The roles of marginal lagoons in the maintenance of genetic diversity in the Brazilian migratory fishes Prochilodus argenteus and P . costatus

The rio São Francisco basin contains many endemic species, such as Prochilodus argenteus and P. costatus, which have great commercial importance. However, information about the main recruitment sites and genetic studies containing extensive sampling of these species are scarce. To investigate the roles of the marginal lagoons in the maintenance of genetic variability and in the population structure, we analyzed six microsatellite loci in nine sampling groups of P. argenteus and five sampling groups of P. costatus. Our results showed high levels of genetic variability and low values of genetic differentiation for P. argenteus (FST = 0.008, P < 0.05) and for P. costatus (FST = 0.031, P < 0.05). In addition, high values of gene flow combined with a small genetic distance suggest the presence of a single population for each species in the middle rio São Francisco basin. Moreover, putative migration routes involving marginal lagoons during the reproductive season could be detected, confirming the importance of these nurseries in the lifecycle of these species. Our results also indicate the necessity of adequate management of the fish resources and the conservation of the floodplains in the rio São Francisco basin.


Introduction
The rio São Francisco basin constitutes an area that covers 619,543 km², ca.7.5% of Brazil, with ecological domains ranging from Atlantic rainforest to Cerrado and Caatinga.In recent years, human impact, including the destruction of wetlands and marginal lagoons for agriculture and the construction of large dams, such as the Três Marias dam (TMD), for hydroelectric power generation, has affected the basin (Sato et al., 1987;Menezes, 1996).The central portion of the rio São Francisco basin, including Três Marias region, has greater investment in the fishery (Camargo & Petrere, 2001) and it is characterized by the presence of many floodplains and marginal lagoons.These marginal lagoons are important nursery habitats for migratory species recruitment because they provide an ideal habitat for the growth of juveniles with abundant food and relatively high temperatures (Moojen, 1940;Pompeu & Godinho, 2003).Unfortunately, important aspects of the fish migrations in the rio São Francisco, such as feeding and reproductive habitats, shoal composition and structure, homing and distances traveled between habitats, remain poorly studied (Sato & Godinho, 2003).
Prochilodontids occur abundantly in the major drainages of South America and are considered one of the most important components of the commercial and subsistence fisheries in Neotropical freshwater environments (Lowe-McConnell, 1975;Goulding, 1981;Vari, 1983;Castro & Vari, 2004).The family Prochilodontidae comprises three genera (Ichthyoelephas, Prochilodus, and Semaprochilodus) with highly restructured lips, teeth and jaws; these characteristics distinguish this family externally from all other characiforms (Castro & Vari, 2004).Prochilodus argenteus, an endemic species from the rio São Francisco basin, nowadays introduced in other Neotropical drainages (Castro & Vari, 2004), is the largest member of the Prochilodontidae family and is among the most important recreational and commercial fish species in the basin (Camargo & Petrere, 2001;Godinho et al., 2003).Another endemic species, P. costatus, has an importance in the subsistence fishery, as one of the most captured species in that region (Camargo & Petrere, 2001).
Genetic studies using variable markers such as microsatellites and mitochondrial DNA have been conducted to understand the population structure and dynamics of a considerable number of animal species (Zhang & Hewitt, 2003).Microsatellites, or simple sequence repeats (SSRs), are codominant nuclear markers with Mendelian inheritance.Microsatellites are abundantly distributed along genomes and demonstrate high levels of allelic polymorphism (DeWoody & Avise, 2000).The molecular structure and genetic variability of microsatellites are extensively exploited in evolutionary studies of a wide variety of fish species (see Chistiakov et al., 2006), including some from Neotropical freshwater (Hrbek et al., 2007;Abreu et al., 2009;Calcagnotto & DeSalle, 2009;Matsumoto & Hilsdorf, 2009;Pereira et al., 2009;Sanches et al., 2012).
Population genetic studies have been performed with both Prochilodus species from the rio São Francisco basin using Random Amplified Polymorphic DNA (RAPD) technique (Hatanaka & Galetti, 2003) and microsatellites (Hatanaka et al., 2006;Carvalho-Costa et al. 2008;Sanches et al., 2012;Barroca et al., 2012).However, there is no knowledge about the main nursery areas of the populations of either species or the importance of the marginal lagoons in the maintenance of the genetic variability of these or other migratory species.Considering the threatened marginal regions of the middle rio São Francisco basin and the ecological and economic importance of these species, the goals of this study were to access the main recruitment sites of both species and provide a robust dataset regarding the population genetic structure of these species.

Sampling sites
Migratory fishes reproduce in the mainstream or tributaries of the rio São Francisco basin, and their eggs and larvae are carried out downstream reaching marginal lagoons (Moojen, 1940) that stay unconnected of the mainstream in the dry season.Thus, samples collected in marginal lagoons were originated from spawning sites located upstream to the lagoons (e.g., samples collected in marginal lagoons from rio Paracatu was derived from eggs spawned only in upstream of the rio Paracatu).Thus, we consider samples from different lagoons of the same tributary as a single sampling group (Fig. 1, Table 1).We divided samples from the mainstream of rio São Francisco in the Três Marias region in three sampling groups to test previous ecological and genetic hypotheses that suggested a population differentiation among them (Godinho & Kynard, 2006;Hatanaka et al., 2006).rio São Francisco (SFR, n = 33); two sampling groups were collected downstream of the Três Marias dam, one during rainy season (TMDr, n = 22) and one during dry season (TMDd, n = 33); and one sampling group was collected downstream rio Abaeté during the rainy season (ABAr, n = 33) (Fig. 1, Table 1).In reference to P. costatus, a total of 156 individuals subdivided into five sampling groups were collected along the middle rio São Francisco basin.One sampling group was collected in the marginal lagoons of rio Paracatu (PAR, n = 33); one sampling group was from marginal lagoons of the rio São Francisco basin (SFR, n = 33); two sampling groups were collected downstream of the Três Marias dam, one group was obtained during the rainy season (TMDr, n = 26) and one was obtained during the dry season (TMDd, n = 33); and one sampling group was from downstream rio Abaeté during the rainy season (ABAr, n = 31) (Fig. 1, Table 1).All of the fish used for this study were collected in accordance with Brazilian laws under a permanent scientific collection license.Fin clips or muscles were fixed and preserved in 95% ethanol and deposited in the Laboratório de Biologia e Genética de Peixes, Universidade Estadual Paulista "Júlio de Mesquita Filho", câmpus de Botucatu, São Paulo, Brazil.Molecular analyses.Genomic DNA was extracted using a saline solution with proteinase K described by Aljanabi & Martinez (1997).Amplifications were carried out using six microsatellite loci developed by Barbosa et al. (2008) for Prochilodus argenteus: Par66, Par69, Par71, Par76, Par83 and Par85; the same loci were used for P. costatus, exception the locus Par80 was used instead of Par76 (monomorphic for this species).The reactions were performed in a total volume of 12.5 µl with 7.3 µl of ultrapure H 2 O, 1.25 µl of Taq DNA buffer (10X), 0.45 µl of MgCl 2 (50 mM), 0.4 µl of dNTP (2 mM), 0.5 µl of each primer (10 µM), 0.1 µl of Platinum Taq DNA Polymerase enzyme (Invitrogen; www.invitrogen.com)(5 U/µl) and 2.0 µl of genomic DNA (10-50 ng).Polymerase chain reactions (PCR) consisted of an initial denaturation step (5 min at 95ºC) followed by 35 cycles of chain denaturation (30 s at 95ºC), primer hybridization (30 s at 52-54ºC) and nucleotide extension (30 s at 72ºC).After these cycles, a final extension was performed at 72ºC for 10 minutes.Microsatellite amplified fragments were subjected to electrophoresis on 6% polyacrylamide gels for approximately 12 hours at 150 v.The gels were then stained with silver nitrate.The allele lengths were identified by reference to a 10 bp ladder (Invitrogen) using Kodak Digital Science 1D software.
The F ST (Wright, 1951) assuming the infinite allele model (IAM, Kimura & Crow, 1964), the R ST (Slatkin, 1995) assuming the stepwise mutation model (SMM, Kimura & Otha, 1978) and the analysis of molecular variance (AMOVA) (Excoffier et al., 1992) calculated for all loci and for all populations with 16,000 permutations (Guo & Thomson, 1992) were performed using the Arlequin 3.1 program (Excoffier et al., 2005).Initially, the samplings of Prochilodus argenteus were considered to be a single group for the AMOVA.Then, the samplings were subdivided into two groups: group 1 (adult fishes collected at mainstream of Três Marias region -ABAr, TMDd, and TMDr) and group 2 (juveniles collected inside unconnected marginal lagoons -SFR, VEL, JEQ, PAR, URU, and CAR).Similarly, the populations of P. costatus were first considered to be a single group and then subdivided into two groups: group 1 (adult fishes collected at mainstream of Três Marias region -ABAr, TMDd, and TMDr) and group 2 (juveniles collected inside unconnected marginal lagoons -LSF and PAR).This criterion was adopted to answer where are the main recruitment areas of the fishes found in the mainstream of Três Marias region (the most fishery activity of the São Francisco basin).In addition, rates of chord genetic distance (Cavalli-Sforza & Edwards, 1967) corrected for sample size were calculated with Geneclass2 (Piry et al., 2004), and these values were used to construct an unrooted neighbor-joining tree (Saitou & Nei, 1987) using the software Paup 4.0b10 (Swofford, 2003).

Prochilodus argenteus
Genotypes of 273 specimens of nine sampling groups were analyzed.All microsatellite loci were highly polymorphic.A total of 99 alleles were detected for all loci in all populations.The number of alleles per locus ranged from three (Par76) to 24 (Par85), with an overall average of 9.6 alleles (Table 2).Nineteen private alleles were found in low frequency (< 0.05) with the allele 239 (Par85 -URU), presenting a frequency of 0.052.Ho and He ranged from 0.379 (Par69) to 0.937 (Par71) and 0.473 (Par69) to 0.951 (Par85), respectively (Table 2).From the 54 pairwise F IS estimates, 41 showed positive values (heterozygote deficiency), and only 13 showed negative values (an excess of heterozygotes).Positive F IS values indicating heterozygote deficiency was predominant in all loci with a positive overall average (F IS = 0.123, P < 0.008) (Table 2).Only six out of 54 HWE tests performed for P. argenteus (6 microsatellite loci in 9 sampling groups) were significant.In all departure occurrences, the significant values of disequilibrium were associated with the presence of null alleles (Table 2).Null alleles were detected in 14 estimates and stutters were verified only in Par85 in the population TMDr.No evidence of linkage disequilibrium was detected in any loci (P > 0.05).
Low indices of F ST obtained by means of pairwise values among local populations of Prochilodus argenteus for all loci were significantly detected (F ST = 0.008, P < 0.05), with pairwise F ST for all loci ranging from -0.054 (ABA -JEQ) to 0.007 (ABAr -TMDd) (Table 3, above diagonal), clearly indicating an absence of population structure.Additionally, the overall R ST also showed significant lower values (R ST = 0.059, P < 0.05), ranging from -0.460 (ABAr -PAR) to 0.184 (TMDr -PAR) (Table 3, bellow diagonal).When we considered all specimens to be a single group (see Material and Methods), the AMOVA values (Table 4) revealed that only 0.82% of the total genetic variance was due to differences among populations (F ST = 0.008, P < 0.05; R ST = 0.069, P < 0.05) and 99.18% of the genetic variation within populations.When we considered the putative existence of two groups (see Material and Methods), the hierarchical AMOVA also revealed that 99.13% of the total variance was found within populations (values are shown in the Table 4).
The gene flow parameter Nm was calculated from the mean F ST value.The mean value obtained was Nm = 9.318,  5), and the dendrogram of genetic distance showed a low genetic differentiation among sampling groups analyzed (Fig. 2).

Prochilodus costatus
Five sampling groups with a total of 156 specimens were analyzed.All loci used in the analysis were highly polymorphic with a total of 112 alleles for all loci in all sampling groups.The number of alleles per locus ranged from four (Par69) to 29 (Par85), with an overall average of 12.9 alleles (Table 6).Nineteen private alleles were found at low frequency (d" 0.04).The observed heterozygosity ranged from 0.231 (Par69) to 0.958 (Par71), and the expected heterozygosity ranged from 0.580 (Par69) to 0.960 (Par85).The F IS indexes showed 24 comparisons with positive values (heterozygote deficiency) and six comparisons with negative values (excess of heterozygotes).The deficiency of heterozygotes was predominant in all loci with a positive overall average (F IS = 0.113, P < 0.008) (Table 6).Only seven out of 30 HWE tests performed for P. costatus (6 microsatellite loci in 5 sampling groups) were significant, and of these, six estimates were associated with the presence of null alleles (Table 6).Presence of stutters was verified in the Par71 loci on PAR.No linkage disequilibrium was detected (P > 0.05), and the allelic variation was treated independently.
Our results indicate a significant absence of the population structure based on F ST = 0.031, P < 0.05 and R ST = 0.044, P < 0.05.In addition, pairwise F ST values do not indicated structuring in all estimatives, ranging from -0.005 to 0.049.Median R ST values were detected between SFR and TMDd (R ST = 0.113, P < 0.05) (Table 7).Considering no predefined hierarchical models, the AMOVA showed that 96.82% of the genetic variance was found within   8).
The mean value of gene flow obtained for Prochilodus costatus was Nm = 5.911, and the values within each group (group 1: ABAr, TMDd, TMDr; and group 2: PAR, SFR) were Nm = 6.698 and Nm = 13.010,respectively, showing a higher number of migrants per generation.The chord distance ranged from 32.6% (TMDr -PAR) to 39.6% (SFR -ABAr) (Table 9).The SFR sample group has more discrepant indexes, which is shown in the dendrogram (Fig. 3).

Genetic diversity and lack of genetic structuring
In the present study, we did not find any genetic evidence of the existence of population structure in the species Prochilodus argenteus or P. costatus from the middle rio São Francisco basin.The population differentiation indexes F ST and R ST , the chord genetic distance associated to a high gene flow and previous studies (Carvalho-Costa et al., 2008;Sanches et al., 2012) corroborate the panmictic unit model for both species.The migration behavior, which is common in prochilodontids (Castro & Vari, 2004) likely plays an important role in the continuous gene flow in these species, keeping them as a single unit in the studied area.Variability levels in microsatellites with a mean number of 9.6 alleles per locus in Prochilodus argenteus and 12.9 alleles per locus in P. costatus were similar to those reported in freshwater fish with an average of 9.1 alleles per locus (DeWoody & Avise, 2000) and were similar to or higher than previous studies with both Prochilodus species (Hatanaka et al., 2006;Carvalho-Costa et al., 2008;Sanches et al., 2012).High genetic diversity was previously identified in P. lineatus (Sivasundar et al., 2001), which can be explained by the capacity of Prochilodus species to perform great migrations and by their large occurrence along the rivers.In the rio São Francisco basin, shoals of P. argenteus are capable of migrating up to 1,100 km upstream to spawn (Pinheiro, 1981), while in the rio Mogi-Guaçu (upper rio Paraná basin), P. lineatus migrates a maximum round trip distance of 1,300 km (Godoy, 1975;Toledo et al., 1986).
Significant deviations from the HWE were found in only six microsatellite loci of Prochilodus argenteus.These deviations are common in microsatellite regions (Alam & Islam, 2005;Carreras-Carbonell et al., 2006;Chevolot et al., 2006).Hatanaka et al. (2006) found populations with significant deviations from HWE in two of the four analyzed loci.The occurrence of null alleles and stuttering seems to be the most probable cause of these deviations because all of the deviations were associated with the occurrence of null alleles.Additionally, F IS values in P. costatus were positive in 80% of the estimates, also observed previously in the same species (Carvalho-Costa et al., 2008).However, we found low levels of population genetic differentiation among local populations of Prochilodus argenteus, as indicated by all population analyses, including F ST values that were lower (0.008, P < 0.05) than R ST values (0.059, P < 0.05).The R ST values under the SMM are expected to be greater than the F ST values under the IAM (Slatkin, 1995), although the inverse has been observed in some fish species (e.g., Pereira et al., 2009).Low indexes of F ST have been observed in other studies with characiforms, such as in Colossoma macropomum (Santos et al., 2007), Piaractus mesopotamicus (Calcagnotto & DeSalle, 2009;Iervolino et al., 2010), Prochilodus argenteus (Hatanaka et al., 2006;Sanches et al., 2012), Prochilodus costatus (Carvalho-Costa et al., 2008), Prochilodus lineatus (Revaldaves et al., 1997) and Salminus brasiliensis (Lopes et al., 2007).
The overall F ST found in P. argenteus (0.008, P < 0.05) is similar to that found for the same species (F ST = 0.008, P = 0.0002, Hatanaka et al., 2006).However, these authors suggested a model of structured population based on the occurrence of private alleles and high levels of heterozygosity found in one local population.Our results do not show any genetic structure (pairwise F ST ) among sampling groups from the Três Marias region, any differences in the heterozygosity, or the presence of exclusive alleles.Thus, we suggest the existence of only one panmictic unit of P. argenteus in the central portion of the rio São Francisco basin based on F ST , R ST and chord genetic distance.
Similarly, our results do not show the existence of a genetic structure in Prochilodus costatus.The F ST and R ST indexes found here were significantly low (F ST = 0.031, P < 0.05; R ST = 0.044, P < 0.05) with a high levels of gene flow (mean of Nm = 5.911), confirming the null hypothesis of the absence of a population structure.This hypothesis, evidenced by fixation rates is also supported by the homogeneous genetic distance (Table 9).These results corroborate those of Carvalho-Costa et al. (2008), who found low levels of pairwise structuring (F ST = -0.009and 0.006) in populations collected downstream of Três Marias dam, the rio Abaeté and the confluence among the two rivers at the reproductive season.

Migratory routes
Migratory fishes from the rio São Francisco basin spawn only in the reproductive season, which extends from November to February and coincides with the occurrence of rain and flooding, higher temperatures, and long photoperiods (Sato et al., 1996;Sato & Godinho, 2003).After total spawning in the mainstream river or in tributaries, eggs and larvae of the migratory fishes are carried downstream by water flow reaching marginal lagoons, which provide essential habitats for juveniles with an abundant food and relatively high temperatures (Moojen, 1940).
During the dry season, the marginal lagoons become isolated from the main channel.At the next rainy season, flooding provides water connections, and juveniles are ready to return to the main river (Sato & Godinho, 2003).
Migrations to spawn can explain the existence of a higher number of migrants per generation observed in our analysis (Prochilodus argenteus, Nm = 9.318; P. costatus, Nm = 5.911).Wright (1931) suggests that values greater than one migrant per generation prevent the substantial local differentiation by genetic drift.In fact, the existence of a single panmictic population for each species associated with the highly migratory behavior support the hypothesis of extensive gene flow among shoals.The convergence between rio São Francisco and rio Abaeté is considered the most important spawning site of Prochilodus argenteus (Godinho & Kynard, 2006), mainly because of the limnological conditions offered by the junction of the two rivers, such as water flow above 600 m³/ s, water temperature over 24ºC and concentrations of dissolved oxygen above 5 mg/l (Sato et al., 2005).As suggested by Hatanaka & Galetti (2003), one fraction completes its migration toward the dam during the reproductive period, and the majority possibly migrates to locations with environmental conditions more favorable for reproduction.Sato et al. (2005) found individuals with greater body size and weight and with better reproductive conditions downstream of the rio Abaeté compared with individuals that were collected upstream under the influence of the Três Marias dam.Observations made over the past 25 years as well as reports from local fishermen also relate the downstream confluence of the rio Abaeté and rio São Francisco as a preferential spawning site of other migratory fishes from this basin, such as Brycon orthotaenia, Conorhynchos conirostris, Pseudoplatystoma corruscans, and Salminus franciscanus (Sato et al., 2005).
Considering two hypothetic scenarios for Prochilodus argenteus, first as a single population and second as divided into two groups, the degree of variation detected by AMOVA among populations was minimal (F ST = 0.008, P < 0.05), and variation among the groups was not detected (F CT = 0.001, P > 0.05).These results show that the samples belonging to the Três Marias region (group 1) are not isolated from those recruited in the marginal lagoons (group 2).Based on these data, we can suggest that the populations of P. argenteus, which are recruited in the tributary lagoons, migrate preferentially to the mainstream of the rio São Francisco in the feeding season and then migrate to the Três Marias region or randomly return to the headwaters of tributaries during the reproductive season.In addition, the F ST , R ST , AMOVA and chord distance could not detect a population genetically distinct from the other populations that possessed similarity with individuals collected downstream rio Abaeté during the rainy season.Thus, this important area (downstream of the rio Abaeté) does not receive individuals of P. argenteus from one exclusive region but instead receives individuals from many regions along the middle rio São Francisco basin, that it must lead to a population homogenization.

Contributions of the marginal lagoons for migratory fishes
The floodplain area from the central rio São Francisco basin was estimated to be about 2,000 km² (Welcomme, 1990).The importance of the marginal lagoons in the recruitment of fishes was previously estimated by Sato et al. (1987), identifying 37 juveniles including migratory species such as Prochilodus argenteus (< 70 g) and P. costatus (< 80 g).Our results indicate a high gene flow with low genetic differentiation among samples from both marginal lagoons and mainstream rio São Francisco, suggesting a direct connectivity among these shoals.The continuous and rapid advancement of agriculture and the subsequent sediment deposition in some of the important tributaries from the central basin (Sato & Godinho, 2003) are factors that directly affect the conservation of the lagoons.Notwithstanding, Pompeu & Godinho (2006) observed a gradual reduction in fish richness and an abundance of lagoons that did not receive annual flooding because of the water flow control by the Três Marias hydroelectric dam.Local extinction with a reduction of almost 70% of the native fish fauna was also observed in a lagoon from the rio São Francisco basin (Pompeu & Alves, 2003).Considering these fish nurseries in the rio São Francisco basin (Sato & Godinho, 2003) and upper rio Paraná (Agostinho et al., 2000), the Brazilian conservation policy should include these marginal lagoons in the floodplains as a priority for the maintenance of the genetic variability in migratory fishes.

Fig. 1 .
Fig. 1.Map showing the central portion of the rio São Francisco basin and the distribution of the samples of Prochilodus argenteus (yellow) and of P. costatus (black).The circles represent marginal lagoons from tributaries, squares represent marginal lagoons from the rio São Francisco, and triangles represent places in the mainstream rio São Francisco in the Três Marias region.ABA = rio Abaeté; CAR = rio Carinhanha lagoons; JEQ = rio Jequitaí lagoons; PAR = rio Paracatu lagoons; SFR = rio São Francisco lagoons; URU = rio Urucuia lagoons; VEL = rio das Velhas lagoons; TMD = Três Marias Dam.

Table 1 .
Samples of Prochilodus argenteus and P. costatus from the rio São Francisco basin collected between July/2008 and July/2010.Samples from TMDr and ABAr were collected during the rainy season (November to February), and samples from TMDd were collected during the dry season (March to October).ABAr = rio Abaeté at rainy season; CAR = rio Carinhanha lagoons; JEQ = rio Jequitaí lagoons; PAR = rio Paracatu lagoons; SFR = rio São Francisco lagoons; URU = rio Urucuia lagoons; VEL = rio das Velhas lagoons; TMDd = Três Marias Dam at dry season; TMDr = Três Marias Dam at rainy season.

Table 4 .
Analysis of Molecular Variance (AMOVA) among sampling groups of Prochilodus argenteus.Structure tested: All samplings as a single group and divided in two groups: group 1 (Mainstream ABAr, TMDd and TMDr) and group 2 (Marginal lagoons of CAR, JEQ, PAR, SFR, URU and VEL).

Table 8 .
Analysis of Molecular Variance (AMOVA) among sampling groups of Prochilodus costatus.Structure tested: All samplings as a single group and divided in two groups: group 1 (Mainstream ABAr, TMDd and TMDr) and group 2 (Marginal lagoons of PAR and SFR).