Introduction
The apparent weakness of geographic barriers over large distances in marine environments hinders the identification of cladogenesis and several cases of cryptic lineages might remain overlooked (Rocha, Reference Rocha2003; Luiz et al., Reference Luiz, Madin, Robertson, Rocha, Wirtz and Floeter2012; Da Silva et al., Reference Da Silva, Marceniuk, Sales and Araripe2016). Moreover, considering that speciation processes might take place without major changes in morphology, the actual number of species might be underestimated when external body features are analysed in isolation (Winker, Reference Winker2005). Failure of species diagnosis has a particularly negative impact on economically important species in fisheries where distinct morphotypes putatively related to a single species might encompass multiple evolutionary units (Mayden, Reference Mayden, Claridge, Dawah and Wilson1997; Winker, Reference Winker2005; Da Silva et al., Reference Da Silva, Schneider, Sampaio, Angulo, Brito, de Santos, Santos, de Carvalho-Filho and Santos2018).
The mojarras (genus Eucinostomus, family Gerreidae) comprise 11 nominal taxa (Froese & Pauly, Reference Froese and Pauly2019) widespread in the Western Atlantic (WA; seven species) and Eastern Pacific (EP; four species) that have commercial relevance to local fisheries. These fish are found in coastal areas, including estuaries and hypersaline lakes because of their high osmoregulation adaptability (Nelson et al., Reference Nelson, Grande and Wilson2016). Morphologically, they are characterized by dorsal compression, protractile mouth, smooth gill bones, cycloid scales on the head and ctenoid scales over the body (Nelson et al., Reference Nelson, Grande and Wilson2016). From a systematic viewpoint, the genus Eucinostomus is considered one of the most problematic genera of coastal fishes from the New World since their putative interspecific morphological traits are subtle and usually overlapped, while their evolutionary interrelationships remain poorly known (Matheson & McEachran, Reference Matheson and McEachran1984; De La Cruz-Agüero & Galvan-Magana, Reference de la Cruz-Agüero and Galvan-Magana1993; De La Cruz-Agüero, Reference De La Cruz-Agüero2013). Because of their controversial taxonomy, some species in this group, such as Ulaema lefroyi (E. lefroyi), are regarded as inquirendae (Nelson et al., Reference Nelson, Grande and Wilson2016). In addition, other taxonomic methods such as cytogenetic analyses have been inefficient in discriminating mojarra species (Calado et al., Reference Calado, Bertollo, Cioffi, Costa, Jacobina and Molina2014).
The difficulties in recognizing Eucinostomus species based on traditional taxonomy were recently reported in a wide study of DNA barcoding in fish species from north-eastern Brazil, WA (Brandão et al., Reference Brandão, de Bitencourt, Santos, Watanabe, Schneider, Sampaio and de Affonso2016). These authors verified that 82 specimens of mojarras morphologically identified as Eucinostomus melanopterus or only to the genus level (Eucinostomus sp.) encompassed five molecular units with deep genetic divergence, representing E. melanopterus (N = 5), E. lefroyi (N = 7), E. gula (N = 2), E. harengulus (N = 64) and E. jonesii (N = 4). Moreover, this work represented the first report about the occurrence of E. harengulus and E. jonesii in north-eastern Brazil.
The accurate diagnosis of taxa is a key step to biodiversity conservation (Bickford et al., Reference Bickford, Lohman, Sodhi, Ng, Meier, Winker, Ingram and Das2007). The utilization of mitochondrial markers such as the Cytochrome c oxidase subunit I (COI) as DNA barcodes has been particularly useful in the identification of cryptic species and in the resolution of taxonomic uncertainties, with several examples in fishes (Ferreira et al., Reference Ferreira, Kavalco, de Almeida-Toledo and Garcia2014; Hyde et al., Reference Hyde, Underkoffler and Sundberg2014; Winterbottom et al., Reference Winterbottom, Hanner, Burridge and Zur2014; Barreira et al., Reference Barreira, Lijtmaer and Tubaro2016; Nirchio et al., Reference Nirchio, Paim, Milana, Rossi and Oliveira2018). This approach has also been improved over the last decade by the incorporation of distinct algorithms to test the hypothesis of evolutionary independent lineages and to avoid synonyms in the dataset (Brown et al., Reference Brown, Collins, Boyer, Lefort, Malumbres-Olarte, Vink and Cruickshank2012; Puillandre et al., Reference Puillandre, Lambert, Brouillet and Achaz2012; Zhang et al., Reference Zhang, Kapli, Pavlidis and Stamatakis2013; Luo et al., Reference Luo, Ling, Ho and Zhu2018).
Considering the problematic systematics, the wide range of mojarras and the efficiency of DNA-based identification, we assessed the genetic diversity of all Eucinostomus representatives with available COI sequences throughout most of their geographic distribution. Based on distinct analytical methods, including species delimitation algorithms, we reviewed their systematic relationships within a phylogeographic context, providing evidence of overlooked phylogenetic diversity and variation in the range of evolutionary units, with emphasis in the WA.
Materials and methods
Sampling
A total of 400 COI sequences of Eucinostomus, comprising 11 recognized taxa were analysed: E. argenteus (N = 90), E. cf. gula (N = 6), E. currany (N = 7), E. dowii (N = 6), E. entomelas (N = 10), E. gracilis (N = 2), E. gula (N = 49), E. havana (N = 2), E. harengulus (N = 88), E. jonesii (N = 86), E. lefroyi (N = 11) and E. melanopterus (N = 20), as well as other sequences named as Eucinostomus sp. (N = 23). These sequences were downloaded from the public datasets, for example Bold Systems and GenBank (NCBI) (supplementary material 1). We also added 16 specimens of mojarras, E. argenteus (N = 1), E. gula (N = 8), E. jonesii (N = 2), E. lefroyi (N = 5), collected in the state of Rio Grande do Norte (05°05′26″S 36°16′031W″), north-eastern Brazilian coast, that were identified according to Woodland (Reference Woodland, Carpenter and Niem2006). Fish were collected by gillnets and euthanasia was accomplished by immersion in cold water for 10–15 min. Afterwards, muscle tissues were removed from each individual and stored in ethanol at −20 °C (Blessing et al., Reference Blessing, Marshall and Balcombe2010).
DNA isolation, amplification and sequencing
Total DNA was isolated from stored muscle tissues, using the DNeasy kit (QIAGEN). A fragment of 645 base pairs (bp) of the COI gene was amplified via PCR using the primers VF2_t1 (5′TGTAAAACGACGGCCAGTCAACCAACCACAAAGACATTGGCAC3′) and FishR2_t1 (5′CAGGAAACAGCTATGACACTTCAGGGTGACCGAAGAATCAGAA3′) as described by Ward et al. (Reference Ward, Zemlak, Innes, Last and Hebert2005). The reactions encompassed 12 μl of 2× Taq master mix (Vivantis), 2 μl of template DNA solution at 40 ng μl−1, 0.5 μl of each primer (10 mM) and ultrapure water to a final volume of 25 μl. The PCR steps (adapted from Ward et al., Reference Ward, Zemlak, Innes, Last and Hebert2005) included a first denaturation at 95°C for 2 min, 35 cycles at 94°C (30 s), 57°C (30 s) and 72°C (2 min), plus a final extension at 72°C for 7 min. The PCR products were purified with ExoSap IT enzymatic system (Affimetrix). The sequencing of COI fragments was carried out using the BigDyeTM Terminator v 3.1 Cycle Sequencing Ready Reaction kit (Applied Biosystems) using the M-13 initiator followed by reading in ABI 3130 automatic sequencer (Applied Biosystems).
The electropherograms were checked and edited in the software Geneious (Kearse et al., Reference Kearse, Moir, Wilson, Stones-Havas, Cheung, Sturrock, Buxton, Cooper, Markowitz, Duran, Thierer, Ashton, Meintjes and Drummond2012), followed by visual inspection of consensus sequences for final edition adjustments. In addition, the COI sequences were uploaded and deposited in the BOLD platform under the project ‘Assessing the genetic diversity of species Eucinostomus – EUCI’, being automatically assigned to a Barcode Index Number – BIN (group of sequences that should correspond to a single taxon), following the analytical procedures of Ratnasingham & Hebert (Reference Ratnasingham and Hebert2013). Afterwards, these sequences were aligned with those available in Bold Systems and GenBank by using the ClustalW method.
Phylogenetic and distance analyses
Phylogenetic reconstructions were carried out based on Maximum likelihood (ML) and Bayesian inference (BI). The best evolutionary model for both ML and BI trees was HKY + I as indicated in the software PartitionFinder (Lanfear et al., Reference Lanfear, Calcott, Ho and Guindon2012). Branch support in ML analysis was based on 1000 bootstrap replicates using RAxML (Stamatakis et al., Reference Stamatakis, Aberer, Goll, Smith, Berger and Izquierdo-Carrasco2012). In the case of BI, ultrametric trees following the Yule Speciation prior model were generated in the software BEAST 1.8.4 (Drummond & Rambaut, Reference Drummond and Rambaut2007), based on 20 million generations with sampling at every 2000 generations. The convergence of Markov chains was inspected in Tracer 1.6 (Drummond et al., Reference Drummond, Suchard, Xie and Rambaut2012). All values of Effective Sample Size (ESS) were above 200. Based on 10% of burn-in, the remaining trees were used to obtain a consensus tree and the branch support was based on the posterior probability values.
Based on BI and ML results, we calculated the genetic distance of each cluster with support values higher than 1 of probability or 95% of bootstrap to build a Neighbour-joining (NJ) tree in the software MEGA6 (Tamura et al., Reference Tamura, Stecher, Peterson, Filipski and Kumar2013) using the Kimura 2-parameter (K2P) evolutionary model as identified by the Barcode of Life initiative (www.boldsystems.org). All trees were visualized using FigTree v. 1.4.1 (Rambaut & Drummond, Reference Rambaut and Drummond2009).
Species delimitation methods
To establish a potential threshold among Eucinostomus species, we built a distance matrix based on the K2P model using the function sppDistMatrix available in the R package SPIDER v1.3–0 (Brown et al., Reference Brown, Collins, Boyer, Lefort, Malumbres-Olarte, Vink and Cruickshank2012). Based on this matrix, we created a density object with the minimum local function which disregards any previous knowledge about the species identity to indicate potential thresholds to infer intra- and interspecific variation levels (Brown et al., Reference Brown, Collins, Boyer, Lefort, Malumbres-Olarte, Vink and Cruickshank2012). We also used three widely used algorithms for species delimitation from molecular data: the Bayesian Poisson Tree Process (bPTP) (Zhang et al., Reference Zhang, Kapli, Pavlidis and Stamatakis2013), the generalized mixed Yule-coalescent model (GMYC) (Pons et al., Reference Pons, Barraclough, Gomez-Zurita, Cardoso, Duran, Hazell, Kamoun, Sumlin and Vogler2006) and the Automatic Barcode Gap Discovery (ABGD) (Puillandre et al., Reference Puillandre, Lambert, Brouillet and Achaz2012).
The bPTP method was carried out in the web server <http://species.h-its.org/ptp/> using as input a non-ultrametric tree based on ML inference. This analysis was performed with 500,000 generations with sampling at every 500 generations and 10% of burn-in. The GMYC was implemented in the web server <http://species.h-its.org/gmyc/>, based on an ultrametric tree obtained in BEAST (Drummond et al., Reference Drummond, Suchard, Xie and Rambaut2012).
The ABGD was carried out based on the pairwise genetic distances (based on p-distance, K2P and Jukes–Cantor models). The analysis was done using a gap width value of 1.0 for all distances using the software available in <http://wwwabi.snv.jussieu.fr/public/abgd/>. The congruence among the delimitation of MOTUs was evaluated by comparing the clusters inferred from each algorithm.
Population analysis
Haplotype networks were built in the software Popart (Leigh & Bryant, Reference Leigh and Bryant2015) using the Median-joining network algorithm to elucidate the genealogical relationships of E. argenteus and E. gula, because both species encompassed a high number of sequences from distinct localities throughout most of their range.
Results
The final alignment of COI sequences comprised 593 bp. Insertions, deletions or stop codons were absent, indicating that all sequences correspond to functional COI genes and not to putative pseudogenes or nuclear mitochondrial DNA segments (numts).
Phylogenetic analyses combined with distance methods and tools of molecular identification were helpful in identifying 34 potentially misidentified sequences in the analysed dataset of Eucinostomus, related to E. argenteus, E. harengulus, E. jonesii, E. dowii and E. gula (see supplementary material 2). The genetic variation in these samples ranged from 0.2–1.7%. Therefore, these sequences were reallocated to their respective groups and analysed according to their actual taxonomic classification to avoid biased evolutionary inferences. Sequences with divergence above 2%, which did not fit into any of the described taxonomic species of Eucinostomus were included with the acronym ‘cf.’. They are: E cf. argenteus, E. cf. melanopterus, E. cf. harengulus and E. cf. gula.
The BI and ML analyses recovered 15 MOTUs distributed into highly supported clades 1/>95, as follows: E. jonesii, E. dowii, E. cf. argenteus, E. harengulus, E. cf. harengulus, E. gracilis, E. gula, E. cf. gula, E. entomelas, E. argenteus, E. lefroyi, E. melanopterus, E. cf. melanopterus, E. currani and E. havana (Figure 1). The species delimitation algorithms (GMYC, bPTP, BINs and ABGD) also separated these clusters, besides revealing four additional MOTUS within E. gula, E. argenteus, E. melanopterus and E. harengulus (Figure 1). However, only the barcode index (BINs) was not found for E. cf. harengulus, since these sequences are only available in NCBI.
Fig. 1. Species delimitation using GMYC, bPTP, ABDG and BINs within Eucinostomus. The density related to the genetic distances among representatives of the genus Eucinostomus is shown below on the left side.
According to the distance method based on the K2P evolutionary model, the highest genetic distance was observed between E. havana and E. argenteus (21.8%), while the lowest value was detected between E. cf. melanopterus and E. melanopterus (2.9%). Two sequences erroneously assigned to E. argenteus were closely related to E. harengulus with 3.3% genetic divergence, being thus referred to as E. cf. harengulus. The highest values of intraspecific genetic variation were observed in E. gula (1.6%) and E. cf. gula (1.2%), while values close to 0% were observed in most of the Eucinostomus representatives (Table 1). The threshold potential from intra- to interspecific variation in mojarras as inferred by the minimal local function using Spider was established as 2.7% (Figure 1). This value was in agreement with the species delimitation algorithms where sequences with genetic distances above 2.7% were recovered as distinct MOTUs.
Table 1. Matrix of genetic distance based on K2P model among Eucinostomus species
HAR, E. harengulus; HAR1, E. cf. harengulus; JON, E. jonesii; ARG, E. argenteus; ARG1, E. cf. argenteus; GUL, E. gula; GUL1, E. cf. gula; CUR, E. currani; MEL, E. melanopterus; MEL1, E. cf. melanopterus; DOW, E. dowii; ENT, E. entomelas; GRA, E. gracilis.
Haplotype networks based on the most representative taxa with information about their geographic origin (E. argenteus and E. gula) revealed two haplogroups within each. The haplogroups composed of E. argenteus + E. cf. argenteus and E. gula + E. cf. gula were separated by 103 mutation steps (genetic distance = 20.7%) and 23 mutation steps (genetic distance = 4.4%), respectively (Figure 2). Particularly within E. argenteus/E. cf. argenteus, the range of each haplogroup corresponds to the Caribbean and Brazilian provinces.
Fig. 2. Map and haplotype network based on COI sequences in populations of E. gula, E. cf. gula, E. argenteus and E. cf. argenteus throughout their range in the Western Atlantic. The bars over lines indicate the number of mutations between the haplotypes.
Discussion
Besides identifying cryptic diversity, DNA barcoding analyses have been particularly effective in detecting synonyms and misidentifications among fish (Ferreira et al., Reference Ferreira, Kavalco, de Almeida-Toledo and Garcia2014; Hyde et al., Reference Hyde, Underkoffler and Sundberg2014). Accordingly, the pairwise divergence in COI sequences of Eucinostomus associated with comparative tools (BLASTn and Species Level Barcode Records) and phylogenetic inferences detected several cases of misidentification in the sequences available from Bold Systems and GenBank, as observed in E. argenteus, which has four BINs representing different species with the same morphological identification. It is possible to observe that two of these BINs include the species E. jonesii, E. dowii, E. harengulus and Eucinostomus sp. as well as a possible cryptic lineage for the coast of Brazil. Two clusters were also found for E. gula, E. melanopterus and E. harengulus species. For the latter sequences are only available on GenBank (see supplementary material and Figure 1). Furthermore, some sequences were identified only at genus level. In general, these sequences are derived from reports about larvae and juveniles of Caribbean fish without any previous information about COI sequences for comparative analysis, thus hindering their precise identification (Valdez-Moreno et al., Reference Valdez-Moreno, Vásquez-Yeomans, Elías-Gutiérrez, Ivanova and Hebert2010; Weigt et al., Reference Weigt, Baldwin, Driskell, Smith, Ormos and Reyier2012). Indeed, even the morphological identification of adult stages in Eucinostomus species is still under debate (Matheson & McEachran, Reference Matheson and McEachran1984; De La Cruz-Agüero & Galvan-Magana, Reference de la Cruz-Agüero and Galvan-Magana1993).
Another issue commonly observed in studies using DNA barcodes is the restricted range of sampling from putative widespread taxa (Ward et al., Reference Ward, Zemlak, Innes, Last and Hebert2005; de Ribeiro et al., Reference de Ribeiro, Caires, Mariguela, Pereira, Hanner and Oliveira2012; Brandão et al., Reference Brandão, de Bitencourt, Santos, Watanabe, Schneider, Sampaio and de Affonso2016). This issue restrains further biogeographic inferences, including the potential detection of phylopatric evolutionary lineages (Neves et al., Reference Neves, Lima, Mendes, Torres, Pereira and Mott2016). In this sense, the phylogenetic and species delimitation analyses carried out in the present study were particularly informative in assessing the actual diversity in Eucinostomus, since 15 distinct lineages (MOTUs) were clearly identified. Four MOTUs were related to cryptic forms in the formal taxa E. argenteus, E. gula, E. melanopterus and E. harengulus from the Western Atlantic (Caribbean and Brazilian provinces). Moreover, the tree topologies suggest distinct evolutionary histories for these lineages (Figure 1), as also supported by their high genetic diversity (20.7% for E. argenteus, 4.4% for E. gula, 3.3% for E. harengulus and 2.9% for E. melanopterus).
Similarly, molecular studies in distinct animal groups have shown the remarkable isolation of Caribbean lineages when compared with other regions (De Biasse et al., Reference De Biasse, Richards, Shivji and Hellberg2016; Fields et al., Reference Fields, Feldheim, Gelsleichter, Pfoertner and Chapman2016; Hurtado et al., Reference Hurtado, Mateos and Liu2017). Furthermore, recent reports have revealed that the biogeographic patterns in fishes from the Brazilian Province are quite heterogeneous, with growing evidence for high levels of endemism according to body size, dispersal routes and environmental features (Argolo et al., Reference Argolo, Ramos, Barreto, Bitencourt, Sampaio, Schneider and Affonso2018; Pinheiro et al., Reference Pinheiro, Rocha, Macieira, Carvalho-Filho, Anderson, Bender, Carlos, Ferreira, Figueiredo-Filho, Francini-Filho, Gasparini, Joyeux, Luiz, Mincarone, Moura, de Nunes, Quimbayo, Rosa, Sampaio, Sazima, Simon, Vila-Nova, Floeter and Treml2018). Therefore, some of these vicariant effects could be responsible for the distinction among the lineages within the complex E. argenteus. In the Caribbean province, the great biodiversity reflects ancient and recent patterns of circulation and topography from the Miocene, when the Americas were not yet connected. Climatic changes during this period conditioned the opening and closing of currents at different depths that modified the circulation pattern, and influenced global climate changes during the Pliocene (Williams & Duda, Reference Williams and Duda2008; Williams et al., Reference Williams, Smith, Herbert, Marshall, Warén, Kiel, Dyal, Linse, Vilvens and Kano2013; Thacker, Reference Thacker2017). In addition, change in the flow of the Amazon on the north coast of South America between 10–6 mya and closure of the Isthmus of Panama between 5.5–3.6 mya might have, although temporarily, affected other taxa, such as E. gula, E. melanopterus and E. harengulus. Therefore, major diversification events during the Miocene–Pliocene could have become more and more accentuated during Pleistocene sea level fluctuations, leading to population isolation and divergence of cryptic MOTUs in E. cf. argenteus, E. cf. melanopterus, E. cf. gula and E. cf. harengulus (Lambeck & Chappell, Reference Lambeck and Chappell2001; Müller et al., Reference Müller, Sdrolias, Gaina, Steinberger and Heine2008).
It should also be pointed out that the divergence between the cryptic lineages in E. argenteus was seven times higher than that established as the threshold from intra- to interspecific variation (2.7%). Similar cases of deep variation in COI sequences have also been reported in marine fish such as Scorpaena nonata (18%; Landi et al., Reference Landi, Dimech, Arculeo, Biondo, Martins, Carneiro, Carvalho, Lo Brutto and Costa2014), lantern fishes (17–25%; Pappalardo et al., Reference Pappalardo, Cuttitta, Sardella, Musco, Maggio, Patti, Mazzola and Ferrito2015), Gonostomatidae, Sternoptychidae and Myctophidae (16–23%; Kenchington et al., Reference Kenchington, Baillie, Kenchington and Bentzen2017). In fact, E. argenteus is a controversial taxon within Eucinostomus, being regarded as a putative species complex (Matheson & McEachran, Reference Matheson and McEachran1984). Moreover, the present data diagnosed that most of misidentified sequences in public datasets are related to E. argenteus, bringing more noise to their taxonomic status that remains to be resolved. Besides these aspects, the close relationship between the cryptic lineages of E. argenteus from the Atlantic in relation to E. entomelas and E. downi from the Pacific (~10–12% of divergence) is also intriguing. This pattern reinforces the role of the Panama isthmus in cladogenetic events of the ichthyofauna from the Eastern Pacific and Western Atlantic (Bacon et al., Reference Bacon, Silvestro, Jaramillo, Smith, Chakrabarty and Antonelli2015; O'Dea et al., Reference O'Dea, Lessios, Coates, Eytan, Restrepo-Moreno, Cione, Collins, de Queiroz, Farris, Norris, Stallard, Woodburne, Aguilera, Aubry, Berggren, Budd, Cozzuol, Coppard, Duque-Caro, Finnegan, Gasparini, Grossman, Johnson, Keigwin, Knowlton, Leigh, Leonard-Pingel, Marko, Pyenson, Rachello-Dolmen, Soibelzon, Soibelzon, Todd, Vermeij and Jackson2016; Thacker, Reference Thacker2017).
In Eucinostomus gula a different scenario was observed. Even though the sampled region of E. gula overlaps with the cryptic lineages of E. argenteus, the genetic difference between the lineages in the former is 50% higher than the threshold of 2.7% for interspecific distinction, suggesting a more recent evolutionary history for the cryptic lineages of E. gula. The close relationship between both lineages in this formal taxon indicates a recent expansion of their range in the Brazilian Province.
The lack of shared haplotypes in the complex E. gula provides strong evidence for phylogeographic disjunction caused by allopatry. Therefore, E. gula is likely to encompass two Evolutionary Significant Units (ESUs), one of them representing an undescribed taxon since a divergence of 4.4% was observed in COI sequences in relation to the other lineages. Putatively, a Brazilian endemic lineage evolved in allopatry as a result of major past vicariant events previously mentioned for E. argenteus. A similar pattern was also reported in Chromis multilineata (Rocha et al., Reference Rocha, Claudia, Ross and Brian2008), a reef fish species of similar distribution to E. gula.
The hypothesis of allopatric evolution of distinct lineages in E. gula could be weakened by the lack of geographic isolation of both lineages. However, the evidence combined with reciprocal monophyly, geographic coexistence and lack of a shared haplotype reinforces that both ESUs of E. gula in the tropical Atlantic evolved in allopatry followed by dispersal that determined their secondary contact. This evidence combined with reciprocal monophyly, geographic coexistence and lack of a shared haplotype reinforces that both ESUs of E. gula in the tropical Atlantic evolved in allopatry followed by dispersal that determined their secondary contact.
Unfortunately, the lack of georeferenced data in the other cryptic lineages of E. melanopterus and E. harengulus restrains further inferences about their diversification processes. Therefore, further phylogeographic studies focusing on both formal taxa are highly recommended.
Final remarks
Our phylogenetic analyses combining species molecular delimitation in a biogeographic context revealed new operational taxonomic molecular units (MOTUs) in Eucinostomus species from the Atlantic. Based on extensive barcode datasets we were able to recover the intraspecific diversity for this genus, estimating a potential threshold of 2.7%. Our data also corroborated the efficiency of COI markers in detecting cryptic lineages of mojarras that could be useful in resolving the taxonomic uncertainties in this fish group by providing a database to further support integrative approaches involving morphology, ecological features and other molecular markers. Subtle and overlapped morphological differences have served as proxies for species discrimination within Eucinostomus, however, in future systematic studies are strongly encouraged to consider genetic evidence as well.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0025315419001206.
Financial support
UPJ thanks Fundação de Amparo à Ciência e Tecnologia do Estado de Pernambuco (FACEPE) for research funding (BCT-0125-2.04/15), Conselho Nacional de Desenvolvimento Científico e Tecnológico CNPq (425080/2016-1) and Fundação de Amparo do Estado de Alagoas, FAPEAL (APQ – 1026/2016).
