Dmrt1 induces the male pathway in a turtle species with temperature-dependent sex determination

Temperature-dependent sex determination (TSD) exists in many reptiles that lack heteromorphic sex chromosomes, in which the incubation temperature of the developing embryos determines gonadal sex (Bull and Vogt, 1979; Charnier, 1966; Ferguson and Joanen, 1982; Pieau et al., 1999). However, the molecular mechanism underlying TSD has been a long-standing mystery. In particular, little is known about (1) which genetic components are responsible for triggering the differentiation of bipotential primordium into either a testis or ovary; and (2) how the physical signal of temperature is transduced into a biological signal. Phylogenetic analyses suggest frequent and repeated evolutionary transitions between TSD and genetic sex determination (GSD) in environmentally sensitive lineages, including reptiles (Quinn et al., 2011; Sarre et al., 2011). In many cases, both GSD and TSD simultaneously exist in the same species (Chen et al., 2014; Radder et al., 2008; Yamamoto et al., 2014). Most recently, a rapid transition from a mixed GSD-TSD system to TSD was experimentally induced in one generation in the dragon lizard Pogona vitticeps, which exhibits ZZ/ZW heterogamety that is susceptible to sex reversal by temperature (Holleley et al., 2015; Quinn et al., 2007).

Some genes (or gene networks) involved in GSD systems have been identified in the gonads of TSD species during the thermosensitive period (TSP), and several of those exhibit temperature-dependent expression patterns prior to gonadal sex differentiation (Shoemaker-Daly et al., 2010; Shoemaker et al., 2007b). The idea that TSD and GSD systems share common genetic components is strongly supported by recent transcriptomes of TSD taxa from the red-eared slider turtle (Czerwinski et al., 2016), the painted turtle (Radhakrishnan et al., 2017) and the alligator (Yatsu et al., 2016). Although RNA interference has been applied on in vitro gonads of TSD turtles to knockdown Sox9 expression (Sifuentes-Romero et al., 2013; Shoemaker-Daly et al., 2010), a functional analysis of candidate sex-determining genes has not yet been performed in vivo in any reptile with TSD, which is largely due to lack of efficient genetic manipulation techniques in TSD species.

Dmrt1 (doublesex and mab3-related transcription factor 1) encodes a transcription factor that contains a DNA-binding motif (DM domain), which is an ancient and conserved component of the vertebrate sex-determining pathway (Matson and Zarkower, 2012). DM-domain genes have been identified as the master sex-determining gene in several non-mammalian GSD species, including dmy in the medaka, DMRT1 in chicken and dmw in frog (Matsuda et al., 2002; Smith et al., 2009; Yoshimoto et al., 2008). Although Dmrt1 is not required for male sex determination in the mouse, it is essential for maintaining the Sertoli cell phenotype in postnatal mammalian testes (Matson et al., 2011; Raymond et al., 2000). In the red-eared slider turtle Trachemys scripta (T. scripta), a TSD species, Dmrt1 exhibits a differential expression pattern in the early developing gonad between MPT (male-producing temperature) and FPT (female-producing temperature) early in the TSP (from stage 15 to stage 20) (Czerwinski et al., 2016; Kettlewell et al., 2000; Murdock and Wibbels, 2003). The sex-determination period of T. scripta ranges from stage 15 to 19 (Matsumoto and Crews, 2012); therefore, Dmrt1 might be involved in determining the fate of the bipotential gonad in this TSD species. Although several investigators have shown that temperature influences Dmrt1 expression, whether Dmrt1 has a critical functional role in the male sex determining pathway has not been investigated in a TSD model system.

We confirmed that Dmrt1 exhibited a temperature-dependent expression pattern in T. scripta early embryonic gonads prior to sexual differentiation, and showed a rapid response to temperature shifts. In addition, the expression of Dmrt1 was induced in FPT embryonic gonads that were masculinized by aromatase inhibitor treatment. We developed an efficient lentiviral vector-mediated gene-modulating approach for both in vivo and in vitro functional analysis of the gene. Most importantly, loss- and gain-of-function analyses provided strong evidence that Dmrt1 is both necessary and sufficient for testicular differentiation in T. scripta. We show that DNA methylation in the Dmrt1 promoter responds rapidly to temperature shifts, and might account for the temperature-dependent dimorphic expression of Dmrt1. Our findings provide the first functional evidence for a genetic trigger for maleness in a TSD system, and suggest that further studies to determine whether methylation patterns are a cause or consequence of changes in Dmrt1 expression could lead to insight into the elusive link between temperature and sex determination.

In a comprehensive transcriptome analysis of MPT and FPT T. scripta gonads at stages 15-21, Dmrt1 was among a small group of transcripts that showed a consistent MPT-specific pattern from stage 15 onwards (Czerwinski et al., 2016). The full-length coding sequence of T. scripta Dmrt1 was obtained by RACE. The complete cDNA sequence of T. scripta Dmrt1 was 2448 base pairs (bp) (accession number KY945220), with a 241 bp 5′ untranslated region (UTR), an 1107 bp 3′ UTR and an open reading frame (ORF) of 1100 bp, which encodes a protein of 369 amino acids (Fig. S1A). The DM domain that is present in mice and chicken Dmrt1 was also highly conserved in T. scripta Dmrt1. The deduced amino acid sequence of T. scripta Dmrt1 shared 80.7%, 75.3% and 53.9% identity with that of chicken, mice and zebrafish, respectively (Fig. S1B). The phylogenetic tree showed that T. scripta Dmrt1 was evolutionarily more closely related to chicken and mice, and distantly related to fish (Fig. S1C).

The mRNA and protein expression of Dmrt1 were first examined in different tissues of the adult turtle. RT-PCR showed that Dmrt1 mRNA was abundantly expressed in testis and was not detected in ovary, heart, liver, spleen, lung, kidney and muscle (Fig. S2A). Dmrt1 protein was localized in the nucleus of Sertoli cells surrounding the spermatogonia in postnatal testis, and was not detected in ovary (Fig. S2B). These studies imply that Dmrt1 is involved in testicular development in T. scripta, as in other species (Raymond et al., 2000; Smith et al., 2009).

qRT-PCR analysis showed that the Dmrt1 transcript was expressed in MPT gonads throughout the period of sex determination, from as early as stage 15 to stage 17, during which time Dmrt1 transcript was not detected in FPT gonads. From stage 19, Dmrt1 expression increased dramatically, reaching a peak at stage 23. By contrast, FPT gonads exhibited extremely low expression of Dmrt1 throughout embryogenesis (Fig. 1A), consistent with other transcriptome results (Czerwinski et al., 2016). A highly MPT-specific Dmrt1 mRNA localization in gonadal medulla at stage 21 was observed via in situ hybridization (Fig. 1B). We also examined the cellular localization of Dmrt1 protein in turtle embryonic gonads. Immunofluorescence revealed that Dmrt1 protein was detected in MPT gonads as early as stage 14. From stage 15 to 17, Dmrt1 protein was strongly expressed throughout the medulla of MPT gonads, whereas it was not detected in FPT gonads (Fig. 1C). Dmrt1 protein was abundantly expressed and mainly localized in the nuclei of precursor Sertoli cells surrounding the primordial germ cells in MPT gonads. In contrast, Dmrt1 protein signal was not detectable in FPT gonads throughout embryogenesis (Fig. 1C). These data reveal that the male-specific expression of Dmrt1 precedes the initiation of gonadal sex differentiation in T. scripta, suggesting an upstream role for Dmrt1 in this TSD system.

To determine the involvement of Dmrt1 in temperature-induced sex determination, we performed in vivo and in vitro experiments to investigate whether Dmrt1 expression was regulated by temperature in T. scripta. In vivo experiments revealed that the mRNA expression of Dmrt1 in turtle gonads at stages 17 and 21 was thermosensitive. The expression level significantly decreased with the increase in temperature from 27°C to 31°C (Fig. 2A, Fig. S3A). We found a similar temperature-dependent pattern of Dmrt1 expression in isolated gonads cultured for 5 and 20 days (Fig. 2B, Fig. S3B). In addition, we examined the response of Dmrt1 expression to temperature shifts from either MPT→FPT or FPT→MPT both in vivo and in vitro. In gonads developing in vivo at MPT and transferred to FPT at stage 16 (MPT→FPT), Dmrt1 expression responded rapidly to the new FPT temperature. Expression decreased significantly below MPT-typical levels by early stage 17 and remained at baseline levels from stage 19 onwards. In gonads shifted from FPT to MPT in vivo, significant upregulation of Dmrt1 expression occurred rapidly by stage 17, reaching the MPT-typical level by stage 18, and stabilizing there for the duration of gonadogenesis (Fig. 2C). In cultured gonads shifted from MPT to FPT, Dmrt1 expression was repressed immediately by day 2, and dropped to baseline levels typical of constant FPT temperature from day 5 onwards. In the opposite shift (FPT→MPT), Dmrt1 expression exhibited a twofold increase in response to temperature shift by day 2, and climbed to peak levels at day 20 (Fig. 2D, Fig. S3C,D). We also compared the response times of Dmrt1, Amh and Sox9 expression to temperature shifts in vitro. The response times for significant (>twofold) downregulation of Dmrt1, Amh and Sox9 expression in MPT→FPT gonads were day 2, day 3 and day 7, respectively. In FPT→MPT shifts, Dmrt1, Amh and Sox9 expression elevated to twofold higher levels by day 2, day 3, and day 8, respectively. These results indicate a more rapid response of Dmrt1 to temperature shifts than Amh or Sox9 (Fig. S3C,D) and that Dmrt1 expression is an upstream responder to temperature shifts in T. scripta.

Previous studies demonstrated that sex determination in the T. scripta embryo is susceptible to application of exogenous estrogen and to the aromatase inhibitor (AI) during the TSP (Crews, 1994; Crews and Bergeron, 1994). Masculinized FPT embryos induced by AI exhibited male-like morphology with a dense medulla and a degenerated cortex (Fig. 3A-F). A significant increase in Amh and Sox9, and downregulation of Cyp19a1 and Foxl2 were detected in masculinized FPT embryos at stage 25 by qRT-PCR (Fig. S4). Immunochemistry showed that aromatase (Cyp19a1) was highly expressed in the medulla of FPT gonads at stage 25; however, its expression decreased dramatically and disappeared in masculinized FPT gonads induced by AI (Fig. 3G-H). Sox9 and Dmrt1 proteins were robustly expressed in the medulla of masculinized FPT gonads, similar to levels in MPT gonads (Fig. 3J-O). The response timecourse analysis revealed that Dmrt1 expression responded rapidly to AI treatment and was significantly upregulated from stage 17 onwards (Fig. 3P). These observations indicate that Dmrt1 is an early responder to the induction of male differentiation in T. scripta.

To solve the problem of the lack of available genetic manipulation techniques in T. scripta, we established an efficient gene-modulating method for functional analysis. We used lentiviral vectors carrying Dmrt1-specific shRNAs with a GFP reporter gene to knockdown endogenous Dmrt1 transcripts in T. scripta embryos (Fig. S5A). To test the efficacy of lentivirus delivery, mortality rate and GFP expression were examined in turtle embryos injected with virus at stage 14. Approximately 38% (305/800) of treated embryos survived after stage 21, and 55% (167/305) of these survival embryos showed global GFP reporter expression, including widespread expression in gonad tissue (Fig. S5B-F). Robust GFP immunofluorescence was detected in gonadal sections, indicating effective lentivirus infection (Fig. S5G-I). We next examined Dmrt1 expression in MPT gonads carrying LV-Dmrt1-shRNAs to determine whether lentivirus-mediated RNA interference efficiently knocked down target gene expression. qRT-PCR showed that Dmrt1-shRNA#1 exerted the most powerful repression at stages 19 and 21 (Fig. S6A). Dmrt1 mRNA in MPT gonads with LV-Dmrt1-shRNA#1 was significantly downregulated from early stage 16 onwards, compared with Dmrt1 mRNA in control MPT gonads (LV-NC-shRNA). The average decrease in Dmrt1 expression throughout gonadogenesis induced by lentivirus-mediated RNA interference was 73.7% (Fig. S6B). Immunochemistry further demonstrated that in MPT gonads with LV-Dmrt1-shRNA#1, no Dmrt1 protein expression was detected in the obviously feminized cortical region, and Dmrt1 protein signal was weak or almost lost in the medulla (Fig. S6C).

Overexpression of Dmrt1 in turtle FPT embryos was also performed using a lentivirus-mediated expression system; however, no embryos survived to stage 19, presumably owing to global effects. As an alternative approach to investigate whether overexpression of Dmrt1 was sufficient to drive male development, we used in vitro electroporation of cultured FPT gonads with lentiviral vector carrying the Dmrt1 ORF (LV-Dmrt1-OE). Fifty-four percent (188/350) of cultured gonads removed from stage 16 embryos were viable to day 20 and 65% (123/188) of these survival gonads were GFP positive. qRT-PCR showed that ectopic expression of Dmrt1 in FPT cultured gonads carrying LV-Dmrt1-OE was dramatically increased by day 5 of culture, and remained more than fivefold higher than FPT-typical levels during 30 days of gonad culture in vitro (Fig. S7A). Ectopic activation of Dmrt1 protein was found in the medulla of FPT-cultured gonads infected with LV-Dmrt1-OE (Fig. S7B). These results indicate that we successfully developed an effective gene up- and downmodulating method in T. scripta that works both in vitro and in vivo.

To verify that Dmrt1 plays a role in sex determination or gonadal differentiation in T. scripta, we infected embryos at stage 14 with a scrambled virus shRNA (LV-NC-shRNA) or a silencing virus (LV-Dmrt1-shRNA) and compared phenotype and marker gene expression by gonadal histology, immunofluorescence and qRT-PCR. Control MPT embryos treated with the non-silencing scrambled virus exhibited the typical male phenotype, consisting of round, cylindrical testes with degenerated oviducts, whereas scrambled control FPT embryos displayed typical long and flat ovaries along with oviducts located in the mesonephric tissue (Fig. 4A,D). In MPT embryos infected with LV-Dmrt1-shRNA and showing high GFP expression, gonads became elongated and nonvascularized, which was accompanied by the differentiation of oviducts in the mesonephros, exhibiting varying degrees of female-like morphology (Fig. 4B,C, Fig. S8A). Hematoxylin and Eosin stains of gonadal sections showed that control MPT gonads had a dense medulla with seminiferous cords and a reduced cortex (Fig. 4E,E′), whereas the gonads of control FPT embryos showed a well-developed outer cortex, populated with primordial germ cells, and a degenerated medulla (Fig. 4H,H′). In contrast, most of the MPT gonads with Dmrt1 knockdown were strongly feminized, characterized by a thickened outer cortex with a number of primordial germ cells and a vacuolated medulla (Fig. 4G,G′, Fig. S8B). Interestingly, some Dmrt1-knockdown MPT gonads exhibited partial redirection of sexual trajectory (ovotestis), showing an ovarian-like developed cortex and a dense medulla with testis cords simultaneously present (Fig. 4F,F′, Fig. S8B). Overall, 54 (44 ovaries and 10 ovotestes) out of 63 knockdown MPT embryos (85.71%) exhibited varying degrees of male to female shift in sexual trajectory (Table 1). We also performed the Dmrt1 knockdown experiments in turtle eggs incubated at the threshold or pivotal temperature (PvT) that produces an even ratio of males and females. At PvT, 95% of embryos treated with LV-Dmrt1-shRNA showed phenotypic evidence of female development (38/40, 31 ovaries and seven ovotestes) (Table S1).

To confirm the activation of the female pathway in MPT embryos with Dmrt1 knockdown, we analyzed the expression of sex-specific marker genes and germ cell distribution patterns at stages 19, 21, 23, 25 and 26. At the mRNA level, strong downregulation of testicular differentiation markers Amh and Sox9, and significant upregulation of ovarian development regulators Cyp19a1 and Foxl2, were observed in Dmrt1-knockdown MPT gonads relative to controls at different developmental stages (Fig. 4I, Fig. S9A). At the protein level, Sox9 protein was expressed specifically in the nuclei of precursor Sertoli cells in control MPT gonads, whereas control FPT gonads lacked Sox9 expression. In most MPT gonads following knockdown of Dmrt1, Sox9 protein expression was sharply reduced (Fig. 4J,J′,M,M′, Fig. S9B). However, some Sox9 protein was still present in the medulla of partially sex-reversed MPT gonads (Fig. 4K,K′). Following Dmrt1 knockdown, MPT embryos simultaneously exhibited female-like cortical expression of γH2ax, a protein that is expressed in female germ cells entering meiotic prophase and is not present in male germ cells throughout embryogenesis (Fig. 4J,J′,M,M′). Aromatase, the key enzyme involved in estrogen synthesis, was strongly expressed in the medullary region of control FPT gonads, and was never detected in control MPT gonads. However, Dmrt1-knockdown MPT embryos showed ectopic activation of aromatase in the gonadal medulla (Fig. 4N-Q′). In some MPT embryos with partial feminization, a small amount of aromatase was ectopically expressed in medullary cells outside the remaining testis cord structures (Fig. 4O,O′). Vasa staining showed medullary cord distribution of germ cells in control MPT gonads, whereas control FPT gonads displayed cortical localization of germ cells. In MPT embryos with Dmrt1 knockdown, Vasa-positive germ cells exhibited a female-like distribution pattern, mainly enriched in the developed outer cortex with few germ cells localized in the medulla (Fig. 4R-U′, Fig. S9C). These results provide solid evidence that Dmrt1 is required for testis determination in T. scripta.

To determine whether Dmrt1 was sufficient to initiate the male pathway in this TSD system, we investigated whether ectopic activation of Dmrt1 in FPT individuals would result in a redirection of sexual trajectory. Unfortunately, no embryos infected in ovo with the lentiviral vector carrying the Dmrt1 gene survived past the period of gonadal differentiation. Therefore, we used a whole-organ in vitro culture system to analyze the effects of overexpression of Dmrt1 on FPT gonadal development. Gonads from stage 16 embryos incubating in ovo at FPT or PvT were dissected, electroporated with the lentiviral vector carrying the Dmrt1 ORF (LV-Dmrt1) and cultured either at FPT or PvT (in accordance with previous in ovo culture temperature). At day 30 of in vitro culture, the control gonads cultured at FPT with empty lentiviral vector exhibited an oval shape, a thickened cortical region populated with germ cells and a reduced medulla with no evidence of sex cords. FPT-cultured gonads overexpressing Dmrt1 showed an obvious female-to-male redirection of gonadal morphology, characterized by well-organized testis cord structure in the medulla and a degenerated cortex, similar to normal males (Fig. 5A-D′). Of gonads cultured at FPT, 42 out of 51 (82%) infected with LV-Dmrt1-OE displayed complete or partial female-to-male redirection of sexual trajectory (Table 2). Under the condition of PvT, 51 out of 54 (94%) gonads overexpressing Dmrt1 developed into testes or ovotestes (Table S2).

qRT-PCR for male and female marker genes showed that Amh and Sox9 expression increased, and Cyp19a1 and Foxl2 expression decreased relative to controls in cultured FPT gonads overexpressing Dmrt1 at days 10, 15, 20, 25 and 30 (Fig. 5E). Maintenance of Sox9 protein in sex cords of Dmrt1-overexpressing FPT-cultured gonads was confirmed by immunofluorescence (Fig. 5F-I′). γH2ax protein expression (a marker of female meiosis) was totally lost in completely sex-reversed FPT gonads. Although overexpression of Dmrt1 in FPT gonads resulted in the relocalization of many germ cells from the cortical region to sex cords within the medulla (Fig. 5J-M), some γH2ax-positive germ cells were still present in the remaining cortex in partially sex reversed FPT gonads (Fig. 5F-I′). These findings suggest that Dmrt1 is sufficient to initiate male development in the TSD system.

DNA methylation in gene promoters is a conserved epigenetic modification that influences transcript expression. Recently, differences in methylation patterns at MPT and FPT have been shown to correlate with the sex-specific expression of genes in organisms with TSD (Piferrer, 2013). To investigate whether temperature is correlated with the methylation pattern of the Dmrt1 promoter in T. scripta, we analyzed the DNA methylation signature in the promoter region of Dmrt1 in T. scripta gonads incubated at MPT versus FPT. Bisulfite sequencing revealed that the overall percentage of CpG methylation in the Dmrt1 promoter was significantly lower in MPT gonads relative to FPT gonads from stage 15 onwards throughout the TSP (Fig. 6A). To test whether the DNA methylation signature shifts in accordance with temperature shifts during the TSP, eggs were shifted either MPT→FPT or FPT→MPT at stage 16 and the methylation status of Dmrt1 promoter was examined at stages 17, 18, 19, 20 and 21. Results showed that the methylation level of the Dmrt1 promoter in MPT→FPT gonads was dramatically increased by stage 17 relative to that of MPT gonads. By contrast, FPT→MPT gonads exhibited a rapid decrease in DNA methylation of Dmrt1 promoter region by stage 17 in response to temperature shifts (Fig. 6B). These data revealed that the Dmrt1 transcript expression fluctuation was negatively correlated with the methylation dynamics of Dmrt1 promoter region during the temperature shifts (Fig. 2B), raising the possibility that DNA methylation directly responds to temperature and is responsible for the temperature regulation of Dmrt1.

Even though TSD has been studied for several decades, the molecular mechanism underlying this mode of sex determination has remained elusive. Here, we demonstrate that Dmrt1 can act as a male sex-determining gene in the TSD species T. scripta. We show that Dmrt1 has a temperature-dependent dimorphic expression pattern preceding the initiation of gonadal sex differentiation. Using a novel in vivo viral transduction system, we provide solid functional evidence that Dmrt1 is both necessary and sufficient to trigger testicular differentiation. This is the first such functional study to identify and test a genetic factor that can determine the fate of the bipotential gonad in a TSD species.

A number of genes involved in the GSD system have been previously identified in the gonads of some TSD species (Rhen et al., 2007; Shoemaker-Daly et al., 2010; Shoemaker et al., 2007a; Czerwinski et al., 2016). In the red-eared slider turtle, Dmrt1 is a gene that exhibits an early sexually dimorphic expression pattern in the developing gonad between MPT and FPT (Kettlewell et al., 2000; Shoemaker et al., 2007a). In this study, MPT-specific expression of Dmrt1 transcript was detected as early as stage 15, just at the onset of sex determination and the TSP. Sexually dimorphic expression of Dmrt1 preceded Amh and Sox9, two well-known factors for male sexual development, and was maintained throughout the gonadal differentiation period, which is very similar to the expression pattern of Dmrt1 in chicken (Lambeth et al., 2014; Smith et al., 2009). Interestingly, we found that Dmrt1 also exhibited early male-specific embryonic expression before the onset of gonadal sex differentiation in the Chinese soft-shelled turtle, a GSD species with ZZ/ZW sex chromosomes (W. Sun, H. Cai, G. Zhang, H. Zhang, H. Bao, L. Wang, J. Ye, G. Qian and C. Ge, unpublished). We show that the Dmrt1 protein is present in nuclei of pre-Sertoli cells surrounding primordial germ cells in the MPT gonad as early as stage 14. Pre-Sertoli cells are the first cell type to differentiate in the developing testis and send organizing signals to other nascent cell types, thereby directing testis formation. These features of Dmrt1 expression suggest that it is important for both primary sex determination and subsequent gonadal differentiation in T. scripta. To determine whether Dmrt1 is a master TSD gene, we examined the effect of temperature on Dmrt1 expression both in vivo and in vitro. Dmrt1 transcript expression in individual embryos was highly temperature dependent between 25 and 33°C, during the sex-determination period and TSP. The temperature-shift assays showed a rapid response of Dmrt1 to temperature shifts from MPT to FPT or FPT to MPT both in vivo and in vitro, preceding changes in Amh and Sox9, which is consistent with previous reports (Shoemaker-Daly et al., 2010; Shoemaker et al., 2007b). These findings confirm the thermosensitivity of Dmrt1 in T. scripta gonadal cells.

In T. scripta, exogenous estrogen and its synthetase aromatase can override the temperature effect if applied during the TSP (Matsumoto and Crews, 2012; Ramsey and Crews, 2009). In chicken, ZW eggs treated with aromatase inhibitor (AI) exhibit upregulation of Dmrt1, leading to masculinization of ZW embryos (Smith et al., 2003). It has been proposed that exogenous steroid hormones may redirect the gonadal trajectory by interacting with the candidate sex-determining genes (Matsumoto and Crews, 2012). In this study, turtle embryos incubating at FPT exhibited a rapid increase in Dmrt1 transcript and protein expression in response to AI treatment. This response of Dmrt1 expression occurred at early stages of TSP, clearly preceding the first signs of morphological differentiation of testes or ovaries in T. scripta. The change of Dmrt1 transcript in response to AI treatment was earlier than that of Sox9, which occurred at stage 19 (Matsumoto et al., 2013b). By contrast, treatment of MPT eggs with estrogen caused a rapid downregulation of Dmrt1 transcript expression in T. scripta (Murdock and Wibbels, 2006), which preceded downregulation of Sox9 (Barske and Capel, 2010). It is possible that the reversed expression pattern of Dmrt1 induced by exogenous ligands is responsible for the ultimate sex-reversal in T. scripta. At the least, these results suggest that Dmrt1 acts upstream in the testis pathway.

Genetic research in TSD species has been hindered by lack of available genetic manipulation techniques. A number of studies have focused on establishing effective loss- and gain-of-function methods in turtle species. Sifuentes-Romero et al. reported that transfection of in vitro cultured MPT gonads of the olive ridley sea turtle with siRNAs specific to Sox9 caused a significant reduction of Sox9 mRNA and protein expression (Sifuentes-Romero et al., 2013). On the other hand, in vitro cultured gonads of T. scripta electroporated with a fusion GFP:Sox9 plasmid exhibited a certain amount of ectopic expression of Sox9 at FPT (Shoemaker-Daly et al., 2010). However, no in vivo genetic modification has been achieved in turtle embryos to date. In this study, we have developed both in vivo and in vitro methods to efficiently knockdown and overexpress Dmrt1 transcript in the gonads of T. scripta, achieved by injection of embryos in ovo with lentiviral vectors carrying Dmrt1-specific shRNAs and in vitro electroporation of cultured gonads with lentiviral vector carrying Dmrt1 ORF. To our knowledge, this is the first time that an in ovo gene-modulating approach has been established in turtle embryos, which opens the door to elucidate the function of the molecular cascade underlying TSD. Using this approach, we have clearly demonstrated that Dmrt1 knockdown caused complete feminization of MPT embryos. Similarly, knockdown of Dmrt1 in chicken embryos leads to feminization of genetic male individuals, with a decline in Sox9 expression and an increase in Cyp19a1 and Foxl2 expression (Smith et al., 2009). In medaka, XY Dmy knockout fish exhibit a fertile male-to-female sex reversal (Matsuda et al., 2002). Most importantly, our gain-of-function experiment revealed that ectopic expression of Dmrt1 in cultured FPT turtle gonads resulted in a female-to-male redirection of sexual trajectory, evidenced by formation of sex cord-like structures. Likewise, overexpression of Dmy in XX medaka results in testicular differentiation (Matsuda et al., 2007), and ZW chicken gonads overexpressing Dmrt1 display an evident masculinized morphology, as well as activation of Amh and Sox9 (Lambeth et al., 2014). These findings demonstrate that, similar to its role in chicken and medaka, Dmrt1 is both necessary and sufficient for testicular differentiation in T. scripta.

Both Amh and Sox9 were upregulated following overexpression of Dmrt1 in FPT-cultured gonads, raising the possibility that these two genes respond to elevated Dmrt1. The order of their expression at MPT also implies that Dmrt1 lies upstream of Amh and Sox9 in T. scripta, similar to chicken. Recently, the Dmrt1 ortholog Doublesex (Dsx) has been identified as a key regulator of the male phenotype in the branchiopod crustacean Daphina magna, which exhibits environmental sex determination (ESD) (Kato et al., 2011). In beetles, Dsx is involved in nutrition-dependent male sexual trait development (Gotoh et al., 2014; Kijimoto et al., 2012). All these observations suggest that, in spite of the diversity and plasticity of sex-determination mechanisms, DM domain genes are highly conserved in male sexual differentiation pathways between GSD, TSD and the broader category of ESD systems.

We have demonstrated that Dmrt1 is a strong candidate master TSD gene in T. scripta, as evidenced by temperature sensitivity of Dmrt1 during the sex-determination period, and that it is necessary and sufficient for testis differentiation. However, the mechanism by which temperature regulates Dmrt1 remains unclear. Dmrt1 might not directly respond to temperature because, in some TSD species, lower temperatures produce males, whereas in others, higher or intermediate temperatures promote males. It is possible that there is a temperature sensor or chromatin regulator present during the TSD initiation period that is responsible for regulating temperature-induced activation or repression of sex-determining genes. A recent study in the American alligator revealed that the male gonad-typical TRPV4 channel may influence the male gonadal sex determination pathway by modifying Amh and Sox9 expression during TSP (Yatsu et al., 2015). However, TRPV4 alone was not sufficient to initiate gonadal sex determination, because activation of the TRPV4 channel in FPT gonads did not cause redirection of gonadal sexual trajectory. Schroeder et al. reported that a SNP in the cold-inducible RNA-binding protein (CIRBP) loci was highly associated with gonadal sex phenotype in the snapping turtle with TSD, indicating CIRBP is involved in gonadal differentiation (Schroeder et al., 2016). Nevertheless, additional factors are likely involved in sensing or transducing the thermal influence because there was not perfect concordance between CIRBP genotype and sex. Considering the diversity of TSD patterns, it is probable that the transduction of temperature into a molecular signal (gene) for gonadal differentiation is polygenic and complex, as opposed to being dependent on a single common factor.

Recently, epigenetic mechanisms have been demonstrated to regulate genes involved in sex determination and gonadal differentiation (Kuroki et al., 2013; Piferrer, 2013; Shao et al., 2014; Tachibana, 2015; Zhang et al., 2013). Epigenetic mechanisms, including DNA methylation, histone modification and non-coding RNAs, can integrate environmental information with the regulation of gene expression, and have been emerging as a promising regulatory mechanism for TSD (Matsumoto et al., 2013a; Navarro-Martín et al., 2011; Parrott et al., 2014; Piferrer, 2013; Venegas et al., 2016). The first example of an epigenetic mechanism mediating temperature effects on sexual development in a vertebrate came from a study on the European sea bass, a fish with a polygenic system of sex determination where temperature and genetics contribute equally to sexual fate. In this example, DNA methylation of the gonadal aromatase (Cyp19a1) promoter was associated with temperature-induced sex ratio shifts (Navarro-Martín et al., 2011). In T. scripta, the DNA methylation level of the gonadal Cyp19a1 promoter was higher at MPT than FPT by stage 19 onwards, and was reduced in response to temperature shifts from MPT to FPT (Matsumoto et al., 2013a). Similarly, differential incubation temperatures resulted in dimorphic DNA methylation patterning of the Cyp19a1 and Sox9 promoters in gonads of American alligator embryos (Parrott et al., 2014). We show that the DNA methylation status of the Dmrt1 promoter in gonads of T. scripta displayed significant temperature-dependent dimorphism from stage 15 onwards, and responded rapidly to temperature shifts both from MPT to FPT and FPT to MPT. This temperature-induced dimorphic DNA methylation patterning of the Dmrt1 promoter occurred at the beginning of the TSP, prior to the initial time of differential DNA methylation in the Cyp19a1 promoter in the middle of TSP. These findings raise the possibility that DNA methylation functions as a key mediator that integrates temperature into a molecular trigger that determines sex in the TSD system. However, the DNA methylation status of the Dmrt1 and Cyp19a1 promoters at both constant and shifted temperatures were perfectly correlated with transcript expression patterns; thus, it is unclear whether changes in methylation are a cause or consequence of gene expression changes. If temperature can modulate the DNA methylation status of a putative sex-determining gene, elucidation of the mechanism will be a key next step.

Recent transcriptomes of TSD taxa show a series of genes exhibiting earlier differential expression between MPT and FPT, before the onset of TSP or even the formation of a gonad, such as Sf1 (Valenzuela et al., 2006; Valenzuela, 2008), Igf1r, Insr (Radhakrishnan et al., 2017), Fdxr, Pcsk6, Kdm6b, Twist1, Hsp6b and TRP channel genes (Czerwinski et al., 2016). These genes might be involved in regulating temperature-induced activation or repression of putative sex-determining genes.

To summarize, we have characterized a conserved DM domain gene, Dmrt1, in the red-eared slider turtle, and have found that Dmrt1 exhibits a temperature-dependent sexually dimorphic expression pattern from the earliest stages of the sex-determination period, and is capable of responding rapidly to temperature shifts and aromatase inhibitor treatment. Most importantly, in ovo and in vitro loss- and gain-of-function analyses provide solid evidence that Dmrt1 is both necessary and sufficient to initiate the male development in T. scripta. Furthermore, DNA methylation of the Dmrt1 promoter might act as a crucial mediator in the regulation of Dmrt1 by temperature. These findings demonstrate that Dmrt1 is a candidate master TSD gene in T. scripta, consistent with the strong conservation of this DM domain gene during the evolution of sex-determination mechanisms (Herpin and Schartl, 2015; Matson and Zarkower, 2012). In short, we have functionally identified the first genetic trigger for maleness in a TSD system, thereby shedding new light on the elusive TSD molecular mechanism.

Freshly laid red-eared slider turtle (T. scripta) eggs were obtained from the Hanshou Institute of Turtles (Hunan, China). Fertilized eggs were randomized in trays of moist vermiculite and placed in incubators at 26°C (MPT) or 32°C (FPT), with humidity maintained at 70-80%. In this species, incubation of eggs at 26°C (MPT) produces all males, whereas incubation at 32°C (FPT) generates all females. For in vivo temperature-shift experiments, 200 eggs were shifted at developmental stage 16 from an incubator kept at 26°C to an incubator kept at 32°C, and vice versa. For in vivo experiments regarding temperature-dependent expression of Dmrt1, groups of 200 eggs were placed in humidified incubators held at 25°C, 26°C, 27°C, 28°C, 29°C, 30°C, 31°C, 32°C and 33°C. The temperature-sensitive period (TSP) in T. scripta extends from approximately stage 15 to stage 20, when the embryo is environmentally sensitive and when sex determination occurs (Wibbels et al., 1991). Embryos were staged according to criteria established by Greenbaum (2002). At stages 14, 15, 16, 17, 19, 21, 23 and 25, embryos incubated at different temperatures were removed from eggshells, decapitated and placed in PBS for dissection. Gonads of stage 16 embryos incubated at different temperatures were dissected for whole-gonad in vitro organ culture at the constant or shifted temperatures. Experiments were carried out according to a protocol approved by the Zhejiang Wanli University.

Total RNA isolated from MPT embryonic gonads at stage 25 was treated with DNase I and then reverse transcribed using SuperScript III reverse transcriptase and oligo(dT). Based on the published partial sequence of Dmrt1 (Accession Number AY316537.1), a pair of PCR primers were designed to amplify a fragment of turtle Dmrt1 cDNA. 5′ RACE and 3′ RACE were performed according to the manufacturers' protocols for the SMART RACE cDNA Amplification kit (Clontech) and a CapFishing Full-length cDNA Premix kit (Seegene). The sequences of primers for amplification are listed in Table S3.

Gonads from embryos in each group were microdissected from the mesonephros, and individual pairs of gonads were harvested for RNA extraction using TRIzol (Invitrogen) or RNeasy Plus Micro kit (Qiagen). The cDNA was generated from 0.5-2 μg RNA using the SuperScript First-Strand Synthesis System (Fermantas) based upon the manufacturer's protocol, followed by DNase treatment. Real-time PCR was carried out in triplicate with a SYBR Green Supermix (Bio-Rad) in a Bio-Rad iCycler system. After normalization with Gapdh, relative RNA levels in samples were calculated by the comparative threshold cycle (Ct) method (Schmittgen and Livak, 2008). The sequences of primers for PCR are listed in Table S3.

Gonad-mesonephros complexes were dissected from turtle embryos at stages 23 and 25, immediately frozen in OCT embedding medium, stored at −80°C, and subsequently sectioned at 20 μm and thaw-mounted onto SuperFrost Plus slides (Erie Scientific). Sections were fixed in ice-cold 4% paraformaldehyde (PFA)/PBS, and incubated in 0.25% acetic anhydride /triethanolamine. After washes in 2× standard saline citrate (SSC), slides were dehydrated through a series of ethanol solutions, air dried, and stored at −80°C. In situ hybridization for Dmrt1 mRNA expression was performed as described previously (Shoemaker et al., 2007a). Briefly, sections were rehydrated, pre-hybridized for 2 h at 65°C and then hybridized overnight under the same conditions in the presence of a digoxygenin-labeled Dmrt1-specific antisense riboprobe, generated from MPT gonadal cDNA. After RNase A treatment at 37°C, sections were washed in a decreasing series of SSC and equilibrated in Tris buffer at room temperature before incubation in 1:5000 anti-DIG-alkaline phosphatase Fab fragments (Roche) for 2 h. Sections were washed in 100 mM Tris and incubated in 5 mM levamisole. Chromogenic product was formed using BCIP/NBT in 100 mM Tris (Sangon, Shanghai) at 30°C until the desired darkness was achieved, then terminated simultaneously for all slides. Sense controls did not show any color reaction. Sections were dehydrated, delipidated, mounted in AquaMount and photographed.

Gonad-mesonephros complexes or cultured gonads were fixed in 4% PFA overnight at 4°C, then embedded in paraffin wax and sectioned. Paraffin sections (5-8 μm) were deparaffinized prior to immersion in 10 mM sodium citrate buffer for 15 min for antigen retrieval at a sub-boiling temperature (99°C). Sections were covered with primary antibodies and incubated overnight at 4°C. The primary antibodies used in this analysis included rabbit anti-Dmrt1 (produced privately by Sangon Biotech, 1:250), rabbit anti-Sox9 (Chemicon, AB5535, 1:1000), mouse anti-β-catenin (Sigma, C7207, 1:250), rabbit anti-aromatase (Abcam, ab18995, 1:150), rabbit anti-Vasa (Abcam, ab13840, 1:50) and mouse anti-γH2AX (Abcam, ab26350, 1:250). Primary antibodies were detected using secondary antibodies Alexa Fluor 488 donkey anti-rabbit IgG (Invitrogen, A21206), Alexa Fluor 488 donkey anti-mouse IgG (Invitrogen, A21202), Alexa Fluor 594 donkey anti-rabbit IgG (Invitrogen, A21207) and Alexa Fluor 594 donkey anti-mouse IgG (Invitrogen, A21203), all diluted at 1:250. Nuclei were stained with DAPI. Gonad sections were imaged using a fluorescence microscope (Ti-E, Nikon) or confocal microscope (A1 Plus, Nikon).

Turtle gonad culture in vitro was carried out mainly according to the methods described previously (Mork and Capel, 2013; Shoemaker-Daly et al., 2010). Briefly, gonad-mesonephros complexes were immediately removed from stage 16 embryos incubating in vivo at different temperatures. Gonads were carefully dissected from the adjacent mesonephros and placed on 0.4 μm transparent, low-protein-binding Biopore membrane (Millipore) floating on 2 ml of Leibovitz's L-15 medium (Gibco) supplemented with 10% charcoal-stripped fetal bovine serum (FBS) and 0.2% penicillin-streptomycin solution. Isolated gonads were cultured in sterile culture plate wells (Corning) placed in cell incubators, in which the same temperatures were maintained as before dissection, monitored daily with HOBO data loggers and verified with calibrated thermometers. Culture medium was refreshed by replacing 750 μl every day for the duration of culture. Gonads were grown at either constant temperatures or shifted temperatures (26°C→32°C and 32°C→26°C). Using this approach, we cultured gonads explanted at stage 16 for up to 30 days. On day 5 of in vitro culture, the gonad was found to have reached late stage 18. Gonads cultured for 10, 15, 20, 25 and 30 days at FPT were found to have reached stages early 20, late 21, late 22, 23 and late 23, respectively. Gonads cultured at MPT for 5, 10, 15, 20, 25 and 30 days corresponded to stages late 17, late 18, late 19, late 20, 21 and late 21, respectively. Following 1, 2, 3, 4, 5, 10, 15, 20, 25 and 30 days of culture, gonads at MPT (26°C) and FPT (32°C) were collected for histology, immunohistochemistry and gene expression analysis. For in vitro experiments regarding the relationship between Dmrt1 expression and temperature, isolated gonads were cultured in cell incubators at different temperatures (23°C, 24°C, 25°C, 26°C, 27°C, 28°C, 29°C, 30°C, 31°C, 32°C, 33°C), and harvested on day 5 or 20 for RNA extraction.

A non-steroidal aromatase inhibitor letrozole (PHR1540, Sigma) was administered to eggs incubating at FPT (32°C). Letrozole was dissolved in 95% ethanol at a concentration of 10 μg/μl, and 10 μl of the drug was applied topically to the eggshell in the region adjacent to the embryo at stage 15. Controls were treated with 10 μl of 95% ethanol. Gonad-mesonephros complexes were dissected from treated and control embryos at stage 25 for histology and immunohistochemistry. Gonads were separated from the adjacent mesonephros at stages 17, 19, 21, 23 and 25 and hatching time, and preserved for qRT-PCR analysis.

Three shRNAs targeting turtle Dmrt1 mRNA were designed to give rise to siRNA, using the shRNA designer website (http://rnaidesigner.thermofisher.com/rnaiexpress/design.do). The lentivirus vector was used to deliver shRNAs directed specifically against turtle Dmrt1 mRNA. The designed shRNA construct contained a unique 21 nt double-stranded Dmrt1 sequence that presented as an inverted complementary repeat, a loop sequence (5′-CTCGAG-3′) and the RNA Pol-II terminator (5′-TTTTTT-3′). Annealed oligonucleotides were ligated into pGP-U6 (GenePharma) between the BbsI and XhoI sites by T4 DNA ligase (TaKaRa) to produce pGP-U6-Dmrt1-shRNA. The pGP-U6-Dmrt1-shRNA construct was digested with AgeI-EcoRI and inserted into the EcoRI site of pGLV-U6-GFP (GenePharma). The recombinant vector pGLV-GFP-Dmrt1-shRNA was termed LV-Dmrt1-shRNA. The negative control vector (pGLV-GFP-NC-shRNA, termed LV-NC-shRNA) contained a nonsense shRNA insert in order to control any effects caused by non-RNAi mechanisms. The sequences of the shRNA are as follows: Dmrt1-shRNA#1,5′-GGTGGCAGCTCCTGTTTATTG-3′; Dmrt1-shRNA#2, 5′-GGATGCTCATTCAGGACATTC-3′; Dmrt1-shRNA#3, 5′-GCAGTCAAGATTCTGGCTTAA-3′; negative control, 5′-TTCTCCGAACGTGTCACGTAT-3′.

For the generation of lentivirus, 293T 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 construct by lipofectamine. Twenty-four hours post-transfection, the transfection mix was replaced with a fresh culture medium (without antibiotics). The virus-containing supernatant was harvested 72 h post-transfection, cleared by centrifugation (1000 g for 15 min at 4°C) and then filtered through a 0.45 μm filter (Millipore). Viruses were titrated by adding serial dilutions to fresh 293T and assessing GFP expression after 48 h. Viral titers of ∼4×10⁸ infectious units/ml were obtained. Lentivirus aliquots were stored at −80°C before infection of turtle embryos.

Total RNA was isolated from MPT embryonic gonads at stage 25, whereupon reverse transcription was carried out to prepare cDNA. A full-length turtle Dmrt1 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. The empty vector pGLV-GFP-empty was used as a negative control (LV-empty). High-quality proviral DNA was used to transfect 293T cells. Virus propagation was carried out as described above. A viral titer of 4×10⁸ infectious units/ml was obtained for in vitro electroporation of cultured gonads.

A high-titer virus of LV-Dmrt1-shRNA (at least 1×10⁸ infectious units/ml) was injected into stage 13.5-14 turtle embryos at MPT (26°C) or PvT (29.2°C, a threshold temperature that produces an even ratio of males and females). Eggs were swabbed with alcohol swabs prior to injection using a fine (0-25 µl) metal Hamilton needle. Approximately 5 μl was injected per embryo and a total of 500 eggs were injected in each treated group. Three-hundred control embryos at MPT, PvT and FPT were injected with scrambled control virus of LV-NC-shRNA. Eggs were sealed with parafilm and incubated for the indicated time points (stage 23 and 25). Ratio of survival to stage 25 was 40-60%. Embryos showing GFP fluorescence in the urogenital system were chosen for further analysis. More than 30 pairs of GFP⁺ gonads were sampled per treatment.

To assess the effects of Dmrt1 overexpression in FPT embryonic gonads, we used in vitro electroporation of FPT-cultured gonads with the lentiviral vector carrying the Dmrt1 ORF (LV-Dmrt1). High-quality LV-Dmrt1 DNA was prepared and used at final concentration of 1 μg/μl diluted in electroporation mix (0.16% carboxymethyl cellulose, 1 mM MgCl₂ in PBS). In vitro electroporation was performed according to methods described previously (Lambeth et al., 2014; Shoemaker-Daly et al., 2010). Briefly, gonads from stage 16 embryos that were incubated in vivo at FPT or PvT were placed on solidified Sylgard in a sterile petri dish in 10 μl of 1 μg/μl LV-Dmrt1 DNA and electroporated under the following condition: 40 V, three pulses of 50 ms each, with a 100 ms pulse interval. Control gonads at stage 16 at FPT, MPT or PvT were electroporated with LV-empty DNA. After electroporation, each gonad was immediately transferred to fresh culture medium for 5-10 min. Electroporated gonads were then transferred to a 0.4 μm Millicell membrane (Millipore), and cultured as above for up to 30 days. More than 30 cultured gonads showing GFP fluorescence were selected for further analysis.

Genomic DNA from MPT, FPT, MPT→FPT or FPT→MPT individual pairs of gonads at stages 15, 16, 17, 18, 19, 20 and 21 was isolated using the QIAamp Fast DNA Tissue kit (Qiagen). Six pairs of gonads were sampled at each stage/temperature for bisulfite sequencing. The isolated DNA was treated with Proteinase K and bisulfite and purified from a Methylamp DNA modification kit (Epigentek) according to the manufacturer's protocol. PCR for Dmrt1 promoter CpG sites was carried out on bisulphite-treated DNA using primers specific to the converted DNA around the TSS (designed based on the published promoter sequence of T. scripta Dmrt1, Accession Number KJ583239, the forward primer 5′-TTTTTAGTTTTGGAGTTAAGGTAGTA-3′ and reverse primer 5′-AAAAAATAACACTAACCACACCAAC-3′). The purified gDNA was amplified by nested PCR under the following PCR conditions: 94°C for 5 min, followed by 35 cycles at 94°C for 30 s, 57°C for 30 s, 68°C for 30 s and final extension at 72°C for an additional 10 min period. The PCR products were gel-purified and cloned into the pGEM-T vector (Promega). The resulting sequences were examined for conversion efficiency and accuracy, using BiQ analyzer software. Sequences with less than a 97% conversion rate were not analyzed. Ten clones from each pair of gonads were sequenced.

Each experiment was independently repeated at least three times. All data are expressed as the mean±s.d. Student's unpaired t-test was used to test significance (*, ^#P<0.05; **, ^##P<0.01; ***, ^###P<0.001; n.s., no significance).

We thank Ceri Weber (Duke University) for her comments on the manuscript, and Mr Dongdong Pan, Han Cai and Haisheng Bao (Zhejiang Wanli University) for assistance in egg incubation.