Structure and genetic variability of golden mussel (Limnoperna fortunei) populations from Brazilian reservoirs

Abstract The golden mussel, Limnoperna fortunei a highly invasive species in Brazil, has generated productive, economical, and biological impacts. To evaluate genetic structure and variability of L. fortunei populations present in fish farms in the reservoirs of Canoas I (CANFF), Rosana (ROSFF), and Capivara (CAPFF) (Paranapanema River, Paraná, Brazil), eight microsatellite loci were amplified. Five of those eight loci resulted in 38 alleles. The observed heterozygosity (Ho) was lower than the expected heterozygosity (He) in all populations, with a deviation from the Hardy–Weinberg equilibrium (HWE). The average value for the inbreeding coefficient (Fis) was positive and significative for all populations. There was higher genetic variability within populations than among them. The fixation index (Fst) showed a small genetic variability among these populations. The occurrence of gene flow was identified in all populations, along with the lack of a recent bottleneck effect. The clustering analysis yielded K = 2, with genetic similarity between the three populations. The results demonstrate low genetic structure and suggest a founding population with greater genetic variability (ROSFF). Our data point to the possible dispersal of L. fortunei aided by anthropic factors in the upstream direction. It was concluded that the three populations presented a unique genetic pool for Paranapanema River, with occurrence of gene flow.

valves, and shows high filtration rates and facility to fecundate and form colonies, reaching a density of more than 150,000 individuals/m 2 (Cataldo, Boltovskoy, Hermosa, & Canzi, 2005). Giordani (2013)also highlighted that this organism has a gland that secretes protein filaments, known as byssus, which allows its fixation on practically all types of natural or artificial substrates, being nowadays a matter of great concern for all sectors that develop activities associated with the use of water.
In Brazil, the occurrence of golden mussel was first reported at the end of 1998 and beginning of 1999, in the State of Rio Grande do Sul, in the Jacuí River delta, and in the Guaíba Lake basin, respectively (Mansur, Richinitti, & Santos, 1999;Mansur et al., 2003). Recent studies have demonstrated a wide territorial distribution of this species, which encompasses several South American river basins, such as the basins of rivers Paraguay, Paraná, Uruguay, La Plata (Pessotto & Nogueira, 2018), and even in water bodies in the Northeast region of Brazil, such as the São Francisco River Basin (Barbosa et al., 2016). In the Paranapanema River, a tributary of the Paraná River, the first occurrence of the species was recorded in 2006, in the Canoas I reservoir (Garcia, Orsi, Casimiro, & Kurcheski, 2009). Its dispersion occurs in several ways, involving different stages of its life cycle, both larvae and adult (MMA-Ministério do Meio Ambiente, 2004).
The presence of the golden mussel in the Brazilian reservoirs has promoted significant environmental and economic impacts, which require frequent investment in maintenance and control. Damage to hydroelectric plants pipelines, pumps, turbines, boat hulls (Mansur et al., 2003), and net cages in fish farms (Oliveira, Ayroza, Castellani, Campos, & Mansur, 2014) are some of the main impacts caused by the spreading of this species. Considering all the detrimental effects caused by this species and its abundance in the invaded environments, it is crucial to find ways of controlling its populations that satisfactorily solve the incrustation problems without affecting the health of the local populations or causing environmental impacts.
Genetic studies using molecular biology techniques might provide additional information about the golden mussel, as information about the dispersal pattern, population genetic structure, as well as the possible influence of environmental factors on these characteristics, enabling the development of new technologies that contribute to the control and understanding of the invasion mechanisms of the species (Ghabooli et al., 2013;Oliveira et al. 2014;Zhan et al., 2012). According to the MMA (2004), one of the greatest obstacles for the implementation of measures to control the dispersion of golden mussel is lack of genetic information.
Genetic structuring analyses performed in previous studies by Zhan et al. (2012) and Ghabooli et al. (2013) reinforce the idea of "jump" dispersal dynamics of L. fortunei in South America.
According to the authors, human-mediated transport of propagules (e.g., abandonment of lines and hooks, ballast water discharge, recreational activities) are important factors that contributed to the dispersal of the mussel along the La Plata and Parana River basins. Another invasive mollusk, Zebra mussel (Dreissena polymorpha), has shown low genetic differentiation among populations of the Great Lakes, North America (Astanei, Gosling, Wilson, & Powell, 2005). The authors also pointed out that ballast water discharge contributed to the invasion of this species. In this study, we test the hypothesis that the dispersal of L. fortunei in the Paranapanema River may occur mainly via anthropogenic factors, which would provide gene flow even among populations isolated by dams. Thus, we evaluate the genetic structure and variability of golden mussel (L. fortunei) populations in three reservoirs of the Paranapanema River, Paraná, collaborating to the understanding of invasion patterns in the assessed regions.

| Collected biological material
The samples were collected in December 2014 from fish farms located in three reservoirs of the Paranapanema River, State of Paraná ( Figure 1): F I G U R E 1 Location of the Canoas I (1), Capivara (2), and Rosana (3) reservoirs, indicated by the points, and the CANFF collection sites (22°56′25.63″S; 50°24′49.86″ W), CAPFF (22°41′17.16″S; 51°17′51.30″W), and ROSFF (22°39′25.20″S; 52°46′52.78″W), indicated by the stars, along the Paranapanema River Samples were collected from fish farms that use the open-net tank production system to grow the Nile tilapia (Oreochromis niloticus). The mollusks were mechanically removed from the colonies, which remained fixated to the nets of the tanks. After removal, the mussels were placed in buckets and taken to nearby locations, where they were cleaned to remove any type of superficial dirt (slime, mud, etc.). The animals were subsequently kept for 2 hr in recipients containing ice (approximately 5°C) to induce numbness before they were killed. At this stage, they were cleaned again to remove dirt that remained between the mussels and were placed into labeled plastic bags. These bags were sealed and placed in thermos boxes filled with ice, before transporting them to the laboratory for analyses.

| DNA extraction and quantification
DNA was extracted using a protocol based on that described The mussel shell was opened, the animal was labeled, and the adductor muscle was removed with the help of tweezers by sectioning at the insertion region (basis). The sample was washed with absolute ethanol and then placed into a sterile microtube, where it was kept at room temperature for 10 min for the residual ethanol to evaporate. Next, the lysis solution (700 μl lysis buffer, 50 μl 20% SDS, and 15 μl 200 μg/ml proteinase K) was added to the samples, which were kept in a water bath at 50°C for 17 hr. The tubes were subsequently removed from the water bath and added with 700 μl of 5 M NaCl.
The contents were then mixed by inversion before centrifugation at 11,270 g at 4°C for 10 min.
After centrifugation, 800 μl of the supernatant was removed from each sample and placed into a new sterile microtube, before the addition of 700 μl cold absolute ethanol for DNA precipitation.
In order to increase the efficiency of the process, the microtubes were stored at −20°C for 2 hr. The samples were then centrifuged at 11,270 g at 4°C for 10 min. The ethanol (supernatant) was discarded, and the samples were dried at room temperature for 20 min.
The samples were added with 35 μl TE (Tris/EDTA) and 5 μl RNAse (30 μg/ml) and kept in a water bath at 37°C for 40 min before storage at −20°C. DNA was quantified using a PICODROP ® spectrophotometer (Picodrop Limited, Hinxton, UK). The samples were diluted to a final concentration of 30 ng/μl. In order to assess DNA quality, an electrophoresis run in 1% agarose gel was conducted at 70 V for 1 hr.

| DNA amplification and capillary electrophoresis
The amplification was performed at the Laboratory of Molecular

Markers and Plant Cytogenetics at the Department of Biology, State
University of Londrina (UEL). Eight primer pairs (Lf04, Lf06, Lf07, Lf19, Lf21, Lf22, Lf23, Lf38) previously developed for L. fortunei (Zhan et al., 2012) were tested in the analysis of the microsatellite regions (unique primers developed for this species until the present time). An indirect labeling with fluorophores was used for genotyping, using a system based on the addition of three primers to the PCR reaction, according to Schuelke (2000). In this method, a tail of the M13 sequence (TGTAAAACGACGGCCAGT) is added to the 5′ end of the primer. The amplification reactions were prepared with 4.5 μl GoTaq Green Master Mix (2X reaction buffer, pH 8.5, 1600 μM dNTP, and 3 mM MgCl 2 , Promega, Winchester-USA), 0.08 μl 5 p.m.
The PCR reactions were performed in a PTC200 thermocycler (MJ Research, Massachusetts-USA) using a gradient program.
All eight primer pairs were tested, but only five of them were used (Lf06, Lf07, Lf21, Lf22, Lf23) as they presented the expected results: they were polymorphic and amplified the microsatellite alleles consistently and reproducibly. The respective PCR reactions using these five primer pairs were performed using the following program: 94°C

| Statistical analyses
The results were displayed using data matrices and analyzed with the software FSTAT (Goudet 2005), to find the number of alleles per locus (Na), the allele richness (Ra), and the inbreeding coefficient (Fis). The number of effective alleles (Ae), observed (Ho) and expected heterozygosity (He), deviations from the Hardy-Weinberg equilibrium (HWE; p < 0.05; calculated by chi-square test for Hardy-Weinberg equilibrium) for each locus, and genetic distance (Nei, 1978) were calculated using the software POPGENE (Yeh, Boyle, & Xiyan, 1999 (Cornuet & Luikart, 1996)  .25 indicate small, moderate, high, and elevated genetic differentiation, respectively. BAYESASS (Wilson & Rannala, 2003) was employed to estimate gene flow using the Bayesian method. To identify the number (K) of genetically similar population clusters, we used the Structure v. 2.3.3 software (Hubisz, Falush, Stephens, & Pritchard, 2009) with a no-admixture, length of burn-in period of 50,000 and 500,000 repetitions of MCMC (Markov chain Monte Carlo), and 20 replicates per K, with K ranging from 1 to 6. The number of clusters was determined using the website Structure Harvester (Earl, 2012).

| RE SULTS
Three of the eight loci tested (Lf04, Lf19, and Lf38) did not amplify or presented unspecific products, and were therefore removed from the analyses. The five loci used (Lf06, Lf07, Lf21, Lf22, and Lf23) were polymorphic and amplified consistent and reproducible microsatellite alleles, with the expected sizes between 57 bp (Lf23) and 291 bp (Lf21). In total, 38 alleles were detected for 75 individuals from the three natural populations of L. fortunei. The locus that presented the largest number of alleles (Na) was Lf06 (12 alleles), followed by Lf07 and Lf23 (7 alleles), and Lf21 and Lf22 (6 alleles). A low frequency of null alleles (5 alleles = 13.1%) was identified in the L. fortunei populations. Two loci showed null alleles in CANFF (Lf06 and Lf07), two in ROSFF (Lf06 and Lf22), and one (Lf06) in CAPFF.
Considering the results presented to verify the estimates of genetic diversity parameters of the five loci (Table 1) Using the analysis of molecular variance (AMOVA), a greater genetic variability was identified within than between the studied populations (   Figure S1).

| D ISCUSS I ON
The five investigated loci were highly polymorphic and amplified  fragmentation caused by natural and artificial barriers. However, the low genetic structure verified by Structure and the low Fst values practically rule out the influence of the Wahlund effect on the heterozygote deficit. Therefore, the effect of an inbreeding process appears to be the most suitable explanation for the observed heterozygote deficit. Mating between related individuals has already been observed in bivalves (Li & Hedgecock, 1998), furthermore, two-thirds of the L. fortunei population are composed by females, that is, the lower proportion of males increases the chance of recurrent inbreeding (Ricciardi, 1998;Zhan et al., 2012). In bivalves such as Placopecten magellanicus, Kenchington et al. (2006)  there is approximately 13 km. In this river, in a region close to the CANFF point, the mussel was first observed in 2006 (Garcia et al., 2009). Based on this information and taking into account that larvae F I G U R E 2 (a) Graphic representation of the first two factorial correspondence analysis (FCA), blue squares-ROSFF, white squares-CAPFF, and yellow squares-CANFF; (b) Clustering analysis of (Limnoperna fortunei) from Brazilian reservoirs and adults of L. fortunei present limited swimming abilities against currents, with its dispersion being carried out mainly by passive diffusion (Ricciardi, 1998), it can be concluded that the migration occurred upstream via anthropogenic factors. This hypothesis, which is supported by the dispersal dynamics by "jump", defended by Zhan et al. (2012) and Ghabooli et al. (2013) in the South American basins, is corroborated by structuring analyzes and gene flow. Based on the above, we conclude that the migration of L. fortunei through the Paranapanema River was mediated by one or more vectors, which allowed them to overcome the natural (current) and artificial (hydroelectric dam) barriers.
According to Holland (2001), when a population with a low genetic variability invades a heterogeneous and unknown habitat, Among the proposals that can support the control of the golden mussel, there is the importance of conducting new genetic research, collecting samples beyond the points already analyzed, still in the Paranapanema River, to clarify mechanisms of gene flow, genetic structure, and differentiation processes involved in the colonization of new habitats, besides the development of new specific molecular markers for L. fortunei from this region, considering the absence of amplification detected in some microsatellite primers. It is also considered important to carry out researches to identify the main vectors causing the dissemination of the mussel in the Paranapanema River, which will enable the implementation of inspection measures, readjustment, or even elimination of these vectors, as well as mapping of risk areas to avoid new colonization.
In conclusion, it was observed high variability with low genetic structure and occurrence of gene flow in both directions (upstream and downstream). The three populations presented a unique genetic pool for the entire stretch of the sampled river.

ACK N OWLED G M ENTS
The authors thank the fish farmers' availability of physical infrastructure in this work and Capes for the financial support.

CO N FLI C T O F I NTE R E S T
The authors reported no potential conflict of interest.

AUTH O R CO NTR I B UTI O N S
The study was coordinated by NMLB; samples collection was per-