Are genes faster than crabs? Mitochondrial introgression exceeds larval dispersal during population expansion of the invasive crab Carcinus maenas

Biological invasions offer unique opportunities to investigate evolutionary dynamics at the peripheries of expanding populations. Here, we examine genetic patterns associated with admixture between two distinct invasive lineages of the European green crab, Carcinus maenas L., independently introduced to the northwest Atlantic. Previous investigations based on mitochondrial DNA sequences demonstrated that larval dispersal driven by advective currents could explain observed southward displacement of an admixture zone between the two invasions. Comparison of published mitochondrial results with new nuclear data from nine microsatellite loci, however, reveals striking discordance in their introgression patterns. Specifically, introgression of mitochondrial genomes relative to nuclear background suggests that demographic processes such as sex-biased reproductive dynamics and population size imbalances—and not solely larval dispersal—play an important role in driving the evolution of the genetic cline. In particular, the unpredicted introgression of mitochondrial alleles against the direction of mean larval dispersal in the region is consistent with recent models invoking similar demographic processes to explain movements of genes into invading populations. These observations have important implications for understanding historical shifts in C. maenas range limits, and more generally for inferences of larval dispersal based on genetic data.


Summary
Biological invasions offer unique opportunities to investigate evolutionary dynamics at the peripheries of expanding populations. Here, we examine genetic patterns associated with admixture between two distinct invasive lineages of the European green crab, Carcinus maenas L., independently introduced to the northwest Atlantic. Previous investigations based on mitochondrial DNA sequences demonstrated that larval dispersal driven by advective currents could explain observed southward displacement of an admixture zone between the two invasions. Comparison of published mitochondrial results with new nuclear data from nine microsatellite loci, however, reveals striking discordance in their introgression patterns. Specifically, introgression of mitochondrial genomes relative to nuclear background suggests that demographic processes such as sexbiased reproductive dynamics and population size imbalancesand not solely larval dispersal-play an important role in driving the evolution of the genetic cline. In particular, the unpredicted introgression of mitochondrial alleles against the direction of mean larval dispersal in the region is consistent with recent models invoking similar demographic processes  For the purposes of genetic diversity analyses, sampling sites (green) have been divided into southern (hexagons), northern (circles) and central (squares) regions; the latter corresponds to the primary admixture zone. Full locality names and additional details can be found in table 1. Arrows show the spread of the two mitochondrial haplotypes (red = southern; blue = northern) through time. Large circles give the likely sites where C. maenas was first introduced. Adapted from [19][20][21]. Arrows are meant to indicate geographical extent of haplotype clusters only, and not actual movement of haplotypes; for instance, simultaneous arrival of southern and northern haplotype clusters to Newfoundland in 2007 has been ascribed to anthropogenic translocation from the admixture zone near Halifax [20].
Cma14EPA and Cma16EPA were amplified from genomic DNA using previously described molecular protocols [17,23]. Samples that failed to amplify at three or more loci after two attempts were dropped from the dataset.

Genetic analysis
Standard genetic diversity indices were generated for all samples using ARLEQUIN v. 3.0 [24] for mitochondrial data and FSTAT v. 2.9.3.2 [25] for microsatellite data. Significant differences (α = 0.05) in mean diversity between southern, central and northern sample regions were determined using the Tukey-Kramer method to account for unequal sample sizes, implemented in the R statistical programming environment v. 3.0.2 (http://www.R-project.org). Regional boundaries were drawn in the south between Lubec, ME (L) and Chance Harbor, NB (CH) and in the north between Chester, NS (C) and Halifax, NS (H) (figure 1). These regions correspond roughly to the southern and northern limits of the observed nuclear admixture zone in the earliest temporal samples. Analyses were run separately on each year's data (i.e. 2000, 2002 and 2007). Comparisons between mitochondrial and nuclear loci were not made with the data from the 2000 collection due to limited sampling in the southern region in that year.
To assess nuclear admixture, we adopted the Bayesian inference model implemented in STRUCTURE v. 2.3.4 [26] to assign individual crab microsatellite genotypes to population clusters. Models were run independently for each collection year. Initial runs with the number of population clusters ranging from K = 1 to K = n (where n was the total number of sample sites for that year) found that the optimal number of population clusters was K = 2 in all collection years, corresponding to the northern and southern invasion (not shown). To assess individual admixture, subsequent model runs were conducted with K = 2 (five runs of 10 4 generations burn-in followed by 10 5 generations of data collection) and admixture between the two clusters was estimated using the individual co-ancestry coefficient Q of the highest likelihood model run. All analyses assumed uncorrelated allele frequencies allowing for admixture.
Effect of geography on admixture was determined based on changes in allele frequency against alongshore linear distance from the southernmost sampling site at Rye Playland, NY (RP). Distances were Table 1. Sampling site details. Distance is along-shore distance from the southernmost site at Rye Playland, NY. N is the number of individuals for which data was collected at either mitochondrial or microsatellite loci.     [27]. The HZAR algorithm takes both allele frequencies and sample sizes into account, and maximum-likelihood clines and 95% intervals for cline centre and cline width were determined separately at both nuclear and mitochondrial loci for years 2002 and 2007. Limited sampling in the southern portion of the range in 2000 prevented direct comparison of cline shapes across all years. The MCMC was run for 10 6 iterations with 2.5 × 10 5 discarded as burn-in. Clines were based on frequency of southern haplotypes in the case of mitochondrial data and on mean population co-ancestry (Q) in the southern STRUCTURE cluster in the case of microsatellite date.
To identify individuals with cytonuclear discordance, we conducted additional STRUCTURE runs employing the USEPOPINFO flag to pre-assign individuals to either the northern or the southern mitochondrial haplotype cluster. Crabs identified by STRUCTURE as 'immigrants' in this analysis suggest individuals in which northern mitochondrial haplotypes have been introduced into southern nuclear backgrounds or vice versa.

Genetic diversity
In 2002, genetic diversity was significantly lower in the southern part of the range (points from Lubec, ME (L) south) for both nuclear and mitochondrial loci (figure 2). By 2007, this regional difference in mitochondrial gene diversity had eroded (i.e. non-significant pairwise differences across all regions). However, there remained a significant deficiency in nuclear allelic richness in the southern region.   (StP). At nuclear loci, extensive introgression of southern alleles appears to be limited mainly to Halifax and points south, even in 2007 (although one individual at St Peter's did exhibit a nuclear genotype consistent with co-ancestry predominantly in the southern cluster).

Discussion
The unexpected discordance between patterns of genetic variation at nuclear and mitochondrial loci strongly suggests that larval transport is not the only factor shaping temporal dynamics of genetic variation in this system. Specifically, our analyses indicate that the introgression of mitochondrial genomes from the southern to the northern populations, against the direction of mean larval transport and beyond a previously established range limit for C. maenas, may be explained in part by underlying demographic processes. An earlier examination of mitochondrial haplotype frequencies of C. maenas in the northwest Atlantic [21] observed a pronounced southward displacement in the mitochondrial haplotype frequency cline observed in this study ( figure 3). This temporal shift was shown to be consistent with larval dispersal driven by predominantly southward advective coastal currents in the region, and larval transport models accurately predicted the evolution of the mitochondrial genetic cline over time. In this study, the southward expansion of northern haplotypes is reflected in a 213 km southward shift in the mitochondrial cline centre between 2002 and 2007, as well as significant elevation of mitochondrial gene diversity in the southern portion of the range in the same time period ( figure 2). Similarly, the shift in the nuclear genetic cline is consistent with an overall displacement of the admixture zone to the south (figure 3). However, microsatellite data revealed only a modest broadening of cline width compared with the mitochondrial data (figure 3), and there was no significant increase in microsatellite allelic richness in the southern part of the range beyond the primary admixture zone in western Nova Scotia and New Brunswick (figure 2). Assuming that sex-biased larval dispersal is highly unlikely for C. maenas, advective currents would be expected to drive similar shifts in both nuclear and mitochondrial cline centres, consistent with our observations. However, the relatively rapid bi-directional expansion of the mitochondrial cline suggests an important role for demographic processes that, unlike larval dispersal, are expected to have differential influences on nuclear and mitochondrial genomes.
Discordance between genomes is revealed most strikingly by patterns of cytonuclear mismatch in individual crabs. Nuclear admixture between northern and southern genetic clusters remains largely constrained to the primary admixture zone, with some modest admixture in samples from northern New The bidirectional expansion of cytonuclear mismatches is particularly conspicuous given the overall shift in cline centre to the south. Notably, from 2001 to 2007, Pringle et al. [21] documented spread of southern mitochondrial haplotypes (i.e. those with previous northern range limits near Halifax, NS [22,23]) towards the northeast and against the direction of advective currents. They accounted for this unanticipated result by introducing into their model substantial variation in larval transport (i.e. L diff ), which would allow larval dispersal northward and expansion of the southern mitochondrial haplotypes. However, the magnitude of L diff was roughly 3.5 times greater than mean larval transport and substantially higher than what would be predicted by known regional current patterns. The authors suggested that variation in larval dispersal may have been shaped by long-distance anthropogenic transport or by differences in currents encountered by larval cohorts spawned at different times.
A number of recent theoretical and empirical studies reveal that demographic dynamics at expanding range margins may drive surprisingly large stochastic shifts in allele frequency [2]. Rapid population expansions can lead to genetic 'surfing' of alleles at the expansion front, often leading to dramatic genetic differentiation between peripheral and core populations [11,12]. In cases of contact between established and newly expanding populations, these dynamics can result in extensive genetic introgression across population or even species boundaries [15,28,29]. Particularly interesting in this regard are recent simulations demonstrating dramatic introgression driven by demographic imbalances between rapidly expanding invasive populations and relatively static, established ones [13]. Under these circumstances, initial crosses in the contact zone are likely to be interpopulational, as the first few invading individuals are unlikely to find mates from their own population given its low density. Thereafter, genes from the established population that became introgressed into the invading population via interbreeding would experience subsequent amplification through rapid demographic expansion at the invasion front, ultimately leading to dramatic asymmetric introgression of 'native' genes into the invasive gene pool. Such introgression is predicted to be especially pronounced for organellar markers based simply on differences in effective population size and associated sensitivity to neutral drift [13,28,30].
These demographic processes may be at the heart of the observed spread of C. maenas mitochondrial haplotypes in the northwest Atlantic. Both southern and northern lineages are invasive in the region, but the former had been established nearly 200 years before the latter and appeared to have reached a stable range limit near Halifax, NS [18,19]. Contact between the two invasion fronts could thus reasonably be modelled as an encounter between a (southern) native and (northern) invasive population, as in Currat et al. [13]. Genetic introgression from south to north would thus be driven both by initial imbalance in population size at the invasion front and by rapid subsequent demographic expansion of the northern population, as the latter became abundant throughout the Canadian Maritimes within two decades after initial introduction [19]. This demographically driven introgression of mitochondrial genomes from southern to northern populations would relax demands on previous models to account for northward mitochondrial expansion through larval dispersal, presumably enabling incorporation of L diff values more consistent with regional circulation statistics. In addition, the demographic models invoked here may help explain why the southern population failed to expand beyond its northern range limit prior to the second, northern invasion. If substantial northward larval transport is unnecessary to explain the bidirectional spread of mitochondrial haplotypes, then it seems likely that previous range limits were set largely by the lack of sufficient northward dispersal to establish stable populations in northern Nova Scotia, consistent with observations of ephemeral populations north of Halifax prior to the northern invasion [19]. Pringle et al. [21] proposed that Allee effects may have prevented establishment of crabs beyond that previous range limit until occupation of northern habitats by the more recently invading population, and our findings support this hypothesis. In effect, the arrival of a new and abundant population in the north allowed alleles from the naturalized nineteenth century population to expand beyond their previous, demographically imposed range limit, eventually reaching the provinces of Prince Edward Island, Quebec, and even Newfoundland. This 'genetic stickiness' in newly occupied sites was facilitated by demographic imbalances in the zone of contact, allowing southern alleles to expand northward much more rapidly than dispersal alone would have allowed. Although there has been considerable recent interest in the implications of demographic processes for genetic dynamics at range margins, our understanding of these dynamics remains imperfect [31]. Rapid introgression of mitochondrial haplotypes relative to nuclear backgrounds is a common observation [28][29][30][32][33][34][35], and a number of factors may drive these patterns. For instance, the effects of demographic imbalances on mitochondrial introgression may be affected by sex-biased reproductive dynamics. In cases of highly disparate relative abundances, female mate discrimination may drive the capture of maternally inherited genomes by the more abundant population; in such circumstances, female choice and male competition renders matings by rare males unlikely, whereas rare females will be likely to accept interpopulational fertilizations [33,36,37]. Introgression is further facilitated in such cases because first generation hybrids are more likely to mate with members of the more abundant population [35]. Studies of C. maenas reproductive behaviour do suggest an important role for female mate choice and male competition [38], and a recent analysis of interspecific hybridization between C. maenas and its sibling species C. aestuarii concluded that these drivers best explain the observed introgression of mitochondrial genomes across species boundaries [39]. Such female-biased reproductive behaviour could further exaggerate introgression of mitochondrial genomes relative to nuclear background.
Advective systems are complex, with multiple mechanisms (i.e. larval dispersal, demographic imbalances and the potential for sex-biased reproduction) likely to be shaping the temporal development of genetic clines. In our study system, C. maenas shows biased introgression of mitochondrial genomes. This pattern occurs in both directions, but is most dramatic from south to north. Such introgression may be driven by different processes in different areas (or at different times) depending on demographic dynamics. Without additional data, particularly from sample sites within the initial contact zone and closer to the time of first contact, it is difficult to say for certain which models are the most appropriate for fully understanding admixture dynamics in C. maenas. Yet we do believe it is clear that existing knowledge of invasion history and observed genetic patterns are most consistent with selectively neutral drivers of introgression. Cytonuclear discordance has sometimes been attributed to selection on mitochondrial-encoded genes [32], as mitochondrial genes play central roles in oxidative phosphorylation, and the bioenergetic consequences of mitochondrial function may confer selective advantages for traits ranging from life history to thermal tolerance [34,40]. Among intertidal species, in particular, thermal adaptation is frequently driven by selection on mitochondrial genomes [41]. In fact, recent studies have revealed that there may be phenotypic differences between C. maenas populations that potentially confer selective advantage, including significantly higher physiological tolerance to low temperature stress among northern lineages in the northwest Atlantic [42,43]. Nevertheless, selective explanations are difficult to square with the bi-directional character of the mitochondrial introgression; it seems especially unlikely, for instance, that northern mitochondrial genes would be advantageous in the southern portion of the range, whereas southern genes are favoured in the north. Still, the potential role of adaptation clearly deserves future scrutiny. In particular, it has yet to be determined whether there may be some adaptive benefits to admixture that further drives evolutionary dynamics in C. maenas. For example, a number of studies have indicated that admixture may promote colonization and population expansion by facilitating response to novel selective pressures [44][45][46]. To date, observed C. maenas expansion patterns in the northwest Atlantic have been attributed entirely to neutral mechanisms [20,21], but this should not be taken as a reason to rule out the possibility for adaptive consequences of admixture in this system.
Beyond providing insights into the invasion history of C. maenas in the northwest Atlantic, our results may have more general implications for studies aimed at better appreciating the role of dispersal at population range margins. Some authors have recently attempted to estimate such dispersal by analysing neutral genetic cline decay across dynamic admixture zones [8,10]. These studies represent significant methodological advances and offer novel possibilities for indirect estimation of larval dispersal in certain systems. Our findings serve as an important reminder that dispersal may not be the only factor shaping genetic dynamics, particularly at advancing or interacting range margins. Demographic processes can have strong and varied impacts on the spatio-temporal evolution of genetic clines, and if not accounted for, may have unforeseen influences on dispersal estimates. At the edge of species ranges in particular, demographic factors and processes such as gene surfing and genetic stickiness can result in mean displacement of genes outpacing mean larval dispersal. Better appreciation of the demographic factors driving selectively neutral shifts in allele frequency could substantially improve future models of dispersal dynamics at population range limits.
Data accessibility. All genetic data are included in the electronic supplementary material, file D1.