Anthropic pressures, such as pollution, habitat degradation and loss, climatic changes, and, especially, harvesting, are the major agents of marine defaunation, reducing the effective population size (Pan et al., 2013; Pinsky, Palumbi, 2014; Allendorf et al., 2014; McCauley et al., 2015; Martínez-Candelas et al., 2020). Since small populations are more susceptible to genetic drift, compromising genetic diversity and evolutive potential, which can lead several species to extinction, understanding the genetic diversity distribution along the adaptive landscape is crucial for successful management plans (Allendorf et al., 2014; Cadrin et al., 2014; Cadrin, 2020).

Due to the supposed absence of clear gene flow barriers in marine environments, the management of marine species is generally based on panmixia (Cowen et al., 2006). In addition, a large number of marine species have high dispersion capacity, through both active, such as adult migrations, and passive pathways, during the larval phase, which is strongly associated with genetic homogeneity (Palumbi, 1992; Selkoe et al., 2014, 2016). However, ocean currents, convergence zones, and oceanic gyres, as well as differences in temperature, salinity, philopatry, and historical phenomena, can promote isolation, reducing the gene flow, and leading to diversification events (Grant, Bowen, 1998; Cowen et al., 2000; Machado-Schiaffino et al., 2010; Laurrabaquio-A et al., 2019; Lehnert et al., 2019; Chen et al., 2020; Faria et al., 2020; McKeown et al., 2020; Torrado et al., 2020; Zhao et al., 2020; Sadeghi et al., 2021). To better characterize different management units and identify barriers between species and populations, molecular tools are effective (e.g., Haney et al., 2010; da Silva et al., 2016; Healey et al., 2018; Azpelicueta et al., 2019; Jacobina et al., 2020; Klanten et al., 2020; Neves et al., 2020; Andrade et al., 2021; Labrador et al., 2021; Zarei et al., 2021), especially when using both mitochondrial and nuclear data, to more closely establish the species history, obtaining more robust results.

In this context, the genus Hippocampus (Syngnathidae), commonly known as seahorses, stands out. The relevance of the current study is related to observed decreases in several seahorse populations, especially linked to overfishing and habitat loss (Foster, Vincent, 2004). Currently, most seahorse species are listed in the International Union for Conservation of Nature (IUCN) red list, and the entire Hippocampus genus is listed in Appendix II of the Convention on the International Trade in Endangered Species of Wild Fauna and Flora (CITES). Until 2014, the occurrence of two species was confirmed along the Brazilian coast: H. reidi Ginsburg, 1933, and H. erectus Perry, 1810. However, using both molecular and morphological methods, Silveira et al. (2014) provided evidence of the presence of H. patagonicus Piacentino & Luzzatto, 2004. These three species are classified as Vulnerable (VU) in the Brazilian Red List of the Brazilian Ministry of Environment (ICMBio, 2018), and the planning for the species conservation was recently established at a meeting of experts promoted by IUCN in the Northeastern Brazil (RAT, 2024, participation).

Among the Hippocampus species, the long-snout seahorse H. reidi is the most abundant species, inhabiting coastal and shallow waters, in the Atlantic Ocean, from North Carolina (United States), Gulf of Mexico and Caribbean Sea, to Southeast Brazil (Lourie et al., 1999; Musick et al., 2000; Hercos, Giarrizzo, 2007; Silveira, 2011; Silveira et al., 2014). As it is the most common seahorse in Brazilian estuaries, H. reidi is a tourist attraction, which includes direct interactions with the humans (Ternes et al., 2016). Some life history strategies make the long-snout seahorse extremely vulnerable to these practices, as well as to harvesting and habitat loss, such as the formation of stable reproductive pairs, low mobility, patchy distribution, small home ranges, low reproductive rates, and a long period of parental care. For these reasons, this species is globally classified as Near Threatened (NT) by IUCN (Foster, Vincent, 2004; Oliveira, Pollom, 2017).

In the Brazilian Northeast, nuclear genetics of the Inter Simple Sequence Repeat (ISSR) showed a structure pattern between Maracaípe (Pernambuco) and Jericoacora (Ceará) (Montes et al., 2018). In addition, Carmo et al. (2022) showed a stable population structure between two bays of Rio de Janeiro state, and no differences between seasons. However, despite the species being charismatic, crucial information for conservation plans (e.g., population/genetic structure, genetic diversity, connectivity, and number of management units) remains poorly explored, especially in Brazil, which is considered the largest seahorse exporter for the international aquarium trade in Latin America as well as the largest consumer market (Baum, Vincent, 2005; Rosa, 2005),

Thus, the present study aimed to characterize the H. reidi genetic structure in the Southwest Atlanticthrough a multiloci approach, using both mitochondrial (Cytochrome b and control region) and nuclear (first intron of the ribosomal protein S7) data. More specifically, we asked how many H. reidi management units exist along the Brazilian coast and if there are any priority areas for conservation, using genetic diversity, population genetics, and demographic parameters. These informations can and will be directly used for the management plan of long-snout seahorses in Brazil.

Sample collection and molecular procedures. In total, 362 tissue samples of H. reidi were obtained from trade (dried specimens; Fig. S1) or collected from individuals found in situ, along the Brazilian Atlantic coast, in estuarine and reef environments, from Pará to Santa Catarina states (Fig. 1), between 2012 and 2015. For the in situ sampling, the non-destructive methods of collecting the dorsal fin (fin-clipping) or dermal filaments were used (Lourie, 2003; Planas et al., 2007). As it is a Near threatened (NT) species, vouchers were not collected. All individuals were photographed (Fig. 2) and returned to the same location. For trade seahorses, tissue was removed from the base of the tail for DNA analysis. All tissue samples were stored in 96% ethanol and kept at -20 ºC.All details about the samples, including geographical coordinates, can be found in supplementary material Tab. S2.

The total genomic DNA was extracted using the DNeasy Blood and Tissue (Qiagen®) kit, following the protocol suggested by the manufacturer. The extracted DNA was visualized in 1% electrophoresis gel and stained with GelredTM, and posteriorly quantified using a nano spectrophotometer Nanodrop 2000 (Thermo Scientific).

Three regions were amplified via PCR: the Cytochrome b gene (Cytb) and the control region (CR), representing the mitochondrial genome, and the first intron of the ribosomal protein S7 (S7), representing the nuclear genome. The Cytb fragment was amplified using shf2 and shr2 primers (Lourie et al., 2005), following the cycle presented by the authors. The CR region was amplified using the HCAL2 and HCAH2 primers (Teske et al., 2003) following the cycle presented by the authors. For each region, the PCR reactions were carried out in a total volume of 25 µL, using: 12.5 µl of 2X Taq Master Mix (Vivantis®) (1.25U of Polymerase Taq, 1X of buffer, 0.2 mM of dNTPs and 1.5 mM of MgCl2), 0.5 µl of MgCl2 (50 µM), 1.0 µl of each primer (10 µM), 2.5 µl of DNA (2-10 ng/µl), and 8 µl of ultrapure water. For the S7, the universal primers S7RPEX1F and S7RPEX2R were used (Chow, Hazama, 1998), following the amplification protocol described by Teske et al. (2004): 94 °C for 5 min, followed by 35 cycles of 30 s at 94 °C, 1 min at 60 °C, 1 min at 72 °C, and a final extension for 10 min at 72 ºC. The PCR reaction was carried out in a total volume of 25 µL using: 12.5 µl of 2X Taq Master Mix (Vivantis®) (11.25U of Polymerase Taq, 1X of buffer, 0.2 mM of dNTPs and 1.5 mM of MgCl2), 0.5 µl of MgCl2, 0.5 µl of each primer (10 µM/µL), 3 µl of DNA (2–10 ng/ µl), and 8 µl of ultrapure water.

The PCR products were purified using the ExoSAP-IT kit (QIAquick® PCR Purification Kit), following the protocol suggested by the manufacturer. The sequencing was carried out in a forward direction using the Bigdye Terminator v. 3.1 cycle Sequencing Ready Reaction kit (Applied Biosystems), in an automatic sequencer ABI 3500-Applied Biosystems. All sequences were deposited in the GenBank database under accession codes PQ134117 - PQ134478 (Cytb), PQ127407 - PQ127768 (CR), and PQ130575 - PQ131134 (S7).

Data analysis. All sequences were edited and aligned using the ClustalW algorithm (Thompson et al., 1994) implemented in BioEdit Sequence Alignment Editor v. 7.0. (Hall, 1999). The mitochondrial fragments (Cytb and CR) were concatenated as a single non-recombinant marker and, from now on, will be called mtDNA. Due to the presence of polymorphic sites in diploid regions, such as intron S7, the alleles were reconstructed with the PHASE v. 2.1 tool (Stephens et al., 2004) implemented in DnaSP v. 6.0 (Librado, Rozas, 2009), using default parameters and considering only allelic states with probabilities higher than 70% (Stephens et al., 2004). The analyses described below were performed for each marker (mtDNA and intron S7) separately. The genetic diversity indices [number of haplotypes (H) and polymorphic sites (S), private haplotypes (%Hp), haplotype (h) and nucleotide (π) diversity] was assessed by sample sites and cluster/genetic groups identified by other analysis, using the software Arlequin v. 3.5 (Excoffier, Lischer, 2010).

The population structure of H. reidi was tested through three approaches. First, the genealogical relationship between the haplotypes and their sample sites was investigated through a haplotype network using the TCS method in PopART (Clement et al., 2002; Leigh, Bryant, 2015). Second, the population structure of H. reidi was tested by the Bayesian Analysis of Population Structure – BAPS v. 6.0 (BAPS; Corander, Marttinen, 2006; Corander et al., 2008), firstly using a genetic mixture analysis with the sequences, followed by a population admixture analysis with a total of 10,000 interactions.

Lastly, to directly associate the geographic information with the genetic structure, the Geneland package (Guillot et al., 2005) on the R platform (http://www.R-project.org) was used. Due to limitations in the analysis regarding the sample size, only 300 individuals were included (Tab. S2). The number of groups (k) analyzed was 1–12 groups, with 9 independent runs, 1 million Markov chain interactions (MCMC), and a Thinning value = 1000.

The phylogenetic relationships of H. reidi were reconstructed through Bayesian Inference topologies in MrBayes v. 3.2.6 (Ronquist, Huelsenbeck, 2003). The nucleotide substitution model was estimated in jModelTest v. 2.1.7 (Darriba et al., 2012) under the Akaike Information Criterion [HKY+I (mtDNA) and HKY+G (intron S7)]. GenBank sequences of H. erectus [NC_022722.1 and KF557652.1 (mtDNA); KX646492.1 and KX646493.1 nuDNA)], H. trimaculatus Leach, 1814 [MF579378.1 and MF579379.1 (nuDNA)], Syngnathus typhle Linnaeus, 1758 [NC_030279.1 and KU925872.1 (mtDNA)], S. schlegeli Kaup, 1853 [AP012318.1 and KP861226.1 (mtDNA)] and S. temminckii Kaup, 1856 [AY277308.1 (nuDNA)] were used as external groups. Each database was analyzed with a burn-in of 10% and 10 million MCMC.

The genetic differentiation was assessed through the pairwise FST, among all sample sites and genetic clusters in Arlequin v. 3.5 (Excoffier, Lischer, 2010) using 1,000 permutations (p < 0.05). The Analysis of Molecular Variance (AMOVA) tested different hypotheses based on the population structure results for mitochondrial data: (a) Null hypotheses of one single group (Brazil); (b) Two groups, without the Bahia samples: (1) North/Northeast (NNE – PA, PI, CE, RN, PB, PE, AL) and (2) South/Southeast (SES – ES, RJ, SP, SC); (c) Two groups, with Bahia samples in the NNE group: (1) NNE + BA and (2) SES; (d) Two groups, with Bahia samples in the SES group: (1) NNE and (2) SES+BA; (e) Three groups: NNE (PA, PI, CE, RN, PB, PE, AL), (2) BA and (3) SES (ES, RJ, SP, SC).

Oscillations in population size were investigated through the Bayesian Skyline Plot analysis (BSP; Drummond et al., 2005) in Beast v. 2.6.4 (Bouckaert et al., 2014). Based on the population structure results, three groups were defined: (1) NNE, (2) BA, and (3) SES. For each mitochondrial marker, the nucleotide substitution model was estimated in jModelTest v. 2.1.7 (Darriba et al., 2012) under the Akaike Information Criterion, selecting the HKY model, except for control region data from NNE (JC model) and Cytb from BA (HKY+G model). In Beauti, both markers were linked. The mutational rates used for calibration were 1 x 10-8 per site per year (Mobley et al., 2010) and 5 x 10-8 per site per year (Bowen et al., 2006), for Cytb and CR, respectively. Three independent runs of 10 million (NNE and SSE groups) and 15 million (BA group) MCMC were performed with a burn-in of 25%. The log and tree files were combined using the LogCombiner tool implemented on Beast v. 2.4.6, and the parameters convergence (ESS>200; Effective Sample Size) was checked on Tracer v. 1.7.1 (Rambaut et al., 2014). In addition, the traditional neutrality tests [Fu’s FS (Fu, 1997) and Tajima’s D (Tajima, 1989)] in Arlequin v. 3.5. (Excoffier, Lischer, 2010), and the Mismatch Distribution analysis in DNAsp v. 5.1 (Rogers, Harpending, 1992; Librado, Rozas, 2009) were performed as complementary approaches.

Voucher specimens. As the long-snout seahorse Hippocampus reidi is globally classified as Near Threatened (NT), and Vulnerable (VU) in Brazil, vouchers were not collected. Thus, all individuals were photographed and, after collecting tissue samples, they were returned to the same location.

For mitochondrial data, concatenated fragments of 1,162 bp from 362 H. reidi individuals defined 69 haplotypes, of which 54 are private to some sample sites. The genetic diversity ranged from 0.308 (PB) to 0.952 (BA and RJ), for haplotype, and from 0.0007 (RN and PB) to 0.0059 (BA), for nucleotide diversity (Tab. 1).

The intron S7 data recovered the alleles from 341 H. reidi individuals, of which 120 was homozygote and 221 was heterozygote, with fragments of 520 bp, defining 11 haplotypes. The genetic diversity ranged from 0.576 (ES) to 0.879 (SP), for haplotype, and form 0.0032 (SC) to 0.0046 (PB), for nucleotide diversity (Tab. 1).

The mitochondrial haplotype network recovered two lineages separated by 6 mutational steps. In general, these groups correspond to the North/Northeast (NNE) and South/Southeast (SES) regions, and the Bahia state presented both lineages. Furthermore, one haplotype from Piauí state grouped with the SES samples (Fig. 3A). The individual results of each mitochondrial marker recovered a similar pattern and can be found in Fig. S3. The nuclear data from intron S7 did not recover any population structure pattern, with all haplotypes being shared among all sample sites, except for the Hap6, exclusive from Alagoas state (Fig. 3B).

The BAPS analysis identified 3 genetic profiles (k = 3; p = 1) for the mtDNA data. The green and blue/red profiles are almost exclusively of the sample sites from the NNE and SES groups, respectively. The BA group presented the 3 genetic profiles (Fig. 4A). The nuclear data recovered 8 genetic profiles (k = 8; p = 1), shared, in general, among all sample sites, in different frequencies, except by the blue/green, which is exclusive of the samples from the NNE group, and pink, which is exclusive of SES samples (Fig. 4B). These three genetic profiles are present in the BA group.

For the mtDNA, the Geneland data showed a higher probability of K = 4: Cluster 1 – Bahia, Cluster 2 –Espírito Santo, Cluster 3 –Rio de Janeiro, São Paulo, and Santa Catarina; and Cluster 4 –Pará, Piauí, Ceará, Rio Grande do Norte, Paraíba, Pernambuco, and Alagoas (Figs. 5A–E). The nuclear data recovered a similar result, with a higher probability of K = 3: Cluster 1 –Bahia and Espírito Santo, Cluster 2 –Rio de Janeiro, São Paulo, and Santa Catarina, Cluster 3 –Pará, Piauí, Ceará, Rio Grande do Norte, Paraíba, Pernambuco, and Alagoas (Figs. 5F–I).

The mitochondrial Bayesian topology sustains H. reidi from the Brazilian coast as a monophyletic group, also recovering the two lineages identified by the haplotype network. The monophyletic reciprocity between these groups presented a high branch support (posterior probability > 0.9; Fig. S4). The nuclear data failed to recover any clades (Fig. S5).

For mitochondrial data, the significant pairwise FST ranged from moderate [PE vs. PI (FST = 0.045)] to very high [ES vs. AL (FST = 0.862)]. The ES samples presented high and significant values in all comparisons, ranging from 0.237, when compared to BA samples, to 0.862 when compared to AL samples (Fig. 6A). In addition, the BA samples presented high and significant values in all comparisons, except when compared to SP samples (FST = 0.053; p > 0.05) (Fig. 6A). The individual results of each mitochondrial marker recovered a similar pattern and can be found in Tab. S6. The nuclear data revealed a similar scenario, with significant pairwise FST ranging from low [CE vs. AL (FST = 0.001; p > 0.05)] to high [BA vs. SC (FST = 0.198; p > 0.05)]. The ES and BA samples presented moderate-high values in all comparisons, except when compared to SP and CE, respectively (Fig. 6B).

Considering the null hypothesis, the AMOVA presented a high and significant FST value for mitochondrial (FST = 0.59; p < 0.05), and low for nuclear data (FST = 0.04; p < 0.05) (Tab. 2). In the different scenarios tested for mitochondrial data, based on the results of the population structure analyses, the highest genetic differentiation between groups (FCT) was found when Bahia (BA) samples were excluded, considering the North/Northeast (NNE) and South/Southeast (SSE) as distinct groups (FCT = 0.79; p < 0.05). Similar FCT values were found considering the following groups: NNE, BA, ES, SES (FCT = 0.69; p < 0.05), NNE, SES+BA (FCT = 0.67; p < 0.05), and NNE, BA+ES, SES (FCT = 0.67; p < 0.05) (Tab. 2).

The demographic analysis of BSP indicates expansion events at 5 and 20 thousand years ago for the NNE and BA groups, respectively (Fig. 7). The SES group seems to have suffered a recent contraction in population size (ca. 5 thousand years ago; Fig. 7).

The neutrality tests were negative and simultaneously significant only for mitochondrial data for total BR samples, RN and RJ (Tab. 1). The mismatch distribution analysis for both NNE and SES presented a unimodal curve, and for BA a bimodal curve (Fig. S7).

Population structure, genetic diversity and demographic parameters of H. reidi. The concatenated mitochondrial data reveal the presence of two H. reidi lineages in Brazilian coast. The first one represents the North and Northeast (NNE) samples and the other represents the South and Southeast (SES) samples. In addition, these lineages occur in sympatry in Bahia state, which seems to be a contact zone between them. Despite detecting some signals of structure (e.g., population differentiation in BA and ES population by pairwise FST and BAPS; three genetic clusters by Geneland), the nuclear data did not recover the two lineages identified by mitochondrial data.

These incongruences could be related to mutational rate differences between the molecular markers, since the mtDNA has higher mutational rates than nuDNA (Zink, Barrowclough, 2008; Calcagnotto, 2012; Toews, Brelsford, 2012). In addition, the subtle genetic differentiation in nuDNA could be influenced by male movements, acting as gene flow mediators, allowing a higher miscegenation between NNE and SES (Hellberg et al., 2002; Murray et al., 2017; Green et al., 2018; Day et al., 2019; Roycroft et al., 2019). However, more accurate data about the movement patterns of H. reidi are necessary in order to explain why males juveniles possibly disperse more, as suggested by nuDNA data observed herein.

The genetic structure in two lineages is reinforced by several mtDNA analyses (haplotype network, Bayesian topology, BAPS, and pairwise FST). These lineages appear to be distributed in three groups: (1) North/Northeast (except the BA samples), representing Lineage A, (b) South/Southeast, representing Lineage B, and (c) Bahia (BA), representing a mixture of both lineages, suggesting being a contact zone between them. This hypothesis is supported by AMOVA (FCT = 0.67)and Geneland results. However, it is important to highlight that the BA samples are genetically closer to the South/Southeast samples, presenting lowest values of pairwise FST. The absence of shared haplotypes between NNE and SSE may indicate a reduced gene flow.

This structure pattern can be partially explained by the Isolation by Distance (IBD) hypothesis, since a positive correlation between the genetic differentiation and geographic distance was found (unpublished data; Defavari, 2016). The IBD can be related to long-snout seahorse life history strategies, such as the closer relationship with the estuarine/mangrove environments (Lourie et al., 1999). However, two geographically closer populations presented a moderate and significant pairwise FST (PA vs. PI; FST = 0.134), while populations separated by 2,000 km presented a negative FST (PE vs. PA; FST = -0.023). In this way, the genetic similarity between populations from the same group can be explained by the high dispersion of H. reidi during the larval or juvenile phases as previously observed (Foster, Vincent, 2004; Lourie et al., 2004). The absence of clear physical gene flow barriers in the marine environment could also facilitates this phenomenon (Cowen et al., 2000, 2006), and the broad-scalehomogeneity occurs in several marine species with planktotrophic larvae (Hellberg et al., 2002). Thus, this pattern can be used to explain the absence of H. reidi genetic structure in several areas along the Brazilian coast.

The NNE and SSE represent different biogeographic sub-provinces of the Brazilian province (Pinheiro et al., 2018), and their sample sites represent different marine Ecoregions (Spalding et al., 2007). This geographic division seems to be deeply related with the split of the South Equatorial current in the North-Brazil and Brazil currents, species composition, and the São Francisco River mouth. It is important to highlight that, despite being the most common seahorse species along the Brazilian coast, presenting tolerance to soft changes in the salinity levels (Tseng et al., 2020), abrupt changes in salinity levels due to discharge of freshwater can affect the survival of the H. reidi individuals (da Hora et al., 2016) and can act as gene flow barrier.In addition, while the NNE presents warmer waters, the SES presents colder waters, and differences in temperature can reduce the gene flow, allowing local adaptation and isolation (Santos et al., 2003; Cunha et al., 2014). Thus, these temperatures gradients could be able to explain the genetic pattern found in H. reidi.

For both mitochondrial and nuclear data, the major pairwise differentiation was related to two sample sites: Espírito Santo (ES) and Bahia (BA). The ES presented only two shared haplotypes by mtDNA data [with BA (Hap32) and with BA and SP (Hap27)], revealing possible gene flow loss. Despite presenting significant FST values when compared with the SES samples, the ES samples is genetically closer to this group. These samples were collected below the Doce River mouth, and freshwater discharge as this one can limit the seahorses’ movements in NNE direction, as argued above. Additionally, in ES, the continental platform is narrow and contains a submersed mountain chain, the Vitória-Trindade (VTC), which could have acted as a glacial refuge during the Pleistocene (Pinheiro et al., 2015), and is being associated with genetic differentiation of other marine species (e.g., Santos et al., 2006; Pinheiro et al., 2015; dos Santos Freitas et al., 2017; Nauer et al., 2019; Souza et al., 2019). Thus, these features could favor the reduction of the gene flow between SES and NNE longsnout seahorses.

The BA samples presented high differentiation levels by pairwise FST, including when compared to NNE group. Of 28 haplotypes, only five are shared. However, it is genetically closer to SES group. Pinheiro et al. (2018) grouped the Bahia state into the same sub-province that SES populations, and Spalding et al. (2007) considered the BA state as a different marine Ecoregion, grouped with the ES into Eastern Brazil, which can explain the genetic similarity found between them. Despite that, BA presented both H. reidi lineages, reinforcing the idea of a contact zone mentioned above.

This sympatry can be explained by the demographic expansion that occurred after the Last Glacial Maximum (LGM; ca. 15 thousand years ago). The use of different refugia followed by gene flow during the glacial (sea level retraction) and interglacial (sea level expansion) cycles, respectively, has been associated with marine species diversification (e.g., Chen et al., 2020; Neves et al., 2020). These phenomena can be potentialized in areas with a narrow continental shelf (Dolby et al., 2016, 2018), such as Bahia, which presents the narrowest continental shelf on the Brazilian coast (ca. 14 km wide; Dominguez et al., 2012). Thus, after the LGM, the two H. reidi lineages may have had secondary contact in the Bahia coast, which is reinforced by the bimodal pattern of the mismatch distribution analysis, suggesting two episodes of population expansion.

The high degrees of genetic diversity were similar to those found for other Hippocampus species (Lourie et al., 2004, 2005; Goswami et al., 2009; Panithanarak et al., 2010; Saarman et al., 2010). Despite the highest sample size, the NNE group presented the lowest diversity level, and signals of population size contractions were identified by mismatch distribution analysis. However, the BSP did not recover any contraction events; on the contrary, indicated a recent populational expansion (ca. 5 thousand years ago). The SES group presented an opposite scenario. Despite the high genetic diversity, the BSP revealed a recent populational contraction (ca. 5 thousand years ago), after a long expansion period. These events could be related to anthropic actions, such as trade, tourism, and habitat degradation and loss.

Conservation implications. Solve taxonomic uncertainties, identify management units, and investigate the genetic diversity are crucial steps to management success. Here, we discuss some H. reidi conservation issues considering the presence of three management units, despite the presence of two lineages. This evidence justifies their separate management, setting different protection and sustainable actions. However, for sustainable use of the longsnout seahorse, we suggest that periods of non-harvesting be established during the months of October to February, which are the reproductive peaks of the species in Brazil.

Management Unit I consist of the NNE group, represented by North and Northeast populations, ranging from Pará to Alagoas states. This unit contains only the H. reidi Lineage A and is characterized by low genetic diversity (mtDNA), absence of genetic structure, shared haplotypes, and a low-moderate pairwise FST, suggesting a gene flow between the populations. This unit is considered the most vulnerable due to both low genetic diversity and possible contraction in population size by mismatch distribution analysis. Thus, we recommend that new conservation units should be created, or existing units should be amplified, to allow gene flow between the populations, avoiding the erosion of genetic variation. Furthermore, since H. reidi is traditionally used in the Brazilian Northeast as ornamental fishes, supervision efforts should be concentrated in trade and transport.

Management Unit II consists of the BA group, made up of specimens from the Bahia coast, representing a contact zone between Lineages A and B. This unit is characterized by the highest genetic diversity levels (mtDNA). In addition, the highest differentiation levels in all comparisons [except when compared to São Paulo (mtDNA and nuDNA) and Espírito Santo (nuDNA)] suggest a gene flow loss. The BA unit represents a priority area for conservation due to the high genetic levels and sympatry of the two lineages, representing a significant gene pool portion of H. reidi from the Brazilian coast. Thus, we suggest the integral protection of the longsnout seahorse along the Bahia coast, without allowing sustainable uses, and the supervision efforts should be concentrated in trade and transport.

Management Unit III consists of the SSE group, represented by South and Southeast, ranging from Espírito Santo to Santa Catarina states. This unit contains only the H. reidi Lineage B. Although the AMOVA hypothesis that considered the ES as a distinct group showed a high and significant differentiation between groups (FCT), the shared Lineage and exclusive genetic profiles with the others SES populations indicates a collaborative management.Contradicting the recent population contraction observed by the BSP analysis, the SES unit is characterized by high genetic diversity and the absence of an interpopulation structure, despite presenting some high pairwise FST values. Thus, this unit is in a reasonable conservation state, and the existing preservation efforts seem to be effective. Nevertheless, we reinforce the need to maintain the existing conservation units along the SES coast, especially due to the presence of another seahorse species, the H. patagonicus (unpublished data; Defavari, 2016).