Dmrt1 is required for primary male sexual differentiation in Chinese soft-shelled turtle Pelodiscus sinensis

In vertebrates, the primary sex-determining signals that initiate sexual development are remarkably diverse, ranging from complete genetic to environmental cues. However, no sex determination-related genes have been functionally identified in reptiles. Here, we characterized a conserved DM domain gene, Dmrt1, in Chinese soft-shelled turtle Pelodiscus sinensis (P. sinensis), which exhibits ZZ/ZW sex chromosomes. Dmrt1 exhibited early male-specific embryonic expression, preceding the onset of gonadal sex differentiation. The expression of Dmrt1 was induced in ZW embryonic gonads that were masculinized by aromatase inhibitor treatment. Dmrt1 knockdown in ZZ embryos by RNA interference resulted in male to female sex reversal, characterized by obvious feminization of gonads, significant down-regulation of testicular markers Amh and Sox9, and remarkable up-regulation of ovarian regulators, Cyp19a1 and Foxl2. Conversely, ectopic expression of Dmrt1 led to largely masculinized genetic females, production of Amh and Sox9, and a decline in Cyp19a1 and Foxl2. These findings demonstrate that Dmrt1 is both necessary and sufficient to initiate testicular development, thereby acting as an upstream regulator of the male pathway in P. sinensis.

Chinese soft-shelled turtle Pelodiscus sinensis (P. sinensis), belonging to the Reptilia, Testudines, turtle family, is an important aquaculture species in southern China. This species exhibits a sex-dependent dimorphic growth pattern. Compared with females, male individuals are characterized by larger body size, faster growth, thicker and wider calipash, and less fat 1 . Recently, researchers have been exploring the sex control approach to give rise to all male turtle offspring. The sex determination mode of P. sinensis is generally categorized as genetic sex determination (GSD), as evidenced by the existence of heteromorphic ZZ/ZW micro-sex chromosomes 2,3 . However, little is known about the genetic components involved in primary sex determination and gonadal differentiation in P. sinensis.
In vertebrates, the primary sex-determination signals that initiate gonadal differentiation vary remarkably, ranging from complete genetic to environmental cues (for example, temperature-dependent sex determination, TSD). Most studies of primary genetic signals have been focused on the identification of sex-determining genes and understanding its downstream molecular genetic network. For example, Sry, the first discovered SD gene, has been proven to function as the master initiator for testicular development by activating its direct target Sox9 transcription during a crucial period in eutherian mammals [4][5][6][7][8] . Besides Sry, several other SD genes or candidates have been identified sequentially, such as Dmrt1 in chicken 9,10 , Dmw in frog 11,12 , and Dmy 13,14 , Amhy 15 , Amhr2 16 , SdY 17 , Gsdf 18 , Sox3 19 and Gdf6Y 20 in fish 21 . However, functional characterization of SD-related genes has not yet been reported in reptilian species.
In spite of the large diversity in primary genetic sex determination signals, the Dmrt1 gene that codes for a transcription factor with a DNA-binding motif (DM domain), seems to be an evolutionary conserved factor that participates in primary sex determination and postnatal testicular differentiation 22 . Dmrt1 exhibits a sexually dimorphic expression pattern and knock out/down of this gene causes male to female sex reversal in many taxa 9,13,23,24 . Indeed, in some non-mammalian vertebrates lacking Sry, Dmrt1 or its homologous gene acts as a regulator that is analogous to Sry in primary sex determination and sexual differentiation. For example, in chicken 9,10 and medaka fish Oryzias latipes 13,14 , Dmrt1 and Dmy, located on Z and Y sex chromosome respectively, have been demonstrated to be necessary and sufficient to initiate male development. In the frog Xenopus laevis, a duplicated and truncated Dmrt1 gene on the female-specific W chromosome, Dmw, is required for female sex determination 11 . Interestingly, the Dmw gene triggers ovarian development via blocking the ability of the autosomal Dmrt1 gene to determine testis fate 12 . In reptiles, DM domain genes have mainly been cloned and characterized in several species exhibiting TSD. Male-specific up-regulation of Dmrt1 in early developing gonads, prior to the onset of gonadal differentiation, has been observed in the red-eared slider turtle 25,26 , suggesting a possible upstream role of Dmrt1 in determining the fate of the bipotential gonad in turtles. However, a complete functional study of Dmrt1 in turtles has not yet been performed, which is largely due to lack of an effective genetic manipulation approach.
Herein lentiviral vector-mediated RNA interference and over expression system was introduced into embryos of P. sinensis to manipulate the ectopic expression of Dmrt1, elucidating the functional role of Dmrt1 on primary sex determination in P. sinensis. The male-specific expression of Dmrt1 was identified in the nucleus of pre-Sertoli cells during the sex determination period, which preceded the initiation of gonadal sex differentiation. The expression of Dmrt1 was induced in genetic female embryonic gonads that were masculinized by aromatase inhibitor treatment. Loss-and gain-of-function analyses provided the solid evidence that Dmrt1 is necessary for primary male sexual differentiation, and its ectopic expression is sufficient to trigger testicular development, thereby functioning as an upstream regulator in male sexual differentiation in P. sinensis.  Fig. 1b). The phylogenetic tree showed that P. sinensis Dmrt1 was evolutionarily most closely related to the red-eared slider turtle, followed by chicken, human and mice, and distantly related to fish ( Supplementary Fig. 1c).

Characterization of
In this study, the mRNA and protein expressions of Dmrt1 were first examined in different tissues of adult turtle. Using RT-PCR, qRT-PCR and Western blot analyses (Fig. 1a,b,c and Supplementary Fig. 2a,b), we determined that Dmrt1 was abundantly expressed in the testis, but was not detected in the ovary, heart, liver, spleen, lung, kidney, and muscle. In addition, we examined the cellular localization of Dmrt1 protein in 1-year, 2-year and 3-year old testes of P. sinensis by immunofluorescence. Dmrt1 protein was localized in the nucleus of Sertoli cells surrounding the spermatogonia in testes but was not detected in germ cells of different developmental stages. Meanwhile, we examined the Dmrt1 localization in 3-year old ovaries. As expected, no Dmrt1 fluorescence signal was observed in ovaries of this stage (Fig. 1d).
Dimorphic expression pattern of Dmrt1 in early embryonic gonads. To find out whether Dmrt1 gene is involved in male sexual differentiation in P. sinensis, we first analyzed the expression pattern of Dmrt1 mRNA in early embryonic gonads of different developmental stages by qRT-PCR. Here, each embryo's sex was identified by karyotype analysis and qPCR of 18S rRNA gene. Dmrt1 transcript was detected in male gonads throughout the period of gonadal sex differentiation from as early as stage 15, and increased dramatically to a peak of stage 23. In contrast, female gonads exhibited extremely low expression of Dmrt1 throughout embryogenesis (Fig. 2a).
We then proceeded to examine the expression and cellular localization of Dmrt1 protein in early embryonic gonads through immunofluorescence. Here, β-catenin (CTNNB1) immunofluorescence was performed to better display the structure of turtle embryonic gonads 27 . The embryonic gonads were morphologically undifferentiated and appeared identical between the sexes at stage 16 and 17 (prior to the onset of sexual differentiation). At this time, Dmrt1 protein was already strongly expressed throughout the medulla of male gonads, whereas it was not detected in females (Fig. 2b). At stage 19, formation of the primitive seminiferous cord occurred only in medullar region of male gonads. Dmrt1 protein was abundantly expressed and mainly localized in the nuclei of precursor Sertoli cells surrounding the primordial germ cells in male gonads. From stage 21 to 27, Dmrt1 protein continued to be widespread in the medulla of male gonads, and was maintained at a high expression level. However, Dmrt1 protein signal was not detectable in female gonads throughout embryogenesis (Fig. 2b).
In addition, we analyzed the chronological expression of Dmrt1, Amh, and Sox9, which are key factors involved in male development. qRT-PCR analyses revealed that initiation of Dmrt1 and Amh expression in male gonads occurred before gonadal sex differentiation, and the onset of Dmrt1 at stage 15 preceded that of Amh at stage 16. Sox9 transcript was first detectable at stage 18, when gonadal sex differentiation had begun (Fig. 2c).
Up-regulation of Dmrt1 in ZW embryonic gonads during female to male sex reversal. Previous studies in our laboratory have demonstrated that treatment of exogenous estradiol-17β (E 2 ) at periods of primary sex determination were able to induce ZZ embryos to develop towards the female phenotype (male to female sex reversal), whereas aromatase inhibitor (AI) was capable of masculinizing ZW embryos (female to male sex reversal) 28 . Here, genotype of all E 2 /AI-treated embryos was determined by karyotype analysis and qPCR of 18S rRNA gene. The masculinized ZW embryos exhibited male-like morphology with dense medulla and degenerated cortex ( Supplementary Fig. 3a). A significant increase in Amh and Sox9, and down-regulation of Cyp19a1 and Foxl2 were detected in masculinized ZW embryos by qRT-PCR ( Supplementary Fig. 3b). Meanwhile, a thick outer cortex and degenerated medulla, as well as decrease of Amh and Sox9 and up-regulation of Cyp19a1 and Foxl2, were observed in feminized ZZ embryos ( Supplementary Fig. 3).
Here, we analyzed the expression level of Dmrt1 in gonads with sex reversal to further determine whether Dmrt1 expression is associated with testis differentiation. qRT-PCR analysis showed that the mRNA expression of Dmrt1 decreased dramatically in ZZ embryonic gonads during male to female sex reversal induced by E 2 . Dmrt1 expression in feminized genetic males was down-regulated from stage 17, and reached its lowest expression level by stage 25 (Fig. 3a). Whereas in female to male sex reversed embryos after AI treatment, Dmrt1 expression was induced from stage 18, peaked at stage 22 and was maintained at that high level (Fig. 3b). The higher level of Dmrt1 expression in masculinized ZW gonads represented a >20-fold difference (versus that in control ZW gonads), implying active up-regulation of Dmrt1 in males. Stimulatory effects of AI or E 2 treatment on Dmrt1 expression were studied in more detail at the protein level. In mammals, female germ cells enter meiotic prophase where they remain until adulthood, whereas male germ cells arrest in G0/G1 phase of mitosis and do not enter meiosis until sexual maturity 29 . Here, no Foxl2 antibody is available to be successfully used in turtles and that, alternatively, we used meiosis-specific protein γH2AX as the embryonic ovarian marker. The dual immunofluorescence of Dmrt1 and γH2AX showed that in the control ZZ embryonic gonad at stage 27, Dmrt1 protein was abundantly expressed in the nuclei of precursor Sertoli cells, while no γH2AX protein signal was detectable in germ cells. Whereas, γH2AX protein was highly expressed in germ cells, and no Dmrt1 protein fluorescence signal was detected in control ZW embryonic gonads. In E 2 -treated ZZ embryonic gonads, the expression of Dmrt1 protein was almost completely lost, while γH2AX induction was localized in the gonadal cortex, exhibiting a female-like distribution pattern. In ZW embryonic gonads during male to female sex reversal induced by AI, Dmrt1 protein expression was strikingly up-regulated in the medulla, with γH2AX protein expression almost disappearing (Fig. 3c).

Establishment of a lentivrius-mediated gene-modulating method in turtle embryos.
In this study, lentiviral vectors carrying Dmrt1 specific shRNAs or the ORF of Dmrt1 were utilized to knockdown endogenous Dmrt1 transcripts in ZZ gonads or induce ectopic expression of Dmrt1 in ZW gonads, respectively. To test the efficacy of lentivirus delivery into the turtle embryos, GFP expression in turtle embryos injected with virus at stage 14 ( Fig. 4a) were examined at stage 25. Approximately 55% of treated embryos showed global GFP reporter expression, including widespread expression in gonad tissues (Fig. 4b,c). GFP immunofluorescence of gonadal sections further confirmed that embryonic gonads with virus treatment exhibited robust GFP expression, indicating the effective infection of lentivirus in turtle gonadal tissues (Fig. 4d,e,f).
We then examined Dmrt1 expression in embryonic gonads carrying LV-Dmrt1-shRNA (#1, Supplementary  Fig. 4) or LV-Dmrt1-OE to figure out whether the lentivirus-mediated RNA interference or over-expression systems are able to efficiently modulate candidate genes expression. Here, genotype of all lentivirus-treated embryos with high GFP expression was determined by karyotype analysis and qPCR of 18S rRNA gene. qRT-PCR analysis showed that Dmrt1 mRNA in genetic male gonads with LV-Dmrt1-shRNA was significantly reduced from stage 17, compared with that in control ZZ gonads (LV-NC-shRNA). The average inhibiting rate of Dmrt1 mRNA expression by lentivirus-mediated RNA interference was 83.26% (Fig. 4g). In contrast, ectopic expression of Dmrt1 mRNA in ZW gonads carrying LV-Dmrt1-OE was activated at stage 17 and sharply increased from stage 18, with the highest expression level exceeding 20-fold than that in the negative control ZW gonads (LV-empty) (Fig. 4h). Immunofluorescence analysis further revealed that in ZZ gonads with LV-Dmrt1-shRNA, no Dmrt1 protein expression was detected in the feminized cortical region, while the medullar region showed relatively weak Dmrt1 protein signals. In contrast, localized Dmrt1 protein expression was present in the medulla of ZW gonads with LV-Dmrt1-OE (Fig. 4i). Feminization of the ZZ gonads with Dmrt1 knockdown. To investigate the function of Dmrt1 on male development of P. sinensis, phenotype and marker genes expression of embryonic gonads with Dmrt1 knockdown were assessed through gonadal histology, immunofluorescence and qRT-PCR. Gonadal development in embryos treated with non-silencing scramble shRNA or empty vector was normal. Control ZZ gonads were short and cylindrical, while ZW gonads were relatively long and flat. In ZZ embryos infected with LV-Dmrt1-shRNA and showing high GFP expression, gonads became elongated and flat, exhibiting varying degrees of female-like morphology (Fig. 5a). Histological analysis of gonadal sections by H&E staining showed that the control ZZ gonads of stage 27 had a dense medulla with seminiferous cords and a reduced cortex. Whereas the gonads of control ZW embryos possessed a well-developed outer cortex, populated with primordial follicles and a vacuolated medulla. In contrast, the Dmrt1 knockdown ZZ gonads were strongly feminized, characterized by a thickened outer cortex with a large number of germ cells and a highly degenerated medulla (Fig. 5b). The statistic data showed that male proportion in the LV-Dmrt1-shRNA treated group was 14.7% (vs. 48.2% in controls) (Fig. 5c), and 71.8% (28 of 39) of ZZ gonads showed male to female sex reversal (Table 1).
Male and female marker genes were examined at different embryonic stages to confirm the activation of the female developmental pathway in ZZ gonads with Dmrt1 knockdown. At the mRNA level, remarkable down-regulation of testicular differentiation markers Amh and Sox9, and significant up-regulation of ovarian development regulators Cyp19a1 and Foxl2 were observed in ZZ gonads of different stages with Dmrt1 knockdown relative to controls (Fig. 6a). Amh and Foxl2 responded as early as stage 17 to Dmrt1 knockdown, while Sox9 and Cyp19a1 showed a relative delayed response to Dmrt1 knockdown, which occurred from stage 19 onward (Fig. 6a). At the protein level, Sox9 was expressed specifically in the nuclei of precursor Sertoli cells in control male gonads, while control female gonads lacked Sox9 expression. In ZZ gonads with Dmrt1 knockdown,   Sox9 expression almost disappeared; simultaneously germ cells in the feminized cortex were γH2AX positive, exhibiting a female-like distribution pattern (Fig. 6b).
In vertebrates, sex-steroids play an important role in gonadal differentiation. Our above data showed that Dmrt1 expression could be affected by exogenous estradiol-17β (E 2 ) treatment (Fig. 3) and Dmrt1 knockdown was able to regulate the expression of Cyp19a1, which encodes the key enzyme for estrogen synthesis (Fig. 6a). Therefore, it is necessary to analyze whether Dmrt1 knockdown can alter steroid hormone production in turtle embryos. We measured the E 2 and testosterone content in control and treated ZZ embryos at different stages (21, 23, 25 and 27). No significant difference was observed in E 2 and testosterone production between control and Dmrt1 knockdown ZZ embryos at stage 21. Up-regulation of E 2 level in ZZ embryos with Dmrt1 knockdown was observed from stage 23. This is consistent with Cyp19a1 expression that had been remarkably increased in ZZ embryos with Dmrt1 knockdown at this stage (Fig. 6a). At stages 25 and 27, significant differences in both E 2 and testosterone levels were detected in control male and female embryos. Dmrt1 knockdown in ZZ embryos resulted in up-regulation of E 2 level and down-regulation of testosterone level (Fig. 7a,b).

Masculinization of the ZW embryos over-expressing Dmrt1.
To determine if Dmrt1 was sufficient to initiate male development, we investigated the effect of ectopic expression of Dmrt1 in genetic females. In ZW embryos over-expressing Dmrt1, gonads exhibited a short cylindrical structure, a feature specific to control ZZ gonads (Fig. 5a). Histological analysis for gonadal sections showed varying degrees of sex reversal phenotypes in ZW gonads over-expressing Dmrt1, including testes and ovotestes of variable proportions. The ovotestes exhibited a well-developed medulla with seminiferous cord-like structure and germ cells, as well as a partially degenerated cortex still containing germ cells (Fig. 5b). Testes and ovotestes were identified in 84.7% of virus treated embryos (Fig. 5c), and 67.6% (23 of 34) of genetic female embryos showed female to male sex reversal (Table 1).
Up-regulation of Amh and Sox9 and down-regulation of Cyp19a1 and Foxl2 were observed in ZW gonads with Dmrt1 over-expression, determined by qRT-PCR (Fig. 6a). However, these sex-specific genes differed in the earliest response time to Dmrt1 over-expression, with the expression change of Amh, Foxl2, Sox9 and Cyp19a1 occurring from stages 17, 17, 19 and 21, respectively (Fig. 6a). Ectopic activation of Sox9 protein in treated ZW gonads was confirmed by immunofluorescence. Sox9 protein expression was induced and concentrated in the nuclei of Sertoli cells within the masculinized regions (testis cords) of ZW gonads over-expressing Dmrt1, but generally was lower and less widespread compared to control males. γH2AX protein expression was significantly decreased in some partially sex reversed ZW gonads, compared to that in female controls. However, a small number of γH2AX-positive meiotic germ cells still occurred in the cortial region (Fig. 6b). Co-existence of Sox9 and γH2AX further confirmed the partial masculinization of ZW gonads, consistent with the histological analysis results. For sex steroids, over-expression of Dmrt1 in ZW embryos resulted in significant down-regulation of E 2 production and markedly increased testosterone levels at stage 25 and 27 (Fig. 7a,b).

Discussion
In the present study, we provide solid evidence that Dmrt1 gene is both necessary and sufficient for primary testicular differentiation in Chinese soft-shelled turtle, through a newly developed lentivrius-mediated gene-modulating approach.
In P. sinensis, specific expression of Dmrt1 has already appeared in male gonads as early as stage 16, prior to the onset of gonadal sex differentiation, which occurs at stage 17. Subsequently, robust and consecutive expressions of Dmrt1 followed by Amh and Sox9, were detected only in the male gonads throughout embryogenesis. This dimorphic expression pattern indicates the possible involvement of Dmrt1 in primary male sex determination. These results are consistent with previous studies in red-eared slider turtle, Trachemys scripta, that reported Dmrt1 mRNA expression in early male embryos was higher than that in females, preceding Sox9 expression 30,31 . In non-mammalian vertebrates, estrogen plays an important regulatory role in early gonadal sex differentiation, while exogenous estrogen can significantly suppress Dmrt1 expression in male embryos 23,32,33 . In T. scripta, Dmrt1 mRNA expression was significantly inhibited in estrogen-treated male embryos, while in female embryos with AI treatment, Dmrt1 expression was dramatically up-regulated 34,35 . In this study, we found that Dmrt1 expression was significantly reduced in feminized ZZ embryos of P. sinensis treated with E 2 , whereas remarkable up-regulation of Dmrt1 expression was observed in the distinctly masculinized ZW embryos of P. sinensis induced by AI, implying that Dmrt1 associated with male gonadal differentiation of P. sinensis, and repression of Dmrt1 by estrogen signaling is highly conserved in non-mammalian vertebrates, although this process may be indirect. These observations suggest that Dmrt1 gene participates in male sexual development of P. sinensis, likely located upstream of the male pathway.
To date, the complete analysis of Dmrt1 function in sexual differentiation and testicular development has never been performed in reptiles, including turtles, largely due to the unavailability of gene knockout and transgenic techniques. Recently, a number of studies have been focusing on the novel establishment of effective gene-modulating methods in turtle species. Itzel et al. 36 reported a 70-80% reduction of Sox9 mRNA and protein expression in in vitro cultured male gonads of Olive ridley sea turtles, Lepidochelys olivacea, by using RNA interference. In contrast, GFP: Sox9 expression vector was introduced into in vitro female cultured gonads of red-eared slider turtles by electroporation, resulting in a certain amount of ectopic Sox9 expression 37 . However, no in vivo genetic manipulation methods were achieved in turtle embryos. In this study, high-titre lentiviral vectors carrying Dmrt1 specific shRNA or Dmrt1 ORF were injected into P. sinensis embryos at stage 14, leading to the global GFP reporter expression in 55% of treated embryos at stage 25, as well as efficient knockdown of Dmrt1 in ZZ gonads (inhibiting rate >80%) and robust over-expression of Dmrt1 in ZW gonads (over 20 times higher) throughout the embryogenesis of P. sinensis. This lentiviral vector expression system showed obvious infection of turtle embryos and in vivo efficient gene-modulating ability, which provides a powerful tool for gene functional analysis in turtle species.
Using a novel method of genetic manipulation in P. sinensis embryos, we revealed that Dmrt1 knockdown in ZZ embryos caused complete feminization of male gonads, with down-regulation of testicular markers Amh and Sox9 and up-regulation of ovarian regulators Cyp19a1 and Foxl2. Moreover, ZZ gonads with Dmrt1 knockdown exhibited γH2AX positive meiotic germ cells in the thickened outer cortex, which only occurred in female gonads. These observations are quite similar with that in other vertebrates. In ZZ chick embryos, downregulation of Dmrt1 by RNA interference led to the feminization of embryonic gonads, a decline in Sox9 expression and an increase in Cyp19a1 and Foxl2 expression 9 . Dmy knockdown in XY medaka larvae caused increased number of meiotic germ cells, suppressed the male pathway (Gsdf, Sox9a2, etc.) and favored the female cascade (Rspo1, Foxl2, etc.), resulting in a fertile male to female sex reversal 38,39 . Most importantly, in the converse experiment, Dmrt1 over-expression in ZW gonads of P. sinensis triggered the formation of sex cord-like structures, activated the ectopic expression of Amh and Sox9, and antagonized Cyp19a1 and Foxl2 expression, suggesting that Dmrt1 is capable of initiating the male developmental pathway. Likewise, over-expression of Dmy in XX medaka resulted in testicular differentiation 40 . It is known that Dmy can activate expression of autosomal Dmrt1, which is involved in testicular differentiation as a male sex maintainer. Furthermore, evidence has shown that Dmrt1 mutations cause male to female sex reversal after the sex is determined by Dmy in the medaka [41][42][43] . In chick embryos, ZW gonads over-expressing Dmrt1 exhibited a masculinized morphology, as well as activation of Amh and Sox9 10 . These findings demonstrated that Dmrt1 is both necessary and sufficient to initiate male development in P. sinensis. In addition, Amh and Sox9 were upregulated following over-expression of Dmrt1 in ZW gonads, suggesting that ectopic expression of these two genes may be a response to elevated Dmrt1. Based on this expression pattern, implication arises that Dmrt1 lies upstream of Amh and Sox9 in P. sinensis. Taken together, all these observations suggest that Dmrt1 exerts a key upstream regulator role in primary male sexual differentiation of P. sinensis, although it is uncertain whether a Z-linked gene exists upstream of Dmrt1, acting as the male sex-determining gene. Of course, it cannot be excluded that a W-linked gene functions as the female sex-determining gene in P. sinensis, similar to Dmw in Xenopus laevis 11,12 . Thus further research is needed to elucidate the full sex-determining mechanism of P. sinensis.
In summary, we demonstrate here for the first time that Dmrt1 is not only necessary for primary male sexual differentiation, but also sufficient to trigger testicular development in P. sinensis, thereby acting as an upstream regulator of the male pathway. In addition, the lentivirus-mediated genetic manipulation approach worked well in turtle embryos, which provides a powerful foundation for the functional studies of interesting genes in turtle species.  [44][45][46] , were removed from eggshells, decapitated, and placed in PBS for dissection. Gonad-mesonephros complexes (GMCs) and whole-gonads adjacent to mesonephros were dissected from treated and control embryos, respectively. GMCs were fixed in 4% paraformaldehyde (PFA) overnight at 4 °C, dehydrated through a methanol gradient, and stored in 70% ethanol at 4 °C until paraffin embedding and sectioning was performed. Other gonads were broken up thoroughly and stored in TRIzol reagent (Invitrogen, USA) for total RNA isolation. Meanwhile, these embryos were treated by liquid nitrogen grinding and then stored at −80 °C for genomic DNA extraction. Additionally, the tissue samples of heart, liver, spleen, lung, kidney, muscle, brain, testis and ovary from adult turtles were prepared and stored at −80 °C for cDNA cloning and expression analysis. Metaphase chromosome spreads of P. sinensis were prepared from fibroblast cells for karyotype analysis according to previous studies in our laboratory 47 (Supplementary Fig. 5). All animal experiments were carried out following the guidelines approved by the Institutional Animal Care and Use committee at Zhejiang Wanli University, China.
Alignment of deducted amino acid sequences were carried out by Clustal X software and MegAlign program in DNAStar software, while the phylogenetic tree was constructed using the Neighbour-Joining(N-J) method in Mega 6.0 software. The values on the tree represent bootstrap scores of 1,000 trials, indicating the credibility of each branch. The sequences of nucleotide and amino acid used in the phylogenetic analysis were obtained from GenBank (NCBI).

Estradiol-17β and Aromatase Inhibitor Treatments.
Estradiol-17β (E 2 , Sigma, USA) or a non-steroidal aromatase inhibitor (AI) letrozole (Sigma) were administered to eggs before sexual differentiation. The E 2 /AI was dissolved in 95% ethanol at a concentration of 20 μg/μl, and 5 μl of drug was topically applied to the eggshell in the region adjacent to the embryo at developmental stages 15 and 16 (gonadal differentiation normally begins at stage 17). Controls were treated with 5 μl of 95% ethanol. Egg incubations for both treated and control groups were performed under the same condition as previously described. Gonad-mesonephros complexes were dissected from treated and control embryos at stage 25 for histology and immunohistochemistry use. Gonads were separated from adjacent mesonephros at stage 16,17,18,19,20,21,22,23,24,25,26 and hatching time, and preserved for qRT-PCR analysis.
For the generation of lentivirus, 293 T producer cells were transfected with optimized packaging plasmids (pGag/Pol, pRev and pVSV-G) along with pGLV-Dmrt1-shRNA or pGLV-NC-shRNA expression clone constructs by lipofectamine. 24 h post transfection, the transfection mix was replaced by a fresh culture medium (without antibiotics). The virus-containing supernatant was harvested 72 h post transfection, cleared by centrifugation (3000 rpm/min, 15 min, and 4 °C), and then filtered through a 0.45 μm filter (Millipore, USA). Viruses were titrated by adding serial dilutions to fresh 293 T, and assessed using GFP expression after 48 h. Viral titres of approximately 1 × 10 9 infectious units/ml were obtained. Lentivirus aliquots were stored at −80 °C before infection of turtle embryos.
Construction of LV-Dmrt1-OE Vector System. Total RNA isolation was performed on the testis of adult Chinese soft-shelled turtle and then followed by reverse transcription to prepare the cDNA (methodology found above). Based on the complete cDNA sequence of P. sinensis Dmrt1 gene, the open reading frame (1107 bp) was PCR amplified from cDNA using forward primer 5′-CCCCAAATTGTAGAGGCGAACC-3′ and reverse primer 5′-TGAGGGCAGGGCAGAGGAGG-3′. The PCR product was digested with EcoRI and cloned to pGLV-EF1a-GFP (LV-4, GenePharma). The recombinant vector pGLV-GFP-Dmrt1 was named LV-Dmrt1-OE. The empty vector pGLV-GFP-empty was considered as a negative control (LV-empty). High quality proviral DNA was then used to transfect 293 T cells. Virus propagation was carried out as described above.
Infection of Turtle Embryos. High titre virus of LV-Dmrt1-shRNA or LV-Dmrt1 (at least 1 × 10 9 infectious units/ml) was injected into turtle embryos at stage 14, using a metal Hamilton needle(the diameter of the needle: 0.5 mm). Approximately 5 μl was injected per embryo, and a total of 500 eggs were injected in every treated group. Every 300 control embryos were injected with scrambled control virus of LV-NC-shRNA or LV-empty. Eggs were sealed with parafilm and incubated for the indicated time points (stage 15 to 27). Embryos showing GFP fluorescence in the urogenital system were chosen for further analysis.
Quantification of 18S rRNA Repeat Copy Number. Copy number of the 18S rRNA repeats was quantified in each embryo by qPCR to identify the individual sex except using karyotype analysis in this study. This examination was performed as previously described 48 . The genomic DNA was first extracted from all tested embryos, and subsequently copy number of the 18S gene was normalized against GAPDH using the comparative Ct method of normalization (Ratio (18 S/GAPDH) = 2 Ct GAPDH − Ct 18S ). A higher value of ΔCt (GAPDH -18S) indicates a larger number of copies of 18S rRNA repeats. The sequences of primers are as follows: 18S rRNA (F: 5′-GAGTATGGTTGCAAAGCTGAAA-3′, R: 5′-CGAGAAAGAGCTATCAATCTGT-3′); GAPDH (F: 5′-GGCTTTCCGTGTTCCAACTC-3′, R: 5′-GACAACCTGGTCCTCCGTGTATC-3′). The W chromosome in P. sinensis has a much larger number of copies of rRNA genes compared to the Z chromosome according to FISH mapping results 49 , thus female individuals have more copies of 18S rRNA genes than males ( Supplementary  Fig. 6).
Quantitative RT-PCR. Total  Western Blot Analysis. Total protein extraction from different tissues of adult Chinese soft-shelled turtle was extracted with RIPA lysis buffer (Santa Cruz, USA). Western blot was performed to examine the Dmrt1 protein expression in adult tissues. Equal amounts of denatured protein samples were separated on 10% SDS-PAGE (sodium dodecyl sulfonate-polyacrylamide gel electrophoresis) and transferred onto a PVDF (polyvinylidene difluoride) membrane. The membrane was incubated in 5% dry skim milk at room temperature for 1 h and subsequently with primary antibody (rabbit anti Dmrt1) at dilutions recommended by the supplier (dilution 1:1000, Millipore, USA) overnight at 4 °C. Antibody recognition was detected with the secondary antibody linked to horseradish peroxidase (goat anti-rabbit IgG-HRP, 1:5000, Santa Cruz, USA) at room temperature for 60 min. β-actin bands were used as an internal control. The immunoreactive bands were visualized with DAB (3,3N-Diaminobenzidine Tertrahydrochloride) Horseradish Peroxidase Color Development Kit (Beyotime, china).