Temporal dynamics of genetic clines of invasive European green crab (Carcinus maenas) in eastern North America

Abstract Two genetically distinct lineages of European green crabs (Carcinus maenas) were independently introduced to eastern North America, the first in the early 19th century and the second in the late 20th century. These lineages first came into secondary contact in southeastern Nova Scotia, Canada (NS), where they hybridized, producing latitudinal genetic clines. Previous studies have documented a persistent southward shift in the clines of different marker types, consistent with existing dispersal and recruitment pathways. We evaluated current clinal structure by quantifying the distribution of lineages and fine‐scale hybridization patterns across the eastern North American range (25 locations, ~39 to 49°N) using informative single nucleotide polymorphisms (SNPs; n = 96). In addition, temporal changes in the genetic clines were evaluated using mitochondrial DNA and microsatellite loci (n = 9–11) over a 15‐year period (2000–2015). Clinal structure was consistent with prior work demonstrating the existence of both northern and southern lineages with a hybrid zone occurring between southern New Brunswick (NB) and southern NS. Extensive later generation hybrids were detected in this region and in southeastern Newfoundland. Temporal genetic analysis confirmed the southward progression of clines over time; however, the rate of this progression was slower than predicted by forecasting models, and current clines for all marker types deviated significantly from these predictions. Our results suggest that neutral and selective processes contribute to cline dynamics, and ultimately, highlight how selection, hybridization, and dispersal can collectively influence invasion success.

However, determining the mechanisms ultimately responsible for genetic clines can be challenging, as structure can be a product of both adaptation and historical vicariance.
Species invasions therefore provide a unique opportunity to investigate clines formed by contemporary interactions and observe these evolutionary processes in action. Several marine invasive species are characterized by genetic structure in their invaded range that can arise through different mechanisms such as multiple divergent source introductions, hybridization, and/or range expansion with subsequent divergence (Herborg, Weetman, Van Oosterhout, & Hänfling, 2007;Pringle et al., 2011;Richardson, Sherman, Lee, Bott, & Hirst, 2016;Roman & Darling, 2007;Saarman & Pogson, 2015). For example, in blue mussels (genus Mytilus), genetic clines have been formed through contact and introgression between invasive (M. galloprovincialis) and native (M. trossulus) mussels in the Pacific Northwest (Saarman & Pogson, 2015). Genetic clines have also been formed when invasion fronts from two separate introductions from previously allopatric lineages come into secondary contact in the invaded range, and such a scenario has occurred in the invasive European green crab (Carcinus maenas) (Darling, Tsai, Blakeslee, & Roman, 2014;Pringle et al., 2011;Roman, 2006).

Carcinus maenas is native to Europe and northern Africa and
is among the most notorious marine invasive species worldwide (Darling, Bagley, Roman, Tepolt, & Geller, 2008;Roman, 2006).

C. maenas have invaded the waters of multiple continents includ-
ing the Atlantic and Pacific coasts of North America (Darling et al., 2008;Roman, 2006), and their introductions can have significant impacts on invaded ecosystems, including damage to eelgrass beds (Garbary, Miller, Williams, & Seymour, 2013;Malyshev & Quijon, 2011), agonistic and predatory interactions with native species (Floyd & Williams, 2004;Rossong, Williams, Comeau, Mitchell, & Apaloo, 2006) and overall alterations to community structure (Lutz-Collins, Cox, & Quijón, 2016; Matheson et al., 2016). Along the coast of eastern North America, molecular data support that the current distribution of C. maenas is the result of at least two separate introductions. The initial introduction was reported in 1817 near New York, United States (Say, 1817) and likely originated from southwestern Europe (Roman, 2006). Following the initial US introduction, C. maenas expanded northward to Nova Scotia, Canada (NS) during the 1900s, but further movement appeared to halt along the eastern Scotian Shelf near Halifax, NS (~44.6°N) (Roman, 2006). A second genetically distinct introduction originating from northern Europe likely occurred in the 1980s in northeastern NS and the Gulf of St. Lawrence (Blakeslee et al., 2010;Darling et al., 2008;Roman, 2006). Over the last two decades, this northern lineage has subsequently spread through NS and New Brunswick (NB) and has made secondary contact with the original southern introduction. As a result, hybridization between the lineages has occurred in these areas resulting in latitudinal genetic clines (Darling et al., 2014;Jeffery et al., 2017a;Pringle et al., 2011). In addition, anthropogenic transport of individuals from the zone of secondary contact has resulted in a recent introduction of admixed individuals in southeastern Newfoundland (NL) (Blakeslee et al., 2010).
In this system, we take advantage of the strong genetic differentiation between lineages from the past and recent introductions, and the interaction of invasion fronts, as a platform to examine dynamics in the hybrid zone and study the temporal changes in the genetic clines.
Such investigations are possible because the clinal genetic structure of C. maenas in eastern North America has been well characterized in previous studies (Darling et al., 2014;Jeffery et al., 2017a;Pringle et al., 2011), allowing for a historical perspective and thus the ability to investigate temporal dynamics. Previous studies exploring C. maenas genetic structure between 1999 and 2007 documented a southward progression of the genetic clines (Darling et al., 2014;Pringle et al., 2011), consistent with the predominant southward circulation on the eastern Scotian Shelf (DFO, 2003;Wu, Tang, & Hannah, 2012) and the broad dispersal potential of larval C. maenas, which can remain in the water column for >50 days (Behrens Yamada et al., 2005;Klassen & Locke, 2007;Pringle et al., 2011). Considering the temporal movement of the clines, it was predicted that continued patterns of dispersal and connectivity would result in a persistent southward shift (Pringle et al., 2011). Darling et al. (2014), however, suggested that larval dispersal alone could not explain the displacement of C. maenas clines, where introgression occurred more rapidly for mitochondrial DNA relative to nuclear markers, suggesting demographic processes were influencing invasion dynamics. Indeed, predictions made by Pringle et al. (2011) do not explicitly account for demographic processes such as gene surfing (Slatkin & Excoffier, 2012) or the role of hybridization in countering Allee effects (Mesgaran et al., 2016), nor do they account for selection associated with potential adaptive differences among invasions Tepolt & Somero, 2014). While demographic processes have been implicated for changes in C. maenas genetic clines, recent evidence has suggested there might be differences in physiological thermal tolerance and behaviour between crabs from different regions (Rossong et al., 2011;Tepolt & Palumbi, 2015;Tepolt & Somero, 2014) that could influence invasion and range expansion success (Williams, Nivison, Ambrose, Dobbin, & Locke, 2015).
Here, we examine the distribution patterns of C. maenas lineages in eastern North America from New Jersey, United States, to western NL, Canada (~39 to 49°N) between 2011 and 2015 and compare to previous studies. We used single nucleotide polymorphisms (SNPs) selected to discriminate between northern and southern lineages, building directly on previous studies of population structure using SNPs (Jeffery et al., 2017a(Jeffery et al., , 2017b by expanding the scale of sampling, both in number of samples and by life stage (juveniles and adults). In addition, to evaluate the temporal progression of the clines, we incorporated genetic data collected over several time points from 2000 to 2015 from both mitochondrial and microsatellite markers. With these samples, we quantified the temporal changes in cline position and shape across different marker types, thus providing a new perspective into the evolutionary processes that shape the current clines. This work represents the most comprehensive spatial genetic census using multiple marker types in the C. maenas invasive western Atlantic range to date. Our ability to determine the current distribution and movement patterns of lineages over the last decade is critical to the ongoing management and monitoring of this invasive species as lineages could have differential impacts on ecosystems given differences found in physiology, reproduction and competitive behaviour of crabs from different regions (Best, McKenzie, & Couturier, 2017;Rossong et al., 2011;Tepolt & Palumbi, 2015;Tepolt & Somero, 2014 Notes. Juvenile crabs were collected from a subset of locations in 2015. a Indicates that some or all adult samples were previously genotyped by Jeffery et al. (2017b). b We note that these sites are located in close proximity but considered separately here.
were also collected from a subset of 15 sites in 2015 to quantify genetic differences between life stages, to obtain a current understanding of the species' population structure and clinal patterns, and potentially detect natural selection acting as one mechanism regulating the spread of lineages. The use of 15 sampling locations for both adults and juveniles also allowed us to compare changes in the genetic clines between life stages. All juvenile crabs were <25 mm in carapace width and likely represent juveniles that have already experienced one winter, as previous research found that juvenile crabs in Maine were 3-10 mm at the end of their first winter and 13-28 mm in carapace width after the second winter (Berrill, 1982 Table S1).

| SNP genotyping
Although loci were not fixed between north and south, these SNPs were capable of discriminating between the northern and southern lineage, as the mean pairwise F ST value between the southernmost and northernmost site was 0.344 (range 0.071-0.802) (see Supporting Information Figure S1B). gBlocks (Integrated DNA Technologies, Coralville, IA, USA) were designed and synthesized for use as positive controls (Richards-Hrdlicka, 2014 Pompanon, Bonin, Bellemain, and Taberlet (2005). This is consistent with the range (0%-0.2%) found in previous studies (Hess et al., 2015;Larson et al., 2013;Petrou et al., 2013). After this quality checking and prior to genetic analyses, genotypes generated from the SNP assays were merged with genotypes for the 96 SNPs from the previous RAD-seq data set using the R package genepopedit (Stanley, Jeffery, Wringe, DiBacco, & Bradbury, 2017).

| Population structure
First, to test for genetic differences between life stages in our SNP  (Table 2; see also Section 3 and Supporting Information   Figures S2, S3, S4, and Supporting Information Table S2), we pooled all samples for subsequent analyses of SNP data.
Spatial genetic structuring across the sampling range was determined using Bayesian clustering analysis in the program STRUCTURE v2.3.4 (Pritchard, Stephens, & Donnelly, 2000). STRUCTURE runs were completed through the R package parallelstructure (Besnier & Glover, 2013) to perform three independent Markov Chain Monte Carlo (MCMC) runs using 100,000 burn-in and 500,000 iterations for K = 2 (i.e., two genetic clusters). We chose K = 2 because we expected to identify two genetic clusters (derived from northern and southern lineages) with hybridization occurring in parts of the range based on previous C. maenas studies, their known invasion history and the informative SNPs chosen for our study (Blakeslee et al., 2010;Darling et al., 2008;Jeffery et al., 2017a;Pringle et al., 2011;Roman, 2006

| Hybrid assignment
To assign individuals as hybrids between the northern and southern lineage within our sample, we used the program NEWHYBRIDS v1.1 (Anderson, 2008). Following methods described for C. maenas hybrid assignment in Jeffery et al. (2017a), we tested the ability of our 96 SNPs to accurately identify hybrids and then assigned individuals to hybrid and pure classes in our data set (see Supporting Information for full details of hybrid analyses). Hybrid assignments were evaluated at two levels, where our primary analysis included three genotype classes (pure north, pure south and hybrid). For the secondary analysis, we report hybrid classes as two groups: F 1 hybrids and recombinant hybrids, with the latter class including F 2 , backcrosses and potentially later generation hybrids (e.g., F 3 and later generations). We chose to consider only a single recombinant hybrid class because these later generation hybrids are unreliably detected. Next, we validated our hybrid results using GENODIVE (Meirmans & Van Tienderen, 2004) (see Supporting Information for full details).
Hybridization patterns in the hybrid zone and the introduced hybrid region in southeastern NL were further investigated using the R package INTROGRESS (Gompert & Buerkle, 2010

| Temporal changes in the genetic clines
In addition to the informative SNPs described above, we also included two additional types of marker (mitochondrial and microsatellite loci) to provide further information about the processes shaping the genetic clines and to compare marker types, particularly because different marker types have been shown in C. maenas to differ in their population frequencies and clinal progression over time (Darling et al., 2014). In addition, these two marker types provide a historical understanding of changes in C. maenas population structure over time (Darling et al., 2014;Pringle et al., 2011;Roman, 2006 were compiled from previous studies (Blakeslee et al., 2010;Darling et al., 2014;Pringle et al., 2011;Roman, 2006 Figure   S3). The results show that ML clines generally overlapped for these four data sets (except in parts of the southern range where sampling was limited in 2015). While the clines did shift southward between 2011/2013 and 2015, we found no significant differences observed in cline centres or widths among time periods and life stages (Supporting Information Table S3). However, when all data sets were combined (the full SNP data set with all life stages and years together), cline width was significantly larger than the cline width for 2011/2013 adult data set alone, and this may be attributed to the inclusion of finer-scale sampling in the hybrid zone for later time periods. For all models, we calculated "coastal" distance from the southernmost site (Tuckerton, NJ; TKT) to each site using least-cost distance in the R package marmap (Pante & Simon-Bouhet, 2013).
Least-cost distances were estimated with depth restricted between 0 and 30 m to give a best estimate of distance along the shoreline.
However, to allow movement across greater depths, maximum depth was increased to 60 m for the site located on Grand Manan Island, NB. The same was performed for sites located beyond Sydney Harbour, NS (SYH), because some of these sites were on islands (Brudenell River, PE, and Baie de Bassin, QC), and thus, we increased maximum depth for all sites in this region to provide distances consistent with C. maenas movement. At last, for sites in western NL, maximum depth was increased to 300 m to allow distance calculations to these sites (i.e., across the Laurentian Channel). The four sites in southeastern NL within Placentia Bay (NOH, BTH and FRH) and Fortune Bay (FTB) were excluded from cline analyses, as these sites could skew our analyses because they do not represent an area of natural hybridization between lineages (Blakeslee et al., 2010) and are beyond the linear coastline of the two invasion fronts.
In hzar, clines were fit with a null model and 15 different models In addition, to quantify the temporal progression of the cline, the rate of change for the cline centre was calculated from its movement between each consecutive sample year for 2000, 2002, 2007 and 2015. These rates of change were then compared with predicted rates from a previous study modelling this from COI data (Pringle et al., 2011).  Figure S2A). Pairwise F ST values were also lower among nearby locations relative to more distant locations (Supporting Information Figure S2A). For site-specific comparisons between adults and juveniles, only two sites (Fortune Bay, NL; FTB and Hampton, NS; HMP) showed a significant genetic divergence between life stages (both p < 0.05; Table 2). The limited differences between adult and juvenile samples were further visualized using a neighbour-joining tree; adult and juvenile samples from the same site generally grouped close to each other on the tree, including FTB adults and juveniles (Supporting Information Figure S2B).  Figure S4). The only exceptions were two southeastern NL sites (Boat Harbour, NL [BTH] and FTB), which both grouped with southern sites; however, bootstrap support was low for these nodes (see Supporting Information Figure S4).  and extending eastward along the south shore of NS to East River Point (ERV) (Figure 3a). Extensive hybridization was also detected in southeastern NL (Figure 3a).

| Hybrid assignment
Individuals were also assigned maximum-likelihood hybrid indices using GENODIVE (Meirmans & Van Tienderen, 2004) where results agreed with NEWHYBRIDS assignment. Hybrid indices showed variation across most site locations, with more hybrids observed between STA and ERV, as well as in southeastern NL (Figure 3b).
Further, in southeastern NL, the limited number of pure individuals in most locations supports the introduction of admixed individuals to this region (Blakeslee et al., 2010;Jeffery et al., 2017a). The large variation in hybrid indices within many sites is consistent with the assignment of recombinant hybrids rather than first-generation hybrids in the hybrid regions.
F I G U R E 3 (a) Map of 25 European green crab (Carcinus maenas) sampling locations with the proportion of individuals assigned to pure or hybrid genotype classes by the program NEWHYBRIDS based on 96 informative SNPs. Map shows assignment to two pure (north and south) and a single hybrid class. Inset bar plots show the proportion of individuals assigned to each genotype class across all sites. The asterisk (*) denotes that all hybrids were assigned as recombinant hybrids (i.e., no first-generation hybrids) using subsequent analyses in NEWHYBRIDS.
(b) Boxplot of individual hybrid indices for sampling locations calculated using methods described by Buerkle (2005 INTROGRESS (Gompert & Buerkle, 2010) provided further support that hybrids represented recombinant hybrids; however, only a few individuals appear to show signatures reflective of potential third-generation hybrids (see Supporting Information Figure S6).
The hybrid zone in NB/NS showed similar levels of introgression as the introduced hybrid region in southeastern NL (Supporting Information Figure S6).

| Temporal changes to the genetic clines
Clines were modelled for several time periods between 2000 and 2015 for COI and microsatellite markers (Figure 4a,b) across all sites with the exception of locations in southeastern NL. Results of all maximum-likelihood (ML) cline models are provided in Table 3 and  Table 3). In addition, the cline width increased significantly between 2000 and 2015 for the COI marker (Table 3) (Table 3; Figure 4c). The estimates of cline centre for all marker types differed significantly from the predicted cline centre; however, the cline width did not differ significantly between observed and predicted clines. Therefore, observed and predicted clines are noncoincident but concordant.
Microsatellite and COI markers showed similar clines (i.e., coincident and concordant) for 2015 (see Supporting Information Figure   S7 for a map of current population structure for both markers), but the SNP Q-value data showed less of a southward shift over the range relative to these markers (Figure 4c). However, when data for the SNP panel were separated by year and life stage (see Supporting   Information Table S3 and Supporting Information Figure S3), a temporal shift in cline centre and an increase in cline width were observed over time (i.e., from adults in 2011-2013 to adults in 2015 to juveniles in 2015); however, confidence intervals overlapped indicating no significant difference between estimates (Supporting Information Table S3) although differences in sampling range could influence this interpretation because 2015 adults and juveniles did not contain samples located south of St. Andrews, NB (STA).
Furthermore, although the SNP Q-value cline showed less of a shift than COI and microsatellites, when we examined individual SNP loci (allele frequencies), many SNP clines were coincident with the 2015 COI and microsatellite cline (45% and 51%, respectively) (Supporting Information Figure S8A). However, 48% and 43% of the SNP clines were noncoincident and located further northward relative to the COI and microsatellite clines, respectively (Supporting Information Figure S8A). For cline widths, the widths of 64% and 65% of the SNP clines were concordant with COI and microsatellite clines (Supporting Information Figure S8B). Variation in ML cline centre and width for SNPs were not significantly related to

| D ISCUSS I ON
The repeated invasion of C. maenas into eastern North America provides an unprecedented opportunity to explore secondary contact as it occurs and examine the factors that promote temporal stability of genetic clines. Consistent with previous studies, clinal analyses revealed two genetically distinct groups of C. maenas corresponding to southern and northern lineages derived from two independent introductions (Blakeslee et al., 2010;Darling et al., 2014;Jeffery et al., 2017a;Pringle et al., 2011;Roman, 2006). Our results confirm the presence of extensive hybridization in the zone of secondary contact in southern Nova Scotia and the Bay of Fundy (Jeffery et al., 2017a) with potentially diminishing direct interactions of the pure lineages (i.e., no first-generation hybrids were detected). In an interesting manner, displacement of the clines was slower relative to expectations of passive dispersal (Pringle et al., 2011), which may support the hypothesis that other mechanisms, such as selection, nonrandom dispersal and/or other demographic processes, are influencing cline dynamics. Our results demonstrate the difficulties in forecasting invasion dynamics and highlight the possible interaction of multiple processes including dispersal, hybridization and selection in regulating invasion success.

| Temporal progression of the clines
Our data set provides an opportunity to explore the evolution with Darling et al. (2014), this may suggest that mechanisms other than larval dispersal alone may be contributing to temporal changes in the clines (discussed below). Instead, it may suggest the involvement of neutral processes not accounted for by Pringle et al. (2011), such as atypical hydrodynamic features that drive nonrandom larval dispersal and can thus generate steep clines in the absence of selection (Hare, Guenther, & Fagan, 2005).
In an interesting manner, we also observed differences in the speed of cline movement between the COI and microsatellite markers. Both marker types displayed a slower southwards progression than predicted by the neutral, advective dispersal model (Pringle et al., 2011), yet the progression of the microsatellite cline was less than the COI cline over time (as also observed in the 2000-2007 data in Darling et al., 2014). Discordance between nuclear and mitochondrial markers is not uncommon, although the mechanisms responsible are often difficult to ascertain and sex-biased processes, F I G U R E 4 Maximum-likelihood (ML) genetic clines for European green crab (Carcinus maenas) for different time points and genetic markers where clines were modelled using hzar (Derryberry et al., 2014) with either the southern haplotype frequency (mitochondrial marker COI) or mean admixture coefficient (Q-value; microsatellite and SNP markers) against the distance from the southernmost site in Tuckerton, New Jersey (TKT). Clines are shown with their associated fuzzy cline region (95% credible cline region) and the ML estimate of cline centre is indicated (dot) with two log-likelihood low and high estimates ( mito-nuclear incompatibility and/or adaptive introgression are frequently proposed drivers of mito-nuclear discordance (Toews & Brelsford, 2012;Wolff, Ladoukakis, Enríquez, & Dowling, 2014).
Specific to C. maenas, Darling et al. (2014) suggested that gene surfing (Currat, Ruedi, Petit, & Excoffier, 2008) and genetic "stickiness" (i.e., protection from Allee effects through hybridization) (Mesgaran et al., 2016) likely contributed to the rapid bi-directional expansion of the mitochondrial cline relative to nuclear markers, and these potential mechanisms may also explain temporal changes observed here. Further, the enhanced influence of drift due to haploidy, maternal inheritance and background selection in the absence of recombination could also result in different rates of movement for the COI marker relative to diploid markers. Despite differences between marker types, the COI marker represents the most diagnostic marker for tracking the distribution of the genetic lineages, thus providing more weight to our conclusions regarding this species' spread over time (Darling et al., 2014;Pringle et al., 2011;Roman, 2006). Our study thus provides a rare, long-term perspective of species invasion distribution and spread over time, from its inception to the present.

| Potential mechanisms contributing to current cline dynamics
All marker types resolved a common spatial genetic pattern, where current clines for COI, microsatellites, and SNP allele frequencies (mean overall loci and many individual SNPs) overlapped in cline centre and width, and the widths of these clines were also concordant with that of the SNP cline based on Q-value. However, current clines generated for all marker types did not coincide with previous predictions potentially resulting from asymmetric connectivity and/or limitations of previous forecasting models (Pringle et al., 2011). Instead, deviations from predictions may implicate selective processes in shaping the current clines where cline dynamics may be influenced by physiological differences related to environmental conditions such as temperature Tepolt & Somero, 2014). Latitudinal clinal patterns exist in the native range of C. maenas (Roman & Palumbi, 2004), and differences in thermal tolerance (i.e., cardiac threshold) exist between crabs from northern and southern parts of the native range, with higher latitude populations being less tolerant of warmer conditions and vice versa (Tepolt & Somero, 2014 similar thermal differences exist between parts of the invaded range, it is unclear whether these physiological differences reflect adaptive divergence or genomewide divergence carried over from the native range that has been maintained through neutral processes (Tepolt & Somero, 2014). Recent work suggests associations between genetic structuring and winter sea surface temperature  that are the result of recent secondary contact that has resulted in a hybrid zone that coincides with a temperature gradient in the Atlantic Ocean (Bierne, Welch, Loire, Bonhomme, & David, 2011) (see Figure 5). In this case, the coupling of endogenous (genetic incompatibilities) and exogenous (environmental) barriers may drive signals of local adaptation when genetic incompatibilities may be more likely (Bierne et al., 2011). Here, we cannot determine to what extent this coupling hypothesis (sensu Bierne et al., 2011) influences the C. maenas clines, although genetic incompatibilities between lineages appear minimal. In fact, the temporal broadening of the cline widths across marker types suggests that selection against hybrids is weak (Sotka & Palumbi, 2006) and this is evident by large numbers of later generation hybrids in our data set. Further, the presence of an apparently stable admixture zone in southeastern NL suggests the absence of strong selection effects against hybrids as well as limited selection related to thermal tolerance in these colder waters.
Indeed, patterns of hybridization in the zone of secondary contact may provide insight into the mechanisms that may be operating to slow the progression of the clines relative to predictions (Pringle et al., 2011

| Future of the green crab clines
Our study indicates that the genetic clines of C. maenas have progressed southward slower than predicted by Pringle et al. (2011), perhaps implicating processes such as environmentally associated selection, later generation hybrid advantage, nonrandom larval dispersal and demographic processes in shaping cline dynamics in this system. Experimental studies are needed to resolve the role of these mechanisms in influencing C. maenas clines. Moreover, future cline dynamics may be further altered by human-mediated global change (Taylor, Larson, & Harrison, 2015). For example, climate change associated ocean warming and acidification could lead to range shifts and expansions in C. maenas lineages (Compton, Leathwick, & Inglis, 2010;Gibson, Atkinson, Gordon, Smith, & Hughes, 2011). Further, recent predictions from Stanley et al. (2018) suggest that the genetic cline centres of multiple species in the Northwest Atlantic, including C. maenas, will shift northward under future climate scenarios with both lineages of C. maenas experiencing increases in the extent of suitable habitat. In addition, the anthropogenic transport of individuals via ship traffic and other anthropogenic vectors (e.g., see Blakeslee et al., 2010;Cohen, Carlton, & Fountain, 1995;Fowler et al., 2016) could also influence the future distribution and population structure of C. maenas. It also remains unclear how C. maenas distributions could be influenced by competition with the more recent invader the Asian shore crab (Hemigrapsus sanguineus) (Lord & Williams, 2017). Altogether, several factors may contribute to ongoing changes in spatial genetic structure, and given the reported differences in behaviour, physiology and reproduction between C. maenas from different regions (Best et al., 2017;Rossong et al., 2011;Tepolt & Somero, 2014), temporal genetic sampling over 5-to 10-year intervals should be implemented to monitor these potential changes and help facilitate more appropriate management strategies.

| CON CLUS IONS
In our study, multiple genetic marker types suggest a slowed southward movement of C. maenas genetic clines over time counter to F I G U R E 5 Maximum-likelihood (ML) genetic clines for European green crab (Carcinus maenas) for 96 single nucleotide polymorphisms (SNPs) where clines (black lines) were modelled using hzar (Derryberry et al., 2014) with allele frequency against the distance from Tuckerton, New Jersey (TKT) for each site. The mean ML estimate of cline centre is indicated by the red line. Mean winter sea surface temperature (blue line) from Jeffery et al. (2018) was added on an inverse scale on the right-hand axis for comparison against cline models. Clines for each SNP were weighted based on their effective number of alleles as described in the Section 2 Winter sea surface temperature (°C) SNPs (n = 96) Mean cline centre Temperature dispersal-based predictions. It is likely that both selective and neutral processes shape the current clines, and future work using complex realistic models of dispersal and experimental work on quantifying the competitive ability and physiology of hybrid and genetically pure individuals are necessary to better understand the processes restricting the range limits of each lineage and help predict future dynamics of the cline. In addition, climate change and human-mediated transport may lead to continued alterations and range shifts in C. maenas distributions. Our study highlights the current challenges and complexities in forecasting invasion dynamics and emphasizes the importance of temporal monitoring using genetic markers to understand the spread of invasive species and inform management decisions.

ACK N OWLED G EM ENTS
The authors wish to thank staff at Fisheries and Oceans Canada for their assistance with sample collection, in particular Jennifer

CO N FLI C T O F I NTE R E S T
None declared.

DATA A RCH I V I N G
Data available from the Dryad Digital Repository: https://doi. org/10.5061/dryad.130vb65.