CRISPR/Cas9 knockout of amhy caused male-to-female sex reversal in XY fish

If amhy is the SD gene in threespine stickleback, it should be both necessary and sufficient for male sex determination. To test this, we first knocked out function of amhy in XY fish using CRISPR/Cas9. We designed two sgRNAs targeting exon 1 and exon 3 of amhy (Fig 1). We injected stickleback embryos from lab-reared fish derived from two independent marine populations (Port Gardner Bay and Japan Pacific Ocean) with CRISPR ribonucleoprotein (RNP) made with one of the two sgRNAs. 8/17 (47%) of F0 XY fish injected with RNP exhibited male-to-female sex reversal. For all eight XY sex-reversed fish, we observed ovaries with all cell types seen in wildtype ovaries (S2A Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. CRISPR/Cas9 knockout of amhy caused male-to-female sex reversal in F1 XY fish.

We used CRISPR/Cas9 to target exon 1 (A) or exon 3 (B) of amhy. CRISPR target site and PAM sequence are indicated by orange lines above wildtype reference sequence. Mutations relative to wildtype are bolded. We identified four loss of function amhy alleles in F1 progeny. One allele (-2282, + 8) deleted the first six exons of amhy. Other alleles featured small frameshift indels. C) We identified male-to-female sex reversal using gonad morphology and histology. Ovaries in XX wildtype and XY amhy-KO fish feature oocytes throughout multiple stages of development. P – primary oocyte, C – cortical alveolus stage, V – vitellogenic stage, M – Mature oocyte, G – Granulosa cells, T – Theca cells. Scale bars = 200 μm.

https://doi.org/10.1371/journal.pgen.1011932.g001

F0 crispants are often mosaic for mutations induced by CRISPR/Cas9 gene editing (S2B Fig), which can cause variation in observed phenotypes. To generate non-mosaic F1 gene-edited fish, we used two F0 XY males that had heritable germline mutations in amhy but were not sex reversed, indicating they likely did not have amhy mutations in cell-lineages important for sex determination. Sperm from these fish were used to fertilize eggs from wildtype XX females. F1 offspring from these crosses inherited frameshift mutations in amhy, identified by Sanger sequencing (S3 Fig). We analyzed five F1 males from one F0 crispant with mutations in exon 1 and 21 F1 males from one F0 crispant with mutations in exon 3. All 26 F1 fish exhibited complete male-to-female sex reversal and developed ovaries morphologically and histologically indistinguishable from wildtype XX females (Fig 1C). No F1 XY fish showed development of testis or ambiguous gonad tissue with any visible spermatocyte production. We identified three loss of function alleles in exon 1 of amhy (Fig 1A) and one in exon 3 (Fig 1B). In ten F1 fish, the region of amhy in exon 3 targeted by CRISPR/Cas9 could not be amplified. PCR of exons 1 and 7 showed the presence of these distal portions of amhy, but PCR primers spanning a large putative deletion all failed to amplify, suggesting a chromosomal rearrangement disrupted amhy in these fish. As a specific mutation could not be positively identified, these samples were excluded from further analysis and final counts.

We analyzed the expression of canonical male differentiation genes amh and dmrt1 and female differentiation genes foxl2 and cyp19a1a using in situ hybridization on wildtype and F1 XY amhy-KO ovaries (Figs 2 and S4). We also analyzed gsdf as a positive control, as this gene is expected to be expressed in somatic gonadal cells of both sexes [65]. Both wildtype and sex-reversed ovaries showed expression of foxl2 and cyp19a1a in somatic cells. Expression of amh and dmrt1 was minimal and restricted to oocytes when present as has been observed in some species of fish [21,66–69].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Gene expression in sex-reversed gonads resembles expression in wild type gonads.

In situ hybridization of gonadal genes in wildtype and sex-reversed stickleback gonads. A positive control gene, gsdf, was expressed in somatic cells of both testes and ovaries. Wildtype XY and sex-reversed XX testes show strong expression of the male differentiation genes amh and dmrt1 but not the female differentiation genes foxl2 and cyp19a1a. Wildtype XX and sex-reversed XY ovaries strongly express female differentiation genes but not male differentiation genes. Scale bars = 200 μm.

https://doi.org/10.1371/journal.pgen.1011932.g002

Knockout of amh or amhrII in teleost species without amh-based sex determination has caused male-to-female sex reversal, including in lab strains of zebrafish which lack sex chromosomes (reviewed in [70]) and medaka which have an XX-XY system with a dmrt1 derived SD gene [32]. Thus, it is possible that off-target mutations of the paralogous amh could contribute to the sex reversal observed in F1 amhy mutants. Both sgRNAs have multiple mismatches to their homologous region in amh, and we verified that all F1 XY amhy-KO females did not have mutations within amh by Sanger sequencing (S5 Fig). This demonstrates that the sex reversal phenotype was not caused by mutations of amh. We also investigated whether the sex reversal phenotype was caused by a mutation at an unknown autosomal locus. Such an autosomal mutation must be dominant since we observed male-to-female sex reversal in heterozygous XY F1 fish. F1 XX siblings of sex-reversed XY fish would also be heterozygous for this mutation, and half of their F2 XY progeny would be sex reversed. We crossed two F1 XX females with wildtype XY males and phenotyped F2 offspring for sex reversal. All 16 F2 XY fish were phenotypically male, indicating that sex reversal in F1 XY amhy-KO fish was not caused by an off-target autosomal mutation.

Introduction of an amhy transgene caused female-to-male sex reversal in XX fish

To determine if amhy is sufficient for male sex determination, we created a construct (pT2-5amhy2.5,Xla.Ef1a:EGFP) containing an 11 kb region of the Y chromosome which includes amhy, 5 kb of upstream sequence, and 2.5 kb of downstream sequence (S6 Fig). By including the surrounding non-coding sequences, we aimed to recapitulate the endogenous expression patterns of amhy. An Xla.Ef1a:EGFP cassette downstream of amhy was included to facilitate the screening of injected embryos for transgene integration (Fig 3C). These components were flanked by Tol2 transposon sequences to facilitate random integration into the genome by the Tol2 transposase [71,72]. 16 F0 XX fish showed positive GFP expression and successful transgene integration, detected through PCR genotyping (Fig 3D). At reproductive age, all 16 F0 XX Tg(amhy,EGFP) were female-to-male sex reversed, developing testes rather than ovaries (Fig 3A). All 10 control F0 XX fish injected with the control Xla.ef1a:EGFP transgene showed positive presence of the transgene but did not show any signs of sex reversal and developed cytologically normal ovaries (Fig 3B). We also analyzed the expression of amh, dmrt1, foxl2, cyp19a1a, and gsdf using in situ hybridization on wildtype and F0 XX Tg(amhy,EGFP) testes (Figs 2 and S7). Both wildtype and sex-reversed testes showed strong expression of amh and dmrt1 in somatic cells and no discernible expression of foxl2 or cyp19a1a.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. XX fish with an amhy transgene exhibit female-to-male sex reversal.

A) XX Tg(-5amhy3.5,Xla.Ef1a:EGFP) fish developed testes with varying cell composition. Fertile fish had testes with clearly defined cysts of mature sperm as is found in wildtype fish. The subfertile testis had meiotic spermatocytes and immature round spermatids while hypotrophic testes had only undifferentiated spermatogonia. G – spermatogonia, P – primary spermatocyte, S – secondary spermatocyte, R – round spermatid, Z – Spermatozoa Scale bars = 100 μm. B) Control XX Tg(Xla.Ef1a:EGFP) fish developed ovaries identical to wildtype XX counterparts. Scale bars = 100 μm. C) Embryos injected with pT2-5amhy3.5,Xla.Ef1a:EGFP seven days after fertilization under bright field (top) and EGFP excitation (bottom). Embryos positive for transgene integration show somatic EGFP expression. We removed embryos with no EGFP expression or EGFP expression only in the yolk, which indicates the yolk was injected rather than the blastomere [41]. D) We used PCR of the sex-linked marker idh to genotype XX and XY fish. PCR of EGFP and amhy show the presence of the associated transgene. 1 – wildtype XX, 2 – wildtype XY, 3 – XX Tg(EGFP), 4 – XX Tg(amhy,EGFP).

https://doi.org/10.1371/journal.pgen.1011932.g003

We attempted to constitutively express amhy using a synthetic exon-only coding sequence with an N-terminus FLAG tag joined to a 3’ EGFP by a P2A self-cleaving peptide [73,74]. Plasmids were made using the D. rerio ubiquitin promoter [75] or the X. laevis ef1a promoter [76] to drive expression of the amhy,EGFP construct (S6 Fig). Transgenes with the respective promoter driving expression of only EGFP were used as controls. Somatic EGFP expression was observed with both control transgenes indicating these commonly used vertebrate constitutive promoters can drive expression in stickleback. Neither the Dre.Ubi:amhy,EGFP nor the Xla.Ef1a:amhy,EGFP transgene produced any larvae with visible EGFP expression, indicating that the amhy transgenes were failing to integrate or be actively expressed in any embryos. Amh has been overexpressed successfully in mouse [77], chicken [78], and Nile tilapia [16], and the P2A peptide has been successfully used in stickleback overexpression constructs [79], indicating that overexpression alone or the presence of a P2A peptide was likely not responsible for the failure of this construct. Introns of amhy may contain essential functional elements needed for gene expression of amhy in threespine stickleback. The presence of introns has been shown to impact gene expression, even under the regulation of non-native constitutive promoters [80,81].

Sex-reversed amhy mutants produce viable gametes but are subfertile

Female-to-male sex-reversed F0 XX fish had varied fertility. Four of the 16 XX Tg(amhy,EGFP) males had testes morphologically resembling those of a fertile, wildtype male. 12 XX males had abnormally small and underdeveloped testes. We performed crosses with all four males with mature testes with wildtype XX females. We identified viable, successfully fertilized embryos by the presence of a gastrula and active epiboly 12–24 hours post fertilization (hpf). One male was extremely subfertile and only 3% of embryos were viable (n = 911). The remaining three males showed normal viability rates (80% n = 97; 94% n = 89, 98% n = 235), similar to viability rates of three wildtype males (73% n = 314, 87% n = 252, 97% n = 291) (S1 Table). One fertile male also produced offspring with somatic EGFP expression at a very low frequency (7%, n = 231), indicating a heritable germline integration of the transgene containing the constitutive Ef1a:EGFP cassette had occurred. Histology of F0 XX Tg(amhy,EGFP) males showed key differences between fertile, subfertile, and hypotrophic testes (Fig 3A). Fertile males had normal testis histology with visible mature spermatozoa. The subfertile male testis lacked spermatozoa, instead containing meiotic spermatocytes and immature round spermatids concordant with its reduced fertility. Histology of the hypotrophic testes showed the germ cells were still early undifferentiated spermatogonia. This phenotypic variation may be a result of variable amhy expression influenced by transgene copy number and insertion location.

Sex-reversed XY amhy-KO females produced mature oocytes and became gravid, but they were subfertile compared to wildtype females. To assess fertility, we crossed three F1 XY amhy-KO females to wildtype XY males. These clutches all showed extremely low viability rates (2% n = 43, 4% n = 51, and 6% n = 63) (S2 Table). One of the three F1 fish was crossed four additional times and showed improvement in viability with the subsequent crosses. The viability rates of the fourth and fifth cross were 78% (n = 100) and 79% (n = 89) respectively, though this is still lower than the average viability of eggs from nine wildtype females (85% n = 857) (S1 and S2 Tables). Three days after fertilization, 50 of the 148 viable F2 eggs had arrested development. We were able to genotype 48 of the failed embryos. 36 of the 48 arrested embryos were YY and carried both the paternally transmitted Y chromosome (Y⁺) and the maternally transmitted amhy-KO Y chromosome (Y^KO) (Fig 4D). The remaining embryos were reared to adulthood and 92 were genotyped for their sex chromosome complement. No adult Y⁺Y^KO fish were observed, indicating that the YY genotype is embryonic lethal due to the missing X chromosome (Fig 4A and 4B). XY^KO females were found to transmit the X and Y^KO chromosome with equal frequency, indicating there was not any biased transmission of the X or Y chromosome during female meiosis (36 Y⁺Y^KO embryos out of 148 total does not deviate from a 1:3 Mendelian ratio; chi-squared = 0.0360; p = 0.8494; of the 92 adult F2s genotyped 36 XX, 27 XY⁺, and 29 XY^KO does not deviate from the expected 1:1:1 ratio; chi-squared = 1.457; p = 0.4827). We also measured fertility in an F2 cross which contained sex-reversed XY^KO females and wildtype XX female siblings. We crossed 11 F2 females (5 XX and 6 XY^KO) with wildtype XY males. The clutch size did not differ significantly between XY^KO and XX females (Fig 4C and S3 Table). However, the viability rate was substantially lower in the F2 XY^KO females, similar to the F1 generation (Fig 4C and S3 Table). Two F2 females were also eggbound and were unable to be stripped of eggs. Both eggbound females were XY^KO, and no eggbound XX females were observed at the time crosses were performed.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. XY amhy-KO females can reproduce and transmit an X or Y chromosome.

YY embryos (A) at 48 hours post fertilization show arrested development compared to wildtype embryos (B). Arrowheads denote anterior end of body axis. C) F2 XX and XY amhy-KO female siblings do not show a significant difference in clutch size (Wilcoxon test, p = 0.54). Clutches from XY amhy-KO females have a significantly lower viability rate than those from XX female siblings (Wald z-test of beta regression coefficients, z = -2.405, p = 0.0162). D) PCR of the sex-linked marker IDH shows the absence of the X chromosome in YY embryos. Both the paternal (wildtype) and maternal (+9, + 5) loss of function alleles can be detected in XY offspring. YY embryos show the presence of both amhy alleles as well as a third band, likely a heterodimer of the maternal and paternal amplicons.

https://doi.org/10.1371/journal.pgen.1011932.g004

Male nuptial coloration is present in both sex-reversed genotypes

Male nuptial coloration is one of the most prominent sexually dimorphic characteristics in threespine stickleback. During their breeding season, males from many stickleback populations develop red throats and dark blue eyes and bodies [82]. To investigate the separate contributions of gonadal and genotypic sex on this trait, we compared coloration in mature sex-reversed XX Tg(amhy,EGFP) males and XY amhy-KO females to wildtype XX and XY fish (Fig 5). Neither XX Tg(amhy,EGFP) males nor XY amhy-KO females appeared to fully recapitulate wildtype XY male coloration, but both were distinct from wildtype XX females (Fig 5A). XX males had blue coloration on their eyes and bodies, but none were as dark as completely colored XY males. XY females were varied in appearance, but some had discernibly blue eyes and darkened bodies not seen in XX females. We measured the proportion of body area that was blue or dark for each genotype (S8 Fig and S4 Table). As red throats are not prominent even in wildtype males under our laboratory conditions, we were not able to quantify this trait. Both proportion of dark and blue body area showed wide variation within genotypes (Fig 5B). We saw significantly more blue (p < 0.0001) but not dark coloration (p = 0.1161) in XY males compared to XX females. XX males were significantly less blue (p = 0.0003) and dark (p = 0.0007) than XY males. XY females were significantly more blue than XX females (p = 0.0348), but there was no difference in darkness (p = 0.1074). Together, these findings show that male nuptial coloration is influenced both by hormones produced by the testis and by sex chromosome genotype, but both male gonad differentiation and an XY genotype is needed to present full male nuptial coloration.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Both male and female sex-reversed stickleback show partial male nuptial coloration.

A) Representative images of wildtype XX females and XY males and sex-reversed XX Tg(amhy,EGFP) males and XY amhy-KO females. XX females are silver and may darken slightly when gravid. XY males have dark blue eyes and bodies and red throats. Sex-reversed XX males and XY females show intermediate aspects of male nuptial coloration. B) Proportion of body area that is blue or dark for each genotype. Differences in coloration between genotypes were tested using estimated marginal means from a beta regression with Wald z-tests and Tukey adjustment for pairwise comparisons (* p < 0.05, ** p < 0.01, *** p < 0.001) (S4 Table).

https://doi.org/10.1371/journal.pgen.1011932.g005

amhy contributes to higher total amh dosage early in development

Since Amh shows high specificity for its dedicated receptor, AmhrII, [28–30] we predicted that amhy would be functionally constrained regarding coding sequence evolution and would instead evolve novel function through regulatory changes. We investigated the degree of sequence conservation within putative regulatory regions (2 kb upstream and downstream of the coding sequence), of amhy relative to autosomal amh. We also compared sequence similarity between amh and the orthologous amh in the ninespine stickleback (Pungitius pungitius) which do not have an amhy (Fig 6 and S5 Table). The divergence time between threespine and ninespine stickleback is approximately 26 mya [83], similar to the 22 mya divergence time between the threespine amh and amhy [20]. There was less sequence identity between the coding sequences of threespine amh and amhy (0.780) than between threespine amh and ninespine amh (0.879). Lower sequence identity despite shorter divergence time is consistent with weaker purifying selection acting on the Y chromosome due to a lack of recombination [84]. This was also supported by a higher overall dN/dS value between threespine amh and amhy (0.780) than between threespine amh and ninespine amh (0.668), though this pattern did vary across exons (S5 Table). Alternatively, reduced sequence identity and elevated dN/dS could be due to positive selection acting on coding sequence changes to amhy. The sequence identity of upstream and downstream regions was much lower between threespine amh and amhy (0.365 and 0.376) than between threespine amh and ninespine amh (0.595 and 0.660), indicating amhy has undergone significant regulatory evolution since its duplication onto the Y chromosome, which may cause differences between amh and amhy expression.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. amh and amhy show no sequence conservation outside of the coding sequence.

Vistaplots showing conserved sequence identity of threespine amh to threespine amhy (top) and to ninespine amh (bottom) with sequence identity on the y axis and nucleotide position of threespine amh on the x axis. Exon positions are indicated along the x axis in blue. Conserved exon and noncoding sequences share 70% sequence identity in a 100 bp window.

https://doi.org/10.1371/journal.pgen.1011932.g006

To determine whether expression of these genes differ throughout development, we sequenced the total transcriptomes from prehatching embryonic stages 17 (early somitogenesis, ~ 43 hpf), 20 (onset of pigmentation, ~ 73 hpf), and 23 (completion of yolk sac circulation, ~ 122 hpf), stage 26 larvae (0 days post hatching (dph), around 200 hpf), as well as larvae 4 and 8 dph [85]. The earliest identified difference between male and female threespine stickleback is an increase in primordial germ cell count in females relative to males at 3–4 dph, followed by differences in germ cell morphology at 7 dph [86]. Sex determination should begin sometime prior to this stage. We predicted that the unique regulatory architecture of amhy resulted in earlier expression timing relative to amh. Thus, we predicted amhy would be expressed at higher levels relative to amh early in development in order to initiate sex determination. As the gonad differentiates, amh expression should increase relative to amhy due to the canonical role of amh in male differentiation. This pattern has been observed in Japanese flounder [15] and Patagonian pejerrey [13], which also have amhy SD genes. To compare amh and amhy expression, we first scaled amhy read counts by the difference in transcript length which is not accounted for by normalized read counts (S6 Table). We then tested whether amhy expression was different from one half of autosomal amh, as males have two copies of amh and one copy of amhy. Average expression of amhy was higher than one half amh at stage 23 and all timepoints post hatching, but this difference was only significant at stage 23 (p = 0.0454). We found amhy expression trended towards higher expression at later developmental stages, consistent with a role in sex determination. However, this difference was only significant between stage 20 and 4 dph (p = 0.0398, Tukey HSD = 0.0361). We then hypothesized that male sex determination could be initiated by a higher overall dose of amhy when expressed simultaneously with amh. We combined amh and transcript length adjusted amhy normalized read counts at each stage and found total amh dosage was higher in males at 0 dph (p = 0.0448), 4 dph (p = 0.0347), and 8 dph (p = 0.0297) (S6 Table), across the time where germ cell count diverges in males and females. Together, these results suggest that increased total dosage of amh provided by amhy is key to its role in male sex determination.

Sex differentiation genes are not differentially expressed between males and females early in development

Following sex determination, we expected to see divergence in the expression of overall male and female transcriptomes. However, principal component analysis of autosomal gene expression indicated that male and female larvae had similar expression profiles. Autosomal genes were clustered by developmental time rather than separating by sex at any stage of development (Fig 7A). Only sex chromosomes showed clear separation by both sex and developmental time (S9 Fig), since the X chromosome is expressed two-fold higher in females, relative to males, and the Y chromosome is only expressed in males. The overall lack of sex-specific expression on the autosomes suggests that many aspects of sex differentiation occur well after the earliest noted phenotypic difference between males and females: the proliferation of germ cells in the primordial gonad at 3–4 dph [86].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Expression of gonadal growth and differentiation genes in embryonic and larval stickleback.

(A) PCA plots of normalized read counts for all genes cluster by stage rather than sex. (B) and (C) Normalized read counts of genes in embryonic stages 17, 20, 23, and larvae 0, 4, and 8 days after hatching. Stages with significant differences between males and females are marked with an asterisk and bolded (p < 0.05; Tables 1 and S6). The orange arrow (B) indicates stages where differences in germ counts have been observed between males and females.

https://doi.org/10.1371/journal.pgen.1011932.g007

Download:

• PNG
  larger image
• TIFF
  original image

Table 1.

Benjamini Hochberg adjusted P values for differential expression between males and females.

https://doi.org/10.1371/journal.pgen.1011932.t001

We examined whether any developmental stages had genes with known functions in sex differentiation that were significantly differentially expressed between males and females. Many male and female differentiation genes are highly conserved across vertebrates [2] and should also be expressed during development of the threespine stickleback. For instance, we expected differential expression of key transcription factors canonically involved in differentiation, including the male biasing factors dmrt1 and sox9b or the female biasing factors foxl2a, followed by divergence of downstream genes such as the aromatase enzyme cyp19a1a and the TGF-b hormones amh and gsdf. In total, we found 455 differentially expressed genes (DEGs) between males and females on the autosomes (S1 File). Key regulators of differentiation should be differentially expressed throughout development and into adulthood in order to actively maintain gonadal fate [2]. However, 430 of the DEGs were only differentially expressed in a single stage (S1 File), suggesting these genes do not play a substantial, ongoing role in sex differentiation. Additionally, while we expected to see an increase in the number of DEGs over time due to activation of male and female specific gene networks, the number of DEGs was highest at stage 17 (156 DEGs), decreasing in later stages until a small increase from 4 dph (41 DEGs) to 8 dph (65 DEGs) (S10 Fig). None of the canonical differentiation genes of interest were differentially expressed between males and females at any timepoint examined in this study (Fig 7 and Table 1). Male differentiation genes dmrt1 and amh and female differentiation genes foxl2 and cyp19a1a were differentially expressed in adult gonads of both wildtype and sex-reversed fish shown by in situ hybridization (Figs 2, S4, and S7), indicating their function in differentiation are conserved in stickleback. Of the identified DEGs, only a small handful have known associations with gonadal function, development, or differentiation. Two genes associated with germ cell specification and maintenance [87], ddx4 and piwil1, showed female biased expression at 8 dph (Fig 7C). Interestingly, amhrII, which shows female biased expression in embryonic mouse gonads [88], was also expressed more highly in females at 8 dph. Although cyp19a1a was not differentially expressed, the steroidogenic enzyme hsd17b1, which plays a role in estrogen synthesis [89], showed female biased expression at 8 dph. This gene acts as the SD gene in Seriola fishes [90] and is upregulated in ovaries of Siberian sturgeon before the activation of foxl2 or cyp19a1a [91]. A lack of differential expression in many of these key genes indicates that gonad differentiation occurs after 8 dph.

We identified some other DEGs that have known reproductive functions in other vertebrates but do not have clear connections to sex differentiation and are only differentially expressed at a single timepoint. Gonadotropin-releasing hormone II receptor (LOC120811940) has known roles in the stimulating steroidogenesis in both testes and ovaries across mammals [92] and had male biased expression at stage 17 (logFC = 2.74, P = 1.6E-3). TGF-beta 3 hormone is expressed more highly in male mice during gonadogenesis [93], and a TGF-beta-3 proprotein-like gene (LOC120807995) was expressed slightly higher in females at stage 17 (logFC = -0.63, P = 0.042). Piwil2, another germ cell associated gene [87], was expressed higher in males at stage 20 (logFC = 1.71, P = 4.0E-3). Cdc25b is expressed at low levels in mouse ovaries and testes [94] and was male biased at stage 23 (logFC = 0.52, P = 0.012). Separin/separase (espl1) has known roles in mitotic and meiotic chromosome segregation [95] and was expressed at higher levels in 8 dph females (logFC = -0.56, P = 1.4E-4). Additional work will be necessary to understand what function, if any, these genes have in threespine stickleback gonad development.

amhy is the sex determination gene in threespine stickleback fish

Our findings demonstrate that amhy is the SD gene in threespine stickleback, as it is both necessary and sufficient for sex determination (Table 2). Independently derived amhy genes have been functionally demonstrated to have a role in sex determination in three other clades of teleost; however, threespine stickleback offers a novel perspective as their amhy is more diverged, and genetic sex determination is more canalized in this species. Nile tilapia and Japanese flounder have only one and nine SNPs, respectively, that differentiate their X- and Y-linked copies of amh [15,16], and amhy in Patagonian pejerry shares 94% nucleotide identity with its autosomal paralog [13]. In all three of these species, genetic sex determination can be overridden to varying degrees by temperature [96–98]. In contrast, the threespine stickleback, amh and amhy only share 78% nucleotide identity (S5 Table). There is no evidence that temperature can disrupt sex determination in threespine stickleback, and occurrences of sex-reversed or intersex stickleback in wild populations is rare [99,100]. This indicates that genetic sex determination is highly canalized in threespine stickleback and less susceptible to environmental signals than many other teleost species, making stickleback a robust model to investigate how amh regulates gonadal development. Comparative studies across these taxa will be invaluable for understanding how amh repeatedly evolves this novel function, and the presence of an independently derived male associated amh duplication in the closely related brook stickleback [22,40] makes the stickleback family an especially powerful tool for studying the convergent evolution of amh as an SD gene.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Phenotypes of mutant stickleback from this study.

https://doi.org/10.1371/journal.pgen.1011932.t002

Potential mechanisms of sex determination in stickleback

Germ cell density has been shown to affect sex determination in several teleost species, with the male fate associated with lower germ cell count and the female fate with a higher count. Common lab zebrafish strains are undifferentiated gonochorists, where all individuals begin ovary development, but the gonad transitions to testis development in some individuals [101]. Complete ablation of germ cells causes zebrafish to undergo male development [102], and a threshold of germ cells at the time of gonadal differentiation is necessary to sustain ovary development [103]. O. latipes, like sticklebacks, are differentiated gonochorists where either testis or ovary development begins from a common undifferentiated primordium [104]. Ablation of germ cells in XX O. latipes causes female-to-male sex reversal [105], and overproliferation of germ cells in XY fish causes male-to-female sex reversal [106]. The association of germ cells with sex has also been observed outside of teleosts in the red eared slider turtle (Trachemys scripta elegans) which has temperature dependent sex determination. Embryos incubated at female promoting temperatures have higher germ cell counts than those incubated at male promoting temperatures, and depleting germ cells at intermediate temperatures biased the sex ratio towards males [107]. Even after primary sex determination, germ cell depletion in mice [77] and rats [108] leads to masculinization of the ovary, which produces Sertoli-like cells and seminiferous tubule-like structures before its eventual degradation. This suggests the presence of a widely conserved, ancestral mechanism in vertebrates by which germ cell density governs gonad identity. A new duplication of amh could increase overall amh dosage, reducing germ cell proliferation to co-opt this underlying mechanism and become a male determining gene.

Consistent with this mechanism, we found evidence that amhy contributes to higher amh dosage in post hatching stages where differences in germ cell proliferation between males and females are first observable in the developing gonad [86]. We also found increased expression of germ cell specific genes (e.g., ddx4 and piwil1) in females at 8 dph, corroborating previous histological observations of increased germ cell proliferation in female stickleback gonads beginning at 3 dph [86]. Overexpression of Amh in mice [77,109] causes reduced germ cell count and gonad size in males and females, and loss of Amh alters primordial follicle recruitment in females [110]. Similar findings have also been observed in chicken [78,111]. Studies misexpressing amh in teleosts have primarily focused on sex determination; however, gonadal hypertrophy observed in amh knockouts in other teleosts [32,34] alongside gonadal hypotrophy with a lack of mature germ cells observed in some XX Tg(amhy,EGFP) males in this study demonstrate the conserved function of amh suppressing germ cell proliferation in stickleback and other teleosts.

Somatic cells may also be affected by amhy in other ways. In mice, low concentrations of Amh promote proliferation and growth factor production in somatic cells, but increased concentrations induce apoptosis [112]. Altering somatic cell proliferation could lead to differences in the levels of hormones and growth factors that they produce, shaping gonadal development. While Amh is not necessary for initial gonadal differentiation, it is important to maintaining Sertoli cell fate in fetal mice [113]. Estrogen production in somatic cells also plays a major role in sex differentiation and ovary development in teleosts (reviewed in [114,115]). In tilapia, Amh signaling directly inhibits cyp19a1a production via Smad transcription factors [116]. While we did not detect a difference in cyp19a1a expression between males and females in embryos or larvae, it is expressed more highly in adult ovaries than in testes, so amhy could play a reinforcing role later in development to suppress estrogen synthesis in males.

The evolutionary changes that have given rise to novel function of amhy in stickleback sex determination remain unclear. While we do see substantial sequence divergence in the putative regulatory regions of amhy relative to amh, we only detected a significant difference in expression at a single developmental stage. As we are unable to isolate embryonic gonadal tissue for expression analysis, our results in this study are limited to characterizing total expression in the whole embryo. This likely reduces our ability to detect differences in the expression of genes like amh and amhy which are restricted to a small population of cells. Alternative approaches, like single cell RNA-seq, spatial transcriptomics, or HCR in situ will be useful to survey gene expression in the early differentiating gonad. Still, we should not rule out the significance of coding sequence evolution for amhy. In Nile tilapia, amhy expression is lower than the X-linked amh throughout development, and misexpression of amhy but not amh induces male sex determination [16]. This implies that the single missense SNP between amh and amhy in tilapia is important to the function of amhy in sex determination. In Spotted knifejaw (Oplegnathus punctatus), a Y-linked copy of amhrII shows higher affinity for binding amh than its X-linked paralog [23]. In zebrafish which lack amhrII, amh may instead use the type II TGF-β receptor bmpr2a [117,118]. Mutations that increase the stability of Amhy, increase its affinity for the receptor, AmhrII, or promote interaction with a non-canonical receptor could also affect gonadal development. Manipulating amh and amhrII could provide further insight into the evolution of amhy and its mechanism of inducing male development. We would predict knockout of amh or amhrII to result in overproliferation of germ cells and some level of male-to-female sex reversal in XY fish as is seen in medaka amhrII mutants [32]. This would demonstrate that germ cell proliferation regulated by the canonical amh signaling pathway can influence sex in stickleback, making protein evolution or interaction with non-canonical receptors less parsimonious avenues by which amhy has evolved its novel function. Misexpression of amh and amhy using regulatory elements from the opposite gene (i.e., expressing the amh genic sequence with amhy regulatory elements) or non-native promoters could also elucidate whether coding sequence evolution is important to the function of amhy.

Sex differentiation occurs later than anticipated

We observed no differential expression in canonical vertebrate sex differentiation genes, including dmrt1, sox9b, foxl2a, and cyp19a1a. While the order and timing of expression varies across taxa, these core genes are reliably differentially expressed between males and females before the onset of morphological gonadal differentiation [119–121]. These genes show expected dimorphic expression patterns in adult gonads, shown both by our in situ data as well as gonadal transcriptomic data [122]. This indicates that there is some delay between the initial establishment of germ cell differences between males and females and the later differentiation of the primordial gonad into either a testis or ovary. Expression analysis of whole embryos and larvae could obscure gonad specific changes in gene expression as previously discussed. However, we did identify differential expression of ddx4 and piwi1l which are highly germ cell specific [87] as well as hsd17b1 and amhrII which are not gonad specific [89,123] but show early female biased gonadal expression in other species [88,91]. This demonstrates some ability to detect sex differences in expression for significant genes. Female biased expression of hsd17b1 precedes differential expression of these canonical differentiation genes in sturgeon, where hsd17b1 is thought to increase estrogen, upregulating of foxl2 and cyp19a1a in females [91]. In stickleback, female biased hsd17b1 expression may be a key difference between males and females, initiating expression changes in canonical differentiation genes. Further analysis of gene expression at later timepoints will be necessary to determine if this is the case.

We propose that sex determination in threespine stickleback occurs in two phases. In the early critical period of sex determination, germ cell proliferation in the undifferentiated gonad diverges between males and females due to the presence of amhy in males. Then, during the sensitive period of gonadal differentiation, the difference in germ cell count between males and females dictates gonadal fate, marking the end of the sex determination and the beginning of differentiation. Unlike Sry in mammals which is expressed in a brief window and directly initiates gonadal differentiation [124], amhy in stickleback appears to be expressed throughout this early period of germ cell proliferation. These two phases of germ cell proliferation followed by gonad differentiation are also present in other species with SD genes including Nile tilapia (reviewed in [125]), and medaka (reviewed in [104]) as well as species without SD genes such as red eared slider [107] where germ cell proliferation is affected by temperature. This suggests that this two-phase system of differential germ cell proliferation followed by gonadal differentiation is broadly conserved across vertebrates, but what combination of environmental and/or genetic factors influences germ cell proliferation has changed between taxa. The ability of amh to immediately integrate with this preexisting sex determination network through its existing function in regulating germ cell proliferation along with possible reinforcing roles in regulating somatic cell identity and estrogen synthesis would explain the prevalence of amh as an SD gene.

Fertility of sex-reversed threespine stickleback

Both XY amhy-KO females and XX Tg(amhy,EGFP) male sex-reversed fish were fertile in our study. Fertility of sex-reversed mutants is not uncommon in other teleost species; however, it is often observed in species with younger or less diverged sex chromosomes. In these instances, the X and Y chromosomes contain a similar complement of genes. In medaka, the Y specific region is 280 kb and the only functional gene it contains is the sex determining dmrt1by [12]. In Nile Tilapia, the sex determining region on linkage group 23 is 1.5 Mb and contains 51 annotated genes, including the sex determining amhy [126]. In both species, sex-reversed fish are fertile and YY offspring are viable [11,127]. In contrast, the threespine stickleback Y chromosome has a ~ 15 Mb non-recombining region [20]. Within this region, the Y chromosome has lost half of the almost 1200 ancestral gametologs, causing inviability of YY embryos, and has accumulated many Y-specific genes not present on the X chromosome [20]. Sex chromosome-specific gene content evolution could lead to improper expression in the opposite sex, affecting gametogenesis. Given this divergence, the ability of both XY amhy-KO females and XX Tg(amhy,EGFP) males to reproduce is surprising. In mouse, reduced fertility of XY females is a result of meiotic errors between the X and Y chromosomes as well as the effect of aberrant Y-chromosome genes in the ooplasm on early development (Reviewed in [128]). As clutch size in XY females did not differ from XX siblings and some crosses from XY females have normal viability, it seems unlikely that widespread meiotic failure occurs in XY female stickleback. Furthermore, it is clear that Y chromosome transcripts or dosage differences from X chromosome genes do not prevent embryo development. Threespine stickleback are batch spawners that produce multiple clutches of eggs over a single breeding season [82]. If ovulated eggs are not passed, they can become overripe and inviable [129,130]. Overripe eggs can also harden and obstruct the fish, causing them to become eggbound. Many of the clutches from XY amhy-KO females had this hardened, overripe appearance that is uncommon in our lab-reared, wildtype fish. In addition, some XY females we attempted to cross had no cloaca dilation and were eggbound. We suspect that subfertility of XY females is not due to an inability to produce viable eggs, but due to an inability to pass these eggs once mature. Whereas wildtype fish can pass eggs on their own and begin another cycle of ovulation, XY females retain these eggs until they are manually removed. Repeated and frequent crossing would prevent overripe eggs from building up and allow the harvest of freshly ovulated, viable eggs. Hormones produced by the gonad and pituitary gland appear to play a role in egg maturation and overripening for stickleback [130,131], so while XY females are histologically similar to XX females, variation in hormone production could inhibit proper production and release of eggs.

The mammalian Y chromosome has several genes essential for spermatogenesis and sperm development [132–136]. Single cell sequencing has shown that many Y-specific genes are expressed throughout spermatogenesis in stickleback [137], but these genes are apparently not essential for this process. Several XX Tg(amhy,EGFP) males from this study had testes with many fully mature spermatozoa that were able to fertilize eggs and generate viable progeny at similar rates to wildtype sperm. This raises a question of the importance of the Y chromosome to fitness and reproduction. Testes from fertile XX males did not show an obvious reduction in sperm number indicating that widespread meiotic arrest does not occur in XX males. Stickleback are external fertilizers in which sperm competition frequently occurs [138]. Variation in sperm morphology [139] and motility [140] can contribute to reproductive success, so there may be differences in sperm function that our ex vivo fertilization assays do not capture. Some XX Tg(amhy,EGFP) males did have hypotrophic testes with a reduction in identifiable germ cells, but since this phenotype was not universal, we suspect that variation in transgene copy number and insertion location led to overexpression of amhy in these fish. The germline Tg(amhy,EGFP) offspring we produced will allow us to determine if this phenotypic variation exists in a consistent genetic background.

A four core genotypes stickleback model

Sex chromosomes impact a wide range of traits outside of fertility that are still critical to fitness. By comparing XY males and XX females to sex-reversed XY females and XX males, the Four Core Genotypes mouse model has allowed researchers to dissect the role of genotypic and gonadal sex to a number of behavioral and physiological sex differences in mammals (reviewed in [141,142]) including mating behavior [143] and parental care [144]. The creation of reproductively mature XX males and XY females in this study offers a new vertebrate model to study the separate contributions of gonadal sex and sex chromosomes to sexual dimorphism.

Many traits are sexually dimorphic in the threespine stickleback, including shape and size of the head and body [43–46], bony armor features [43,44], and brain size [47,48] as well as feeding behaviors [49,50] and boldness [51]. Male stickleback show complex courtship and mating behavior [52,53] and male-only parental care [54], and females show a variety of mate choice preferences [55–60]. Male nuptial coloration is a particularly interesting trait, as strong coloration is beneficial for courtship [60,61], but has a cost of increased risk of predation [62,63]. Therefore, sexually antagonistic selection would select for stronger, more conspicuous coloration in males and inconspicuous coloration in females. Androgens, secreted from the gonads, are shown to influence nuptial coloration in stickleback [64]. These observations are supported by the presence of nuptial coloration in XX males in our study, which lack all Y-linked genes, but have functioning testes. In contrast, male nuptial coloration in XY females that lack testes shows that there are also sex-linked genetic factors that contribute to nuptial coloration independent of gonadal sex. Future research with the Four Core Genotypes stickleback will provide many insights into the genetic basis of sexual dimorphism and how sexually antagonistic selection shapes the evolution of sex chromosomes.

Ethics statement

All animal procedures were approved by the University of Georgia Animal Care and Use Committee (protocols A2021 07-031-Y3 and A2024 08-009-A12).

Animal husbandry

All experiments used lab-derived progeny from wild-caught threespine stickleback fish from Port Gardner Bay (Washington, USA) and Japan Pacific Ocean (Akkeshi, Japan). We collected fish from Washington State over multiple years using Scientific Collection Permit numbers 21-122, 22-150, 23-147, and 24-143. All experiments were performed using fish crossed and reared in laboratory conditions. Fish were maintained in 3.5 ppt Instant Ocean Sea Salt (Spectrum Brands, Blacksburg, VA, USA) in reverse osmosis water at 18 °C and pH 8.0 and a summer photoperiod (16L:8D). After hatching, fry were fed freshly hatched Grade A Brine Shrimp (Brine Shrimp Direct, Ogden, UT, USA) twice daily.

CRISPR/Cas9 knockout of amhy

We microinjected stickleback embryos following standard procedures [41], with the addition of KCl to the injection mix at a final concentration of 0.2 M [145]. Cas9 protein was purchased from QB3 MacroLab (University of California, Berkeley). Synthetic sgRNAs were designed using CHOPCHOP [146–148] and purchased from Synthego: amhy_sgRNA_77r (5’- GATCAGCTGCTGTCCATGCA-3’) and amhy_sgRNA_600r (5’- GGTCCTCTCTCGTCTTCGTT -3’). We combined Cas9 protein and sgRNA at a final concentration of 10 μM each and incubated for at least ten minutes at room temperature to allow the ribonucleoprotein (RNP) to form. The injection mix was made to a final concentration of 4 μM RNP.

Synthesis of constructs and transgenesis

Plasmids pBT2 [149], pUbi:Switch [75], and pCS-zT2TP [8] were kindly shared by Mary Goll. Plasmid pT2AL200R150G [71] (referred to as pT2Xla.Ef1a:EFGP) was kindly shared by Tyler Square and Craig Miller. We purchased a synthetic amhy,EGFP construct from TWIST Bioscience (San Francisco, CA, USA). This construct consisted of a Kozak consensus sequence (5’-GCCGCCACCATGG-3’), N-terminus flag tag, the threespine stickleback amhy coding sequence, a P2A self-cleaving peptide [73,74], EGFP, and an SV40 poly(A) sequence, all flanked by 5’ BamHI and 3’ NotI restriction sites to facilitate cloning. We performed all restriction digests following the manufacturer’s guidelines (New England Biolabs). We isolated all of the plasmids using a Monarch Plasmid Miniprep Kit (NEB #T1010) and gel purified all DNA fragments using the Monarch Gel Extraction kit (NEB #T1020). DNA fragments were ligated using T4 DNA Ligase following manufacturer recommended protocols (NEB #M0202). We transformed all plasmids into DH10B competent cells (Thermo Scientific EC0113). All plasmids were sequenced by Azenta Life Sciences (South Plainfield, NJ, USA) (S2 File).

We assembled each construct as follows (S6 Fig):

1. pT2Dre.Ubi:amhy,EGFP: The D. rerio ubiquitin promoter was excised from pUbi:Switch and cloned into pBT2 using XhoI and BamHI. The synthetic amhy,EGFP construct was excised from pTwist_amhy-P2A-EGFP and cloned into the intermediate pT2Dre.Ubi backbone using BamHI and NotI.
2. pT2Dre.Ubi:EGFP: the D. rerio ubiquitin promoter was excised from pUbi:Switch and cloned into pT2Xla.Ef1a:EGFP using XhoI and BamHI.
3. pT2Xla.Ef1a:amhy,EGFP: the synthetic amhy,EGFP construct was excised from pT2Dre.Ubi:amhy,EGFP and cloned into pT2Xla.Ef1a:EGFP using BamHI and BglII.
4. pT2-5amhy2.5,Xla.Ef1a:EGFP: an 11.5 kb region of the Y chromosome containing amhy along with 5 kb of upstream and 2.5 kb downstream DNA was excised from bacterial artificial chromosome STB26-N21 from the CHORI-215 library [150] and cloned into pBT2 using BamHI and XhoI. The Xla.Ef1a:EGFP cassette was excised from pT2Xla.Ef1a:EGFP by digestion with XhoI and HpaI and ligated into the intermediate pT2-5amhy2.5 plasmid following digestion with XhoI and SwaI.

We synthesized Tol2 mRNA using the mMessage mMachine SP6 Transcription kit (Invitrogen AM1340) and purified the product with the MEGAclear Transcription Clean-Up Kit (Invitrogen AM1908). We injected stickleback embryos as in [41] with the addition of KCl to a final concentration of 0.2 M [145]. All embryos were screened for EGFP expression five to seven days after injections before hatching had occurred. We removed all embryos without discernible somatic EGFP fluorescence.

Genotyping and sanger sequencing

We extracted DNA from fin clips or whole embryos using a HotSHOT isolation protocol [151]. We performed all PCR reactions on an Analytik Jena Biometra Tone in a total volume of 20 µl using 0.8 μL of genomic DNA, 0.2 μM each primer, 0.2 mM each dNTPs, 0.5 units DreamTaq DNA polymerase (Thermo Scientific EP), and 2 µl 10x DreamTaq Green Buffer with 20 mM MgCl₂. We genotyped fish for sex using the sex-linked marker idh [39] with the following cycling conditions: 1 cycle of 95°C for 2 min; 30 cycles of 95°C for 30 s, 57.3°C for 30 s, and 72°C for 30 s; 1 cycle of 72°C for 5 minutes. All other reactions used: 1 cycle of 95°C for 2 min; 35 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s; 1 cycle of 72°C for 5 minutes. Primer sequences for all reactions are listed in S7 Table. We purified the PCR product using a Monarch PCR & DNA cleanup kit (NEB #T1030). All samples were Sanger sequenced through Azenta Life Sciences (South Plainfield, NJ, USA).

Phenotyping and crossing

To phenotype gonads, we euthanized adult stickleback fish in 0.05% MS-222 buffered to neutral. We dissected out the gonads and fixed one or both gonads for 24–48 hours in 10% neutral buffered formalin. The remainder of the fish was preserved in 75% ethanol. The gonads were processed, embedded in paraffin, sectioned at 5 µm, and stained with hematoxylin and eosin by the University of Georgia College of Veterinary Medicine Histology Lab. We used a Zeiss Axio Scope A1 microscope and Zeiss Axiocam 305 Color camera to capture all images at the University of Georgia Biomedical Microscopy Core.

To make crosses, we euthanized mutant or wildtype male stickleback and gently macerated testes in 500 µl of Hanks’ Balanced Salt Solution. We stripped eggs from gravid females and applied approximately 50 µl of sperm solution with a disposable transfer pipette.

We photographed adult stickleback with a Samsung Galaxy S22 under standardized lighting. We quantified coloration using FIJI (v. 1.54p) [152]. We measured total fish body area using the SIOX: Simple Interactive Object Extraction plugin. We used the Threshold Color plugin to measure blue (Y = 0 – 254, U = 100 – 255, V = 0 – 131) and dark (Y = 0 – 31, U = 0 – 255, V = 0 – 255) body area in masked images. Pixel counts are presented in S3 File. We calculated the proportion of body area for each color and performed a beta regression for differences in means in R (v. 4.4.1) and RStudio (v. 2024.04.2 + 764) with the packages betareg (v. 3.2-4) and emmeans (v. 1.11.2-8), allowing dispersion to vary by genotype (color ~ genotype | genotype).

Riboprobe synthesis

We generated riboprobe templates for gsdf, dmrt1, foxl2, and cyp10a1a from cDNA from adult gonads. We extracted total RNA from mature testis and ovary tissue using TRIzol Reagent (Invitrogen 15596026) according to manufacturer’s instructions. We synthesized cDNA using the iScript cDNA Synthesis Kit (Bio-Rad 1708891) and amplified the targeted region for each gene through PCR as described above (S7 Table). We cloned the PCR product into plasmid pJC53.2 (kindly shared by Kendall Clay and Rachel Roberts-Galbraith) after digestion with Eam1105I (Thermo Scientific FD0244) as in Collins et al. 2010 [153]. We verified each riboprobe template after cloning with sanger sequencing by Azenta Life Sciences (South Plainfield, NJ, USA). For amh, we ordered a synthetic gene fragment from TWIST Bioscience (San Francisco, CA, USA) containing a 307 bp region of the amh cds flanked by a 5’ T3 promoter and 3’ T7 promoter to facilitate riboprobe synthesis (S2 File).

To synthesize riboprobes, we PCR amplified linear template from the appropriate plasmid or used the synthetic gene fragment directly. We performed in vitro transcription from these templates using T7 (New England Biolabs M0251S), T3 (New England Biolabs M0378S), or SP6 (New England Biolabs M0207S) RNA polymerase according to manufacturer’s instructions, substituting 0.2 mM of UTP for DIG-labeled UTP (Roche 11209256910). We purified riboprobes with a DNase treatment (New England Biolabs M0303S) and an ammonium acetate precipitation.

In situ hybridization

We adapted a colorimetric in situ hybridization (ISH) protocol from Square, et al. 2021 [154]. We fixed the dissected gonads for 24–48 hours in 10% neutral buffered formalin. The gonads were processed, embedded in paraffin, and sectioned at 5 µm by the University of Georgia College of Veterinary Medicine Histology Lab. Unless otherwise stated, we performed all steps in Lock-Mailer microscope slide jars in a volume of 9-11mL and at room temperature. To deparaffinize, we warmed the slides in a 65°C incubator for five minutes, allowed them to cool to room temperature, then washed the slides in Hemo-De once for 5 minutes then again for 10 minutes. We performed prehybridization washes indicated with an asterisk in a Tissue-Tek II Slide Staining Station rather than Lock-Mailer jars: 100% EtOH for 5 minutes*, 80% EtOH for 10 minutes*, Milli-Q water for 10 minutes*, PBST for 5 minutes*, 15 ug/mL proteinase K in PBST for 5 minutes, PBST rinse*, 4% PFA for 20 minutes, PBST twice for 10 minutes each*, 67°C pre-heated hybridization buffer (50% formamide, 5X SSC, 0.1% Tween, 5mg/mL CHAPS, 0.1 mg/mL yeast RNA, 0.1mg/mL heparin, pH 6.0 with citric acid) twice for 5 minutes each, and pre-heated hybridization buffer for 1–4 hours on a rotator at 67°C. We performed hybridization with 100–500 ng/mL of riboprobe in hybridization buffer incubated overnight rotating at 67°C. We stored hybridization solution at 20°C and reused up to 3 times.

Post-hybridization, we washed the slides 6 times with hybridization wash (50% formamide, 5X SSC, 0.1% Tween) rotating at 67°C for 20–90 minutes each, totaling 5–6 hours. We rinsed the slides in MABT (maleic acid buffer with Tween) and washed them in MABT twice for 20 minutes each, first at 67°C, then at room temperature. We added blocking solution (2% Blocking Reagent (Roche 11096176001) in MAB) to each slide, covered the slides with parafilm, and incubated for 1–3 hours in a humidor. We poured off the blocking solution and added 1:2000 anti-digoxigenin alkaline phosphate antibody (Roche 11093274910) in blocking solution, covered the slides with parafilm, and left them in a humidor overnight at 4°C. We performed five 20–50-minute post-antibody washes hybridization wash solution (50% formamide, 5X SSC, 0.1% Tween, citric acid to pH 6.0) on a rotator for a total of three to four hours, followed by an overnight MABT wash at 4°C.

Prior to coloration, we performed three 5-to-10-minute washes in NTMT (0.1 M NaCl, 0.1 M Tris pH 9.5, 0.05 M MgCl2, 0.1% Tween). We moved the slides into coloration solution (blocking solution plus 25 ug/mL NBT and 175 ug/mL BCIP) and incubated 2–30 hours covered to block light. We removed the slides from coloration solution once the color was stable in the positive control (gsdf), but before any background color was seen in the negative control (sense gsdf). Once coloration was complete, we rinsed the slides with PBST and washed for 10 minutes in PBST. We fixed the slides in 4% PFA in PBS at 4°C for one to five days.

To prepare slides for mounting, we performed three 5-minute washes in PBST and three 5-minute washes in deionized water. We mounted and coverslipped the slides using Vectashield antifade mounting medium with DAPI. We used a Zeiss Axio Scope A1 microscope and Zeiss Axiocam 305 Color camera to capture all images at the University of Georgia Biomedical Microscopy Core.

Sequence alignments

We retrieved sequences for amh (Gene ID: 120823390) and amhy (Gene ID: 120812167) from threespine assembly GAculeatus_UGA_version5 (GCF_016920845.1) and amh (Gene ID: 119216504) from ninespine assembly fPunPun2.1 (GCF_949316345.1). We performed all alignments and calculated sequence identity between amh homologues with Muscle 5.1 [155] within Geneious Prime (v2024.0.5) using default parameters. We defined intron and exon positions based upon the threespine amh sequence in each alignment. We calculated dN/dS with the R package ape v5.8-1 [156] using the translation aligned coding sequences of amh (XM_040183306.1), amhy (XM_040167960.1), and ninespine amh (XM_037469420.2). One additional base was included at the 5’ end of exon 7 of the ninespine amh cds in order to correct the predicted annotation which contains multiple premature stop codons in this final exon. Alignment files are presented in S2 File. We generated Vistaplots [157,158] using LAGAN [159] alignments and a 100 bp window size with a 70% sequence identity threshold.

RNAseq

We staged embryos according to established stickleback developmental stages [85], preserved the whole embryos in RNAlater Stabilization Solution (Invitrogen AM7021) following manufacturer guidelines, and stored at -80 °C until analysis. We extracted RNA using TRIzol Reagent (Invitrogen 15596026) and the Direct-zol RNA Microprep Kit (Zymo R2062). We created sequencing libraries using the Illumina TruSeq Stranded mRNA Library Prep kit (20020594). We pooled the libraries from all samples and 150 bp paired end reads were sequenced on two lanes using a NovaSeq X Plus by Azenta Life Sciences (South Plainfield, NJ, USA). All of the samples were sequenced to an average of 34.7 M read depth (min 11.6 M, max 71.3 M, SD = 10.4 M) (S1 File). We trimmed raw reads for quality and sequencing adapters with Trimmomatic (v. 0.39) [160] using leading and trailing quality threshold of 3, a sliding window of 4 with quality threshold of 15, and minimum length cutoff of 36. We aligned paired and orphan reads to the threespine stickleback reference genome (GAculeatus_UGA_version5, NCBI RefSeq assembly GCF_016920845.1) using HISAT-3N [161] with default parameters except for reversing RNA strandedness (R for Read 1 and F for Read 2). All samples had an alignment rate between 66% and 87%, counting only uniquely mapped reads (S1 File). We generated Indexed BAM files of the alignments with Samtools (v. 1.17). We generated transcript counts against the NCBI annotation GTF file (release 100) for GCF_016920845.1 with HTseq (v. 2.0.2) [162] using HTSeq-count with default parameters except for reversing RNA strandedness (R for Read 1 and F for Read 2). We performed all remaining analyses in R (v. 4.4.1) and RStudio (v. 2024.04.2 + 764). We combined the counts from paired and orphaned reads for each sample and used DESeq2 (v. 1.44.0) [163] to generate normalized read counts and perform differential expression analysis. We identified DEGs using the standard workflow with stage and sex combined as a single factor. We used the contrast function to calculate log2 fold changes (logFC) and adjusted p values for differential expression between males and females at each timepoint. We identified DEGs of interest using GO terms from the NIH gene2go database (https://ftp.ncbi.nlm.nih.gov/gene/DATA/).