Introduction

Sciaenidae (Acanthuriformes) is a large family of fishes, comprising 68 genera and 298 species1, widespread in the Atlantic, Pacific, and Indian oceans, inhabiting either tropical, subtropical or temperate regions2,3. Nonetheless, the diversification processes and the interrelationships among several taxa in Sciaenidae remain largely unknown. Even though previous phylogenetic inferences invariably support Sciaenidae as monophyletic4,7, their intergeneric and interspecific relationships are poorly resolved, such as observed in taxa of the subfamily Stelliferinae8,9.

Stelliferinae was recognized as a supragenus10, being later elevated to the subfamily level by Sasaki11. A total of 51 species distributed in the coast and estuaries along the western Atlantic and eastern Pacific belong to Stelliferinae, thus representing the third richest subfamily in Sciaenidae. Formerly, this subfamily comprised six genera, as follows: Stellifer Oken, 1817, Ophioscion Gill, 1863, Bairdiella Gill, 1861, Odontoscion Gill, 1862, Corvula Jordan and Eigenmann, 1889, and Elattarchus Jordan and Evermann, 1896. However, molecular phylogenetic inferences revealed the non-monophyly and the close relationship between Stellifer and Ophioscion, suggesting that both genera should be synonymized8,9. Recently, Chao et al.12 followed this suggestion and recognized Ophioscion as a junior synonym of Stellifer, resulting in five valid genera of Stelliferinae.

All morphological and molecular phylogenetic studies confirmed the monophyly of Stelliferinae6,8,9,10,11. Nonetheless, the genetic and morphological similarities among taxa hindered the definition of reliable intergeneric and interspecific relationships within the subfamily, probably as a result to their rapid adaptive radiation8,9. Furthermore, to date, morphological and molecular phylogenetic analyses have evaluated only a limited number of Stelliferinae taxa, with just one study including all genera within the subfamily11, thereby restricting the scope for comprehensive comparative analyses. A summary of previous phylogenetic studies is provided in Fig. 1.

Fig. 1
figure 1

Previous morphological (a and b) and molecular (c–f) phylogenetic hypotheses for Stelliferinae.

Full size image

Phylogenies based on morphological traits support the monophyly of the analyzed genera and confirm a close relationship between Stellifer and Ophioscion10,11. However, these studies differ regarding the placement of Bairdiella, which is identified as the sister group of Odontoscion in Chao10, while is closely related to the Stellifer/Ophioscion clade in Sasaki11. Furthermore, Sasaki11 proposed that Odontoscion is closely related to Elattarchus and that these genera form a clade with Corvula.

In contrast, molecular phylogenies including Stelliferinae suggest the non-monophyly of both Stellifer and Ophioscion4,6,8,9. The placement of Bairdiella also remains contentious in molecular analyses. Santos et al.4 placed Bairdiella as closely related to the Stellifer/Ophioscion clade, while Barbosa et al.8 identified it as the sister group of Odontoscion. In contrast, Lo et al.6 and Silva et al.9 proposed that Bairdiella forms a clade with Corvula/Odontoscion. Additionally, Lo et al.6 reported a closer relatedness between one specimen of Corvula macrops Steindachner, 1875 and Odontoscion xanthops Gilbert, 1898 than to co-specific samples, suggesting the need for a re-evaluation of the taxonomic status of these taxa.

Another relevant aspect about the taxonomy of Stelliferinae is the hypothesis of hidden diversity in Odontoscion. A study based on DNA barcode revealed at least two lineages of Odontoscion dentex Cuvier, 1830 along the Atlantic, with a genetic divergence as high as 12.9%13. Likewise, species delimitation methods, using DNA barcoding, identified three cryptic lineages within O. dentex (two in the Caribbean coast and another one in the Brazilian coast), whose genetic differences ranged from 3.95 to 9.33%14. These results suggest that the taxonomic status of these lineages should be further investigated, given that only O. dentex is recognized for the western Atlantic1.

The Stelliferinae subfamily has an amphi-American distribution, with species inhabiting estuarine and coastal regions of the western Atlantic and eastern Pacific, except for O. dentex, which is associated with coral reefs. Several Stelliferinae taxa are widely distributed across regions characterized by putative barriers to gene flow, which have been identified as key drivers of genetic differentiation and diversification in marine fishes. These barriers include marine circulation patterns, variations in salinity and sea surface temperature, freshwater and sediments plumes discharged from large rivers, and the formation of the Isthmus of Panama15,16,17,18. In the western Atlantic, the Amazon plume, formed between 9.4 and 2.4 Ma, discharges approximately 6,300 km3 of freshwater and 900 × 106 tons of sediment annually into the ocean, extending about 2300 km along the northeastern coast of South America19,20. This plume acts as a significant barrier, preventing gene flow between populations and driving speciation among fishes in the Caribbean and Brazilian provinces15,21,22. Moreover, for amphi-American groups, the formation of the Isthmus of Panama (17 − 3 Ma) is considered a major event that triggered the diversification of marine biota in the region. It altered marine circulation, salinity, and temperature patterns in the western Atlantic and eastern Pacific, leading to isolation and diversification of marine species. Given that Stelliferinae taxa inhabit areas influenced by several historical and extant physical and ecological processes acting as barriers, it is necessary to evaluate the biogeographical structure of this group and investigate the potential barriers to gene flow and associated processes driving its diversification.

Therefore, considering: (i) the non-monophyly of Stellifer, (ii) the discrepancies in the phylogenies regarding the positioning of Bairdiella and Odontoscion, (iii) the lack of molecular phylogenies including Elattarchus, (iv) the need to evaluate the monophyly of Corvula and Odontoscion and (v) the debate about the taxonomic status of lineages in O. dentex, we carried out an extensive multilocus analysis in Stelliferinae based on mitochondrial and nuclear regions, totaling 13 genetic markers. The present data were used to infer the evolutionary relationships among the valid genera in this subfamily, to test the monophyly of Stellifer, Corvula and Odontoscion, to evaluate the positioning of Bairdiella, Elattarchus and Odontoscion, and to test the hypothesis of cryptic speciation in O. dentex. In addition, we also estimated the time of divergence, the biogeographic history, and phylogeographic patterns in Stelliferinae to infer the processes that shaped their diversification patterns.

Results

Phylogenetic relationships and the time of the most recent common ancestor (TMRCA)

The phylogenetic reconstruction was based on a dataset of 9,041 base pairs (bp) from the 13 genetic markers, referring to the three mitochondrial DNA and 10 nuclear DNA loci (Supplementary Table S1). A total of 66 specimens were used, representing 24 species distributed among the five valid genera of Stelliferinae (47% taxon coverage). In all analyses, Larimus acclivis and Nebris microps were used as outgroups.

In general, the topologies of concatenation-based (Maximum Likelihood - ML, Bayesian inference - BI and species tree in the StarBEAST) and the summary-coalescent species tree (ASTRAL) phylogenies were congruent, except for some discrepancies in some clades, and all our results supported the monophyletic status of Stelliferinae as well as the presence of four groups, each with high bootstrap, and posterior probabilities supports (Figs. 2, 3, 4 and 5). Group 1 indicates a close relatedness among Bairdiella, Elattarchus, Corvula, and Odontoscion while the groups 2 to 4 encompass species of Stellifer, suggesting that this is a non-monophyletic genus (Figs. 2, 3, 4 and 5). In ML and BI trees, groups 2 and 3 were closely related and formed a clade with group 1 (Figs. 2 and 3). On the other hand, in the concatenation-based species tree, group 2 was closely related to group 1 (Fig. 4), whereas the summary-coalescent species tree displayed a polytomic arrangement among groups 1, 2, and 3 (Fig. 5).

Fig. 2
figure 2

Phylogenetic tree of Stelliferinae inferred from Maximum Likelihood (ML) based on the concatenated database of 13 genomic regions. The species Larimus acclivis and Nebris microps were used as outgroups. In the nodes are the bootstrap support values. The number after the vertical bars along each clade indicate the posterior probability values based on the Bayesian species delimitation.

Full size image
Fig. 3
figure 3

Phylogenetic tree of Stelliferinae inferred using Bayesian Inference (BI) based on a concatenated database of 13 genomic regions. The species Larimus acclivis and Nebris microps were used as outgroups. The numbers on the nodes represent the posterior probability values.

Full size image
Fig. 4
figure 4

Bayesian species tree with time-calibrated maximum clade credibility generated using StarBeast3 analysis, based on a concatenated dataset of 13 genomic regions from Stelliferinae. The species Larimus acclivis and Nebris microps were used as outgroups. Node squares represent ancestral range estimations, inferred using the BayArea-like + j identified as the best-fitting model by the biogeographic estimation. The time of the most recent common ancestor (TMRCA) is shown above the nodes, with bars representing the 95% highest posterior density intervals. Values below the nodes correspond to Bayesian posterior probabilities. The map shows the six marine biogeographic regions used to code the geographic distribution of extant species. H in the time bar refer to the Holocene.

Full size image
Fig. 5
figure 5

Phylogenetic tree of Stelliferinae inferred using a summary-coalescent approach in ASTRAL-III based on a dataset of 13 genomic regions. The species Larimus acclivis and Nebris microps were used as outgroups. Node values represent bootstrap support greater than 30%.

Full size image

There was topological discordance among gene trees and phylogenies using the concatenation-based and the summary-coalescent approaches (Figs. 2, 3, 4 and 5; Supplementary Figs. 1–13). The mitochondrial markers support the four groups observed in the phylogenies based on concatenation or summary-coalescent approaches (Supplementary Figs. 1–3). In contrast, the nuclear markers yielded poorly resolved trees, with all genera in a polytomic arrangement, and some gene trees grouping Bairdiella, Elattarchus, Odontoscion and Corvula in a clade with high bootstrap support (Supplementary Figs. 4–13).

In group 1, Bairdiella was recovered as the sister group of Elattarchus, while Corvula was closely related to Odontoscion. However, Odontoscion was found to be non-monophyletic, as O. xanthops was more closely related to C. macrops than to its congener O. dentex, except in the summary-coalescent species tree, where the Odontoscion species formed a clade but with low posterior probability (Figs. 2, 3, 4 and 5). In both concatenation-based and summary-coalescent species trees, Bairdiella armata Gill, 1863 was closely related to Bairdiella goeldi Marceniuk et al., 2019, although this relationship was well supported only in the concatenation-based species tree (Figs. 4 and 5). Conversely, B. armata was closely related to Elattarchus archidium Jordan and Gilbert, 1882 in the ML analysis, but with low bootstrap support (Fig. 2), and formed a polytomy with B. goeldi and E. archidium in the BI tree (Fig. 3). In addition, all phylogenetic inferences revealed two well-supported reciprocally monophyletic lineages within O. dentex, one occurring along the Mexican coast (Caribbean Province) and the other in the Brazilian Province (Figs. 2, 3, 4 and 5).

The non-monophyly of Stellifer was evident in all methods of phylogenetic reconstruction, as species from this genus grouped into three clades (Figs. 2, 3, 4 and 5). Within group 2, we identified two clades. In the first one Stellifer punctatissimus Meek and Hildebrand, 1925/Stellifer scierus Jordan and Gilbert, 1884 are sister taxa, and a close relationship was observed between Stellifer gomezi Cervigón, 2011/Stellifer menezesi Chao et al., 2021, both recovered as non-monophyletic taxa. Furthermore, this group also includes Stellifer strabo Gilbert, 1897, Stellifer simulus Gilbert, 1898, and Stellifer typicus Gill, 1863; however, the relationship among these taxa remains unresolved (Figs. 2, 3, 4 and 5). In the second clade of group 2, Stellifer illecebrosus Gilbert, 1898 and Stellifer mancorensis Chirichigno F., 1962 are sister taxa, forming a clade closely related to Stellifer oscitans Jordan and Gilbert, 1882/Stellifer rastrifer Jordan, 1889, and Stellifer pizarroensis Hildebrand, 1946, while Stellifer stellifer Bloch, 1790 is resolved as the earliest-diverging taxon in this group (Figs. 2, 3, 4 and 5).

In group 3, Stellifer microps Steindachner, 1864 was the sister taxon of Stellifer naso Jordan, 1889, forming a clade closely related to Stellifer brasiliensis Schultz, 1945 (Figs. 2, 3, 4 and 5). Finally, group 4 represents the most ancient clade within Stelliferinae in which Stellifer ericymba Jordan and Gilbert, 1882 and Stellifer vermicularis Günter, 1867 were closely related in both IB and concatenation-based species trees, with high support (Figs. 3 and 4). In contrast, in the ML and summary-coalescent species trees the group 4 contains only S. vermicularis, while S. ericymba is closely related to group 2 (Figs. 2 and 5).

The estimated TMRCA values indicate that the origin and diversification of Stelliferinae took place around 16.7 Ma (HPD: 13.2–20.3) during Early Miocene (Fig. 4), when most taxa diverged. Some taxa diverged later during the Pliocene while the diversification of S. gomezi/S. menezesi and S. punctatissimus/S. scierus occurred in the Pleistocene (Fig. 4).

Coalescent species delimitation

The tests carried out using distinct population size (θ) and root age (τ0) parameters produced the same results, confirming that most nominal taxa in Stelliferinae correspond to valid species with a posterior probability of 1 (Fig. 2). In addition, the two phylogenetic lineages identified in O. dentex were recovered as distinct species by BPP. Similarly, C. macrops and O. xanthops were also identified as unique species by this method. On the other hand, S. gomezi and S. menezesi were attributed to a single species.

Biogeographic history of Stelliferinae

The BioGeoBEARS analysis selected the BayArea-like + j as the best-fitting model for reconstructing the ancestral area of Stelliferinae, based on the corrected Akaike Information Criterion (Supplementary Table S2). According to this model, our analysis suggests that the Panamic and Peruvian provinces in the eastern Pacific were likely the center of origin for Stelliferinae during the Early Miocene. All groups appear to share an ancestor that inhabited the eastern Pacific before colonizing the western Atlantic. Our ancestral area reconstruction indicates at least five independent dispersal events from the eastern Pacific to the western Atlantic via the Central American Seaway (CAS) between the Miocene and Pliocene, prior to the closure of the Isthmus of Panama (Fig. 4). The ancestor of group 3 was the first to cross the CAS during Early Miocene, subsequently diversifying in the western Atlantic during Late Miocene. For group 1, the ancestral range was inferred to be in the Panamic and Peruvian provinces, followed by dispersal to the western Atlantic. This included ancestors that gave rise to the B. goeldi and Odontoscion dentex lineages, which occurred in distinct periods during the Late Miocene. The crown node of group 2 was estimated in the Panamic and Peruvian provinces, with two dispersal events to the western Atlantic. These events involved the ancestors of S. menezesi/S. gomezi and S. punctatissimus during the Late Miocene, and S. rastrifer during the Early Pliocene.

Discussion

Phylogenetic relationships and species delimitation in Stelliferinae

This study represents the most comprehensive molecular evaluation of the evolutionary relationships within Stelliferinae, considering both the taxa and molecular markers sampled. Our results support the monophyly of Stelliferinae, consistent with previous morphological and molecular phylogenetic studies4,6,8,9,10,11.

The discordance between the topologies of gene trees and phylogenies inferred using concatenation and summary-coalescent approaches can be attributed to the limited phylogenetic signal present in individual markers. Moreover, topological incongruence may arise from incomplete lineage sorting, a phenomenon commonly observed in taxa with large effective population sizes or recent rapid divergence events23,24, as exemplified by Stelliferinae8,9. These processes frequently result in short, poorly resolved branches in phylogenetic trees23,24. While it remains challenging to determine which evolutionary processes have contributed most significantly to the observed phylogenetic patterns in Stelliferinae or to ascertain which method, coalescent or concatenation, is more prone to bias, evidence suggests that concatenation approaches have consistently provided robust insights into the phylogeny of taxa undergoing recent rapid radiations. These methods have been widely applied in systematic studies23,24. Accordingly, given that analyses of concatenated datasets yielded more congruent topologies, our discussion of taxonomic relationships within Stelliferinae is based on the results derived from these phylogenies.

In relation to group 1, all inferences corroborated a close relationship between Bairdiella and Elattarchus, while Corvula was placed as the sister group of Odontoscion. These results diverge from Chao10 and Barbosa et al.8 who suggested a close relationship between Bairdiella and Odontoscion. Nonetheless, representatives of Corvula and Elattarchus were not evaluated by these authors. Our results also differ from the proposal of Sasaki11, who carried out the most comprehensive morphological phylogenetic analysis in Sciaenidae so far placing Bairdiella as the sister group of the clade Stellifer/Ophioscion (currently referred to as Stellifer). On the other hand, our results agree with Lo et al.6 and Silva et al.9 in relation to the close relationship between Corvula and Odontoscion and between this clade and Bairdiella (the genus Elattarchus was not included in their analyses). Therefore, considering that the present study is the only multilocus phylogenetic analysis in Stelliferinae including Elattarchus, we suggest that the relationships recovered in group 1 may be the most likely evolutionary scenario for this clade.

The monophyly of Bairdiella was well supported only in the concatenation-based species tree. In fact, this genus was regarded as monophyletic by Lo et al.6, even though these authors have not analyzed representatives of the genus Elattarchus. Since Bairdiella encompasses seven valid species and only two of them were included in the present study, as well as in Lo et al.6, further efforts in sampling and utilization of multilocus data are needed to properly assess the monophyletic status of the genus.

Our results indicate that Odontoscion is a non-monophyletic genus since O. xanthops was more closely related to C. macrops than to the congeneric O. dentex. Lo et al.6 also found O. xanthops and C. macrops as closely related species based on their genetic similarity. Furthermore, a recent species delimitation study suggested that these taxa may represent a single species14. In fact, these genera are morphologically similar. According to Sasaki11, the genus Odontoscion is characterized by two autapomorphies, the basisphenoid separated ventrally from parasphenoid and the levator operculi originating from the posttemporal, while Corvula lacks autapomorphic traits. Nevertheless, the main diagnostic trait used to distinguish these groups is the presence of canines in Odontoscion, which has determined cases of misidentification within Odontoscion and Corvula [see 3, 25]. Although our phylogenetic and species delimitation (BPP) analyses support O. xanthops and C. macrops as distinct species, the morphological characters used to delimit the genera appear to be inconsistent and unreliable.

A detailed morphological comparison reinforces this view, revealing extensive overlap between species of Odontoscion and Corvula in nearly all traits traditionally used in generic diagnoses within Sciaenidae (Supplementary Table S3). Both genera share key features such as the absence of mental barbels, a two-chambered swim bladder without appendages, and a finely serrated preopercular margin lacking strong bony spines26,27. They also exhibit similar body forms, with elongate profiles and terminal to subterminal mouths, as well as longitudinal stripes below the lateral line25,26,27. Dorsal fins are continuous in both groups, with overlapping counts of spines (10–13) and soft rays (21–29), and anal fins consistently bear two spines and 7 to 11 soft rays26,27. While differences such as the presence of canines in Odontoscion exist, they likely reflect intraspecific variability or ecological adaptations rather than deep phylogenetic divergence. Given the strong morphological congruence and the close evolutionary relationships inferred from molecular data, we propose that Corvula should be considered a junior synonym of Odontoscion.

Moreover, two reciprocally monophyletic lineages were identified in O. dentex following a disjunct distribution range since each lineage was restricted to distinct provinces (Brazilian and the Caribbean). Likewise, a report based on DNA barcoding13 recovered two genetically divergent groups (12.9%) between samples from the Caribbean and Brazil in this taxon. Later, species delimitation analysis using DNA barcoding identified three putative species in O. dentex; one of them endemic to the Brazilian province while the remaining two were distributed along the Caribbean province14. So far, three valid species are recognized in Odontoscion, including two taxa from the eastern Pacific (O. xanthops and Odontoscion eurymesops Heller and Snodgrass, 1903) and a single taxon described for the western Atlantic (O. dentex). Thus, the present results and DNA barcoding strongly indicates a cryptic speciation process within O. dentex. Accordingly, a detailed taxonomic review based on morphological data and increased sampling efforts is necessary to validate the new species of Odontoscion along the western Atlantic.

The non-monophyly in the genus Stellifer was evident by all analyses based on molecular data4,6,8,9, thereby diverging from morphological phylogenetic inferences10,11. Based on a multilocus approach, Silva et al.9 proposed that Stellifer and Ophioscion were closely related groups that should be synonymized, while Chao et al.12 placed Ophioscion as a junior synonym of Stellifer. However, our phylogenetic analysis revealed three groups within Stellifer, where the groups 2 and 3 would be more related to the species in group 1 (along with Bairdiella, Elattarchus, Corvula, and Odontoscion) than to group 4. In addition, the group 4 represented the most divergent clade within Stelliferinae, as also reported by Lo et al.6 and Silva et al.9. Based on the congruent results between this study and previous reports4,6,8,9, we recommend that these groups should be reclassified as distinct genera of Stelliferinae. For that, all recognized species in Stellifer should be sampled to carry out a wide taxonomic revision in this subfamily, focused on the classification of new or previous valid genera.

In addition, S. gomezi and S. menezesi are non-monophyletic species, as they formed a highly supported polytomous branch in the present study. Stellifer menezesi is a recently described species12, differentiated from S. gomezi by variations in body height, size of nostrils, length of snout, pectoral fins, and second spine of the anal fin; even though, most of morphological traits are overlapped in both taxa12,28. Furthermore, the Bayesian Phylogenetics and Phylogeography (BPP) test also recovered S. gomezi and S. menezesi as a single species, consistent with previous findings from other species delimitation methods14. Therefore, we provide additional evidence that S. menezesi should be considered a junior synonym of S. gomezi.

In group 2, S. menezesi/S. gomezi, S. scierus/S. punctatissimus, S. strabo, and S. simulus were recovered as closely related species, forming a well-supported clade, corroborating the phylogenetic inferences by Silva et al.9. On the other hand, S. typicus was closely related to this clade, while Silva et al.9 proposed that S. rastrifer/S. stellifer is a sister clade of the group comprising S. menezesi/S. gomezi, S. scierus/S. punctatissimus, S. strabo, and S. simulus. Nevertheless, the support values of these evolutionary relationships were low in both reports, thus hindering the determination of the most likely phylogenetic arrangement among such species.

Furthermore, the group 2 revealed a close and highly supported relationship among S. oscitans, S. rastrifer, and S. pizarroensis. These taxa formed a clade along with S. illecebrosus and S. mancorensis, while S. stellifer was the most divergent species in this group. These results differ from previous phylogenetic studies based on a few species of the genus, which have shown incongruent results. For instance, Santos et al.4 recovered S. rastrifer as the sister group of Stellifer sp. B (currently referred to as Stellifer collettei), while S. stellifer was the sister group of S. punctatissimus (formerly Ophioscion punctatissimus). On the other hand, Barbosa et al.8 suggested a close relationship between S. stellifer and Stellifer sp. B, both forming a clade with S. rastrifer. Instead, Lo et al.6 suggested that S. rastrifer is closely related to S. oscitans, while Silva et al.9 allocated S. rastrifer as the sister group of S. stellifer. Considering the reduced number of species in previous studies and the strong statistical support obtained in our analyses, we suggest that the present phylogenetic arrangement should be the most likely scenario for these taxa.

In the case of group 3, we observed that S. microps and S. naso formed a closely related clade, grouped along with S. brasiliensis, as previously reported in other molecular phylogenies including these species4,8,9. Therefore, our results reinforce that such clade of Stellifer should be referred to as a distinct genus.

Time of diversification, biogeographic history and phylogeographic patterns in stelliferinae

Our results suggested that Stelliferinae fishes originated in Early Miocene, about 16.7 Ma (HPD: 13.2–20.3 Ma). In addition, most taxa from this subfamily have also diverged during this period, with a few records of speciation during the Pliocene and Pleistocene. Similarly, Lo et al.6 and Silva et al.9 reported that the diversification of Stelliferinae has taken place during the Miocene, at 15.7 and 15.6 Ma, respectively.

The subfamily Stelliferinae exhibits an amphi-American distribution. From a phylogeographic perspective, we identified closely related species inhabiting distinct oceans within groups 1 and 2. Range estimations suggest that the center of origin for Stelliferinae was in the eastern Pacific. Our findings are consistent with those of Lo et al.6, who also proposed an eastern Pacific origin for Stelliferinae. Based on our ancestral reconstruction, we propose that the ancestors of Stelliferinae species diversified in the Pacific and that multiple independent dispersal events to the western Atlantic occurred between the Miocene and Pliocene, via the CAS, prior to the closure of the Isthmus of Panama. Several studies have suggested that the CAS served as the primary dispersal route for marine biota between the Pacific and Atlantic prior to the closure of the Isthmus16,18,29.

The formation of the Isthmus of Panama was a gradual process that began in the Miocene (~ 17 Ma) and culminated in the complete isolation of the Pacific and Atlantic Oceans approximately 3 Ma during the Pliocene17,18,30. Throughout the Miocene and Pliocene, tectonic movements drove the uplift of the Isthmus and the progressive closure of the CAS, the Pacific-Atlantic connection route in Central America. The reduced connectivity between these oceans led to significant changes in oceanographic conditions, including alterations in sea surface circulation patterns, salinity, and temperature17,18,30,31,32, which would have favored the occurrence of diversification events within Stelliferinae. However, a partial connection between the eastern Pacific and western Atlantic persisted until the total closure of the Isthmus of Panama (~ 3 Ma), being putatively sufficient to enable species dispersal between the two regions. This scenario could explain the formation of closely related clades among species now distributed in geographically disjunct areas.

It is noteworthy that the divergence between S. punctatissimus and S. scierus dated back to nearly 2.1 Ma during the Pleistocene, with the 95% highest posterior density (HPD: 0.7–3.3 Ma) overlapping the diversification times reported in previous studies6,9. Accordingly, we suggest that the closure of the Isthmus of Panama might have interrupted the gene flow between ancestral populations, leading them to independent evolutionary pathways that promoted the speciation, as proposed by Silva et al.9. In fact, several authors suggest that the closure of the Isthmus of Panama was the most important vicariant event to the speciation processes among marine sister taxa of amphi-American distribution16,33,34,35,36.

The divergence between the two lineages of O. dentex occurred approximately 3.3 Ma (HPD: 1.4–4.9), during the Pliocene, following a disjunct distribution, with one lineage occurring in the Brazilian province and another in the Caribbean province. It is noteworthy that O. dentex is a species associated with coral reefs, being found at depths of up to 30 m37, inhabiting areas of high salinity. It is likely that this taxon has low tolerance to reduced salinity levels, a characteristic commonly observed in coastal and estuarine environments where most species of Sciaenidae occur. Moreover, the two lineages of O. dentex are separated by the Amazon plume, a massive outflow of freshwater and sediment from the Amazon River into the Atlantic that represents a major biogeographic barrier for Atlantic reef fishes15,21,22. Interestingly, the estimated divergence time between the two lineages coincides with the period of increased Amazon River outflow (6.8 to 2.4 Ma) during the Late Miocene and Early Pliocene19. This enhanced outflow likely acted as a barrier, isolating populations of O. dentex in the Caribbean and Brazilian provinces, thereby preventing gene flow and promoting speciation.

The period from 6.8 to 2.4 Ma is recognized as the second stage of Amazon River formation, marked by an increase in freshwater and sediment discharge into the western Atlantic, with sedimentation rates reaching ~ 0.3 m/ka19. From 2.4 Ma onwards, sedimentation rates into the Atlantic increased significantly to ~ 1.2 m/ka, coinciding with the Amazon River attaining its current conformation19. Today, the Amazon River discharges approximately 6300 km3 of freshwater and 900 × 106 tons of sediment annually into the western Atlantic20. The area of influence of the Amazon River plume extends roughly 200 km from its mouth, reducing salinity levels up to an isobath of 30 m deep in the Atlantic Ocean38,39,40. Consequently, the Amazon River plume has prevented the formation of reef habitats at these depths, acting as a potential barrier to the dispersal of O. dentex lineages. Supporting this hypothesis, several studies on coral reef fishes have demonstrated the role of the Amazon plume in driving diversification among sister species distributed between the Caribbean and Brazilian provinces15,21,41,42,43. Therefore, our results suggest the existence of at least two species within O. dentex, underscoring the need for additional sampling across its distribution range to refine biogeographic inferences and conduct a comprehensive taxonomic review of this nominal taxon.

Conclusions

Our study represents a substantial advancement in the systematics of Stelliferinae by combining broader taxonomic sampling, expanded genetic data, and integrated phylogenetic and biogeographic analyses. In summary, our results confirm the monophyly of Stelliferinae, while the genera Odontoscion and Stellifer proved to be non-monophyletic, and cryptic diversity was detected within O. dentex. Conversely, S. gomezi and S. menezesi are likely to represent a single species, and the genera Corvula and Odontoscion should be synonymized. Additionally, our ancestral area reconstruction suggests an eastern Pacific origin to Stelliferinae, where ancestral lineages diversified and undertook independent dispersal events to colonize the western Atlantic. Moreover, the formation of the Isthmus of Panama appears to have played a role in the diversification of Stelliferinae. The Amazon plume outflow was also identified as a likely barrier driving diversification in O. dentex. Therefore, this study provides evidence supporting the need for a comprehensive taxonomic review of Stelliferinae.

Materials and methods

Sample collection

We sampled 24 of the 51 valid species of Stelliferinae, representing 47% taxon coverage and including all currently genera recognized in this subfamily: Stellifer (18/31 species), Bairdiella (2/7 species), Odontoscion (2/3 species), Corvula (1/3 species), and Elattarchus (1/1 species). Samples were collected from sites across the western Atlantic and eastern Pacific (Table 1). Species selection was guided by morphological diversity, taxonomic breadth, and the availability of high-quality tissue samples. Although not comprehensive, this sampling provides a robust framework for phylogenetic analyses and supports the taxonomic reassessment of the subfamily.

Table 1 Samples and genetic markers used in the phylogenetic analyses of stelliferinae. The accession number of sequences deposited in GenBank are shown below each marker. The abbreviations WA and EP refer to Western Atlantic and Eastern pacific, respectively; n = number of individuals used in the analyses; 16S rDNA = 16S ribosomal RNA gene; coi = cytochrome oxidase C subunit I; cytb = cytochrome b; rhod = rhodopsin; EGR1 = early growth response 1 gene; RAG1 = recombination activating protein 1 gene; ENCI = ectodermal-neural cortex 1 gene; SREB2 = super conserved receptor expressed in brain 2; glyt = glycosyltransferase; PLAG2 = pleiomorphic adenoma gene-like 2; MYH6 = myosin, heavy polypeptide 6; zicI = zic family member I; TBRI = T-box brain I. asequences obtained from GenBank while (-) indicates missing data.
Full size table

In total, 66 specimens were analyzed. These included 24 samples obtained from the tissue bank of Sciaenidae available in the Laboratory of Fish Microbiology at the Institute of Coastal Studies, Federal University of Para (UFPA), Brazil; 29 samples from the fish collection at the Museum of Zoology, University of Costa Rica (UCR); and 13 samples from the fish collection at the Universidad Michoacana de San Nicolás de Hidalgo (UMSNH), Mexico. As outgroups, we included samples of Larimus acclivis Jordan and Bristol, 1898 and Nebris microps Cuvier, 1830 (Table 1). All specimens were identified based on identification keys10,44,45,46.

No threatened or protected species were used in the present study. The specimens sampled along the Brazilian coast were collected by artisanal fisheries under license provided by the Brazilian Environment Ministry (Permit number 18401-3) on behalf of Dr. Simoni Santos. The samples from Costa Rica were collected according to the licenses obtained from the National System of Conservation Areas (Permit code R-SINAC-SE-DT-PI-003-2021) and the National Council of Biodiversity Management (Permit code R056-2015-OT-CONAGEBIO) as well as the resolution No. 377 of the Vicerectoría de Investigación of the UCR. The specimens from Mexico were collected under permits PPF/DGOPA-035/15 and CONAPESCA-PPF/DGOPA-262/17, while the fishes from Ecuador at the UMSNH were collected according to the license 013/2012 PNG/N21-2017-EXP-CM-2016-DNB/MA.

Ethics declaration

For the sampling, approval by the ethics committee was not requested because the fish were purchased from artisanal fishermen and were already dead at the time of collection.

DNA extraction, polymerase chain reaction amplification, and sequencing of genomic regions

The whole genomic DNA was isolated from muscle tissue using the Wizard Genomic DNA Purification kit (PROMEGA). The concentration and the purity of DNA samples were evaluated in a Nanodrop 2000 spectrophotometer (Thermo Scientific).

The 13 DNA markers were amplified via Polymerase Chain Reaction (PCR), comprising three mitochondrial (16S rDNA, COI, and CYTB) and ten nuclear (RHOD, EGR1, RAG1, ENCI, SREB2, GLYT, PLAG2, MYH6, ZICI, and TBRI) loci. Out of this total, 12 markers refer to exons while one of them (16S rDNA) represents a non-protein-coding gene. The PCR was carried out 2.4 µl of dNTPs at 1.25 mM, 1.5 µl of 10x buffer solution, 0.5 µl of MgCl2 at 50 mM, 0.3 µl of each primer at 10 pmol/µl, 1–3 µl of template DNA (100 ng/µl), 0.12 µl of Taq DNA polymerase at 5U/µl (Invitrogen - Thermo Fisher Scientific) and ultrapure water to a final volume of 15 µl. A nested PCR was carried out to amplify the RAG1 marker, as described by Silva et al.9. Details about the primers and the amplification conditions for each marker are shown in Table 2.

Table 2 Primers and amplification conditions used in analyses of genomic regions in stelliferinae.
Full size table

The quality of amplicons was evaluated by electrophoresis in 1% agarose gel stained with GelRed and visualized under ultraviolet light. The successfully amplified PCR products were purified using the polyethylene glycol-8000 M protocol55. The sequences were generated by Sanger’s method56 using the Big Dye 3.1 terminator kit (Applied Biosystems) according to the manufacturer’s instructions. The sequencing was performed in ABI 3500XL Genetic Analyzer (Applied Biosystems).

Phylogenetic inferences and time of the most recent common ancestor (TMRCA)

The sequences were edited using the software Bioedit 5.0.657 and automatically aligned in Clustal W58, available in Bioedit. Manual adjustments in the alignments were carried out when necessary. Degenerated bases were used in the heterozygous sites of nuclear regions. The database was concatenated in SequenceMatrix59, while PartitionFinder 260 was used to select the best evolutionary models and partitions for Bayesian inference (BI).

Phylogenetic reconstruction based on Maximum Likelihood (ML) inference was performed with IQ-TREE 2.1.1361 for the 13 individual markers and the concatenated dataset. Coding regions were partitioned by codon position, while a single partition was applied to the 16S rDNA, using evolutionary models selected by ModelFinder62 (Supplementary Table S1). Branch support was evaluated through nonparametric bootstrapping, with 1000 replicates for individual markers and 2000 replicates for the concatenated dataset.

The BI tree was generated in the software MrBayes GPU63,64. In this case, the 12 protein-coding regions were partitioned according to their codon positions using the models selected by PartitionFinder, whereas the 16S rDNA sequences were considered a single partition (Supplementary Table S1). Four independent runs (30 million generations each) were carried out using the default parameters from each model as starting values. Clade posterior probabilities were calculated by using Metropolis coupled Markov Chain Monte Carlo algorithm (MCMCMC), assuming a burn-in of 10%. The run parameters over generations and the data convergence were evaluated in the software Tracer 1.7.165, where only ESS (Effective Sample Size) > 200 were accepted.

The species tree and the TMRCA in Stelliferinae were determined using StarBEAST366, available in the software BEAST 2.7.567. For that, we considered the hierarchical BI based on HKY + I + G substitution model. The tree was built assuming an uncorrelated relaxed molecular clock and Yule prior. The time of divergence was calibrated based on the fossil record of Stelliferinae and Larimus Cuvier, 1830, a Sciaenid genus closely related to this subfamily. The TMRCA of the clade encompassing Larimus and Nebris Cuvier, 1830 was estimated from the fossils Larimus henrici Nolf and Aguilera, 1998 and Larimus steurbauti Nolf and Aguilera, 1998 found in the Cantaure formation, Mexico, dating back to 16 and 23 Ma (Early Miocene)68. The fossil record of Bairdiella sp69. from the Tortonian (7.2 to 11.6 Ma) was selected to estimate the TMRCA of the clade comprising Bairdiella. The posterior probability parameters were evaluated using MCMC with 130 million generations in which the log parameters were sampled every 13,000 generations, assuming a burn-in of 10%. The run parameters and the convergence of the chains were verified in Tracer 1.7.165 and only ESS > 200 were considered. The consensus tree along with their support values and estimates of time of divergence was generated using the software TreeAnnotator 1.8.

To account for discordances between gene trees and the species tree, the gene trees estimated in IQ-TREE were analyzed using the multispecies coalescent species tree method in ASTRAL-III70. ASTRAL-III has been demonstrated to be an efficient summary method under various levels of incomplete lineage sorting70, although it is sensitive to gene tree estimation error. Therefore, to minimize such errors, we analyzed the data after collapsing poorly supported relationships in the gene trees (bootstrap < 30%) using TreeGraph 271.

Finally, all trees were visualized in FigTree 1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/) and saved as SVG files for edition in the software InkScape 1.1 (https://inkscape.org/pt-br/).

Coalescent species delimitation

The diversity and delimitation of the interspecific barriers within the Stelliferinae were assessed based on the multispecies coalescent model available in the software Bayesian Phylogenetics and Phylogeography (BPP)72,73 using the reversible-jump Markov Chain Monte Carlo (rjMCMC) algorithm72 and the topology of ML tree as input.

We also tested different combinations of ancestral population size (θs) and root age (τ0) priors in order to test their influence on species delimitation. Therefore, we evaluated scenarios characterized by either a small ancestral population (theta prior = 2, 2000) and deep divergences (tau prior = 1, 10) or a large ancestral population size (theta prior = 1, 10) and shallow divergences (tau prior = 2, 2000). In these cases, we carried out two runs combining the priors for ancestral population size and root age, based on 9 × 105 MCMC generations, sampled every two generations with a burn-in of 9 × 104 generations.

Biogeographic analyses

We used the R package BioGeoBEARS74,75 to estimate the ancestral biogeographic ranges of Stelliferinae based on the time-calibrated phylogeny inferred in StarBEAST3. We reconstructed historical biogeography by considering the distribution of species included in our phylogeny across six biogeographic provinces: the Carolinian, Caribbean, and Brazilian provinces in the western Atlantic31, and Cortez, Panamic, and Peruvian provinces in the eastern Pacific76. We employed the corrected Akaike Information criterion (AICc) to compare six biogeographic models: the dispersal, extinction and cladogenesis (DEC) model77; the dispersal-vicariance (DIVA-like) model78; the Bayesian estimation (BAYAREA-like) model79; and the three models above, incorporating the jump dispersal (j) parameter80 (DEC + J; DIVA-like + J; BayArea-like + j).