Population connectivity and larval dispersal of the exploited mangrove crab Ucides cordatus along the Brazilian coast

Introduction

The mangrove crab Ucides cordatus is one of the most typical inhabitants of mangrove ecosystems in Central and South Americas (Melo, 1996). It is abundant along Brazilian coast, where crab fishing is one of the most important activities in several estuaries, and a source of employment and income for many traditional crab collectors (Diele, Koch & Saint-Paul, 2005). This decapod crustacean is also considered a key species for the ecological balance within benthic communities of mangrove forests, and have the largest biomass among mangrove invertebrates (Wolff, Koch & Isaac, 2000). The species has a strong impact on litter turnover, nutrients cycling, soil aeration and forest productivity (Pülmanns et al., 2014, 2016). Despite its importance to the ecosystem and abundance throughout the year, U. cordatus is currently on the Brazilian list of species threatened by overfishing (Pinheiro & Rodrigues, 2011). In addition to exploitation and habitat destruction, a fungal disease has affected crab populations in several locations (Boeger et al., 2005, 2007; Schmidt, Theil & Galli, 2008; Vicente et al., 2012). Together, these events could reduce genetic diversity and eliminate well-adapted stocks, which would cause the species to decline (Avise & Hamrick, 1996).

This crab belongs to the infra-order Brachyura, family Ucididae (Crustacea: Decapoda) and has been extensively studied in Brazil. Its high fecundity are likely to cause direct impact on the species ecology and conservation. Recorded values of fecundity per female ranged from 36.081 to 250.566 eggs (Pinheiro, Baveloni & Leme-Terceiro, 2003) while fertility ranged from 71.200 to 220.800 larvae (Hattori & Pinheiro, 2003). Also, the efficient larval dispersion is important to sustain high levels of genetic diversity. The process begin with crabs performing a dramatic synchronized mating movements that allow larval emission specifically on the strongest syzygy ebb tide of the month (around new moon or full moon), following the “syzygy tide inequality cycle,” which provides efficient larval transport to ocean waters (Schmidt, Bemvenuti & Diele, 2012). The euryhaline planktonic U. cordatus zoea I and II larvae are able to tolerate a wide range of salinities. However, after these early larval phases, which take about eight days, the final development stages do not tolerate low salinities. Therefore, export of newly hatched larvae is an important strategy to avoid stressful salinities (Diele & Simith, 2006; Simith & Diele, 2008). Studies have also shown that the development until megalopal stage lasts 20–69 days, depending on the temperature and salinity conditions (Rodrigues & Hebling, 1989; Abrunhosa, Melo & Abrunhosa, 2003; Diele & Simith, 2006), which is enough time for long range dispersion. The plankton-to-benthos transition begins with the reinvasion of the estuary by the last larval stage—the megalopa. Megalopae settlement and metamorphosis occurs mainly at conspecific burrows (Schmidt & Diele, 2009), triggered by odors emanated by juveniles and adults (Simith, Abrunhosa & Diele, 2013). Adults reach the legal minimum capture size after 6.13 years in males and 7.38 years in females (Diele & Koch, 2010).

Several studies have reported high genetic diversity for the U. cordatus using molecular markers as Random Amplified Polymorphic DNA (RAPD), Polymerase Chain Reaction-Restriction Fragment Length Polymorphism (PCR-RFLP) (Oliveira-Neto et al., 2007a), control region mtDNA sequences (Oliveira-Neto et al., 2007b; Pie et al., 2008), and Inter-Simple Sequence Repeats (ISSR) (Britto et al., 2011). These studies have also not provided any evidence of population structure among surveyed sites. However, in widespread species, microsatellite markers present better resolution to detect any trace of genetic differentiation among populations that live apart, and may be under unequal forces of natural selection. Usually these markers have exceptional degree of polymorphism and informativeness for evolutionary and ecological studies at large or even fine scale surveys (Abdul-Muneer, 2014). More recently, a research using these markers in U. cordatus (Oliveira-Neto et al., 2014) evidenced weak but significant genetic differences among populations (F_ST = 0.03, P = 0.0014). Even though, no quantitative information on the gene flow magnitude or its possible directivity were evaluated until now. Bayesian analysis with Markov chain Monte Carlo methods can be used to provide reliable gene flow estimates without the unrealistic assumption of symmetry of migration rates or equal population sizes (Faubert, Waples & Gaggliotti, 2007). A method proposed by Beerli & Felsenstein (2001) uses molecular data to determine the sample genealogy and provide migration estimates even in cases where the differentiation among populations is assumed insignificant. Besides, different models regarding migration direction can be tested to detect possible routs of larval dispersion along the ocean.

Due to the high ecological and social value of this species, a more extensive survey coverage also needs to be undertaken, particularly in estuaries under the possible influence of the North Brazil Current (NBC) (Fig. 1). Oliveira-Neto et al. (2014) concentrated most of their sample (77%) along a small portion of the Brazilian northeastern coast (over approx. 370 km), with only two collecting sites located on its lower and upper distribution limits. In this article we covered other sampling localities and used other set of microsatellites markers aiming to: (i) assess the spatial distribution of U. cordatus genetic diversity along the Brazilian coastline, including estuaries under the influence of the NBC and Brazil Current (BC), two major oceanic circulation systems flowing in opposite directions, away from each other; (ii) test different models of migration to estimate larval dispersal directivity; and (iii) estimate the migration rates among localities. A new perfect tetranucleotide microsatellite locus is also reported for the assessment of genetic variability among U. cordatus populations.

Materials and Methods

Results

Population structure

Pairwise F_ST and Jost’s D evidenced low genetic differentiation between localities (Mean F_ST = 0.013; s.d. = 0.004/Mean Jost’s D = 0.034; s.d. = 0.021) (Table 3). Mantel test performed with genetic distances and linear geographical coastline distances (Supplemental Information 1, “Coastline_Distance”) showed no significant correlation (Mantel_FST: R² = 0.04, P = 0.220; Mantel_{Jost′s D}: R² = 0.10, P = 0.140).

Table 3:

Triangular distance matrices (used for computation of Mantel test with 1,000 permutations) among U. cordatus sample localities.

	CAN	CAR	FOR	CAU	CUR	SOU
CAN	—	0.009	0.022	0.012	0.011	0.012
CAR	0.029	—	0.020	0.011	0.009	0.014
FOR	0.082	0.069	—	0.010	0.016	0.018
CAU	0.038	0.026	0.019	—	0.010	0.011
CUR	0.030	0.018	0.044	0.019	—	0.005
SOU	0.034	0.037	0.053	0.021	−0.004	—

DOI: 10.7717/peerj.4702/table-3

Note:

Above the diagonal are linearized pairwise F_ST. Below the diagonal are linearized pairwise Jost’s D. Values in bold represent significant differences (P < 0.05) between localities.

Regardless of these findings, pairwise comparisons of the genetic distances showed statistical significant differences among several localities. The southernmost sample (CAN) showed significant genetic differences from all other localities. As well, samples from FOR were statistically different from most other localities, with exception to its neighbor CAU. Neighboring localities CUR and SOU have also not shown significant differences (Table 3). CAR samples, which were under the influence of BC, showed no significant differences from CAU and CUR (samples under the NBC influence; see Fig. 1), however, significant differences were found when it was compared to other samples. Despite of the significant differences between the most distant localities, and the absence of differences between the closest ones, no pattern of differentiation could be observed in general.

The assumption of panmixia events was statistically discarded after running SASha, which has shown a non-panmictic condition (P = 0.011), albeit with a small difference between the observed mean distance distribution of alleles (2,055.0 km) when compared with the expected mean distance distribution (2,079.7 km).

AMOVA has shown that only 0.71% of the total genetic variance is explained by differences among sample sites, although the variance was significant (F_ST = 0.019; P < 0.001). Considering the variation between BC and NBC groups, no differences were found (F_CT = 0.007; P = 0.141), as well as for comparisons among BC vs NBC-1 vs NBC-2 (F_CT = 0.011; P = 0.069). However, Samova results suggested an alternative grouping arrangement with significant differentiation for samples organized in four groups, named hereafter as: BC-1 (“CAN”); BC-2 (“CAR”); NBC-1 (“FOR and CAU”); NBC-2 (“CUR and SOU”). Following this indication F_CT value remained low, but was significant (F_CT = 0.037; P = 0.023).

Bayesian clustering using Structure suggested lack of genetic differentiation among Brazilian samples with one single genetic group being detected (Posterior probabilities ≈ 100% for K = 1). However, another clustering solution suggested by DAPC analysis estimated the occurrence of six genetic groups (Fig. 2). The model was defined after retaining 35 principal components, which explained 85% of the total genetic variance. None of the six groups were completely isolated, with clusters 2, 4 and 6 being totally overlapped; their connectivity was also supported by minimum spanning tree (Fig. 2C), which presented three branches, all connected to Cluster 4. Clusters 1, 3 and 5 were the most divergent with more isolated samples (Fig. 2C). None of the clusters was exclusively represented by a single locality (Figs. 2B and 2C) and no pattern of genetic structure due to geographical position was observed.

Discussion

U. cordatus displays average genetic multilocus diversity with number of alleles, observed heterozygosity and PIC mean values of 12, 0.66 and 0.69, respectively. One caveat needs to be taken into consideration regarding the samples obtained: although the species is present along the Brazilian coast, we could only obtain samples from a few locations. That was main due to limited available survey budget. The present results fill in a gap in knowledge on this species.

According to Botstein et al. (1980), markers with PIC values higher than 0.5 are very informative and here, only locus UcSSR-01 presented lower value (PIC_{UcSSR − 01} = 0.453). Among all loci, only UcSSR-04 was not in Hardy–Weinberg equilibrium (HWE), which may be due to the occurrence of null alleles and therefore, heterozygous individuals would have been scored as homozygous causing HWE deviations (Callen et al., 1993). Nevertheless, overall genotyping data revealed high quality of these microsatellites, and all other markers have passed in quality control. Excluding locus UcSSR-04, differences between H_O and H_E estimates were small, averaging 0.66 and 0.67 respectively. Oliveira-Neto et al. (2014) reported similar values for U. cordatus samples from other estuaries with mean values of 0.74 and 0.73 for H_O and H_E, respectively.

Multilocus estimates for Weir and Cockerham’s θ, including or excluding locus UcSSR-04, differed slightly (0.015 vs 0.016). In both situations estimates showed low levels of genetic differentiation among localities and no evidence of population structuring. Low levels of inbreeding in the species was detected after F and f calculations. Linearized pairwise F_ST and Jost’s D have also evidenced weak genetic differentiation among localities, in agreement with θ estimates and with previous results (Oliveira-Neto et al., 2007a, 2007b; Britto et al., 2011). In addition, non-significant results were obtained from Mantel test, supporting the hypothesis of extensive gene flow promoting weak genetic variation among populations, a possible consequence of an efficient larval dispersal strategy of the species. Additionally, Structure also suggested the absence of preferential clusters along the Brazilian coastline, in agreement with previous finding (Oliveira-Neto et al., 2014).

Contrasting with Structure results, the approach based on DAPC analysis suggested the occurrence of six genetic groups, albeit without correlation with sample localities. Janes et al. (2017) advise for problems of many studies over- or underestimating population genetic structure when only Structure results are considered. Jombart & Collins (2015) pointed that the concept of “true k” is fairly hypothetical and does not mean that clustering algorithms should necessarily be discarded, but surely the reality is more complex than a few clear-cut, isolated populations. These considerations are important since population dynamics of U. cordatus seems to be more complex than it was supposed before DNA molecular data was available. Furthermore, despite of the indication of six groups by DAPC results, each one of them should not be considered an isolated genetic entity, as clear overlaps could be graphically observed among them.

High gene flow between estuaries due to species pelagic larval export strategy is a plausible explanation for our findings. In numerous crab species with similar dispersal strategy larvae are thought to be transported up to 100 km seaward from its spawning location (Epifanio, Masse & Garvine, 1989; Diele, 2000; Epifanio & Garvine, 2001). Our study and the Oliveira-Neto et al. (2014) found statistical differences when southernmost samples were compared with other from Brazilian Northeast and North estuaries. CAN location is situated in São Paulo state, which is located in Southeast Brazil, while southernmost samples from Oliveira-Neto et al. (2014) belongs to Paraná state, which is located in South region of Brazil. Despite of the low values of genetic differences found, southernmost samples were significant different from other localities, suggesting that irrespective of the absence of a clear population structuring pattern, the partitioning of the genetic diversity among Brazilian estuaries is not random distributed. Samova showed significant differences among groups (P = 0.023) when localities were arranged as BC-1, BC-2, NBC-1 and NBC-2, despite of the low F_CT (0.037). SASha results also pointed to a non-panmictic situation for U. cordatus crabs, albeit with a small difference between the observed mean distance distribution of alleles (2055.0 km) and the expected mean distance distribution (2079.7 km).

The gene flow models run in Migrate have also not suggested a scenario with completely free exchange of alleles among all sampling localities, and panmixia model was as well discarded. These two current branches run in opposite directions along the Brazilian coast (Fig. 1), and larger genetic differentiation between northern and southern populations might be expected if only the presumed effects of these currents on migration are taken into account. BC flows at about 0.5–0.6 m/s (Evans, Signorini & Miranda, 1983) and NBC at 0.6–1.0 m/s (Arnault et al., 1999), hence organisms living under the influence of these currents (i.e., larval U. cordatus specimens) could potentially be transported over several 100 km in only one month. However, there is no evidence to support the main Brazilian currents (BC and NBC) as having a strong barrier effect on larval dispersion.

The model that best described U. cordatus migration patterns was the island model. The Bayesian approach of Migrate allows the estimation of gene flow without the assumption of symmetric migration rates or equal population sizes (Beerli & Felsenstein, 2001), nevertheless, our results support symmetric dispersal in this species. We could not identify any location as a potential source or sink of individuals. Results show an average high number of individuals (≈24) per generation in a species with overlapping generations (Alcântara-Filho, 1978; Schmidt & Diele, 2009).

The dispersal of the U. cordatus larvae is yet poorly studied in the natural environment. Further studies on oceanographic transport processes and dispersion patterns should elucidate what happens with the planktotrophic larvae when exported through the estuarine plume. If larvae are transported only by ocean surface currents, one would expect a predominantly southward dispersal along the Brazilian Current and northward dispersal along the North Brazilian Current. Another possible explanation is that the larvae are transported by subsurface currents in deeper waters, which could flow in the opposite direction to the surface current (Schott, Fischer & Stramma, 1998). This hypothesis suggests an active migration of larvae under deeper water currents. In fact, U. cordatus larvae perform downward migration, but only during the later development stages, when the pelagic-benthic transition is successfully completed and the larvae dive toward the bottom sediment (Silva, 2007). Therefore, during the first zoeal stages, U. cordatus larval dispersal is still dependent on surface currents. On the other hand, local wind-driven currents might also explain larval connectivity (Roman & Boicourt, 1999). Although this hypothesis seems more plausible for micro- and meso-tidal regimes, in which estuarine plumes are more susceptible to local wind-driven currents, it is less likely for macro-tidal regimes in which estuarine plumes can reach long distances from the coast. The given possible explanations do not exclude each other, and they might occur simultaneously to sustain high levels of multidirectional larval transport that can explain the genetic connectivity observed in this study.

Is the dispersal rate estimated for U. cordatus sufficient to maintain only weak differentiation between locations? According to Wright (1931) statement, only one migrant per generation (the One-Migrant-per-Generation rule—OMPG) would be enough, under certain circumstances, to guarantee low level of genetic differences among populations. This rule can be applied to many conservation and management purposes, however, this concept does not imply panmixia among individuals of different subpopulations. OMPG should be interpreted as a qualitative evaluation of tradeoffs between losses of genetic diversity within populations versus homogenization among populations. The rule results from a simple model of population structure based upon a host of unrealistic assumptions, which could be better applied considering 1–10 migrants per generation (Mills & Allendorf, 1996) and effective number (Ne) rather than sample size (N) (Wang & Whitlock, 2003). In addition, little information is available on natural selection of U. cordatus and its impact on the genetic diversity of the species.

In the absence of extreme selective forces, the homogenizing effects of dispersal, even for small levels of effective migration, can be sufficient to overcome population differentiation (Thorpe, Solé-Cava & Watts, 2000). Nevertheless, despite of the high gene flow among localities, the significant genetic differences among them (according to Jost’s D, F_ST and F_CT) demonstrate that migration rates are actually not counterbalancing the loss of diversity due to genetic drift or selection events. Therefore, the reported levels of exploitation (anthropic selection) may promote a negative impact on populations (Duarte et al., 2014). Under these circumstances, even with high levels of larval dispersal, a more specific conservation policy is necessary to be provided by the Brazilian government in order to maintain healthy crab stocks.

Apart from whether exploitation of crab stocks in Brazil is a sustainable fishing activity, divergence of adult traits can occur due to differential pressure induced by strong fishing efforts. In several species, exploitation concentrated on larger individuals, has been demonstrated to result in constant reduction in age and length of maturity (Haugen & Vøllestad, 2001; Sinclair, Swain & Hanson, 2002). U. cordatus populations in Brazil, under uncontrolled fishing pressure, are already assuming a different size profile, in which the average crab size was reduced considerably (Jankowsky, Pires & Nordi, 2006). For example, a survey of the morphological pattern of crabs from Espírito Santo (Brazil) revealed that the mean carapace width of the population was 4.89 cm, below the minimum legal harvesting size (6.13 cm) (Conti & Nalesso, 2010). This type of variation could not be assessed with the present molecular markers. Moreover, because the commercial capture is allowed only for the collection of male individuals, the predominance of females over males at some sites (especially close to villages) brought out the advanced level of overfishing of this population and highlighted the urgent need of management actions (Conti & Nalesso, 2010; Goês et al., 2010).

Thus, even with crab populations presenting homogeneous genetic distribution along the Brazilian coast and with high migration rates, the significant differences found among some localities cannot be rejected. U. cordatus populations are not panmictic and despite of the weak population structure, it seems that gene flow does not completely bypass the effects of drift, natural selection (e.g., lethargic crab disease) or human exploitation, which is excessive and should be taken into account for designing species conservation programs. Statistical models suggest that migration events occur following island model, with symmetrical rates of migration among locations. Thus, the connectivity observed between crab populations evidenced the importance of a cooperative work over a wide geographic range aiming at the conservation and management of this fishery resource.

Supplemental Information