The pangenome of Acanthopagrus provides genomic variation evidence for genetic diversity

Summary

Acanthopagrus species are ecologically and economically important sparid fishes with strong environmental adaptability. To characterize genomic architecture underlying their adaptation and phenotypic diversity, a population-scale survey of intraspecific variation in yellowfin seabream (A. latus) was conducted using whole-genome resequencing data from 80 wild individuals. Among the genomic variations identified, structural variations (SVs) contribute disproportionately high to genomic diversity, defining 11,463 variable genes associated with immunity, ion transport, and environmental adaptation. Compared to the core genes with conserved functions in basic metabolism, these variable genes exhibit lower expression levels but higher transcriptional variance. Furthermore, a pangenome graph for Acanthopagrus species identifies a 24-bp deletion in the gch2 promoter of blackhead seabream (A. schlegelii) as a candidate variant for the loss of xanthophore pigmentation. This deletion disrupts a conserved motif, potentially impairing Pax7a-mediated regulation. These findings uncover genetic mechanisms driving adaptation and phenotypic divergence in Acanthopagrus species with high aquaculture values.

Graphical abstract

Highlights

• •A graph pangenome of Acanthopagrus reveals genomic diversity beyond the reference
• •Structural variants, rather than single-nucleotide polymorphisms (SNPs), dominantly shape the variable gene landscape
• •Transcriptional flexibility of variable genes is a key source of adaptive plasticity
• •Variants in gch2 may contribute to xanthophore pigmentation divergence in Acanthopagrus

Zoology; Ichthyology; Genetics; Genomics

Introduction

The genus Acanthopagrus, a cornerstone of the family Sparidae, possesses substantial economic and ecological value.¹ These fishes are euryhaline and adapt to a broad range of water salinities across coastal and estuarine habitats,^2,3 and most species are protandrous hermaphrodites.^4,5 Furthermore, Acanthopagrus species are primary targets for commercial fisheries and recreational angling, underpinning significant socio-economic interests.^6,7,8 Within this genus, yellowfin seabream (A. latus) and blackhead seabream (A. schlegelii) are of paramount economic importance. While they share many biological traits, they can be distinguished by their fin color: Yellowfin seabream features vivid yellow fins, while blackhead seabream displays black-colored fins. This phenotypic divergence is not only taxonomically significant but also serves as a critical quality trait that directly influences consumer preference and market valuation. However, the genetic architecture underlying these distinct color phenotypes remains largely uncharacterized.

Recent advancements in genomics have revolutionized our understanding of genomic variations in Acanthopagrus and other marine taxa. Linear reference genomes, assembled via next-generation sequencing (NGS) and long-read sequencing technologies, have greatly propelled population genomics research.^9,10,11 These resources facilitate precise identification of single-nucleotide polymorphisms (SNPs), the most prevalent form of genetic variation, and have been instrumental in deciphering population structure, adaptive processes, and the genetic basis of key traits.^12,13,14 For instance, SNP-based approaches have successfully identified genetic markers associated with growth rate, disease resistance, and environmental adaptation of aquaculture species, offering critical insights into aquaculture and conservation management.^15,16,17 However, while SNPs are highly informative, they capture only a fraction of total genetic diversity; other forms of genomic variations remain comparatively understudied despite their potential functional significance.

Beyond SNPs, genomic variations also encompass insertions and deletions (INDELs, <50 bp) and structural variations (SVs, >50 bp). The latter include deletions (DELs), insertions (INSs), duplications (DUPs), and inversions (INVs).^18,19,20,21 These variations have been shown to substantially influence an organism’s phenotypes by increasing genome diversity, rewiring epigenetic modifications, and altering gene transcription, thereby playing a pivotal role in adaptation and speciation.^22,23,24,25 For instance, in Chrysophrys auratus, SVs and INDELs collectively accounted for three times more variation than SNPs; this suggests they may play a more significant role in driving the genetic divergence in marine teleost genomes than SNPs.²⁶ However, single-reference genome approaches may introduce bias by overlooking non-reference alleles, leading to underestimated genetic diversity and flawed biological interpretations across species.^27,28,29

To address the limitations, pangenomics has emerged as a powerful framework for capturing the full spectrum of genetic diversity.^30,31 A pangenome represents the collective sequence entities of multiple individuals within a species or closely related taxa. It comprises core components, representing the genome regions shared across all individuals, and the accessory genome, which consists of sequences specific to certain individuals or lineages.^32,33 Among the mathematical frameworks developed for pangenome analysis, the pangenome graph has been widely used to integrate multiple genomic datasets into a single data structure.³⁴ In this model, individual-specific genetic variations are represented as unique paths, which significantly reduces the reference bias inherent in linear reference frameworks while facilitating the discovery of genetic variations associated with adaptation, disease resistance, and other critical traits.^35,36,37,38 However, the application of pangenome graphs remains largely unexplored in marine biology, despite its extensive use in studies of humans, crops, and livestock.

Here, we present the pangenome graph for the genus Acanthopagrus, integrating SNPs, INDELs, and SVs to provide a holistic view of its genetic landscape. Using the PanGenome Graph Builder (PGGB) pipeline, we constructed a reference-free, bidirected sequence variation graph incorporating genomic data from yellowfin seabream and blackhead seabream.³⁹ This resource facilitated the identification of associations between pigmentation genes and phenotypic divergence between the two species, shedding light on the genetic mechanisms driving adaptive evolution and speciation within Acanthopagrus. These findings represent a significant advancement in marine genomics, laying a strong foundation for understanding species divergence and supporting genomic-assisted selection and breeding in aquaculture.

Results

Characterization of SV in wild yellowfin seabream

From the 80 yellowfin seabream individuals, we initially detected 61,215,259 genomic variations (STAR Methods, Figure S1), including 54,304,264 SNPs, 6,709,422 INDELs, 189,852 DELs, 9,277 DUPs, and 2,444 INVs. Following filtration for a minor allele frequency (MAF) > 0.1, 33,662,926 SNPs (62.0%), 3,205,257 INDELs (47.8%), 65,898 DELs (34.7%), 4,428 DUPs (47.8%), and 1,376 INVs (56.3%) were retained for downstream analyses (Figures 1A–1F). A total of 1,639 high-confidence chromatin interaction regions were found to co-localize with SVs. Notably, 1,920 SVs were anchored at one end of an interaction, while their corresponding distal anchors overlapped with gene promoter regions (≤2 kb upstream of the transcription start site, TSS). This indicates that these SVs may affect chromatin spatial architecture and gene regulation (Figure 1G). Across most variation types, the majority (∼60.5–71.9%) occurred in intergenic and intronic regions. Only 18.1% of the variants (11,094,093) were identified in exonic and promoter regions of genes, potentially exerting direct or indirect effects on their protein coding and expression regulation (Figure 1H). In contrast, DUPs exhibited a distinct genomic distribution, with a pronounced enrichment (∼70%) within promoter regions (Figure 1H). Regarding size distribution, over 80% of the identified SVs were relatively small (<1,000 bp); nonetheless, DUPs displayed a significantly greater average length than DELs and INVs (Figure 1I).

Figure 1.

Landscape of genomic variations in yellowfin seabream (Acanthopagrus latus) population

(A–F) Circos plot illustrates the genomic distribution of five variant types (SNP, INDEL, DEL, DUP, and INV) across the 24 chromosomes. All densities were calculated using a 100 kb non-overlapping sliding window. Tracks from outer to inner: (A) Gene density (black), with darker shades representing higher density.

(B) SNP density (orange-red).

(C) INDEL density (pink).

(D) Deletion (DEL) density (orange histogram) and minor allele frequency (MAF) (scatterplot, range: 0.1–1.0).

(E) Duplication (DUP) density (green histogram) and MAF (scatterplot, range: 0.1–1.0).

(F) Inversion (INV) density (blue histogram) and MAF (scatterplot, range: 0.1–1.0).

(G) Chromatin interaction map shows significant inter- and intra-chromosomal interactions (links) identified by Hi-C data. Link colors correspond to the source chromosomes.

(H) Distribution of the five variant types across various genomic features (e.g., promoter, UTR, and exon). The y axis represents the relative proportion of each variant type within specific regions.

(I) Density plots show the length distribution for large structural variants (SVs), including DELs (orange), DUPs (green), and INVs (blue). The x axis represents the SV length (bp), and the y axis represents the density estimation.

Functional enrichment of core genes and variable genes

Core genes were defined as those lacking any high-impact variants (STAR Methods) across all individuals (all GT values are “0/0”), whereas variable genes were defined as those harboring high-impact variants (GT value ≠ “0/0”). We then performed iterative random sampling to simulate the dynamic fluctuations in core and variable gene counts relative to population size (rarefaction analysis, Figure S2). However, the resulting saturation curves failed to reach a smooth convergence, regardless of whether SNPs or SVs were utilized (Figures S3 and S4). To investigate the heterogeneity, we reconstructed a neighbor-joining (NJ) phylogenetic tree based on high-impact SNPs to screen for potential outliers. The analysis revealed 21 significant outliers— predominantly from the Zhanjiang (ZJ) population (18/21)—that formed a divergent clade (Figure S5). It suggests that these individuals experienced distinct evolutionary dynamics from the rest of the cohort, leading to the failure of rarefaction analysis.¹⁷ After removing these outliers, the rarefaction analysis yielded smooth saturation curves.

SVs achieved faster convergence in rarefaction analysis than SNPs (Figures 2A and 2B), and we obtained the final set of core genes and variable genes at a sample size of 50. Among the 11,463 variable genes identified (38.1% of total genes), 55.7% (6,380 genes) were exclusively influenced by SVs and 25.7% (2,949 genes) were only affected by SNPs (Figure 2C). Only a small proportion (2,134 genes, 18.6%) was affected by both SNPs and SVs. Functional enrichment analysis revealed distinct biological roles for core and variable genes in the growth, development, and environmental adaptation of yellowfin seabream (Figure 2D). Core genes were primarily involved in fundamental processes such as the respiratory chain, energy metabolism, ribosome biogenesis, and protein synthesis. In contrast, variable genes were significantly enriched for Gene Ontology (GO) terms related to cell junctions, the nervous system, immunity, and ion channels. Notably, there were no overlapping terms between the two groups (Figure 2D). Furthermore, core genes exhibited significantly higher expression levels and lower inter-individual variance than variable genes (Figures 2E and 2F). A linear mixed-effects model (LMM) confirmed this trend: Elevated expression of core genes persists after controlling for sample- and gene-specific random effects (estimate = 0.092, t = 4.965, p = 6.87e−7, Figure 2F).

Figure 2.

Characterization and functional divergence of core and variable genes

(A and B) Rarefaction analysis (saturation curves) depicts the relationship between the number of identified core and variable genes and sample size (n = 1 to 50). Genes are defined by (A) SNPs and (B) SVs, respectively. The curves demonstrate faster convergence and greater stability when utilizing SVs.

(C) Venn diagram illustrates the overlap of variable genes defined by SVs and SNPs. Values indicate the number of genes and their corresponding percentages relative to the total number of variable genes.

(D) Dot plot of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment for core and variable genes. The x axis denotes the gene category, dot size indicates the number of genes, and color intensity represents statistical significance (log₁₀(adjusted P-value)).

(E) Density plot shows the distribution of expression variance for core and variable genes across 56 transcriptome samples. Dashed lines indicate the mean variance for each group.

(F) Raincloud plots compare the mean expression levels (log₂(TPM+1)) between core and variable genes. Boxplots show the median (center line) and interquartile ranges (box limits). Statistical significance was determined using a linear mixed-effects model (LMM), accounting for random effects for samples and genes (Estimate = 0.092, t = 4.965, p = 6.87e−7).

Construction of the pangenome of Acanthopagrus

A linear pangenome was constructed for yellowfin seabream and its close relative, blackhead seabream. Over 70% of the genomic regions exhibited synteny across the four assemblies (Figure S6). Notably, large-scale INVs were identified at chromosome termini between the two species. For example, on chromosome 3, a 10 Mb fragment and 74 loci exceeding 10 kb displayed inverted orientations in blackhead seabream relative to yellowfin seabream. These discrepancies may stem from either authentic SV or assembly artifacts in the blackhead seabream genome, necessitating further validation (Figure 3A). Among the three yellowfin seabreams, we identified a higher density of syntenic regions with high sequence conservation, encompassing over 95% of core genes.

Figure 3.

Architecture and genomic features of the Acanthopagrus pangenome

(A) Visualization of a large-scale inversion (∼10 Mb) identified on chromosome 3 between yellowfin seabream (A. latus, fAcaLat) and blackhead seabream (A. schlegelii, fAcaSch). The linear pangenome alignment depicts syntenic blocks, inversions, and other structural rearrangements. The top track represents the coordinates along chromosome 3.

(B) Structural evaluation of the pangenome graph. The plot displays the distribution of node counts (y axis) and total sequence length (color gradient) across different genome combinations. The “U-shaped” distribution indicates that a large number of nodes are either shared by all genomes (core) or specific to a single genome (unique). The bottom panel matrix indicates the presence (black dot) or absence of each assembly within specific node subsets.

Following rigorous correction of chromosomal orientations to ensure consistency (Figures S7–S11), we constructed a pangenome graph with a total length of 806,572,107 bp, comprising 38,639,018 nodes and 52,724,247 edges. Each node is connected by an average of 1.36 edges, with a mean edge length of 21 bp. While only 32.8% of the nodes (12,656,023) are shared across all individuals, these nodes accounted for the majority (74.4%) of total nucleotides. As expected, blackhead seabream harbored the highest proportion of species-specific sequences. Intriguingly, one yellowfin seabream assembly contained approximately 46.8 Mb of unique sequences (5.9%) absent in the other two assemblies (Figure 3B). These findings underscore that integrating multiple genomes can provide a far more comprehensive landscape of sequence variation than a single reference.

Pangenome graph reveals genetic loci associated with xanthophore pigmentation of yellowfin seabream and blackhead seabream

The primary phenotypic distinctions between yellowfin seabream and blackhead seabream lie in the distribution of yellow pigmentation in pelvic, anal, and lower caudal fins (Figures 4A and 4B). To elucidate the genetic basis of this difference, we leveraged the pangenome graph to examine sequence divergence in candidate genes associated with xanthophore formation and yellow pigmentation, specifically gch2, csf1ra, pax7b, and bco2b.^40,41,42,43 While the promoter and exon regions of csf1ra, pax7b, and bco2b were highly conserved (Figure S12), several genomic variations were identified in the gch2 locus of blackhead seabream (Figures 4C–4H), including a notable 24 bp INDEL within its promoter region (Figures 4E and 4F).

Figure 4.

Genomic variation at the gch2 locus associated with fin pigmentation

(A and B) Representative photographs and close-up views of the pelvic, anal, and lower caudal fins of (a) yellowfin seabream (A. latus, yellow fins) and (B) blackhead seabream (A. schlegelii, black fins).

(C) Linear genome alignment of the gch2 locus. The top schematic depicts the gene structure, while the bottom tracks show alignment coverage for three yellowfin seabream genomes (orange) and one black seabream genome (gray). Yellow highlights denote two major genomic variations: a 24-bp deletion (INDEL) in the promoter region and a 70-bp insertion (INS) near the transcription end site (TES).

(D) Bandage visualization of the local pangenome graph topology for the gch2 gene. Blue loops represent complex variation bubbles.

(E) Close-up of the graph bubble corresponding to the 24-bp INDEL in the promoter. The two paths represent the reference (upper) and deletion (lower) alleles.

(F) Haplotype sequences aligned to the graph paths in (E), illustrating the specific 24-bp INDEL in blackhead seabream.

(G) Close-up of the graph bubble corresponding to the 70-bp INS.

(H) Haplotype sequences aligned to the graph paths in (g), showing the insertion sequence.

To determine if the sequence divergences identified by the pangenome graph were evolutionarily conserved, we extended our analysis to the genome resequencing data from an additional 24 yellowfin seabream and blackhead seabream individuals. The 24 bp INDEL exhibited complete fixation between the two species, characterized by distinct, non-overlapping distribution patterns (Figures 4E, 4F, 5A, and 5B). This finding was further validated by the sequencing results from fin rays (Figure S13). In contrast, other variants displayed polymorphic or ambiguous patterns (Figures 5C and 5D). These results highlight the superiority of graph-based genotyping in mitigating biases inherent in linear pangenomes, providing a more rigorous framework for precise variant identification (Figures 4C, 4G, 4H, 5C, and 5D).

Figure 5.

Genotypic distribution of genomic variations at the gch2 gene across yellowfin seabream and blackhead seabream populations

(A and B) Sequence Tubemap shows read alignment support for the 24-bp INDEL. Rows represent individual samples. The green gradient represents alignment scores for the reference (non-deletion) allele. (A) In yellowfin seabream populations, high alignment scores indicate the absence of the deletion (yellowfin-type: fAcaLat_1, fAcaLat_2, and fAcaLat_3 genotype).

(B) In blackhead seabream populations, high alignment scores indicate the presence of the 24-bp deletion (blackhead-type: fAcaSch_1 genotype).

(C and D) Sequence Tubemap illustrates read alignment support for the 70-bp downstream insertion, revealing a complex and ambiguous pattern.

(C) In yellowfin seabream, reads with high alignment scores are mapped to both the insertion and non-insertion paths, indicating that this locus is polymorphic and not fixed within the yellowfin seabream population.

(D) In blackhead seabream, while the highest-scoring reads supported the insertion (blackhead-type), reads supporting the non-insertion (yellowfin-type) contained numerous SNPs (indicating low sequence identity). This conflicting alignment pattern suggests genotyping ambiguity for this locus, precluding a definitive conclusion regarding its fixation.

The SV in the promoter region of gch2 as a candidate driver of xanthophore pigmentation divergence in Acanthopagrus

To assess the functional significance of the 24 bp INDEL in blackhead seabream, we performed a transcription factor binding site (TFBS) scanning within the gch promoter. A binding motif “ATTCAT” for the transcription factor Pax3/7 was identified to overlap with the 24 bp INDEL, with a JASPAR specificity score >0.85 (Figure 6A). Four domains were identified in the Pax7a protein, specifically Pax3, Homeodomain, Pax7, and OAR domains (Figure 6B). The 3D structure of the Pax3 domain was then modeled using AlphaFold3. The resulting model demonstrated high reliability (ranking score = 0.84; local pLDDT >90) and excellent stereochemical quality, with 95.4% of residues in favored regions and none in disallowed regions. Structural alignment against the established paired domain crystal structure (PDB: 1MDM) confirmed a highly conserved topological fold (TM-score >0.70; RMSD = 2.58 Å). This TM-score significantly exceeds the 0.5 threshold for structural homology, thereby validating the model for subsequent docking (Figures 6C and 6D). Molecular docking analysis showed favorable electrostatic potential compatibility at the binding interface, where the Pax3 domain forms multiple hydrogen bonds with the “ATTCAT” motif (Figures 6E and 6F). These data support the model in which Pax7a binds directly to the 24 bp INDEL sequence, suggesting that the loss of this binding site might drive the observed divergence in gch2 expression between the two species and the resulting differences in xanthophore pigmentation.

Figure 6.

Proposed molecular mechanism of gch2 cis-regulation by the Pax7a transcription factor

(A) Motif scanning of the 24-bp sequence deleted in blackhead seabream. The highlighted “ATTCAT” sequence matches the binding motif of the Pax3/7 transcription factor family (JASPAR ID: MA2114.1) with a high specificity score (0.86).

(B) Domain architecture of the yellowfin seabream Pax7a protein. Domains are color-coded: Pax3 DNA-binding domain (orange), homeodomain (green), Pax7 domain (blue), and OAR domain (yellow).

(C) Predicted 3D structure of the interaction between Pax7a protein and the 24-bp DNA fragment via molecular docking, where the Pax3 domain (orange) directly interacts with the DNA major groove.

(D) Close-up of the interface between the Pax3 domain residues and the DNA helix.

(E) Electrostatic potential surface of the Pax3 domain (ranging from −5 kT/e [red) to +5 kT/e [blue)), illustrating charge complementarity with the negatively charged DNA backbone.

(F) Detailed view of hydrogen bond interactions. Black dashed lines denote hydrogen bonds between specific amino acid residues of Pax7a and the nucleotide bases of the 24-bp deletion. Bond distances range from 2.47 Å to 3.74 Å.

Discussion

SVs mediate sequence and regulation divergence between yellowfin seabream individuals

In this study, we performed a genome-wide investigation of SVs in yellowfin seabream populations, identifying 70,712 SVs across 80 individuals. The reliability of our dataset was established through rigorous cross-validation against external datasets and PCR experiments (Figures 5A, 5B, and S13). Compared to SNPs, although non-SNP variations were less numerous than SNPs, they affect a disproportionately larger number of genes (Figures 1A–1F and 2C), which is consistent with observations in other species.^44,45,46,47 In humans, for example, rare SVs are 841- and 341-fold more likely to be strongly deleterious than rare SNPs and rare INDELs, respectively.⁴⁸ Despite their lower frequency, SVs exert a profound biological impact because a single structural event can span thousands of base pairs, frequently disrupting entire exons or critical regulatory elements. Studies in C. auratus revealed that SVs are significantly enriched in genomic regions exhibiting strong signatures of selection (e.g., extreme Tajima’s D and π values) and possess substantial predictive power for complex traits such as growth.^26,49 These findings reinforce the hypothesis that SVs are not merely genomic noise but are potent drivers of phenotypic adaptation, often carrying greater functional weight than the more numerous but smaller-scale SNPs.

In the yellowfin seabream’s genome, 13,499 SVs were located within the gene body, which may disrupt coding sequences or regulatory elements, leading to direct alterations in protein structure and function.⁵⁰ Functional enrichment analysis revealed that these SV-associated genes are predominantly involved in cell junctions, nervous system, immunity, and ion channels (Figure 2D). This enrichment pattern aligns with the “genomic island of divergence” hypothesis observed in other marine teleosts, where structural variants preferentially target loci mediating environmental interactions to drive rapid adaptation.⁵¹ Given that yellowfin seabream is a euryhaline species, the high prevalence of SVs in the genes of ion channels and cell junctions likely reflects an adaptive requirement for flexible osmoregulation and signal transduction under fluctuating salinity.⁵² Similarly, the high variability in immune-related genes suggests a mechanism for maintaining high allelic diversity to cope with diverse pathogen pressures.⁵³ Thus, SV-mediated alterations in these genes likely provide the genomic plasticity necessary for population survival in heterogeneous environments.

In contrast, core genes are primarily associated with fundamental “housekeeping” processes, including the respiratory chain, energy metabolism, ribosome biogenesis, and protein synthesis (Figure 2D). Our transcriptomic analysis further revealed that these core genes exhibited significantly higher expression abundance but lower variance than the genes harboring SVs (Figures 2E and 2F). This distinct architecture mirrors the evolutionary theory of expression level-evolutionary rate (E-R) anticorrelation, which posits that highly expressed genes are subject to intense selective constraints against deleterious mutations to avoid putative protein misfolding and non-specific interactions, which are harmful to the cell’s survival.^54,55,56

The majority of SVs in this study were located in non-coding regions (Figure 1H), which is consistent with the observation in humans, where more than 88% of causal SVs reside in non-coding regions.²³ These non-coding SVs may cause the gain or loss of cis-regulatory elements (e.g., enhancers or promoters) or alter the 3D chromatin conformation, ultimately changing gene transcription.^22,57 Specifically, the co-localization of SVs with promoter regions and chromatin interaction anchors highlights a physical mechanism by which these variants modulate long-range gene regulation. By potentially disrupting the spatial contact between enhancers and promoters, these SVs could alter the transcriptional output and expression plasticity (Figure 1G). Furthermore, such variants may affect the binding activities of TFs to promoters or distal elements, thereby driving expression differences between individuals and contributing to the rapid adaptation across diverse habitats.

Potential role of genomic variations in mediating pigmentation differences between yellowfin seabream and blackhead seabream

Phenotypic diversity in teleosts is a key driver of adaptation to heterogeneous habitats and lineage divergence. In shallow-waters environments, where short-wavelength light penetrates more effectively, fish in vibrant blue, green, and yellow colors are often highly conspicuous, facilitating their intraspecific signaling, reproductive ornamentation, or aposematic displays.⁵⁸ Gaining a comprehensive understanding of the genomic mechanisms underlying these color variations is essential for elucidating the dynamics of phenotypic evolution and environmental adaptation between different species.

In the case of yellowfin seabream and blackhead seabream, the primary phenotypic difference is localized to fin coloration (Figure 4A). We hypothesize that this pigmentation divergence is driven by distinct selective pressures associated with habitat partitioning and visual communication. The bright yellow fins of yellowfin seabream likely serve as critical visual cues for schooling or mate recognition in clear, shallow coastal waters. Similar adaptive strategies have been widely documented in other teleosts; for instance, in Lake Victoria cichlids (e.g., Pundamilia nyererei) and bluefin killifish (Lucania goodei), bright yellow and red pigmentations are strongly favored by sexual selection to enhance conspecific signaling in well-lit environments.^59,60 Conversely, the loss of yellow coloration in blackhead seabream—resulting in predominantly dark or black fins—may represent an adaptation to rocky reefs or deeper habitats. Reduced conspicuousness in these environments enhances camouflage and crypsis, thereby mitigating predation risk and conferring a fitness advantage. This strategy mirrors the evolutionary trajectories observed in other benthic and reef-associated marine organisms; for instance, juvenile shore crabs (Carcinus maenas) exhibit environmentally induced color changes that improve background matching, while reef-associated gobies (Gobiidae spp.) modulate body coloration and melanin distribution to enhance crypsis in visually complex rocky habitats.^58,61

Teleost color diversity is predominantly driven by six types of neural crest-derived chromatophores: xanthophores (yellow), erythrophores (red), melanophores (black), cyanophores (blue), leucophores (white), and iridocytes (iridescent).⁶² Among these, xanthophore pigmentation is primarily regulated by the pteridine biosynthetic pathway,⁶³ in which the gch gene serves as a key rate-limiting factor.⁶⁴ Studies in red tilapia (Oreochromis spp.) demonstrated that the spatial expression of gch1 was strongly correlated with xanthophore distribution.^65,66 Its paralog, gch2, is essential for xanthophore differentiation and pigment synthesis; gch2 deficiency in zebrafish larvae impairs pteridine production, causing xanthophores to appear pale gray despite the physical presence of these cells.⁴⁰ Conversely, gch2 overexpression in these mutants partially rescues the wild-type phenotype. In this study, we observed high sequence conservation of gch1 between yellowfin seabream and blackhead seabream, but identified seven SVs within the promoter and gene body of gch2 between the two species. Notably, a 24 bp INDEL in the blackhead seabream genome causes a loss of Pax7 transcription factor binding motif (“ATTCAT”) in the gch2 promoter, which may alter the expression of gch2 and suppress xanthophore pigmentation in blackhead seabream (Figure 6A). This finding underscores the critical role of genomic variations in driving phenotypic differentiation between the two species, and largely broadens our understanding of evolutionary divergence within the Acanthopagrus genus.

Beyond providing evolutionary insights, the high-resolution SVs catalog and pangenome graph developed in this study can also provide a technical foundation for addressing practical challenges in aquaculture.⁶⁷ This resource helps resolve taxonomic ambiguities and preserve species integrity through the identification of species-diagnostic markers. For example, the fixed 24 bp INDEL in the gch2 promoter can be utilized to verify species purity and monitor hybrid stocks.⁶⁸ Furthermore, the pangenome framework enhances stock management and traceability; by offering higher sensitivity than traditional SNP-based methods, it enables the detection of population-specific signatures necessary to distinguish wild stocks from farmed escapees.²⁶ The pangenome graph also aids in mitigating inbreeding depression in hatchery-reared populations by identifying deleterious SVs and large-scale genomic erosions, which often accumulate under intense selection and are frequently overlooked by single linear reference genomes.⁶⁹ Finally, this work accelerates targeted molecular breeding by localizing SVs within functional gene clusters associated with cell junctions, the nervous system, and immunity, alongside identifying candidate loci for pigmentation. These SV-anchored genes represent a prioritized candidate pool for marker-assisted selection and fast-tracking genetic improvement in the aquaculture of yellowfin seabream and black seabream.⁷⁰

Limitations of the study

Despite the insights provided by this pangenomic analysis, several limitations remain. First, the pangenome graph was constructed exclusively using short-read sequencing data, which may limit the discovery and characterization of SVs in complex repetitive regions of low sequence mappability. The future integration of long-read sequencing (e.g., PacBio or Oxford Nanopore) will be essential to overcome this constraint and improve the structural resolution of the pangenome. Second, while our current sample size provided a robust overview, particularly for yellowfin seabream, it may lack sufficient power to capture low-frequency accessory variants in black seabream. Expanding the sample size across a broader geographic range would enhance the detection of rare genetic variations and provide a higher-fidelity map of the evolutionary landscape of these species. Finally, our findings are primarily based on in silico bioinformatic analyses. Further larger-scale PCR and gene editing validation will be necessary to confirm the functional impact of these identified SVs in the adaptation and speciation of Acanthopagrus species.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jianguo Lu (lujianguo@mail.sysu.edu.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data: All 80 whole-genome resequencing datasets reported in this study have been deposited at NCBI and are publicly available. Accession numbers are listed in the key resources table. The Hi-C dataset is available under the SRA accession number SRR12328045. The 56 transcriptome datasets of yellowfin seabream are shown in Table S3. The remaining 24 resequencing datasets consist of 11 yellowfin seabreams and 13 blackhead seabreams, as shown in Table S4.
• •Code: All original code is available in this paper’s supplemental information (Data S1).
• •All other items: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported in part by the R&D Project for Jinwan Yellowfin Seabream Breeding System Construction [no. K20-42000-018], Project supported by Innovation Group Project of Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai) [no. 311021006], R&D Project for Jinwan Yellowfin Seabream Breeding System Construction, China, Innovation Group Project of Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai), China, and National Natural Science Foundation of China [no. 31902427]. The funding bodies are not involved in the design of the study, the collection, analysis, and interpretation of data, and the writing the manuscript.

Author contributions

Conceptualization: J. L. and Y. Y. Investigation: Y. H., W. W., and X. Z. Data analysis and visualization: Y. H., H. W. and Z. G. Funding: J. L. Writing: Y. H. and J. L. All authors read and approved the final manuscript.

Declaration of interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Yellowfin seabream (A. latus) fin tissues	This paper	Sampling sites: Zhuhai, Guangdong, China
Blackhead seabream (A. schlegelii) fin tissues	This paper	Sampling sites: Zhuhai, Guangdong, China
Chemicals, peptides, and recombinant proteins
Eugenol	Sigma-Aldrich	N/A
AMPure XP beads (Agencourt SPRIselect)	Beckman Coulter	N/A
Critical commercial assays
Magnetic bead–based genomic DNA extraction kit	Tiangen Biotech (Beijing)	DP705
NEBNext® Ultra™ II DNA Library Prep Kit	New England Biolabs (NEB)	E7645
E.Z.N.A.® Tissue DNA Kit	OMEGA Bio-tek	D3396
2 × Taq Plus Master Mix (Dye Plus)	Vazyme	P211
FastPure Gel DNA Extraction Mini Kit	Vazyme	DC301
Deposited data
Raw whole-genome resequencing data (A. latus)	This paper	NCBI BioProject: PRJNA1380438, Table S1
Genome assembly (A. latus & A. schlegelii)	NCBI GenBank	See Table S2
Transcriptome datasets (56 samples)	NCBI SRA	See Table S3
Published resequencing data (A. latus & A. schlegelii, 24 samples)	NCBI SRA	See Table S4
Hi-C datasets (A. latus)	NCBI SRA	SRR12328045
Software and algorithms
iSeq (v1.1.0)	Chao et al.71	https://github.com/BioOmics/iSeq
fastp (v0.23.4)	Chen et al.72	https://github.com/OpenGene/fastp
BWA-MEM (v0.7.17)	Li,73	http://bio-bwa.sourceforge.net/
Sambamba (v1.0.1)	Tarasov et al.74	https://github.com/biod/sambamba
DeepVariant (v1.5.0)	Poplin et al.75	https://github.com/google/deepvariant
GLnexus (v1.4.1)	Yun et al.76	https://github.com/dnanexus-rnd/GLnexus
SpeedSeq (v0.1.2)	Chiang et al.77	https://github.com/hall-lab/speedseq
Lumpy-SV (v0.3.1)	Layer et al.78	https://github.com/arq5x/lumpy-sv
CNVnator (v0.3.3)	Abyzov et al.79	https://github.com/abyzovlab/CNVnator
SVTyper (v0.7.1)	Chiang et al.77	https://github.com/hall-lab/svtyper
Svtools (v0.5.1)	Larson et al.80	https://github.com/hall-lab/svtools
Bcftools (v1.21)	Li,81	http://samtools.github.io/bcftools/
snpEff (v5.1)	Cingolani et al.82	http://pcingola.github.io/SnpEff/
GenomicRanges (R package, v1.54.1)	Lawrence et al.83	https://bioconductor.org/packages/GenomicRanges
Circos (v0.69.9)	Krzywinski et al.84	http://circos.ca/
ChIPseeker (v1.42.0)	Yu et al.85	https://bioconductor.org/packages/ChIPseeker
Trim Galore (v0.6.10)	Babraham Bioinformatics,86,87	https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/
HiC-Pro (v3.1.0)	Servant et al.88	https://github.com/nservant/HiC-Pro
Armatus (v2.3)	Filippova et al.89	https://github.com/kingsfordgroup/armatus
Chrom3D (v1.0.2)	Paulsen et al.90,91	https://github.com/Chrom3D/Chrom3D
EggNOG-mapper (v6.0))	Hernández-Plaza et al.92	https://github.com/eggnogdb/eggnog-mapper
AnnotationForge (R package, v1.48.0)	Carlson & Pagès,93	https://bioconductor.org/packages/AnnotationForge
clusterProfiler (R package, v4.18.1)	Xu et al.94	https://bioconductor.org/packages/clusterProfiler
ggplot2 (R package, v4.0.0)	Wickham,95	https://ggplot2.tidyverse.org/
Trimmomatic (v0.40)	Bolger et al.96	https://github.com/usadellab/Trimmomatic
HISAT2 (v.2.2.1)	Kim et al.97	https://daehwankimlab.github.io/hisat2/
StringTie (v3.0.0)	Pertea et al.98	https://ccb.jhu.edu/software/stringtie/
lme4 (R package, v1.1-37)	Bates et al.99	https://cran.r-project.org/web/packages/lme4/
lmerTest (R package, v3.1-3)	Kuznetsova et al.100	https://cran.r-project.org/web/packages/lmerTest/
minimap2 (v2.27)	Li,101	https://github.com/lh3/minimap2
fixchr (default version)	Goel et al.102	https://github.com/schneebergerlab/fixchr
SyRI (v1.7.1)	Goel et al.102	https://github.com/schneebergerlab/syri
plotsr (v1.1.0)	Goel et al.103	https://github.com/schneebergerlab/plotsr
PGGB (v0.7.2)	Garrison et al.104	https://github.com/pangenome/pggb
PanSN-spec (v0.1.0)	N/A	https://github.com/pangenome/PanSN-spec
MashMap (v3.1.3)	Jain et al.105	https://github.com/marbl/MashMap
wfmash (0.9.2)	Guarracino et al.106	https://github.com/waveygang/wfmash
Seqwish (v0.7.11)	Garrison,107	https://github.com/ekg/seqwish
smoothxg (v0.6.6)	Garrison et al.108	https://github.com/pangenome/smoothxg
GFAffix (v0.2.1)	codia-lab,109	https://github.com/codialab/GFAffix
vg toolkit (v1.57.0)	Hickey et al.110	https://github.com/vgteam/vg
Bandage-NG (v2025.4.1)	Wick et al.111	https://github.com/asl/BandageNG
ODGI (v0.9.3)	Guarracino et al.112	https://github.com/pangenome/odgi
SequenceTubeMap (default version)	Beyer et al.113	https://github.com/vgteam/sequenceTubeMap
Miniport (v0.18)	Li,114	https://github.com/lh3/miniprot
BLAST+ (v2.16.0)	Camacho et al.115	https://blast.ncbi.nlm.nih.gov/Blast.cgi
gffread (v0.12.7)	Pertea and Pertea,116	https://github.com/gpertea/gffread
JASPAR (database online service)	Castro-Mondragon et al.117	https://jaspar.elixir.no/
InterPro (online service)	Paysan-Lafosse et al.118	https://www.ebi.ac.uk/interpro/
AlphaFold3 (online service)	Abramson et al.119	https://alphafoldserver.com/
SAVES (v6.1)	UCLA-DOE LAB	https://saves.mbi.ucla.edu/
TM-align (online service)	Zhang et al.120	https://aideepmed.com/TM-align/
UCSF ChimeraX (v1.10.1)	Meng et al.121	https://www.cgl.ucsf.edu/chimerax/

Experimental model and study participant details

The yellowfin seabream and blackhead seabream samples used in this study were sourced from fishery production. All experimental fish were 2-year-old individuals at the functional hermaphroditic stage, characterized by the simultaneous presence of both testicular and ovarian tissues. All animal handling and experimental procedures complied with the regulations and were approved by the Experimental Animal Ethics Committee of Sun Yat-sen University.

Method details

Sampling and DNA sequencing

A total of 80 yellowfin seabreams were sampled from four coastal sites (Table S1). Fish were anesthetized with 10 mg/L eugenol, and pectoral fin tissues were collected and preserved in ethanol for subsequent DNA extraction. Genomic DNA was extracted using a Magnetic Universal Genomic DNA Kit (DP705, Tiangen Biotech, Beijing, China). DNA quality and concentration were assessed via NanoDrop spectrophotometer and Qubit 3.0 fluorometer. Samples yielding >1.5 μg of DNA were fragmented to an average size of ∼350 bp using a Covaris ultrasonicator. Sequencing libraries were constructed using the NEBNext® Ultra™ II DNA Library Prep Kit (New England Biolabs, E7645, Ipswich, MA, USA), purified with AMPure XP beads (Agencourt SPRIselect, Brea, CA, USA), and PCR-amplified. Library quality was verified by quantitative PCR, and libraries with an effective concentration of 3 nM were then sequenced on the Illumina HiSeq 2500 platform (150-bp paired-end reads). The average sequencing depth per individual exceeded 10× with a genome coverage rate of at least 96%, ensuring the high data quality for downstream analyses.

Genomic variations identification using whole-genome resequencing data

Genome resequencing datasets for yellowfin seabream and blackhead seabream were retrieved from the NCBI, with data integrity verified using iSeq. Raw reads were filtered for quality using fastp (-z 4 -q 20 -u 30 -n 5), and aligned to the reference genome of yellowfin seabream (GCF_904848185.1, Table S2) using BWA-MEM. Alignment results were sorted, deduplicated, and indexed using Sambamba. SNPs and INDELs were called for each sample using DeepVariant (--model_type=WGS) and merged via GLnexus for joint variant calling across the population. To ensure robust SVs (DEL, DUP, and INV) discovery, the SpeedSeq pipeline was employed, which utilizes a multi-signal integration strategy to minimize false positives and negatives. Specifically, Lumpy-SV was used to integrate discordant read-pair (RP) and split-read (SR) signals to cross-validate breakpoints. Split reads were extracted from BAM files using the extractSplitReads_BwaMem script. Read-depth (RD) analysis was conducted to capture copy number variants (CNVs) potentially missed by RP/SR signals. Sample genotypes were inferred using SVTyper, and the resulting variants were merged and sorted using the lmerge and lsort commands of svtools. Finally, svtools was used for population-scale genotyping, copy number annotation for non-breakends (BND) variants, and redundant variants filtering.

Filtering and annotation of population variation

For each yellowfin seabream population, rare variants with a MAF <0.1 were filtered using Bcftools. SNPs and INDELs were annotated using snpEff, with a custom database constructed from the NCBI RefSeq genome assembly GCF_904848185.1 and its corresponding gene annotation. SVs, except for BND variants, were annotated using the findOverlaps function of the GenomicRanges R package. Genomic distribution and allele frequency profiles of all five variant types (SNPs, INDELs, DELs, DUPs and INVs) were analyzed via custom scripts, and visualized using Circos. Variant locations in the genome were obtained using the ChIPseeker package.

Inference of significant interactions from Hi-C data

Hi-C data (data and code availability) of yellowfin seabream were first quality-controlled using Trim Galore. To achieve a balance between genomic resolution and signal-to-noise ratio, a 40 kb resolution was used for detecting genome-wide chromatin interactions using the HiC-Pro suite. The Armatus tool was then applied to call topologically associating domains (TADs) for each chromosome. These identified TADs were converted into a segmented genome using Chrom3D, and only the significant interactions between genomic segments (FDR < 0.05) were retained for visualization.

Identification of core genes and variable genes

SNPs and INDELs categorized as ‘HIGH IMPACT’ by snpEff, along with SVs overlapping gene promoters or exons, were defined as high-impact variants. These variants were used to classify genes across populations. As illustrated in Figure S2, rarefaction analysis was performed to robustly distinguish core and variable genes. For each sample size (n, ranging from 2 to 50), we randomly drew individuals from the total population (N=80) 1,000 times. In each iteration, genes maintaining a reference genotype (GT=0/0) across all sampled individuals were identified. The final core gene set (Cⁿ) was defined as the intersection of these sets across all 1,000 iterations ($\cap L_i^n$), ensuring the rigorous identification of stable genes. The variable gene sets (Vⁿ) was obtained as the set different between the total gene universe and the core set (U-Cⁿ). The convergence of core and variable gene counts was evaluated as sampling size (n) increased; the final core and variable gene sets were obtained at the saturation point of n=50 and utilized for all subsequent analyses.

Enrichment analysis of gene function

Functional annotation of yellowfin seabream genome was performed by searching protein sequences against the eggNOG database, followed by a construction of a custom gene annotation database using the AnnotationForge package. GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were conducted for genes of interest using the clusterProfiler R package, with adjusted p-value cutoff of 0.05 and q-value cutoff of 0.01. Enrichment results were visualized using the ggplot2 package.

Expression analysis of core genes and variable genes

Fifty-six transcriptome datasets of yellowfin seabream under normal growth conditions were retrieved from NCBI (Table S3). For each sample, raw read quality was assessed with fastp, and low-quality bases and adapter sequences were trimmed using Trimmomatic. High-quality reads were aligned to the reference genome of yellowfin seabream using HISAT2. The mapped reads were further assembled and quantified via StringTie. Gene expression levels were normalized to Transcripts Per Million (TPM) to ensure comparisons across samples and genes. TPM values were log2-transformed to stabilize variance. To rigorously assess the expression differences between core and variable genes, we constructed an LMM using the lme4 and lmer Test R packages. In this model, gene category (core vs. variable) was specified as a fixed effect, while sample identity and gene ID were included as random effects (formula: expr ∼ category + (1|sample) + (1|gene)) to account for inter-individual variability and baseline expression heterogeneity.

Construction of the liner pangenome of Acanthopagrus

Autosome sequences were extracted from three chromosome-level genome assemblies of yellowfin seabream and one gap-free genome assembly of blackhead seabream (Table S2). All-versus-all pairwise genome alignments were conducted using minimap2 (-ax asm5 –eqx). Homologous chromosomes between each pair of genomes were identified using fixchr, and the chromosomes were reoriented where necessary (Figures S4–S8). Synteny and SVs were then detected for each genome pair using SyRI, and the linear pangenome was visualized using plotsr.

Construction of the pangenome graph of Acanthopagrus

The Acanthopagrus pangenome graph was established using the PGGB pipeline, which fulfills two primary objectives: (1) creating a reference-free, unbiased graph structure derived from symmetric all-to-all alignments for accurate representation of non-reference sequences, and (2) preserving base-level resolution throughout the graph. The four assemblies were renamed according to PanSN-spec conventions (Table S2), and partitioned by chromosomes using the partition-before-pggb. Corresponding chromosomes from the four assemblies were aligned using the MashMap of wfmash (-s 5000 -l 25000 -p 90 -n 1 -k 19 -H 0.001 -Y) to assess sequence similarity. Base-level alignments were performed using the wavefront alignment algorithm. Pangenome variation graph was built for each chromosome using seqwish (-k 23 -f 0 -B 10000000), and subsequently normalized using smoothxg (--chop-to 100 -I .9000 -R 0 -j 0 -e 0 -l 700,900,1100 -P 1,19,39,3,81,1 -O 0.001 -Y 300) and GFAffix to refine the graph structure. These graphs were converted to VCF files using the vg deconstruct with default parameters, and indexed for downstream genotyping. Short-read data from 24 samples (Table S4) were mapped to these graphs using vg giraffe with default parameters to generate Graph Alignment Map files. Read support was computed using vg pack, filtering for mapping quality ≥5 and base quality ≥5 to ensure data quality. Finally, variants were called for each sample using vg call, leveraging the graph snarl structure and specifying 'GCF_904848185.1′ as the reference path. Graphs were visualized using Bandage-NG, ODGI and SequenceTubeMap.

Verification of gene structure of gch2

Genomic DNA (gDNA) was extracted from the caudal fins of three yellowfin seabreams and three blackhead seabreams, respectively, using the E.Z.N.A.® Tissue DNA Kit (D3396, OMEGA, USA). DNA concentration and quality were assessed using a NanoDrop-2000 spectrophotometer. PCR amplifications were performed using 2 × Taq Plus Master Mix (Dye Plus) (P211, Vazyme, China). Primers used for PCR are listed in Table S5. Amplification was performed in a 40 μL reaction volume containing 50 ng of gDNA, 0.5 μM of each primer, and 20 μL of 2 × Taq Plus Master Mix. The thermal cycling conditions included an initial denaturation at 95°C for 3 min; followed by 40 cycles of denaturation at 95°C for 15 s, annealing at 58°C for 15 s, and extension at 72°C for 30 s; and a final extension at 72°C for 10 min. Target products were verified via 1.5% agarose gel electrophoresis, purified using the FastPure Gel DNA Extraction Mini Kit (Vazyme, DC301, China), and subjected to Sanger sequencing using the corresponding primers.

Protein sequence prediction and structural modeling

Using the protein sequence of GCF_904848185.1 as a reference (Table S2), sequence alignment and homology-based annotation was performed for the fAcaSch_1 assembly using miniprot. Protein sequences of blackhead seabream were derived from the predicted CDS using the gffread utility from the Cufflinks. Transcription factor binding motifs were scanned against the JASPAR database, and protein domains were predicted via the InterPro web service. Three-dimensional protein structures were modeled using AlphaFold3, with model reliability evaluated through predicted Local Distance Difference Test (pLDDT) and ranking scores. The stereochemical quality of the predicted structures was validated using Structure Analysis and Verification Server (SAVES, UCLA), and structural conservation was assessed by aligning the models against the reference crystal structure (PDB: 1MDM) using TM-align. Molecular docking simulations and structural visualizations, including electrostatic potential analysis and hydrogen bond identification, were performed using UCSF ChimeraX.

Quantification and statistical analysis

For the population-scale genomic variation survey, the sample size n=80 represents the number of wild yellowfin seabream individuals. For the gene expression analysis, n=56 represents the number of independent transcriptome datasets. To compare the expression levels between core and variable genes (Figure 2F), a Linear Mixed-Effects Model was implemented using the lme4 and lmerTest R packages to account for sample identity and gene ID as random effects. Statistical significance for functional enrichment (GO and KEGG) was determined using the clusterProfiler R package, applying an adjusted p-value cutoff of 0.05 and a q-value cutoff of 0.01. In all relevant figures (e.g., Figures 1I, 2E, and 2F), the definition of center and dispersion is provided: box plots represent the median (center line) and interquartile ranges (box limits). Exact p-values, estimates, and t-values are specified directly within the figures or their corresponding legends.

Published: March 30, 2026