The Serra Spanish mackerel, Scomberomorus brasiliensis (Collette et al., 1978), is a pelagic marine fish distributed in the Caribbean Sea and western South Atlantic, between southern Belize and the extreme south of Brazil, in Rio Grande do Sul (Collette and Russo, 1984). This mackerel is common near islands and open beaches and appears to be able to migrate long distances on a seasonal basis (Collette and Russo, 1984; Gold et al., 2010). The diet of S. brasiliensis is omnivorous, and includes a range of fish, crustaceans, and mollusks (Fonteles-Filho, 1988). Reproduction is oviparous, with pelagic eggs and larvae (Gesteira and Mesquita, 1976).

The Serra Spanish mackerel is one of the most important fishery resources in the western Atlantic. In Brazil alone, between 2009 and 2011, the annual catch of this fishery resource was approximately 9,500 t (Maia et al., 2015). In Brazil, 2,733,000 tons of S. brasiliensis were caught between 1970 and 2000 (FAO, 2000), with the state of Ceará contributing 40.7% of the total catch. The Brazilian bulletin of fishery and agricultural statistics reported a mean annual catch of 433,000 tons between 2009 and 2011 (MPA, 2011).

Nóbrega and Lessa (2009) found evidence of the overfishing of S. brasiliensis off the coast of the Brazilian Northeast, observing declining catches of this species and its congener, the King mackerel, Somberomorus cavalla (Gonçalves et al., 2003; Nóbrega and Lessa, 2009). In this case, the analysis of phylogeographic patterns and the genetic diversity of commercially exploited organisms is essential for the development of adequate conservation and management policies for these species (Oliveira et al., 2009, 2014; Reiss et al., 2009; Silva et al., 2015). There are a few studies of the genetic diversity of fish species exploited commercially in Brazil (Santa Brígida et al., 2007; Rodrigues et al., 2008; Silva et al., 2015) although none has contemplated S. brasiliensis. Using microsatellite markers, Gold et al. (2010) found evidence of structuring in the S. brasiliensis populations between the coast of Venezuela (Margarita Island and Cumana) and Trinidad and Tobago, which was corroborated by mitochondrial sequences. Up until now, however, there has been no molecular investigation of the Brazilian populations of S. brasiliensis, so it is not known whether these stocks are genetically distinct from those of the Caribbean region, or whether the populations of the western South Atlantic are genetically contiguous.

The present study evaluated in detail the phylogeographic patterns, population dynamics, and genetic diversity of S. brasiliensis on the western coast of the South Atlantic, based on the sequences of two distinct regions of the mitochondrial genome (including previously published data) – MT-CYB and MT-ND4 and compared the diversity indices with those available for the individuals from Brazil, Venezuela, and Trinidad and Tobago.

The sequences were aligned using CLUSTAL W (Thompson et al., 1994), implemented in BioEdit V.7.0.4 (Hall, 1999). The phylogenetic relationships were analyzed in PAUP^∗ 4.0b10 (Swofford, 2002) using the Neighbor-Joining (NJ) and Maximum Likelihood (ML) approaches. The nodes were estimated using 1,000 bootstrap pseudo-replicates (Felsenstein, 1985) and heuristic searches. The sister species Scomberomorus cavalla was used as the outgroup (Supplementary Material 1). The ML tree generated here was also used to construct the haplotype network in the Haploviewer program (Salzburger et al., 2011). The most adequate evolutionary models (TIM1 + I for MT-ND4 and HKY for MT-CYB) for the data were selected using jModelTest 2 (Darriba et al., 2012), based on the Akaike Information Criterion (AIC). We used the ARLEQUIN 3.5 software (Excoffier and Lischer, 2010) to estimate the haplotype (h; Nei, 1987) and nucleotide diversity (π; Nei, 1987) of each population and the neutrality tests. We calculated Tajima’s D (Tajima, 1989) and Fu’s Fs (Fu, 1997), which are the most sensitive indices for the detection of deviations from neutrality due to recent population expansion or bottlenecks. The genetic differentiation between pairs of populations was estimated using the Fst fixation index (Excoffier et al., 1992), with its significance being tested by 10,000 permutations. The genetic variability among populations was tested using a hierarchical Analysis of Molecular Variance, AMOVA (Excoffier et al., 1992) run in Arlequin v.3.5.

A Bayesian Skyline Plot (BSP) was run in BEAST2 v2.0.3 (Bouckaert et al., 2014) for 200 million generations, with samples taken at intervals of 10,000 generations, using the default parameters and priors (HKY substitution model) for both markers (MT-CYB and MT-ND4), with the strict molecular clock model, tree prior and 10% of which were discarded as burn-in. The convergence and mixing of the chains were inspected visually in TRACER v1.6 (Rambaut et al., 2014). The convergence and mixing were appropriate when all of the Effective sample size (ESS) values for each of the parameters analyzed were above 200. For the MT-CYB gene with three groups and the piecewise-constant skyline model, auto-optimize operators, and log parameters every 10,000 interactions. For the MT-ND4, the plot had seven groups and a piecewise-constant skyline model, auto-optimize operators, and log parameters every 10,000 interactions. For this, two simulations were run, one based on a strict molecular clock, calibrated by the mutation rate, and the other with a relaxed, log-normal non-correlated clock. The results of the two analyses (strict and relaxed clocks) were compared using the Bayes factor, run in TRACER v1.6 (Rambaut et al., 2014). The strict model was the most adequate for our data (lnBayes Factor = 3.72) (Suchard et al., 2001), using a substitution rate of 1% per million years (Donaldson and Wilson, 1999; Craig et al., 2006), with a standard deviation of 10% for the two genes separately, MT-ND4 and MT-CYB.

We obtained 105 MT-CYB sequences with total length of 410 bps from the S. brasiliensis specimens representing three Brazilian populations. Overall, 19 of the nucleotides were variable, and six sites were parsimonious. A total of 19 haplotypes were detected, including 11 singletons and eight shared haplotypes, one of which was very common, being shared by 54 (51%) of the specimens and all of the Brazilian populations analyzed (Figure 2A and Supplementary Material 2).

The three S. brasiliensis populations presented haplotype diversity (Hd) of between 0.701 and 0.808, while nucleotide diversity (π) ranged from 0.004 to 0.002 (Table 1). Only the total population and that from Macapá presented significantly negative deviations from neutrality, whereas the other populations returned negative, but non-significant values (Table 1). Divergence (P) varied between 0.0 and 1.5%, with a mean population divergence of 1%. Values for intra- and inter-population divergence were all below 1%. According to the PhiST values, none of the comparisons between populations were significant (Table 2). This is corroborated by the AMOVA, which indicated no significant structural arrangement. The overall analysis revealed that almost 99% of the observed variance was derived from within-population differentiation (Table 2).

A dataset of 188 sequences of 500 bps was compiled for the MT-ND4 gene, based on the 188 S. brasiliensis specimens representing the South American distribution of the species (Figure 1B and Supplementary Material 2). In this case, haplotype diversity (Hd) ranged between 0.523 and 0.866, and nucleotide diversity (π) varied from 0.002 to 0.013 (Table 1). A total of 48 haplotypes were identified, one of which was quite common, being shared by 88 (47%) of the specimens and found in all of the populations analyzed (Figure 2B and Supplementary Material 2). The individuals from Cumana were the most variable of all the study localities, with some haplotypes being separated by up to 12 mutations (Figure 2B). The Fst (Table 3), AMOVA, and haplotype network all indicate the existence of a discrete genetic structure in this region.

All the results of the Analysis of Molecular Variance (AMOVA) indicated that the variation occurs primarily within populations, rather than between them, with low and non-significant PhiST values being obtained (Table 2). However, in the specific case of the MT-ND4 gene, when we tested Cumana and all the other regions (AMOVA 1), 12.9% of the variation was explained by the differences between this region and the other localities (Table 2). This result is supported by the Fst indices between localities, which were significant (Table 3).

In the case of the neutrality tests, both Fu’s F and Tajima’s D were significant only when the S. brasiliensis individuals were grouped within a single population for both mitochondrial genes. In accordance with Fu’s F and Tajima’s D for the total population, the Bayesian Skyline Plot presented evidence of demographic expansion in the S. brasiliensis population, dated to the Holocene in both markers. The phylogenetic analyses were run only for the MT-ND4 gene. The phylogeographic reconstructions based on the NJ and ML approaches identified two strongly supported groups of haplotypes. The first clade was formed by the four haplotypes from Cumana, which is represented by the yellow circles to highlight that there are two divergent groups (bootstrap NJ 97% and 80% ML), while the second clade contains the remaining S. brasiliensis haplotypes (Figure 1B).

Scomberomorus brasiliensis is one of the most important fishery resources in the central Atlantic and northern Brazil, which may be suffering the effects of overfishing in some Brazilian regions (Nóbrega and Lessa, 2009). The present study analyzed datasets obtained from two distinct regions of the mitochondrial genome of S. brasiliensis specimens from the Brazilian coast, MT-CYB and MT-ND4. The MT-ND4 sequences were obtained from 188 S. brasiliensis specimens representing the known distribution of the species at South American.

The MT-ND4 sequences were structured in the S. brasiliensis specimens from Cunama, as indicated by the formation of two well-supported haplotype groups in the NJ and ML trees (Figure 1B). The Fst indices (Table 3) also indicated a greater degree of differentiation of these haplotypes, in comparison with the other localities. Gold et al. (2010) found genetic structuring in the Cumana specimens, a pattern like that observed in the present study, which was supported by the phylogeographic reconstructions (Figure 1B). One reason for the divergence of the Cumana population may be its geographic location (Figure 1A), which is relatively isolated from the influence of adjacent marine environments, a situation that may limit its contact with other populations. Even so, fidelity to spawning sites, adaptations to specific oceanographic conditions, or feeding preferences may also be relevant here (Gold et al., 2010; López et al., 2010; Zhu et al., 2014; Córdova-Alarcón et al., 2019).

It is important to note, however, that no evidence from either the population analyses or the haplotype network indicated any structuring among the other sites analyzed. According to the mitochondrial data, then, there is a single genetic stock of S. brasiliensis between eastern Venezuela and southern Brazil (Figure 1B and Table 3). In the MT-ND4 network, the most frequent haplotype was found in 47% of the specimens, and in all the S. brasiliensis populations (Figure 2B). As for the MT-ND4 gene, no population structuring was observed in the MT-CYB sequences, and the extremely low Fst values recorded among the three localities further reinforce the lack of population structure.

In the case of the MT-CYB, all three populations presented high levels of haplotype diversity (Hd: 0.701–0.808), but while low levels of nucleotide diversity were recorded in Macapá and Fortaleza, this value was relatively high in Paranaguá (Table 1). High levels of haplotype diversity (Hd: 0.523–0.866) were also recorded in the MT-ND4 gene, with low levels of nucleotide diversity (π: 0.002–0.013). All these values are relatively high in comparison with the sister species, S. cavalla, for which, Santa Brígida et al. (2007) recorded a Hd of 0.704.

This observed genetic homogeneity may be related to the dispersal capacity of the members of the family Scombridae (Collette and Russo, 1984; Moyle and Cech, 2004; López et al., 2010). The results of the present study are consistent with those of Batista and Fabré (2001), who concluded that the coast of the Brazilian state of Maranhão is a part of the species’ migratory circuit, which may exceed 300 nautical miles. Few physical barriers to dispersal exist in the ocean, and the dispersal of pelagic eggs and larvae on ocean currents, such as the Brazilian and North Brazil currents (see Rocha-Olivares et al., 2000; Broughton et al., 2002; Córdova-Alarcón et al., 2019), may also have contributed to the genetic homogeneity observed in the present study.

A similar pattern of genetic connectivity has been observed in other fish species, such as the King mackerel, Scomberomorus cavalla, on the northern and northeastern coasts of Brazil (Santa Brígida et al., 2007), and the red snapper, Lutjanus campechanus, and Acoupa weakfish, Cynoscion acoupa, populations of the Brazilian coast, in which no structuring was found in mitochondrial markers (Broughton et al., 2002; Gomes et al., 2008; Rodrigues et al., 2008; Silva et al., 2015). Furthermore, Siccha-Ramírez et al. (2018) utilizing SNPs (Single Nucleotide Polymorphisms) analyze samples from two commercially important species (S. brasiliensis and Thunnus albacares), where the results suggest the absence of any genetic structure among the local populations for both species, despite the sampling points are distant approximately 3,000 km.

As migratory pelagic fish generally present low levels of geographic differentiation across their distribution in the oceans due to their way of life as pelagic larval life with external fertilization and the swimming ability of the adults, strong genetic heterogeneity has nevertheless been detected at the regional scale in Scomberomorus commerson (López et al., 2010; Fauvelot and Borsa, 2011). Further studies of the Brazilian S. brasiliensis populations, based on more variable markers, such as microsatellites and new generation sequencing, may nevertheless reveal yet undetected structuring.

The BSP reconstructions indicate that the population of S. brasiliensis expanded between approximately 7 and 10 million years ago. The MT-CYB data indicate that the Ne of the Brazilian population of S. brasiliensis increased around 10,000 years ago (Figure 3A). In the case of the MT-ND4 gene, the whole population (Ne = 29 × 10¹⁴) expanded at approximately 7.5 million years ago (Figure 3B). It is important to treat the estimates obtained in the present study with caution, however, because the divergence rates of the mitochondrial genes are highly variable in fish (Donaldson and Wilson, 1999; Waters et al., 2007).

The Holocene is characterized by an interglacial period that began approximately 10,000 years ago, when sea levels began to rise following the glaciations of the Pleistocene (Lambeck et al., 2002). The Serra Spanish mackerel is associated with areas of high productivity in which larval growth is favored (Lauth and Olson, 1996), and the increase in sea levels may have contributed to an increase in primary productivity, favoring the expansion of the S. brasiliensis populations into the tropical and subtropical waters of the Atlantic (Lee et al., 1995; Lambeck et al., 2002). This probable expansion of the population is supported by the negative and highly significant Fs and D values, which indicate a highly stable population with a long evolutionary history (Table 1).

Given the fact of S. brasiliensis is one of the most harvested marine fish species on the Brazilian coast we suggest that population of the southwestern Atlantic may correspond to a single genetic stock, and should thus be treated as a unit. The genetic variability observed in the S. brasiliensis population sampled in the present study is important for the definition of effective conservation strategies, with the primary objective of mediating the decline of the population and guaranteeing the sustainability of this fishery resource (Ward, 2000; López et al., 2010; Souza-Shibatta et al., 2018; Córdova-Alarcón et al., 2019). Given this, further research on the S. brasiliensis populations should adopt a multi-locus approach integrating microsatellites, SNPs, and RADseq (see Siccha-Ramírez et al., 2018), which would complement the findings of the present study, based on two mtDNA loci. The findings of the present study provide important insights into the genetic diversity and the potential distribution of the stocks of S. brasiliensis and provide a fundamental contribution to the development of effective, long-term management strategies for the populations of the species.

The datasets generated for this study can be found in the online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

Ethical review and approval was not required for the animal study because all Brazilian samples of S. brasiliensis were obtained directly from local fish markets or other retail outlets. The species is not subject to any legal restrictions in Brazil, according to the Brazilian Environment Institute (IBAMA).

DC: acquisition of samples, laboratory procedures, analysis and interpretation of the data, development of the intellectual content, drafting and writing the manuscript, and the approval of the final version. PR: collected samples and drafted the manuscript. CQ: acquisition of samples, laboratory procedures, analysis and interpretation of data, draft of the manuscript, and the development of intellectual content. LR-F: analysis and interpretation of data, development of intellectual content, manuscript writing, and final version approval. JL: analysis and interpretation of data, draft work, development of intellectual content, writing of the manuscript, and approval of the final version. IS: interpretation of data, development of intellectual content, manuscript writing, and approval of the final version. MV: interpretation of data, development of intellectual content, manuscript writing, and final version approval. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.