Loss of dmrt1 restores zebrafish female fates in the absence of cyp19a1a but not rbpms2a/b

Although sex determination is a common biological process crucial to reproduction and, thus, species survival in sexually reproducing organisms, the mechanisms controlling this process vary extensively among animal species. The decision to develop as one sex or the other is triggered in different species by many environmental factors, including temperature and nutrition status, presence of sex chromosomes and the function of many sex-associated genetic factors that converge upon regulation of downstream sexual differentiation factors, for example, Sox9 (SRY-box 9), Foxl2 (forkhead box L2), dmrt1 (doublesex and mab-3 related transcription factor 1) and Cyp19a1a (aromatase; cytochrome P450, family 19, subfamily a) (Kossack and Draper, 2019; Matson et al., 2011; Uhlenhaut et al., 2009; Lau et al., 2016). In domesticated laboratory strains of zebrafish, sex determination appears to be polygenic (Anderson et al., 2012; Liew et al., 2012; Moore and Roberts, 2013; Wilson et al., 2014), being influenced by numerous genetic factors. Like most teleost fish, zebrafish are a gonochoristic species, eventually differentiating into one of two sexes, and in general do not switch sex as adults (Avise and Mank, 2009). However, the zebrafish juvenile gonad, which is initially undifferentiated, further develops into a bipotential ovary that contains early-stage oocytes but remains poised to develop as either an ovary or a testis (Takatsu et al., 2013). This plasticity provides an excellent model for studying genetic factors and interactions between essential regulators of sex-specific differentiation among vertebrates.

Factors regulating sex differentiation and maintenance vary in function, from transcription factors regulating sex-specific programs, to regulators of signaling molecules, and even signaling molecules themselves. Doublesex and Mab-3 related transcription factor 1 (Dmrt1) is a transcription factor crucial for male gonad differentiation and maintenance across vertebrates (Matson et al., 2010, 2011; Webster et al., 2017; Kopp, 2012; Kim et al., 2007; Raymond et al., 2000). In mammals, Dmrt1 acts by directly activating male-determining genes and simultaneously suppressing female-determining genes (Matson et al., 2011; Murphy et al., 2010). Recent studies have verified that dmrt1 is expressed in both the germline and somatic gonad in zebrafish and, as in mammals, is required for male differentiation, as mutants disrupting Dmrt1 cause female-biased sex ratios (Webster et al., 2017; Lin et al., 2017). Aromatase, encoded by cyp19, required for the final steps of estrogen synthesis (Bulun et al., 1994; Simpson et al., 1994), is a crucial factor for sex differentiation in fish (Takatsu et al., 2013; Dranow et al., 2016). Zebrafish have two copies of cyp19 (ovarian cyp19a1a and brain cyp19a1b) and, in contrast to dmrt1, knockout of ovarian cyp19a1a results in an all-male phenotype that is characterized by early transition to male gonad features with an apparent absence of a bipotential ovary stage of development (Yin et al., 2017).

Bone morphogenetic protein 15 (Bmp15) is a signaling molecule expressed primarily by the oocyte in mice and zebrafish (Dranow et al., 2016; Clelland et al., 2006; Dube et al., 1998). Bmp15 has recently been shown to be required in zebrafish for maintenance of the female gonad, after sex has been determined, and has been postulated to promote specification of cyp19a1a, producing somatic fates that promote further oocyte and follicle development (Dranow et al., 2016). Knockout of bmp15 in zebrafish causes early switching of mutant sex from female to male, and thus results in recovery of only fertile adult males (Dranow et al., 2016). Although work in mammalian systems indicates that TGF-β signaling, particularly GDF-9 and BMP15, are crucial for granulosa cell fate and development (Peng et al., 2013a; Otsuka et al., 2011; Su et al., 2004), Gdf9 appears to be dispensable in zebrafish because loss of function causes no overt phenotypes and does not worsen phenotypes caused by loss of bmp15 (Dranow et al., 2016).

Recent work from our laboratory has identified an RNA-binding protein, RNA-binding protein of multiple splice forms 2 (Rbpms2), as a crucial germline-expressed factor for female sex differentiation in zebrafish (Kaufman et al., 2018). Rbpms2 is conserved among vertebrates and has been shown to regulate Bmp in the gastrointestinal tract and smooth muscle plasticity in mammalian systems (Nakagaki-Silva et al., 2019; Sagnol et al., 2014; Notarnicola et al., 2012), and to play a significant role in zebrafish cardiac development (Kaufman et al., 2018). More specifically, RBPMS2 is expressed in chick visceral smooth muscle cell (SMC) precursors and blocks SMC differentiation by promoting expression of the BMP inhibitor noggin RNA (Notarnicola et al., 2012). In addition, based on studies in HEK 293T cells, RBPMS has been postulated to promote nuclear accumulation of Smad2/3 and transcriptional activation of Smad2/3 targets by a mechanism that requires Smad2/3 phosphorylation (Sun et al., 2006). Although rbpms2 is expressed in many cell types, including primary oocytes in mammals (Bgee version 14.1; Ensembl ID: ENSMUSG00000032387), the function of Rbpms2 in mammalian sex determination has not been investigated. Our laboratory has demonstrated that simultaneous loss of both redundant copies of rbpms2 [rbpms2a and rbpms2b, i.e. in rbpms2 double mutants (DMs), referred to here as rbpms2DMs] results in recovery of exclusively fertile males (Kaufman et al., 2018); however, how Rbpms2 promotes female fates is not clear. It has also been shown that both the somatic gonad and meiotic oocytes contribute to establishing and maintaining the female gonad in domesticated zebrafish (reviewed by Kossack and Draper, 2019). Therefore, it is expected that the sex-specific identities of the somatic gonad and germ cells must match to develop a functional gonad successfully.

Although crucial roles for each of the above-mentioned genes in determining sexual fate in zebrafish have been established, and it is known that female and male pathways function antagonistically (Kossack and Draper, 2019; Herpin and Schartl, 2015), precisely where each factor resides in the hierarchy of genes regulating germ cell and somatic gonad sex and their interactions in gonadal differentiation remains incompletely understood. In this work, we surveyed known regulators of sex determination to identify potential Rbpms2 targets. Specifically, we used genetic epistasis experiments and cell biological approaches to tease apart the genetic hierarchy of these crucial factors in sex determination. We provide evidence that TGF-β signaling is activated in early germ cells of the bipotential ovary and in wild type becomes activated in the somatic gonad upon sexual differentiation, and that in rbpms2DMs initial activation of TGF-β in the bipotential germline is intact. Because zebrafish first establish a bipotential ovary, based on the presence of histologically evident early oocytes that eventually differentiate as an ovary or a testis, we reasoned that antagonism of male promoting factors such as Dmrt1 would be key to female-specific differentiation.

Surprisingly, we found that loss of dmrt1 was not sufficient to suppress the male-only differentiation phenotype of the rbpms2DMs, thus rbpms2 is epistatic to dmrt1 in terms of adult sex. Moreover, Rbpms2 has a role beyond simply repressing or antagonizing Dmrt1, as it is required to promote female fates and for female sex-specific differentiation even in the absence of Dmrt1. Interestingly, unlike either rbpms2DMs, which develop as fertile males, or dmrt1^uc27 mutants, which develop as fertile females, the triple mutants (rbpms2;dmrt1TMs) were sterile in breeding and immunohistological assays, indicating that germ cells might be lost as a consequence of cell autonomous failure to establish sex-specific identity, later requirement for dmrt1 in germline maintenance, nonautonomous defects in somatic gonad development, or mismatched germ cell and somatic gonad sex-specific identity. Further demonstrating the distinct contribution of Rbpms2 to promoting female fates, we found that loss of dmrt1 was sufficient to suppress the early male-differentiating phenotype of cyp19a1a mutant gonads, as evidenced by the presence of oocytes containing the female-specific marker Buckyball and morphologically normal Balbiani bodies in cyp19a1a;dmrt1DM gonads (Heim et al., 2014). In contrast, early cyp19a1a single mutant gonads prematurely take on a male identity and fail to develop any oocytes. Taken together these findings indicate that cyp19a1a acts during at least two steps of female-specific differentiation. First, cyp19a1a-mediated suppression of dmrt1 is key to establish a bipotential ovary and initiate female fate acquisition in zebrafish, possibly by promoting rbpms2, which is required for female-specific differentiation, even in the absence of Dmrt1. Ultimately, female fates are not maintained in cyp19a1a;dmrt1DMs, probably due to the later Bmp15-dependent expression of Cyp19a1a that is required for subsequent follicle differentiation and oocyte maintenance by a mechanism that is independent of inhibition of Dmrt1. In support of this notion, maintenance of female fate can be prolonged in bmp15 mutants that also lack dmrt1; however, these oocytes fail to progress to vitellogenic stages and bmp15;dmrt1DMs eventually switch to male fate.

Because rbpms2DMs can develop as fertile males that complete meiosis and spermatogenesis but fail to complete female differentiation (Kaufman et al., 2018), we reasoned that rbpms2 is not required for meiosis per se, but instead is required to prevent the germline from adopting a male fate. This could occur by promoting expression of female factors, repressing male factors or a combination of both. Because Rbpms2 is an RNA-binding protein, we surveyed genes required for meiosis, oogenesis and sex determination for presence of the previously identified Rbpms RNA-binding site consensus sequence, CAC(n3-12)CAC (Farazi et al., 2014). The presence of a Rbpms RNA-binding site indicates that an RNA might be a potential target for Rbpms2 regulation. Targets involved in gonad development (particularly those with similar phenotypes to rbpms2) are compelling targets for investigation. We identified five or more Rbpms RNA-binding sites within the untranslated regions (UTRs) and exons of the male fate regulator dmrt1; cyp19a1a; the TGF-β signaling molecule, bmp15; and bmp receptors, bmpr1ab, bmpr1bb and bmpr2b (Table 1), making them compelling targets for regulation by Rbpms2.

Because Rbpms2 regulates BMP signaling in other contexts, RNAs encoding Bmp15 ligand and Bmp receptors were identified as potential targets of Rbpms2 regulation based on the presence of Rbpms RNA-binding sites, and because Bmp15 is required in zebrafish for maintenance of the female gonad later in development, we examined bmpr expression in rbpms2 mutants. We previously showed that cellular features of rbpms2DM juvenile gonads are comparable to wild-type gonads (Kaufman et al., 2018). To determine whether rbpms2DM juvenile gonads resemble the wild type molecularly, with respect to Bmp-mediated signaling, we prepared cDNA from trunks at day 21 post fertilization (d21), prior to sex-determination, and used RT-PCR to examine zebrafish bmp receptors known to be expressed in the gonad, bmpr1ab, bmpr1bb, bmpr2b (Dranow et al., 2016), and ef1alpha (also known as eef1a) as control. We observed comparable expression of bmpr transcripts between Rbpms2 mutants and wild-type siblings (Fig. 1A).

In mammalian systems, TGF-β signaling mediated by GDF-9 and BMP15 form an oocyte-granulosa cell feedback loop (Otsuka et al., 2011; Su et al., 2004; Matsumoto et al., 2012; Tsukamoto et al., 2011a,b; Peng et al., 2013b; Wigglesworth et al., 2013). By contrast, in zebrafish, Gdf9 appears to be dispensable because gdf9 loss neither causes overt phenotypes nor worsens phenotypes caused by loss of bmp15 (Dranow et al., 2016); thus, the role of TGF-β in the differentiating zebrafish gonad remains unclear. To investigate potential involvement of TGF-β, we examined the localization of p-Smad2, an indicator of active TGF-β signaling (de Sousa Lopes et al., 2003; Nakao et al., 1997; Zhang et al., 1996). In bipotential (d21) wild-type (Fig. 1B) and rbpms2DM (Fig. 1C) juvenile gonads, p-Smad2 was detected with DAPI in the nuclei of germ cells marked by ziwi-GFP (Leu and Draper, 2010). As differentiation proceeded in wild-type females, p-Smad2 remained strong in the nuclei of early oocytes and became activated in the nuclei of somatic gonad cells (Fig. 1D). Nuclear p-Smad2 was also detected in the somatic gonad of wild-type (Fig. 1E) and rbpms2DM (Fig. 1F) males, although its presence appeared weaker than in females and declined in differentiating spermatogonia. Based on these observations, we conclude that TGF-β is active in the early germline and soma of the differentiating gonads of wild-type zebrafish, reminiscent of the patterns observed in mammals (Xu et al., 2002); however, the relevant ligand(s) in zebrafish remain to be identified. Moreover, although initial activation of p-Smad2 in the germline was intact, the strong signals achieved in germline and soma of wild-type females were not detected in rbpms2DMs. Because Rbpms2 binds to and has been implicated in potentiating activation of Smad2/3 targets (Sun et al., 2006), it is possible that Rpbms2 promotes female fates by boosting TGF-β signaling. However, based on the expression and immunohistochemical (IHC) data herein and the earlier sex-reversal phenotypes of rbpms2DMs compared with bmp15 and bmpr1bb (also known as alk6b) mutants (Dranow et al., 2016; Neumann et al., 2011), altered Bmp signaling is not likely to be responsible for the initial failure of oocyte development observed in rbpms2 mutants.

A key and evolutionarily conserved regulator of male-specific gene expression and antagonist of female fates that is necessary for male-specific development is the Doublesex and Mab3 related transcription factor, Dmrt1 (Webster et al., 2017; Ge et al., 2017; Huang et al., 2017; Zhang et al., 2016; Zhao et al., 2015; Lindeman et al., 2015; Koopman, 2009; Raymond et al., 1999; Koopman and Loffler, 2003). Zebrafish mutants lacking dmrt1 develop mostly as fertile females, the opposite phenotype of rbpms2DMs (Webster et al., 2017); rbpms2 and dmrt1 are both expressed as RNAs in stage I oocytes (Webster et al., 2017; Kaufman et al., 2018). Because female development initiates but ultimately fails in rbpms2DMs, and because we identified Rbpms2 binding sites in the 3′UTR of dmrt1 transcripts, it is possible that Rbpms2 promotes female differentiation by repressing dmrt1 directly and/or by antagonizing the male-specific program mediated by Dmrt1. We reasoned that if rbmps2 mutants develop as fertile males because of failed antagonism of the Dmrt1-mediated male pathway, then loss of dmrt1 should restore female development in rbpms2DMs. However, if rbpms2 acts downstream of dmrt1 or acts in a distinct pathway, then dmrt1 mutants would fail to differentiate as females even in the absence of Rbpms2. To test this hypothesis, we conducted genetic epistasis analysis between mutants lacking rbpms2 and dmrt1 functions. We generated adults lacking rbpms2a/b and dmrt1 and screened for sex-specific secondary sex traits and fertility (Fig. S1). To determine the sex-specific identity of germ cells in the early gonads, and to resolve the epistatic relationship between rbpms2 and dmrt1, we performed IHC analysis of d28 gonads. In triple heterozygous (HHH) siblings, gonads had numerous germ cells and early-stage oocytes (Fig. 2A) or were testis-like gonads (Fig. 2B) (Heim et al., 2014; Bontems et al., 2009). The rbpms2DM gonads were either intersex with some early oocytes and clusters of spermatogonia-like cells (Fig. 2C), or were testis-like (Fig. 2D) as we previously reported (Kaufman et al., 2018). All dmrt1 single mutant gonads examined had early oocytes (Fig. 2E), whereas rbpms2;dmrt1TMs (Fig. 2F) resembled their rbpms2DM siblings. Based on these data, rbpms2 is required for female-specific differentiation, even in the absence of dmrt1, and is therefore epistatic to dmrt1.

Like rbpms2DMs (Fig. 3A,B; Fig. S1), rbpms2;dmrt1TM adults developed exclusively as males, indicating that rbpms2 is also epistatic to dmrt1 in terms of adult sex. However, unlike rbpms2DMs, which were fertile males, or dmrt1 mutants with some Rbpms2 intact (compound genotypes with various rbpms2 combinations), the TM males were sterile in fertility assays (failed to induce spawning in mating assays) (Fig. 3B,C; Fig. S1). Interestingly, dmrt1 mutants that were also rbpms2b mutants, with just one copy of rbpms2a intact (compound HMM), were sterile males like the TMs (Fig. 3B). Thus, although rbpms2a and rbpms2b have redundant functions, requiring loss of both to observe the male-only phenotype, it appears that rbpms2b is more important for sustaining female fates because, even in the absence of dmrt1, one copy of rbpms2b but not rbpms2a is sufficient to maintain female fates. Identifying the basis of this difference in rbpms2 activity is an important area of future investigation. Unlike the ovary or testis of wild-type zebrafish (Fig. 3D,E), rbpms2DMs or the ovary of dmrt1 mutants (Fig. 3F), TM adult males were found to lack germ cells upon inspection of the dissected gonad (Fig. 3G). It appears that germ cells of the TM gonad are lost as they attempt to transition to a male fate. In previous studies, roughly 30% of dmrt1 mutants recovered were sterile males due to failed spermatogenesis and somatic gonad deficits (Webster et al., 2017; Wu et al., 2020), raising the possibility that the sterile testes observed in TMs could simply be a result of loss of dmrt1. Because we only recovered dmrt1 mutant females, we were unable to compare testis development in dmrt1 single mutants and TMs. Because we did not detect Rbpms2 protein in testis (Fig. S2) or in somatic cells in the ovary (Kaufman et al., 2018), Rbpms2 probably acts cell autonomously to promote female germline fate, which could secondarily impact somatic gonad fate. However, dmrt1 transcripts have been detected in both germ and somatic cells (Webster et al., 2017), thus germ cells might be lost in rbpms2;dmrt1TMs because of a cell autonomous failure to establish sex-specific identity (e.g. the germ cells are neither truly male or female without Rbpms2 and Dmrt1), nonautonomous defects in somatic gonad development, as a consequence of mismatched germ cells and somatic gonad identity, or as a result of a later germ cell autonomous requirement for Dmrt1 in germline maintenance or development of the somatic gonad (Webster et al., 2017; Matson et al., 2010; Wu et al., 2020; Lei et al., 2007; Masuyama et al., 2012). Nonetheless, these results indicate that Rbpms2 is required to promote expression of factors required for female fates, despite absence of dmrt1.

Loss of the germ cells in rbpms2;dmrt1TM adults could simply be a consequence of lack of sex-specific identity or a mismatch between germ cells and somatic cell sex-specific identities. If so, the gonads lacking two opposing sex-specific differentiation factors should disrupt the sex-specific identity of the gonad, potentially force a mismatch between somatic and germ cell identity, and recapitulate the sterile male phenotype observed in rbpms2;dmrt1TMs. Somatic-gonad derived factors such as Dmrt1 and Cyp19a1a are continually expressed in adult gonads and loss of either gene disrupts stochastic gonad differentiation as either male or female (Webster et al., 2017; Dranow et al., 2016), with loss of dmrt1 resulting in female sex bias (Webster et al., 2017) and loss of cyp19a1a resulting in recovery of only male adults (Dranow et al., 2016), indicating that the sexual phenotype of the somatic gonad must be actively maintained by the functions of these proteins. However, the genetic relationship between these two essential factors had not been appreciated in zebrafish. If bipotential ovary formation fails in cyp19a1a mutants due to failure to antagonize the Dmrt1-mediated male pathway, then loss of dmrt1 should restore development of a bipotential ovary and potentially later female fates in cyp19a1a mutants. Conversely, if male gonadogenesis requires dmrt1 antagonism of cyp19a1a, then loss of cyp19a1a should restore male-specific fates in dmrt1 mutants. However, if dmrt1 and cyp19a1a independently contribute to sex-specific differentiation, then cyp19a1a;dmrt1DMs lacking both sex-specific identity factors would fail to confer a sex-specific identity to the somatic gonad or result in a mismatch of soma and germ cell identity, leading to sterile males as observed in rbpms2;dmrt1TMs. To test this hypothesis, we conducted genetic epistasis analysis between mutants lacking cyp19a1a and dmrt1 and determined the sex-specific identity of the double mutant germ cells using IHC analysis of d45 gonads (Fig. 4; Fig. S3). In cyp19a1a^−/+;dmrt1^−/− (referred to as cyp19a1a;dmrt1HM) siblings, all gonads examined had numerous early-stage oocytes (Fig. 4A), consistent with the previously reported dmrt1 mutant female-bias phenotype (Webster et al., 2017; Wu et al., 2020). As expected, all cyp19a1a^−/−;dmrt1^−/+ (cyp19a1a;dmrt1MH) sibling gonads examined were testis-like and lacked the oocyte marker, Buc (Fig. 4B) (Heim et al., 2014; Bontems et al., 2009). By contrast, cyp19a1a;dmrt1DMs either contained early oocytes with normal Balbiani bodies marked by Buc (Fig. 4C) or were testis-like with clusters of undifferentiated cells that lacked Buc expression (Fig. 4D). Based on these data, it appears that the loss of dmrt1 is sufficient to restore development of a bipotential ovary and initial female differentiation in the absence of cyp19a1a, possibly by preventing dmrt1 inhibition of rbpms2.

To investigate the long-term effects of the absence of cyp19a1a and dmrt1 on differentiation and fertility, we screened adult mutants for sex-specific secondary sex traits and fertility (Fig. 5 and Fig. S4). The cyp19a1a;dmrt1MH (Fig. 5A) and cyp19a1a;dmrt1HM (Fig. 5B) adults appeared male and female, respectively, based on external secondary sex traits. cyp19a1a;dmrt1MMs appeared male, based on inspection of secondary sex traits (Fig. 5C,D) similar to rbpms2;dmrt1DMs. Similarly, in an independent study using different alleles, cyp19a1a;dmrt1 were all masculinized (Wu et al., 2020). In that study, masculinization of cyp19a1a;dmrt1MMs was attributed to lower serum estrogen levels, which are required for development of female features (Wu et al., 2020; Brion et al., 2004). As expected, male cyp19a1a;dmrt1MH siblings developed testes (Fig. 5E,I) and female cyp19a1a;dmrt1HM siblings developed ovaries (Fig. 5F,J) with late stage oocytes, including pre-vitellogenic (PV), early vitellogenic (EV) and vitellogenic (V) oocytes. In contrast, despite their male secondary sex traits, D120 cyp19a1a;dmrt1MM gonads (n=4/5) (Fig. 5G,H,K,L; Fig. S4) were surprisingly ovary-like and contained early oocytes that progressed as far as the PV stage of oogenesis. The absence of vitellogenic growth suggests that, although the bipotential ovary can be established and oogenesis can progress through prophase I without cyp19a1a in the absence of dmrt1, further oocyte differentiation requires cyp19a1a. Failure of cyp19a1a;dmrt1MM oocytes to progress beyond PV stages is reminiscent of bmp mutant ovary phenotypes and is consistent with a role for cyp19a1a in promoting oocyte and follicle development downstream of bmp15 (Dranow et al., 2016). Accordingly, at d145, half of the recovered cyp19a1a;dmrt1MM gonads (Fig. 5O; Fig. S4) were ovaries with oocytes stalled at the PV stage, whereas the other half no longer contained oocytes. Consistent with failed oocyte differentiation in the absence of cyp19a1a and a requirement for dmrt1 in spermatogenesis or somatic gonad development, double mutants were sterile in fertility assays (absence of sperm or eggs in mating assays) (Fig. 5M). Similarly, in the study by Wu and colleagues, histological analysis revealed cyp19a1a;dmrt1MMs with well-developed ovaries despite the masculine secondary sex traits and observed failure to maintain oocytes and progressive sex-reversal of double mutants (Wu et al., 2020). Thus, Cyp19a1a function antagonizes Dmrt1 to establish the bipotential ovary, but is also required later in a dmrt1-independent manner to maintain female fate and promote oocyte differentiation in adulthood.

Bmp15 is an oocyte-derived protein required to promote oocyte differentiation beyond the PV stage and is required for development of cyp19a1a-positive granulosa cells (Dranow et al., 2016). In bmp15 mutants, initial cyp19a1a is intact, but later cyp19a1a expression is bmp15 dependent. Therefore, to explore late cyp19a1a function in the context of loss of dmrt1 and investigate the regulatory relationship between bmp15 and dmrt1, we examined secondary sex characteristics and sex ratios in mutants lacking bmp15 and dmrt1 (Fig. 5P-U; Fig. S5). As expected, bmp15;dmrt1MH adults (90-145 d) were exclusively male based on secondary sex traits (Fig. 5P) and primary gonad sex (Fig. 5S), and bmp15;dmrt1HM adults were female (Fig. 5Q,T). Although bmp;dmrt1MMs initially appeared feminized based on their body shape, they had male papillae (Fig. 5R). Like cyp19;dmrt1MMs, bmp;dmrt1MM gonads were ovaries with early stage oocytes that did not progress beyond PV stages (Fig. 5U), as expected if Bmp15 promotes oocyte development by inducing Cyp19a1a-producing somatic gonad fates. That oocytes persist longer in bmp15;dmrt1MM ovaries (Fig. 5U) than in bmp15;dmrt1MH single mutants (Fig. 5S), indicates that dmrt1 contributes to sex reversal of bmp15 mutants, probably by promoting or reinforcing expression of factors required for spermatogenesis or male-specific differentiation of the somatic gonad.

Previous work showed that cyp19a1a is required to establish a bipotential ovary containing early oocytes (Dranow et al., 2016). Within the bipotential ovary, Cdk21 promotes mitotic divisions of gonocytes (Webster et al., 2018) and Amh limits gonial proliferation and promotes maturation (Lin et al., 2017). In contrast to cyp19a1a mutants, rbpms2, bmp15 and amh mutants develop a bipotential ovary and begin to develop oocytes that ultimately fail prior to prophase I in the case of rbpms2 (Kaufman et al., 2018) and after diplotene arrest in the case of bmp15 (Dranow et al., 2016) and cdk21 (Webster et al., 2018), leading to eventual differentiation of an adult testis in these cases (Fig. 6A-D). This suggests that Cyp19a1a is required earlier in gonad development than Rbpms2, Amh or Bmp15 (Fig. 6B-D). Here, we provide genetic evidence that mutation of dmrt1 (Fig. 6E) restores bipotential ovary development in the absence of cyp19a1a and allows development of a gonad that contains early oocytes (Fig. 6G). That Rbpms2 is not expressed in cyp19a1a single mutants (Fig. S6), which develop exclusively as males, and that normal Balbiani bodies were detected in cyp19a1a;dmrt1DMs indicate that inhibition of Dmrt1 allows expression of Rbpms2 and female development in the absence of Cyp19a1a. Moreover, these results support a regulatory relationship between dmrt1 and cyp19a1a and suggest that Cyp19a1a is required to establish the bipotential ovary and does so by antagonizing Dmrt1 activity (Fig. 6). That bipotentiality is restored indicates that this relationship is probably mutually antagonistic, with Dmrt1 antagonizing early Cyp19a1a activity under normal conditions, poising the system for male differentiation. However, despite the ability of cyp19a1a;dmrt1DMs to initially form a bipotential ovary that can develop as male or female, subsequent oogenesis and ovary maintenance fails, leading to recovery of all male cyp19a1a;dmrt1DM adults (Fig. 6G). Because the primary function of Cyp19a1a enzyme is to catalyze the final step of estrogen (estradiol) synthesis (Bulun et al., 1994; Simpson et al., 1994), the antagonistic relationship between Cyp19a1a and dmrt1 is probably an indirect relationship mediated by the estrogens produced by Cyp19a1a.

Indeed, recently another group independently discovered the same epistatic relationship between cyp19a1a and dmrt1 and provides evidence in support of estrogen involvement (Wu et al., 2020). Additionally, that cyp19a1a;dmrt1DMs can establish a bipotential ovary, but eventually resolve as males, indicates that Cyp19a1a is dispensable for establishment of the bipotential ovary and initiation of female development in the absence of Dmrt1. However, this work and recent work from Wu and colleagues (Wu et al., 2020) indicate that Cyp19a1a is also required for subsequent oocyte maturation and maintenance of female fates by a mechanism that does not depend on antagonism of Dmrt1. Consistent with this notion, earlier work demonstrated that bmp15 produced in oocytes is required to promote oocyte differentiation and development of cyp19a1a-positive granulosa cells (Dranow et al., 2016). Moreover, in human ovaries and granulosa-like cells, activin promotes CYP19 expression via SMAD2 (Mukasa et al., 2003; Nomura et al., 2013); thus, it seems likely that subsequent failure of cyp19a1a;dmrt1DM oocytes to mature is due to lack of the later granulosa cell-derived source of cyp19a1a, which appears to utilize a Dmrt1 inhibition independent mechanism as maturation fails and cannot be restored in the absence of Dmrt1 whether all Cyp19a1a is lacking (cyp19a1a;dmrt1DM) or only the early source is present (bmp15;dmrt1DM) (Fig. 6G,H). Finally, Rbpms2 probably promotes female fates temporally upstream of Bmp15 (Dranow et al., 2016) and estrogen (estradiol) (Wu et al., 2020) because Rbpms potentiates transcription mediated by Smad2/3 (Sun et al., 2006) and mutant oocytes lacking Bmp15 or estrogen (Wu et al., 2020) reach the diplotene stage of oogenesis or beyond in the absence of Dmrt1, whereas rbpms2 mutants do not.

Our study identified potential Rbpms2 targets and used genetic epistasis experiments and cell biological approaches to decipher the genetic hierarchy of crucial factors involved in sex-specific differentiation of the germline and somatic gonad. We show that TGF-β signaling is activated in early germ cells of the bipotential ovary, and in the wild type becomes activated in the somatic gonad by d28, during the early stages of sexual differentiation. In rbpms2DMs, initial activation of TGF-β in the germline is intact. We provide genetic evidence that dmrt1 antagonizes the crucial female factor rbpms2, acting either upstream or in a parallel mutually antagonistic pathway. Based on their opposite mutant phenotypes and the presence of Rbpms2 binding sites in the dmrt1 3′UTR, Rbpms2 may promote female fate by preventing translation of Dmrt1. Accordingly, loss of Rbpms2 would lead to production of Dmrt1 and male differentiation (Fig. 6). However, our genetic data indicate that, once produced, Dmrt1 or its targets antagonize or promote elimination of Rbpms2 and other factors required for acquisition of female fates, as Rbpms2 protein is not detected in testis.

Moreover, Rbpms2 function in promoting female fates extends beyond simply repressing or antagonizing dmrt1, as it is essential for female sex-specific differentiation, even in the absence of Dmrt1, indicating that additional Rbpms2 targets are required for female fates (Fig. 6F). In contrast to loss of rbpms2 function, female fates can be restored or prolonged in mutants lacking cyp19a1a and dmrt1 or bmp15 and dmrt1.

Taken together, this work and the work of Wu et al. (2020) indicate that cyp19a1a acts during at least two steps of female-specific differentiation. Early cyp19a1a-mediated suppression of dmrt1 is required to establish a bipotential ovary and initiate female fate acquisition in zebrafish, possibly by promoting expression of rbpms2, which is required for female-specific differentiation, even in the absence of Dmrt1. Finally, once female fates have been established, Cyp19a1a is required for subsequent oocyte maturation and maintenance of female fates by a mechanism that does not depend on antagonism of Dmrt1.

Wild-type zebrafish embryos of the AB strain were obtained from pairwise crosses and reared according to standard procedures (Westerfield, 2000). Embryos were raised in 1× embryo medium at 28.5°C and staged as described (Kimmel et al., 1995). All procedures and experimental protocols were in accordance with NIH guidelines and approved by ISMMS (protocol # 17–0758 INIT) IACUC. The zebrafish rbpms2a^ae30 allele was previously generated in our laboratory using CRISPR-Cas9 mediated mutagenesis (Kaufman et al., 2018) and the rbpms2b^sa9329 allele was obtained from the Sanger Institute's Zebrafish Mutation Project (Kettleborough et al., 2013). The zebrafish dmrt1^uc27, bmp15^uc31 and cyp19a1a^uc38 alleles were obtained from Bruce W. Draper, University of California, Davis (Webster et al., 2017; Dranow et al., 2016).

Genomic DNA was extracted from adult fins, juvenile fins or single embryos using standard procedures (Westerfield, 2000). The genomic region surrounding rbpms2a^ae30 was amplified using the primers 5′-TTTGCTAAAGCCAACACGAA-3′ and 3′-ATTCACCCTGGCCAGAGTTT-5′, followed by digestion of the wild-type strand with the enzyme BsurI (New England Biolabs, R0581S) (Kaufman et al., 2018). The genomic region surrounding rbpms2b^sa9329 was amplified using the dCAPs primers 5′-CACTTATCAAGCTAACTTCAAAGCAGA-3′ and 3′-TGAAAGGGGACAAATAAGTCA-5′, followed by digestion of the mutant strand with the enzyme MboII (New England Biolabs, R0148S) as described previously (Kaufman et al., 2018). The genomic region surrounding bmp15^uc31 was amplified using the primers 5′-AGCCTTTCAGGTGGCACTCG-3′ and 3′-GGGAGAAAGTGTTTTTCAGTGG-5′, and dmrt1^uc27 using primers 5′-GTTGTAACTGGCAGCTGGAGA-3′ and 3′-GGCGATGAGTCTGCATTTCT-5′, but these products were resolved without digestion. After 40 cycles of PCR at 60°C annealing, samples were digested for 30 min using specified restriction enzymes. Digested PCR products were resolved using a 1.5% ultrapure agarose (Invitrogen) and 1.5% Metaphor agarose (Lonza) gel.

PCR and melting curve analyses were performed as described (Parant et al., 2009). PCR reactions contained 1 μl of LC Green Plus Melting Dye (BioFire Diagnostics), 1 μl of Taq buffer, 0.8 μl of dNTP mixture (2.5 mM each), 1 μl of each primer (as indicated below) (5 μM), 0.05 μl of Taq (Genscript), 1 μl of genomic DNA and water up to 10 μl. PCR and melt curve analyses were performed in a Bio-Rad CFX96 Real-Time System, using black/white 96-well plates (Bio-Rad HSP9665). PCR reaction protocol was 98°C for 1 min, then 34 cycles of 98°C for 10 s, 60°C for 20 s, and 72°C for 20 s, followed by 72°C for 1 min. After the final step, the plate was heated to 95°C for 20 s and then cooled to 4°C. Melt curve analysis was performed over a 72–92°C range and analyzed with Bio-Rad CFX Manager 3.1 software. All mutations were confirmed by TA cloning and sequencing. Primers used were as follows: rbpms2a^ae30, 5′-ACACGAAGATGGCGAAGAGT-3′ and 3′-CAGGGTGCAGGTTGGAAG-5′; rbpms2b^sa9329, 5′-ATGAGGGTTCACTTATCAAGCTA-3′ and 3′-TCCGGTCAGCTGTAATGTCTAA-5′; dmrt1^uc27, 5′-CTCTCGCTGCAGAAACCAC-3′ and 3′-GGCGATGAGTCTGCATTTCT-5′; cyp19a1a^uc38, 5′-CCAACTGACCTGGAATGTGTG-3′ and 3′-AGCTACAATACTGCTGCTGCTA-5′.

Trunks were dissected at d21 from the specified genotypes and placed into Trizol (Life Technologies). RNA was extracted using standard phenol-chloroform-isoamyl alcohol (PCI) extraction and was used for oligo(dT) cDNA preparation (using Invitrogen SuperScript III reverse transcriptase). RT-PCR was performed using the following primers: bmpr1ab, 5′-GATGCCACAAACAACACCTG-3′ and 3′-GCAACCAAAGTGAAGCAACA-5′; bmpr1bb, 5′-GAGGCAGATGGGTAAACTG-3′ and 3′-CTCCTGTGTTCTGTTGAG-5′; bmpr2b, 5′-CGGCCTCTGGGAGAAAACAC-3′ and 3′-TGGCCTCATCTCTGTGTATAG-5′; ef1alpha, 5′-AGCCTGGTATGGTTGTGACCTTTCG-3′ and 3′-CCAAGTTGTTTTCCTTTCCTGCG-5′. PCR products were resolved using a 1.5% ultrapure agarose (Invitrogen) gel and visualized with a Biorad gel imager.

For whole-mount immunofluorescence (IF) of gonads, trunks or tissue at d28, d45 and d120 were fixed in 4% paraformaldehyde overnight at 4°C, dehydrated in MeOH and placed at −20°C. In the case of trunks, an incision was made to open the body cavity during staining and gonads were dissected for imaging. Rabbit anti-phospho-Smad2 antibody (Millipore AB3849-I) was used at 1:200. Rabbit anti-Buckyball antibody y1165 was used at 1:500 (Heim et al., 2014). Rabbit anti-Rbpms2 antibody (Abcam A170777) was used at 1:500. Chicken anti-Vasa antibody was a gift from Bruce W. Draper (Blokhina et al., 2019) and used at 1:3000 dilution. AlexaFluor488, CY3 (Molecular Probes) secondary antibodies were diluted at 1:500. Images were acquired using a Zeiss Axio Observer inverted microscope equipped with ApotomeII and a CCD camera. Images were processed in ImageJ/FIJI, Adobe Photoshop and Adobe Illustrator.

Fish were sexed based on morphologically distinct secondary features of females and males, such as body shape, fin coloration, genital papillae and tubercle characteristics. Females are typically larger than males, have a more rounded abdomen and have a pale body and anal fin stripes, protruding genital papillae and smooth pectoral fins lacking spiky tubercles. In contrast, males appear smaller than females, have a slender body shape, display dark yellow body stripes and anal fins, lack protruding genital papillae and have spiky tubercles on their pectoral fins. Images were acquired using a Zeiss Stemi 508 dissecting scope equipped with a color CCD camera. Images were processed in Zeiss ZenBlue, ImageJ/FIJI, and Adobe Illustrator.

Gonads were dissected from either d90, d120 or d145 adult fish. Gonads were either imaged live in brightfield at 40× magnification and fixed for further analysis, or fixed and stained with Vasa antibody (1:3000) then imaged in brightfield and GFP at 40× magnification. Gonad overview images were acquired using a Zeiss Axio Zoom dissecting scope equipped with an Apotome II and a CCD camera. Images were processed in Zeiss ZenBlue, ImageJ/FIJI, and Adobe Illustrator.

Fertility was assessed in mating trials where couples for a given genotype were paired and bred for several repeat mating trials. For each cross, if a male triggered spawning, the number (in parentheses in Fig. 3C) of fertilized eggs per trial was counted. For sperm acquired through in vitro fertilization (IVF), the number of eggs successfully fertilized was also counted. n/a indicates that fish were not available for subsequent mating trials.

Statistical analyses were performed using Graphpad Prism 8 and Excel; we performed a chi² test to compare the observed sex ratios to the expected 50:50 sex ratio. We then compared each group to the heterozygote siblings using a chi² goodness of fit test with a Bonferroni correction for multiple comparisons, using an adjusted significance value determined by the number of comparisons to be calculated. Where groups were pooled, expanded data is available in the Supplementary material.

We thank members of the Marlow lab for helpful discussions, our animal-care staff for fish care (Einstein and CCMS at ISMMS) and the Microscopy CoRE at Icahn School of Mount Sinai and at Einstein. We thank Daniel Dellal and Paloma Bravo for technical assistance.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.190942.reviewer-comments.pdf