and Bmp15 in mediating Nobox regulation of the follicle cells. Both genes were significantly downregulated in the double mutant ovary (nobox−/−;dmrt1−/−). Further evidence for the role of Bmp15 in relaying signals of oocyte specific Nobox to the surrounding follicle cells came from recent studies on bmp15 in zebrafish. Knockout of zebrafish bmp15 gene resulted in a complete arrest of follicles at PV stage with normal formation of cortical alveoli but no yolk accumulation52,58. Our genetic and pharmacological experiments provided clear evidence that such arrest of follicle development in bmp15-null zebrafish was due to reduced expression of cyp19a1a and therefore estrogen production52. However, more evidence is needed to confirm the regulation of Gdf9 and Bmp15 (or other molecules) by Nobox and their roles in mediating oocyte regulation of follicle cell functions including cyp19a1a expression and estrogen biosynthesis (Fig. 8).

Fig. 8: A working model on differential roles of Figla and Nobox in zebrafish folliculogenesis.
figure 8

Figla plays a critical role in controlling follicle formation in the event of cyst breakdown or follicle assembly, while Nobox is more involved in regulating follicle development after its formation, including such events as follicle activation (PG-PV transition) and vitellogenic growth (PV-EV transition). Nobox controls vitellogenic growth by regulating aromatase (cyp19a1a) expression in the follicle cells, which may be mediated by oocyte-secreted signaling molecules such as Gdf9 and Bmp15.

In summary, we performed genetic analysis for roles of oocyte-specific transcription factors figla and nobox in zebrafish folliculogenesis. By attenuating the male-promoting pathway via deleting dmrt1 gene, we were able to evaluate the functions of both Figla and Nobox in controlling folliculogenesis without being interrupted by early sex reversal to males. Our data confirmed that Figla is essential for germ cell cyst breakdown or follicle assembly whereas Nobox is important for the subsequent progression of follicle development, especially the transition from previtellogenic stage to vitellogenic growth (PV-EV transition). We provided further evidence that the Nobox-dependent vitellogenic growth was due to a deficiency in estrogen production. This study has provided strong evidence in a fish model for the important roles of oocyte in orchestrating folliculogenesis (Fig. 8).



The mutant zebrafish lines for cyp19a1a (umo5), figla (umo14), dmrt1 (umo15), nobox (umo36) were produced in our previous studies34,35,36,42. The fish were reared at 28 ± 1 °C with a photoperiod of 14-h light and 10-h dark in a flow-through aquarium system (Tecniplast, Buguggiate, Italy). The larvae were reared in nursery tanks with paramecia and artemia before transfer to the aquarium system, and the adults were fed with artemia and Otohime fish diet (Marubeni Nisshin Feed, Tokyo, Japan). Zebrafish husbandry and all experiments were conducted in full accordance with animal care and use guidelines with ethical approval by the Research Ethics Committee of the University of Macau (Approval. No. AEC-13-002).


Genomic DNA from an embryo or a small piece of the caudal fin was extracted by alkaline lysis68. The sample was incubated in 30–50 µl NaOH (50 nmol/µl) at 95°C for 10 min to extract the genomic DNA. Then, 3–5 µl Tris-HCl (pH 8.0) was added for neutralization. The extract was used for high-resolution melting analysis (HRMA), and the melt curves were analyzed with the Precision Melt Analysis software (Bio-Rad, Hercules, CA)68. The primers used for genotyping are listed in Supplementary Table 1.

Sampling and histological examination

The fish were sampled at different time points for phenotype analysis. The fish were sacrificed after anaesthetization with MS222 (Sigma, St. Louis, MO). The gross morphology of each fish was photographed with a digital camera (Canon EOS 700D). The pectoral fins and cloaca of each sampled fish were examined on the Nikon SMZ18 dissecting microscope (Nikon, Tokyo, Japan) for breeding tubercles and genital papilla and photographed with the Digit Sight DS-Fi2 camera (Nikon).

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

Follicle staging

Follicles were staged according to size and morphological features such as cortical alveoli and yolk granules as reported69,70,71,72. We divide follicles into six stages: primary growth (PG, <150 µm), previtellogenic (PV, ~250 µm), early vitellogenic (EV, ~350 µm), mid-vitellogenic (MV, ~450 µm), late vitellogenic (LV, ~550 µm) and full-grown (FG, >650 µm).

Fertility assessment

Female fertility was assessed by the number of eggs spawned in natural breeding with males. The female fish from different groups were paired with wild-type (+/+) males individually for natural spawning. After spawning in the morning, the number of eggs released by each female was counted.

Sex identification

In general, the sex was identified according to dimorphic morphological features, such as body shape, fin color, and genital papilla73, and, if necessary, by dissection under the Nikon SMZ18 stereomicroscope (Nikon). The sex identity was confirmed by histology at the end of sampling.

RNA extraction and quantitative real-time PCR

Total RNA was extracted from isolated PG and PV follicles74 using TRIzol (Invitrogen, Waltham, MA) according to the manufacturer’s protocol. 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 Supplementary Table 1. Melting curve analysis was performed to demonstrate primer specificity. A standard curve was included in each PCR assay for quantification. The relative gene expression level was normalized to the level of ef1a in each sample and expressed as fold change compared to the control group.

Measurement of serum E2

After anaesthetization with MS222, the blood was gently collected from the heart of each fish using a 10-μL tip and transferred into a 1.5-mL tube. The samples were left at room temperature for 1 h to separate the serum. The supernatants were collected after centrifugation (3000 rpm, 30 min, 4 °C). The levels of E2 in the serum were measured using an ELISA kit (Neogen Corporation, Lansing, MI; RRID:AB_2935669) according to the manufacturer’s instructions.

Oral administration of E2

Different groups of females were treated with E2 for 15 days from 70 to 85 dpf. In brief, the fish were fed twice a day with the E2-containing Otohime fish diet (0 or 2 μg/g), each at 5% of the total body weight in the tank (10% per day in total). During the treatment period, the water was renewed daily to maintain good water quality.

Statistics and Reproducibility

All values are presented as the mean ± sem. The data were statistically analyzed by Student’s t-test or one-way ANOVA using Prism 9 (GraphPad Prism, San Diego, CA). The significance is shown by the P value in the figures (P > 0.05 or ns, not significant). The sample sizes represent independent biological replicates. All experiments were performed at least twice.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.