Sex determination is a basic process in the evolution of sexual reproduction and diverse sex-determination systems (SDSs). Genetic sex determination systems, multifactorial polygenic sex-determining mechanisms, and environmental sex determination systems are found across taxa (Ezaz et al., 2016). Teleosts show a large diversity of SDSs, whereas turnovers or transitions often occur in many vertebrates (Pennell et al., 2018). ZZ/ZW and XX/XY are the common SDSs in the female and male heterogametic systems, and they have evolved repeatedly and independently in the lineage. This results in the presence of rapidly evolving SDSs in closely related species (Gammerdinger and Kocher, 2018; Nguyen et al., 2021a; Nguyen et al., 2021b). Thus, a study of sister taxa retaining divergent sex chromosome systems would be informative when examining SDS evolution. Among the 115 species of clariid fish (Clariidae), some have either male heterogametic or female heterogametic sex determination systems (Pandey and Lakra, 1997; Nguyen et al., 2021a). Both heterogametic systems are present together, including multiple loci and sex chromosomes in the lineage. Varying SDSs have been observed in African catfish (Clarias gariepinus, Burchell, 1822) based on their geographical location: a ZZ/ZW system in Africa, a XX/XY system is indicated in some populations from Israel, Hungary, and China, or both systems in the same population in Thailand (Teugels et al., 1992; Eding et al., 1997; Eyo and Effiong, 2005; Nguyen et al., 2021a). The SDS of African catfish might be influenced by a multifactorial, polygenic sex-determination (PSD) system that occurs in varied species. The occurrence of natural, multiple and independent “switch” loci or alleles specify the segregation of sex within a species. PSD can also occur by modifying current sex chromosomes to generate a third functional sex chromosome at the same locus, or by modifying autosomal loci in other regions of the genome to generate a novel process for gonad development regulation. A well-known example is observed in platyfish (Xiphophorus maculatus, Günther, 1866), with some populations showing segregation of male Y chromosome alleles and female W chromosome alleles at the same chromosome pair (Volff and Schartl, 2001). However, the genetic architecture contributing to variations in SDSs is not completely understood, especially in the few known examples of PSD in teleosts (Roberts et al., 2016). Clarification of the genetic structure of SDSs is essential to understand the evolutionary mechanisms and their influence on the processes of speciation (Qvarnström and Bailey, 2009), particularly in species without heteromorphic sex chromosomes.

Interspecific hybridization has successfully improved the growth rate, body size, disease resistance, and tolerance to stressors in aquaculture (Suprapto et al., 2017). For instance, male African catfish (C. gariepinus, 2n = 56) have been crossed with female bighead catfish (Clarias macrocephalus, Günther, 1864, 2n = 54 with XX/XY) to produce fast growth and improved disease resistance traits in the resulting hybrids (2n = 55) (Na-Nakorn, 1995; Chaivichoo et al., 2020; Nguyen et al., 2021a; Nguyen et al., 2021b). However, reproductive failure can occur as a result of chromosomal incompatibility with parental genomes. This can culminate in spermatogenic breakdown in F₁ male hybrids, leading to subsequent elimination by apoptosis and limiting mass production (Ponjarat et al., 2019). By contrast, F₁ female hybrids can produce copious backcross offspring with low embryo mortality (Na-Nakorn et al., 2004; Abol-Munafi et al., 2006) due to differences in the checkpoint systems in meiotic cells. This may be more stringent during the development of mature spermatozoa than production or development of an ovum or as a result of complex interactions between two different SDSs (Ponjarat et al., 2019). Thus, the male C. gariepinus × female C. macrocephalus hybrid is a good animal model to study the evolutionary process of SDSs in teleosts. In this study, we proposed the following two hypotheses: 1) the hybrid might possess the dominant SDS from male African catfish when Y or W sex chromosomes is present in different individuals, or 2) the hybrid might exhibit new candidate loci or a combination of different alleles due to the influence of PSD found in male African catfish. To determine how sex determination affects the viability of hybrid catfish, genome-wide single-nucleotide polymorphism (SNP) analyses were performed using Diversity Arrays Technology (DArTseq™) of captive-bred individuals scored with phenotypic sex, as described by Nguyen et al. (2021a) and Nguyen et al. (2021b). Results will provide insights to improve the understanding of SDSs in clariid fish.

Adult individuals were collected as 15 male and 15 female hybrids, chosen randomly from a breeding stock (Kasetsart University, Bangkok, Thailand) to reduce the possibility of a high occurrence of siblings in the genetic pool. The hybrids were produced from pure breeding stock maintained for 10 years. Six generations of the breeding stock were bred in captivity. The sex of each individual was identified based on external morphology and internal examination of gonadal morphology (Esmaeili et al., 2017; Ponjarat et al., 2019). Animal care and all experimental procedures were approved by the Animal Experiment Committee, Kasetsart University, Thailand (Approval no. ACKU61-SCI-026) and concurred with the Regulations on Animal Experiments at Kasetsart University. The dorsal fins of each individual were removed for DNA extraction and total genomic DNA. DArT sequencing and genotyping, marker selection, DArT sequencing analysis, estimation of expected sex-linked markers, comparison of potential sex-linked loci, and homology searching processes were performed.

The DArTseq methodology identified 50,348 SNP loci and 51,294 PA loci. Polymorphism information content values ranged from 0 to 0.50 for all loci, suggesting that the overall distribution of PIC values was asymmetrical and skewed toward higher values. Several SNP and PA loci were selected with a varying set of criteria (Table 1; Figure 1 and Figure 2) and compared to determine the existence of XX/XY or ZZ/ZW SDS in the hybrid. In the case of ZZ/ZW type, filtering using the criterion of 30:70 male:female showed in seven SNP loci and 128 PA loci as moderately female-linked (Supplementary Table S2, Figure 2). Hamming distances between male and female individuals using the moderately sex-linked SNP and PA loci indicated within-sex distances of 0.370 ± 0.032 in males and 0.382 ± 0.026 in females for SNP loci, and 0.613 ± 0.011 in males and 0.397 ± 0.019 in females for PA loci. Between-sex distances were 0.669 ± 0.019 for SNP loci and 0.712 ± 0.008 for PA loci. The CATT results showed a statistically significant association with the phenotypic sex for seven SNP loci (χ² = 5.231–16.425, p < .022) and 128 PA loci (χ ² = 4.320–13.715, p < .037) (Table 1; Supplementary Table S2, Figure 1). The criterion of 20:80 male:female involved the designation of one SNP locus and sixteen PA loci as moderately female-linked (Supplementary Table S2, Figure 2). Proportional pairwise Hamming distances between males and females using the moderately sex-linked SNP and PA loci (under the null exclusive model) showed within-sex distances of 0.133 ± 0.033 in males and 0.343 ± 0.047 in females for SNP loci, and 0.603 ± 0.017 in males and 0.338 ± 0.018 in females for PA loci. Between-sex distances were 0.760 ± 0.029 for SNP loci and 0.745 ± 0.009 for PA loci. The CATT results verified a significant association with phenotypic sex for one SNP locus (χ ² = 16.425, p < 0.000) and 16 PA loci (χ² = 6.250–13.715, p < 0.012) (Table 1; Supplementary Table S2, Figure 1).

For the XX/XY determination system, filtering using the criterion of 70:30 male:female yielded seven SNP loci and 97 PA loci that were male-linked (Supplementary Table S3, Figure 2). Hamming distances between male and female bighead catfish using sex-linked SNP and PA loci (under the null exclusive model) showed lower within-sex distances of 0.367 ± 0.02 in males and 0.457 ± 0.021 in females for SNP loci, and 0.403 ± 0.022 in males and 0.517 ± 0.033 in females for PA loci. Between-sex distances of 0.680 ± 0.014 for SNP loci and 0.719 ± 0.018 for PA loci were observed. CATT results verified the significance of associations, with phenotypic sex for seven SNP loci (χ ² = 6.533–10.995, p < .001) and 97 PA loci (χ ² = 4.572–16.250, p < .032) (Table 1; Supplementary Table S3, Figure 1). No SNP loci and seven PA loci were associated with males based on the criterion of 80:20 male:female (Supplementary Table S3, Figure 2). Hamming distances between male and female hybrid catfish using sex-linked PA loci showed lower within-sex distances 0.318 ± 0.022 in males and 0.539 ± 0.038 in females for PA loci. Between-sex distances of 0.778 ± 0.020 for PA loci were observed. CATT results verified significant associations with phenotypic sex for seven PA loci (χ ² = 9.458–16.250, p < .002) (Table 1; Supplementary Table S3, Figure 1). A glPlot showed similarity between sexes in the sample group for both XX/XY and ZZ/ZW SDSs when considering moderately sex-linked loci (Figure 3). Female sex-linked loci and male sex-linked loci of hybrid catfish indicated sequence homology with vertebrate genomes, based on global BLAST analyses of NCBI databases (Supplementary Tables S2,3). Regarding the ZZ/ZW type, ten of the 135 SNP and PA loci were homologous to the putative genes. Moreover, two SNP loci and 18 PA loci demonstrated partial homology with transposable elements (TEs) (Supplementary Tables S4,5). For PA loci, chi-square tests showed that the 30:70 and 20:80 filtering criteria demonstrated insignificant differences in males (χ ² = 5.388 × 10^–32, p = 1) and females (χ ² = 1.959 × 10^–33, p = 1). For SNP loci, Kruskal–Wallis tests indicated that these filtering criteria produced insignificantly different percentages of heterozygosity in males (H = 2.860, p = .091) and females (H = 6.840, p = .009) (Supplementary Figure S1). Calculation of pairwise comparisons using the Nemenyi-tests with chi-square approximation for independent samples revealed that the 30:70 and 20:80 filtering criteria resulted in insignificant differences in heterozygosity compared with other filters for males (p = .143) and females (p = .571). By contrast, with regard to the XX/XY sex chromosome system, six of 104 SNP and PA loci were homologous to the putative genes. Two SNP loci and twenty-one PA loci showed partial homology with TEs (Supplementary Tables S6,7). Chi-square tests showed that the 70:30 and 80:20 filtering criteria indicated insignificant differences in males (χ ² = 3.962 × 10^–32, p = 1) and females (χ ² = 2.518 × 10^–32, p = 1) for PA loci. The hybrid sex-linked loci were searched to find the genome-wide SNP of their parents (C. gariepinus × C. macrocephalus) and detect the loci inherited from their parents. The genome-wide SNP of their parents (C. gariepinus × C. macrocephalus) were used from our previous study, as described by Nguyen et al. (2021a); Nguyen et al. (2021b). Twenty-six candidate sex-linked markers were found in bighead catfish and three candidate sex-linked markers were found in African catfish, while 75 of the total 84 candidate sex-linked loci of the bighead catfish were found in the hybrids (Supplementary Table S8).

Across a range of sample sizes and loci, 30 phenotypically sexed individuals are essential to minimize the probability of selecting less than one spurious sex-linked marker. The probability (P _i ) of a single locus exhibited a sex-linked pattern by chance was 9.31 × 10^–10. For the full data set (no filtering), out of 101,642 loci (including SNP and PA loci), 9.463 × 10^–5 were expected to spuriously show perfect sex-linked loci based on chance alone.

Although the SDS of the last common ancestor of clariid fishes is not known, the hybrid catfish in this study provided a unique opportunity to study the various SDSs that are present in clariid fishes. Most SDSs are XY or ZW homomorphic sex chromosomes (Nguyen et al., 2021a). Due to the lability of sex determination processes in teleosts, closely related species sometimes possess different SDSs (Kitano and Peichel, 2012). Both sex determination systems might have undergone collateral evolution, or a transformation might have occurred in the common ancestor, which separated the lineages. Here, the genome-wide SNP of their parent (C. gariepinus x C. macrocephalus) was used from our previous research, as described by Nguyen et al. (2021a); Nguyen et al. (2021b). Female bighead catfish can produce only X gametes, whereas male African catfish can produce possible X, Y, and Z gametes. If both systems are present in the African catfish population in Thailand, their sex-linked SNP loci can be independently transferred from the male parent to the hybrid, thereby resulting in the presence of different SNP profiles in each individual. In this study, 26 candidate sex-linked markers came from the bighead catfish and three candidate sex-linked markers came from the African catfish. Seventy-five of the total 84 candidate sex-linked loci of the bighead catfish were observed in the hybrids but they were not all sex-linked loci in the hybrid (Supplementary Tables S8). This suggests that several sex-determining loci or clusters remain in the parental African catfish. Developmental pathways that are downstream of the primary sex-determining genes are conserved among various organisms having vastly different SDSs (Ezaz et al., 2016). These results collectively suggest that the variation in the SDSs of the parents might result in PSD in the hybrid. Male F₁ hybrids are sterile because of the influence of different numbers of diploid chromosomes in their parents or because of the presence of complex SDSs. Sex-linked genes are important to determine the incompatibility between hybrids and the presence of non-homologous sex chromosomes could indicate reproductive isolation and increase the probability of speciation (Qvarnström and Bailey, 2009). In our previous study, we have shown that all sex-linked loci in African catfish and bighead catfish were not found in the same linkage group (Nguyen et al., 2021a; Nguyen et al., 2021b). Therefore, we predict that African catfish exhibit PSD on different chromosomes. PSD results in the generation of multiple phenotypic or reproductive classes within one sex. In the African catfish, more than two sex types have been observed in diverse populations (Teugels et al., 1992; Eyo and Effiong, 2005; Nguyen et al., 2021a). The presence of multiple classes may promote fitness benefits as natural selection in the population, suggesting that PSD engenders an evolutionarily stable scenario. Moreover, there is a noticeable diversity in how PSD has evolved separately across teleosts (Sandra and Norma, 2010). Genetic conflicts may also markedly impact hybrid sex determination and gametogenesis, with outcomes as intersex or sterile individuals, respectively and this can build strong post-zygotic barriers. PSD systems are evolutionarily unstable, because of sex-specific natural selection; they have generally been evaluated as only a midway step in the evolution of SDSs (Bachtrog et al., 2014).

Several reasons for the emergence of differences in SDSs in two closely related species have been postulated. With no sex chromosomes in the last common ancestor, both systems developed independently or one lineage retained the ancestral sex chromosomes, while transition occurred in other lineages (Mank et al., 2006). Such transitions can occur on the same pair of sex chromosomes or involved as an autosome that then became a new sex chromosome (Franchini et al., 2018). Sex-determining gene(s) at the molecular level may change their mode of action from female to male determination. Alternatively, a novel sex-determining (SD) gene might have initiated SD turnover in the evolving lineage. This change in mode of action of the genes may occur in the lineage of clariid fish, where both XX/XY and ZZ/ZW systems are found (Nguyen et al., 2021a; Nguyen et al., 2021b). By contrast, the platyfish has XY system, which is an ancestral SDS. One X chromosome acquired a female sex-determination allele dominant to the Y chromosome. This resulted in WY individuals developing ovaries. Several species of cichlid fish from Lake Malawi exhibited both an XY locus and WZ locus on different chromosome pairs. With occurrence on the same individual, the W female determines the dominant locus and the selected individual consequently develops as female (Ser et al., 2010). Ultimately, the dominant locus determines the fate of the gonad in the system being examined. Modification of existing sex chromosomes leads to multifactorial mechanisms, creating a functional third sex chromosome at the same or different locus. In this study, the female hybrids were fertile and able to produce large numbers of backcross progeny (Na-Nakorn et al., 2004; Ezaz et al., 2016). This suggested that the systems remain partially compatible (if one supersedes the other). We determined seven SNP or PA markers for male-linked loci and 17 SNP or PA markers for female-linked loci, with the criterion of 80:20 male:female across all examined hybrid specimens. Comparison of sex-linked loci of hybrids and the parent species showed no tendency of inheritance. Many false-positive signals might be desirable from such specimens due to their diverse genetic backgrounds (Gamble et al., 2015). Current data exhibited the possibility of this approach. The probability that a single locus exhibited sex linkage pattern by chance was 9.31 × 10^–10, and when the full data set (without filtering) of 101,642 loci (including SNP and PA loci) was considered, 9.463 × 10^–5 were expected to spuriously indicate sex-linkage. Thus, identification of any erroneously sex-linked loci by chance seems unlikely. This suggests the possibility of the presence of novel sex-linked genetic loci as a consequence of combination polygenes. Combining independently evolved sex chromosomes could also stimulate new combinations of sexual characters (Tanaka et al., 2007). Sex determination in zebrafish and cichlids likely results from a combination of additive and epistatic interactions at many loci (Parnell and Streelman, 2013). Surprisingly, the PA35637844 female-linked locus of the hybrid in the present study showed homology with the B4GALNT1 gene, and this locus was derived from a non-sex-linked locus found in the bighead catfish. Our gene ontology search in the Ensembl v103 database confirmed that the B4GALNT1 gene encodes functions related to the spermatogenesis pathway and hatchability (Izumi et al., 2019). Hence, the PA35637844 locus might have a putative role in the sex determination of the hybrids. An allele that wins in one combination of sex determination loci genotypes may lose in another, as observed from sex-linked loci of the bighead catfish. However, the influence of PSD remains in hybrids and African catfish. These outcomes illustrate the difficulty in predicting the consequences of SDSs on the speciation process that has only been studied in some cases.

Hybrid catfish is of considerable economic importance for aquaculture as they can improve productivity through hybrid vigor, produce sterile animals or transfer desirable traits. Our findings revealed that SDS type ZZ/ZW can co-exist with XX/XY types as PSD in the same individuals of hybrid catfish. These models can be used to study the evolution of gene networks and epistasis, thereby allowing us to investigate the developmental regulation of genes that are traditionally thought to be members of core sex signaling networks. Identifying additional sex-determining genes and their interactions can offer unique insight to infer the process of sexual development. However, our results need to be extrapolated to other catfish species to investigate the evolution of SDSs in clariid fish. Chromosome mapping using fluorescence in situ hybridization on sex-linked loci and high-quality whole genome assemblies should be performed in catfish to identify the positions of sex-determining genes/loci.