A single intronic single nucleotide polymorphism in splicing site of steroidogenic enzyme hsd17b1 is associated with phenotypic sex in oyster pompano, Trachinotus anak

Abstract

Teleosts show varied master sex determining (MSD) genes and sex determination (SD) mechanisms, with frequent turnovers of sex chromosomes. Tracing the origins of MSD genes and turnovers of sex chromosomes in a taxonomic group is of particular interest in evolutionary biology. Oyster pompano (Trachinotus anak), a marine fish, belongs to the family Carangidae, in which 17b-hydroxysteroid dehydrogenase 1 (hsd17b1) has repeatedly evolved to an MSD gene. Whole-genome resequencing identified a single nucleotide polymorphism (SNP) at chromosome 24 to be strictly associated with phenotypic sex, with females being the heterozygous sex. This SNP is located in a splicing site at the first exon/intron boundary of hsd17b1. The Z-linked SNP results in malfunction of all spliced isoforms, whereas the W-linked isoforms were predicted to have open reading frames that are conserved among vertebrates, suggesting that hsd17b1 is a female-determining gene. The differential alternative splicing patterns of ZZ and ZW genotypes were consistently observed both in undifferentiated stages and differentiated gonads. We observed elevated recombination around the SD locus and no differentiation between Z and W chromosomes. The extreme diversity of mutational mechanisms that hsd17b1 evolves to an MSD gene highlights frequent in situ turnovers between sex chromosomes in the Carangidae.

1. Background

Sex determination (SD) systems are amazingly plastic in teleosts, involving both environmental and genotypic SD mechanisms [1,2]. Genotypic sex determination (GSD) systems have evolved independently and repeatedly, involving various master sex determining (MSD) genes [3,4]. In teleosts, almost all the identified MSD genes are members of gene families involved in gonad development, including DMRT1, members of the SOXA and B1 protein families and TGF-β signalling pathway as reviewed by Pan et al. [5] or genes that interact with these pathways [6,7], except for a few exceptions [8,9]. A recent study in the fish genus Seriola revealed that a steroidogenic enzyme gene, hsd17b1 evolved to be an MSD gene, providing an exceptional case of the diversity of MSD genes [10]. MSD genes arise in two major ways: mutations creating new alleles (allelic diversification), and neo-functionalization of duplicated genes [11,12]. In the case of allelic diversification, a mutant variant determines sex either by affecting the gene's protein function, or by altering its expression pattern [11,13]. In the case of an MSD gene originating from a duplication, recombination may be locally suppressed to allow sequence divergence, owing to lack of homology at the SD locus between a pair of sex chromosomes as reviewed by Kratochvil et al. [14], although such cessation of recombination can be also a result of heterozygosity connected with allele differentiations [15]. By contrast, mutant MSD alleles involving a single nucleotide polymorphism (SNP) difference might not, or less affect recombination, and divergence between its alleles can remain minimal, even in species whose SD systems have evolved over dozens of millions of years [10,11,13,16].

The oyster pompano, Trachinotus anak, belonging to the family Carangidae, is a marine fish of great commercial value in the Asian Pacific countries, with morphologically indistinguishable sex chromosomes (homomorphism) [17]. It has to be mentioned that almost all the commercially cultured oyster pompano (T. anak) in China, were misidentified either as Trachinotus ovatus with a native distribution in the east Atlantic, or its closely related sister species Trachinotus blochii distributed in the west Pacific [18,19]. A recent phylogenetic study based on mitochondrial DNA, classified specimens as Trachinotus anak [18,19]. In aquaculture practice, the sex ratio is 1 : 1, suggesting GSD. However, our preliminary study using RADseq-based genome scans, detected no SD locus (data not shown), raising the question: whether sex is determined by environmental factor or by an MSD locus in a physically small genomic region that was not discovered by low-resolution genetic markers, and required a whole-genome scan.

Locating the SD gene and understanding the mechanism of SD are not only intriguing in evolutionary biology but also valuable for the molecular selective breeding of this fish. Based on whole-genome resequencing in a captive population and transcriptome analysis, we identified a single intronic SNP at the first exon/intron boundary of 17b-hydroxysteroid dehydrogenase 1 (hsd17b1) that is strictly associated with sex, with females being the heterozygous sex (ZW SD system). This SNP affects splicing of Z-linked hsd17b1, leading to female development. No other divergent site was detected between the two alleles. Our data provide novel insights into the evolution of MSD genes and the distinct pattern of sex chromosome differentiation since turnover in teleosts.

2. Methods

(a) . Mapping populations

A brood stock obtained by the capture of wild oyster pompano in the South China Sea, was first employed to identify the sex determining locus. Ninety-six adults at more than 2 years old were randomly collected from this stock for whole-genome resequencing. The sex of each fish was determined by gonadal histological analysis. To evaluate the association between the candidate variant identified by resequencing and phenotypic sex, 25 wild adult fish were collected from the South China Sea and 233 fish at 2 years old were randomly selected from three large commercial populations (128, 63 and 42 fish, respectively), from Hainan Island and Guangdong, China.

(b) . Whole-genome resequencing and association mapping

Genomic DNA was isolated from the fin using TIANamp Marine Animals DNA Kit (Tiangen, China). DNA libraries with 500 bp insert size were constructed using an Illumina Truseq DNA PCR free Library Preparation Kit (Illumina, USA), and sequenced for 2 × 150 bp reads on an Illumina NovaSeq platform. Raw reads were filtered with process_radtags in the Stacks package (-c -r -q) [20]. BWA-mem [21] was used to align the cleaned reads to the reference genome of T. anak, which was previously misidentified as T. ovatus [17], with default parameters. Variants calling was performed according to the workflow of the best practices of Picard/GATK v. 4.0 [22]. SNPs and indels were filtered with the following parameters: ‘QD less than 2.0 || FS greater than 60.0 || MQ less than 40.0 || MQRankSum <−12.5 || ReadPosRankSum <−8.0 || SOR greater than 4.0’ and ‘QD less than 2.0 || FS greater than 200.0 || ReadPosRankSum <−20.0 || SOR greater than 10.0’, respectively, which kept 14 623 889 variants. After ruling out the variants with a minor allele frequency of less than 0.05 and missing genotypes of greater than 0.2, 2 082 987 variants were retained for further analysis.

The genetic background of the mapping population was assessed by principal component analysis using R package GAPIT2 [23]. Association between genotypes and phenotypic sex was assessed with Fisher's exact test, using Plink [24]. F_ST scan between females and males was carried out using VCFtools [25]. The identified candidate variant was genotyped using Sanger sequencing in selected commercial populations of oyster pompano.

(c) . Evolutionary analysis

Linkage disequilibrium (LD) among variants around the SD locus was estimated with r² in 10 kb windows using the program Plink [24] and viewed in Haploview [26]. Recombination rates (ρ = 4 Ner per kb) along sex chromosomes were estimated in 10 kb windows, using the program LDhat [27]. Phylogeny of the identified MSD gene was constructed based on the protein sequences of selected vertebrate species. Sequences were aligned using Muscle [28]. The phylogenetic tree was constructed using IQ-Tree2 [29], under the protein mutation model of JTT + I + G4, which was estimated using ModelFinder with Bayesian information criterion [30]. Bootstrapping was carried out by 1000 times.

(d) . Transcriptome sequencing and gene expression analysis

Fish at different developmental stages were collected and genotyped to determine the genotypic sex with primers in the electronic supplementary material, table S1. DNA isolated from samples acquired from heads, was used for genotyping. Both undifferentiated and differentiated gonads of both genotypes were determined by histology. For histological observations, either trunk or gonad sections were stained with haematoxylin and eosin according to a previous study [31]. For samples with undifferentiated gonads at 15 days post fertilization (dpf), six trunk samples from each genotype were pooled for transcriptome sequencing. For samples with differentiated gonads at 60 dpf, gonads from three samples of each genotype/sex were separately used for transcriptome sequencing. RNA was isolated using TRIzol (Invitrogen, USA) and messenger RNA (mRNA) libraries were constructed using TruSeq RNA Library Prep Kit v. 2 (Illumina, USA) for 2 × 150 bp sequencing using Illumina NovaSeq. For gonadal samples, more than 20 million reads were produced for each sample, while for pooled trunk samples, approximately 120 million reads were produced separately for female and male genotypes. Raw reads were filtered with process_shortreads (-c -r -q) [20]. Program Star [32] was used to align the cleaned reads to the reference genome [17], with default parameters. The expression level of genes of interest was quantified by HTSeq-count [33] and further normalized by transcripts per kilobase million. Differentially expressed genes were determined with EdgeR [34] using the following criterion: fold change (FC) > 2 and significance cut-off value of 0.01 after corrections with Benjamini–Hochberg false discovery rate.

Real-time quantitative polymerase chain reaction (qPCR) was used to validate the relative expression of the candidate gene. Total RNA of 2 μg from each sample was first digested with DNase I (Roche, Switzerland). The product was then used for synthesizing cDNA with Super M-MuLV Reverse Transcriptase (Sangon Biotech, China). Gene specific primers were designed using Primer3 [35]. The housekeeping gene, β-actin was used as a reference to normalize the relative expression of genes using 2^−ΔΔCT method, according to a previous method [36]. qPCR was carried out using SYBR^® Green Real-time PCR Master Mix (TOYOBO, Japan), with ABI 7900HT Fast Real-Time PCR System (ABI, USA). Three biological replications, each with three technical repeats were used for the study.

(e) . Examination of alternative splicing

Alternative splicing was first examined by analysing the assembled transcripts based on mRNA sequencing as described above. Spliced isoforms of interest were then amplified using reverse transcription PCR (RT-PCR) and separated by running 3% agarose gel electrophoresis. To estimate the richness of individual spliced isoforms, RT-PCR products were then cloned into pGEM-T vectors (Promega, USA) and sequenced by Sanger sequencing. Ninety-six clones for each genotype were randomly selected for sequencing.

3. Results

(a) . A single nucleotide polymorphism associated with phenotypic sex

We randomly collected 96 fish from a sample captures in the South China Sea. This ‘brood stock’ has 1 : 1 sex ratio (46 females and 50 males). Using whole-genome resequencing, we performed a whole genome-wide association study using 2 082 987 variants, with minor allele frequency greater than 0.05, including both SNPs and indels, with an average of 18× sequence coverage for each individual sample (electronic supplementary material, figure S1). Surprisingly, the association test found only a single SNP, at position 7 826 766 bp on chromosome 24, significantly associated with phenotypic sex (figure 1a,b). This SNP was heterozygous G/A in all 46 females and homozygous A/A in all 48 males, suggesting female heterogamety (figure 1c). Interestingly, we identified two alleles at this SNP in males: A and a single-base deletion (figure 1c). We first sequenced this SNP in a wild population consisting of 15 females and 10 males that were collected from the South China Sea, and also found perfect association with phenotypic sex. We further sequenced 233 individuals randomly collected from three large commercial populations, from Hainan Island and Yangjiang, Guangdong province, which also had a sex ratio of 1 : 1, and the perfect association with phenotypic sex was maintained in these samples (figure 1c). Approximately 1% of males in the mapping population had the allele of single-base deletion, and this was not found in the wild population and any of the three commercial populations. In order to test for other candidate variants that might have been filtered out as missing data, we reduced the cut-off value for missing data to 0.5 across the mapping population, but still found no other variant associated with phenotypic sex.

Figure 1.

Whole genome-wide association study identifies a single SNP associated with phenotypic sex in oyster pompano. (a) A photo of oyster pompano used in this study and the population structure of 96 captured oyster pompano, consisting of 46 females and 50 males, used for the whole genome-wide association study. (b) Only a single SNP at chromosome 24: 7 826 766 was identified to be associated with phenotypic sex in the whole genome-wide association test, which is highlighted with a red dot with a p value of E-27 (Fisher's exact test). (c) Strict correlations between genotypes at the SNP (chromosome 24: 7 826 766) and phenotypic sex separately in the whole-genome resequencing population and three commercial populations. Three alleles (G, A and one-base deletion (−), respectively) were identified in the population for the whole genome-wide association test. (Online version in colour.)

(b) . No differentiation between sex chromosomes

Annotation against the reference genome showed that the sex-specific SNP is located in the intron 1 of hsd17b1. Interestingly, F_ST analysis detected no evidence of differentiation between a pair of sex chromosomes, other than the sex-specific SNP (figure 2a). We then estimated the distribution pattern of recombination rates along the chromosome carrying the sex-specific SNP. Recombination rates were found to be elevated around the sex-specific SNP and show a peak at this SNP (figure 2b). In line with the above observations, linkage disequilibrium analysis yielded low pairwise r²-values between the sex-specific SNP and each of its flanking variants (less than 0.11) and significantly lower than the values for pairs of variants within either flanking region (one-way ANOVA, p < 0.0001) (figure 2c). We detected three additional intronic SNPs in the genomic region of hsd17b1 (position 7 825 351 bp, 7 826 643 bp and 7 826 734 bp, respectively), and the three SNPs were not linked with the sex-specific one (figure 2c). Although there is no morphologically distinct W chromosome or haplotype defined by multiple variants, we refer to the G allele as W-linked below. There is no evidence of differentiation around the sex-specific SNP as well as throughout the sex chromosome, and there is also no evident differentiation detected in the sex chromosome in comparison to the autosomes (electronic supplementary material, figure S2).

Figure 2.

Genetic differentiation and recombination signature around the sex-specific SNP. (a) F_ST analysis of sequence divergence around the sex-specific SNP between females and males, where annotated protein coding genes and the sex-specific SNP were indicated. (b) Distribution of recombination rates, estimated as 4 Ner per kb, along sex chromosomes, where the sex–specific SNP was indicated. (c) Plot of linkage disequilibrium, in the form of r²-values, around the sex-specific SNP, where pairwise r² involving the sex-specific SNP and the genomic position of hsd17b1 were highlighted. (Online version in colour.)

(c) . Z-specific alternative splicing leads to malfunction of hsd17b1 isoforms

Next, we checked that the protein sequences of the oyster pompano hsd17b1 is most similar to sequences of fishes from the same family Carangidae (figure 3a; electronic supplementary material, table S2). We examined the genome of the species and found only a single copy of hsd17b1. Sanger sequencing of the sex-specific SNP showed that the A and G alleles had same peak height in the female (electronic supplementary material, figure S3a). We further mapped the whole-genome resequencing reads of ZW and ZZ genotypes separately to the sex locus and found there was no difference in sequence coverage between genotypes (electronic supplementary material, figure S3b). Thus, the MSD gene hsd17b1 is not likely to have originated from duplication. The sex-specific SNP was located at the first base of intron 1, altering the 5’ splicing site of intron 1 from the W-linked GT sequence (a sequence that is conserved at the 5’ splicing site of most introns in eukaryotes) to the Z-linked allele's AT or TA for single-base deletion (figure 3b). We therefore hypothesized that the W-linked GT sequence is the ancestral state, and the Z-specific allele leads to alternative splicing of hsd17b1 and malfunction of the spliced isoforms.

Figure 3.

A single SNP at the first base of intron 1 of hsd17b1, is associated with phenotypic sex. (a) Phylogeny among vertebrates constructed based on protein sequences of paralogous genes: hsd17b1 and hsd17b3, where the position of oyster pompano was highlighted with a red dot and bootstrapping confidence was also indicated for each branch. Oyster pompano was most closely clustered with Seriola fishes including the amberjack (Seriola dumerili), yellow tail (Seriola quinqueradiata) and yellowtail kingfish (Seriola lalandi). Accessions of these protein sequences are listed in the electronic supplementary material, table S2. (b) Exon/intron boundaries and splicing sites of hsd17b1. The W-copy hsd17b1 (OK358625) has a conserved splicing site (GT) at the first exon/intron boundary, while the Z-copy hsd17b1 shows either an SNP (Z1, OK358626) or a single-base deletion (Z2, OK358627), altering the splicing site from GT to AT and from GT to TA, respectively. The splicing sites are highlighted in shade and intron sequences are shown with lower case letters. (Online version in colour.)

To examine this hypothesis, we sequenced the transcriptomes of both genotypes with G/A and A/A alleles at the sex-specific SNP, at different gonadal developmental stages (electronic supplementary material, figure S4a). Transcriptome sequencing of pooled trunks at 15 dpf, when the gonads were undifferentiated (electronic supplementary material, figure S5), showed no overall expression difference between the ZW and ZZ genotypes. At this stage of gonadal development, the marker genes of sex differentiation, e.g. the ATP-dependent RNA helicase of the DEAD-box family (vasa), gonadal soma derived factor (gsdf) and anti-Müllerian hormone (amh) also showed no major expression differences between these genotypes (electronic supplementary material, figure S6), suggesting that gonadal differentiation has not started. However, low expression of hsd17b1 transcripts yielded too few reads to examine alternative splicing of hsd17b1. By contrast, at 60 dpf, transcriptome sequencing of ovaries and testes revealed high basal expression of hsd17b1 in both, with approximately 40-fold higher expression in ovaries than testes (electronic supplementary material, figure S7). All isoforms involving splicing of sequences from exon 1 to exon 3 of hsd17b1 from ZZ genotypes, included frameshift and premature termination codons (electronic supplementary material, figures S4b,c and S8). By contrast, in ZW genotypes, approximately 70.4% of the isoforms were predicted to have an intact open reading frame (ORF) encoding a functional protein that is conserved among vertebrates (electronic supplementary material, figure S9), while the remaining approximately 29.6% of isoforms were sequences with the Z-specific SNP and all were non-functional, as in the ZZ genotypes (electronic supplementary material, figures S4b,c and S10).

(d) . Differential expression of hsd17b1 isoforms between ZW and ZZ genotypes

The transcriptome data showed that the alternative splicing of hsd17b1 affected the first two introns and three exons (electronic supplementary material, figure S4b). We next estimated the proportion of correctly spliced W-specific hsd17b1 transcripts using separated qRT-PCRs for ZW and ZZ genotypes and sequencing of the products. In ZW genotypes at 15 dpf when gonads were undifferentiated, approximately 65% of the total transcripts were correctly spliced W-specific transcripts, whereas in ZZ genotypes the value was close to zero (electronic supplementary material, figure S11a,b). However, the overall low expression of hsd17b1 did not differ significantly between ZW and ZZ genotypes (electronic supplementary material, figure S11b). In differentiated gonads at 60 dpf, approximately 58% of the total transcripts were correctly spliced W-specific in ZW individuals and again the value of correctly spliced transcripts was almost zero in ZZ genotypes (electronic supplementary material, figure S11c). Thus, the splicing difference between ZW and ZZ was similar in undifferentiated and differentiated gonads.

Finally, we amplified and sequenced the transcripts from exon 1 to exon 3 separately for ZW and ZZ genotypes at 15 dpf. Based on transcriptome data, we predicted a total of eight alternatively spliced isoforms, presented in ZW and/or ZZ genotypes (figure 4a). In line with the above results, RT-PCR showed that isoform X1 that was predicted to be present only in ZW fish and have a functional product was indeed observed only in ZW samples (figure 4a,b). We detected five additional alternatively spliced isoforms in ZW and ZZ genotypes that were produced only when the Z-specific SNP was present (figure 4b). Sequencing of the RT-PCR products revealed that isoform X1 accounted for 70.8% of the total isoforms in ZW genotypes (figure 4c; electronic supplementary material, figure S12). Because mRNAs containing premature stop codons decay more rapidly than those containing ORFs [37], there may be no difference in the quantity of Z- and W-linked hsd17b1 transcripts.

Figure 4.

Expression of hsd17b1 alternatively spliced isoforms detected and quantified using reverse transcription PCR (RT-PCR). (a) Diagram of all possible isoforms for sequence splicing from exon 1 to exon 3 of hsd17b1, predicted from gonadal transcriptomes of both ZW and ZZ genotypes. Positions of primers used for RT-PCR examination are indicated. Prediction of the length of RT-PCR products and the presence in different genotypes for individual isoforms, were indicated in the right. Frameshift and stop codon for individual isoforms were indicted with circled F and S, respectively. Each number represents one specific type of isoform, corresponding to that in (b). (b) The presence of individual isoforms as predicted in (a), in trunk samples between ZW and ZZ genotypes (n = 3), at 15 dpf. Each band represents one specific isoform, which is numbered corresponding to the isoform in (a). (c) Proportion (%) of each isoform separately within the trunk samples of ZW and ZZ genotypes at 15 dpf. Each number represents one specific type of isoform and is corresponding to that of the isoform in (a) and (b). (Online version in colour.)

4. Discussion

(a) . A gene, hsd17b1 for interconversion of sex steroids is the master sex determining gene in oyster pompano

In the past nearly twenty years, MSD genes have been discovered in at least a dozen teleosts, including dmrt1bY in medaka (Oryzias latipes) [38], dmrt1 in Chinese tongue sole (Cynoglossus semilaevis) [39], a duplicated copy of amh in Patagonian pejerrey (Odontesthes hatcheri) [40] and Nile tilapia (Oreochromis niloticus) [41], amhr2 in the tiger pufferfish (Takifugu rubripes) [13] and gdsf in one medaka species (Oryzias luzonensis) [11]. These MSD genes are members or related to the TGF-β superfamily signalling cascade. The rainbow trout (Oncorhynchus mykiss) MSD gene, sdY, encoding an immune-related protein homologous to the C-terminal domain of interferon regulatory factor 9 [42], seemed to provide an exception, but was later shown to interact with the female-determining transcription factor foxl2 and suppress the synthesis of oestrogens, leading to male development [7]. In this study, our data strongly suggest that hsd17b1 is the MSD gene in oyster pompano. hsd17b1 is responsible for interconversion between oestrone (E1) and oestradiol (E2), and between androstenedione (A-dione) and testosterone (T) [43,44]. Our study therefore provides an unusual teleostean MSD gene, one involved in controlling synthesis of steroid hormones. This is not implausible. A recent study of Seriola fish including Seriola dumerili and Seriola quinqueradiata showed that a sex-specific missense SNP in the Z-linked hsd17b1 decreased its conversion efficiency from oestrone and oestradiol and led to male development, making this gene a candidate for the MSD gene in these species [10]. Here, our study provides the second case that hsd17b1 evolves as the MSD gene in vertebrates, although the mutational mechanisms are different. Given that sex steroids can not only maintain gonadal differentiation but also induce frequent sex reversal even after the SD stage [45,46], these studies suggest that genes regulating the relative expression and/or functions of sex steroids could also evolve to become MSD genes.

hsd17b1 also plays a critical role in conversion from androstenedione to testosterone, which is necessary for testicular development [43,44]. Our observation that the Z-linked hsd17b1 allele is a loss of function mutation raises the question of how androgens are available to maintain testis development in males in oyster pompano. We speculate that the paralogous gene of hsd17b1, hsd17b3 allows interconversion between androstenedione and testosterone in males to sustain spermatogenesis, as occurs in another fish, the turbot (Scophthalmus maximus) [47,48]. We detected only one copy of hsd17b3 in the genome of oyster pompano, which diverged from hsd17b1 owing to ancient duplication (figure 3a). Interestingly, in the gonadal transcriptomes at 60 dpf, hsd17b3 was almost undetectable, whereas in the transcriptomes of trunks at 15 dpf, we detected substantial expression of this gene in both ZW and ZZ genotypes (electronic supplementary material, figure S13), suggesting that hsd17b3 has a different expression pattern from hsd17b1 and could have taken over hsd17b1's function in producing androgens in ZZ genotypes.

(b) . No sequence divergence and recombination suppression between Z and W chromosomes

As two different modes of MSD origins, allelic diversification and gene duplication probably have distinct effects on the evolutionary fate of sex chromosomes [14]. Here, both Sanger sequencing and whole-genome resequencing data suggest only a single copy of hsd17b1 in oyster pompano, and this MSD gene originated from allelic diversification by a point mutation, providing a rare opportunity to study sex chromosome evolution. Sex chromosome differentiation has been observed in many teleosts with MSD genes originated from larger sequence variation than a point mutation [6,12,49], even in species with recently evolved sex chromosomes [50,51]. The similar sequences of the Z- and W-linked alleles in the oyster pompano is not surprising, given that only a single SNP is fixed between the two alleles, and the SD gene is in a recombining region of the genome. The oyster pompano SD system may also have evolved too recently for variants to accumulate between the two alleles. In Seriola fishes, the genomic region that includes the oyster pompano candidate MSD gene, hsd17b1, has remained undifferentiated for over 30 million years, owing to a high recombination rate around the SD locus [10]. In the tiger pufferfish, the sex chromosomes also showed few SNPs associated with the SD variant, but slight differentiation was detected by an LD analysis; this system is estimated to have been maintained for over 30 Myr [13]. Recent studies in flathead grey mullet showed that sex chromosome differentiation was limited to a few SNPs within the follicle-stimulating hormone-receptor (fshr) gene [16,52]. MSD genes in genome regions with frequent recombination are probably more difficult to map in captive or wild populations with large effective population size than in a single family, even using high-resolution markers such as SNP microarrays [53] and RAD-based sequencing (RADseq) [54]. The evolution of MSD alleles by mutations in existing genes may therefore be underestimated in comparison to evolution involving gene duplications.

(c) . Frequent in situ turnover of sex chromosomes in the Carangidae

To date, hsd17b1 has been exclusively identified or suggested as the MSD gene in five Carangidae fishes, including oyster pompano as revealed in this study and four Seriola fishes [10,55]. Among these five fishes, S. dumerili, S. quinqueradiata and S. lalandi share the same ancestral Z-linked missense SNP in hsd17b1 that leads to SD in these three species and evolved at least 30 Ma [10], while in oyster pompano, a different single intronic SNP affecting alternative splicing of Z-linked hsd17b1 determines sex. In another Seriola fish, S. dorsalis, a W-specific 61-base deletion upstream of hsd17b1 disrupts a candidate silencer and potentially enhances oestrogen production, making it a candidate MSD gene [55]. By contrast, the protein sequences of the W-linked copies are conserved in oyster pompano, S. dumerili, S. quinqueradiata and S. lalandi, and match sequences in other vertebrates [10]. This suggests that the W-linked hsd17b1 is ancestral while the Z-linked alleles are derived by mutations. However, in S. dorsalis, the Z-specific hsd17b1 allele sequence is conserved in all five species [55], which suggests that the W-specific allele is a derived gain-of-function mutant (electronic supplementary material, figure S14). In particular, S. dorsalis was suggested to diverge from S. lalandi at approximately 2.2 Ma [56], indicating the turnover from W- to Z-specific allele occurred in this lineage recently (electronic supplementary material, figure S14). Nevertheless, the recurrent evolution of hsd17b1 to be the master sex determiner in the Carangidae suggests frequent in situ turnover of sex chromosomes, as well as that the regulatory network of SD in this taxon is sensitive to mutations at this gene.

In the four Seriola fishes, hsd17b1 appears to be a dosage-sensitive female-determining gene, and the higher expression allele ensures ovarian development. This is similar to the dosage-sensitive male determining gene, dmrt1. In birds [57] and the Chinese tongue sole [39], both with ZW SD systems, reduced expression of one dmrt1 copy relative to the other leads to female development. In the frog species (Xenopus laevis), also with a ZW SD system, a mutant copy of dmrt1 with reduced functional activity is also a female-determining gene [58]. Genes controlling sex steroids may be especially likely to evolve into dosage-sensitive MSD genes than other genes, owing to the antagonism between oestrogens and androgens [59,60]. In Nile tilapia, increasing androgen can reduce the synthesis of oestrogen by inhibiting aromatase activity and induce conversion of females to males [45,61]. In medaka, E2 and 17α-methyldihydrotestosterone can suppress each other to induce sex reversals, and alteration of the relative level of the two sex steroids can induce sex reversal before gonadal differentiation [62]. These data suggest that variants affecting the relative expression of hsd17b1, could allow sex chromosome turnovers in the Carangidae. In kutum fish (Rutilus frisii), changes in plasma levels of E2 were closely correlated to ovarian development [63]. Therefore, alteration in the relative expression level of hsd17b1, as a result of sequence variation, may affect sexual maturation. In this regard, the candidate gene identified may also control the balance between male and female development. As hsd17b1 is probably a dosage-sensitive female determiner [55], we hypothesize that an ancestral system in which males were the heterogametic sex (which could be termed as XY), was existing in the Carangidae and maintained by the relatively higher expression of hsd17b1 in XX females than in XY males. An X-linked mutation in the ancestral hsd17b1 allele could create a new female determiner (or W-linked allele) and lead to turnover from an XY to ZW SD system.

Ethics

All procedures for the handling of fish were according to the instructions of the Animal Care and Use Committee of Yangjiang Polytechnic of Guangdong Province, China.

Data accessibility

Raw sequencing reads for both whole genome-sequencing and transcriptome sequencing are archived in the China National GeneBank DataBase (CNGB) with BioProject accession CNP0001493.

Authors' contributions

B.F.: conceptualization, data curation, formal analysis, funding acquisition, investigation, project administration; D.X., Ya.L., Z.M. and X.W.: resources, validation; X.Q. and S.L.: resources, validation, visualization; X.C., J.P. and; Y.Y.: resources; Yu.L.: funding acquisition, resources, supervision, writing—original draft; L.W.: conceptualization, data curation, formal analysis, investigation, methodology, project administration, software, supervision, visualization, writing—original draft, writing—review and editing. All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

This project was supported by the China Agriculture Research System of MOF and MARA (CARS-47) and Guangdong Modern Agricultural Industrial Park (grant no. GDSCYY2020-011), China.