Native and invasive taxa on the Pacific coast of South America: Impacts on aquaculture, traceability and biodiversity of blue mussels (Mytilus spp.)

Gaining new knowledge of the native distributions of species (phylogeography) is more and more difficult in a world affected by anthropogenic disturbance, in particular by species translocations. Increasingly, molecular markers are required to support decisions about the taxonomy of native vs. introduced species, and the existence of their hybrids, to answer phylogeographic questions. In many fields, including aquaculture, traceability and food security, taxonomic and phylogeographic knowledge is key to the successful management and conservation of biodiversity. The Pacific coast of Chile is one of the last regions without a clear and agreed understanding of the taxonomy and systematics of smooth‐shelled blue mussels of the genus Mytilus. A panel of 49 bi‐allelic single nucleotide polymorphisms (SNPs) was genotyped in 338 Mytilus individuals collected from nine Chilean and five reference populations. All analyses confirmed the hypothesis that the native Chilean blue mussel is genetically distinct from the reference species M. edulis, M. galloprovincialis and M. trossulus. These results support the hypothesis of a unique evolutionary history of the native Chilean blue mussel on the Pacific coast of South America. It is therefore concluded that the native blue mussel from Chile should be recognized as M. chilensis Hupé 1854. We confirmed a recent Mediterranean origin of introduced M. galloprovincialis on the coast of Chile. This knowledge advances the understanding of global phylogeography of blue mussels and their bioinvasions and harmonizes taxonomy in the context of aquaculture production, seafood traceability, labelling and trade.

& Amorim, 2008). The determination of distributions of native species is not only important in its own right, but is applied in a variety of different management aspects, such as aquaculture, food security, traceability and labelling. Food fraud is now a bigger problem than ever, including dilution, mislabelling, ingredient substitution and tampered with products. This problem is addressed in recent international legislation (E.U. 2013;FDA 2011). Related to aquaculture, marine bioinvasions that may affect cultivated species are now being detected in many production areas and also in remote regions including offshore islands and Antarctica (Gardner, Zbawicka, Westfall, & Wenne, 2016;Lee & Chown, 2007;Shaw, Terauds, Riddle, Possingham, & Chown, 2014). It has been suggested that we are no longer in a position to wait for the establishment of native species baselines and that immediate management action is required (Ojaveer et al., 2015). Thus, taxonomy has a key role to play in the protection and sustainable exploitation of species (Mace, 2004;Seddon, Parker, Ostrander, & Ellegren, 2005).
The taxonomic status of native smooth-shelled blue mussels inhabiting the coast of Chile has been unclear and disputed for a number of years, with authors advancing different suggestions for its nomenclature, including M. edulis-like (Koehn, 1991;McDonald et al., 1991), M. edulis chilensis (Toro, 1998), M. galloprovincialis chilensis (Cárcamo, Comesaña, Winkler, & Sanjuan, 2005), M. galloprovincialis (Toro, Ojeda, Vergara, Castro, & Alcapán, 2005), M. edulis platensis (Borsa, Rolland, & Daguin-Thiébaut, 2012), M. galloprovincialis lineage of Southern Hemisphere origin (Westfall & Gardner, 2013), M. planulatus and M. platensis (Astorga, Cardenas, & Vargas, 2015). These different taxonomic designations may have arisen from a lack of clarity about which mussel species occur at specific sites and because the number and type of markers used in each research project are different (Borsa et al., 2012;Larraín et al., 2015). Despite the lack of agreement about the taxonomic status of the native Chilean mussel, the term M. chilensis has long been employed on food product labels and in scientific articles (Araneda, Larraín, Hecht, & Shawn, 2016;Astorga, 2014;Larraín, Díaz, Lamas, Uribe, & Araneda, 2014;Ouagajjou, Presa, Astorga, & Pérez, 2011;Oyarzún et al., 2016). The name is also used in aquaculture production statistics (FAO 2016) and on good aquaculture practice certifications (GAA 2013). However, as several authors have pointed out, the name has historically had no formal taxonomic standing. This controversy and the history of the discussion are reflected in changes listed in the Word Register of Marine Species (Horton et al., 2017).
Mytilus species in Chile, both native and introduced, have been studied extensively using nuclear ITS) and/or mitochondrial (16s rDNA RFLP, COI, COIII) molecular markers (Borsa et al., 2012;Fernandez-Tajes et al., 2011;Gérard, Bierne, Borsa, Chenuil, & Féral, 2008;Larraín, Diaz, Lamas, Vargas, & Araneda, 2012;Tarifeño et al., 2012;Toro, 1998;Toro et al., 2005;Westfall & Gardner, 2010. The primary problem has been that these markers target different regions of the genome that have different evolutionary rates, giving nonequivalent results when a few are used simultaneously (Kijewski et al., 2011;Rawson, Agrawal, & Hilbish, 1999;Zbawicka et al., 2012). Whilst the monolocus approach has the advantage of being easy to apply, it has the drawback that one locus represents a low power approach to delimit a species. A secondary problem arises from the fact that all smooth-shelled Mytilus taxa interbreed extensively (Michalek, Ventura, & Sanders, 2016). Such hybridization may complicate taxonomic resolution by blurring species boundaries and is not reflected in maternally inherited mtDNA markers that are often used for species identification, and also the evolutionary history of extensive interbreeding and introgression is not well represented by only a few nuclear DNA markers. What has been missing until recently is nuclear DNA markers that are speciesspecific and that can resolve ancestry when hybridization occurs either naturally or as a result of anthropogenic transfer of mussel types to non-native regions. Recent advances in next-generation sequencing methods, along with the increase in sequence data deposited in public databases, now permit the development of genomewide single nucleotide polymorphism markers (SNPs) that may be applied in mussel studies to resolve many of these issues because they cover multiple regions of the genome (Araneda et al., 2016;Mathiesen et al., 2017;Saarman & Pogson, 2015;Zbawicka et al., 2012;Zbawicka, Sanko et al., 2014. Blue mussels have long been farmed in many regions of the world (Kijewski, Wijsman, Hummel, & Wenne, 2009;Molinet et al., 2015).
Chile is now the world's second largest mussel (family Mytilidae) aquaculture producer (FAO 2016). Production is concentrated in the Gulf of Reloncaví and along the coastline of Chiloé Island (Los Lagos region) and is based on the native blue mussel, nominally Mytilus chilensis. However, other species in Chile have also been reported, including M. galloprovincialis in the Gulf of Arauco-Bío-Bío region Daguin & Borsa, 2000;Tarifeño et al., 2012;Westfall & Gardner, 2010) and M. edulis in the Strait of Magellan-Magallanes region (Oyarzún et al., 2016), although this last species is named as M. platensis by Gaitán-Espitia et al. (2016). In addition, at a limited number of locations along the Chilean coast, allele characteristics of M. trossulus (but not M. trossulus mussels themselves) have also been reported (Larraín et al., 2012;Oyarzún et al., 2016). Because accurate identification of mussel species produced by aquaculture is necessary for labelling, traceability, food security and marketing purposes (Larraín et al., 2012(Larraín et al., , 2014, it is important to understand which species is being grown where, which are native mussels, if native and introduced mussels interbreed, and if/how invasive mussels may affect aquaculture. Hybridization between Mytilus taxa has been reported in all geographic areas where two or more species coexist (Gardner, 1997) and hybridization patterns are complicated additionally by the occurrence of two mitochondrial lineages, female and male, and their recombination and introgression (Filipowicz, Burzyński, Śmietanka, & Wenne, 2008;Zbawicka, Wenne, & Skibinski, 2003;. Significantly, a number of studies have now demonstrated that hybridization of Mytilus taxa in aquaculture can cause unwanted or harmful effects to the industry and/or to native mussel populations (Crego-Prieto et al., 2015;Dias, Fotedar, & Snow, 2014;Michalek et al., 2016).
In this article, we describe molecular genetic analyses of blue mussels from the coast of Chile, one of the last biogeographic regions without a clear and universally agreed understanding of its mytilid taxonomy. Using a 49 SNP panel against blue mussels from multiple locations in Chile and reference species, we test the hypothesis that the native Chilean blue mussel (M. chilensis) is an endemic species, with its own unique evolutionary history in the Southern Hemisphere. Our aim was to contribute to the understanding of the taxonomic status of native smooth-shelled blue mussels and to test for the presence of other Mytilus taxa in Chile. This knowledge has practical applications by providing tools to solve issues related to aquaculture policy and management and highlights the potential threats to native mussel populations and aquaculture posed by introduced taxa along the Chilean coast. The SNP panel has direct global uses in seafood labelling, traceability and food security, and regionally by setting the basis for a protected origin designation for native Chilean mussels.

| Sample collection and DNA extraction
Samples (n = 338) of Mytilus spp. were collected from nine locations in Chile spanning almost all of the native mussel's distributional range (~2,500 km) and from five regions as reference samples: Pacific coast of Canada, Northern Ireland, Italy, New Zealand and Spain (Table 1, Figure 1). Provisional species identification of each individual was determined using PCR of the nonrepetitive region of the polyphenolic adhesive protein gene with the RFLP AciI method with primers Me15-16 (Santaclara et al., 2006) or directly by genotyping the equivalent SNP locus BM151A .

| SNP genotyping
In total, 338 Mytilus samples were genotyped using the Sequenom MassARRAY iPLEX genotyping platform (Gabriel, Ziaugra, & Tabbaa, 2009). Assays were designed for 79 candidate SNPs, selected from 385 putative SNPs, which were tested on 300 specimens of Mytilus collected from geographic regions including Europe, North and South America and New Zealand. Mussel SNPs were genotyped as previously described and following earlier testing of their reproducibility Zbawicka et al., 2012;Zbawicka, Sanko et al., 2014.

| Data analysis
Observed (H O ) and expected (H E ) heterozygosities and exact tests of departure from Hardy-Weinberg equilibrium (HWE) were determined using the R package adegenet 3.1-1 (Jombart, 2008). Significance was determined by Markov chain Monte Carlo with 100,000 simulations, and the Benjamini and Yekutieli false discovery rate (FDR-BY) was used to correct significance (p) values after multiple testing (Benjamini & Yekutieli, 2011). Genetic differentiation amongst populations was determined using global and pairwise F ST (theta) values (Weir & Cockerham, 1984), and the 95% confidence intervals for F ST were estimated by bootstrapping with 10,000 replicates using the R package diveRsity (Keenan, McGinnity, Cross, Crozier, & Prodöhl, 2013). The Cluster analysis was performed using three unsupervised methods: (i) discriminant analysis of principal component (DAPC) performed with adegenet 3.1-1 (Jombart, Devillard, & Balloux, 2010) where the number of clusters (K) was identified using the Bayesian information criteria (BIC). DACP variation was plotted using a K = 14 to match mussel populations with clusters; (ii) the nonparametric method implemented in AWclust (Gao & Starmer, 2008) that identifies the number of clusters based on a gap statistic; and (iii) the Bayesianbased method implemented in STRUCTURE with no prior information about the origin of individuals (Pritchard, Stephens, & Donnelly, 2000) assuming admixture and allowing for the correlation of allele frequencies between clusters. The tested number of clusters (K) ranged from 1 to the number of sampling locations plus 1. The length of burn-in period and the number of MCMC cycles after burn-in was 1,000,000 iterations each. Six runs were carried out for each K, and we used the Evanno, Regnaut, and Goudet (2005) method to identify the single value of K which captures the uppermost level of structure in Structure Harvester (Earl, 2012).
Genetic assignment was performed to assign or exclude sampled populations as possible origins of individuals, using the frequency criteria of Paetkau, Calvert, Stirling, and Strobeck (1995) in a selfassignment test with the leave-one-out (LOO) procedure, implemented in GeneClass2.0 (Piry et al., 2004). In this supervised method, individuals were considered to be correctly assigned to their location of origin if the assignment probability to that group was higher than any other assignment probability to any other group.
To identify loci with high information content for individual assignment to species, different ranking criteria were tested as follows: (i) F ST outlier loci above the upper limit of the 95% confidence interval (CI 0.95) were identified by use of LOSITAN (Antao, Lopes, Lopes, Beja-Pereira, & Luikart, 2008) with 1,000,000 simulations, a false discovery rate of 0.1 and a subsample size of 50; (ii) loci with minor allele frequency (MAF) above 0.1, 0.2, 0.3 and 0.4 (Hess, Matala, & Narum, 2011); and (iii) the most informative loci selected with backward elimination locus selection (BELS) version 1.0 (Bromaghin, 2008). This program assesses the power of all loci and sequentially eliminates the locus that makes the smallest contribution to individual assignment performance, thereby providing a ranking order for all loci.

| SNP markers, genetic diversity and Hardy-Weinberg equilibrium
Initially, 79 SNPs were used to genotype 338 Mytilus individuals from the 14 locations. Of these, 24 did not provide an acceptable quality score, four were monomorphic in all samples, and two were tri-allelic. These 30 loci were removed from further analysis.   Table S3) between native mussels from Chile

| Population genetic structure
The DAPC identified three clusters ( determined from the gap statistic (Fig. S1b, Table S4). Both methods correctly matched 99.7% of the individuals (337 of 338 mussels) to the three groups.
The DAPC plot of variation considering all locations (K = 14) also revealed the three major groups as previously detected using K = 3 ( Figure 3). As clusters are abstract objects that are not necessarily coincident with sampling sites, to draw the plot, the colour and shape of the symbols used for all individuals included in the cluster were given by the location that contributed highest number of individuals to that cluster. The number of individuals from each location in each of the 14 clusters is presented in Table 2 triangles. In this last group, the mussels from Spain to Atlantic (CAES) were divided into two clusters, the first one containing only CAES individuals and the second one containing mussels from CAES but also from ORIT (Mediterranean M. galloprovincialis) and Northern Ireland (LFGB) individuals (M. edulis). The Cocholgue (COCL) mussels were mainly clustered with the ORIT individuals, but because they were few in number in each cluster, no red triangles are shown in Figure 3. For the 49 SNP panel, the LOO method (Piry et al., 2004)  of assignment success, respectively, showing the mixture of individuals between these two European locations. The mussels from VACA and WENZ were assigned with 100% accuracy to their sampling locations (Table 3b).

| Highly informative locus panel
The ranking criteria selected different numbers of most informative loci, ranging from 6 to 25 (MAF > 0.4 to MAF > 0.1, respectively). The panel that performed best at assigning individuals to species was selected using the ranking criterion of MAF > 0.2 and included 19 highly informative SNP loci that correctly assigned 336 (99.4%) individuals to species (Table S5 and Table 4). Considering this MAF criterion along with F ST outlier loci and results from BELS, three loci (BM106B, BM151A and BM6C) were included in the group of most informative loci (Fig. S2). Two of these three loci have known mRNA functions, with the polyphenolic adhesive foot protein (BM151A) and the elongation factor G (BM6C), whilst the function of BM106B is presently unknown.

| DISCUSSION
Single nucleotide polymorphisms (SNPs) are powerful markers to monitor organisms at the individual, population and species levels.
They have been developed recently for mussels of the genus Mytilus to study hybrid zones, introgression and adaptive genetic variation with traceability purposes (Araneda et al., 2016;Saarman & Pogson, 2015;Zbawicka, Sanko et al., 2014. The availability of a SNP panel now allows for a multi−locus scan of the genome, adding more confidence to Mytilus species identification. Because not all SNP loci are equally informative based on their performance answering a specific research question, different ranking criteria have been employed to identify loci with high information content (Hess et al., 2011;Storer et al., 2012). These criteria can be used to create minimum panels that maximize individual assignment success to test hypotheses about species identification.

| The taxonomic status of the native blue mussel in Chile
The question of the taxonomic status of the native Chilean blue mussel goes back more than 150 years. In the present study, the taxonomic priority and also the validity of the descriptions of the two South  priority, M. platensis d'Orbigny (1846) holds for the native Atlantic smooth-shelled blue mussels, whereas M. chilensis Hupé (1854) holds for native Pacific smooth-shelled blue mussels. Interestingly, this situation is often not reflected in current taxonomic websites such as WoRMS and ITIS. Subsequently, but much later on, proteinbased (allozyme) and morphometric assessments of blue mussels led McDonald et al. (1991) to conclude that mussels from South America (both coasts), the Falkland Islands and the Kerguelen Islands should be included tentatively in M. edulis. More recently, Borsa et al. (2012), in their review of allozyme and morphometric variation of Chilean blue mussels, confirmed that the Southern Hemisphere form of M. edulis occurs "along the shores from the North Patagonia region of Chile to the southern tip of the South American continent" (p. 52, Borsa et al., 2012) and concluded that native Chilean blue mussels should be assigned subspecific rank and named M. edulis platensis d'Orbigny 1846.

CAESothers CAES WENZ
Numerous authors, using a range of different molecular markers, have reported molecular genetic differences between the native Chilean blue mussel and reference Northern Hemisphere M. edulis and/ or M. galloprovincialis (e.g., Astorga et al., 2015;Gérard et al., 2008;Śmietanka & Burzyński, 2017;Westfall & Gardner, 2010). Regardless of the taxonomic recommendation made by these (and several other) research groups, the common theme is that native Mytilus from South America, and often specifically from Chile, are genetically different from other smooth-shelled blue mussels anywhere in the world. These results challenge the interpretation of native Chilean mussels as being M. edulis-like and also of being like South American Atlantic mussels.
As our SNPs data reveal, there are pronounced nuclear genetic differences between the native blue mussels of Chile and all other mussels that we tested, including reference Northern Hemisphere M. edulis Similarly, Śmietanka and Burzyński (2017), who analysed the complete sequence of the female mitogenome of the native Chilean blue mussel, reported that the genetic distance between M. edulis and M. galloprovincialis (~2.5%) was half the distance that separates M. chilensis from either of these two species (5%), a result that indicates that the native Chilean mussel is a separate taxon within the genus Mytilus.
The use of SNPs in the present study has revealed that M. trossulus is the most differentiated of the taxa examined here, consistent with the suggestion that it is the oldest (ancestral) form (Chichvarkhin, Kartavtsev, & Kafanov, 2000;Kafanov, 1987;Vermeij, 1992)  Loci whose removal caused the assignment performance measure to drop down 0.55 in the BELS software program (Bromaghin, 2008).
T A B L E 3 Number of individuals (%) correctly reassigned to (A) species and (B) region of origin by the leave-one-out procedure implemented in GeneClass2.0 (Piry et al., 2004) using the frequency criteria of Paetkau et al. (1995) ( on the Atlantic coast of South America (e.g., Borsa et al., 2012;Hilbish et al., 2000;McDonald et al., 1991). In addition, M. trossulus alleles have been described in Chile in very low frequencies using the Me15-16 marker (Larraín et al., 2012;Oyarzún et al., 2016). We found no evidence of either M. edulis or M. trossulus in our samples from Chile using the 49 SNP panel. To address the question of the taxonomic status of native blue mussels from the Atlantic Ocean coast of South America (M. edulis, M. platensis or another species), additional studies analysing mussels from this area (i.e., Argentina, Uruguay, Falkland Islands) will be needed. Introductions of alien species are one of the most important environmental issues today (Ojaveer et al., 2015). Considering the differences between M. chilensis and the other commercial blue mussel species revealed in this study, the potential biosecurity risk posed by the anthropogenic introduction and spread of M. galloprovincialis needs further attention. Given that the Mediterranean mussel is listed amongst the hundred worst invasive species, and taking into account the dispersal capacity of this invasive mussel (Branch & Steffani, 2004;McQuaid & Phillips, 2000), for the Chilean aquaculture sector the spread of M. galloprovincialis to other mussel production area poses a threat to the native Chilean blue mussel. This has led local producers to express concerns about a negative effect on production of the native M. chilensis. In Chile, protection and control measures to avoid the introduction of marine species that constitute pests, to isolate their presence on occurrence and to prevent their spread and promote their eradication are regulated by the gen-  (Michalek et al., 2016), the recognition of the name M. chilensis to designate the native Chilean blue mussel will contribute to traceability, authenticity and compliance with seafood labelling regulations, promoting transparency in the seafood trade.

| The role of taxonomy in protecting and exploiting blue mussels
This recognition also provides an opportunity to the local mussel industry to apply for a geographically protected origin indication for the Chilean blue mussel.

| Highly informative loci
The best ranking criterion to select the most informative loci for species identification was minor allele frequency (MAF) > 0.2. This selection criterion includes 19 loci that showed a power comparable to the total panel of 49 SNPs for species identification. In practical terms, these 19 loci constitute a reduced multilocus SNP panel with high performance that permits identification of commercial Mytilus taxa without the inconsistencies associated with the use of a single or a few molecular markers. The performance of the reduced SNP panels selected with the MAF and BELS criteria, to identify mussel geographic origin region, was always lower than for species identification and shows a small increment only when F ST outlier loci were included (Table 4). This finding is probably explained by the fact that loci used to perform species identification must ideally be variable amongst species but fixed amongst populations within species. On the other hand, loci for identification of geographic origin must be more variable amongst populations within species, because they can reflect adaptation to local conditions, as demonstrated recently by Araneda et al. (2016).

| CONCLUSIONS
A new multilocus SNP marker panel reveals a clear separation between M. chilensis and all other Mytilus taxa, a degree of separation that is greater than that between M. edulis and M. galloprovincialis.
Because M. edulis and M. galloprovincialis are already considered different species, this finding provides conclusive evidence of the status of smooth-shelled blue mussels native to the Chilean coast as a species within the Mytilus genus. Taxonomic priority indicates that these mussels should be recognized as Mytilus chilensis Hupé, 1854.
The SNP markers also corroborate the previously reported presence of the highly invasive M. galloprovincialis in the Gulf of Arauco, revealing its probable recent introduction from the Mediterranean Sea.
Recognition of this invasion poses challenges to avoid environmental and economic damage in Chile. The 49 SNP panel was able to assign 99.7% of individuals correctly to species, whilst a comparable (99.4%) assignment success was obtained with a reduced panel of the 19 most informative SNP loci. These SNPs are a particularly valuable tool in terms of increasing our understanding of (i) Mytilus phylogeography, population genetics and connectivity, (ii) elucidating evolutionary processes such as natural hybridization and introgression, (iii) helping to enforce aquaculture and conservation policies, and (iv) increasing transparency in seafood labelling, authenticity and traceability field.