All experiments were done with eggs and milt from the domesticated and commercially available Aquagen strain, Aqua Gen AS, Trondheim, Norway. Figure 4 shows the timeline for production of the different fish groups used in the current study.
Genotyping – genetic background/variation, sex and vgll3
To determine the genetic variation in the Golden Fish, sires 1–3, and dams 1–5, eighteen microsatellite DNA markers were genotyped using standard isolation and amplification protocols previously described in detail [50, 51].
Genotyping for sex served to distinguish potentially successfully sex-reversed fish (exposed to EE2) and distinguish YY from XY males. Total DNA was purified from whole adipose fins using Qiagen DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer’s recommendations. Sex was validated by a PCR-based method aimed to detect the presence of the sdY gene [18]. Individuals showing amplicons of exon 2 and 4 were designated as males. As a positive PCR control, and for species determination, we used the presence of the 5SrRNA gene [52]. PCR amplifications were performed using reaction mixtures containing approximately 50 ng of extracted Atlantic salmon DNA, 10 nM Tris–HCl pH 8.8, 1.5 mM MgCl2, 50 mM KCl, 0.1% Triton X-100, 0.35 μM of each primers, 0.5 Units of DNATaq Polymerase (Promega, Madison, WI, USA) and 250 μM of each dNTP in a final volume of 20 μL. PCR products were visualized in 3% agarose gels. For distinguishing between XY and YY males we used the method recently described by [32].
Genotyping of the vgll3 locus was performed using allelic discrimination assay for the two missense SNPs in vgll3 according to [14] and served to distinguish three different genotypes: (1) homozygous early (EE), (2) homozygous late (LL), and (3) heterozygous early/late (EL).
Sex reversal
On the 23rd May 2012, newly hatched alvelins (out bred from the commercially available and highly domesticated Aquagen strain) were immersed in a water bath for 2 h in 400 μg/L ethynylestradiol-17α (EE2) in 0.04% EtOH [30]. The bath treatment was applied 3 days after 50% of the embryos hatched. The fish were first fed in one 1 m ø tank under continuous light and 12 °C. On the 10th December 2012, fish were transferred to one 1.5 m ø tank. The temperature was changed to natural temperature on the 21st June 2012 and photoperiod changed from continuous to natural on the 1st October 2012. The water was changed from fresh- to salt-water on the 8th May 2013. The seawater temperature was stable at 9 °C and the photoperiod was simulated natural (60o N). On the 12th February 2014, all fish were tagged with an electronic transponder for individual recognition (i.e. PIT tag), had the adipose fin removed (stored in ethanol) for DNA extraction and genotyping, and transferred to one 5 × 5 m (7 m deep) sea-cage with natural photoperiod. Fish were genotyped for sex, vgll3, and genetic variation (microsatellites).
After 916 days in seawater, and 636 days in the sea-cage, on the 10th November 2015, all mature fish were checked for the relationship between genetic sex and external phenotype. Only one fish (hereafter called the ‘Golden Fish’) had a mismatch, with a male (XY) genotype, but a female phenotype. The Golden Fish was heterozygous for the early (E) and late (L) maturing vgll3 genotype (EL). Thus, this fish should produce X and Y eggs of both the E and L maturing genotype.
Production of YY supermales
Upon stripping the Golden Fish (killed by an overdose of anaesthetic; Finquel vet. 0.5 g L− 1), both milt and ova were released as it turned out to be a hermaphrodite. Subsequently, we first put the self-fertilized (self) eggs into an incubation tray and then gently dissected out the remaining ova to avoid further self-fertilization. There were no traces of milt inside the abdomen – the fish must have had a functional sperm duct – and the dissected ova were presumably un-fertilized. In order to produce double haploid (dh) offspring from the Golden Fish, the surgically removed eggs were fertilized with UV-irradiated milt, incubated at 8 °C for 4700 minC, subjected to a hydrostatic pressure of 655 bar for 5 mins (TRC-APV, Aqua Pressure Vessel, TRC Hydraulics inc., Dieppe, Canada) [29], and then transferred to an incubation tray. Eggs were incubated at 6 °C. Fish were first fed under a light dark (LD) regime of 12:12 and 12 °C in order to induce parr maturation in males [53]. Fish were pit-tagged and genotyped for sex, vgll3, and genetic variation (microsatellites) on the 7th September 2016. It was confirmed that we had YY supermales among the fish, both from the self (self-YY) and dh production (dh-YY). On the 17th November 2016, when grading out fully mature male parr, we found two mature self-YY’s (sire 1 and 2), both EL for vgll3, and one mature dh-YY (sire 3), EE for vgll3 (Supplementary Table 1). Sire 1 and 2 were killed by an overdose of anaesthetic (Finquel vet. 0.5 g L− 1), and had their testis dissected and homogenized in Cortland solution (124 nM NaCl, 5.1 mM KCl, 2.9 mM Na2HPO4). Sire 3 was anesthetized (Finquel vet., 0.1 g L− 1), stripped for milt, and kept alive for one more year until December 2017 when it was euthanized (Finquel vet. 0.5 g L− 1) for sperm cryopreservation. Sperm from other supermales (not used in the present study) maturing in December 2017 and 2018 was also cryopreserved.
Production of double haploid females
Dh females (dh-XX) were produced according to the procedure described by Hansen et al. [29]. In brief, on 18 December 2012, ova from six diploid females (Aquagen AS) were mixed and fertilized with UV irradiated sperm and subjected to a high hydrostatic pressure at the first meiotic division. Eggs were incubated following standard production procedures and fish produced as yearling smolts that were transferred to seawater in May 2014. At that stage they were also pit-tagged and tissue sampled. The samples were later genotyped (sex, vgll3, and microsatellites). The temperature in seawater was stable at 9 °C and the photoperiod was simulated natural (Western Norway). On the 17th November 2016, the second year in seawater, 5 fully mature ovulated females (dams 1–5, Supplementary Table 1) were selected based on vgll3 genotype, killed by an overdose of anaesthetic (Finquel vet. 0.5 g L− 1) and had their eggs stripped. Dams 1, 2, and 5 had the EE vgll3 genotype, and dams 3 and 4 had the LL vgll3 genotype. The microsatellite data in Supplementary Table 1 suggests dams 1–5 were all progeny of the same female as they share the same two alleles for all 18 markers. This female must have been vgll3 heterozygous early/late (EL).
All-male production
Eggs from dams 1–4 were each split in two equal parts and fertilized with milt from sire 1 or 2, creating two half sibling groups per dam, and a total of 8 different family groups, each with 50/50 occurrence of the two different vgll3 genotypes since sires 1 and 2 were heterozygous for vgll3. Milt from sire 3 was used to fertilize eggs from dam 5.
Each of the 9 family groups (8 from the self-YY (sires 1 and 2) × dh female (dams 1–4) cross, 1 from the dh-YY (sire 3) × dh female (dam 5) cross) were incubated in single trays in a flow-through system at 6 °C. Eggs were mechanically shocked at the eye egg stage on the 9th January 2017 and dead eggs removed. Hatching took place between the 4th and 16th February 2017 and first feeding was on the 22nd March 2017. Each family group were first fed in duplicate start feeding tanks (1 × 1 m, n = 18 tanks in total) under continuous light and a stable temperature of 12 °C. The fish were reared in these tanks until the 21st June 2017 when each family group was subsequently transferred to single 3 m tanks (n = 9 tanks in total). Here the fish were reared under natural temperature and the photoperiod was changed from continuous light to natural light on 1st October 2017.
Experimental set up: all-male - vgll3 genotype and jacking
On the 1st December 2017, 180 fish from each of the eight different sire 1 and 2 x dam 1–4 crosses, and 90 from the sire 3 x dam 5 cross, were pit-tagged and distributed in common garden between six 3 m ø tanks, with the same number of individuals from each group in each tank (totally 1530 fish; 255 per tank). Fish were kept under natural light and 6 °C in these tanks until the 8th January 2018 when they were anesthetized (Finquel vet., 0.1 g L− 1), had their pit-tag number recorded, measured for fork length and body weight, and moved to six new 3 m ø tanks. On the 9th January 2018, photoperiod was shifted to continuous light and the water temperature was gradually adjusted to 16 °C over a 3-day period to induce maturation [11]. Fish were kept under these conditions until the 6th March 2018, when they were all killed by an overdose of anaesthetic (Finquel vet., 0.5 g L− 1), had their pit-tag recorded, were sexed by visual examination of the gonad, and measured for fork length and gonad and body weight, and had their adipose fin sampled (on ethanol) for DNA extraction and vgll3 genotyping.
Statistics
Data were transferred to R version 3.6.1 (R Development Core Team 2018, http://www.r-project.org). All the raw data (“vgll3.csv”) and the R script (“vgll3.pdf”) used to analyse the data can be found in the supplementary material. In the analyses described below, the 44 fish that were phenotypic female were excluded. In addition, 4 fish with skeletal deformities were omitted due to their negative effect on growth [54]. One fish had an EE phenotype, even though the dam was LL, and was therefore omitted.
The fish were categorized as immature or pubertal (i.e. jacks) based on GSI. An initial histogram of GSI demonstrated a continuum between 0.01–0.20 and then those > 0.34 (Supplementary Figure 1a). Here, we expect the lower cluster to be immature individuals based on previous studies (immature fish generally have a GSI value of < 0.11: [8, 10, 55]. From previous work we know that in addition to larger testes, jacks have a high growth rate and an increase in body condition during early puberty, above that of immature males [12]. Therefore, we used unsupervised clustering to assess our GSI cut-off. Subsequently, principal component analysis (PCA) using the variables body mass and length at days 0 and 58, and gonad size, confirmed each genotype formed two clusters based on PC1 vs PC2 for which there was no overlap in those fish we had identified as jacks vs immature (Supplementary Figure 1b-d).
Our hypothesis was that EE males would be more likely to mature than LL males, with EL intermediate, following rearing under a maturation stimulating regime (LL and 16C). As the GSI was bimodal, we used a two-step or hurdle model to assess for genotype effects on puberty. The first part of the model assessed the prevalence of pubertal vs immature males within each genotype using a generalised linear mixed modal (GLMER) with a bimodal response, whereas the second part assessed GSI using a linear mixed effect (LME) model depending on genotype within pubertal males only. For the GLMER, jacking (two levels, Y/N) was the dependent variable, genotype (three levels, EE/EL/LL) was set as an independent variable, with tank as a random effect. For the LME, GSI was the dependent variable. Following this, as larger fish with higher energy reserves are expected to mature earlier, we expected EE fish to be the largest fish with the highest body condition and LL fish to be smaller with a lower body condition. To test this hypothesis, we generated linear mixed effect (LME) models with body mass or body condition as dependent variables, genotype (three levels, EE/EL/LL) as the independent categorical variable, and tank as a random effect. Here, we only used body size data from time zero, immediately prior to entering the environmental conditions known to induce puberty.
Following the above general models, the whole analysis was repeated whilst correcting for potential family effects. Here, the cross between dam 5 and sire 3 was not included as they produced only EE offspring. Furthermore, due to the experimental design, parental effects could only be assessed within EL fish, or between EE and EL or LE and LL fish, as dams 1 and 2 and 3 and 4 produced only EE and EL, or LE and LL offspring, respectively. This meant that EE genotypes could not be compared to LL genotypes when assessing family effects within the current experimental design. We then used the same approach as for the general models, but sire and dam were included as categorical independent variables. When comparing for family effects on the prevalence of jacking in the EE vs EL and the EL only comparisons, dam and sire were not allowed to interact, and tank could not be included as a random effect in these models as not all family/maturity types were represented. However, in the EL vs LL comparison and for the analysis of GSI within EE vs EL, EL vs LL and EL fish only, dam and sire were allowed to interact, and tank was included as a random effect.
Model fit was assessed by examination of model residuals (i.e. standardised vs fitted residuals, histograms, and/or q-q plots). Type II sum of squares were used for models without interactions, whereas type III sum of squares were used when interactions were present. The marginal R2 (R2m) is reported for all models, using the “r.squareGLMM” command within the “MuMIn” library, and the conditional R2 (R2C) is also reported for all models with a random effect. Significance was assigned at p < 0.05. Post hoc tests were done using lsmeans within the “emmeans” library.