Genetic structure and diversity of Nodularia douglasiae (Bivalvia: Unionida) from the middle and lower Yangtze River drainage

The Yangtze River drainage in China is among the most species rich rivers for freshwater mussels (order Unionida) on Earth with at least 68 species known. The freshwater mussels of the Yangtze River face a variety of threats with indications that species are declining in abundance and area of occupancy. This study represents the first analyses of the genetic structure and diversity for the common and widespread freshwater mussel Nodularia douglasiae based on microsatellite DNA genotypes and mitochondrial DNA sequences. Phylogenetic analysis a fragment of the COI mitochondrial gene indicated that N. douglasiae collected from across the middle and lower Yangtze River drainage are monophyletic with N. douglasiae from Japan, Russia, and South Korea. The results of the analysis of both the mtDNA and microsatellite datasets indicated that the seven collection locations of N. douglasiae in the middle and lower Yangtze River drainage showed high genetic diversity, significant genetic differentiation and genetic structure, and stable population dynamics over time. Moreover, we found that the connections among tributaries rivers and lakes in the Yangtze River drainage were important in maintaining gene flow among locations that N. douglasiae inhabits. An understanding of the genetic structure and diversity of a widespread species like N. douglasiae could be used as a surrogate to better understand the populations of other freshwater mussel species that are more rare in the Yangtze River drainage. At the same time, these results could provide a basis for the protection of genetic diversity and management of unionid mussels diversity and other aquatic organisms in the system.


Introduction
Freshwater mussels (Bivalvia: Unionida) are one of the most important faunas in freshwater ecosystems, for their potential to enhance biodiversity and ecosystem functioning (e.g., PLOS  (e.g., the Yangtze River drainage) has numerous natural and anthropogenic features that may have shaped the genetic structure and diversity of this species [25]. Analyses of mitochondrial DNA sequence data provide estimates of the phylogenetic relationships and population evolution in unionid mussels [26][27][28]. Microsatellites or simple sequence repeats (SSR) are useful markers for the study of deeper genetic diversity patterns in freshwater mussels because of their co-dominance, high mutation rate, and ease of scoring [24,[29][30][31]. Combining analyses of mtDNA sequence data and microsatellite genotypes can help to reveal both the course-scale and fine-scale evolutionary history and genetic structure of a species. This study is the first to study the genetic structure and diversity of N. douglasiae in the middle and lower reaches of the Yangtze River, and among the first broad-scale studies for any freshwater mussel in the region. Given the historically interconnected nature of the Yangtze River and its tributary rivers and lakes, we hypothesize that the analyses will reveal high levels of genetic diversity and relatively little genetic structure with high levels of gene flow across sampled locations. This study will provide a basis for the protection and management of diversity in unionid mussels in this large river watershed.

Ethics statement
All necessary permits were obtained for the described field studies from the Yangtze River Fishery Administration of China. The handling of mussels was conducted in accordance with the guidelines on the care and use of animals for scientific purposes set by the Institutional Animal Care and Use Committee (IACUC) of Nanchang University, Jiangxi, China.

Sample collection and DNA extraction
Specimens of N. douglasiae (n = 197) were collected in 2014 and 2016 from Poyang Lake (PY), Donting Lake (DT), Xiannv Lake (XN), Gan River (GJ), Liangzi Lake (LZ), Hongze Lake (HZ), Taihu Lake (TH) in the middle and lower reaches of Yangtze River, China (Table 1 and Fig 1). Tissues of individual specimens were preserved in 95% ethanol and stored at -20˚C until DNA extraction. Specimens were deposited in the Nanchang University Museum and assigned acces-  individuals of N. douglasiae. The PCR reaction was carried out in a 20 μL volume containing 10 μL 2x Taq PCR MasterMix (TianGen); 6.4 μL water; 0.6 μL of 10 μM HEX-, TAMRA-, or 6-FAM-labeled M13 universal primer; 1.0 μL of 10 μM forward primer with an M13 tag on the 5 0 end; 1.0 μL of 10 μM reverse primer; and 2 μL genomic DNA (about 50 ng/μL). PCR amplifications were conducted with the following touchdown thermal cycling program: an initial denaturation at 94˚C for 5 min, followed by 10-15 cycles of 94˚C for 30s, locus-specific annealing temperature ranging between 52˚C and 65˚C for 45 s, 72˚C for 45 s, at the same time, 20 cycles of 94˚C for 30 s, locus-specific annealing temperature 53˚C for 45 s, 72˚C for 45 s, and a final extension at 72˚C for 10 min. All loci were run separately on an ABI3730 automated sequencer and alleles were scored using a TAMRA-labeled size standard [33] using GENE-MAPPER v. 3.7 (Applied Biosystems).

Data analyses
Mitochondrial DNA. The sequences of the mtDNA COI fragment in 64 samples were aligned using Clustal X1.81 [34]. DNASP 5.0 [35] was used to analyse nucleotide composition, number of polymorphic sites (S), average numbers of pairwise nucleotide differences, haplotype diversity (Hd) and nucleotide diversity (π) for each collection location.
To test the monophyly of the N. douglasiae COI haplotypes from the middle and lower Yangtze drainage, a phylogenetic analysis using Bayesian inference was performed using MRBAYES v.3.2.2 [36]. The initial model of evolution (HKY+G) was determined by comparing 24 models of evolution in MRMODELTEST v.2.2 [37]. MRBAYES was run using 3,000,000 generations and six concurrent Markov Chains and 2 hot chains sampled at intervals of every 100 generations for a total of 30,000 trees. A 25% burn-in (7500 trees) was used to ensure stationarity of the log likelihood values. In addition to the 37 COI haplotypes from the middle and lower Yangtze, COI sequences from other putative N. douglasiae from Russia, Japan, and South Korea [18] were included (S4 Table). Also included in the phylogenetic analyses were COI sequences for N. nuxpersicae [11], N. nipponensis [18], and N. sinuolata [18] (S4 Table). As outgroups for the Nodularia dataset, COI sequences available in GenBank from 10 species were used (S4 Table).
To visualize the relationships among the N. douglasiae COI haplotypes from the middle and lower Yangtze, a haplotype network was constructed using a TCS algorithm [38] in POPART [39] with a 95% connection limit and gaps defined as missing data.
Patterns of genetic structure in the COI dataset were evaluated using a hierarchical analysis of molecular variance (AMOVA). The AMOVA was used to partition variance components to populations and to individuals within each collection location, where 1000 permutations were performed to test the significance of each pairwise population comparison. A test for isolationby-distance was conducted by testing the significance of a correlation between pairwise Nei's D (calculated using ARLEQUIN [40]) and geographic distance among sampling locations. The correlation computations between pairwise genetic and geographic distances between populations were analysed using a Mantel test [41]. Geographic distance was meansured among collection sites by measuring distances following waterways in ArcMap GIS (ESRI).
Tajima's D and Fu's Fs tests were conducted through Arlequin 3.5 [40], to examine deviations from neutrality. DNASP 5.0 [35] was used to analyse mismatch distribution analysis (MDA). A Bayesian Skyline Plot (BSP) [42] analysis was computed in BEAST 1.4.7 [43]. The BSP was used to reconstruct the effective population size fluctuations since the time of the most recent common ancestor (TMRCA). MCMC was run for 500 million steps, with sampling every 1000 generations and following a 'burn-in' of the initial 10% cycles. Inspections of the results and construction of the BSP were conducted using TRACER 1.5 [44]. The fit of the constant size population model and Bayesian Skyline coalescent models to dataset was assessed using the Model Comparison function in TRACER 1.5.
Microsatellite DNA. The number of alleles (N A ), the effective number of alleles (N E ), observed heterozygosity (H O ) and expected heterozygosity (H E ), and tests for deviation from Hardy-Weinberg Equilibrium (HWE) were calculated using POPGENE, v1.32 [45]. CERVUS 3.03 [46] was used to calculated polymorphism information content (PIC). Using MICRO-CHECKER v. 2.2.3 [47] to detect possible null alleles from each collection location.
In order to detect any recent genetic bottlenecks (within 2N e − 4N e generations), four tests with varying degrees of sensitivity were conducted using BOTTLENECK v. 1.2.02 [48]. Wilcoxon sign rank tests were carried out using three models of evolution: the infinite alleles model (IAM), two-phase model (TPM), and stepwise mutation model (SMM). A mode-shift test was conducted to identify significant changes in allelic frequency caused by a genetic bottleneck.
Using STRUCTURE v. 2.3.3 [49] population structure was assessed in the study area. Ten iterations, allowing for admixture among genetic groups (K) and assuming correlated allele frequencies, were run for each value of K (number of clusters) which was defined by the number of collection locations for each species: the maximum value of K was calculated by adding 3 to the number of collection locations (i.e., K = 1-10) to allow detection of substructure within sampling locations. Each trial used an initial burn-in period of 200,000 replicates, followed by an additional 200,000 replicates after burn-in to ensure stationarity. To determine optimal solutions for potential numbers of genetic groups (K) within each species [50], calculating ΔK from STRUCTURE output in combination with the log likelihood of the solution for each value of K using STRUCTURE HARVESTER v. 0.6.8 [51]. To further evaluate and visualize the geographic genetic structure among collection locations, a principal coordinates analysis (PCoA) was conducted using GenAlEx 6.5 [52] to ordinate genetic distance estimates [53] calculated for the genotypic data of individuals used.
An analysis of molecular variance (AMOVA) [54] was run using GenAlEx to test the statistical significance of genetic divergences within and among collection locations in each population. Pairwise analyses of genetic divergence (F ST and Jost's D) [55] among sampling locations were calculated using GenAlEx. Geneflow was estimated by calculating number of migrants per generation (N m ) using GenAlEx. A test for isolation-by-distance was conducted by testing the significance of a correlation between pairwise Nei's D (calculated using GenAlEx) and geographic distance among sampling locations. The correlation computations between pairwise genetic and geographic distances between populations were analysed using a Mantel test [41]. Geographic distance was meansured among collection sites by measuring distances following waterways in ArcMap GIS (ESRI).

Mitochondrial DNA
From the 64 sequenced individuals collected from 7 locations in the middle and lower Yangtze River drainage, 37 unique COI haplotypes were identified (GenBank Accession Nos. MG210495-MG210558). The Gan River had the greatest variation with 13 haplotypes, and the lowest was Poyang Lake with 4 haplotypes (Table 2). Haplotypes diversity values at each population varied between 0.857 and 0.975 (Table 2). The Gan River had the greatest haplotype diversity (0.975), and the lowest was Poyang Lake (0.857). Nucleotide diversity values ranged from 0.00726 to 0.04592 (Table 2). Poyang Lake had the greatest nucleotide diversity (0.04592), and the lowest was Liangzi Lake (0.00926).
Phylogenetic analyses (Fig 2) showed strong support for the monophyly of Nodularia (posterior probability = 1.00). Nodularia nuxpersicae and N. nipponensis were found to nest within the 37 N. douglasiae haplotypes sequenced from the Yangtze and additional N. douglasiae COI sequences from Japan, Russia, and South Korea. The 37 COI haplotype sequences from the Yangtze drainage in combination with Nodularia COI sequences from GenBank formed three clades: three haplotypes restricted to Xiannv Lake, Poyang Lake, and the Gan River (H15, H16, and H29); a single haplotype found in two individuals from Hongze Lake (H31); N. nuxpersicae, N. nipponensis and the remaining 33 haplotypes formed the largest clade. Nodularia sinuolata from South Korea was sister to all of the remaining Nodularia COI seqeuences.
Using the TCS algorithm, POPART software produced a single haplotype network (Fig 3). The most frequent haplotype (H16) occurred in 6 individuals and was shared by individuals in the Gan River, Poyang Lake, and Xiannv Lake. Twenty-three haplotypes were rare and occurred in just a single individual. As seen in the phylogeny (Fig 2), for the most part, the haplotypes show little geographic structure. However, a unique group of haplotypes (H32, H33, H34, H35, and H37) were found in Hongze Lake and Taihu Lake, the easternmost collections.
The AMOVA results showed that 11.99% of the total genetic variance was among the seven collection locations, and among sampling location differentiation was significant (overall F ST = 0.1874, p<0.001, Table 3). Pairwise F ST ranged from -0.025 to 0.393 among the collection locations (Table 4).
Genetic differentiation as represented by pairwise genetic distance values among the seven collection locations was not correlated with geographic water distance indicating that more geographically distant site combinations did not produce higher levels of genetic differentiation (p = 0.5020).
The mismatch distribution of analysis pairwise differences was significantly different from the expected distribution of the expanding population model (Fig 4). Similarly there was a lack of statistical significance of Tajima's D test (p<0.01), and non-significant Fu's FS (p<0.01). Moreover, when all samples were pooled together, Tajima's D and Fu's FS test were not significant (p<0.01, Table 5). Additionally, the BSPs showed that N. douglasiae has had a stable historical population size with a small recent expansion event occurring between 250,000 and 300,000 years ( Fig 5). However, the model comparison analysis showed that constant population size was the best fit for the model to the data set, suggesting that there was not much support for the recent expansion trend.  (Table 6). Deviations from HWE were found at only 8 of 91 locus-collection site combinations loci after a Bonferonni correction and were not consistently found at any site or locus (S2 Table). Significant tests for null alleles occurred in 8 of the loci used, however the estimated null allele frequencies were generally low ranging from 0.000 to 0.2792 at any given collection location-locus combination (S3 Table).
The only locus where null alleles were consistently detected at levels that could potentially affect the outcomes of population-level results [56][57] was locus Udo14. Analyses (e.g.,  Wilcoxon tests showed evidence for a recent genetic bottleneck at all of the locations except for Liangzi Lake using the SMM model ( Table 7). The Gan River also showed a significant bottleneck using the TPM model (p<0.05).
Using the data generated from the STRUCTURE analysis, the Evanno et al. [50] ΔK method indicated that K = 2 was the most likely (S1 Fig). Under K = 2, Dongting Lake, Poyang Lake, and the Gan River formed one genetic population and Hongze Lake, Taihu Lake, Liangzi Lake, and Xiannv Lake formed a second genetic population. While, K = 2 was most probable using the Evanno et al. [50] method, K = 3 had a slightly higher log-likelihood score (S1 Fig). With K = 3, Xiannv Lake became a distinct group (Fig 6). The PCoA showed a similar pattern of genetic structure (Fig 7) to the STRUCTURE analysis, with two clusters appearing along axis 1 and Xiannv Lake showing differentiation along axis 2. The PCoA explained 46.9% of the genetic variation across the 13 microsatellite loci in the first two axes.
The AMOVA results showed that 36.0% of the total genetic variance was among the seven collection locations, and among sampling location differentiation was significant (p<0.0001, Table 8). Pairwise F ST and Jost's D was congruent with the pattern of genetic structure revealved by analyses done in STRUCTURE and the PCoA showed that genetic differentiation was significant was moderate to high among the seven collection locations with the exception of the comparisons among Dongting Lake, Poyang Lake, and the Gan River and between the Hongze Lake and Taihu Lake (Table 9). Using a Mantel test, genetic differentiation (F st and Jost's D) among the collection locations using microsatellites was not significantly correlated with pairwise F st values calculated for the mtDNA (p>0.05). Estimates of gene flow (N m ) were generally low (i.e., N m <1) except for among Dongting Lake, Poyang Lake, and the Gan River and between the Hongze Lake and Taihu Lake (Table 10). Genetic structure and diversity of Nodularia douglasiae from the middle and lower Yangtze River drainage Genetic differentiation as represented by pairwise genetic distance values between all seven sample sites were shown to be not correlated with geographic water distance indicating that more geographically distant site combinations did not produce higher levels of genetic differentiation (p = 0.485).

Discussion
The results of this study show clear genetic structure in N. douglasiae across the middle and lower Yangtze River drainage. Two clear genetic groups are revealed using analyses of microsatellite genotypes consisting of 1) Dongting Lake, Poyang Lake, and the Gan River and 2) Liangzi Lake, Xiannv Lake, Hongze Lake, and Taihu Lake. The pattern of genetic structure found using the mtDNA dataset is only partially congruent with the pattern revealed by the microsatellites. The main difference between patterns reveled by the microsatellite and  Genetic structure and diversity of Nodularia douglasiae from the middle and lower Yangtze River drainage mtDNA datasets was that specimens from Liangzi Lake grouped with specimens from Hongze Lake and Taihu Lake with the microsatellite data, but grouped more closely with Dongting Lake and the Gan River with the mtDNA dataset. The overall resulotion of the pattern of geographic structure among the collection locations was fairly poor and inconsistent using the mtDNA dataset. This poor resolution and inconsistency may be a result of the very high haplotypic diversity among the specimens used (37 haplotypes from 64 individuals) and relatively small sample size of the mtDNA dataset.  Genetic structure and diversity of Nodularia douglasiae from the middle and lower Yangtze River drainage The microsatellite dataset is robust and of high quality with few loci out of Hardy-Weinberg equilibrium and relatively low numbers of null alleles predicted to be present in the dataset. The estimated null allele frequencies were generally below thresholds that would impact the results or interpretations of population-level analyses [56][57]. Null alleles in microsatellite datasets are frequently encountered and appear common in bivalves [58][59].

Evolutionary history and genetic structure
The geographic structure among the sampling locations using the mtDNA dataset was somewhat ambiguous. This ambiguity was likely a result of the high haplotype diversity (37 haplotypes), but relatively small sample size (n = 64). The COI phylogeny did not resolve N. douglasiae to be monophyletic due to the inclusion of N. nuxpersicae from Vietnam and N. Table 1.  completed without a priori populations assigned, admixture and correlated alleles were assumed, The most probable number of populations was K = 2 using the Evanno et al. [33] method, but K = 3 had slightly higher ln likelihood scores. Collection location codes are as in Fig 1. https://doi.org/10.1371/journal.pone.0189737.g006 nipponensis from Japan. Klishko et al. [18] found that N. nipponensis was sister to the COI sequences of N. douglasiae that they used in their analysis and thus chose to maintain N. nipponensis as a valid taxon. The phylogeny resolved in this study found that N. nipponensis and N. nuxpersicae were nested within the N. douglasiae COI sequences from the Japan, Russia, South  Given that the N. douglasiae COI lineages from the middle and lower Yangtze were distributed across several collection locations, it seems that N. douglasiae is a single species in the study area. One lineage was found in Poyang Lake, the Gan River, and Xiannv Lake, and Hongze Lake (H15, H16, H29, H31; Fig 3), while the second highly diverse lineage was broadly distributed across all collection locations. The haplotype network results (Fig 3) indicated that haplotype diversity was high, but that closely related haplotypes were broadly distributed across most of the sampling locations. An AMOVA of the COI sequence data showed that genetic differentiation was significant among many of the collection locations and that some sampling locations were significantly differentiated from one another. The general pattern of haplotype differentiation among sampling locations indicated the following groups: Hongze Lake and Taihu Lake; Liangzi Lake and Dongting Lake; and Ponyang Lake and Gan River.

Table 7. Results of tests for genetic bottlenecks in N. douglasiae from seven populations in the middle and lower reaches of Yangtze River using Wilcoxon tests with three different models of evolution and a mode-shift test. Collection location codes as in
Analyses of microsatellite dataset shows that high levels of genetic differentiation exist among the collection locations of N. douglasiae [60], but do not support the existence of more than one species of Nodularia in the study area. The analyses of the microsatellites were much better than the results of the mtDNA sequence data at resolving geographic structure among Table 9. Analysis of genetic differentiation coefficient (Fst) (below diagonal) and Jost D est (above diagonal) calculated using genotypes from 13 microsatellite loci among seven collection locations of Nodularia douglasiae from the middle and lower Yangtze River drainage. Bold type indicates statistical significance after Bonferonni correction (α = 0.002381). Genetic structure and diversity of Nodularia douglasiae from the middle and lower Yangtze River drainage the collection locations. Despite this pattern, genetic differentiation was low and gene flow was high among Dongting Lake, Poyang Lake, and the Gan River and among the Hongze Lake, Taihu Lake, and Liangzi Lake. The Xiannv Lake collection location was most similar to the Hongze Lake, Taihu Lake, and Liangzi Lake, but was found to be genetically distinct. This pattern of overlap can also be seen in the STRUCTURE plots (Fig 6) and the PCoA (Fig 7). All of the analyses of genetic structure, genetic differentiation, and gene flow were congruent. The pattern of genetic structure revealed by the analyses of the microsatellites is geographic in nature and suggests that the divergent COI lineages freely interbreed as the lineages were found to occur across the groups revealed by the microsatellites (e.g., Xiannv Lake, Poyang Lake, and Gan River).

LZ
The results suggest that the connectivity of rivers and lakes in the Yangtze River drainage was very important for dispersal in N. douglasiae. Historically, all seven collection locations were hydrologically connected, however, habitat alterations as a result of dam construction, dredging for canals and sand/gravel mining, and a major increase in urbanization in the last 100 years have had considerable effects of the connectivity among the mainstem of the Yangtze River and its tributaries rivers and lakes [15]. Of the areas sampled in this study, only Poyang Lake (and the Gan River) and Dongting Lake continue to have direct and natural connections to the mainstem of the Yangtze River (i.e., no dams) [12,61]. The Gan River is the largest river running from north to south in Jiangxi Province, China, flowing into Poyang Lake, and is the seventh largest tributary of the Yangtze River. In 1958, a dam was constructed at the outlet of Xiannv Lake that blocked upstream connectivity from the Gan River (and Poyang Lake) [62]. Between 1942 and 1963, hydrologic connectivity between Liangzi Lake and the Yangtze River was also blocked [63][64]. Dredging and canal contruction over the last several hundred years has increased the connections among Hongze Lake, Taihu Lake, the Yangtze River, and the East China Sea [65], but in 1954 a flood control dam was contructed at the outlet of Hongze Lake blocking upstream connectivity from the Yangtze River [66]. Taihu Lake is a geologically recent waterbody, being a large embayment of the East China Sea as recently as 1 million years ago. Gradually it became separated from the sea and is now the third largest freshwater lake in China. The hydrological connectivity among Taihu Lake, the Yangtze River, and the East China Sea has been considerably altered over the last few hundred years as a result of dam construction, canal construction, dredging, and major urbanization [67][68][69].
In theory, an increase of geographic distance should correlate with a gradual reduction of gene flow, resulting in genetic differentiation among populations, i.e., isolation-by-distance [70][71][72]. However, our results showed pairwise genetic distance values between all seven sample sites were not correlated with geographic water distance based on analyses of both microsatellites and mtDNA. Some of the patterns of genetic structure revealed is quite puzzling given that the most geographically proximate collection locations were not always the most genetically similar (e.g., Xiannv Lake and the Gan River). Some of these patterns may have been the result of anthropogenic movement of fish parasitized with glochidia of N. douglasiae for stocking and aquaculture across the region [15,73]. Given the long history of fish stocking and aquaculture in China [15,73], it is plausible that some of the observed pattern of genetic structure in N. douglasiae is the result of host fish stocking (moving parasiting larval mussels), but this is at best speculative until more is known about host use by N. douglasiae. It is also plausible that some of the pattern could be attributed to the movement of adult mussels for use as a human food resource, but again this is speculative.
Adult unionids disperse relatively little as adults, with long-distance dispersal being facilitated by hosts during their larval (glochidial) stage [74][75]. Given the pattern of genetic structure for N. douglasiae in the system (i.e.; similarities among Poyang Lake, Gan River, and Dongting Lake; and between Hongze Lake and Taihu Lake), it seems that at least some of the host fish species for N. douglasiae are highly vagile and capable of long distance movements through the highly hydrologically interconnected (before the active damming of rivers over the past 50 years) large river and lake system in the Yangtze River drainage, thus the maintainance of gene flow and a high degree of connectivity among habitats may be important. While there are no studies of unionids across the middle and lower Yangtze River drainage, some fish (and potential hosts for N. douglasiae) show similar patterns of genetic structure across the region [76][77][78]. Genetic studies of North American unionids have shown that unionids that use hosts with limited dispersal capabilities like Epioblasma triquetra (Rafinesque, 1820) using Percina caprodes [Rafinesque, 1818;Logperch] show high levels of population divergence and structure even at relatively short geographic distances [79], while unionids using highly vagile host fish capable of moving through large river systems like Quadrula quadrula (Rafinesque, 1820) using Ictalurus punctatus [Rafinesque, 1818; Channel Catfish] show lower levels of population divergence and structure and only show strong divergence and structure in populations separated by relatively high geographic distances [20,24,80]. Given that the host fish species of N. douglasiae are currently unknown, inferences about the dispersal abilities is impossible until experiments to determine potential host fish are conducted.
The mismatch distribution analysis and neutrality tests of the mtDNA sequence data suggest that N. douglasiae across the seven collection locations did not have a recent population expansion, and suggested that the current distribution is quite ancient. These analyses also indicate that the population dynamics of N. douglasiae are quite stable. This is not a suprising result given that N. douglasiae is a widely distributed species and often the most abundant unionid species in the region [8,12,61,81]. With the Yangtze River drainage in China being among the most biodiverse regions in the world for unionid mussels [9,61] and other aquatic organisms [12,15,81], this study represents an important first step for understanding the population-level diversity and structure of unionids at a regional scale.

Genetic diversity
Higher levels of genetic diversity among populations of aquatic organisms could improve evolutionary potential for dealing with habitat change, effects of pathogen infection, and other selective forces [82][83][84]. The results analyses of both the mtDNA and microsatellites suggest that there is robust genetic diversity among the populations of N. douglasiae in the middle and lower Yangtze River drainage.
The analyses of the COI seqeuences showed that the haplotype diversity of N. douglasiae among the seven collection locations was high. There were more haplotypes found in N. douglasiae compared with other some rare and imperiled unionids [24,69], but similar to other widespread species [24,30].
The genetic diversity estimated from the microsatellite DNA results showed similar levels of genetic diversity to other unionids in the Yangtze River drainage. Mean observed heterozygocity (H O ) expected heterozygosity (H E ) estimated for N. douglasiae was somewhat lower than that of the widely distributed Sinohyriopsis cumingii (Lea, 1852; heterozygosity: 0.617-0.750) [85] and Solenaia oleivora (H O : 0.501-0.620, H E : 0.598-0.701) [86]. However, levels of heterozygosity calculated for N. douglasiae were somewhat higher than that estimated for Solenaia carinata (H O : 0.472, H E : 0.478) [32], an endemic species found only in Poyang Lake. However, these differences may simply be artifacts of the species-specific microsatellite loci used.
Virtually all N. douglasiae collection locations across the middle and lower Yangtze River drainage showed evidence of a recent bottleneck. It is unclear if these bottlenecks resulted from a founder effect due to colonization by a small founding population with low genetic diversity, or if these were the result of severe demographic reductions followed by subsequent recovery in population size. The moderate levels of genetic diversity as revealed by the polymorphic information criterion (0.25<PIC<0.5), may also be evidence for a recent genetic bottleneck.

Conservation implications
This study represents the first analyses of the genetic structure and diversity for this widespread freshwater mussel and the first for a unionid mussel in the middle and lower Yangtze River drainage. Large-scale patterns of genetic structure occasionally differ among unionid species in the same geographic region [21,25,87]. Therefore, elucidating the commonalities in genetic structure and diversity among species will be necessary for making broad conservation inferences. Future research must include studies to determine dispersal capabilities of Yangtze basin unionids during all life stages [74], and studies that develop a clear understanding of the complex patterns displayed by a variety of freshwater mussel species [24]. While still poorly understood, declines in freshwater mussel populations are occurring in China [8,61]. Unionid populations in the Yangtze River region are especially vulnerable and with drastic reductions in abundance and diversity following the human disturbance and habitats fragmentation [8,12,61]. Currently, only two lakes (Poyang Lake and Donting Lake) remain connected with the Yangtze River. While status assessments have not been completed, it is estimated that approximately 80% of freshwater mussel species in the Yangtze River region could fall into an endangered or threatened status using IUCN criteria [12,61]. Conservation efforts should attempt to keep individuals with similar genetic profiles together and avoid mixing of individuals from distinct genetic groups [25,88].
In this study, N. douglasiae in the Yangtze River region showed robust genetic diversity, and significant and often high genetic differentiation (e.g., some pairwise F st >0. 15) and limited gene flow among the seven collection locations. Moreover, although the historical population dynamics of N. douglasiae appear stable, loss of hydrologic connectivity among rivers and lakes in the Yangtze River drainage may lead to increased isolation of populations and possibly leading depression and population declines. Genetic structure of common species have been shown to be useful surrogates for predicting genetic structure of rare species in North American unionids [23][24]. Therefore, studies on the genetic structure and diversity of common and widespread species like N. douglasiae may assist in understanding general patterns for freshwater mussel populations in the Yangtze River drainage. At the same time, we also propose the urgent need for research on the life history of N. douglasiae and other Chinese unionids with an emphasis on characterizing habitat preferences and host-testing experiments to identify potential host fish species.   [30] . Bold type indicates significant probability for the presence of null alleles. (DOCX) S4 Table. List of all individual Nodularia sp. and outgroups used, collection sites, and Gen-Bank accession codes. (DOCX)