Dmrt1 is the only male pathway gene tested indispensable for sex determination and functional testis development in tilapia

Sex is determined by multiple factors derived from somatic and germ cells in vertebrates. We have identified amhy, dmrt1, gsdf as male and foxl2, foxl3, cyp19a1a as female sex determination pathway genes in Nile tilapia. However, the relationship among these genes is largely unclear. Here, we found that the gonads of dmrt1;cyp19a1a double mutants developed as ovaries or underdeveloped testes with no germ cells irrespective of their genetic sex. In addition, the gonads of dmrt1;cyp19a1a;cyp19a1b triple mutants still developed as ovaries. The gonads of foxl3;cyp19a1a double mutants developed as testes, while the gonads of dmrt1;cyp19a1a;foxl3 triple mutants eventually developed as ovaries. In contrast, the gonads of amhy;cyp19a1a, gsdf;cyp19a1a, amhy;foxl2, gsdf;foxl2 double and amhy;cyp19a1a;cyp19a1b, gsdf;cyp19a1a;cyp19a1b triple mutants developed as testes with spermatogenesis via up-regulation of dmrt1 in both somatic and germ cells. The gonads of amhy;foxl3 and gsdf;foxl3 double mutants developed as ovaries but with germ cells in spermatogenesis due to up-regulation of dmrt1. Taking the respective ovary and underdeveloped testis of dmrt1;foxl3 and dmrt1;foxl2 double mutants reported previously into consideration, we demonstrated that once dmrt1 mutated, the gonad could not be rescued to functional testis by mutating any female pathway gene. The sex reversal caused by mutation of male pathway genes other than dmrt1, including its upstream amhy and downstream gsdf, could be rescued by mutating female pathway gene. Overall, our data suggested that dmrt1 is the only male pathway gene tested indispensable for sex determination and functional testis development in tilapia.


Introduction
Sex determination (SD), which is controlled by genetic or environmental factors, or both in vertebrates [1,2], is a hot topic in developmental and reproductive biology.Since the discovery of the first fish master sex determining (MSD) gene dmy/dmrt1by in medaka (Oryzias latipes) in 2002 [3,4], with the development of genome sequencing and genome editing technologies, a number of MSD genes have been identified in fish species in the past two decades [2,5].In contrast to most mammals that sex is genetically determined by SRY/Sry on the Y chromosome [6,7], the MSD genes in fish exhibit diversity and rapid turnover even in closely related species [5].It is worth noting that most of the fish MSD genes identified belong to the TGF-β signaling pathway, including amhy, amhr2y, gdf6y, bmpr1bby and gsdfy [5,8].In Nile tilapia (Oreochromis niloticus), a gonochoristic teleost with an XX/XY SD system, a tandem duplicate of amh on the Y chromosome, named as amhy, is identified as the MSD gene by our group [9].Although the MSD genes vary extensively among fish species, the downstream factors are more or less conserved [10,11].The dmrt1, gsdf as male pathway genes and foxl2, foxl3 and cyp19a1a as female pathway genes have been identified in Nile tilapia [12][13][14] and some other fishes [10,11].It is well accepted that antagonistic actions between female and male pathway genes determine and maintain the gonadal sex in vertebrates [1].However, to date, the genetic interactions between the identified male and female pathway genes in fish species are largely unknown.
Estrogen plays an important role in female ovarian differentiation in non-mammalian vertebrates [15].It is produced through the conversion of androgens by steroidogenic enzyme CYP19A1 (aromatase) [16].Administration of aromatase inhibitor (AI) or exogenous estrogen (E2) induces sex reversal (SR) in a large number of fish species, including tilapia [15,17,18].Consistently, mutation of cyp19a1a, which is only expressed in somatic cells, leads to testicular development in zebrafish (Danio rerio), tilapia and medaka [13,19,20].In addition, mutation of gonadal somatic cell expressed transcription factor foxl2, an evolutionarily conserved female pathway gene that directly regulating cyp19a1a expression and E2 synthesis [21], results in testicular development in zebrafish, tilapia and gibel carp (Carassius gibelio) [13,22,23].Furthermore, studies have shown that mutation of germline specific expressed foxl3, a paralog of foxl2 found in most vertebrates except placental mammals [24][25][26], results in germ cell SR in ovary of XX tilapia and medaka [14,27].In zebrafish, mutation of foxl3 (also named as foxl2l) results in all male development [28].However, the critical role of foxl3 in sperm-egg fate decision has not been demonstrated outside of the fish clade.So far, in tilapia, our transcriptional regulation and loss-of-function analyses reveal a possible female pathway foxl2-cyp19a1a-foxl3 in controlling and maintaining ovary fate.
Transcription factor dmrt1 is a highly conserved regulator involved in male SD and sex differentiation across vertebrates [2].In mammals, although Dmrt1 is not required for SD because the gonads of XY Dmrt1 mutants still develop as testes in mouse (Mus musculus) [29], ectopic expression of Dmrt1 in XX gonads results in female-to-male SR, suggesting that Dmrt1 retains the capacity of SD [30].A recent study in rabbit (Oryctolagus cuniculus) has shown that dmrt1 is the sex determining gene as the gonads of the XY Dmrt1 mutants differentiate into ovaries [31], indicating that Dmrt1 has a SD role in certain mammals.In bird, loss of one dmrt1 copy in ZZ chicken (Gallus gallus) leads to ovary in place of testis development [32].In reptile, dmrt1 knockdown in ZZ or overexpression in ZW red-eared slider turtle (Trachemys scripta) embryos reverse their gonadal phenotypes [33].In fish, the gonads of the zebrafish and tilapia dmrt1 mutants directly develop into ovaries [14,34], while in medaka, the gonads of the dmrt1 mutants first develop into testes and then transdifferentiate into ovaries [35], probably due to the compensation of dmy/dmrt1by.Anyway, the gonads of the dmrt1 mutants finally develop as ovaries.Additionally, dmrt1 and its duplicates are even identified as the MSD genes in several fish species [3,4,[36][37][38].However, it is still unknown why dmrt1 is so important for males in vertebrates.
The establishment of animal models with disruption of male and female pathway genes simultaneously is crucial for elucidating their genetic interaction in sex differentiation.It is particularly noteworthy that the ovary of dmrt1 mutants cannot be rescued to a functional testis by mutation of cyp19a1a or AI treatment in fish and bird.For example, 1) In zebrafish, mutations of cyp19a1a and/or estrogen receptors (esr1/esr2a/esr2b) fails to rescue the ovaries of dmrt1 mutants to testes [39,40].2) In tilapia, the ovaries of dmrt1 mutants cannot be rescued to functional testes by AI treatment or mutation of foxl2 or foxl3 [14].3) In chicken, administration of AI is unable to rescue the SR of ZZ Dmrt1 mutants [32].Most importantly, these studies were conducted in different species, and up to now, it is necessary to study how the gonads will develop when dmrt1 and key genes of the female pathway, including foxl2, foxl3 and cyp19a1a, are mutated simultaneously in one species.Additionally, teleost fish possess two aromatase genes, cyp19a1a (encoding ovarian aromatase) and cyp19a1b (encoding brain aromatase) [41].There is no doubt that brain aromatase can catalyze the conversion of testosterone to estrogen [42].In zebrafish, the follicles of dmrt1;cyp19a1a double mutants could develop into previtellogenic stage [39,40], which cannot rule out the role of Cyp19a1b that can produce E2.Therefore, it is unknown whether this is caused by the indispensable role of dmrt1 in male SD, or the compensation of cyp19a1b for the insufficient estrogen caused by cyp19a1a deficiency.
Besides Dmrt1, the members of TGF-β signaling pathway, especially amh/amhr2 and gsdf, play critical role in male SD and sex differentiation in a number of fish species as well [8].In XY tilapia, mutation of amhy or gsdf results in up-regulation of Cyp19a1a and male-to-female SR [9,12].This sex reversed ovary of XY amhy and gsdf mutants can be rescued to functional testis by AI treatment [12,43].Consistently, mutation of cyp19a1a rescues the SR in the XY amhy mutant tilapia at 90 days post fertilization (dpf) [43].Similarly, AI treatment rescue the male-to-female SR caused by gsdf mutation in gibel carp and amhr2 mutation in Japanese flounder (Paralichthys olivaceus) [44,45].Therefore, induction of Cyp19a1a expression is responsible for driving male-to-female SR in amhy/amhr2 and gsdf mutants.Nevertheless, it is unknown whether mutation of other female pathway genes, such as foxl2 and foxl3, can rescue the male-to-female SR caused by mutation of amhy and gsdf, both of which were exclusively expressed in somatic cells [9,12].
It is well known that the vertebrate sex is highly plastic.Females with differentiated or even mature ovary are masculinized to functional males by long-term AI treatment in tilapia, medaka and zebrafish, indicating that estrogen is also critical for female sex maintenance [46][47][48].Recent study in our group demonstrated that AI treatment fails to induce testis development in XX dmrt1 mutants and E2 treatment fails to induce oocyte differentiation in XY foxl3 mutants in tilapia [14].Therefore, sex maintenance is suggested to be regulated by transcription factors (dmrt1 and foxl3) and estrogen, which are required to be expressed continuously in the gonads.Like the other MSD genes [1][2][3][4][5], amhy expression is transiently up-regulated during the period of SD (from 3 to 7 days after hatching, dah) [43].How these male and female pathway genes interact and work together to maintain the gonadal fate is unknown.
The roles of male pathway genes amhy, dmrt1 and gsdf and female pathway genes foxl2, cyp19a1a and foxl3 in SD and sex differentiation have been demonstrated by generation of single gene mutation line in tilapia [9,[12][13][14].However, the knowledge about interactions of these genes, especially between genes expressed in somatic cells and germ cells, in gonadal fate decision and maintenance is still limited.The Nile tilapia is a good animal model for studying SD and sex differentiation because genetic all-females and genetic all-males are available [49].A number of genes show sexual dimorphic expression in gonads at 5 dah, the critical time for SD [50,51].For instance, amhy and gsdf are expressed exclusively in somatic cells of the XY gonads, whereas foxl2 and cyp19a1a are expressed exclusively in somatic cells of the XX gonads [9,12,21,43].dmrt1 is expressed in both somatic cells and germ cells in XY testis, while foxl3 is expressed exclusively in germ cells in XX ovary [14].The first morphological differentiation of gonad occurs in tilapia between 20 and 25 dah with appearance of the ovarian cavity in the XX ovary.The efferent duct in the XY testis is observed at around 40 dah.Additionally, difference in germ cell number between sexes is observed during early gonad differentiation.The germ cells continue to proliferate in XX ovary from 9 dah, while germ cell proliferation is not observed in XY testis until 15 dah [52].The germ cells undergo meiosis to generate oocytes in XX gonads from 25 dah onwards, but meiosis does not initiate to generate spermatocytes in XY gonads until 60 dah (S1 Fig).

Gonads of dmrt1 -/-;cyp19a1a -/-double mutants developed as ovaries or underdeveloped testes
Our previous studies showed that Dmrt1 directly binds to the promoter of cyp19a1a to repress its transcription [53] and loss of dmrt1 results in ovary development and up-regulation of cyp19a1a expression in XY tilapia when checked at 15 dah [14].But it is unclear when this upregulation occurs.In this study, the gonads of the XY dmrt1 -/-mutants developed as ovaries with previtellogenic follicles at 60 dah (n = 12), same as those of XX dmrt1 -/-mutants (n = 11) and wild-type (WT) XX (n = 14) (Fig 1A).Whole-mount immunofluorescence (IF) analysis showed that expression of Cyp19a1a was not observed in XY dmrt1 -/-gonads at 5 dah (S2A Fig), indicating that dmrt1 is important for male sex differentiation and maintenance.In contrast, loss of cyp19a1a results in up-regulation of dmrt1 expression and testis development in XX tilapia at 90 dah [13].Consistently, all the gonads of the XX cyp19a1a -/-mutants developed as testes at 60 dah (n = 15), same as the XY cyp19a1a -/-mutants (n = 10) and WT XY (n = 13) (Fig 1A).We found that the majority of the XX cyp19a1a -/-mutants (6/8, 75%) displayed testis development, while a minority of the mutants (2/8, 25%) exhibited testis but with a few dispersed oocytes at 45 dah as demonstrated by IF of Leydig cell (Cyp11c1) and oocyte (42Sp50) markers.All the gonads of the XX cyp19a1b -/-mutants (n = 11) developed as ovaries at 45 dah (S2B and S2C Fig), indicating that cyp19a1b is not implicated in female fate decision.However, all the gonads of the XX cyp19a1a -/-;cyp19a1b -/-double mutants (n = 10) and AI treated-XX cyp19a1a -/-mutants (n = 7) developed as complete testes at 45 dah.Whole-mount fluorescence in situ hybridization (FISH) result showed that expression of dmrt1 was detected in the gonads of the XX cyp19a1a -/-;cyp19a1b -/-double mutants at 5 dah as the WT XY (S2D Fig), suggesting that estrogen is involved in female SD.
How the gonad will develop if both dmrt1 and cyp19a1a are mutated is unknown.The dmrt1 -/-;cyp19a1a -/-double mutants were established by crossing dmrt1 +/-;cyp19a1a +/-double heterozygous males and females, and their gonadal development were analyzed by histological examination.Interestingly, the gonads of the dmrt1 -/-;cyp19a1a -/-double mutants developed as either typical ovaries with ovarian cavity and previtellogenic oocytes (17/27, 63%) (named as type I, as demonstrated by 42Sp50 staining) or underdeveloped testes with no germ cells (10/27, 37%) (named as type II, as reflected by Cyp11c1 and a germ cell marker Vasa staining) irrespective of the genetic sex at 60 dah (Fig 1A and 1B).In agreement with the ovary phenotype (type I), the female specific markers foxl2 (somatic cell), foxl3 (oogonia), and bmp15 (oocyte) were detected, while male specific markers dmrt6, creb1b (spermatocyte) and cyp11c1 were not detected by RT-PCR, similar to those in the WT XX tilapia at 60 dah (Fig 1C).In addition, we examined the gonad development of the dmrt1 -/-;cyp19a1a -/-double mutants at 45 and 120 dah.The ratios of the type I double mutants were 62.5% and 61% at 45 and 120 dah, respectively (S3A-S3C Fig) .Furthermore, whole-mount FISH result showed that foxl3 mRNA was expressed in the gonads of over half (4/7, 57%) of the double mutants at 25 dah.Similar to the WT XY, Amh was expressed in the gonads of the XX/XY dmrt1 -/-;cyp19a1a -/- double mutants at 5 dah, indicating the masculinization of somatic cells.Unlike the clear differences in gonadal morphology and foxl3 expression observed between the type I and type II phenotype in the double mutants at 25 dah, no obvious differences in gonadal morphology and Amh expression were observed between the type I and type II phenotype of the double mutants (n = 7) at 5 dah (Fig 1D).These results excluded the possibility that the underdeveloped testis was transformed from ovary.To our surprise, the Sertoli cell marker Amh and Leydig cell markers 3β-HSD-I and Cyp11c1 were still expressed in the underdeveloped testes of type II double mutants at 120 dah (S3D Fig) .These results demonstrate that the gonads of the dmrt1 -/-;cyp19a1a -/-double mutants developed as ovaries or underdeveloped testes.XY, XX/XY dmrt1 -/-, XX/XY cyp19a1a -/-and XX/XY dmrt1 -/-;cyp19a1a -/-tilapia at 60 dah.(C) RT-PCR analysis of several female and male specific markers expression in the gonads of WT XX, WT XY, XX/XY dmrt1 -/-, XX/XY cyp19a1a -/-, and XX/XY dmrt1 -/-;cyp19a1a -/-tilapia at 60 dah.foxl2, an ovarian somatic cell marker.foxl3, an oogonia marker.bmp15, an oocyte marker.dmrt6 and creb1b, two spermatocyte markers.cyp11c1, a Leydig cell marker.β-actin was used as an internal control.WT, wild-type; dah, days after hatching.(D) Detection of female specific foxl3 mRNA (green) expression in the gonads of WT XX, WT XY, XX/XY dmrt1 -/-, XX/XY cyp19a1a -/-and XX/XY dmrt1 -/-;cyp19a1a -/-tilapia at 25 dah by whole-mount fluorescence in situ hybridization and Amh protein expression in these mutants at 5 dah by whole-mount immunofluorescence.Clear differences in the gonadal morphology and foxl3 expression were observed between the type I and type II gonadal phenotype in the double mutants at 25 dah, while no obvious differences in morphology and Amh expression were observed between the type I and type II gonadal phenotype of the double mutants at 5 dah.Therefore, a panel is missing in Histological examination showed that the follicles of type I double mutants (n = 7) developed into vitellogenic stage at 150 dah, same as the WT XX ovary (S4A Fig) and XY dmrt1 -/- ovary [14].RT-PCR analysis showed that vtg1, vtg2 and vtg3 mRNAs were expressed in the livers of the type I dmrt1 -/-;cyp19a1a -/-double mutants as the XY dmrt1 -/-and WT XX tilapia at 150 dah (S4B Fig) .EIA and UPLC-MS/MS analyses showed that the serum estradiol-17β (E2) level in the double mutants was significantly lower than that of the WT XX fish, but was significantly higher than that of the WT XY fish (S4C Fig) .By IF, the Cyp19a1a was expressed in the granulosa and theca cells of the WT XX, XX cyp19a1b -/-and XY dmrt1 -/-mutants and it was disappeared in the dmrt1 -/-;cyp19a1a -/-double mutants, whereas Cyp19a1b was only expressed in the theca cells of the WT ovary and it was disappeared in the cyp19a1b -/-mutants.It is interesting to note that, in the dmrt1 -/-single and dmrt1 -/-;cyp19a1a -/-double mutants, besides being observed in theca cells, Cyp19a1b was found to be ectopically expressed in the granulosa cells of the ovary at 150 dah (S4D Fig) .These studies reveal that Cyp19a1b was up-regulated and ectopically expressed in the context of dmrt1 loss.
Gonads of dmrt1 -/-;cyp19a1a -/-;foxl3 -/-triple mutants eventually developed as ovaries Given that germline expressed foxl3 is important for female germ cell fate decision [14,27], we focused on the role of foxl3 in the ovary development type I dmrt1 -/-;cyp19a1a -/-double and dmrt1 -/-;cyp19a1a -/-;cyp19a1b -/-triple mutants.Real-time PCR and FISH analyses showed that foxl3 mRNA was expressed in the type I dmrt1 -/-;cyp19a1a -/-double and dmrt1 -/-;cyp19a1a -/-; cyp19a1b -/-triple mutants, similar to the XY dmrt1 -/-mutants and WT XX at 60 dah, but different from the XX cyp19a1a -/-mutants and WT XY tilapia (Fig 3A and 3B).Double mutation of cyp19a1a and foxl3 in XX tilapia resulted in testis development with normal spermatogenesis as demonstrated by IF using male markers Gsdf, Cyp11c1 and Creb1b, and female markers Cyp19a1a and 42Sp50 (n = 8) at 120 dah (S6 Fig) .In addition, in the XX/XY dmrt1 -/-; cyp19a1a -/-;foxl3 -/-triple mutants, the gonads displayed testicular morphology with spermatocyte-like cells as demonstrated by presence of Creb1b expression and absence of 42Sp50 expression at 60 dah (n = 4) (Fig 3C).Transcriptome analysis showed that oocyte markers zar1, gdf9, nanos3 and bmp15 were not expressed in the gonads of the dmrt1 -/-;cyp19a1a -/-; foxl3 -/-triple mutants at 60 dah, similar to the WT XY fish.In contrast, spermatogenic cell markers eef1a1b, dmrt6, fbxo47, creb1b and sox30 were expressed in the gonads of the dmrt1 -/-; cyp19a1a -/-;foxl3 -/-triple mutants compared with the WT XX.Although foxl2 was not up-regulated in the triple mutants, the somatic cells were overall feminized as reflected by the low expression of male markers cyp11c1, cyp17a1 and gsdf compared with the WT XY (Fig 3D ), indicating a misexpression of sex specific genes in somatic cells.However, the gonads of the dmrt1 -/-;cyp19a1a -/-;foxl3 -/-triple mutants reversed to ovaries with most of the areas occupied by oocytes and small areas occupied by spermatocytes-like cells at 120 dah (n = 3) (Fig 3E).Previously, we reported that the gonads of the dmrt1 -/-;foxl3 -/-double mutants developed as ovaries [14], but reversed to testicular morphology with spermatocyte-like cells by AI treatment for 60 [14] or 120 days in this study.However, the gonads will spontaneously revert to ovaries with vitellogenic follicles after withdrawal of AI treatment (S7 Fig) .These data indicate that even if cyp19a1a and foxl3, both of which are absolutely essential for female fate, were disrupted, the gonads still developed as ovaries when dmrt1 was mutated.

Gonads of amhy − ;foxl3 -/-and gsdf -/-;foxl3 -/-double mutants developed as ovaries but with germ cells in spermatogenesis
Previous studies showed that mutation of amhy or gsdf resulted in both somatic and germ cell SR in XY tilapia [9,12].RT-PCR and whole-mount FISH analyses showed that foxl3 mRNA was specifically expressed in the germ cells of XY amhy − and gsdf -/-mutants as the WT XX at 25 dah (Fig 4A and 4B).It is unclear how the gonad will be developed if both amhy or gsdf and foxl3 are lost in tilapia.To answer this question, we established the XY amhy − ;foxl3 -/-double mutants by crossing XY amhy − ;foxl3 +/-phenotypic females with XY foxl3 -/-phenotypic males and XX/XY gsdf -/-;foxl3 -/-double mutants by crossing the gsdf +/-;foxl3 +/-double heterozygous males and females.Consistent with our previous studies [9,12,14], all the gonads of the XY amhy − and XX/XY gsdf -/-tilapia developed as ovaries and mutation of foxl3 in XX tilapia resulted in masculinization of germ cells in ovary at 60 dah.The gonads of the XY foxl3 -/- mutants developed as testes as those of the WT XY fish.However, only somatic cells but not germ cells were feminized in the XY amhy − ;foxl3 -/-(n = 6) and XX/XY gsdf -/-;foxl3 -/-(n = 8)
To examine the early gonad development in amhy − ;foxl3 -/-and gsdf -/-;foxl3 -/-double mutants, we performed whole-mount IF with Vasa antibody to observe the germ cells.The results showed that the number of germ cells in XX foxl3 -/-single, XY amhy − ;foxl3 -/-and XX/ XY gsdf -/-;foxl3 -/-double mutants was similar to that of the WT XX, XY amhy − and XX/XY gsdf -/-tilapia, but different from that of the WT XY tilapia at 15 dah (S8B Fig) .In our previous study, Dmrt1 was found to be expressed in the gonads of the XY gsdf -/-mutants but not in the XX gsdf -/-mutants at 10 dah [12].Therefore, dmrt1 expression at 10 dah was examined to investigate whether germ cells were masculinized in the gonads of the XX gsdf -/-;foxl3 -/-double mutants at this stage using the XX gsdf -/-mutants as control.RT-PCR analysis showed that dmrt1 was not expressed in the gonads of XX foxl3 -/-, XX gsdf -/-, XY amhy − ;foxl3 -/-and XX gsdf -/-;foxl3 -/-double mutants at 10 dah as the WT XX tilapia.However, expression of dmrt1 was observed in the gonads of XX foxl3 -/-single, XY amhy − ;foxl3 -/-and XX gsdf -/-;foxl3 -/-double mutants at 25 dah (Fig 4D).These results indicated that the female fate of germ cells in foxl3 single, amhy;foxl3 and gsdf;foxl3 double mutants could not be maintained due to up-regulation of dmrt1.
As expected, all the gonads of the AI treated-XY amhy − ;cyp19a1a -/-(n = 9) and -XX/XY gsdf -/-;cyp19a1a -/-(n = 8) double mutants developed as testes at 45 dah as demonstrated by Cyp11c1 and 42Sp50 staining (S9B Fig) .Higher expression of Cyp19a1b/cyp19a1b was observed in these two double mutants compared with the amhy − and gsdf -/-single mutants at 45 dah by IF and real-time PCR analyses (S9C and S9D Fig) .To further assess the compensation of cyp19a1b in early ovary development in the two double mutants, we established the XY amhy − ;cyp19a1a -/-;cyp19a1b -/-and XX/XY gsdf -/-;cyp19a1a -/-;cyp19a1b -/-triple mutants.All the gonads of these two triple mutants developed as testes at 45 dah as demonstrated by presence of Cyp11c1 expression, but absence of 42Sp50 expression.In addition, whole-mount FISH analysis revealed that dmrt1 mRNA was detected in the gonads of these two triple mutants at 10 dah (Fig 5E).These two triple mutants exhibited testicular development with all types of spermatogenic cells at 120 dah and finally were fertile at 180 dah in fertility assay (S9E and S9F Fig).These data suggested that loss of both cyp19a1a and cyp19a1b rescued the maleto-female SR of amhy and gsdf mutants.
Gonads of amhy − ;foxl2 KD and gsdf -/-;foxl2 KD double mutants developed as functional testes RT-PCR analysis showed that foxl2 mRNA was highly expressed in the gonads of XY amhy − but not XY gsdf -/-mutants at 10 dah.Its expression was detected in the gonads of both XY amhy − and gsdf -/-mutants at 25 dah (Fig 6A).To study whether disruption of foxl2 can rescue the SR in XY amhy − and gsdf -/-mutants, we injected foxl2 TALEN mRNA in the embryos of amhy − and gsdf -/-mutants to knockdown (KD) its expression (S10A Fig) .Consistent with our previous studies [14,56], the gonads of the F0 foxl2 XX mutants with high mutation rate, named as foxl2 KD , developed as testes at 60 dah (S10B Fig) .Different from the ovary phenotype of XY amhy -and gsdf -/-single mutants, testis development was observed in the XY amhy − ; foxl2 KD (n = 14 for 60 dah, n = 5 for 45 dah) and XX/XY gsdf -/-;foxl2 KD (n = 10 for 60 dah, n = 7 for 45 dah) double mutants, which was demonstrated by the presence of Cyp11c1 and dmrt1 expressions, and absence of female specific Zar1 (oocyte) and Cyp19a1a expressions (Figs 6B, S10C and S10D).In addition, the testes of these two double mutants were filled with all types of spermatogenic cells at 120 dah (S10E Fig) .Furthermore, whole-mount FISH analysis showed that dmrt1 mRNA was expressed in the gonads of XY amhy − ;foxl2 KD double mutants, different from the XY amhy − mutants at 10 dah.dmrt1 mRNA was expressed in the gonads of XY gsdf -/-;foxl2 KD double mutants, same as the XY gsdf -/-mutants at 10 dah.Consistent with our previous studies [12,43], whole-mount IF analysis showed that Cyp19a1a expression was detected in the gonads of XY amhy − but not XY gsdf -/-fish at 10 dah.However, Cyp19a1a was not expressed in the gonads of both XY amhy − ;foxl2 KD and XX/XY gsdf -/-; foxl2 KD double mutants at 10 dah (Fig 6C).Thus, disruption of female pathway gene foxl2 can completely rescue the male-to-female SR of amhy and gsdf mutants.

Discussion
The gonadal fate is controlled and maintained by the antagonistic roles between male and female pathway genes.In recent years, the amhy, dmrt1, gsdf as male pathway genes and foxl2, cyp19a1a, foxl3 as female pathway genes have been identified in tilapia, which provide a good model to elucidate genetic interactions between male and female pathway genes in controlling gonadal fate.Previous work from our group demonstrated that the gonads of the dmrt1;foxl3 double mutants develop as ovaries and the gonads of the dmrt1;foxl2 double mutants develop as dysgenesis testes with no germ cells [14].In this study, we found that the gonads of 4 double mutants, including amhy;cyp19a1a, gsdf;cyp19a1a, amhy;foxl2, gsdf;foxl2 and 2 triple mutants, amhy;cyp19a1a;cyp19a1b and gsdf;cyp19a1a;cyp19a1b, develop as functional testes, and the gonads of 2 double mutants, amhy;foxl3 and gsdf;foxl3, develop as ovaries but with spermatogenesis.In contrast, the gonads of the dmrt1;cyp19a1a double mutants (type I), dmrt1; cyp19a1a;cyp19a1b and dmrt1;cyp19a1a;foxl3 triple mutants develop as ovaries.An intriguing finding of this study is that mutation of the female pathway genes foxl3, foxl2, cyp19a1a partially or completely rescues the SR in amhy and gsdf but not dmrt1 mutants.In line with this, disruption of estrogen synthesis by AI treatment rescues the SR in amhy and gsdf but not dmrt1 mutants.Collectively, our results comprehensively reveal epistatic interactions of male and female pathway genes identified in controlling gonadal fate, and suggest that dmrt1 is the only male pathway gene tested indispensable for SD and functional testis development in tilapia.

Once dmrt1 was mutated, the gonad could not be rescued to a functional testis by mutation of any female pathway gene
It is well accepted that estrogen is essential for female sexual fate decision and maintenance in non-mammals [57][58][59].However, this view is challenged by recent studies in zebrafish, tilapia and even in chicken that the ovary of dmrt1 mutants cannot be rescued into testis by mutation of cyp19a1a or/and estrogen receptor esr1/esr2a/2b or AI treatment [14,32,39,40].Similarly, AI treatment fails to induce female-to-male SR in XX spotted scat (Scatophagus argus) which lacks the functional dmrt1 copy (dmrt1 is located on the sex chromosome, and the X-linked dmrt1b is truncated) [58].In turtle, the gonads are feminized even in male promoting temperature after dmrt1 knockdown [33].In line with this, in this study, we provide new data that the gonads of over half of dmrt1;cyp19a1a double mutants, all dmrt1;cyp19a1a;cyp19a1b triple mutants and AI treated-dmrt1;cyp19a1a double mutants develop into ovaries in tilapia.These results provide additional evidences to support the conclusion that SR induced by blockage of E2 synthesis is dependent on Dmrt1 [14].If dmrt1 is mutated, the gonad cannot be rescued to testis by disruption of cyp19a1a or AI treatment.
foxl3 is identified as a crucial factor for feminizing germ cell in medaka, tilapia and zebrafish [14,27,28].Loss of dmrt1 induces cyp19a1a and foxl3 expression in XY tilapia [14].In the present study, loss of cyp19a1a or cyp19a1a;cyp19a1b disrupted foxl3 expression in XX tilapia, indicating that foxl3 expression is dependent on the presence of E2.However, foxl3 is expressed in the ovaries of dmrt1;cyp19a1a;cyp19a1b triple mutants, suggesting that foxl3 expression can be independent of E2 when dmrt1 is absent.Loss of both cyp19a1a and foxl3 in XX tilapia leads to testicular development, which agrees with our previous study [14], but different from the medaka that AI fails to disrupt oocyte formation in foxl3 mutants [27].However, the gonads of the dmrt1;foxl3 double, dmrt1;cyp19a1a;foxl3 triple and AI treated-dmrt1; foxl3 double mutants eventually developed as ovaries in XX and XY tilapia.Therefore, up-regulation of cyp19a1a and foxl3 in dmrt1 mutants is the consequence, not the cause, for the male-to-female SR in dmrt1 mutants.These results suggested that both cyp19a1a and foxl3 are not essential for female fate in the genetic context of dmrt1 deletion.
In mouse, somatic cells are observed to express Gata4 and Sox9 (Sertoli cell markers) when both Dmrt1 and Foxl2 are conditionally knockout in Sertoli cells [60], indicating the mansculization of somatic cells in the double knockout.The somatic cells are also masculinized in the gonads of XX/XY dmrt1;foxl2 double mutant tilapia [14].Consistently, the somatic cells in the gonads of less than half of XX/XY dmrt1;cyp19a1a double mutants are masculinized in the present study.These results suggest that dmrt1 is dispensable for male somatic cells maintenance when its antagonist foxl2 is mutated in vertebrates.
Besides its role in SD, Dmrt1 is also critical for male germ cell development and survival in vertebrates.In mouse, conditional knockout of Dmrt1 in XY germ cells leads to apoptosis [61].In zebrafish, some dmrt1 single mutants develop into dysgenesis testes with no germ cells [34,40].Loss of rbpms2a;rbpms2b or bmp15 results in testis development [62,63], while triple mutation of dmrt1;rbpms2a;rbpms2b and double mutation of dmrt1;bmp15 result in underdeveloped testis and sterility in adult zebrafish [39].In tilapia, less than half of dmrt1;cyp19a1a and all the dmrt1;foxl2 double mutants have underdeveloped testis with no germ cells [14].These studies reveal that even though the dmrt1 mutants are rescued to males, the germ cells cannot be maintained in male somatic environment due to lack of dmrt1.The molecular mechanism underlying Dmrt1 in maintaining male germ cells deserves further investigations.In contrast, dmrt1 is dispensable for female germ cells in fish, as the ovaries of dmrt1 mutants contain germ cells in zebrafish [34,40], XX tilapia as well as sex reversed XY medaka, XY tilapia [14,35].The germ cells in the ovaries of dmrt1 mutants in chickens and rabbits cannot undergo meiosis and folliculogenesis [31,32], but complete meiosis and folliculogenesis have been observed in the ovaries of dmrt1 mutants in zebrafish, medaka and tilapia, irrespective of their genetics sex (sex reversed or not).The dmrt1 female mutant is fertile in medaka and zebrafish [34,35,40], while its fertility in other fish remains to be investigated.
Overall, it is plausible that as long as dmrt1 is mutated, the mutant gonads could not be rescued to functional testes by mutation of any female pathway gene identified so far.

Once dmrt1 was present, sex reversal caused by mutation of its upstream and downstream male pathway gene could be rescued
Mutation of amhy or gsdf resulted in up-regulation of foxl2 and cyp19a1a/Cyp19a1a expressions and male-to-female SR in tilapia [9,12], which can be rescued to functional testis by AI treatment [43,64].Consistently, we found that the gonads of gsdf;cyp19a1a, amhy;cyp19a1a, amhy;foxl2, gsdf;foxl2 double and amhy;cyp19a1a;cyp19a1b, gsdf;cyp19a1a;cyp19a1b triple mutants developed into functional testes.These results further suggested that induced cyp19a1a and foxl2 expression is responsible for driving the SR in amhy and gsdf mutants.The primary role of TGF-β signal (amhy and gsdf) is to repress female pathway genes, which is supported by the studies in tilapia and gibel carp that amhy and gsdf inhibited cyp19a1a transcription [43,44].It has been demonstrated that expression of dmrt1 is up-regulated in XX cyp19a1a and foxl2 mutants and AI treated-XX tilapia [13,48], but is down-regulated in XY amhy and gsdf mutants [9,12].However, dmrt1 is expressed in the gonads of the amhy;foxl2, gsdf;foxl2 double and amhy;cyp19a1a;cyp19a1b, gsdf;cyp19a1a;cyp19a1b triple mutants at 10 dah.On the other hand, the gonads of dmrt1;cyp19a1a;cyp19a1b triple mutants developed as ovaries due to loss of dmrt1.Based on these results, we speculate that up-and down-regulation of Dmrt1 is the cause, instead of the consequence, of the testicular and ovarian development in these mutants, respectively.Finally, studies have shown that gsdf functions downstream of dmrt1 because the gsdf transcription is directly activated by Dmrt1 in fish species, including tilapia, spotted scat and gibel carp [12,44,65].Overall, the SR caused by mutation of male pathway genes other than dmrt1, including its upstream MSD gene amhy and downstream gsdf, could be rescued to functional testis by either mutation of the female pathway gene or AI treatment to induce Dmrt1 expression.
Single mutation of amhy, dmrt1 and gsdf results in ovary development in XY tilapia [9,12,14].By contrast, the gonads of both XX/XY dmrt1;foxl2, gsdf;foxl2 double mutants and XY amhy;foxl2 double mutants developed as testes.These results suggest that testis development is dependent on amhy, dmrt1 and gsdf in the presence of foxl2, but independent of amhy, dmrt1 and gsdf in the absence of foxl2.There might be an unknown gene that triggers the testis development in these double mutants.In other words, foxl2 is essential for female fate decision in the absence of amhy, dmrt1 and gsdf.
Recently, we have demonstrated that mutation of amhy and gsdf results in down-regulation of dmrt1 expression in XY tilapia [9,12], while mutation of foxl3 leads to spermatogenesis in the ovarian environment in XX tilapia due to up-regulation of dmrt1 in germ cells [14].In the present study, spermatogenesis in ovarian somatic environment and up-regulation of dmrt1 in germ cells was observed in XY amhy;foxl3 and gsdf;foxl3 double mutants, same as the XX foxl3 single mutants.These results further support the notion that up-regulation of dmrt1 is the cause, instead of the consequence, of the spermatogenesis in the ovarian environment, and reveal close crosstalk between somatic and germ cells during SD.Antagonism of Dmrt1 and Foxl3 in germ cell determines its sexual fate and their expressions rely on absence or presence of E2.If foxl3 is absent or mutated, E2 is unable to feminize the germ cells.This is probably supported by the fact that foxl3 gene is lost in the genome of eutherian mammals in which SD is not affected by estrogen.Importantly, sperm was produced in the XX foxl3 mutant of medaka [27] and tilapia (S11 Fig) .Therefore, even in the female somatic cell environment, upregulation of dmrt1 in germ cells produces functional sperm.Our data strongly suggest that germline sexual fate is acquired by foxl3 in females and by dmrt1 in males in tilapia.

Indispensable role of Dmrt1 in male might be attributed to its expression in both somatic and germ cell
A large number of studies have shown in mouse, chicken and fish that Dmrt1/dmrt1 is expressed in both germ cells and somatic cells [29,[66][67][68].In tilapia, female pathway genes foxl2 and cyp19a1a are expressed exclusively in somatic cells [13,69], while foxl3 is expressed exclusively in germ cells [14].Of the tilapia male pathway genes, amhy and gsdf are expressed exclusively in somatic cells [9,12], while dmrt1 is expressed in both germ cells and somatic cells [14,70].Recently, we have demonstrated that amhy inhibits the transcription of cyp19a1a through Amhr2/Smads signaling to disrupt E2 synthesis, which in turn, induces dmrt1 expression in somatic cells in XY tilapia during the critical period of SD [43].In XY tilapia, up-regulation of Dmrt1 in germ cells directly inhibits foxl3 transcription [14], and up-regulation of Dmrt1 in somatic cells directly represses cyp19a1a transcription and directly or indirectly inhibits foxl2 expression to initiate and maintain the testicular fate [53].In XX tilapia, repression of Dmrt1 expression in somatic and germline by estrogen, which is promoted by foxl2 and sf-1 [21], is the crucial step for initiating and maintaining ovarian fate.In this study, by generation of 9 double and 4 triple mutants, we provided strong evidences to support this proposed SD mechanism in tilapia.Based on these results, a model for the regulatory architecture in tilapia SD and sex maintenance is proposed in Fig 7 .Therefore, the indispensable role of Dmrt1 in testicular differentiation and development in male might be attributed to its expression in both somatic and germ cell.Additionally, doublesex (dsx), encodes a transcription factor with DM domain, that determines sex in invertebrates, including the fruit fly (Drosophila), nematode (Caenorhabditis elegans) and silkworm (Bombyx mori) [71][72][73][74].Therefore, dmrt1 is commonly used as a key male pathway gene in vertebrates and even in some invertebrates.

Conclusions
In this study, we analyzed 12 double and 4 triple mutants, including 3 double mutants established previously, in Nile tilapia.Based on the gonadal phenotypes and gene expressions of these mutants, we found that once dmrt1 is mutated, the gonad cannot be rescued to a functional testis by mutation of any female pathway gene.The SR caused by mutation of male pathway genes other than dmrt1 can be rescued by mutation of the female pathway gene or AI treatment (S12 Fig) .Taken together, our results demonstrated for the first time that dmrt1 is the only male pathway gene tested so far indispensable for SD and functional testis development.Our work may shed some light on the genetic interaction between female and male pathway genes in tilapia, which enhances our understanding of SD in vertebrates.

Ethics
All animal experiments were conducted in accordance with the regulations of the Guide for Care and Use of Laboratory Animals and were approved by the Committee of Laboratory Animal Experimentation at Southwest University.

Nile tilapia
The Nile tilapia, Oreochromis niloticus, were reared in recirculating freshwater aquarium at 26˚C with a photoperiod of 14 hours of light and 10 hours of dark.The experimental fish were fed three time a day with commercial dry food (Shengsuo, China).In our cultured conditions, all-XX and all-XY tilapia develop as females and males, respectively.
Previously, our studies showed that mutation of foxl2 in F0 XX tilapia by TALEN induced female-to-male SR [14,56].In this study, in vitro synthesis of mRNA for foxl2 TALENs was carried out with Sp6 mMESSAGE mMACHINE Kit (Ambion, USA).The synthesized mRNA was purified using a MEGAclear Transcription Clean-Up Kit (Ambion, USA) according to the manufacturer's instructions.About 500-800 pg mRNA was microinjected into embryos obtained from crossing gsdf +/-XX females with gsdf -/-XY males or embryos obtained from crossing amhy − XY neo-females with XX neo-males using the WPI Micro-injector.The gsdf -/- XY males and XX neo-males were obtained by administration of letrozole (200 μg/g diet) from 5 to 30 dah in our laboratory.The injected embryos were hatched at 26˚C and collected to extract genomic DNA (gDNA) for mutation analysis at 45 and 60 dah as described previously [14].

DNA extraction, genotyping and genetic sex identification
gDNA was extracted from a piece of caudal fin by proteinase K digestion followed by phenol/ chloroform extraction.The quality and concentration of gDNA were assessed by agarose gel electrophoresis, and measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA).Then, the gDNA was diluted to a concentration of 100 ng/μl for use.DNA fragments spanning the target site of each gene were amplified using the primers listed in S1 Table .The amplified DNA fragments were recovered using a QIAquick PCR Purification Kit (QIAGEN, Germany) according to the manufacturer's instructions.Restriction enzyme digestion was carried out for distinguishing different genotypes of the single, double and triple mutants established in this study, including wild-types, heterozygotes and homozygotes as described previously [9,[12][13][14]42,43].To determine whether foxl2 mutations were generated in the amhy and gsdf mutants, restriction enzyme digestion was performed to evaluate mutation efficiency in each fish as described in our previous studies [14,56].
Genotypic sex of all experimental fish was determined by genomic PCR using tilapia sex specific marker-5 developed previously [76].Genotyping PCR was performed using GoTaq PCR Master Mix (Promega, USA).Cycling parameters were as follows: 95˚C for 3 min, 95˚C for 30 s, 58˚C for 30 s, 72˚C for 40 s, 72˚C for 10 min (steps 2 to 4 run for 36 cycles).Primer sequences were listed in the S1 Table.

Drug treatment
In the treatment group, the newly hatched fry were fed with commercial diet sprayed with 95% ethanol containing letrozole (the third-generation aromatase inhibitor) (Sigma-Aldrich, USA) at a concentration of 200 μg/g feed from 5 to 30 or 60, 120 dah.The control group fish were fed with normal commercial diet sprayed with 95% ethanol only.Later on, all fish were fed with normal commercial diet.The treatment experiments were repeated three times.The gonad phenotype and the gene expressions were determined at different time points.

Gonadal histological examination
The control and mutant fish were sampled at different time points for phenotype analysis.Firstly, the experimental fish were anesthetized by immersion in 0.16 mg/ml tricaine methanesulfonate (MS-222) (Sigma-Aldrich, USA).Secondly, the gonad samples were fixed in Bouin's solution for 24 h at room temperature.They were then dehydrated and embedded in paraffin.Tissue blocks were sectioned at 5 μm thickness using the Leica microtome and stained with Hematoxylin and Eosin (H&E) as described previously [53].The images were taken using the Olympus BX53F microscope.

RNA extraction, RT-PCR and real-time PCR
Different tissues, including liver, ovary, testis and brains, were collected from different genotype individuals (usually three to five parallel samples per genotype) for gene expression analysis.Samples were immediately frozen in liquid nitrogen.Total RNAs were extracted from all samples using a column-based RNA extraction kit (Qiagen, USA) specialized for small quantities of RNA.The concentration of total RNA was quantified by using NanoDrop 2000.DNase I (RNase free) treatment and cDNA preparation were carried out using PrimeScript RT reagent Kit with gDNA Eraser (Takara, Japan) according to the manufacturer's instructions.RT-PCR was performed to check the gene expression according to the methods previously described [69].Real-time PCR was performed with Fast SYBR Green Master Mix (Takara, Japan) on a 7500 Fast Real-Time PCR system (Applied Biosystems, USA).Housekeeping gene β-actin was used as the internal control, which was well working in our previous study [13].The expression of target genes in each genotype was normalized to that of the β-actin.The relative abundance of mRNA transcripts was calculated using the formula: R = 2 -ΔΔCt as described previously [77].At least three samples for each genotype were analyzed.Primer sequences used for RT-PCR and real-time PCR were listed in S1 Table.

Immunofluorescence (IF) and whole-mount IF
For immunofluorescence, the sections were permeabilized with 1% Triton X-100 in PBS for 10 min and then blocked in 5% bovine serum albumin (BSA)/PBS for 30 min at room temperature.The sections were then incubated with primary antibodies in 5% BSA/PBS overnight at 4˚C.For whole-mount immunofluorescence (IF), gonads were collected and fixed in 4% PFA overnight at 4˚C.The gonads were permeabilized with 100% acetone for 20 min before primary antibody incubation.The following rabbit polyclonal antibodies were prepared by our laboratory: Amh (This antibody can recognize both Amh and Amhy proteins), Gsdf, 3β-HSD-I, Vasa, 42Sp50, Zar1, Cyp19a1a, Cyp19a1b, Creb1b, Sox30 and Cyp11c1 (S2 Table ).The dilution and specificity of these antibodies have been analyzed previously [9,[12][13][14]42,43,78,79].Goat anti-rabbit antibody Alexa Fluor 488-and 594-conjugated secondary antibodies (Thermo Fisher Scientific, USA) were diluted to 1:500 in blocking solution and incubated with samples overnight at 4˚C to detect the primary antibodies.The nuclei were stained by 4', 6-diamidino-2-phenylindole (DAPI) using VECTASHIELD Mounting Medium with DAPI (Vector, USA).At least three independent biological replicates for each genotype were analyzed.Fluorescence signals were captured by confocal microscope (Olympus FV3000, Japan).

Fluorescence in situ hybridization (FISH) and whole-mount FISH
For FISH, the gonads of WT and mutant fish were collected at indicated time.The samples were fixed in 4% paraformaldehyde (Sigma-Aldrich, USA) in 0.85× PBS at 4˚C overnight.Specimens were embedded in paraffin and sectioned at 5 μm.For whole-mount FISH, larvae were washed with PBS plus Triton X-100 for three times to increase tissue permeability and improve staining while preserving the antigen epitopes.FISH and whole-mount FISH were performed to examine the gene expression as described previously [80].Probes for foxl3 and dmrt1 antisense or sense digoxigenin-labeled RNA strands were transcribed in vitro from a linearized pGEM-T easy-foxl3 and dmrt1 plasmid using the T7 RNA labeling kit (Roche, Swiss).The sections were deparaffinized, rehydrated, and digested with proteinase K at 37˚C for 20 min.After refixation by 4% paraformaldehyde, the sections were hybridized with DIG-labeled RNA probe at 65˚C overnight.For more sensitive FISH and whole-mount FISH detection, the tyramide signal amplification (TSA) Plus Fluorescence Systems (PerkinElmer Life Science, USA) were used according to the manufacturer 0 s instructions.After mounting with VECTA-SHIELD Mounting Medium with DAPI (Vector, USA), the images were recorded using confocal microscope (Olympus FV3000, Japan).Primers for preparing DIG-labeled probes are list in S1 Table .The fluorescence intensity of Cyp19a1b staining was quantified using Image J software as described previously [81].At least 4 sections from different fish are randomly selected for fluorescence signal intensity quantification.

Measurement of serum hormone level by EIA and UPLC-MS/MS
Blood samples were collected from the caudal veins of experimental fish, and kept at 4˚C overnight.Serum was collected after centrifugation and stored at −80˚C until use.Serum estradiol-17β (E2) level was measured using Estradiol ELISA Kit (Cayman, USA) according to the manufacturer 0 s instructions.Absorbance was measured at a wavelength of 412 nm using a Multiskan GO microplate reader (Thermo Fisher Scientific, USA).To further confirm the serum estrogen level, we adopted Ultra Performance Liquid Chromatography Tandem Mass Spectrometry (UPLC-MS/MS) to measure it as reported previously [83].

Detection of apoptosis by TUNEL
The gonads for Terminal Deoxynucleotidyl Transferase mediated dUTP Nick-End Labeling (TUNEL) experiment were fixed with 4% paraformaldehyde (PFA) for 12 hours at room temperature.They were then dehydrated and embedded in paraffin.Tissue blocks were sectioned at 5 μm and permeabilized by proteinase K for 15 min, and subsequently fixed again with 4% PFA for 30 min at room temperature.For TUNEL assay, the sections were incubated in 1:10 dilution of 50 μl reaction mixture using In Situ Cell Death Detection Kit (Roche, Swiss) at 37˚C for 60 min in the dark followed by washing with 1× PBS.The sections were then incubated in blocking solution for 2 hours at room temperature and stained with anti-Cyp19a1b antibody for 2 hours at 37˚C.Following washes in 1× PBS, the sections were incubated with 1:500 donkey anti-rabbit Alexa Fluor 488 (Thermo Fisher Scientific, USA) for 2 hours at room temperature, followed by washes in 1× PBS.After mounting with VECTASHIELD Mounting Medium with DAPI (Vector, USA), the images were recorded using confocal microscope (Olympus FV3000, Japan).

Sperm morphology and motility analyses
For WT XY and XX foxl3 -/-fish at 120 dah, we dissected and obtained gonads, and crushed them to obtain gonadal suspension, which were collected in a clean and sterile 1.5 mL centrifuge tube.For sperm morphology analysis, sperm of WT XY and XX foxl3 -/-fish at 120 dah were stained with the fluorescent membrane dye PKH26 (Sigma-Aldrich, USA), which incorporates aliphatic reporter molecules into the cell membrane.Approximately ten million cells were suspended in 0.4 ml of diluent C (anisoosmotic aqueous solution provided with the PKH26 dye) in which PKH26 was diluted at a ratio of 4 μl dye to 0.4 ml diluent C. The diluted dye was incubated with the cells (final concentration, 10 μmol/l) for 5 min.The cells were then centrifuged at 100×g for 5 min, washed twice in L-15 (HyClone, USA), resuspended in L-15, and then observed with confocal microscope (Olympus FV3000, Japan) [84].For sperm motility analysis, sperm was collected from WT XY and XX foxl3 -/-fish at 120 dah.After a dilution at 1:10 with phosphatebuffered saline, 10 mL diluted semen was added to the counting chamber of the sperm count plate and captured on the stage of a Leica DM500 light microscope (Leica, Germany) [55].

Statistical analysis
All the experiments were conducted for at least three times.Data were analyzed using Graph-Pad Prism 8 software and expressed as the mean ± SD.Significance of differences among more than two groups were calculated using One-way ANOVA, followed by Tukey test for multiple comparisons.For all statistical analyses, P<0.05 indicated a significant difference labeled by different letters in the figures, and groups labeled with the same letter were not significantly different from each other.

Fig 7 .
Fig 7. A proposed model for tilapia female and male sex determination and maintenance.In XY tilapia, repression of Cyp19a1a expression/E2 synthesis by the MSD gene amhy in somatic cells is the prerequisite for up-regulation of Dmrt1 expression in both somatic and germ cells to initiate testis development.Dmrt1 and its downstream target Gsdf repress cyp19a1a and foxl2/foxl3 expression to maintain male sex.In XX tilapia, induction of Cyp19a1a expression/E2 synthesis by foxl2 is necessary to initiate ovarian differentiation by repressing dmrt1 expression and activating of foxl3 expression.Continuous expression of Cyp19a1a for estrogen production is required to maintain ovarian fate by sustaining high expression of foxl2 in female, suggesting a positive feedback loop between estrogen and foxl2.The signals from somatic cells (amhy and estrogen) determine the fate of germ cells by affecting Dmrt1 and Foxl3 expression.https://doi.org/10.1371/journal.pgen.1011210.g007 Nuclei were counterstained with DAPI.Scale bars = 40 μm.WT, wild-type.dah, days after hatching.(TIF) S6 Fig. Double mutation of foxl3 and cyp19a1a resulted in testis development in XX tilapia.Expression of somatic cell and germ cell markers in the gonads of WT XX, WT XY, XX cyp19a1a -/-, XX foxl3 -/-and XX cyp19a1a -/-;foxl3 -/-tilapia were analyzed by IF at 120 dah.Cyp19a1a, a female somatic specific marker.42Sp50, an oocyte marker.Cyp11c1, a Leydig cell marker.Gsdf, a somatic cell marker expressed highly in males compared with females.Creb1b, a spermatocyte marker.Nuclei were counterstained with DAPI.WT, wild-type.dah, days after hatching.Scale bars = 40 μm.(TIF) S7 Fig.The gonads of dmrt1 -/-;foxl3 -/-double mutants developed as ovaries after withdrawal of AI treatment.(A)