Haemulidae is composed of 136 fish species in 19 genera (Fricke et al., 2021). Two of these genera, Brachygenys Poey, 1868 and Haemulon Cuvier, 1829, are considered important groups of reef fish found in Brazil, given that they occur in large shoals, which are ecologically and commercially valuable (Rocha et al., 2008). The most recent review of the Haemulidae identified 21 species in the genus Haemulon (scaled-fin grunts), of which, 16 occur in the western Atlantic, while five species are found in the eastern Pacific (Tavera, Wainwright, 2019; Fricke et al., 2021). Menezes et al., (2003) recorded the occurrence of nine Haemulon species on the Brazilian coast: Haemulon aurolineatum Cuvier, 1830, H. chrysargyreum (Günther, 1859), H. melanurum (Linnaeus, 1758), H. parra (Demarest, 1823), H. plumieri (Lacepède, 1801), H. sciurus (Shaw, 1803), H. squamipinna Rocha & Rosa, 1999, H. steindachneri (Jordan & Gilbert, 1882) (currently identified as Haemulon atlanticus Carvalho, Marceniuk, Oliveira & Wosiacki, 2021 by (Carvalho et al., 2020), and H. striatum (Linnaeus, 1758).

Tavera, Wainwright, (2019) reassigned Haemulon chrysargyreum to the genus Brachygenys, based on morphological and molecular evidence, with the current valid name Brachygenys chrysargyrea (Günther, 1859) (Fricke et al., 2021). The genus Brachygenys (smallmouth grunts) includes only three species, Brachygenys californiensis (Steindachner, 1875) and B. jessiae (Jordan & Bollman, 1890), which are found in the eastern Pacific (Tavera, Wainwright, 2019), and B. chrysargyrea which occurs in the Western Atlantic, where it is restricted to the oceanic islands of Brazil, the Fernando de Noronha and Atol das Rocas archipelagos (Rocha, Rosa, 1999; Fricke et al., 2021).

The biological characteristics of the Brachygenys and Haemulon species, including their ample geographic ranges, ecological features, genetic patterns, and speciation mechanisms, has been the subject of many taxonomic, evolutionary, and phylogenetic studies (e.g., Rocha et al., 2008; Motta-Neto et al., 2011a, b; Sanciangco et al., 2011; Liang et al., 2012; Tavera et al., 2012, 2018; Bernal et al., 2017; Tavera, Wainwright, 2019). The formation of marine biogeographic barriers, in particular, the Isthmus of Panama, resulted in the establishment of geminal species (Jordan, 1908), that is, allopatric twin species in the eastern Pacific and western Atlantic oceans, as in the case of H. steindachneri from the eastern Pacific and H. atlanticus from the western Atlantic (Carvalho et al., 2020).

The principal objective of this study was to identify the Brachygenys and Haemulon species from Brazil based on the DNA barcode method. We also evaluated the influence of oceanic barriers on the dispersal of the study species.

Sample collection. Three specimens of Brachygenys chrysargyrea and 47 specimens of Haemulon (H. atlanticus = 15 specimens; H. aurolineatum = 17 specimens; H. melanurum = 4 specimens; H. parra = 5 specimens; H. plumieri = 5 specimens; and H. squamipinna = 1 specimen) were collected between 2006 and 2019 off the coasts of Brazil, between the northern extreme of the country and the southeastern state of São Paulo (Fig. 1; S1).

FIGURE 1 |

The species collected were identified based on their morphological characteristics (Lindeman, Toxey, 2002; Marceniuk et al., 2017). Barcode sequences of 149 specimens were obtained from the GenBank and BOLD databases (S1), and were inserted in the distribution map of the study species, in order to obtain a more ample sample of the different coastal regions of the Atlantic.

A small fragment of muscle tissue was removed from each specimen collected during this study and preserved in 95% ethanol at -20°C, before being deposited in the collection of the Fish Genetics and Biology Laboratory (Laboratório de Biologia e Genética de Peixes, LBP) at UNESP, in Botucatu, São Paulo, and the zoological collection of the Universidade Santa Cecília (AZUSC), in Santos, São Paulo (S2).

The species were sampled in accordance with Brazilian legislation, as regulated by the National Council for the Control of Animal Experimentation (CONCEA) and authorized by the Ethics Committee on the Use of Animals (CEUA) of the Biosciences Institute at UNESP through its (protocol 1057/2017).

Extraction of DNA, PCR amplification, and sequencing. The total DNA was extracted from the muscle tissue samples following the protocol proposed by Ivanova et al., (2006). Partial sequences of approximately 650 base pairs (bps) of the COI gene were obtained by PCR amplification using the FishF2 and FishR2 primers (Ward et al., 2005). The PCR reactions were run in a Veriti® 96-well Thermal Cycler (Applied BiosystemsTM or Mastercycler® EPGradient, Eppendorf) using the following temperature cycle: initial denaturation at 94°C for 4 min, followed by 30 cycles of denaturation at 94°C for 30 sec, annealing at 52°C for 30 sec, and extension at 68°C for 1 min, with a final extension at 68°C for 10 min. Each PCR solution comprised of: 7.55 μl of ultrapure water (milli-Q); 1.15 μl of (10X) buffer; 0.5 μl of MgCl₂ (50 mM), 0.5 μl of dNTPs (2 mM); 0.25 μl of each primer (5 mM); 0.2 μl of (5 U/μl) Taq DNA polymerase PHT (Phoneutria Biotechnologies and Services Ltda., Brazil), and 2 μl of the DNA template (50 – 100 ng/ul), for a final volume of 12.5 μl.

The amplification of the target sequence was confirmed by electrophoresis in 1% agarose gel using Blue Green Loading dye I (LGC Biotecnologia). The amplified PCR products were purified with an ExoSap-IT® (USB Corporation) solution, and the purified products were sequenced using the BigDye Terminator v3.1 Cycle Sequencing Ready Reaction kit (Applied Biosystems). This reaction solution consisted of: 3.9 μl of ultrapure water; 1.05 μl of 5X buffer; 0.7 μl of BigDye Terminator mix; 0.35 μl of the FishF2 or FishR2 primers (10 mM), and 1.0 μl of the purified PCR product (50 ng/μl). The amplification cycle was: 2 min at 96°C, and 35 cycles of 30 sec at 96°C, 15 sec at 54°C, and 4 min at 60°C. The purified PCR products were then precipitated in EDTA 125 nM/sodium acetate/alcohol, and the samples were sequenced automatically using an ABI 3130X1 Genetic Analyzer sequencer (Applied Biosystems^TM).

Data analysis. The sequences were edited and aligned using the Geneious Pro 4.8.5 software (Kearse et al., 2012). The edited sequences were compared with those deposited in the National Center for Biotechnology Information (NCBI) GenBank using the BLASTn tool (Johnson et al., 2008). The final matrix had 199 sequences, including 50 obtained in the present study and 149 extracted from GenBank (ncbi.nlm.nih.gov/genbank) or BOLD (boldsystems.org) (S1). The end alignment was exported and analyzed to generate trees in the MEGA v 7.0 software (Kumar et al., 2018), based on the neighbor-joining (NJ) method using the Kimura -2- Parameter model (K2P) (Kimura, 1980) and the maximum likelihood (ML) method with the best Tamura-Nei model (Kumar et al., 2018) and gamma distribution (TRN+G) identified by the PartitionFinder software (Lanfear et al., 2012). All trees were tested by bootstrap, with 1000 pseudoreplicates (Felsenstein, 1985).

Species delimitation analyses. Three methods of species delimitation were used: (1) The Automatic Barcode Gap Discovery, ABGD (Puillandre et al., 2012) which is based on a pairwise genetic distance matrix (generated in MEGA V7.0) run on the ABGD web server (https://bioinfo.mnhn.fr/abi/public/abgd/abgdweb.html) with the Kimura distance model (K2P) and other parameters at default (Pmin = 0.001; Pmax = 0.1); (2) the Bayesian Poisson Tree Process PTP (Zhang et al., 2013) run on the PTP web server (species.h-its.org/ptp), using the best Maximum Likelihood (ML) tree, 10,000 MCMC generations, and a 0.1 burn-in rate as the default settings, and (3) the general mixed Yule-coalescent GMYC (Pons et al., 2006; Fujisawa, Barraclough, 2013), run on the GMYC web server (https://species.h-its.org/gmyc/).

This analysis was conducted using the ultrametric gene tree estimated from the birth-death prior and the relaxed lognormal parameters. The number of polymorphic sites, the number of haplotypes, and the haplotype (HD) and nucleotide diversity (Pi) were estimated using DnaSP v5 (Librado, Rozas, 2009), with the median-joining network being produced using the PopArt program (Leigh, Bryant, 2015), for mutational analyses.

Barcode sequences were obtained from 50 specimens, which were complemented with 149 sequences obtained from GenBank and BOLD (S1), totaling 199 sequences in the final matrix (184 for Haemulon and 15 for Brachygenys) representing the different regions of the western Atlantic and eastern Pacific. The barcode sequences obtained here ranged in length from 440 to 558 bps. The overall mean nucleotide frequencies were 23.1% adenine, 26.6% cytosine, 19.6% guanine, and 30.6% thymine. No stop codons, deletions or insertions were found in any of the sequences.

The intraspecific genetic distances (based on the K2P model) ranged from 0.001±0.001, in both H. melanurum and H. parra, to 0.027±0.005 in H. plumieri (Tab. 1). The interspecific values ranged from 0.0746±0.0121 between H. steindachneri and H. atlanticus to 0.1645±0.0212 between B. jessiae and H. atlanticus (Tab. 1).

The results of the Maximum Likelihood (ML) analyses and the GMYC species delimitation method indicated the presence of 15 MOTUs in the database, while the ABGD and PTP species delimitation methods identified 13 MOTUs (Fig. 2; S3).

FIGURE 2 |

A single MOTU was found in B. californiensis, B. chrysargyrea, B. jessiae, H. atlanticus, H. melanurum, H. parra, H. squamipinna, and H. steindachneri in all analyses. Three MOTUs were identified in H. aurolineatum based on to the PTP and ABGD approaches, while the ML and GMYC methods identified four units in the samples of this species. In H. plumieri, the PTP and ABGD approaches identified two MOTUs, while the ML and GMYC methods identified three (Fig. 2; S3). Given these results, we decided to calculate the genetic distances between these MOTUs using the K2P model (Tab. 2), as described below.

The three species delimitation methods used in the present study (PTP, ABGD, and GMYC) identified only one H. atlanticus MOTU in all the samples analyzed (Fig. 2). However, the ML tree included two lineages, H. atlanticus A, with 19 specimens from the Brazilian states of Pará, Ceará, Alagoas, and São Paulo, as well as Colombia, and H. atlanticus B (5 specimens from Colombia, Guatemala, and Venezuela). The genetic distance between the H. atlanticus A and B lineages was 0.012±0.004. The comparative network analysis of H. steindachneri, H. atlanticus A, and H. atlanticus B further reinforced the presence of three groups (Fig. 3), with 30 mutations separating H. steindachneri (Pi = 0.00271; HD = 0.81667) from H. atlanticus (Pi = 0.00567; HD = 0.65217), and nine mutations between H. atlanticus A Pi = 0.00177; HD = 0.45614) and H. atlanticus B (Pi = 0.00155; HD = 0.70000).

In the case of H. aurolineatum, the ML and GMYC identified four MOTUs, denominated here as H. aurolineatum A (four specimens from Bermuda), H. aurolineatum B (one specimen from Bermuda), H. aurolineatum C (four specimens from the Gulf of Mexico), and H. aurolineatum D, with 49 specimens from Belize, Brazil, Colombia, Jamaica, Venezuela, and the Gulf of Mexico (Fig. 2; S3).

The genetic distances between the pairs of these MOTUs ranged from 0.0183±0.049 between H. aurolineatum lineages C and D to 0.1382±0.0172 between H. aurolineatum lineages A and B (Tab. 2). When the PTP and ABGD methods were applied, however, only three MOTUs were observed - H. aurolineatum A, H. aurolineatum B, and H. aurolineatum C+D (Fig. 2; S3). The genetic distances between these MOTUs were 0.1382±0.0173 between H. aurolineatum A and B, 0.1089±0.0147 between H. aurolineatum A and C+D, and 0.1059±0.0150 between H. aurolineatum B and C+D.

There were 37 mutations between H. aurolineatum A (Pi = 0.00000; HD = 0.00000) and H. aurolineatum D (Pi = 0.00389; HD = 0.72364), five between H. aurolineatum C and H. aurolineatum D (Pi = 0.00281; HD = 0.83333), and 34 mutations between H. aurolineatum B and H. aurolineatum C (Pi = 0.00000; HD = 0.00000) (Fig. 3).

FIGURE 3 |

The ML and GMYC analyses identified three MOTUs in H. plumieri, identified here as H. plumieri A (20 specimens from the Caribbean and Gulf of Mexico), H. plumieri B (five specimens from the Gulf of Mexico), and H. plumieri C, with 10 specimens from Brazil and the Caribbean (Fig 2; S3). The genetic distances between these MOTUs were 0.064±0.011 between H. plumieri lineages A and B, 0.048±0.010 between H. plumieri A and C, and 0.024±0.006 between H. plumieri B and C (Tab. 2). When the PTP and ABGD methods were considered, however, only two MOTUs were observed, one containing H. plumieri A+B and the other, H. plumieri C (Fig. 2; S3). The genetic distance between H. plumieri A+B and H. plumieri C was 0.052±0.009. There were 11 mutations between H. plumieri C (Pi = 0.00194; HD = 0.20000) and H. plumieri B (Pi = 0.00097; HD = 0.40000) and 19 mutations between H. plumieri C and H. plumieri A (Pi = 0.00394; HD = 0.87368) (Fig. 3).

Four of the species analyzed, B. chrysargyrea, H. melanurum, H. parra, and H. squamipinna, presented extremely low intraspecific distances, and all the different analytical approaches indicated that they represented a single MOTU. In the specific case of H. squamipinna, the evidence that the samples represented a single species was expected, given the very restricted distribution of the species off the northeastern coast of Brazil (Rocha, Rosa, 1999). In the other cases, however, the species are much more amply distributed, with H. melanurum being found from southeastern Florida to northern Brazil, including the whole of the Caribbean (Menezes et al., 2003), while H. parra and B. chrysargyrea are distributed from southeastern Florida to southeastern Brazil (Menezes et al., 2003).

In all three cases, the distribution of the species straddles the potential barrier formed by the Amazon-Orinoco Plume. In addition, the southern limit of the distribution of H. parra in Brazil was originally described as being São Paulo (Menezes et al., 2003), and we collected samples in São Paulo during the present study, which indicates that this species also traverses the Vitória Trindade seamount chain, off the eastern coast of Brazil, which was a potential biogeographic barrier during periods of marine regression, in the Quaternary and Tertiary, as observed in other genera and species, such as Orthopristis ruber (Cuvier, 1830) (Marceniuk et al., 2019), Macrodon ancylodon (Bloch & Schneider, 1801) (Santos et al., 2006), and Chaetodipterus faber (Broussonet, 1782) (Machado et al., 2017).

Some of the haemulids from the western Atlantic have ample geographic distributions and have larvae that are able to disperse rapidly on oceanic currents, as well as the ability to migrate vertically in the water column (Majoris et al., 2019). These characteristics would maximize the potential for gene flow between the northern and southern populations of these species in the Western Atlantic, which would minimizing the chances of forming isolated groups (Rocha, 2003; Rocha et al., 2002, 2005, 2007, 2008). The ample geographic ranges and genetic homogeneity detected here in H. melanurum, H. parra, and B. chrysargyrea may thus also be at least partially due to the swimming capabilities of these fishes.

Some Haemulon species have dispersed from the Pacific Ocean to the western Atlantic (Tavera et al., 2012), followed by reverse invasions during the occurrence of vicarious events and the formation of the Isthmus of Panama (Stange et al., 2018). These historical processes resulted in allopatric speciation, which has given rise to twin species (Jordan, 1908), and both sister and novel lineages (Rocha et al., 2007; Luiz et al., 2012; Tavera et al., 2018; Tavera, Wainwright, 2019). In contrast with the species represented by a single lineage in the western Atlantic, as discussed above, the presence of multiple MOTUs in H. atlanticus, H. aurolineatum, and H. plumieri is an intriguing phenomenon, which cannot be accounted for by either the current barriers to gene flow (such as the Amazon-Orinoco Plume) or ancient processes, such as the isolation of the fauna of the Gulf of Mexico and the southern coast of Brazil (Victoria Trindade seamount chain) during the Quaternary and Tertiary glacial cycles.

Until recently, H. steindachneri was believed to occur in both the Pacific and Atlantic oceans, but an ample taxonomic review, supported by cytogenetic and molecular data (Rocha et al., 2008; Tavera et al., 2012, 2018; Bernal et al., 2017, 2019; Motta-Neto et al., 2012, 2019; Tavera, Wainwright, 2019), confirmed that H. steindachneri is restricted to the Pacific, while a new species, H. atlanticus, was described for the Atlantic (Carvalho et al., 2020). Ours results also support the separation of the eastern Pacific H. steindachneri from H. atlanticus, which is found in the western Atlantic (Carvalho et al., 2020), with a genetic distance of 0.0746±0.0121 (Tab. 1), and the two species were separated clearly in all analyses.

However, while all three species delimitation methods used here (PTP, ABGD, and GMYC) indicated that only one MOTU was present in the H. atlanticus samples, the ML tree identified two groups, denominated here as H. atlanticus A (specimens from the Caribbean and the Atlantic coast of South America) and H. atlanticus B (specimens from the Caribbean). The genetic distance between these two MOTUs was 0.012±0.004, which is lower than that usually found between valid marine fish species, i.e., around 2% (Ward, 2009), although it was consistent with that found by Rocha et al., (2008) and Carvalho et al., (2020), which supports the need for further studies to determine whether this difference is due to the simple isolation of populations or the existence of a cryptic species derived from a recent speciation process (Rocha et al., 2007; Motta-Neto et al., 2011a).

In H. aurolineatum, by contrast, some of the analyses identified four MOTUs, that is, H. aurolineatum A and B (specimens from Bermuda), H. aurolineatum C (specimens from the United States), and H. aurolineatum D (specimens from the Gulf of Mexico, Caribbean, and the Atlantic coast of South America), while the other analyses indicated that H. aurolineatum C and D constitute a single MOTU (C+D). In H. plumieri, some of the analyses identified three MOTUs, that is, H. plumieri A (specimens from the Gulf of Mexico and the Caribbean), H. plumieri B (specimens from the Gulf of Mexico), and H. plumieri C (specimens from the Caribbean and the Atlantic coast of South America), while the other analyses indicated that H. plumieri A and B constitute a single MOTU (A+B). Although the genetic distances between some of these MOTU pairs were lower than 2%, many were higher, and reached up to 13%.

In recent years, advances in sequencing technology have supported a substantial increase in the DNA sequences available in databases, such as GenBank, for biodiversity studies. This includes fish, and these advances have contributed to the identification of species from groups with major taxonomic disagreements (Porter, Hajibabaei, 2018; Leray et al., 2019).

However, some authors have questioned the validity of the discrimination of fish species based on DNA sequences, given the frequent misidentification of the sequences available in these databases due to taxonomic inconsistencies, sampling errors, contamination, and hybridization, which reduces their reliability for comparison with other sequences (Locatelli et al., 2020; Pentinsaari et al., 2020). This problem was identified in the case of the five H. aurolineatum GenBank sequences from Bermuda, which we believe have been identified mistakenly as H. aurolineatum, given their considerable genetic distance from the other specimens analyzed, as well as the lack of published reports of this species in the region of the Bermuda archipelago.

Studies of reef fish have identified a number of processes that may have influenced the present-day intra- and inter-specific structuring observed in some groups found in the western Atlantic (Santos et al., 2006; Rodríguez-Rey et al., 2014; Silva et al., 2014, 2015; Ashe et al., 2015; Souza et al., 2015; Bernal et al., 2018, 2019).

These processes include the glacial cycles of the Pleistocene and Miocene, when the Vitória Trindade seamount chain, off the eastern coast of Brazil, isolated the fish fauna of southern Brazil, with a similar process of isolation occurring in the Gulf of Mexico. Another important event was the uplifting of the Andes, around 8 million years ago, which reconfigured the continental drainages of South America, establishing the transcontinental flow of the Amazon River to the Atlantic Ocean.

All these processes may have caused alterations in the gene flow of reef fish populations. In the specific cases of H. atlanticus, H. aurolineatum, and H. plumieri, however, the well-known barriers, such as the Amazon-Orinoco plume and the Victoria Trindade seamount chain would not account for the differentiation of the MOTUs found herein, which implies the influence of alternative phenomena, such as other paleogeographical events or even parametric or sympatric speciation events resulting from processes of ecological specialization.

The DNA barcode method contributed to the identification of the Brazilian reef fish fauna of the genera Brachygenys and Haemulon, and how barriers and ocean currents may influence the population dynamics of these species in the Western Atlantic. Our results indicated that some Haemulon species have been able to traverse the barrier of the Amazon plume and that the action of ocean currents may contribute to the dispersion of these species. However, the evolution of the populations of H. plumieri, H. atlanticus, H. steindachneri, and H. aurolineatum may have been influenced by variations in oceanographic conditions and barriers resulting in the formation of distinct MOTUs, as identified here, that have enriched the diversity of the reef fish species found off Brazil.