Low genetic diversity contrasts with high phenotypic variability in heptaploid Spartina densiflora populations invading the Pacific coast of North America

Abstract Species can respond to environmental pressures through genetic and epigenetic changes and through phenotypic plasticity, but few studies have evaluated the relationships between genetic differentiation and phenotypic plasticity of plant species along changing environmental conditions throughout wide latitudinal ranges. We studied inter‐ and intrapopulation genetic diversity (using simple sequence repeats and chloroplast DNA sequencing) and inter‐ and intrapopulation phenotypic variability of 33 plant traits (using field and common‐garden measurements) for five populations of the invasive cordgrass Spartina densiflora Brongn. along the Pacific coast of North America from San Francisco Bay to Vancouver Island. Studied populations showed very low genetic diversity, high levels of phenotypic variability when growing in contrasted environments and high intrapopulation phenotypic variability for many plant traits. This intrapopulation phenotypic variability was especially high, irrespective of environmental conditions, for those traits showing also high phenotypic plasticity. Within‐population variation represented 84% of the total genetic variation coinciding with certain individual plants keeping consistent responses for three plant traits (chlorophyll b and carotenoid contents, and dead shoot biomass) in the field and in common‐garden conditions. These populations have most likely undergone genetic bottleneck since their introduction from South America; multiple introductions are unknown but possible as the population from Vancouver Island was the most recent and one of the most genetically diverse. S. densiflora appears as a species that would not be very affected itself by climate change and sea‐level rise as it can disperse, establish, and acclimate to contrasted environments along wide latitudinal ranges.


| INTRODUC TI ON
Invasive plant species can respond to environmental pressures in novel environments through genetic change at individual level (local adaptation), gene expression regulation and epigenetic processes (Richards, 2012), and phenotypic plasticity in response to different environmental conditions (Fuller et al., 2010;Miner, Sultan, Morgan, Padilla, & Relyea, 2005;Stomp et al., 2008). In addition, novel genotypes resulting from hybridization between native and invasive plant species may facilitate rapid evolution (Strong & Ayres, 2013). These different evolutionary processes are closely related as phenotypic plasticity may itself evolve through genetic changes (Crispo, 2008) and may contribute to the evolution of traits that support increased niche breadth of invasive species as plasticity facilitates persistence of individuals facing novel environmental pressures (Drenovsky et al., 2012). Phenotypic plasticity is adaptive, affecting the performance and reproductive success of individuals, being a trait itself, and, therefore, is subjected to evolution (Matesanz, Gianoli, & Valladares, 2010). Additionally, intrapopulation trait variation may modulate the effects of genetic diversity on populations (Hughes, 2014). Moreover, polyploidization, and especially hybridization, have been related to high levels of genetic diversity, epigenetic variation, and phenotypic plasticity (Braeutigam et al., 2013;Meyerson et al., 2016) and they are considered important in shaping the geographic range of plant species (Weiss-Schneeweiss, Emadzade, Jang, & Schneeweiss, 2013).
In this context, few studies have evaluated the relationships between genetic differentiation and phenotypic plasticity of invasive plant species to novel, changing environments along wide latitudinal ranges. These studies have shown that species with low genetic variation can colonize wide latitudinal ranges due to high plasticity levels, sometimes related to polyploidy and hybridization, and their spread into dissimilar environments can lead to genetic differentiation (Counts, 1993;De Kort et al., 2016;Meyerson et al., 2016;Quilot-Turion et al., 2013;Zhao, Yang, Xi, Gao, & Sun, 2012). For example, the high level of epigenetic changes recorded in invasive Spartina anglica Salmon, Ainouche, & Wendel, 2005) has been related to its high levels of morphological plasticity and its large ecological amplitude (Thompson, McNeilly, & Gay, 1991), contrasting with low interindividual genetic diversity (Ainouche, Baumel, & Salmon, 2004;Ayres & Strong, 2001;Baumel, Ainouche, & Levasseur, 2001). In contrast, the cosmopolitan glycophyte Phragmites australis (Cav.) Trin. ex Steud. shows high genetic diversity, with some lineages behaving as invasive out of their native range, at the same time that their high phenotypic plasticity improves the responses to global change factors (Eller et al., 2017). Recently, Liu, Strong, Pennings, and Zhang (2017) have showed that both phenotypic plasticity and rapid local adaptation can contribute to the spread of an invasive species, being both processes interrelated and depending on environmental conditions changing with latitude.
The knowledge of these dialectic relationships between genetic and epigenetic changes and trait expression is especially important for making decisions about conservation of endangered species and for management of invasive species in a scenario of globally changing environment that is modifying the distribution of organisms (Kelly & Goulden, 2008). Thus, it is important to link functional traitbased responses to genetic diversity and environmental changes to improve our understanding of mechanisms that promote changes in the species distribution and adaptation (Nicotra et al., 2010). For example, a better understanding of the role of phenotypic plasticity in response to environmental variation will improve our ability to manage weed invasions and improve conservation efforts. In fact, it has been hypothesized that global change and biological invasions are synergistic processes that interact and impart negative ecological and economic impacts (Vilà, Corbin, Ducks, Pino, & Smith, 2007).
However, the levels of genetic diversity in these S. densiflora populations along the Pacific coast of North America experiencing high variability in sedimentary and climatic conditions have not been studied.
In this study, we performed genetic analyses of S. densiflora using chloroplast (cp) and nuclear DNA sequences, involving simple sequence repeats (SSRs). Cp-DNA sequences (useful for identification of maternal lineages in a hybridization context and for phylogeographic analyses) and potentially highly polymorphic nuclear SSRs were employed to explore genetic diversity among populations established along the Pacific coast of North America. Phenotypic variation within and among populations of S. densiflora was also evaluated through analysis of 33 morphological, physiological, and biochemical traits, on plants sampled in the field that were compared to plants maintained in a common-garden. Five invasive populations of S. densiflora along the Pacific coast of North America from San Francisco Bay to Vancouver Island were analyzed to examine the mechanisms allowing this species to deal with changing environmental conditions and the relationships between these mechanisms. Analyzing the level of intrapopulation phenotypic variability and the phenotypic variation of many plant traits between populations of the same species invading contrasted environments and in a common-garden experiment, together with knowledge of genetic diversity, would allow us to know the nature of the above-mentioned phenotypic variation and its level of phenotypic plasticity (Zhao et al., 2016). Combining these different levels of analysis (phenotypic and genetic, contrasted and common environments, and intra-and interpopulation) is particularly relevant and should shed light on the ways that a polyploid species with hybrid origin (intrinsic characteristics that can promote phenotypic variability at different levels; Grover et al., 2012) can respond to spatial and temporal environmental changes.
The quantification of plasticity for several traits in several populations of the same species is a difficult task, and thus, available data in this regard are still very scarce (Molina-Montenegro & Naya, 2012).
Integrated biogeographic approaches involving the study of natural populations along latitudinal gradients paired with common-garden experiments are a way to increase our understanding of plant species responses to environmental change, yet the spatial scale of most studies has been limited (Elsey-Quirk, Seliskar, & Gallagher, 2011;Hierro, Maron, & Callaway, 2005).
Our overall objective was to assess whether trait responses were dependent on genetic differentiation or phenotypic plasticity or a combination of both factors. We hypothesized that (i) S. densiflora would show low genetic diversity and high phenotypic plasticity in response to contrasted environments and that they would be independent from each other, based on previous observations (Castillo et al., 2014Grewell et al., 2016); (ii) due to the polyploid and hybrid nature of S. densiflora, at least certain foliar traits would show both high plasticity and high intrapopulation variability, as leaf (the organ of photosynthesis and transpiration) is highly sensitive to environmental conditions (Stephenson, Oliver, Burgos, & Gbur, 2006); and (iii) genetically based and environmental-based interpopulation differences in response to a changing environment would increase with geographic distance.

| Studied sites
Spartina densiflora populations were sampled in summer of 2010 at five invaded salt marshes along the Pacific coast of North America, covering its entire latitudinal distribution in North America (along 12° of latitude). See Castillo et al. (2014Castillo et al. ( , 2016 for a full description of sampling sites. The southernmost population analyzed was at Corte Madera Creek in San Francisco Bay Estuary (SF; 37°56′33″N, 122°30′55″W; California, USA) in a middle-elevation salt marsh dominated by S. densiflora, which was intentionally introduced from Humboldt Bay to stabilize tidal creek banks in a wetland restoration project in the late 1970s (Faber, 2000).
The next population to the north was located at Vance Marsh, Humboldt Bay (HB; 40°49′53″N, 124°10′17″W; California, USA) in a middle intertidal elevation marsh where the species has been probably introduced accidentally by merchant vessels from Chile during the late nineteenth century (Spicher & Josselyn, 1985). The third population was near the mouth of the Mad River Estuary (MR; 40°56′10″N, 124°7′48″W; California, USA) fringing the narrow intertidal zone at a river bank, just north of Humboldt Bay.
Spartina densiflora populations were visited during low tides.
Ten adult tussocks with between 10 and 80 live shoots and a mean tussock area between 15 and 41 cm 2 (ca. 2-3 years of age) were randomly selected, separated by a minimum of 1 m in attempt to sample potentially different genotypes, and individually marked in each of the five populations. Due to the low genetic diversity of S. densiflora populations recorded previously in San Francisco Bay and Humboldt Bay , we assumed that n = 10 individuals would be representative of each population, a number of plants per population similar to that sampled in previous studies on genetic diversity of invasive species (Andreakis, Kooistra, & Procaccini, 2009;Guo et al., 2015;Moran, Reid, & Levine, 2017).
The external limits of each S. densiflora tussock were clearly visible, so sampling tussocks separated by a minimum of 1 m ensured that plants coming from different establishment events were sampled. The amplification of the trnT-F region was performed in two sections using the primer pairs a and b to recover the trnT-trnL segment and primer pairs c and f to recover the trnL-trnF segment (Taberlet, Gielly, Pautou, & Bouvet, 1991). Sequence alignments and comparisons were performed with the Geneious 9.1.2 software (Kearse et al., 2012). Sequences were compared with previously published S. densiflora sequences , Baumel et al., 2001;Fortune et al., 2008).

| Simple sequence repeats
Neutral molecular markers such as SSRs provide independent and complementary assessment of genetic diversity to phenotypic variation most likely submitted to selective constraints. They are thus commonly used for genotyping in studies of phenotypic variations and fitness (Agarwal, Shrivastava, & Padh, 2008;Chapman, Nakagawa, Coltman, Slate, & Sheldon, 2009). A total of 1435 microsatellite repeats (SSRs) of 4-6 bp were identified from Spartina maritima genomic data obtained by Illumina sequence assemblies (A. Salmon, unpublished), using the MISA (MIcroSAtellite Identification Tool) software (Thiel, Michalek, Varshney, & Graner, 2003). Primer pairs were designed using the default parameters (and optimized TM ranging between 55 and 60°C) of the PRIMER3 software (Koressaar & Remm, 2007;Untergasser et al., 2012), which resulted in a range of primer product sizes of 100-280 bp and SSR motifs of at least −4 to 6 nt long, repeated five to eight times. This resulted in 395 different microsatellite loci with successful primer design (A. Salmon, unpublished data). For the first test, we selected 20 primer pairs considering those with the highest number of expected amplicon size and number of motif repetitions. PCR was performed as described above with these 20 primer pairs in S. densiflora nuclear DNA and analyzed by electrophoresis in agarose gel (1%, 35 min, 100V). Eight primer pairs of which amplification products produced unique and clear bands were chosen for further analysis (Table 1).
The amplified products were diluted to 1/30 before subsequent analysis. An aliquot of 2 μl of these diluted PCR products was mixed with 10 μl of formamide solution (975 μl formamide + 25 μl of GeneScan-500 LIZ size marker) and separated by electrophoresis in an ABI PRISM 16-capillary 3130xl Genetic Analyzer (Applied Biosystems Inc., Waltham, USA). Alleles were identified and binned using GeneMapper 4.1 software (Applied Biosystems Inc., Waltham, USA).

| Genetic diversity analyses
SSR alleles were determined by comparison with the standard marker size (GeneScan-500 LIZ size standard), and the different "genotypes" (harboring different allele combinations) were scored. S. densiflora is a heptaploid species (2n = 7x = 70), so up to seven alleles may be expected per locus. As allelic dosage could not be ascertained from the obtained chromatograms, SSRs were treated as "dominant markers" (i.e., alleles were scored as either present or absent) using the GenAlEx 6.502 software (Peakall & Smouse, 2006, 2012. Genetic diversity analyses are less affected by polyploidy using this approach (Obbard, Harris, & Pannell, 2006).
A cluster analysis was performed using the phangorn R package (Schliep, 2011) with the matrix of pairwise genetic differences and the unweighted pair-group method with arithmetic averages (UPGMA; Sneath & Sokal, 1973). The proportion of genetic diversity among and within populations was assessed by molecular variance analysis (AMOVA) using the GenAlEx 6.502 software for the five sampled populations of S. densiflora. Genetic differentiation among populations (Φ PT ), analogous to Wright's F' ST for dominant data, was obtained following standardization (Meirmans, 2006).

| Phenotypic variation analyses
The following 33 tussock and foliar phenotypic traits were evaluated in the field for the genetically analyzed individuals of S. densiflora: shoot density in (1) alive, (2) spiked, and (3) dead stem density (shoot per m 2 ), shoot (4) length (cm) and (5) diameter (mm), number of (6) live and (7) dead leaves, (8) total nonstructural carbohydrates in rhizomes (mg/g), (9) root biomass (g/m 2 ), (10) rhizome biomass (g/ m 2 ), (11) belowground biomass (BGB, g/m 2 ), (12) dead shoot mass (g/  (Epstein, 1972;Hoagland & Arnon, 1950). Solutions were changed once a week. After 27 months of growth, every tussock trait recorded in the field, except belowground biomass, rhizome TNC, and leaf carbon and nitrogen concentration, was recorded again for the same plants acclimated to common glasshouse conditions (n = 10 tussocks per population). Common glasshouse conditions included substrate redox potential of 234 ± 5 mV, pH 8.3 ± 0.2, electrical conductivity 0.5 ± 0.0 mS/cm, mean monthly air temperature 23 ± 2°C, and mean monthly air relative humidity 62 ± 2%. Light conditions of mean radiation of 700 μmol photon m −2 s −1 at canopy level and a daily photoperiod of 16 hr (that was extended with incandescent lights, Osram Vialox NAV-T (SON-T) 400 W, giving a continuum spectrum) were set up to imitate those recorded at higher latitude sites during summer in relation to photoperiod and light intensity. Spartina densiflora is a facultative halophyte that can germinate, establish, and develop in freshwater conditions (Castillo, Rubio-Casal et al., 2005;. Freshwater was used to avoid salinity effects on tussock and foliar growth. Phenotypic distances between plants in the field were calculated as Gower similarity index ranging between 0 and 1 (Gower, 1971) using the package vegan (Oksanen, 2007) of R software to obtain a pairwise dissimilarity matrix.
Correlations between genetic, phenotypic, and geographic distances in the field were performed using the Pearson's correlation coefficient (p < .05). Comparisons of the dissimilarity matrices described above were tested by Mantel's test for matrix correspondence (Mantel, 1967) using GenAlEx 6.502 software with 1000 permutations, following the method of Smouse and Long (1992); Smouse, Long, and Sokal (1986). A cluster analysis was performed using the package vegan (Oksanen, 2007)  where X max is the maximum mean value for a given trait and X min is the minimum mean value for a given trait at population level (Valladares et al., 2006). This index is adequate for comparison of plasticity sensu lato, which is more focused on the responses of (F-test, p < .05) using the plant trait or the geographic location as the grouping factor, respectively. Dependent variables were tested for homogeneity of variance using Levene's test and for normality using Kolmogorov-Smirnoff test. The relationships between PI v and intrapopulation variability index in the field and in the common-garden experiment were analyzed using Pearson's correlation coefficient (r) and linear regressions. Total intrapopulation variability and total phenotypic plasticity (calculated as arithmetic means for all traits and populations together) were compared using paired-samples t-test. Deviations were calculated as standard error of the mean (SE). The α-level of significance was p < .05 for all tests. These analyses were carried out using SPSS 12.0 (SPSS Inc., Chicago, USA).

| DNA sequencing
The

| Simple sequence repeats
The eight microsatellites primer pairs yielded a total of 19 different alleles. Each primer pair amplified between one and five different alleles per locus, with an average of two alleles per locus. The eight loci examined exhibited different levels of genotype diversity: One single genotype was detected for all populations with the markers MS2, MS7, MS16, and MS17, two different genotypes at the MS13, MS14, and MS18 loci and three genotypes were encountered for MS15 (Table 2). We found seven different multilocus genotypes when all loci recorded were considered, with the most common multilocus genotype (I) present in 40% of the examined individuals (Table 3). The population from Grays Harbor presented two specific multilocus genotypes (II and VI) that were not present in any other population ( Table 3).
The population from Mad River had the lowest genetic diversity (Table 4), showing only three multilocus genotypes (Table 3) (Table 3) and their individual plants scattered in many different clusters Some genetic differentiation was observed among populations (Φ PT = 0.159), but the AMOVA showed that interpopulation variance contributed to only 16% of the total genetic variation, whereas within-population variation represented 84% of the total variation (Table 5). In fact, the five populations of S. densiflora were genetically very similar as showed by Nei's standard genetic distance (from 0 between SF and HB, SF and VI, and HB and VI to 0.15 between SF and GH; Table 4). The UPGMA analysis based on genetic distances revealed that each cluster was composed of individuals from different populations, which agreed with low interpopulation genetic differentiation ( Figure 1). The Mantel test found a significant relationship between genetic distance and geographic distance (p = .030, n = 50; Figure 2a).

| Phenotypic variation
Contrasting to genetic variation, consistent phenotypic differentiation among populations was encountered in the field. The dendrogram using phenotypic distances clustered together every individual of the same population, except for the population from San Francisco Bay of which individuals were grouped in different two clusters with plants from Mad River and Vancouver Island (Figure 1b)  test showed that phenotypic distance increased with geographic distance (p = .010, n = 50; Figure 2b), whereas no significant relationship was found between genetic distance and phenotypic distance in the field (p = .403, n = 50; Figure 2c).
Mean IVI for the 33 recorded plant traits ranged from 7 ± 2% for foliar carbon content in the field to 96 ± 3% for dead shoot mass in the field (Figure 3a). PI v in the field was from 9% for foliar carbon content to 86% for the leaf area index. In general, the traits with higher PI v were those related to tussock aboveground architecture (shoot densities, above-and belowground biomass, and the leaf area index), and leaf adaxial rolling and elongation. In On the other hand, those plant traits that had higher IVI and higher PI v in the common-garden experiment also had higher values for both variables in the field (r = .923, p < .0001, n = 26; r = .688, n = 26, p < .0001, respectively; Figure 5a,b). In addition, the IVI had similar values for every plant trait in the field and in the commongarden experiment (Figure 5a), whereas PI v tended to be lower in the common-garden experiment than in the field (Figure 5b). Thus, total intrapopulation variability and total phenotypic plasticity indices for every plant trait and every population were similar in the field (56 ± 4% and 54 ± 4%, respectively; t-test, p > .05). In contrast, total intrapopulation variability was higher than total interpopulation variability in the common-garden experiment (60 ± 2% and 37 ± 4%, respectively; t-test, t = 8.036, df = 25, p < .0001).  populations was the result of phenotypic plasticity (Castillo et al., 2014Grewell et al., 2016). Nevertheless, the mean interpopulation phenotypic variability recorded for every plant trait in the common-garden experiment was 37% (in comparison with the 56% recorded in the field), which pointed to another process, beyond phenotypic plasticity, that was also responsible for a significant amount of the recorded interpopulation phenotypic variation. In this sense, our study revealed that invasive S. densiflora populations had a similar, TA B L E 5 Analysis of molecular variance (AMOVA) illustrating the proportion of genetic variation distributed within and among populations of Spartina densiflora along the Pacific coast of North America fixed, and high intrapopulation phenotypic variability for many plant traits in the field and in the common-garden experiment (where it was higher than phenotypic plasticity for most of the recorded traits).

| D ISCUSS I ON
Moreover, this intrapopulation phenotypic variability was especially high, irrespective of environmental conditions, for those traits showing also high phenotypic plasticity, suggesting that intrapopulation phenotypic variability in S. densiflora would be regulated through the same mechanisms as those regulating phenotypic plasticity. Ensuring high levels of intrapopulation phenotypic variability for key plant functional traits would allow S. densiflora to respond adequately to the changing environment (Zhao et al., 2016) of salt marshes, where local temporal and spatial changes on salinity, soil moisture, oxygenation, and flooding period are frequent (Adam, 1990;Castillo et al., 2000). The presence of intrapopulation phenotypic variability within species can result in faster range expansions than if only a single phenotype is present in the landscape (Elliott & Cornell, 2012).
Many of the studied plant traits had PI v values higher than 50% in the field, and the mean PI v for every trait was 54 ± 4%, within the range recorded for other invasive plant species (Funk, 2008) and Spartina species (Castillo, Redondo, Wharmby, Luque, & Figueroa, 2005;Elsey-Quirk et al., 2011;Thompson et al., 1991;Trnka & Zedler, 2000). S. densiflora is a heptaploid species of hybrid origin (Fortune et al., 2008), which most likely has played a key role in  (Table 2); more than two alleles were observed at the MS15 locus (the most common heterozygous genotype exhibiting three alleles) and the MS18 locus (heterozygous with 2 or 3 alleles). This is consistent with the hybrid (alloheptaploid) origin of this species (2n = 70, Ayres et al., 2008), which is believed derived from a tetraploid and a hexaploid ancestor in its native South American range Fortune et al., 2008), although more alleles per locus could be expected in a heptaploid genotype. Genes duplicated by polyploidy (homeologs) can exhibit various gene expression evolution patterns which include alternative homoeologous gene silencing, subfunctionalization, neofunctionalization, and nonadditive parental gene expression such as expression of dominance or transgressive expression (Grover et al., 2012). Polyploidy has been repeatedly linked with plant invasion success; in particular, allopolyploidy (hybridization followed by genome multiplication) may lead to substantial epigenetic variation, allowing the expression of novel phenotypes with increased fitness and conferring genetic advantages linked to genetic diversity and gene expression (te Beest et al., 2011;Jackson, 2017). Ecotypic differentiation of S. densiflora has been described for populations in its native range related to environmental changes driven by latitude (Álvarez et al., 2009).
These genome dynamics likely support an increased range of potentially adaptive phenotypic variation and the ability of the species to invade diverse habitats. Our research suggests that the alloheptaploid nature of the S. densiflora genome Fortune et al., 2008) contributes to the responses of this invasive plant species to the changing environments in which it has colonized along the Pacific coast of North America. Phenotypic plasticity has been associated to increased competitive ability and fitness in many taxa, especially when coping with accelerated changing conditions (Hoffman & Sgro, 2011). The above-reported high phenotypic variability would contribute to an overall "jack-of-all-trades" result with less variability across environments in terms of population-level consequences (Richards, Bossdorf, Muth, Gurevitch, & Pigliucci, 2006).
Understanding which traits of an invasive species show higher phenotypic variability may inform us about aspects of its aggressive behavior, which can be important to make predictions about its future spread using invasion models (Martina & von Ende, 2012). The plant traits exhibiting high (>60%) population phenotypic variability were leaf elongation and rolling and those related to aboveground architecture of tussocks, which have been reported as significant factors contributing to fitness (Cornelissen, Song, Yu, & Dong, 2014;Heckathorn & DeLucas, 1991;Kadioglu & Terzi, 2007;Schierenbeck, Fleitas, Miralles, & Simon, 2016). Spartina densiflora has invaded salt marshes in the southwest Iberian Peninsula, where aboveground architecture of its tussocks also varies among different habitats . On the other hand, S. densiflora also occupies a wide variety of habitats in its native range along the South Atlantic coast of South America, where a high degree of phenotypic plasticity has been observed in its root system (Daleo & Iribarne, 2009), which was similar to the interpopulation root system variability observed in our study (ca. 60%). Grewell et al. (2016) studied the response to salinity of the same S. densiflora populations as those analyzed in this study. They found lower phenotypic plasticity values for S. densiflora ecotypic differentiation related to environmental changes driven by latitude and elevation along the tidal gradient has been described for S. densiflora populations in its native range (Álvarez et al., 2009;Di Bella et al., 2014). These results support the hypothesis that the fitness and range potential of invasive species can increase as a result of genetic or epigenetic divergence of populations along broad climatic gradients (Konarzewski, Murray, & Godfree, 2012).
Halophytes need to have efficient medium-and long-distance dispersal mechanisms as they inhabit in salt marshes that are F I G U R E 4 Mean intrapopulation variation index (%) (±SE, n = 5) for five Spartina densiflora populations recorded in the field along the Pacific coast of North America (a) and in a common-garden experiment (b) calculated using their minimum and maximum values from their 33 plant traits (±SE, n = 33). The same letter above the bars denotes nonsignificant differences among populations (ANOVA, p > .05). Source populations are as follows: San Francisco Estuary, California (SF); Humboldt Bay, California (HB); Mad River Estuary, California (MR); Grays Harbor, Washington (GH); and Vancouver Island, British Columbia (VI). Relationship between mean intrapopulation variation index (%) and interpopulation phenotypic plasticity index (%) for 33 traits from five Spartina densiflora populations recorded in the field along the Pacific coast of North America (c) and in a glasshouse common-garden experiment (d). Regression equation: (c) y = 3.38 + 0.98x; (d) y = 28.16 + 0.87x ephemeral ecosystems (Fagherazzi, 2013) due to relatively rapid succession development  and frequent and intense perturbations such as erosion (Castillo, Rubio-Casal, Luque, Nieva, & Figueroa, 2002). The dispersion of many halophytes such as S. densiflora is carried out through hydrochorous propagules rafting on local and regional currents (Morgan & Sytsma, 2013), avoiding habitat fragmentation that may limit species range shift (Davis & Shaw, 2001). S. densiflora populations produce many viable seeds ('Kittleson & Boyd, 1997;Nieva, Diaz-Espejo et al., 2001) and also colonize surrounding sediments by vegetative growth via tillering, producing dense tussocks through "phalanx" clonal growth (Nieva et al., 2005). In this content, the high levels of phenotypic plasticity and intrapopulation phenotypic variability of S. densiflora for key plant traits would promote seedlings to acclimate and to colonize new salt marshes with contrasted environmental conditions. S. densiflora started its invasion of the Pacific coast of North America in the late nineteenth century in Humboldt Bay (Spicher & Josselyn, 1985), being transplanted from there to San Francisco Bay in the late 1970s (Faber, 2000) and then expanding naturally its distribution northward to Grays Harbor in 2001 (Pfauth et al., 2007)  as a species that would not be very affected itself by climate change and sea-level rise as it can disperse, establish, and acclimate to contrasted environments along wide latitudinal ranges. Global climate change may even favor some polyploid taxa that can tolerate stressful environments due to greater phenotypic plasticity (Meyerson et al., 2016).
In summary, our results show that S. densiflora that has invaded a broad latitudinal gradient along the Pacific coast of North America is characterized by low genetic divergence between populations, high phenotypic plasticity related to local conditions, and high and fixed intrapopulation polymorphism in some functional traits, which appears to be a key feature of this allopolyploid invasive species.

ACK N OWLED G M ENTS
The Canada for their assistance.