Laboratory reared medaka (Oryzias latipes, Class Actinopterygii, order Beloniformes, family Adrianichthyidae) were used. For detailed description of this model species and its features see (Kinoshita et al., 2009). All experiments were performed with fish of the Carbio strain. The animals were kept under standard photoperiod cycle of 14/10 h light/dark at 26°C (±1°C). Eggs were collected 1–2 h after starting the light cycle and raised at 26°C in Danieau's medium (17.4 mM NaCl, 0.21 mM KCl, 0.12 mM MgSO₄, 0.18 mM Ca(NO₃)2, 1.5 mM Hepes, pH 7.2). The stages of development were identified according to Iwamatsu (2004).

Animals for colony breeding and embryo production were kept and sampled in accordance with the applicable EU and national German legislation governing animal experimentation, in particular all experimental protocols were approved through an authorization (568/300-1870/13) of the Veterinary Office of the District Government of Lower Franconia, Germany, in accordance with the German Animal Protection Law (TierSchG).

Treatments of embryos and dilutions of the drugs were made in Danieau's medium. To investigate an effect on regulation of sex-related genes, we performed long-term treatments from stage 29, before the sex determination period, and kept in the dark until 1 day after hatching (dah), first meiosis entry period in females. AM580 (10 nM), an agonist of the retinoic acid receptor alpha, and all-trans-retinoic acid ATRA (10 nM) were added to the medium and medium changed every 2 days. The exclude any effect of stress during the treatments, we co-treated the embryos with or without Metyrapone (5 μM, Sigma-Aldrich), a compound that inhibits endogenous cortisol synthesis. The selected drugs concentration for the treatments were based on previous studies (Adolfi et al., 2016, 2019). Specimen were collected at 1 dah and genotyped for sex by PCR for the Y-linked male determining gene dmrt1bY using genomic DNA as template.

The genomic sequence of cyp26a1 (Ensembl gene number ENSORLG00000014516) was retrieved from the Ensembl medaka genome browser (http://www.ensembl.org/Oryzias_latipes). The construction of TALEN expression vectors (left, pCS2TAL3DDD, and right, pCS2TAL3RRR, with both vectors containing the respective TALE fragment, the FokI cleavage domain, and other necessary components) were developed following the standard procedure (Dahlem et al., 2012). The TALEN target sites of cyp26a1 were designed in the second exon, with the right binding site located at the junction of exon 2 and intron 2. The cyp26a1 TALEN recognition sequences were left TALEN 5′ –TCTCCAACATGCACGGAT- 3′ and right TALEN 5′ –TGGAGACTCACCTTTTT- 3′. Between the binding sites, an 18 bp spacer is included, where the FokI nuclease cuts.

In vitro transcription of TALENs was carried out with the Sp6mMESSAGEmMACHINE Kit (Ambion). The resulting mRNA was purified by phenol/chloroform-extraction and then quantified using NanoDrop-2000 (Thermo Scientific). The left and right arm mRNA of each TALEN pair was then mixed at a molar ratio of 1:1, with a final concentration of 100 ng/μL mRNA of each arm, and stored at −80°C until use. About 200 to 600 pg of the mRNA mixture was directly microinjected into medaka embryos at the one-cell stage. The injected embryos were cultivated at 26°C and 10 animals collected at stage 1 dah to extract DNA for mutation efficiency analysis.

To determine the genotypic sex of embryos and adult fish and the presence and absence of mutations, genomic DNA was extracted. Caudal fin clips of the adult fish or whole hatchling were incubated for 1 h at 95°C in 100 μL of Base Solution (25 mM NaOH, 0,2 mM EDTA, pH = 12) and shaking. The solution was cooled down on ice, 100 μL of Neutralization Solution (40 mM Tris-HCl pH = 5.0) added and vortexed. Two microliter of the total volume was used in a 25 μL PCR reaction. The PCR products were resolved on 1% agarose gels.

For determination of genotypic sex, a pair of primers (Supplementary Table 1) was used that amplifies fragments of both dmrt1a (1,100 bp) and dmrt1bY (900 bp), yielding one PCR product (dmrt1a) in XX genotypes, and two PCR products (dmrt1a and dmrt1bY) in XY genotypes. To detect cyp26a1 TALEN mutants, primers were designed flanking the region where the mutations are expected (exon2). PCR product were purified using GenElute^™ Gel Extraction Kit (Sigma-Aldrich) according to the manufacturer's instructions and sequenced using the PCR amplification primers.

HEK 293 cells were cultured in Dulbecco's modified Eagle's medium and 10% fetal calf serum, and maintained at 37°C, 5% CO₂ with 100% humidity. To analyze transcriptional regulation of dmrt1a, an 11875 bp fragment upstream of the start codon was isolated and cloned into pGL4.20 vector containing the firefly luciferase gene (Dmrt1aprom::LucFF) as described (Adolfi et al., 2019). In addition, the responsiveness of the HEK 293 cells to ATRA and AM580 was tested using the pGL3-RARE-luciferase plasmid (Addgene, Cat. 13458), which contains the retinoic acid responsive element (RARE) upstream of the firefly luciferase gene.

Transient transfections in HEK 293 were done at 80% confluency by a polyethylenimine-based procedure. The empty pGL4.20 vector containing the tk promoter and the firefly luciferase gene (pGL4.20-tkmini) was used as negative control. To normalize firefly activity, cells were co-transfected with a Renilla luciferase expressing plasmid (pGL4.74) (Regneri et al., 2015).

For luciferase assays, single wells of a 24-well plate were co-transfected with firefly and Renilla luciferase reporter constructs in a 5:1 molar ratio. The concentration of each construct was calculated in order to obtain a total DNA concentration of 0.5 μg per well. pGL4.20-tkmini and Dmrt1a-prom::LucFF reporter constructs were used with and without co-transfection of the transcriptional activator SF1 of medaka (100 ng). The SF1 expression vector (pcDNA3.1::medakaSF1) was kindly donated from Yann Guiguen (INRA, France). After 16–18 h (day 1), medium was changed. On day 2, cells were incubated for 24 h and with 1 μM ATRA, 10 nM AM580 or DMSO for control. On day 3, cells were harvested in 100 μl of 1 X PLB (Promega).

Renilla and firefly luciferase activities were quantified using the Dual-Luciferase® Reporter Assay System from Promega and the TriStar LB941 microplate multimode reader (Berthold Technologies). Experiments result from at least three replicates and error bars represent the standard error of the mean.

Three individual ovaries and three pools of three testes from wildtype Carbio strain and cyp26a1–/–of medaka were homogenized in TRIzol® reagent (Invitrogen). The total RNA phase was isolated using chloroform and purified using RNeasy® Mini kit (Qiagen) following the manufacturer's instructions. The RNA quality was assessed by measuring the RNA Integrity Number (RIN) using an Agilent 2100 Electrophoresis Bioanalyzer Instrument (Agilent Technologies 2100 Expert). RNA samples with RIN > 8 were used for sequencing.

RNA sequencing libraries were constructed following the standard TruSeq Illumina mRNA library preparation protocol (www.illumina.com; Illumina Inc., BGI, Hong Kong). Read length = 150, sequencing depth for paired end: 65–71 million reads.

Transcriptome sequences were mapped to the O. latipes reference genome (Ensembl Release 93) using the RNA-sequence aligner STAR (https://github.com/alexdobin/STAR/releases). Transcripts were quantified as expected read counts using RSEM (http://deweylab.github.io/RSEM). Differentially expressed genes between testis and ovary were detected by DESeq2 (Love et al., 2014) (Bioconductor/R) for wildtype and mutants. Genes were considered to be differentially expressed, if p.value <= 0.05 AND log2FC ≤ −2 (higher expression in female) and log2FC ≥ +2 (higher expression in male). Histograms for genes with log2FC > 2 AND baseMean > 100 were plotted and genes showing comparable regulation between male and female wildtype and mutant samples were selected. Functional clustering was made using DAVID Bioinformatics Resources 6.8 (https://david.ncifcrf.gov/).

Wildtype and mutant organs of adult (4 months after hatching) males and females and whole embryos of different developmental stages were collected. Total RNA was extracted from 3 pools of adult fish organs (n = 4) or whole embryos (n = 20) using the TRIZOL reagent (Invitrogen) according to the supplier's recommendation. After DNase treatment, reverse transcription was done from 2 μg total RNA using RevertAid First Strand Synthesis kit (Fermentas) and random primers. Real-time quantitative PCR was carried out with SYBR Green reagent and amplifications were detected with a Mastercycler® ep realplex (Eppendorf). All results are averages of at least three independent RT reactions from three independent RNA preparations. Transcript levels of the target genes were normalized against the medaka elongation factor-1 alpha (ef1a) gene (Supplementary Table 1). The ΔCt values presented as means ± standard error of the mean (SEM), were analyzed by one-way ANOVA, Tukey's and Student's t-test. A significance level of P < 0.05 was used for all tests.

Whole larvae and gonads from adult fish were dissected and fixed in Karnovski solution (2% glutaraldehyde and 4% paraformaldehyde in Sörensen buffer [0.1 M, pH 7.2]) for 24 h at 4°C. Then, samples were washed in water, dehydrated in increasing concentrations of ethanol, and embedded in Historesin Technovit 7100 (Kulzer, Hanau, Germany). Serial sections of 2 μm thickness were obtained and counterstained with hematoxylin & eosin (HE).

We performed treatments of medaka embryos at different time points with ATRA and AM580 to activate the RA pathway. From the treated embryos, we analyzed expression of genes known to be involved in sex determination or gonad differentiation. Long-term treatments (stage 29 until 1 dah) of BACdmrt1a::GFP transgenic fish with ATRA resulted in a strong induction of reporter gene expression exclusively in the somatic gonad at hatching stage in both sexes (Figure 1A). Gene expression levels of male-related genes were determined from whole embryos after long-term treatment with AM580 (Figure 1B). The dmrt1bY expression levels were unaffected in males. However, amh and dmrt1a showed significantly increased mRNA levels in both sexes.

To date, the responsiveness of dmrt1a to RA is unknown. Hence, to check whether the treatments had a direct effect by activating dmrt1a transcription, we analyzed the 11,8 kb promoter of dmrt1a after treatments with ATRA or AM580 in HEK 293 cells. The HEK 293 cells were shown to be capable to respond to both ATRA and AM580 when compared to control (DMSO), indicating that the retinoic acid receptors (RAR/RXR) are endogenously produced in this cell line (Supplementary Figure 1). Cells co-transfected with the cofactor SF1 showed a significant increase of the dmrt1a promoter activity. However, treatments with RA had no effect on the promoter of dmrt1a in vitro (Figure 1C).

We showed earlier that cyp26a1 is differently expressed in gonads of medaka, with much higher transcript levels in females than in males (Adolfi et al., 2016) This indicated that cyp26a1 in medaka may perform the role of CYP26B1 in mammals. To evaluate a possible role of cyp26a1 in the timing of meiosis entry and consequently in sex determination/differentiation, we generated knockout cyp26a1 medaka.

To molecularly characterize the cyp26a1 TALEN induced mutations, we amplified the expected target site from F₀ embryos and adult male and female founders and sequenced the 544 bp PCR product (Supplementary Figure 2A). We obtained three different mutant lines (Supplementary Figure 2B) with deletions at the target site. However, only the Δ5 mutation conceptually translates into a protein with a predicted compromised function (frameshift and, premature termination), while the other two presented an in-frame deletion of two or three aa that still could lead to fully functional enzyme (Supplementary Figure 2C). All three mutants present deletions or substitutions within the P450 superfamily domain, however, the Δ5 mutation is predicted to translate into a shorter protein that lacks the Cytochrome P450 cysteine heme-iron ligand signature. Therefore, we generated cyp26a1–/– animals and performed the mutants analyzes only from the Δ5 mutation line.

We then followed the gonad development of wildtype and cyp26a1–/–in male and female larvae at the early meiosis stages (Figure 2). Already at 5 dah, differences in the germ cells are observed in females, in which the cyp26a1–/– present more proliferating germ cells compared to the wildtype (Figures 2A,B), while in male no morphological differences were observed until 15 dah (Figures 2G–J). Mutant females at 10 dah apparently contain more pre-vitellogenic oocytes than wildtype females, indicating increased oogenesis and meiosis entry in the mutant at this stage (Figures 2C,D). At 15 dah, the gonads of both wildtype and cyp26a1–/–females presented no apparent morphological difference anymore (Figures 2E,F). Strikingly, 2 out of 10 15 dah males of cyp26a1–/–had an isolated pre-vitellogenic oocytes inside the undifferentiated gonad, and no sign of germ cell proliferation could be observed (Figures 2K,L). Comparing 4 months old wildtype and mutant mature gonads of medaka no apparent differences in morphology were observed in both sexes (Supplementary Figure 3). Despite the development of oocytes at 15 dah in males of cyp26a1–/– genotype, no sign of any female structure was observed in adult testis.

We performed treatments of cyp26a1 KO embryos with AM580 from embryo stage 29 until 1 dah. Control females of cyp26a1–/–medaka presented morphologically differentiated germ cells already at 1 dah, showing commitment to gametogenesis (Figure 3A). However, XX mutants treated with AM580 had only undifferentiated germ cells (Figure 3B). The developing gonads of both control and treated XY mutants show no sign of germ cell differentiation (Figures 3D,E), indicating that AM580 delays germ cell commitment to gametogenesis in cyp26a1 mutants.

Retinoic acid is an important morphogen, and treatments with AM580 performed during the sex determination period led to several malformations in the embryos (data not shown). Our recent study showed that temperature stress and cortisol treatments lead to masculinization of XX medaka, possibly through a direct activation of dmrt1a in the gonad and repressing germ cell differentiation (Adolfi et al., 2019). Hence, such stress factors in our treatments could possibly interfere with the effect of AM580 in meiosis initiation. To test this hypothesis, we co-treated the cyp26a1 KO embryos with AM580 and Metyrapone, a compound that inhibits the endogenous production of cortisol. Early differentiating oocytes (stage I) were observed in treated females (Figure 3C), and only undifferentiated germ cells were present in males in the same conditions (Figure 3F) showing that RA leads to germ cell differentiation in females.

Expression analyzes of mutant embryos showed that treatments with AM580 resulted in presented upregulation of the male-related gene dmrt1a in both sexes, while the female-related foxl2 gene expression is extremely reduced in females (Figure 3G). Treatments with both AM580 and Metyrapone presented less dmrt1a expression compared to those treated with AM580 alone, while foxl2, foxl3 (oogenesis inducer) and the meiosis marker scp3 were upregulated in females when compared to the control (Figure 3G).

Despite the effect in the early gonad of the cyp26a1 mutants, no apparent effect was seen in the adult animals on the cellular level. Nevertheless, the transcriptome analyses showed significant differences between adult wildtype and KO gonads of both male and female (Figure 4). Differential gene expression analysis (LogFC > 2) revealed that the mutation of cyp26a1 regulated more genes in female than male gonads (Figure 4A; Supplementary Table 2).

The mutant ovaries presented higher number of exclusive upregulated genes (279), some of which are related to retinoic acid metabolism (e.g., aldh1a2, aldh8a1, and rarb), and gametogenesis (spata7, aqp3a, and zar1). In the genes exclusively downregulated in mutant females (264), some factors related to the TGF-beta signaling (smad1, inhbb, and lefty1), retinol metabolism and steroidogenesis (cyp1a) were affected. Interestingly, the gata4 gene, known to be required for gonad formation and testis development in mice (Hu et al., 2013), was strongly downregulated. Despite of a fewer number of genes regulated exclusively in male mutants, the nr4a1 gene, important in testis function (Daems et al., 2014), was highly upregulated.

The genes that were downregulated in the mutant gonads of both sexes (107) are enriched for functions related to mitochondrial electron transport (e.g., cox1, cox2, and cox7a2) and response to oxidative stress (e.g., rsp29). On the other hand, the genes that were upregulated in the mutants of both sexes (41) are related to immune response (e.g., c6) and peptidase activity (e.g., cela1, ela2, and prss1).

The induction of sex-related genes by AM580 treatments already indicated a possible effect of RA on gametogenesis and germ cell differentiation. This could also be observed by the genome wide expression analysis: genes which have a crucial function in germ cell differentiation (ddx4 and dazl) and meiosis (scp3 and dmc1) were slightly upregulated in the mutants of both sexes (Figure 4B). Similarly, growth factors (e.g., amh/amhr2, and sdf1/cxcr4) and transcription factors (e.g., dmrt1 and sox9) related to germ cell differentiation, proliferation and survival were regulated in mutants, especially in females (Figure 4B).