A morphometric and genetic comparison of Sinanodonta woodiana ( Lea , 1834 ) populations : does shape really matter ?

Sinanodonta woodiana (Lea, 1834) is an invasive species which has spread rapidly across Europe over the last few decades. Although it is already well established throughout Italy (both in lakes and in rivers), during a regular monitoring of the mussel communities in Northern Italy, this species was found for the first time in the fluvial natural reserve of Ghiaia Grande (Po River basin). The strong differences in shell shape between these specimens and those from Lake Maggiore prompted a genetic comparison based on DNA barcoding analyses to establish whether specimens from both sites belong to the same species. Our results confirm that the wide shell plasticity may help to explain the role of environmental factors in driving shells’ morphology. Additionally, it may also induce S. woodiana misidentification with the native Anodonta sp. This fact may have strong negative implications for freshwater conservation if on one hand, the native species distribution is falsely over represented; or on the other hand, if the invasive species has a much wider distribution than predicted.


Introduction
The Chinese pond mussel Sinanodonta woodiana (Lea, 1834) (Bivalvia: Unionidae), a species originating from East Asia (South-eastern Russia to Malaysia), spread rapidly across Europe over the last few decades (Kiss 1990;Paunovic et al. 2006;Popa et al. 2007;Munjiu and Shubernetski 2008;Pou-Rovira et al. 2009;Lajtner and Crnčan 2011).Its first detection in Italy (Manganelli et al. 1998) was followed by an increasing number of records.Nowadays, it seems to cover all waters in the Italian Peninsula including the largest Italian Lakes such as Lake Garda (Cappelletti et al. 2009) and Lake Maggiore (Kamburska et al. 2013).The presence of S. woodiana was also recorded in different lakes and rivers across the basins of the Po, Adige, Piave, Reno, Arno, Tiber and Volturno rivers (Manganelli et al. 1998;Bodon et al. 2005;Albano 2006;Solustri and Nardi 2006;Cianfanelli et al. 2007;De Vico et al. 2007) and in Lake Santa Rosalia in Sicily (Colomba et al. 2013).
The accidental introduction of fish specimens bearing glochidia (the parasite larval stage of the bivalve) appears to be a reasonable explanation for the rapid expansion of S. woodiana in most European countries, including Italy (Spyra et al. 2012;Gherardi et al. 2008 andCianfanelli et al. 2007).In Italy, other probable invasion pathways were the commercial trade (e.g. for ponds and aquariums) and its culture for pearls or motherof pearl production (Berni et al. 2003).
S. woodiana is a host generalist, being able to parasitize several native and non-native fish species.This behaviour is believed to increase the opportunities of reaching and establishing in new environments (Douda et al. 2012).The parasitic period is generally short, with development and larvae release in 8-12 days at 20-25°C (Kiss 1990).Characterized by a fast growth, an early maturity (i.e. in the first year of life), a high reproductive output and a lifespan of 10-15 years (Dudgeon and Morton 1983;Afanasjev et al. 2001;Kraszewski and Zdanowski 2007), the species displays many features of a successful invader; high dispersal capacity promotes the impressively rapid diffusion of the species and its wide ecological tolerance allows for the successful establishment in invaded ecosystems.
The species ability to adapt to different habitats, tolerating changes in chemistry and physical parameters of water and sediments, can be a driver of the observed morphological plasticity (e.g.Soroka and Zdanowski 2001;Kraszewski 2006;Douda et al. 2012;Demayo et al. 2012).The native European anodontines are also characterized by a great phenotypic plasticity (Nagel and Badino 2001;Reis et al. 2013).This not only hinders the attribution of a specimen to a particular species, but also promotes S. woodiana misidentification with the native Anodonta species.For instance, the first record of S. woodiana in Lake Maggiore (Kamburska et al. 2013) raised serious doubts about its correct identification (Mienis 2013).Indeed, the unusually elongated shape of Lake Maggiore specimens suggested that they could belong to the native European species Anodonta anatina.
To verify species identification based on morphological characters, a DNA barcoding approach was proposed (Hebert et al. 2003).This method is based on a comparative analysis with a standard reference DNA sequence.In higher animals the most common used DNA barcode sequence is a fragment of the cytochrome oxidase I mitochondrial gene.This molecular approach can provide an accurate identification of mussel species at all life stages (Campbell et al. 2008;Boyer et al. 2011;Zieritz et al. 2012).
The aim of the present study was to compare morphological distinct populations of putative S. woodiana specimens using both molecular and morphometrical approaches, in order to assess if shell shape differences are species specific or more related to environmental variables.
To this aim, freshwater bivalves from the Po River basin and from Lake Maggiore were identified using a Barcoding technique (COI) and these results were compared with a combination of traditional and geometric morphometric techniques, which have been previously used with success to test inter-and intraspecific variations e.g., (Zieritz and Aldridge 2009).

Study area
The Po River is the longest Italian river, with a total length of 650 km.The river collects waters from the whole Northern Italy, the Alps, the great sub-alpine lakes, and, passing through the Padania plain, flows into the Northern Adriatic Sea.The catchment covers an area of 71,000 km², collecting waters from 141 tributaries.The fluvial reserve Ghiaia Grande with its total area of 4.62 km² is part of the large protected area of the Po and Orba systems, with a total surface area of about 140 km² between Piedmont and Lombardy (Figure 1-A).During October 2012, a survey was performed on the Ghiaia Grande reserve to monitor the populations of native unionids (Unio sp., Anodonta sp.).Structured and recruiting Unio sp.populations were found, but no native Anodonta specimens were sampled.Instead, various S. woodiana individuals were caught.These specimens were located along a secondary river branch (Figure 1-B).The entire area is characterized by the presence of meanders, a slow water flow and fine sediments deposition.The upstream (PoUS, 45°09'05''N, 8°19'41''E) and downstream sites (PoDS 45°09'35''N, 8°19'24''E) are defined relatively to the intake of reserve water in the secondary branch.Lake Maggiore is the second deepest and largest subalpine lake in Italy.Samples from Lake Maggiore (LM) were collected at a sampling point (45°50'09.0"N,8°37'25.5"E)located along the southeast coast of the lake during the summer of 2012 (Figure 1-A; Kamburska et al. 2013).This sampling site is characterized by a moderate wave intensity and a high sediment patchiness.However, S. woodiana was found only in soft sediments.

Studied samples
A total of 23 S. woodiana specimens (11 PoUS, 12 PoDS) were collected by visual and tactile search from the secondary branch of the Po River, while 22 specimens were collected from Lake Maggiore.The abundance of mussels was determined by collecting and counting individuals within randomly selected quadrats of 0.25 m 2 .

Linear morphometrics
Sampled organisms were measured with a digital calliper to the nearest 0.01 mm, defining the following four morphometric variables: shell length (L), the maximum antero-posterior dimension of the shell; shell width (W), the maximum leftright dimension with both valves compressed; umbo height (H), the dorsal-ventral dimension of the shell measured perpendicular to the length measured at the level of the umbo; wing height (h), the maximum dorsal-ventral dimension of the shell measured perpendicular to the length (Figure 2).Live animals (only from the River Po) were weighed in triplicate to get the individual weight (BW, g).The measured shell length (L), shell width (W), wing height (h) and umbo height (H) were used to compute the L/W, L/h, L/H, W/h and W/H ratios.The normal distribution of the measured variables, in each population, was calculated using two tests: the Shapiro-Wilk test, optimized for small sample sizes (N<50) and/or a Chi-squared test.T (for equality of means) and F (for equality of variance) tests were performed to assess whether the measurements from each of the three groups of samples (PoUS, PoDS, LM) are significantly different.The multivariate normality of the log-transformed ratios was checked by computing Mardia's multivariate skewness and kurtosis, with tests based on chi-squared (skewness) and normal (kurtosis) distributions.A MANOVA analysis was performed with the populations' location designation as independent variables and the five above mentioned ratios as dependent variables.A Bonferroni correction for multiple testing was applied for p-values.
The growth patterns of the studied populations (allometric or isometric) were analyzed through a curve fitting procedure.This was done with a linear regression of the log-transformed values of our measurements, fitting the points to the allometric equation: For this purpose, the shell length vs. shell height (L-H, L-h), shell width vs. shell length (W-L) and shell width vs. shell height (W-H, W-h) as well as all four linear measurements (L, W, h, H) vs. body weight (BW) were analyzed for all Po river samples.The growth patterns were considered to be isometric if the slope (a) of the regression line was not statistically different from 1 (3, when compared to BW), negative allometric if a<1 (3, when compared to BW), and positive allometric if a>1 (3, when compared to BW).A bootstrapped 95 % confidence interval for a was computed with 2000 replicates.All analyses were performed with the software package PAST (Hammer et al. 2001).

Geometric morphometrics
Specimens were classified in four length classes and assigned to the respective age groups according to (Afanasjev et al. 2001) and (Spyra et al. 2012): i) up to 5 cm (young), ii) from 5 to 10 cm (small), iii) from 10 to 15 cm (medium), and iv) above 15 cm (large) (Appendix 1a,1b,1c).The age of the individuals was determined from shell length (L), since the annual growth rings may not faithfully represent the actual age of the Chinese pond mussel (Kiss 1995).The age of the mussels was estimated from the length-age relationship reported by Dudgeon and Morton (1983).By making the assumption of a similar growth rate of S. woodiana in its native range and in colonized areas, mussel age classification was also performed following Spyra et al. (2012).
The geometric morphometrics and statistical analyses were performed with software MorphoJ (Klingenberg 2011).A dataset comprising 45 pictures (11 from PoUS, 12 from PoDS and 22 from LM) was analysed with MorphoJ after a previous elaboration with Tps series of morphometric programs (Rohlf 2008), where 15 landmarks were defined (5 groups of 3 landmarks each: umbo, wing, ventral edge, anterior and posterior edges, numbered dots on Figure 2).This software provides a platform for a wide range of analyses in geometric morphometrics.Landmark data allow for the visualization of the difference between population means or the deviation of an individual from its population mean.The major variation patterns of size and shape were explored using a principal component analysis (PCA) and centroid size to scale for the configuration of selected landmarks.Landmarks were selected to enlighten visible differences between the rounded and the elongated morphotypes as summarized in Table 1.All pictures were scaled and the differences in shape and dimensions were analyzed.

DNA isolation, amplification and sequencing
For the genetic analysis small samples from the mantle were excised from 26 S. woodiana from the three sites: Lake Maggiore LM. (n=16), Po River Downstream (PoDS) (n=6) and Po River Upstream (PoUS) (n=4).DNA isolation was performed with a NucleoSpin® Tissue kit (Macherey-Nagel GmbH and Co. KG, Düren, Germany), according to the company specifications.
For each DNA sample, a 650 bp fragment of the mitochondrial gene for cytochrome oxidase c subunit I (COI) was amplified, using the primers LCO1490 and HCO2198 as described by Folmer et al. (1994).
The amplification was performed in 25 μl of a solution containing 10 mM Tris-HCl (pH 8.8 at 25°C), 50 mM KCl, 0.08% (v/v) Nonidet P40, 2.5 mM MgCl 2 , each dNTP at 0.1 mM, each primer at 0.1 μM, 1 unit of Taq DNA polymerase (Fermentas UAB, Vilnius, Lithuania), and approximately 10 ng of DNA template.The temperature profile of the polymerase chain reaction for the Table 1.Landmarks position along Sinanodonta woodiana shell (Figure 2) and relative shell shape characters analyzed.

Landmark position Shell shape character
Anterior edge (1), top (2) and posterior edge (3) of umbo relative height and width of umbo Tip (4) and two equidistant points (5, 6) on the edge of wing degree of wing development Posterior margin: point at the maximum distance from the anterior margin (8) and two points (7, 9) at the beginning of curvature degree of posterior margin convexity Ventral margin: point at the maximum distance from the top of umbo (10) and two equidistant points (11,12) along the maximum curvature degree of ventral margin convexity Anterior margin: point at the maximum distance from the posterior margin ( 14) and two points (13, 15) at the beginning of curvature degree of anterior margin convexity COI marker consisted of initial denaturation at 94°C for 30 s, followed by 5 cycles at 93°C for 30 s, 45°C for 45 s, 72°C for 45 s, followed by 30 cycles at 93°C for 30 s, 55°C for 45 s, 72°C for 45s, and a final extension step performed at 72° for 10 min.The PCR products were separated on 1% agarose gel, and the fragments of interest purified using the GeneJET™ Gel Extraction Kit (Fermentas UAB, Vilnius, Lithuania).Unidirectional sequencing reactions were performed by Macrogen Europe (1105 AZ, Amsterdam).
Sequences were aligned using CodonCode Aligner 3.7.1.(CodonCode Corporation, Dedham, MA, USA) and verified manually.The consensus sequence was used as a query to the BOLD SYSTEMS Identification Request (http://www.boldsystems.org).

Morphometric analysis (linear, geometric)
The mean density of S. woodiana in PoUS and in PoDS was of 4.4 ± 1.3 ind m -2 and 9.6 ± 2.2 ind m -2 , respectively.Although in the upstream station both S. woodiana and the native Unio elongatulus were found, downstream only S. woodiana was present.Additionally, a different shell size range was also detected in these two river sections The younger animals, aged from three to five years (mean length 89.42 ± 18.69 mm) were collected upstream (Table 2, supplementary Table 1S), whereas the downstream population was represented by animals belonging to different size classes, with a mean length of 132.44±36.24mm (Table 2, supplementary Table 2S).The smallest collected specimens measured 64.28 mm on the upstream site and 73.73 mm downstream.The upstream site was apparently colonized by a younger population, where the maximum length of the organisms was 115.19 mm, corresponding to an age of 4-5 years.On the other hand, the downstream population included older animals, with a maximum length up to 194.08 mm probably over 10 years of age.The length of animals from Lake Maggiore ranged from 91.27 mm to 164.96 with a mean value of 130.47 ± 22.89 mm (Table 2, supplementary Table 3S).

Univariate T and F-tests
All measurements data exhibited a normal distribution.A comparison between linear measurements and between length/wing height (L/h) and length /width (L/W) ratios of the specimens in our samples, performed with univariate T and F tests, suggested that all collected animals manifested a similar relationship among their measurements at different size classes (Table 3).The p-values of these tests are presented in Table 4.The relation between length and umbo height (L/H) and between length and fresh total weight (L/BW) are represented graphically (Figure 3).

Multivariate Analysis of Variance (MANOVA)
The log-transformed data were tested for normality with a Mardia test.The p-values for skewness and kurtosis exhibited values above 0.05 in both cases.The MANOVA test (after Bonferroni correction for multiple comparisons) showed a significant difference of the multivariate means between the sample from Lake Maggiore and each of the samples from the Po River.However, no differentiation was observed between both Po river samples (Table 5).

Curve fitting analysis
The a values from the allometric equation, calculated for each of the three samples, are presented in Table 6.A linear correlation was found between length and height, and an exponential relation between length and weight for all sampled animals.The analogous trend could suggest that the growth of individuals sampled from two sites was similar.While fitting of regressions were higher for the downstream population (R 2 =0.983), data of the upstream population were slightly more dispersed around regressions, but nonetheless with a high R 2 value (0.961).

Geometric morphometrics analysis
To find out if other differences among animals occurred, pictures with landmarks were analysed with MorphoJ software.The principal component analysis revealed that the first two components Pc1 and Pc2 explained 54.29% and 13.48%, respectively, of the total variability among the landmarks, on all analysed specimens (Figure 4).The population from the Po River comprised two overlapping groups representing the populations from down and upstream.The population from Lake Maggiore is well separated on the right.It was evident that animals from Po River were more rounded in respect to specimens collected in Lake Maggiore.
Younger animals appeared to be more similar than older ones, among all populations.Indeed, using a classifier that labelled organisms with sampling site and the proper length (L), a clear grouping by size (age) could be distinguished by the PCA (Figure 5).The first relative warp clearly distinguished individuals by size, highlighting a wider shape variation among large (> 15 mm) than among small animals.While the younger   individuals (3-4 years old; < 10 mm) are closely grouped, regardless of the sampling site, the distribution of the larger specimens seems to be affected by their origin.In fact, there is an increasingly distinctive grouping by site, with Po River sites on the left and Lake Maggiore on the right of the graph, along with size increase.

Genetic analysis
A 636 bp COI fragment was successfully sequenced from the 26 specimens of S. woodiana analysed.Only one haplotype was identified in the three sampling sites (GenBank accession numbers KF731775-KF731777).The BOLD SYSTEMS Identification request query returned a species match with Sinanodonta woodiana at similarity levels between 99.84% and 100% (9 matches).

Discussion
The two populations of S. woodiana analysed in the present study are morphologically distinct, with specimens from the Po River being more similar in shape to the typical S. woodiana.In contrast, the unusual elongated shape of Lake Maggiore (Kamburska et al. 2013) and Lake Garda (Cappelletti et al. 2009) specimens lead some authors to question whether they were indeed S. woodiana (Mienis 2013).Other authors consider that S. woodiana could easily be confused with other species in the genus Anodonta (Bogan et al. 2011).As a consequence, our first aim was to establish if the populations reported in the present study belong indeed to S. woodiana.In order to answer this question, a DNA barcoding analysis was performed with specimens from Po River, but also with elongated specimens collected from Lake Maggiore.All 26 analyzed specimens exhibited the same mitochondrial haplotype, which returned a species match with S. woodiana from the BOLD SYSTEMS Identification Request.All these matching sequences were obtained from specimens collected in Europe.We consider this finding as a confirmation that the specimens reported in this study, but also those from Kamburska et al. (2013), represent different morphotypes of the species S. woodiana.
The Lake Maggiore specimens are significantly more elongated (i.e. higher length/ height -wing and umbo-ratios) and with less pronounced wings than specimens from both locations of the Po River.Shape differentiation increased with specimen's size: while younger animals were much closer, bigger animals are more dispersed along the first axis of PCA.
Collection of specimens with a wide size range allowed for considerations about the type of growth.
While younger animals are similar in shape, an increased shape differentiation occurs with ageing.Factors like habitat physico-chemical conditions and food availability are likely to drive shell shape on the animals' development and growth.This is also suggested by the fact that growth patterns, either isometric or allometric, are not exactly the same in all populations (see Table 6), which could imply that the species growth pattern is not controlled only genetically, but environmental factors are likely involved.While habitat characteristics are considered the most likely driving factors of morphological divergence between S. woodiana populations (Afanasjev et al. 2001;Soroka and Zdanowski 2001;Kraszewski 2006;Demayo et al. 2012;Urbanska et al. 2013), allometry explains most of the intrapopulation shell shape variation (Demayo et al. 2012).
Several unionid species are reported to display a high intraspecific morphological variability which can reflect an adaptive phenotypic response to habitat factors.For instance, shell shape is influenced by hydrological features and substrate composition, i.e., parameters which are mutually related and apparently induce typical adaptive modifications (e.g.Hinch and Bailey 1988;Zieritz and Aldridge 2009;Hornback et al. 2010).However, attempts to identify a specific habitat parameter which determines a specific change in shell shape gave contradictory results.As an example, more elongated shell shapes were reported to reflect adaptation to strong wave action in lakes (Hinch and Bailey 1988) but also to characterize populations of slow flowing habitats in rivers (e.g.Zieritz and Aldridge 2009).According to Hornback et al. (2010), shell shape reflects the degree of hydraulic variability rather than flow intensity, and slender morphotypes have advantages under high hydraulic disturbance (i.e.hydraulically flashy environments).Variations in growth determined by ontogenetic factors (e.g., feeding and/or metabolic shifts) or parasitic infestation by bucephalid trematodes (Rhipidocotyle sp.) can also significantly alter sagittal and lateral shell shape of unionids (Zieritz and Aldridge 2011).
In the present study, river specimens collected from areas with very slow currents and soft sediments, presented more rounded shells and more pronounced wings than lake specimens, which were collected from a lake area with a moderate degree of wave exposure.Enlargement

Figure 1 .
Figure 1.Study area.A -Po River Basin.Lake Maggiore is highlighted in light blue, Po River in dark blue.B -Natural Reserve Ghiaia Grande.Arrows indicate water flow direction.PoUS -Po UpStream Site, PoDS -Po DownStream Site.

Figure 3 .
Figure 3. Regression analysis of selected parameters: A-Linear regression relation between L/H (length and umbo height in mm); B-Curvilinear regression relation between L/BW (length (mm) and total fresh weight (g)); Po River UpStream Site (PoUS ) and DownStream Site (PoDS ).

Figure 4 .
Figure 4. Principal component analysis of geometric morphometric of Po River UpStream Site (PoUS in blue) and DownStream Site (PoDS in green) populations compared with Lake Maggiore population (LM in red).Confidence ellipses p=0.9.Labels of dots refer to length of organisms in mm.ANOVA test of null hypothesis: organisms are different in shape due to change, p<0.0001.

Figure 5 .
Figure 5. PCA of centroid size and Pc1 classified by age (Young=S in blue, Adult=M in yellow and Old=L in black) and sites (Po Upstream=PoUS (full circle), Confidence ellipse (dotted line in light blue) p=0.9.Po Downstream=PoDS (full quadrat), Confidence ellipse (dotted line in red) p=0.9.Lake Maggiore=LM (full triangle) Confidence ellipse (dotted line in green) p=0.9.Labels of dots refer to length of organisms in mm.

Table 2 .
Morphometric measurements of Sinanodonta woodiana collected in River Po and in Lake Maggiore.PoUS-Po River, UpStream Site; PoDS-Po River, DownStream Site; LM-Lake Maggiore Site.

Table 6 .
Curve fitting analysis.The values in the table represent a values with the bootstrapped 95 percent confidence interval from 2000 replicates (in parantheses).Shaded cells show where an allometric growth pattern was identified.PoUS-Po River, UpStream Site; PoDS-Po River, DownStream Site; LM-Lake Maggiore Site.