Secondary contacts and genetic admixture shape colonization by an amphiatlantic epibenthic invertebrate

Abstract Research on the genetics of invasive species often focuses on patterns of genetic diversity and population structure within the introduced range. However, a growing body of literature is demonstrating the need to study how native genotypes affect both ecological and evolutionary mechanisms within the introduced range. Here, we used genotyping‐by‐sequencing to study both native and introduced ranges of the amphiatlantic marine invertebrate Ciona intestinalis. A previous study using microsatellites analysed samples collected along the Swedish west coast and showed the presence of genetically distinct lineages in deep and shallow waters. Using 1,653 single nucleotide polymorphisms (SNPs) from newly collected samples (285 individuals), we first confirmed the presence of this depth‐defined genomic divergence along the Swedish coast. We then used approximate Bayesian computation to infer the historical relationship among sites from the North Sea, the English Channel and the northwest Atlantic and found evidence of ancestral divergence between individuals from deep waters off Sweden and individuals from the English Channel. This divergence was followed by a secondary contact that led to a genetic admixture between the ancestral populations (i.e., deep Sweden and English Channel), which originated the genotypes found in shallow Sweden. We then revealed that the colonization of C. intestinalis in the northwest Atlantic was as a result of an admixture between shallow Sweden and the English Channel genotypes across the introduced range. Our results showed the presence of both past and recent genetic admixture events that together may have promoted the successful colonizations of C. intestinalis. Our study suggests that secondary contacts potentially reshape the evolutionary trajectories of invasive species through the promotion of intraspecific hybridization and by altering both colonization patterns and their ecological effects in the introduced range.

Despite a great deal of recent research on invasion genetics Bourne, Hudson, Holman, & Rius, 2018;Rius et al., 2015), there remains a dearth of studies investigating how genetic patterns in the native range influence the introduced range.
Biological invasions act as unique experiments in evolution (Yoshida, Goka, Ishihama, Ishihara, & Kudo, 2007), allowing observations of how NIS spread and adapt to novel environments on a human timescale. The genetic study of NIS furthers our understanding on how contemporary gene flow and local adaptation contribute to colonization success (Verhoeven et al., 2011). In addition, studies of NIS have shown that genetic admixture of divergent lineages can affect fitness of colonizing populations through transgressive segregation (Johansen-Morris & Latta, 2006;Wagner, Ochocki, Crawford, Compagnoni, & Miller, 2017), by masking deleterious mutations (Keller & Waller, 2002), and/or by increasing standing genetic variation on which selection can act (Rius & Darling, 2014). Genetic admixture can also disrupt locally adapted gene pools, which may negatively affect colonization success (Gilk et al., 2004). Therefore, understanding how ecological and evolutionary mechanisms influence colonization success is key for unravelling how genetic patterns found in native and introduced ranges relate. Research progress on the evolutionary effects of NIS has largely been dominated by studies conducted in terrestrial ecosystems (Abbott, 1992;Rius & Darling, 2014), with considerably less effort being devoted to study aquatic organisms.
As such, they are relevant and unique models for studying colonization success in marine ecosystems (Zhan, Briski, Bock, Ghabooli, & MacIsaac, 2015). Ciona intestinalis is a solitary ascidian with a disjunct amphiatlantic (i.e., inhabiting both sides of the Atlantic) distribution throughout the North Atlantic Ocean (Bouchemousse, Bishop, & Viard, 2016a). It is generally accepted that the northeast Atlantic coastline is its native range (Bellas, Beiras, & Vázquez, 2003;Bouchemousse, Bishop, et al., 2016a;Gulliksen & Skjaeveland, 1973;Nydam et al., 2017), while the introduced range includes the northwest Atlantic coastline (Bouchemousse, Bishop, et al., 2016a;Nydam & Harrison, 2007). As with all solitary ascidians, C. intestinalis is hermaphroditic and reproduces through broadcast spawning, with external fertilization. The short-lived pelagic larval stage normally lasts <24 hr, though this stage can be extended to five days (Petersen & Svane, 1995). Larvae of C. intestinalis are often retained close to the adults and the production of adhesive mucus strings together with the eggs (Svane & Havenhand, 1993) may result in lower dispersal potential. Consequently, transcontinental dispersal of C. intestinalis is attributed to anthropogenic transport or rafting events of individuals only. Ciona intestinalis shows a high affinity for marine infrastructures (e.g., pontoons and ropes in harbours and marinas), which are known to concentrate NIS (Aldred & Clare, 2014). This propensity to foul can lead to negative economic and ecological impacts when this species is found in aquaculture facilities (Fitridge et al., 2012;Lutz-Collins, Ramsay, Quijón, & Davidson, 2009;Rius et al., 2011). Consequently, most research studying the extensive distribution of C. intestinalis has been performed considering individuals found on artificial structures (e.g., Bouchemousse, Bishop, et al., 2016a;Bouchemousse, Liautard-Haag, Bierne, & Viard, 2016c;Hudson et al., 2016;Zhan, Macisaac, & Cristescu, 2010). This has led to a good understanding of the distribution of C. intestinalis on artificial structures, but there is still limited knowledge of the relative importance of natural and artificial habitats for the spread and establishment of this species in new areas.
The west coast of Sweden is a coastline where C. intestinalis is present on natural substrata from the surface to depths of more than 100m (Dybern, 1965(Dybern, , 1967Svane & Havenhand, 1993). There, the opening of the brackish waters of the Baltic Sea to the Atlantic means that individuals inhabiting shallow water experience a wide range of salinities (10-30 PSU) and variable temperatures (~0-20°C, Dybern, 1965;Renborg, Johannesson, & Havenhand, 2014), whereas individuals at depth live in both more constant temperatures and stable, high salinities (~34 PSU). The difference in density between surface and deeper waters leads to a strong pycnocline separating the less saline surface water of the Baltic Sea from the high salinity bottom water (often more than ~10-15 m in depth) from the Atlantic (Johannesson et al., 2018). There are observable differences in the biology and life history of individual C. intestinalis found in different depths. For example, individuals inhabiting shallow waters (<15 m) have two generations a year (each spawning period lasting a couple of months) during boreal spring and late summer, whereas deeper individuals (>15 m) have one generation per year, with spawning lasting approximately one month during boreal summer (Dybern, 1965).
Additionally, there appears to be slight morphological variation across depths, with shallow individuals being smaller and more heavily pigmented than deeper individuals (Dybern, 1965;Svane, 1983). This may be due to genetically driven phenotypic variation. A recent study using microsatellites showed that the deep and shallow water populations of C. intestinalis along the Swedish west coast are genetically differentiated (Johannesson et al., 2018). Strong pycnoclines can act as distinct barriers to vertical movement of larvae within the water column (e.g., Gallager, Davis, Epstein, Solow, & Beardsley, 1996), and the existence of genetically distinct populations has tentatively been ascribed to the pycnocline present at ~10-15 m acting as a barrier to reproductive exchange. In addition, local adaptation may contribute to the genetic differences between shallow and deep populations as they are exposed to different conditions, including salinity, temperature, food availability and light. Thus, two distinct populations of C. intestinalis separated by an abiotic barrier have evolved along the Swedish west coast.
Here, we used C. intestinalis as a model organism to investigate how understanding genetic variability in the native range can help elucidate mechanisms shaping both colonization success and introduction pathways in new ranges. The objectives of the study were to (a) identify fine-and broad-scale population genomic patterns of C. intestinalis, (b) reveal evolutionary relationships among individuals collected along coastlines across the range of the species, (c) determine the presence or not of genetic admixture and (d) if admixture was present, infer if it could be associated with successful colonization of novel habitats. We hypothesized that the colonization success of C. intestinalis across its introduced range has been affected by the historic divergence of ancestral genotypes, the levels of genetic admixture between divergent lineages, and the intensity of gene flow between native and introduced ranges.

| Field sampling
Tissue samples of 285 C. intestinalis were collected from 20 sites within the putative native and introduced range of the species (Table 1 and Figure 1). Samples from Sweden were collected from shallow natural, deep natural, and shallow artificial substrata (see details in Table 1), whereas sites outside Sweden were all from shallow artificial substrata. Individuals from natural substrata were sampled by either snorkelling, SCUBA diving or a remotely operated underwater vehicle. Artificial substrata were sampled in marinas by pulling up hanging ropes, submerged buoys and checking the undersides of pontoons. We attempted to leave a distance of at least one metre between each sampled individual to limit the chance of collecting closely related individuals. Once collected, tissue was immediately preserved in 95% ethanol which was periodically replaced until tissue pigment no longer leached into the ethanol. Finally, tissue samples were stored at −20°C until DNA extraction.

| DNA extraction and genotyping
DNA was extracted from preserved tissue using the Qiagen DNeasy ® Tissue Kit (Qiagen) according to the manufacturer's protocol. Gel electrophoresis and the QuantiFluor ® dsDNA System (Promega) were used to assess quality and quantity of extracted DNA, respectively. DNA was shipped to the University of Wisconsin Biotechnology Center where it was genotyped using the genotypingby-sequencing methodology (GBS; Elshire et al., 2011). Briefly, GBS reduces the complexity of the sample genome by digesting the DNA using methylation-sensitive restriction enzymes and sequencing the ends of the digested fragments using barcoded adapter regions.

| Analysis of genotyping-by-sequencing data
The GBS assembly was performed using ipyrad v. 0.7.28 (Eaton, 2014), a toolbox for assembly and analysis of restriction site-associated DNA sequencing (RAD-seq) type genomic data sets. We followed the seven sequential assembly steps of ipyrad using parameters based on those recommended for single-end GBS data in the ipyrad documentation (http://ipyrad.readt hedocs.io/). As the C. intestinalis genome is not yet available, we used the de novo assembly method, which requires no prior genomic resources and used ipyrad to trim Illumina adapter reads. As we were working with only one species, we set the level of sequence similarity for clustering to be 90% (I. Overcast, pers. comm.). Following the iterative filtering framework outlined by O'Leary, Puritz, Willis, Hollenbeck, and Portnoy (2018), we used vcftools v.0.1.13 (Danecek et al., 2011) to first filter for loci with a minimum single nucleotide polymorphism (SNP) call quality of 20, a minimum genotype depth of less than five, and a mean minimum depth (across individuals) of <15. Additionally, we chose to remove loci with a minor allele count of less than three, rather than the commonly used minor allele frequency threshold of 5%, because the latter will remove true rare alleles that are important in elucidating fine-scale structure and accurately drawing inference of past demographic events (O'Connor et al., 2015). We then iteratively increased our stringency for allowing missing data (on both loci and individuals separately), so that our final dataset contained loci with at least 50% call rate (i.e., a locus must be present in at least 50% of individuals), and up to 50% allowed missing data per individual. To remove the confounding effects of linkage disequilibrium, we used vcftools to thin markers so that only one SNP per locus was retained in our dataset.
We used BayeScan v.2.0 (Foll & Gaggiotti, 2008) to ensure our dataset contained only putatively neutral loci. BayeScan uses differences in allele frequency between populations to identify candidate loci under natural selection (Foll & Gaggiotti, 2008) and was run using a thinning interval size of 20, with 25 pilot runs of length 10,000 and a burn-in length of 50,000. Prior odds for the neutral model were set to 100 rather than the default 10, to reduce the number of false positives in large datasets (>1,000 SNPs). While commonly used, BayeScan has often been shown to report false positives especially in species undergoing range expansions, while also assuming equal population exchange and evolutionary independence among all populations (Bierne, Roze, & Welch, 2013;Whitlock & Lotterhos, 2015).
We therefore also assessed for candidate loci using two newer meth- did not recover any F ST outlier loci, so we continued our analyses with other software. Regarding the pcadapt, it ascertains population structure using principal component analysis (PCA) to find candidate loci excessively related to population structure. We classified loci that were recovered by both BayeScan and pcadapt as putatively under natural selection and removed them for the following analyses. Finally, we created a more conservative dataset that excluded all loci recovered by BayeScan and pcadapt and ran the whole set of analyses again (see Appendix A within Supporting Information).

| Population structure
We used the software ADMIXTURE v.1.3 (Alexander, Novembre, & Lange, 2009) to estimate the likelihood that an individual comes from one of a certain number of putative sample populations (K).
Like STRUCTURE v.2.3.4 (Pritchard, Stephens, & Donnelly, 2000), ADMIXTURE uses a maximum-likelihood estimation from multilocus SNP genotype datasets, but calculates estimates using a faster numerical optimization algorithm. We performed a discriminant analysis of principal components (DAPC) to visualize between-population genomic variation (Jombart, Devillard, & Balloux, 2010). DAPC transforms the data using PCA before using PCA factors as variables for a discriminant analysis (DA), ultimately maximizing the differences among groups while minimizing variation within groups (Jombart et al., 2010). We used the package adegenet v.2.1.1 (Jombart, 2008) for R (R Development Core Team, 2017) to perform the DAPC. We ran the DAPC with and without a priori knowledge of individual populations. We examined pairwise population genetic differentiation using F ST values and their p values by running 10,000 permutations with Arlequin v.3.5 (Excoffier & Lischer, 2010). We also used Arlequin to measure the inbreeding coefficient F IS and expected heterozygosity (H E ) per population. Finally, we ran an analysis of molecular variance (AMOVA) test using site clusters as inferred by ADMIXTURE and

| Reconstructing invasion routes
To obtain relevant and detailed information and infer the historical relationship among genotypes of C. intestinalis throughout its range,  Figures S3a,b, respectively). Following the results of this initial analysis, we then added in data from Canada sites to infer the colonization history along the introduced range (scenario set 3, Figure S3c). As specific population sizes, divergence times and potential admixture rates were unknown, we used a uniform distribution with a large interval (population sizes and divergence times: 10-10 7 ; admixture rates: 0.001-0.999;  (Cornuet et al., 2014). We used a logistic regression on the 1% simulated datasets that were closest to the observed dataset (using Euclidean distances between simulated and observed datasets) to calculate the posterior probability of each scenario. This approach produces 95% confidence intervals for each scenario's posterior probability, with the most likely scenario defined as the highest estimate without overlapping confidence intervals (Cornuet et al., 2008). For the most probable scenario of scenario set 3 ( Figure S3c), we calculated type I (the probability with which this scenario is rejected although it is the true scenario) and type II (the probability of choosing this scenario when simulating data according to other scenarios) error rates. Finally, we assessed the goodness of fit for the final chosen scenario by implementing the model check- ing feature of DIYABC. We simulated 1,000 datasets using posterior distribution values and compared these with the observed dataset by considering different summary statistics than were used during the generation of the reference table, and visualized this using a PCA (Cornuet et al., 2014).

| Loci assembly and detection of outlier loci
GBS generated a total of 530,157,826 raw reads, with an average of 2,000,596 reads per sample. After filtering and clustering using ipyrad and vcftools, we retained a total of 1,667 putatively unlinked SNPs in the sequence assembly. Twenty individuals were removed from the dataset due to missing data (i.e., >50% missing data), which was likely caused by poor DNA quality or secondary contaminants within the samples (Federman, Donoghue, Daly, & Eaton, 2018  from Shelburne and Sydney. Therefore, the total final number of sites was 18. BayeScan and pcadapt recovered a total of 30 and 61 F ST outlier loci, respectively, of which 14 were found by both software, whereas OutFlank recovered no putative loci under selection. We subsequently removed the 14 loci found in both BayeScan and pcadapt from our analyses, leaving a dataset of 1,653 SNPs. We also performed all analyses on a new dataset that excluded all F ST outlier loci recovered, irrespective of program used (77 loci in total, see Appendix A within the Supporting Information for details).

| Heterozygosity and population structure
Values of F IS ranged from 0.017 to 0.139 (Table 1). Nine sites showed no signs of deviation from Hardy-Weinberg equilibrium, but nine of the sites exhibited significant positive F IS values indicating a deficiency of heterozygotes in these sites (Table 1). Expected heterozygosity ranged from 0.202 to 0.286 (Table 1), with no noticeable differences in genetic diversity between geographical regions (Table 1).  Table S1).

The combination of ADMIXTURE
The AMOVA test using site clusters as inferred by ADMIXTURE and DAPC plots showed that genetic differentiation was significant among groups, among sites within groups and within sites (Tables   S2,S3). The AMOVA test performed using only shallow Sweden sites to test whether there was an effect of substratum (natural vs. artificial) found no significant genetic differentiation between these two groups (Table S4).

| Reconstructing invasion routes
For all of our ABC analyses, our check of priors showed a good match between simulate datasets and the observed data ( Figure S2).
We firstly found that within our northeast Atlantic sampling sites, the ancestral population diverged and formed the deep Sweden and England, Jersey, France groups ( Figure S3a (Table   S7, Figure S4), suggesting that even though this is the most strongly supported scenario, there is some discordance between the scenario posterior combinations and the observed dataset.

| D ISCUSS I ON
Our results showed high levels of genomic differentiation between the main regions of the northeast Atlantic (i.e., English Channel and North Sea) and identified the presence of historical genetic admixture among individuals from these regions. This seems to have resulted in genotypically and phenotypically distinctive individuals that are currently found in shallow sites in Sweden. In addition, we revealed genomic patterns suggesting secondary contacts and postulate that this may have promoted intraspecific hybridization.
Our result supported the presence of genetic admixture during the spread to and colonization of the northwest Atlantic. More specifically, we found evidence of genetic admixture between genotypes from the English Chanel and genotypes from the shallow North Sea.
While we found here no direct evidence that intraspecific hybridization influences colonization, our results indicate that this may be a possible mechanism promoting successful colonization of sites with new environmental conditions, such as trans-oceanic introductions.
This builds on a growing number of studies showing that the mixing of divergent genotypes as a result of human mediated transport of species has the potential to fundamentally alter colonization patterns and to unprecedentedly alter ecological and evolutionary patterns (Bouchemousse, Liautard-Haag, et al., 2016c;Mooney & Cleland, 2001;Pineda, López-Legentil, & Turon, 2011).
The presence of high genetic subdivision among genotypes found in deep sites off Sweden and in England, Jersey and France suggests that individuals found in these sites represent native populations ( Figure 3). This is supported by the ABC analyses, which indicated an initial divergence between these two groups ( Figure 5). This accords with the expectation that native ranges will show a highly defined population structure, often involving two main groups of ancestral genotypes (Boubou et al., 2012;Reusch, Bolte, Sparwel, Moss, & Javidpour, 2010;Rius, Clusella-Trullas, et al., 2014a). Divergence of these C. intestinalis populations may reflect adaptation to differing local conditions and/or earlier periods of allopatric isolation leading to the generation of genetic divergence through selection or genetic drift. Previous research has shown that high plasticity in C. intestinalis allows acclimatization of deep water individuals to shallow water salinities (Renborg et al., 2014), which suggests that local adaptation and primary divergence are less likely. Rather, it seems more likely that during the last glacial maximum deep sites off Sweden and sites in England, Jersey and France were isolated as separate glacial refugia, leading to the divergence that we see today. ABC analyses suggest that secondary contacts leading to genetic admixture between the England, Jersey, France and the deep Sweden genotypes formed the genotypes found in shallow Sweden and Denmark. Thus, either historic artificial transport or postglacial expansion may have promoted such secondary contacts, as reported for other marine invertebrates (Pérez, Nirchio, Alfonsi, & Muñoz, 2006). A similar situation has been identified with the ascidian Pyura chilensis in the southwest Pacific Ocean, where historical divergence of P. chilensis populations occurred due to isolation associated with glacial periods. This was followed by secondary contacts and genetic admixture between these previously isolated populations (Haye & Muñoz-Herrera, 2013). Our results suggest that genetic admixture may have had fitness effects that enabled C. intestinalis to expand to previously uninhabitable substrata and conditions within its native range. The ability of individuals from shallow Sweden populations to survive relatively high temporal variability in environmental conditions such as temperature and salinity compared to individuals found in England, Jersey, France and deep Sweden may be explained by the fitness benefits of genetic admixture (Wagner et al., 2017), allowing survival in the face of strong selective pressures (Verhoeven et al., 2011).
In contrast to the Sweden populations, individuals from England, Jersey and France formed a relatively homogeneous genetic cluster in both ADMIXTURE and DAPC analyses (Figures 2-4). Earlier work showed that samples from these locations were subdivided into two genetic groups (Hudson et al., 2016), but this differentiation was weaker than what we found between samples from deep Sweden and those from England, Jersey and France. The native range of C. intestinalis has been previously described as the northeast Atlantic (e.g., Bouchemousse, Bishop, et al., 2016a;Hudson et al., 2016), and here, we show that this range comprises most of the genomic differentiation among populations, with more complex demographic histories among populations along the northeast coast of the Atlantic than the northwest coast. In line with previous studies that identified admixture within native ranges (Gillis, Walters, Fernandes, & Hoffman, F I G U R E 2 ADMIXTURE plots representing all sampled populations of Ciona intestinalis. The main regions are highlighted above. The different colours represent putative genetic clusters with K ranging from 2 to 7, with K = 4 being found to be the most optimal value ( Figure  S1). Sampling site names are abbreviated as in Table 1 K = 2

F IS D E N H P L T N Q J E R S T M C A N _ 2 C A N _ 1
Deep Sweden Shallow Sweden D e n m a r k E n g l a n d J e r s e y F r a n c e C a n a d a 2009; Rius, Turon, Ordóñez, & Pascual, 2012), our findings suggest that historic artificial transport may have facilitated the admixture of the genotypes from deep Sweden and the English Channel.
It is well established that the recent spread of C. intestinalis has been promoted by the proliferation of man-made structures along coastlines that act both as stationary substrata and as vectors (Clarke Murray et al., 2014). Throughout the study area, this species is widespread in harbours and marinas in much of its current distribution (Bouchemousse, Bishop, et al., 2016a;Hudson et al., 2016), but rare on natural substrata except along the Swedish coast (Johannesson et al., 2018). This raises the issue of the origin and evolutionary background of the many populations living on artificial substrata that may represent either extensions of large natural populations or completely new introduced populations. In the sampled shallow Sweden sites, comparison of genetic differentiation between natural and man-made sites showed no significant differences (Table S4), with individuals sampled on artificial substrata being generally more closely related to nearby shallow natural sites than to individuals from other artificial sites along the Swedish coast ( Figure 4; Table   S1). This suggests that the nature of natural and artificial substrata does not, in itself, create a barrier to local gene flow. In shallow waters off Sweden, the sampled natural substratum included seagrass beds where C. intestinalis lives at modest densities attached to blades of Zostera marina. Below the pycnocline, at depths of 20 m or more, dense populations occur on the natural vertical rock walls (see also Svane & Havenhand, 1993), whereas in the English Channel and south-western North Sea area, very few individuals have been documented on natural substrata. To our knowledge, there are no reports of C. intestinalis inhabiting natural substrata in the English Channel. However, small numbers have been recovered during F I G U R E 3 (a) Discriminant analysis of principal components using unlinked loci with no a priori population information. The first axis explains 58.2% of the variation, and the second axis explains 41.8%. (b) Discriminant analysis of principal components using unlinked loci with a priori population information. The first axis explains 23.1% of total variation, and the second axis explains 18.1%. Sampling site names are abbreviated as in Table 1 dredging estuaries in the English Channel (authors pers. obs.) and the south-western North Sea (Rees, Waldock, Matthiessen, & Pendle, 2001). Such low densities may be due to the effects of predation on different life-history stages as seen for closely related species in other parts of the world (Dumont, Gaymer, & Thiel, 2011;Rius, Potter, Aguirre, & Stachowicz, 2014b).
Our study corroborated the findings of Johannesson et al. (2018) by identifying strong genetic differentiation between shallow and deep populations of C. intestinalis along the west coast of Sweden.
While this genomic differentiation among populations appears surprising as some deep and shallow sites are geographically close to one another, this can be explained due to the effect of the aforementioned pycnocline promoting depth-defined divergence that has also been observed in corals (Prada & Hellberg, 2013). Admixture between deep Sweden and England, Jersey, France genotypes, as indicated by our ABC analysis, suggests that the pycnocline may not have always been the impenetrable barrier to gene flow as currently observed (Johannesson et al., 2018). Taken together, our results suggest that although historically the pycnocline may have allowed the mixing of divergent genotypes, it currently provides a strong barrier to gene flow, maintaining contemporary genomic differentiation between deep and shallow sites.
The northwest Atlantic range of C. intestinalis is restricted to the east coast of North America. It has been documented in eastern Canada since at least the mid-1800s (Carver, Mallet, & Vercaemer, 2006), but its population size and range have only recently expanded (Ramsay, Davidson, Landry, & Arsenault, 2008). Our ABC analyses suggest that the origin of the Canadian sites was due to secondary contact between populations from England, Jersey, France and shallow Sweden. This is supported by ADMIXTURE analysis, which indicates high similarity between the individuals from Canada and the ones from both the English Chanel and shallow sites within the North Sea (Figure 2). In addition, the DAPC analyses indicated that Canadian individuals were similar to individuals found in shallow Swedish waters (Figure 3).  selection against heterozygotes as we excluded loci putatively under selection, and we can also exclude selfing as self-fertilization success is generally low in C. intestinalis (Bouchemousse, Lévêque, Dubois, & Viard, 2016b;Byrd & Lambert, 2000). A more likely explanation is a Wahlund effect (Wahlund, 1928), a reduction of expected heterozygosity due to mixing of two genetically differentiated populations, which has been reported in other studies (Dupont, Viard, Dowell, Wood, & Bishop, 2009;Marescaux et al., 2015) including studies of Ciona spp. (Hudson et al., 2016;Zhan et al., 2010).
The history of the introduction of C. intestinalis to the western Atlantic coast is complex, starting with historical divergence in the native range involving two groups (England, Jersey, France and deep Sweden lineages, Figure 5), which was likely due to genetic drift during a period of isolation (allopatry) in different glacial refugia. More recent historic gene flow between these populations appears to have led to the formation of the admixed genotypes found in shallow Sweden and Denmark sites. Finally, the Canada specimens originated from secondary contacts between individuals from these sites and individuals from the western North Sea and English Channel. Our findings suggest that admixture between genetically diverse native genotypes preceded successful trans-oceanic colonization, in line with previous studies showing that genetic admixture facilitates the colonization of new habitats (Abbott, Barton, & Good, 2016). We suggest that artificial transport of species facilitates secondary contacts and intraspecific admixture among divergent native genotypes, strongly altering NIS evolutionary trajectories and influencing their ecological impacts within the introduced range.

ACK N OWLED G EM ENTS
We would like to thank the guest editors who kindly invited us to contribute to this special issue and also the anonymous reviewers whose comments and suggestions greatly improved the manuscript.
We would like to thank the divers at Tjärnö Marine Laboratory and Kristineberg Marine Station, and Dr Ellen Schagerström, for their help in collecting samples. We would also like to extend our gratitude to Dr Anna-Karin Ring for her assistance in the laboratory.
Finally, we would like to thank Steven Bourne and Rachael Graham for guidance with the bioinformatic analyses. We acknowledge the use of the IRIDIS High Performance Computing Facility, and associated support services at the University of Southampton, in the completion of this work. We would also like to thank Prof Aibin Zhan,