Dnmt3aa but Not Dnmt3ab Is Required for Maintenance of Gametogenesis in Nile Tilapia (Oreochromis niloticus)

Dnmt3a, a de novo methyltransferase, is essential for mammalian germ line DNA methylation. Only one Dnmt3a is identified in mammals, and homozygous mutants of Dnmt3a are lethal, while two Dnmt3a paralogs, dnmt3aa and dnmt3ab, are identified in teleosts due to the third round of genome duplication, and homozygous mutants of dnmt3aa and dnmt3ab are viable in zebrafish. The expression patterns and roles of dnmt3aa and dnmt3ab in gonadal development remain poorly understood in teleosts. In this study, we elucidated the precise expression patterns of dnmt3aa and dnmt3ab in tilapia gonads. Dnmt3aa was highly expressed in oogonia, phase I and II oocytes and granulosa cells in ovaries and spermatogonia and spermatocytes in testes, while dnmt3ab was mainly expressed in ovarian granulosa cells and testicular spermatocytes. The mutation of dnmt3aa and dnmt3ab was achieved by CRISPR/Cas9 in tilapia. Lower gonadosomatic index (GSI), increased apoptosis of oocytes and spermatocytes and significantly reduced sperm quality were observed in dnmt3aa−/− mutants, while normal gonadal development was observed in dnmt3ab−/− mutants. Consistently, the expression of apoptotic genes was significantly increased in dnmt3aa−/− mutants. In addition, the 5-methylcytosine (5-mC) level in dnmt3aa−/− gonads was decreased significantly, compared with that of dnmt3ab−/− and wild type (WT) gonads. Taken together, our results suggest that dnmt3aa, not dnmt3ab, plays important roles in maintaining gametogenesis in teleosts.


Introduction
DNA methylation, a mechanism of epigenetics, plays a crucial role in the control of development-related gene expression during gametogenesis and early embryogenesis [1,2]. During germ cell development, epigenetic reprogramming occurs dynamically, remodeling DNA methylation marks in particular [3,4]. At day 7.5 of early embryogenesis (E7.5) in mice (Mus musculus), global DNA methylation of primordial germ cells (PGCs) is erased. Later, de novo DNA methylation proceeds differentially between male and female germ cells, earlier in spermatogenesis than in oogenesis. In female germ cells, de novo DNA methylation occurs in arrested oocytes in meiotic prophase I. However, in male germ cells, it takes place in mitotically arrested prespermatogonia before birth [5][6][7]. DNA methylation is catalyzed by a group of enzymes called DNA methyltransferases (dnmts), including Dnmt1 and Dnmt3. Dnmt1 is involved in the methylation of hemimethylated DNA and thus called maintenance DNA methyltransferase, while Dnmt3 is able to place methylation marks on previously unmethylated CpGs of DNA and thus is mainly responsible for de novo DNA methylation during development [8,9]. In mammals, the Dnmt3 subfamily is composed of three members, Dnmt3a, Dnmt3b and Dnmt3l [10]. Recently, Dnmt3c, a novel rodent-specific member of the de novo dnmts, has been identified to regulate DNA methylation in the male germline [11]. Of these, Dnmt3a, Dnmt3b and Dnmt3c have been proven to have catalytic activities in vivo, whereas Dnmt3l is a catalytically inactive DNA methyltransferase cofactor [9,12].
Studies on mammals have shown that Dnmt3a, Dnmt3c and Dnmt3l, not Dnmt3b, are required for gametogenesis. During the development of male germ cells in mice, Dnmt3a exhibited dynamic expression patterns, and it is highly expressed in spermatogonia and spermatocytes [13,14]. Male germ cells without Dnmt3a or Dnmt3l undergo meiotic failure, impaired spermatogenesis, which results in no spermatocytes, spermatids or spermatozoa, and significantly reduced testis size in Dnmt3a conditional mutant mice [15][16][17][18]. In female mice, Dnmt3a is expressed in follicles and stromal cells at different developmental stages, and exists in the cytoplasm and nucleus of oocytes and granulosa cells [19,20]. Dnmt3a or Dnmt3l null oocytes fail to acquire methylation during oocyte growth, which leads to abnormal embryonic development after fertilization [15,17].
Teleost-specific whole genome duplication has increased the dnmts copy number [21,22]. Two Dnmt3a paralogous genes (dnmt3aa and dnmt3ab) have been identified in Nile tilapia (Oreochromis niloticus), zebrafish (Danio rerio), flatfish (Solea senegalensis) and ricefield eel (Monopterus albus) [21,[23][24][25]. Studies on dnmt3aa and dnmt3ab in fish have been mainly focused on their expression in gonads. In Nile tilapia and bluehead wrasse (Thalassoma bifasciatum), the expressions of dnmt3aa and dnmt3ab are significantly higher in the testes than in the ovaries, and significantly increased during the sex reversal from female to male induced by Fadrozole [26] or social cues [27,28]. In ricefield eel, dnmt3aa and dnmt3ab are highly expressed in spermatocytes of testes, with the expression of dnmt3aa significantly increased during the female to male sex reversal [25,29]. These studies suggest that dnmt3aa and dnmt3ab may play important roles in the gonadal development of fish; however, their detailed expression profiles during the sex determination and differentiation and gonadal development of teleosts are unclear. Recently, some research groups reported the mutation of dnmt3aa and dnmt3ab in zebrafish, demonstrating their critical function on behavior regulation [30], temperature adaptation [31] and brain neural development [32]. Nevertheless, the roles of dnmt3aa and dnmt3ab in the gonadal development of fish remain unknown.
Nile tilapia belongs to Perciformes, Cichlidae. Fast growth speed, strong adaptability and disease resistance make it an important aquaculture fish in the world. It is considered to be an excellent model for studying gene expression and function related to reproduction and fertility due to its short spawning cycle (14 days), stable XX-XY sex determination system, availability of genetic all-XX and all-XY fish [33] and high-quality genome sequences [34]. In particular, genome editing by CRISPR/Cas9 techniques established in our research group has contributed to the research of gene function in Nile tilapia [35]. Generally, several key biological events occur at different time points during gonadal development in Nile tilapia, such as sex determination and differentiation at 5-10 dah (days after hatching), the initiation of germ cell meiosis and oogenesis in the XX gonads (ovaries) at 30 dah, the initiation of spermatogenesis in the XY gonads (testes) at 75-90 dah and sperm maturation in the XY gonads and vitellogenesis in the XX gonads at 180 dah [36]. In the present study, we clarified the precise expression profiles of dnmt3aa and dnmt3ab during the key stages of gonadal development in females and males and uncovered their roles in reproduction and fertility in teleosts and the possible mechanism involved by homozygous mutant establishment and phenotype analyses in Nile tilapia.

Expression Patterns and Cellular Localization of Dnmt3aa and Dnmt3ab in Developing Gonads
Ontogeny analyses showed that dnmt3aa and dnmt3ab displayed sexually dimorphic expression profiles in developing gonads (Supplementary Figure S1). Dnmt3aa and dnmt3ab were expressed in gonads from as early as 5 dah, with relatively higher expression in the testes than the ovaries. In the testes, dnmt3aa was up-regulated from 5 to 180 dah. In the ovaries, dnmt3aa was up-regulated from 5 dah and reached the highest level at 120 dah, and maintained this level till 180 dah (Supplementary Figure S1a). Dnmt3ab was up-regulated from 5 to 120 dah and maintained at a relatively high level in ovaries and testes at 120 and 180 dah (Supplementary Figure S1b). Generally, higher expression was observed for dnmt3aa than dnmt3ab at all the time points examined.
Cell populations expressing dnmt3aa and dnmt3ab in gonads were identified by fluorescence in situ hybridization (FISH). Dnmt3aa was widely expressed in the gonads of Nile tilapia, highly expressed in oogonia, phase I and II oocytes and granulosa cells of ovaries (Figure 1a-d), and spermatogonia and spermatocytes of testes (Figure 1e-h). Dnmt3ab was mainly expressed in granulosa cells of ovaries (Figure 1i-l) and spermatocytes of testes (Figure 1m-p). In contrast, no signal for dnmt3aa and dnmt3ab mRNA was detected in the gonads using the sense probe (Supplementary Figure S2). dnmt3ab were expressed in gonads from as early as 5 dah, with relatively higher expression in the testes than the ovaries. In the testes, dnmt3aa was up-regulated from 5 to 180 dah. In the ovaries, dnmt3aa was up-regulated from 5 dah and reached the highest level at 120 dah, and maintained this level till 180 dah (Supplementary Figure S1a). Dnmt3ab was up-regulated from 5 to 120 dah and maintained at a relatively high level in ovaries and testes at 120 and 180 dah (Supplementary Figure S1b). Generally, higher expression was observed for dnmt3aa than dnmt3ab at all the time points examined.
Cell populations expressing dnmt3aa and dnmt3ab in gonads were identified by fluorescence in situ hybridization (FISH). Dnmt3aa was widely expressed in the gonads of Nile tilapia, highly expressed in oogonia, phase I and II oocytes and granulosa cells of ovaries (Figure 1a-d), and spermatogonia and spermatocytes of testes (Figure 1e-h). Dnmt3ab was mainly expressed in granulosa cells of ovaries (Figure 1i-l) and spermatocytes of testes (Figure 1m-p). In contrast, no signal for dnmt3aa and dnmt3ab mRNA was detected in the gonads using the sense probe (Supplementary Figure S2). Figure 1. Cellular localization of dnmt3aa (a-h) and dnmt3ab (i-p) in developing gonads by fluorescence in situ hybridization. Dnmt3aa was widely expressed in the gonads of Nile tilapia, especially in oogonia, phase I and II oocytes and granulosa cells of ovaries (a-d) and spermatogonia and spermatocytes of testes (e-h). Dnmt3ab was mainly expressed in granulosa cells of ovaries (i-l) and spermatocytes of testes (m-p). OC, oocytes; OG, oogonia; GC, granulosa cells; SC, spermatocytes; SG, spermatogonia; dah, days after hatching. Red fluorescence represents the dnmt3aa and dnmt3ab signals. Blue fluorescence represents the DAPI signals. Arrows indicate the positive signals. White boxes indicate the regions magnified in (d) and (i).

Establishment of Nile Tilapia Dnmt3aa and Dnmt3ab Mutant Lines by CRISPR/Cas9
The guide RNA (gRNA) sites containing Hpy 188III and Mly I adjacent to protospacer adjacent motif (PAM) were selected in the third and second exon of dnmt3aa and dnmt3ab, respectively (Figure 2a,b). Complete digestion of the PCR products from dnmt3aa and dnmt3ab with Hpy 188III and Mly I, respectively, produced two fragments in the control groups, while an intact DNA fragment was observed in embryos injected with both Cas9 mRNA and target gRNA (Figure 2a,b). Representative Sanger sequencing results from the uncleaved bands were listed. In-frame and frame-shift deletions induced at the target sites were confirmed by Sanger sequencing (Figure 2a  . Cellular localization of dnmt3aa (a-h) and dnmt3ab (i-p) in developing gonads by fluorescence in situ hybridization. Dnmt3aa was widely expressed in the gonads of Nile tilapia, especially in oogonia, phase I and II oocytes and granulosa cells of ovaries (a-d) and spermatogonia and spermatocytes of testes (e-h). Dnmt3ab was mainly expressed in granulosa cells of ovaries (i-l) and spermatocytes of testes (m-p). OC, oocytes; OG, oogonia; GC, granulosa cells; SC, spermatocytes; SG, spermatogonia; dah, days after hatching. Red fluorescence represents the dnmt3aa and dnmt3ab signals. Blue fluorescence represents the DAPI signals. Arrows indicate the positive signals. White boxes indicate the regions magnified in (d,i).

Establishment of Nile Tilapia Dnmt3aa and Dnmt3ab Mutant Lines by CRISPR/Cas9
The guide RNA (gRNA) sites containing Hpy 188III and Mly I adjacent to protospacer adjacent motif (PAM) were selected in the third and second exon of dnmt3aa and dnmt3ab, respectively (Figure 2a,b). Complete digestion of the PCR products from dnmt3aa and dnmt3ab with Hpy 188III and Mly I, respectively, produced two fragments in the control groups, while an intact DNA fragment was observed in embryos injected with both Cas9 mRNA and target gRNA (Figure 2a,b). Representative Sanger sequencing results from the uncleaved bands were listed. In-frame and frame-shift deletions induced at the target sites were confirmed by Sanger sequencing (Figure 2a,b).
F1 generation fish were obtained by crossing chimeric XY F0 males and WT XX females. Heterozygous F1 offspring with a deletion of 4 bp for dnmt3aa and 5 bp for dnmt3ab were selected to breed the F2 generation ( Figure 2c). Further, homozygous mutant fish of dnmt3aa and dnmt3ab were validated by Sanger sequencing (Figure 2d,e). Frame-shift mutations led to premature termination of the translation of dnmt3aa at amino acid 126 and dnmt3ab at amino acid 36 (Figure 2f,g). Restriction enzyme digestion assay identified the dnmt3aa +/+ , dnmt3aa +/− , dnmt3aa −/− , dnmt3ab +/+ , dnmt3ab +/− and dnmt3ab −/− individuals (Figure 2h,i). The loss of dnmt3aa and dnmt3ab mRNA was confirmed by reverse transcription PCR (RT-PCR) using a specific primer on the target site ( Figure 2j). F1 generation fish were obtained by crossing chimeric XY F0 males and WT XX f males. Heterozygous F1 offspring with a deletion of 4 bp for dnmt3aa and 5 bp for dnmt3 were selected to breed the F2 generation ( Figure 2c). Further, homozygous mutant fish dnmt3aa and dnmt3ab were validated by Sanger sequencing (Figure 2d,e). Frame-shift m tations led to premature termination of the translation of dnmt3aa at amino acid 126 an dnmt3ab at amino acid 36 (Figure 2f,g). Restriction enzyme digestion assay identified th dnmt3aa +/+ , dnmt3aa +/− , dnmt3aa −/− , dnmt3ab +/+ , dnmt3ab +/− and dnmt3ab −/− individuals (Figu 2h,i). The loss of dnmt3aa and dnmt3ab mRNA was confirmed by reverse transcription PC (RT-PCR) using a specific primer on the target site ( Figure 2j). sequences of forward primer were designed on the target site, which is indicated by white box. No band was amplified in the homozygous mutants, while one band corresponding to dnmt3aa and dnmt3ab mRNA was amplified in the WT XY testes. gapdh was used as internal control.

Gonadal Morphology and Histology of Dnmt3aa −/− and Dnmt3ab −/− Female Mutants
Morphological observation showed that the dnmt3aa −/− ovaries atrophied at 60 da while there was no difference in gonad morphology between dnmt3ab −/− and WT (Figu 3a-c). We selected three different sampling points (part1, part2 and part3) from dnmt3aa ovaries, including the smaller gonad part, the thicker part and the atrophy part, for hist logical observation. The results showed that at 60 dah, WT and dnmt3ab −/− ovaries we full of oogonia and phase I and phase II oocytes, while dnmt3aa −/− ovaries had only a fe oogonia and oocytes (Figure 3a'-c'). Consistently, statistical analysis showed that the G of dnmt3aa −/− fish was significantly lower than that of WT fish, and the number of follicl at different developmental stages was significantly reduced, while there was no signi cant difference between dnmt3ab −/− and WT fish (Figure 3d,e).
Two more developmental stages at 120 dah and 240 dah were further analyzed. Mo phological observations showed that dnmt3aa −/− ovaries were still atrophied and degene ated, while dnmt3ab −/− ovaries developed normally (Figure 3f-h,k-m). Histological obse vation showed that the WT and dnmt3ab −/− ovaries were filled with phase II and phase I oocytes at 120 dah, while only a few phase II and phase III oocytes existed in dnmt3aa sequences of forward primer were designed on the target site, which is indicated by white box. No band was amplified in the homozygous mutants, while one band corresponding to dnmt3aa and dnmt3ab mRNA was amplified in the WT XY testes. gapdh was used as internal control.

Gonadal Morphology and Histology of Dnmt3aa −/− and Dnmt3ab −/− Female Mutants
Morphological observation showed that the dnmt3aa −/− ovaries atrophied at 60 dah, while there was no difference in gonad morphology between dnmt3ab −/− and WT (Figure 3a-c). We selected three different sampling points (part1, part2 and part3) from dnmt3aa −/− ovaries, including the smaller gonad part, the thicker part and the atrophy part, for histological observation. The results showed that at 60 dah, WT and dnmt3ab −/− ovaries were full of oogonia and phase I and phase II oocytes, while dnmt3aa −/− ovaries had only a few oogonia and oocytes (Figure 3a'-c'). Consistently, statistical analysis showed that the GSI of dnmt3aa −/− fish was significantly lower than that of WT fish, and the number of follicles at different developmental stages was significantly reduced, while there was no significant difference between dnmt3ab −/− and WT fish (Figure 3d,e).
Two more developmental stages at 120 dah and 240 dah were further analyzed. Morphological observations showed that dnmt3aa −/− ovaries were still atrophied and degenerated, while dnmt3ab −/− ovaries developed normally (Figure 3f-h,k-m). Histological observation showed that the WT and dnmt3ab −/− ovaries were filled with phase II and phase III oocytes at 120 dah, while only a few phase II and phase III oocytes existed in dnmt3aa −/− ovaries (Figure 3f'-h'). At 240 dah, the oocytes of WT, dnmt3aa −/− and dnmt3ab −/− females were matured, and less oogonia and phase I, phase II, phase III and phase IV follicle cells were observed in dnmt3aa −/− ovaries than the dnmt3ab −/− and WT ovaries (Figure 3k'-m'). Consistently, statistical analysis showed that the GSI and the number of follicles of dnmt3aa −/− were significantly decreased (Figure 3i,j,n,o). In addition, the number of follicles in the dnmt3aa −/− ovarian smaller part (part1) and atrophy part (part3) was also significantly reduced (Supplementary Figure S3). Taken together, these results suggest that the homozygous mutation of dnmt3aa resulted in reduced follicles and ovarian atrophy and degeneration in Nile tilapia. Consistently, statistical analysis showed that the GSI and the number of follicles of dnmt3aa −/− were significantly decreased (Figure 3i,j,n,o). In addition, the number of follicles in the dnmt3aa −/− ovarian smaller part (part1) and atrophy part (part3) was also significantly reduced (Supplementary Figure S3). Taken together, these results suggest that the homozygous mutation of dnmt3aa resulted in reduced follicles and ovarian atrophy and degeneration in Nile tilapia.

Gonadal Morphology and Histology of Dnmt3aa −/− and Dnmt3ab −/− Male Mutants
At 60 dah, no obvious difference was observed in the testicular morphology among them (Figure 4a

Germ Cell Apoptosis in Dnmt3aa −/− and Dnmt3ab −/− Gonads
Germ cells in the ovaries of dnmt3aa −/− fish were significantly decreased. In order to further evaluate the effect of dnmt3aa and dnmt3ab mutation on the ovary development of Nile tilapia, the total and apoptotic germ cells were examined in WT, dnmt3aa −/− and dnmt3ab −/− fish by Vasa (germ cells marker) and TUNEL immunofluorescence co-staining. (Figure 5a-o). The number of Vasa-positive cells in dnmt3aa −/− ovaries was significantly reduced compared with that of WT fish, but there was no difference between dnmt3ab −/− and WT fish (Figure 5p). TUNEL assay showed that there was a large number of germ cells in apoptosis in dnmt3aa −/− ovaries (Figure 5j), but there was no obvious germ cell apoptosis in ovaries of the dnmt3ab −/− and WT fish (Figure 5e,o). Statistical analysis showed that the number of apoptotic germ cells in the dnmt3aa −/− ovaries was significantly increased compared with that in the dnmt3ab −/− and WT ovaries (Figure 5q). The expression of Cyp19a1a, the key enzyme of estrogen synthesis, was further examined and the positive signals were still observed in the dnmt3aa −/− ovaries (Supplementary Figure S4). It is worth noting that a significant reduction in germ cells in the ovaries was also observed in the F0 generation mutants with a high mutation rate (75%) at another dnmt3aa target site (Supplementary Figure S5).   Figure S4). It is worth noting that a significant reduction in germ cells in the ovaries was also observed in the F0 generation mutants with a high mutation rate (75%) at another dnmt3aa target site (Supplementary Figure S5).
Similar to the situation observed in ovaries, the number of spermatocytes in the testes of dnmt3aa −/− fish decreased significantly at 120 dah. To further evaluate the effect of dnmt3aa and dnmt3ab mutation on the testis development of Nile tilapia, the total and apoptotic spermatocytes in WT, dnmt3aa −/− and dnmt3ab −/− fish were examined by Sycp3 (spermatocyte marker) and TUNEL immunofluorescence co-staining (Figure 6a-o). Statistical analysis showed that the number of Sycp3-positive cells in dnmt3aa −/− fish was significantly lower than that of WT fish, while no significant difference in the number of Similar to the situation observed in ovaries, the number of spermatocytes in the testes of dnmt3aa −/− fish decreased significantly at 120 dah. To further evaluate the effect of dnmt3aa and dnmt3ab mutation on the testis development of Nile tilapia, the total and apoptotic spermatocytes in WT, dnmt3aa −/− and dnmt3ab −/− fish were examined by Sycp3 (spermatocyte marker) and TUNEL immunofluorescence co-staining (Figure 6a-o). Statistical analysis showed that the number of Sycp3-positive cells in dnmt3aa −/− fish was significantly lower than that of WT fish, while no significant difference in the number of Sycp3-positive cells was observed in dnmt3ab −/− and WT fish (Figure 6p). A large number of apoptotic spermatocytes, significantly higher than that of WT and dnmt3ab −/− fish, was observed in dnmt3aa −/− fish (Figure 6j,q).   Similar to the situation observed in ovaries, the number of spermatocytes in the testes of dnmt3aa −/− fish decreased significantly at 120 dah. To further evaluate the effect of dnmt3aa and dnmt3ab mutation on the testis development of Nile tilapia, the total and apoptotic spermatocytes in WT, dnmt3aa −/− and dnmt3ab −/− fish were examined by Sycp3 (spermatocyte marker) and TUNEL immunofluorescence co-staining (Figure 6a-o). Statistical analysis showed that the number of Sycp3-positive cells in dnmt3aa −/− fish was significantly lower than that of WT fish, while no significant difference in the number of Sycp3-positive cells was observed in dnmt3ab −/− and WT fish (Figure 6p). A large number of apoptotic spermatocytes, significantly higher than that of WT and dnmt3ab −/− fish, was observed in dnmt3aa −/− fish (Figure 6j,q).   Semen was obtained from the mature WT, dnmt3aa −/− and dnmt3ab −/− mutants by in vitro extrusion at 240 dah, and analyzed with a computer-assisted sperm analyzer after 1:10 dilution. The sperm from dnmt3aa −/− mutants displayed lower activity compared with those from the WT and dnmt3ab −/− fish (Figure 7a-c). In addition, the sperm concentration of dnmt3aa −/− mutants was significantly lower than that of WT and dnmt3ab −/− fish (Figure 7d). Further analysis showed that the percentage of progressive sperm in dnmt3aa −/− fish was significantly lower than that in WT and dnmt3ab −/− fish (Figure 7e), and the proportion of immotile sperm was significantly higher than that in WT and dnmt3ab −/− fish (Figure 7f). Furthermore, the VSL (straight linear velocity) (Figure 7g

Sperm Quality of WT, Dnmt3aa −/− and Dnmt3ab −/− XY Fish
Semen was obtained from the mature WT, dnmt3aa −/− and dnmt3ab −/− mutants by in vitro extrusion at 240 dah, and analyzed with a computer-assisted sperm analyzer after 1:10 dilution. The sperm from dnmt3aa −/− mutants displayed lower activity compared with those from the WT and dnmt3ab −/− fish (Figure 7a-c). In addition, the sperm concentration of dnmt3aa −/− mutants was significantly lower than that of WT and dnmt3ab −/− fish ( Figure  7d). Further analysis showed that the percentage of progressive sperm in dnmt3aa −/− fish was significantly lower than that in WT and dnmt3ab −/− fish (Figure 7e), and the proportion of immotile sperm was significantly higher than that in WT and dnmt3ab −/− fish (Figure 7f).

Apoptosis Gene Expression and Compensatory Expression of Dnmt Family Genes in
Dnmt3aa −/− and Dnmt3ab −/− Gonads As mentioned above, there was a large number of germ cells in apoptosis in the gonads of dnmt3aa −/− fish. Thus, the expression of apoptosis genes in 60 dah XX and 120 dah XY gonads was further analyzed by quantitative real-time PCR (qRT-PCR). The results showed that the expression of apoptosis genes baxa, baxb, caspase3a, caspase3b and caspase8 was significantly increased in dnmt3aa −/− ovaries at 60 dah, but there was no significant difference between dnmt3ab −/− and WT ovaries (Figure 8a). Gonads of dnmt3aa and dnmt3ab homozygous mutants were analyzed for gene compensatory expression. The homozygous mutation of dnmt3ab caused a compensatory increase in dnmt3aa expression at 60 dah. However, there was no significant difference in the expression of other dnmts between the mutants and WT fish (Figure 8b). At 120 dah, the expression of apoptosis genes baxa, baxb, caspase3b and caspase8 in dnmt3aa −/− males was significantly increased (Figure 8c). Interestingly, dnmt3aa and dnmt3ab were found to compensate for each other in the homozygous mutant males at 120 dah (Figure 8d).

Apoptosis Gene Expression and Compensatory Expression of Dnmt Family Genes in Dnmt3aa −/− and Dnmt3ab −/− Gonads
As mentioned above, there was a large number of germ cells in apoptosis in the gonads of dnmt3aa −/− fish. Thus, the expression of apoptosis genes in 60 dah XX and 120 dah XY gonads was further analyzed by quantitative real-time PCR (qRT-PCR). The results showed that the expression of apoptosis genes baxa, baxb, caspase3a, caspase3b and caspase8 was significantly increased in dnmt3aa −/− ovaries at 60 dah, but there was no significant difference between dnmt3ab −/− and WT ovaries (Figure 8a). Gonads of dnmt3aa and dnmt3ab homozygous mutants were analyzed for gene compensatory expression. The homozygous mutation of dnmt3ab caused a compensatory increase in dnmt3aa expression at 60 dah. However, there was no significant difference in the expression of other dnmts between the mutants and WT fish (Figure 8b). At 120 dah, the expression of apoptosis genes baxa, baxb, caspase3b and caspase8 in dnmt3aa −/− males was significantly increased ( Figure  8c). Interestingly, dnmt3aa and dnmt3ab were found to compensate for each other in the homozygous mutant males at 120 dah (Figure 8d). WT, wild type. The reference gene gapdh was used to normalize the expression values. Results are presented as mean ± SD. Different letters above the error bar indicate statistical differences at p < 0.05 as determined by one-way ANOVA followed by Tukey's post hoc test. dah, days after hatching.

The Global DNA Methylation Level of WT, Dnmt3aa −/− and Dnmt3ab −/− Gonads
Immunoreactive signals of 5-methylcytosine (5-mC) were predominantly present in the nuclei of oocytes and granulosa cells in ovaries (Figure 9a-c) and spermatogonia, spermatocytes and spermatozoa in testes (Figure 9d-f). The immunoreactive signals were strong in spermatocytes and spermatozoa but weak in spermatogonia (Figure 9d-f). Statistical analysis showed that the 5-mC levels in ovaries and testes of dnmt3aa −/− fish were significantly lower than those of WT and dnmt3ab −/− fish. There were no significant differences in the 5-mC levels in ovaries and testes between dnmt3ab −/− and WT fish (Figure 9g,h). Further analysis showed that the 5-mC levels of granulosa cells in ovaries and spermatogonia, spermatocytes and spermatozoa in testes of dnmt3aa −/− fish were significantly lower than those of dnmt3ab −/− and WT fish (Supplementary Figure S7). These results showed   (Figure 9a-c) and spermatogonia, spermatocytes and spermatozoa in testes (Figure 9d-f). The immunoreactive signals were strong in spermatocytes and spermatozoa but weak in spermatogonia (Figure 9d-f). Statistical analysis showed that the 5-mC levels in ovaries and testes of dnmt3aa −/− fish were significantly lower than those of WT and dnmt3ab −/− fish. There were no significant differences in the 5-mC levels in ovaries and testes between dnmt3ab −/− and WT fish (Figure 9g,h). Further analysis showed that the 5-mC levels of granulosa cells in ovaries and spermatogonia, spermatocytes and spermatozoa in testes of dnmt3aa −/− fish were significantly lower than those of dnmt3ab −/− and WT fish (Supplementary Figure S7).
These results showed that the mutation of dnmt3aa significantly reduced the 5-mC levels in Nile tilapia ovaries and testes. that the mutation of dnmt3aa significantly reduced the 5-mC levels in Nile tilapia ovaries and testes.

Discussion
DNA methylation, mediated by dnmts, is required for proper embryonic development and for the formation of mature functional germ cells [37]. Dnmt3a, one of the dnmts, is responsible for de novo methylation of mammalian germ cells and plays crucial roles in mammalian gonad development [17,[38][39][40]. In teleosts, there are two Dnmt3a paralogs, dnmt3aa and dnmt3ab, due to the third round of genome duplication [21]. The expression patterns and roles of dnmt3aa and dnmt3ab in gonadal development remain poorly understood in teleosts. In the present study, the expression patterns of dnmt3aa and dnmt3ab were examined at different development stages of gonads in Nile tilapia. Homozygous mutation of dnmt3aa and dnmt3ab was established and gonadal phenotypes and possible mechanisms were analyzed and discussed.

Different Expression Patterns of Dnmt3aa and Dnmt3ab in Gonads Indicate the Subfunctionalization in Teleosts
Gene expression patterns are important aspects of gene regulation and function analysis. The analysis of Dnmt3a expression patterns in gonads is mainly focused on mammals [19,41], but rarely reported in fish. In humans (Homo sapiens), Dnmt3a mRNA expression is detected in the ovarian follicles from primordial to secondary follicles, granulosa cells, and germinal vesicle (GV) and metaphase II (MII) oocytes [42]. In addition, Dnmt3a is expressed in human spermatogenic cells, including spermatogonia cells, primary spermatocytes, secondary spermatocytes and round spermatids [43]. In rhesus monkeys (Macaca mulatto), Dnmt3a mRNA is expressed in follicles at different developmental stages [44]. In female mice, the Dnmt3a protein is localized around the nucleus of GV oocytes and in the cytoplasm of the MII oocytes [45]. In male mice, Dnmt3a is expressed at mRNA and protein levels in male germ cells, at a high level in type A spermatogonia, slightly

Discussion
DNA methylation, mediated by dnmts, is required for proper embryonic development and for the formation of mature functional germ cells [37]. Dnmt3a, one of the dnmts, is responsible for de novo methylation of mammalian germ cells and plays crucial roles in mammalian gonad development [17,[38][39][40]. In teleosts, there are two Dnmt3a paralogs, dnmt3aa and dnmt3ab, due to the third round of genome duplication [21]. The expression patterns and roles of dnmt3aa and dnmt3ab in gonadal development remain poorly understood in teleosts. In the present study, the expression patterns of dnmt3aa and dnmt3ab were examined at different development stages of gonads in Nile tilapia. Homozygous mutation of dnmt3aa and dnmt3ab was established and gonadal phenotypes and possible mechanisms were analyzed and discussed.

Different Expression Patterns of Dnmt3aa and Dnmt3ab in Gonads Indicate the Sub-Functionalization in Teleosts
Gene expression patterns are important aspects of gene regulation and function analysis. The analysis of Dnmt3a expression patterns in gonads is mainly focused on mammals [19,41], but rarely reported in fish. In humans (Homo sapiens), Dnmt3a mRNA expression is detected in the ovarian follicles from primordial to secondary follicles, granulosa cells, and germinal vesicle (GV) and metaphase II (MII) oocytes [42]. In addition, Dnmt3a is expressed in human spermatogenic cells, including spermatogonia cells, primary spermatocytes, secondary spermatocytes and round spermatids [43]. In rhesus monkeys (Macaca mulatto), Dnmt3a mRNA is expressed in follicles at different developmental stages [44]. In female mice, the Dnmt3a protein is localized around the nucleus of GV oocytes and in the cytoplasm of the MII oocytes [45]. In male mice, Dnmt3a is expressed at mRNA and protein levels in male germ cells, at a high level in type A spermatogonia, slightly decreased in type B spermatogonia, and further decreased in preleptotene and pachytene spermatocytes [13,14]. In general, Dnmt3a is expressed in mammalian follicles and granulosa cells of ovaries and spermatogenic cells of testes. There are few studies on the cellular localization of dnmt3aa and dnmt3ab in teleosts except for a report describing immunoreactive signals for Dnmt3a (Dnmt3aa and Dnmt3ab) in male germ cells, particularly in spermatocytes in ricefield eels [25]. In this study, we comprehensively studied the expression patterns and their cellular localization of dnmt3aa and dnmt3ab in the ovaries and testes of Nile tilapia. Dnmt3aa and dnmt3ab displayed sexually dimorphic expression profiles in developing gonads, with a higher expression in testes than in ovaries. Furthermore, higher expression was observed for dnmt3aa than dnmt3ab at all the time points examined, suggesting that dnmt3aa may play more important roles in gonadal development than dnmt3ab. In Nile tilapia, dnmt3aa was highly expressed in oogonia, phase I and II oocytes and granulosa cells, while dnmt3ab was mainly expressed in granulosa cells of the ovaries. Both were highly expressed in spermatogonia and spermatocytes of the testes. These results demonstrated that the expression of dnmt3aa and dnmt3ab in Nile tilapia was basically consistent with that in mammals. Interestingly, in ovaries, dnmt3aa was expressed in both germ cells (oogonia and oocytes) and somatic cells (granulosa cells), while dnmt3ab was only expressed in somatic cells (granulosa cells). The different expression patterns of dnmts in gonads are probably essential for the acquisition of a sex-specific methylation pattern [46]. The different expression patterns of dnmt3aa and dnmt3ab in the gonads of Nile tilapia suggest that they may play different roles in the gonadal development of Nile tilapia.

Homozygous Mutants of Dnmt3aa or Dnmt3ab Are Viable in Nile Tilapia
The roles of Dnmt3a in early development and gametogenesis in mammals have been investigated by gene targeting disruption. Although Dnmt3a −/− mice develop to term and appear to be normal at birth, most homozygous mutant mice become runted and die at about 4 weeks of age [47]. These results indicate that the deficiency of Dnmt3a in mice is fatal. However, recent studies on zebrafish have shown that the homozygous mutants of dnmt3aa or dnmt3ab are viable. Dnmt3aa and dnmt3ab are demonstrated to have essential and non-overlapped functions in modulating behavioral control, and dnmt3aa or dnmt3ab mutation fish display anxiety symptoms [30]. In addition, some groups have demonstrated the roles of dnmt3aa or dnmt3ab in regulating developmental thermal plasticity, and the phenotypic effects of dnmt3aa and dnmt3ab are additive [31]. In this study, the growth and survival of the homozygous mutants of dnmt3aa and dnmt3ab were grossly normal compared with the WT fish. Different spatial-temporal expression profiles of dnmt3aa and dnmt3ab, the up-regulation of dnmt3ab in dnmt3aa −/− gonads and the upregulation of dnmt3aa in dnmt3ab −/− gonads were observed. These results indicate the sub-functionalization and compensation of dnmt3aa and dnmt3ab in Nile tilapia, which might be possible reasons for the survival of the mutants. As suggested in a previous study [48], the existence of duplicate genes probably increased the survival opportunities of the species. Our study provided another example for this notion.

Deficiency of Dnmt3aa, but Not Dnmt3ab, Results in Decrease in GSI, Reduction in Germ Cells and Decline in Sperm Quality in Nile Tilapia
In mammals, Dnmt3a is essential for both male and female germ lines' de novo methylation [5,17,49]. During follicular development after birth, DNA methylation plays important roles in the regulation of gene expression in oocytes. DNA methylation also plays crucial roles in normal spermatogenesis. Previous studies have demonstrated that male germ cells in mice have a highly distinct epigenetic pattern, characterized by a unique genome-wide pattern of DNA methylation. During spermatogenesis, global remethylation is established from spermatogonia to spermatocytes, and continues in the round spermatids and spermatozoa [50,51]. It has been reported that abnormal DNA methylation in spermatogenic cells due to genetic failure, environmental factors and disturbed expression of the dnmts may lead to spermatogenic impairments [52][53][54]. Indeed, azoospermia studies in humans have shown that significant changes in dnmts expression and DNA methylation levels in spermatogenic cells may lead to male infertility [55]. In this study, mutants of dnmt3aa showed atrophy and degeneration of ovaries and more transparent testes, a lower GSI and a significant decrease in oocytes and spermatocytes, while mutants of dnmt3ab showed normal development of ovaries and testes. Our results demonstrate that dnmt3aa, but not dnmt3ab, plays important roles in gonadal development and gamogenesis.
In humans, low sperm motility is associated with decreased sperm methylation [56]. Compared with WT and dnmt3ab −/− fish, the dnmt3aa −/− mutants displayed a significantly lower sperm concentration and significantly reduced progressive sperm proportion. The VSL, VCL and BCF in dnmt3aa −/− mutants were significantly lower than those of the WT and dnmt3ab −/− fish. Further observation and analysis showed that compared with the WT, tails of some sperm in dnmt3aa −/− fish were shorter and curved, which might account for the decrease in sperm motility.

Low Global DNA Methylation Results in Germ Cell Apoptosis and Abnormal Phenotype in
Gonads of Dnmt3aa −/− Mutants DNA methylation is essential for the normal development of cells, and abnormal DNA methylation can result in cell apoptosis [57]. Previous reports have demonstrated that the conditional knockout of Dnmt3a in mice results in a significant reduction in DNA methylation in oocytes [49,58,59]. In this study, the mutation of dnmt3aa significantly reduced the global DNA methylation levels in Nile tilapia ovaries and testes. Further analysis showed that the 5-mC levels of granulosa cells in ovaries and spermatogonia, spermatocytes and spermatozoa in testes of dnmt3aa −/− fish were significantly lower than those of dnmt3ab −/− and WT fish. Our results indicate that dnmt3aa is essential for the correct establishment of DNA methylation patterns of gonads in female and male Nile tilapia. The deficiency of dnmt3aa results in a significant reduction in DNA methylation, which induces apoptotic gene expression and germ cell apoptosis. In this study, a significant increase in apoptosis signals in oocytes and spermatocytes was observed in dnmt3aa −/− mutants and the expressions of apoptotic genes were significantly increased. The apoptosis of germ cells resulted in a decrease in GSI, reduction in germ cells and gonadal degeneration in female and male mutants.
In female mice, conditional knockout of Dnmt3a in oocytes results in the failure of DNA methylation establishment in oocytes, loss of maternal imprints and death of offspring in the uterus [5,17,39]. However, mouse oocytes with Dnmt3a conditional knockout are capable of being fertilized, different from the phenotype observed in Nile tilapia in which global mutation of dnmt3aa caused oocyte apoptosis and ovarian degeneration. The discrepancy might be attributed to the following reasons. The oocyte does not develop in isolation but is instead highly dependent on surrounding granulosa cells of the intact ovarian follicle. The process of oocyte development is supported by granulosa cells, and bidirectional communication between oocytes and granulosa cells is important for the development and maturation of the oocytes [60][61][62]. Granulosa cells undergo dynamic DNA methylation changes to repress or activate the genes required for their proliferation and differentiation during follicular development [63,64]. Correctly established DNA methylation in granulosa cells is important for the regulation of the expression of genes related to follicular development [65]. Changing Dnmt3a gene expression in granulosa cells causes impaired oocyte maturation in the GV and MII oocytes and subsequent abnormal embryonic development [20,66]. In this study, the mutation of dnmt3aa resulted in a significant reduction in the methylation level of ovarian granulosa cells. The significant decrease in the 5-mC level of granulosa cells affected the development of oocytes and resulted in oocyte apoptosis in Nile tilapia. In addition, global mutation of dnmt3aa might also influence estrogen synthesis via the hypothalamic-pituitary-gonadal axis and result in defects in vitellogenesis, which, in turn, may result in oocyte apoptosis. It is worth noting that in zebrafish, medaka (Oryzias latipes) and Nile tilapia, a significant decrease in ovarian germ cells led to female to male sex reversal [35,[67][68][69]. In this study, positive signals of Cyp19a1a were observed in dnmt3aa −/− ovaries, indicating the mutants were still female. In fact, a small number of oocytes in the ovaries continued to develop at 120 and 240 dah. The remaining germ cells maintained the ovarian phenotype.
In male mice, germ cells without Dnmt3a undergo meiotic failure, impaired spermatogenesis, which results in no spermatocytes, spermatids or spermatozoa, and significantly reduced testis size [17]. However, viable sperm were produced in dnmt3aa −/− fish even though the sperm concentration and motility were significantly lower than those of WT fish, which is different from the phenotype of Dnmt3a conditional mutation in mice. A significant up-regulation of dnmt3ab expression in dnmt3aa −/− testes indicated that the compensatory increase in dnmt3ab might be one of the reasons for these differences.

Animal Rearing Conditions
Nile tilapia (parental fish, one year old) were raised in circulating aerated fresh water tank under natural light at 26 ± 0.5 • C. The water quality parameters were monitored daily (pH: 7.2 ± 0.5; dissolved oxygen: 6.5-7.0 mg/L). Fish were fed three times a day with commercial feed (Shengsuo, Yantai, China). All-XX progenies were obtained by crossing a pseudomale (XX male, producing sperm after hormonal sex reversal) with a normal XX female. All-XY progenies were obtained by crossing a YY super male with an XX female. Larvae and juveniles were raised under the same conditions as the parental fish. Animal experiments were conducted in accordance with the regulations of Guide for Care and Use of Laboratory Animals and were approved by the Committee of Laboratory Animal Experimentation at Southwest University, China. (No. IACUC-20181015-12, 15 October 2018).

Establishment of Dnmt3aa and Dnmt3ab Homozygous Mutant Lines by CRISPR/Cas9
Briefly, the gRNA of dnmt3aa/dnmt3ab and Cas9 mRNA were co-injected into one-cellstage embryos (0-90 min after fertilization) at a final concentration of 250 and 500 ng/µL, respectively. All of the injected embryos were incubated at 26 • C. Twenty injected embryos were collected at 72 h after injection. Genomic DNA was extracted from pooled control and injected embryos and used to assess the mutations. DNA fragments spanning the target site were amplified. The mutated sequences were analyzed by restriction enzyme digestion (Hpy 188III for dnmt3aa and Mly I for dnmt3ab) and Sanger sequencing.
Heterozygous F1 offspring were obtained by F0 XY male founders mated with WT XX females. The F1 fish were genotyped by fin clip assay and the individuals with frame-shift mutations were selected. XY male and XX female siblings of F1 generation, carrying the same mutation, were mated to generate homozygous F2 mutants. The F1 mutant fish that carried 4 and 5 base-pair deletions were used for construction of F2 dnmt3aa and dnmt3ab mutants, respectively. The dnmt3aa and dnmt3ab F2 mutants were screened using restriction enzyme digestion and Sanger sequencing. The genetic sex of each fish was determined by genotyping using sex-linked marker (marker 5) as described previously [33,70].

Gonad Morphological and Histological Analysis
To investigate the roles of dnmt3aa and dnmt3ab in gonadal development, the gross morphology and histology of WT, dnmt3aa −/− and dnmt3ab −/− ovaries and testes were analyzed at 60, 120 and 240 dah. The fish were anesthetized with MS-222 (Sigma-Aldrich, St. Louis, MO, USA) and the gonad morphology of the mutants and WT fish was imaged by stereomicroscope (Leica, Bensheim, Germany) after biopsy. The body and gonad weight were measured for GSI calculating (n = 10 for each genotype). Then, the gonads were fixed in Bouin's solution for 24 h at room temperature. The fixed samples were then processed as follows: serial dehydration in 70, 80 and 90% ethanol for 1.5 h each, 95% ethanol for 2 h and 100% ethanol three times for 1 h each; sequential clearance in xylene and ethanol mixture (1:1) for 30 min and xylene twice for 30 min each; and infiltration in paraffin 2 h. The samples were sectioned at 5 µm thickness using the Leica microtome (Leica Microsystems, Wetzlar, Germany). The sections were stained with hematoxylin and eosin (H&E) as described previously [71]. Photographs were taken under Olympus BX51 light microscope (Olympus, Tokyo, Japan). Sibling WT fish were used as control for phenotype analysis. Germ cells from the median sections of testes (n = 5 for each genotype) and different parts (1-3) of ovaries (n = 5 for each genotype) were counted for statistical analyses. The histological classification of the follicles and spermatogenic cells for Nile tilapia was performed according to the standards described previously [72,73].

Gene Expression Analyses by qRT-PCR
Gonads of the WT, dnmt3aa −/− and dnmt3ab −/− fish were dissected at different developmental stages (5, 30, 60, 90, 120 and 180 dah) for gene expression assay. Three parallel ovary and testis samples with different number of gonads based on the age of fish were prepared at each developmental stage. Total RNA (1.0 µg) was extracted and treated with DNase I to eliminate genomic DNA contamination, and was reverse transcribed using PrimeScript RT Master Mix Perfect Real Time Kit according to the manufacturer's instructions (Takara, Dalian, China). qRT-PCR was performed on an ABI7500 qRT-PCR machine, according to the protocol of SYBR Premix Ex TaqTM II (Takara, Dalian, China). The relative abundance of key genes in the gonad was evaluated using the formula R = 2 − Ct [74]. The reference gene gapdh was used to normalize the expression values. Primer sequences used in this study are listed in Supplementary Table S1.

Immunofluorescence (IF), Terminal Deoxynucleotidyl Transferase-Mediated dUTP Nick End-Labeling (TUNEL) and Immunohistochemistry (IHC)
The Vasa, Cyp19a1a and Sycp3 rabbit polyclonal antibodies were prepared by our laboratory. The dilution and specificity of these antibodies have been analyzed previously [75][76][77]. For IF, Alexa Fluor 488-and 594-conjugated secondary antibodies (Invitrogen, Shanghai, China) were diluted 1:500 in blocking solution and incubated with tissue to detect the primary antibodies. The nuclei were stained by 4 ,6-diamidino-2-phenylindole (DAPI) (Invitrogen, Carlsbad, CA, USA). Apoptosis of germ cells was evaluated by staining paraffin sections of WT, dnmt3aa −/− and dnmt3ab −/− mutant fish at 60 (ovaries) and 120 dah (testes) with in situ cell death detection kit, TMR red TUNEL system (Roche, Mannheim, Germany) according to the manufacturer's protocol. Apoptotic germ cells in the testes (entire median section, n = 5 for each genotype) and ovaries (part 2, n = 5 for each genotype) were counted for statistical analysis. Fluorescence signals were captured by confocal microscopy (Olympus FV3000) (Olympus, Tokyo, Japan). The assessment of global DNA methylation in WT, dnmt3aa −/− and dnmt3ab −/− gonads was performed with immunohistochemistry using the anti-5-mC antibody (MABE146, Merck Millipore) according to previous reports [20,25]. Photographs were taken under an Olympus BX51 light microscope (Olympus, Tokyo, Japan). Finally, the positive signals were quantified using image J software (National Institutes of Health, Bethesda, MD, USA).

Fluorescence In Situ Hybridization (FISH)
The open reading frames of Nile tilapia dnmt3aa and dnmt3ab were amplified with specific primers (Supplementary Table S1), and the amplicons were cloned into pGEM-T Easy Vector. The sense and anti-sense RNA probes were labeled with digoxigenin (DIG) by in vitro transcription using an RNA labeling kit (Roche, Mannheim, Germany). The gonads of fish were sampled at indicated time. The fish gonads were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) and processed for serial paraffin sectioning at 5 µm thickness. The sections were deparaffinized, rehydrated and digested with proteinase K (4 µg/mL; Roche, Mannheim, Germany) at 37 • C for 15 min, followed by hybridization with DIG-labeled RNA probes at 60 • C overnight. The sections were washed with 50% formamide/2 × SSC for 30 min, 2 × SSC for 20 min and 0.2 × SSC for 20 min. The slides were incubated for 30 min at room temperature in a humidified chamber with Anti-DIG-POD (Roche, Mannheim, Germany) diluted by 200 times in DIG2 buffer. Then, the sections were washed with DIG1 buffer. After washing, the TSA Plus Fluorescein System (PerkinElmer, Boston, MA, USA) was used for the amplification of hybridization signals. The nuclei were stained by DAPI (Invitrogen, Carlsbad, CA, USA) staining. Fluorescence signals were captured by confocal microscopy (Olympus FV3000) (Olympus, Tokyo, Japan).

Sperm Mobility Analysis
Sperm concentration, sperm motility (%), curvilinear velocity (VCL), straight line velocity (VSL) and beat frequency of sperm flagella (BCF) were examined by computer assisted sperm analysis using the Sperm Quality Analyzer according to the manufacturer's instructions (Zoneking Software, China). Briefly, semen collected from WT, dnmt3aa −/− and dnmt3ab −/− XY fish (n = 5 for each genotype) at 240 dah was diluted with PBS at 1:10. After sperm activation, one drop of semen was dripped into the counting pool of the sperm counting board, and placed on the operating platform of a Leica DM500 light microscope (Leica, Bensheim, Germany). All parameters were collected according to instrument instructions.

Sperm Papanicolaou Staining and Scanning Electron Microscope Analysis
Semen from WT, dnmt3aa −/− and dnmt3ab −/− fish (n = 5 for each genotype) was collected by in vivo extrusion and then 1 µL of drained semen mixed with 9 µL of double distilled water was applied to clean slides, which were dried naturally and then stained by Papanicolaou staining. Photographs were taken under Olympus BX51 light microscope (Olympus, Tokyo, Japan). To further examine sperm morphology, semen was collected for scanning electron microscope analysis. In brief, sperm specimens were pre-fixed using 2.5% glutaraldehyde, rinsed three times with PBS (pH 7.2) and dehydrated in ascending graded ethanol. Then, the dehydrated samples were put into a drying basket and dried with critical point dryer. The surface of the dried samples was treated with electric conduction, and the specimens were observed under a Zeiss Evo LS10 (Zeiss, Oberkochen, Germany) scanning electron microscope.

Data Analyses
All data are presented as mean ± SD from at least three independent experiments. Different letters above the error bar indicate statistical differences at p < 0.05 as determined by one-way ANOVA followed by Tukey's post hoc test. Statistics analyses were performed using GraphPad Prism 8 software package (GraphPad Software, La Jolla, CA, USA).

Conclusions
In this study, it was demonstrated that both dnmt3aa and dnmt3ab displayed sexually dimorphic expression in developing gonads. Dnmt3aa was highly expressed in oogonia, phase I and II oocytes and granulosa cells in ovaries and spermatogonia and spermatocytes in testes, while dnmt3ab was mainly expressed in ovarian granulosa cells and testicular spermatocytes. The mutation of dnmt3aa resulted in a lower GSI, increased apoptosis of oocytes and spermatocytes and significantly reduced sperm quality, while no obvious phenotype was observed in dnmt3ab homozygous mutants. The level of 5-mC in dnmt3aa −/− mutant testes and ovaries decreased significantly, while there was no difference between dnmt3ab −/− mutants and WT. Our results suggest that dnmt3aa, not dnmt3ab, plays important roles in maintaining normal gametogenesis in teleosts. Our results enrich the understanding of the function of DNA methyltransferase in gonads of non-mammalian vertebrates.