INTRODUCTION

The scombrid species of fish inhabiting the epipelagic realm have streamlined bodies with forked or lunate tails, allowing them to move efficiently over long distances for feeding and migration.¹ The Scombridae forms the largest and state-of-art family in Scombroidei.^2–4 It includes commercially valuable, edible, and recreational fish, and there are currently 51 valid species.^1,5 There are 22 species belonging to 11 genera of the family Scombridae recorded in China.⁶ Previous studies on the taxonomy of scombroid fishes have mainly relied on traditional morphology.^2,7,8 The currently accepted classification is primarily based on the work of Collette and Monsch.^2,9 Collette² summarized available data on the scombroid fishes and divided the Scombridae into two subfamilies, Gasterochismatinae and Scombrinae, with only one genus (Gasterochisma) under Gasterochismatinae. The subfamily Scombrinae consists of four tribes: mackerels (Scombrini), Spanish mackerels (Scomberomorini), bonitos (Sardini), and tunas (Thunnini) (Figure 1a). Monsch⁹ proposed that the family Scombridae consisted of subfamilies Scombrinae, Scomberomorinae, Sardinae (tribes Sardini and Thunnini), Acanthocybiinae and Xiphiinae (billfishes, tribes Xiphiini and Istiophorini) (Figure 1b).

Figure 1.The cladistic hypothesis of the Scombroidei. (a) from Collette (1984)²; (b) from Monsch (2000).⁹

Despite numerous morphological studies, unresolved taxonomic issues remain within the family. For instance, there has been disagreement regarding the hypothesis that Gasterochisma is a sister group to the Scombrinae.^10,11 The earliest differentiated clade within the Scombridae is also subject to debate. Collette and Nauen⁸ considered that the Scombrini was the earliest differentiated group within Scombridae, whereas Johnson¹² alternatively proposed the hypothesis that Grammatorcynus was the earliest differentiated group. Additionally, the monophyletic status of Sardini and Thunnini has been disputed.¹³ The instability of these classifications confounds the management of scombroid fishes as a fisheries resource and the understanding of the diversity and complexity of marine communities.

Many studies have recently been published on the phylogenetic relationships of the Scombridae. For instance, Block et al.¹⁴ amplified the cytochrome b gene (Cytb) from 75 scombrid individuals, supporting the monophyly of endothermic tunas (Auxis, Euthynnus, Katsuwonus, and Thunnus) and suggesting Gasterochisma melampus falls within the radiation that included primitive scombrids (Scomber) and gempylids (Ruvettus). Dalziel et al.¹⁵ discovered that the bonitos (tribe Sardini) were embedded within Thunnini based on the phylogenetic tree constructed by mitochondrial cytochrome c oxidase (COX II). Orrell et al.¹⁶ used a single-copy nuclear Tmo-4C4 gene. They combined Tmo-4C4 with mitochondrial DNA cytochrome b (Cytb) to analyze phylogenetic relationships in the Scombridae, revealing that Gasterochismatinae and Scombrinae were sister groups. Jondeung and Karinthanyakit¹⁷ concluded that the genera Sarda and Thunnus were sister taxa to each other, according to the analysis of mitochondrial genes Cytb and nicotinamide adenine dinucleotide dehydrogenase subunit 2 (nd2). Despite numerous attempts to resolve the evolutionary relationships within the Scombridae using both morphological and molecular data, the phylogenetic relationships of the scombrids remain unresolved.^18–21

Furthermore, previous studies of scombroid relationships have mainly focused on composite taxa corresponding to either genera or tribes. For example, Chow and Kishino²² investigated the phylogenetic relationships among species exclusively within the genus Tunas using Cytb and adenosine triphosphate (ATP) (two mitochondrial genes) and internal transcribed spacer (ITS) region (one nuclear gene). It was reported that gene trees derived from mitochondrial DNA (mtDNA) and nuclear DNA (nDNA) data yielded concordant phylogenies supporting the genus Sarda’s monophyly.²³ Jeena et al.⁵ comprehensively reported phylogenetic relationships within the genus Scomberomorus using eight mitochondrial genes and three nuclear genes. However, to accurately determine the earliest diverging clade within the Scombridae, it is essential to consider the full range of taxa within the family. Focusing solely on a subset of species or genera may lead to incomplete or biased conclusions about evolutionary relationships. At present, only a few studies have exclusively evaluated the whole family. Hence, there is a critical need to investigate the phylogenetic relationships of Scombridae as a whole.

Most polymorphisms in mitochondrial genes are primarily concentrated in the D-loop region, exhibiting 5 to 10 times more significant variability than other mtDNA segments, rendering it an ideal locus for genetic differentiation studies.²⁴ Lee et al.²⁵ observed significant length variations in the D-loop structure when examining closely related teleost fish species, suggesting its suitability for rapid evolutionary analyses of species. Mitochondrial cytochrome b (Cytb) gene and CO1 were widely utilized sequences in research, extensively employed in systematic studies to resolve taxonomic discrepancies across various hierarchical levels.²⁶ Due to their robust applicability, these two genes have been extensively employed in phylogenetic investigations across diverse fish taxa. Due to its rapid variation and abundance of informative sites, the ITS sequence has emerged as a significant molecular marker in the phylogeny and taxonomy of lower taxonomic levels.²⁷ It can provide valuable information for distinguishing between closely related species and resolving evolutionary relationships.

This study aimed to investigate the phylogenetic relationships of Scombridae. For this purpose, we used one nuclear (ITS) and three mitochondrial DNA loci (CO1, Cytb, and D-loop) to reconstruct phylogenetic relationships of 48 Scombridae species from 14 genera combined with many sequences obtained from GenBank. The results provide valuable basic information and a greater understanding of Scombridae’s classification system and phylogenetic relationship.

MATERIALS AND METHODS

Samples Collection and Ethics

Seventeen individuals of Scombridae were sampled. All the fish samples were bought from local fish dealers in Fude Market in Nanning, Guangxi province, China. All samples were preserved in 95% ethanol and transported to the laboratory to study coral reefs in the South China Sea, Guangxi University. They were identified at the family, genus, or species levels. The samples were morphologically identified concerning the “Key to Marine and Estuarial Fish of China” and the FishBase database (https://www.fishbase.se/search.php). As food fishes, no permits were required for sampling. All procedures followed corresponding regulations, by-laws, and Ethics.

DNA extraction, PCR amplification and sequencing

Genomic DNA was extracted from approximately 20 mg of muscle tissue in 95% ethanol using a marine animal tissue genomic DNA kit (Tiangen Biotech, China). Primers were adapted from the literature (Table 1). PCR amplifications were performed in 50 µL volumes of 1 μL of DNA template, 2.5 μL of each primer (10 mmol/L), 25 μL 2× PCR master mix, and 19 μL ddH₂O. We amplified and directly sequenced four genetic markers for all samples: CO1, Cytb, D-loop (mitochondrial markers), and ITS (nuclear marker).^28–31 The details of primers are shown in Table 1. All PCR products were subjected to 1% agarose gel electrophoresis to confirm amplification. Sangon Biotech (Shanghai, China) sequenced PCR products with single intense bands based on Sanger sequencing technology.

Sequence analysis

Additional sequences from 40 species of 14 genera in the family were downloaded from GenBank (Table 2). Amplified sequences were aligned using the Clustal W algorithm,³² manually edited using Seqman v. 7.1.0 to remove gaps and severe base pair mismatches. All generated sequences were compared to the sequences of nucleotide databases using the search tool BLAST.³³ After confirming the accuracy of sequences, all amplified sequences were deposited in GenBank. Base composition, transition/transversion ratios, and the analysis of molecular diversity indicators, such as polymorphic sites, conserved sites, parsimony-informative sites, and singleton sites, were determined using MEGA X.³⁴ Distance matrices from aligned nucleotide sequences were determined by applying the Kimura 2-parameter and the Tamura-Nei model using the pairwise distance between genera calculation of MEGA v. 7.0.^35,36 DAMBE v. 5.3.19 was used to analyze substitution saturation. A scatter plot was constructed with F84 model-corrected distances as the horizontal axis and the number of transitions and transversions as the vertical axis.³⁷

Molecular phylogenetic analysis

We selected Carangidae (Trachurus trachurus) and Percichthyidae (Lateolabrax japonicus), closely related to Scombridae, as the outgroup taxa. A partition-homogeneity test was run in PAUP 4.0b10³⁸ to examine whether the sequences from the four loci should be combined in a single dataset analysis. No conflicting phylogenetic signals between the datasets were detected. The multigene analysis focused on increasing the phylogenetic signals, thereby verifying the validity of the primary investigation results. For phylogenetic analyses, CO1, Cytb, D-loop, and ITS region sequences were concatenated into one partitioned dataset. Three methods, neighbor-joining (NJ), maximum likelihood (ML), and Bayesian inference (BI), were employed to reconstruct the phylogenetic relationships within Scombridae. Phylogenies were constructed using the NJ tree through MEGA 7.0.26, the ML tree via RAxML, and the BI tree executed with MrBayes 3.2.^35,39,40

Three phylogenetic tree optimality criteria were employed. The NJ analysis was done in two-parameter mode.⁴¹ We selected the best-fitting models for ML and BI using the Akaike Information Criterion (AIC) as implemented in MrModeltest 1.0b.^42–44 The model GTR+I+G was suggested as the best fit for CO1 and Cytb; the model GTR+G was the optimal model for D-loop and ITS. BI analysis used four independent MCMC chains run simultaneously for 10 million generations while sampling one tree per 1000 replicates. Bayesian posterior probabilities (BPP), the frequencies of nodal resolution, were mapped on the BI tree. We visually inspected the trace files in Tracer v. 1.7⁴⁵ to verify that the chains had reached convergence. For ML, nodal support was assessed using a nonparametric bootstrap sampling of 1000 replicates. The output trees were further edited by Figtree v. 1.4.3 software (http://influenza.bio.ed.ac.uk/software/Figtree).

RESULTS

Sample identification

According to the morphological characteristics, 17 fish samples representing 5 genera 8 species were identified as Sarda orientalis (Linnaeus, 1758), Euthynnus affinis (Cantor, 1850), Scomber japonicas (Honttuyn, 1782), Scomber australasicus (Cuvier, 1832), Scomberomorus niphonius (Cuvier, 1832), Rastrelliger kanagurta (Cuvier, 1816), Scomberomorus commerson (Lacepède, 1800), and Scomberomorus guttatus (Bloch & Schneider, 1801). The morphology of eight species was presented in the Supporting Information (Supplementary Figure S1)

SEQUENCE FEATURES

The concatenated dataset comprised 1619 aligned sites: 575 bp from CO1, 311 bp from Cytb, 216 bp from D-loop, and 517 bp from ITS. The sequences downloaded by amplification and database in this study included 46 CO1 sequences, 48 Cytb sequences, 28 D-loop sequences and 14 ITS sequences (Table 2). Through the analysis of individual gene datasets (the sequence by PCR amplification and download from GenBank), the mtDNA (CO1, Cytb , and D-loop) exhibited a bias toward A+T in base composition, with percentages of 52.9%, 52.8%, and 62.7%, respectively. The ITS regions displayed A+T base content of 30.1% (Table 3). The results of the saturation analysis revealed that the base mutations in the single gene were suitable for phylogenetic analyses and did not reach saturation (Figure 2).

Figure 2.Saturation plot for transition and transversion of gene sequences. (a) CO1; (b) Cytb; (c) D-loop; (d) ITS

The crosses (×) show the number of transition events, and the triangles (△) show the number of transversion events. The x-axis shows the genetic distance based on the F84 model, and the y-axis shows the proportion of transitions or transversions, which is calculated by multiplying the number of transitions or transversions by the sequence length. The curves show the trends of the variance of transitions and transversions with the genetic distance increasing.

SPECIES DIVERGENCE

Using Tamura-Nei mean values, a genetic pairwise distance matrix was generated between the 14 genera (Table 4). The inter-genera genetic distances ranged from 0.086 to 0.253, with an average genetic distance of 0.154. The smallest genetic distance was observed between Sarda and Allothunnus (0.086), and the largest genetic distance was observed between Rastrelliger and Gymnosarda (0.253).

PHYLOGENETIC RELATIONSHIPS

The trees constructed by analyses of the concatenated data using the methods of NJ, ML, and BI were consistent for well-supported nodes (Figures 3-5). The low nodal support values in Figures 3-5 were mainly reflected in the branch nodes above the genus level. The source file of three phylogenetic trees was in supplementary files. The Scombridae formed three monophyletic groups, one consisting of Grammatorcynus (BI BPP = 1.0, ML BS = 100% and NJ BS =100%), another consisting of Rastrelliger and Scomber (BI BPP = 1.0, ML BS = 80% and NJ BS =68%) and the other comprising Scomberomorus, Acanthocybium, Sarda, Auxis, Katsuwonus, Euthynnus, and Thunnus (BI BPP = 0.563, ML BS = 42%, and NJ BS = 46%) (Figures 3-5).

Figure 3.Neighbor-joining (NJ) tree derived from concatenated sequence dataset (CO1+Cytb+D-loop+ITS). Nodal support values are indicated on the branches. The specific symbol (★) denotes the samples in this study and the symbol (▲) indicates some special species in result. The scale represents the length of the genetic distance.

Figure 4.Maximum likelihood (ML) tree derived from concatenated sequence dataset (CO1+Cytb+D-loop+ITS). Nodal support values are indicated on the branches. The specific symbol (★) denotes the samples in this study and the symbol (▲) indicates some special species in result. The scale represents the length of the genetic distance.

Figure 5.Bayesian inference (BI) phylogenetic hypothesis of Scombridae relationships based on analysis of the concatenated dataset (CO1+Cytb+D-loop+ITS) using Mrbyes 3.2. Nodal support values are indicated on the branches. The specific symbol (★) denotes the samples in this study and the symbol (▲) indicates some unique species in result. The scale represents the length of the genetic distance.

All results revealed that Grammatorcynus bilineatus and Grammatorcynus bicarinatus did not cluster with others in the family Scombridae, forming a distinct genus clade (Figures 3-5). The NJ tree believed that Scombrini was the earliest differentiated group within Scombididae (Figure 3).

Rastrelliger and Scomber were recovered as sister taxa, forming the Scombrini. The monophyly of the Scombrini was highly supported in both sets of analyses (BI BPP = 1.0, ML BS = 80%) (Figure 4 and 5). Thunnus was a highly monophyletic group, and Auxis+Katsuwonus+Euthynnus (BI BPP = 1.0, ML BS = 96%, and NJ BS = 96%) were sister groups to each other, forming the Thunnini (Figures 3-5). Within Thunnus, we recovered a deep split between a clade with T. orientalis and T. alalunga, and a subclade containing all remaining Thunnus species (Thunnini thynnus, Thunnini maccoyii, Thunnini obesus, Thunnini tonggol, Thunnini albacares, and Thunnini atlanticus). Katsuwonus + Auxis + Euthynnus was the next lineage to branch off, followed by a clade formed by all extant Thunnus. The Sardini consists of Sarda, Cybiosarda, and Allothunnus. However, the evolutionary relationship between Acanthocybium solandri and Gymnosarda unicolor was doubtful. The NJ Tree showed that G .unicolor belonged to the Sardini, and A. solandri differentiated separately (Figure 3); the ML tree revealed that A. solandri converged into one clade with G. unicolor (ML BS = 56%) (Figure 4); the BI tree indicated that G. unicolor was clustered with Scomberomorus cavalla, and A. solandri was a sister group to the rest of Scomberomorus. Two major clades were within the Scomberomorus subclade (1.0 BPP) (Figure 5). The topology of the three evolutionary trees was not precisely consistent. Still, they all indicated a common class group, the regalis group: S. tritor, S. maculatus, S. concolor, S. sierra, S. brasiliensis, and S. regalis (Figures 3-5).

DISCUSSION

Gasterochismatinae is not a sister group to Scombrinae

Our results all supported that Gasterochisma and Scombrinae are not sister taxa. The phylogenetic placement of Gasterochisma has been debated due to the presence of distinctive primitive features, as well as uniquely derived characteristics, making it difficult to classify using traditional morphological criteria.¹¹ The currently accepted phylogeny in Figure 1a suggested that the subfamily Gasterochismatinae was closely related to Scombrinae, potentially indicating a sister group relationship.^2,46 Block et al.¹⁴ proposed that Gasterochisma was not closely related to Scombrinae, but the strength of the branch nodes in the phylogenetic tree constructed using the Cytb gene was weak. Finnerty⁴⁷ speculated that the limited resolution of the single-gene informative locus and the high degree of saturation and homogeneity in the Cytb gene might obscure relationships among nodes in the tree. To address the issue of Cytb gene saturation, researchers have suggested two approaches: sequencing slower-evolving genes, such as the 16s ribosomal gene, or using multiple molecular markers for analysis.^48,49 Reeb⁵⁰ utilized 16S rDNA (475 bp) and maximum likelihood distance and neighbor-joining algorithms to construct phylogenetic trees, concluding that Gasterochisma belonged to Scombrinae but were not sister groups. Our research provides further evidence supporting this conclusion, indicating that Gasterochismatinae and Scombrinae are not sister taxa.

Grammatorcynus is the earliest divergent lineage of the Scombridae

This study’s phylogenetic analysis of ML and BI methods indicated that Grammatorcynus was the earliest diverging clade in Scombridae. The NJ tree showed that the phylogenetic position of Grammatorcynus was located between Scombrini and Scomberomorini. There has been controversy about which group of Scombridae was the earliest to differentiate. According to traditional morphology, Scombrini was considered the earliest differentiated group in the Scombridae, while Grammatorcynus occupied an intermediate evolutionary position between Scombrini and Scomberomorini.^12,51 However, Johnson¹² observed that the skeletal morphology of Grammatorcynus differed from other Scombridae species in lacking a fourth pharyngeal dental plate. Consequently, Johnson¹² proposed that Grammatorcynus was the earliest diverging clade within the Scombridae. With the rapid development of molecular biology, the study of phylogenetic relationships on Scombridae is improving. Orrell et al.¹⁶ utilized the maximum likelihood method with the Tmo-4C4 gene sequencing to construct a phylogenetic tree and their findings were consistent with the traditional taxonomic classification which suggested Scombrini as the earliest-differentiated taxon within Scombridae. However, using the same gene, the maximum parsimony method showed Grammatorcynus as the earliest divergent branch.¹⁶ Santini et al.⁵² utilized phylogenetic analysis of seven genetic loci, including both nuclear and mitochondrial genes, to identify Scombrini as the earliest divergent clade within the Scombridae. However, Miya et al.⁵³ reconstructed phylogenetic trees for the mackerel family based on mitochondrial CO1 and Cytb genes, all indicating that Grammatorcynus was the most basal clade. The different branching hypotheses might be caused by the different algorithms and datasets using construct phylogenetic trees. The topological structure of the three algorithms in this study was not completely consistent. Unlike the analysis of ML and BI, the NJ tree believed that Scombrini was the earliest differentiated group within Scombididae. Many factors affected the topologies of phylogenetic trees, including the choice of ingroup representation, the evolution of genes, long-branch attraction (LBA), and the method of tree construction.⁵⁴ Due to different evolutionary patterns between genes, some mitochondrial genes might not have enough information loci to construct interspecific phylogenetic relationships.^55,56 Accordingly, different genetic markers are suitable for phylogenetic analysis at different taxonomic levels. Thus, our results suggested that Scombrini was a group separately differentiated from the rest of Scombridae and Grammatorcynus was the earliest divergent lineage in the Scombridae.

Sardini and Thunnini are non-monophyletic

Our analyses showed that Allothunnus (tribe Thunnini) and Cybiosarda (tribe Sardini) clustered into a clade (BI BPP = 0.688, ML BS = 56%, and NJ BS = 27%). The low nodal support values are shown in Figures 3-5. Overall, the low support rate of nodes in phylogenetic trees was the result of inherent complexities in the evolutionary process, limitations of these genes which their information loci were insufficient to support evolutionary relationships above the genus level and the stochastic nature of molecular evolution.⁵⁷

The ML tree revealed that Acanthocybium (tribe Scomberomorini) and Gymnosarda (tribe Sardini) clustered into a single clade. In addition, the BI tree analysis indicated Auxis+Euthynnus+Katsuwonus did not cluster into a clade with Thunnus. All results consistently supported Sardini and Thunnini as non-monophyletic groups and showed they were clustered. These findings contradicted the conventional analysis of morphological data.^23,58 Only a few studies have performed phylogenetic relationships, specifically on these two taxa, Sardini and Thunnini. The available molecular studies can only support the monophyly of genera Sarda, Auxis, Euthynnus, Katsuwonus, and Thunnus and cannot support the monophyly of Sardini and Thunnini.^23,58 Santini et al.⁵² revealed the clade of Acanthocybium+Gymnosarda and believed that the clade of Katsuwonus+Auxis+Euthynnus was not clustered with Thunnus. Monsch⁵⁹ suggested that Gymnosarda was a link between Sardinae and Scomberomorinae. Monsch²¹ reported that the genus Auxides might be the immediate sister group of Scomber and Rastrelliger. All of these contradicted the monophyletic character of Sardini and Thunnini. Thus, we confirmed the non-monophyly of the Sardini and Thunnini as recognized by Santini et al.⁵² and Monsch.^9,21 Nevertheless, additional morphological data and more comprehensive molecular data are needed to study the two groups in the future deeply.

CONCLUSION

This study investigated the molecular evolutionary relationships of 48 species from 14 genera found within the Scombridae. In summary, we propose that (1) Gasterochismatinae and Scombrinae are not sister groups; (2) Grammatorcynus likely represent the earliest divergent lineage in the Scombridae; and (3) Sardini and Thunnini are non-monophyletic. However, the phylogenetic relationship within the Scomberomorus genus and the classification of tribes require further clarification. Hence, additional morphological and molecular data (e.g., ITS), including an expansion of the molecular dataset through incorporating more genes and taxa, such as Orcynopsis, is necessary to enhance our understanding and refinement of the systematic phylogenetic relationships of this family.

CONFLICT OF INTEREST

The authors declare that they have no conflict of interest.

AUTHORS’ CONTRIBUTION PER CREDIT

Methodology: Xinru Zeng (Equal), Haoyu Yu (Equal). Formal Analysis: Xinru Zeng (Equal), Mengyao Cui (Equal). Investigation: Xinru Zeng (Equal), Haoyu Yu (Equal), Pingzhong Zheng (Equal). Data curation: Xinru Zeng (Equal), Mengyao Cui (Equal), Haoyu Yu (Equal). Writing – original draft: Xinru Zeng (Equal). Writing – review & editing: Xinru Zeng (Equal), Pingzhong Zheng (Equal), Fen Wei (Equal). Conceptualization: Fen Wei (Lead). Supervision: Fen Wei (Lead). Project administration: Fen Wei (Lead). Funding acquisition: Fen Wei (Lead).

ACKNOWLEDGMENTS

We thank the Coral Reef Research Center of China for the infrastructure provided. This work was supported by funding from the National Natural Science Foundation of China (Nos. 42090041 and 42030502) and the Science and Technology Project of Guangxi (Nos.AA17204074, AD17129063, 2018GXNSFBA050023). We thank the reviewers for their meaningful comments.

DATA AVAILABILITY STATEMENT

The data presented in this study are openly available in GenBank with Reference accession numbers OQ781867-OQ781879, OR343904-OR343906, and OR357872-OR357886.