Landscape genetic structure of Scirpus mariqueter reveals a putatively adaptive differentiation under strong gene flow in estuaries

Abstract Estuarine organisms grow in highly heterogeneous habitats, and their genetic differentiation is driven by selective and neutral processes as well as population colonization history. However, the relative importance of the processes that underlie genetic structure is still puzzling. Scirpus mariqueter is a perennial grass almost limited in the Changjiang River estuary and its adjacent Qiantang River estuary. Here, using amplified fragment length polymorphism (AFLP), a moderate‐high level of genetic differentiation among populations (range F ST: 0.0310–0.3325) was showed despite large ongoing dispersal. FLOCK assigned all individuals to 13 clusters and revealed a complex genetic structure. Some genetic clusters were limited in peripheries compared with very mixing constitution in center populations, suggesting local adaptation was more likely to occur in peripheral populations. 21 candidate outliers under positive selection were detected, and further, the differentiation patterns correlated with geographic distance, salinity difference, and colonization history were analyzed with or without the outliers. Combined results of AMOVA and IBD based on different dataset, it was found that the effects of geographic distance and population colonization history on isolation seemed to be promoted by divergent selection. However, none‐liner IBE pattern indicates the effects of salinity were overwhelmed by spatial distance or other ecological processes in certain areas and also suggests that salinity was not the only selective factor driving population differentiation. These results together indicate that geographic distance, salinity difference, and colonization history co‐contributed in shaping the genetic structure of S. mariqueter and that their relative importance was correlated with spatial scale and environment gradient.

Estuary may be the best laboratory to reveal the interaction between divergent selection and gene flow (Bible & Sanford, 2016;McCairns & Bernatchez, 2008). It represents the transitional zone between freshwater from inland and salt water from open sea (Potter, Chuwen, Hoeksema, & Elliott, 2010;Pritchard, 1967), where various processes including physical, chemical, biological, and geological dynamics are immensely complex (Wolowicz, Sokolowski, & Lasota, 2007). The distinguishing attribute of estuaries is the impact of abiotic characteristics, such as the mixing of two water sources, the rise and fall of the tides and ocean currents. This attribute makes the estuary system exceptionally variable in space and in time, especially in salinity gradients and tide levels, and creates dynamic and heterogeneous habitats (Wolowicz et al., 2007;Xin, Wang, Lu, Robinson, & Li, 2015), which may drive divergent selection. On the other hand, estuary is an open system without physical boundaries of dispersal, and the agitation of tide and freshwater connects different regions of an estuary and accelerates dispersals of all floating propagules in water, including seeds, eggs, larva, etc., which are likely to result in high gene flow among populations. It is intriguing what pattern of spatial genetic structure will be displayed in estuarine species under the contrasting effects of selection pressure and gene flow.
Here, we focus on an estuarine species, Scirpus mariqueter Tang & F. T. Wang (Cyperaceae), which is almost limited in the Changjiang (Yangtze) River estuary (CRE) and its adjacent Qiantang River estuary (QRE; Ou, Fang, & Shen, 1992). S. mariqueter grows in the lowest intertidal zone as a pioneer species forming dense meadows usually in front of Phragmites australis and sometimes can consist of single-species communities covering ~100 km 2 (Ou et al., 1992). S. mariqueter has been confirmed to play a key role in accelerating the development of islands and foreshores in the estuaries (Ou et al., 1992;Yang, 1998). In addition, its tubers and achenes are important food sources for several million migratory birds every year (Ma et al., 2003). However, recently, this species has been threatened by an invasive species, Spartina alterniflora Loisel., and also destroyed by frequent reclamation (Chen, Li, Zhong, & Chen, 2004). The reclamation events caused new populations establishing in the outside of isolating seawalls and thereby re-isolating populations in the inside of seawalls. Possibly, colonization or reclamation process also played a role in structuring the genetic make-up of the populations. There is a pressing need for developing protective measures, which also requires detailed information on the genetic structure of this species.
It is challenging to predict genetic structure of S. mariqueter because so many stochastic and potentially complex processes, including selection pressure under environmental variation, dispersal ability, and reclamation process, might affect the genetic structure.
In this study, using amplified fragment length polymorphism (AFLP), we quantified the distribution of genetic variations and the migration among populations of S. mariqueter under heterogeneous environments of two adjacent estuaries, the CRE and the QRE. In addition, this structure may be driven by neutral and/or selective processes; thus, we look for loci potentially affected by selection (outlier loci). To disentangle the relative roles of the different evolutionary forces acting on genetic structure of this species, it is useful to combine the information on levels of gene flow obtained from neutral loci with the information from outliers that are likely to be of selective signs. In this sense, it is hypothesized that if outlier loci do not show the same IBD or IBE pattern as neutral loci, but showing a direct correlation between genetic differentiation and geographical distance or environmental variables, or colonization history, the evidence for the relative importance of above processes in shaping the structure of this species will be found.

| Study species and sample locations
For a long time, S. mariqueter has been considered as an endemic species of China, occurring in the CRE and the QRE (Ou et al., 1992). This species is a perennial clonal herb, which usually expands vegetatively by tubers with rhizome connection and reproduces sexually by seeds. S. mariqueter flowers from June to August. Although its flowers are wind pollinated and protogynous (see in Supporting information Figure S1), usually implying expected outcrossing, it is also highly self-compatible . Achenes are matured in summer-autumn, and dispersal occurs mostly via achenes and tubers by currents and waterfowls and also by boats navigated among seaports. The discrete patches of S. mariqueter tend to be clear away when suffered reclamation which separate tide, but the new established patches occurred soon, and thus, the colonization history of population can be estimated from the constructed time of the latest reclamation.
The Changjiang River, is one of the largest rivers in the world, discharges a large amount of freshwater into the East China Sea through the CRE (Chen, 1988;Yang et al., 2006). The CRE has a complex structure because of the existence of a few alluvial islands. It is divided into the north branch and the south branch by Chongming Island, and the latter is again divided into the north passage and the south passage by another alluvial island, Changxing Island ( Figure 1).
This complex estuary has resulted in highly heterogeneous habitats with dramatically different salinity, tide, and currents (Kong, He, Ding, & Hu, 2004;Wang, Li, Zhou, & Gao, 2011;Xue et al., 2009;Zhang et al., 2013;Zheng, Ding, & Hu, 2008). Temporal and spatial variation in salinity is mainly controlled by the relative importance of river-ocean mixing. For example, only 1%-3% of freshwater discharge input to the north branch; hence, the salinity of north branch is far higher than south branch (Kong et al., 2004). Over all, the order of the salinity of these branches, from high to low, is the north branch, the south passage, and the north passage (Hu, Hu, Gu, Su, & Gu, 1995;Xue et al., 2009;Zheng et al., 2008), and within the same branch, the salinity increases from nearshore to offshore (Figure 1).
In addition, at each location, the salinity changes over timescales of days (tide fluctuation), seasons (river inflow is different in wet season and dry season), and years (annual weather anomalies; Hu et al., 1995;Xue et al., 2009). These results reflect complicate temporal and spatial changes in salinity. Adjacent to the CRE, the QRE is a typical funnel-shaped estuary with higher salinity in most regions than the CRE due to only with about 1/10 freshwater discharge of the latter and has stronger tide influence compared to the CRE (Yang, Zhu, & Zhu, 2001). These two estuaries consist of a highly connected and extremely heterogeneous system.

| Sample and data collection
In this study, samples were collected from fourteen discrete locations along the coastal lines (Figure 1), which cover the known distributional range of S. mariqueter. Thirteen populations were located at the CRE and the QRE and one at the Gaomei wetlands from Jhonggang River Estuary (JRE) of Taiwan, China. Since the locations in the same region were suffered simultaneous reclamation process, and new population could be founded outside the seawall; thus, the populations from the same region have the same colonization history. According to colonization histories of these locations, these 14 populations were defined to eight groups from different geographic region: group 1 (QD1 and QD2 from Qidong), group 2 (CM1, CM2, and CM3 from Chongming Island), group 3 (NH and JS from Shanghai), group 4 (HS from Hengsha Island), group 5 (JD from Jiuduan shoal), group 6 (HZ1 and HZ2 from Hangzhou), group 7 (YY and BL from Ningbo), and group 8 (TW from Taiwan; Table 1).
Approximate coastal geographical distances between population pairs were obtained with Google Earth software. Pairwise geographical distance between populations that located in the CRE and QRE ranged from 20 to 340 km, but more far from TW population (~800 km; Figure 1, Table 2). Taking into account that the growing period of S. mariqueter is May-October, the estimations of average surface salinity in summer of each location were obtained from previous studies ( Table 1). The absolute values of average pairwise salinity differences between sampling points were then calculated.
Considering that S. mariqueter holds the strong ability of clonal reproduction, we sampled randomly with 10-20 m sampling intervals to avoid sampling the same clone. About 50 individuals from each population were collected except for the population of Hengsha Island (HS), where only 13 individuals were available due to very limited population size of this population. Each individual was, respectively, stored in a plastic bag and then dried with silica.
Genomic DNA was isolated from 5-cm dried leaf tissue using CTAB (hexadecyltrimethylammonium bromide)-based method (Doyle & Doyle 1987) from 10-cm long-dried leaf material. DNA quality and concentration were estimated on 1% agarose gels.
We select AFLP genome scans to detect genetic structure. About 50 ng of DNA were used for AFLP analysis according to Vos et al. (1995)  were scored (1 as present, 0 as absent) using the GeneMapper 3.7 software (Applied Biosystems). Only unambiguous fragments were analyzed and transferred into a binary matrix. In order to reduce scoring errors, fragment peaks with fluorescence values >100 were considered as loci, and 48 samples were repeated for all processes to detect differences in allele scoring. The error rate was calculated as the proportion of fragments that could not be reproduced and the locus of error rate over 10% was discarded from the dataset. Finally, the resulting adjusted binary matrix was assembled for subsequent analysis.
F I G U R E 1 The geographic range of Scirpus mariqueter, the salinity level of the Changjiang River estuary (CRE) and the Qiantang River estuary (QRE), and the locations of sampled populations in this study. Photograph of Scirpus mariqueter from the Changjiang River Estuary (photograph credit: Mei Yang). Pie graphs show the genetic clusters and their relative proportion in different populations (the abbreviation of each cluster is corresponded to those designated in Table 4). Probability isoclines and numbers on these lines show the extent of the surface salinity (‰ or ppt) of the two estuaries in summer (redrawn according to Chen, 1988;Kong et al. 1994;Hu et al., 1995 andBao et al., 2013). Population codes are corresponded to those designated in Table 1 2 The program decomposed F ST into locus-and population-specific components and was run by setting sample size to 10,000 and the thinning interval to 50 (Foll & Gaggiotti, 2008). The loci with a posterior probability over 0.99 were retained as outliers, corresponding to a Bayes Factor >2. Secondly, we used the Fdist approach by

.3 | Detection of outlier loci
Beaumont & Nichols implemented in MCHEZA (Antao & Beaumont, 2011), which applies a multitest correction based on false discovery rate (FDR, which is the proportion of false positives among the tests found to be significant) to avoid high overestimation of the percentage of outliers (Caballero, Quesada, & Rolan-Alvarez, 2008

| Genetic diversity, differentiation, and isolation by distance/environments
We characterized the overall level of genetic diversity within population, estimating: the proportion of polymorphic loci (PPL), Nei's gene diversity (H j ), and the Bayesian estimate of gene diversity (h S ).
The first two calculations were carried out using AFLP-SURV 1.0 (Vekemans et al., 2002) with nonuniform prior distribution and assuming Hardy-Weinberg genotypic proportions. AFLP-fragment frequencies were estimated using the reliable square root method (Lynch & Milligan, 1994)   Note. The average value of salinity were measured in the surface water, and data were from Chen (1988), Hu et al. (1995), Kong et al. (2004), and Bao et al. (2013) a model in which θ B = 0, and (d) a f-free model. These models were compared using the deviance information criterion (DIC). The model with the smallest value was chosen. We set default sampling parameters: burn-in = 50,000, sampling = 250,000, thin = 50.
In order to identify effects of colonization history on genetic structure, we used AMOVA. Total genetic diversity was partitioned among groups, among populations, and within populations by carrying out a hierarchy AMOVA on Euclidean pairwise distances among individuals using GENALEX 6.5 (Peakall & Smouse, 2012) with 999 permutations. Eight population groups were defined based on their population colonization history.
To investigate the relative importance of spatial distance and salinity difference on genetic structure, we tested for IBD and IBE based on the three datasets: all loci, positive, and neutral dataset.
For IBD test, we applied traditional method of a Mantel test as implemented in GENALEX 6.5 (Peakall & Smouse, 2012 isolation by environments also been calculated to examine the association between the genetic and salinity difference. The average surface salinity in summer (peak growing season of S. mariqueter) was selected as an environment factor to test the effect of IBE. These correlations were performed using SPSS 15.0 software. All of above analysis were carried out with and without outlier loci to determine whether or not our results were being influenced by loci which may be under selection.

| Genetic structure and contemporary estimates of dispersal
Assignment test is a common genetic method to provide contemporary or short-term estimates of dispersal among populations (Campbell, Duchesne, & Bernatchez, 2003). Several different approaches were used to ascertain populations' genetic structure.
First, the AFLP binary matrix was analyzed using a Bayesian model-based clustering method, as implemented in STRUCTURE 2.2 (Falush, Stephens, & Pritchard, 2007;Pritchard, Stephens, & Donnelly, 2000). We chose a burn-in period of 30,000 iterations and chain length of 100,000, respectively. Independent runs with K (the number of populations) iteratively set from 2 to 14. Each run was parameterized following a model of admixture and correlated allele frequencies. After assessing the distribution of P(X|K) and Ln (K) values, all individuals were partitioned into K clusters based on the probability values. And also, the most likely number of genetic clusters (K) was detected following the approach presented by Evanno, Regnaut, and Goudet (2005).
Given that the true values of K could not be obtained to visualize structure (see in Section 3), furthermore, we used a model-free iterative reallocation method, FLOCK (Duchesne & Turgeon, 2009

| Detection of loci under selection
A total of 641 samples were scored for 434 loci, and these loci were

| Genetic diversity and among-population differentiation
The proportion of polymorphic loci (PPL) ranged from 49.3% to 66.8% (Table 1) based on all loci revealed that 80.51% of the variation within populations, while 6.85% was due to variation between populations and 12.64% was due to variation between groups. When based on 21 outliers, the hierarchical AMOVA revealed that 47.19% of the variation within populations, while 9.21% was due to variation between populations and relatively high proportion (43.60%) was due to variation between groups with different population colonization history (Table 3).

| Genetic structure and ongoing gene flow
The Bayesian clustering method based on STRUCTURE could not infer an optimal structuring into K populations: Ln (K) kept increasing FLOCK analysis, the non-Bayesian approach, however, found a partition into 13 clusters (i.e., genetic types, abbreviated as "C") as the most likely solution (Table 4). According to the assignment test by FLOCK, each individual was assigned to one of thirteen clusters.
There were six narrowly distributed clusters, and they almost oc-   factor, similar results were obtained (when salinity difference ≤4.0 ppt, r = −0.38, p = 0.042; when salinity difference >4.0 ppt, r = 0.29, p = 0.047). However, no significantly positive relationship was found as expected when only considering outlier dataset, but also a con-

| D ISCUSS I ON
In this work, non-Bayesian clustering analysis using FLOCK assigned all individuals to thirteen genetic clusters and revealed a complex genetic structure of S. mariqueter: some clusters were limited in marginal locations compared with very mixing constitution in central populations. In another analysis, only two genetic clusters were found in STRUCTURE following Bayesian clustering method, suggesting high connectivity among populations, and both populations and individuals were varying degrees of admixture. However, a relatively high genetic differentiation value (F ST = 0.1857) among populations was found although all locations hydrologic linked and F I G U R E 3 Bayesian clustering for 14 populations of S. mariqueter STRUCTURE. Each individual is represented by a vertical bar divided into two or three segments corresponding to its membership coefficients in the two (K = 2) or three (K = 3) inferred clusters. Each tonality represents a different cluster and black lines separate the individuals of different localities. (a) and (b), analyses using the entire genetic dataset of 434 loci with K = 2 and K = 3, respectively; (c) and (d), analyses considering outlier dataset of 21 loci with K = 2 and K = 3, respectively; (e) and (f), analyses taking into account only the 316 neutral loci with K = 2 and K = 3, respectively. The K value is determined from the mean estimated ΔK (Evanno et al, 2005). More information found in Supporting information Figure S2 indeed amount of migrants were detected. Considering these results, population genetic structure and its correlates with spatial distance, salinity difference and population colonization history were analyzed. The outlier loci analyses suggested that some of the AFLP loci were putatively under divergent selection, and the analyses of population genetics considering only these outliers revealed a very different pattern of differentiation from results obtained with the entire dataset or neutral dataset according to STRUCTURE, AMOVA, IBD, and IBE tests. The results suggest that the ecological forces, including environmental factors like surface salinity, spatial distance, and colonization history all play important role in shaping population structure of S. mariqueter and the effects may be compound.

| Migration, gene flow and genetic structure
Every specific trait of estuarine habitats, especially the rise and fall of the tides and the exchange between saltwater and freshwater, make the spatially isolated species in estuary to be interconnected to varying degrees through dispersals of seeds or/and other propagules (Potter et al., 2010;Uncles & Stephens, 2011). For S. mariqueter, beyond the transport by water flow, the long-distance dispersal mostly F I G U R E 4 IBD analyses for all 14 population (a) and 13 populations in CRE and QRE when excluding TW population considering different dataset: all loci dataset (b), outlier loci dataset (c), and neutral dataset (d). The mantel test scatter plot shows the relationship between the pairwise genetic differentiation (F ST ) and the geographic distance (km) between populations TA B L E 4 Assignment numbers and allocation of S. mariqueter individuals (n = 641) statistics from 14 locations S. mariqueter is a cross-pollination species (Ou et al., 1992;Yang et al., 2013) and has a narrow adaptive belt at its each distributed site; thus, it is reasonable for us to assume that these migrations are likely to result in gene flow among populations, which tends to reduce genetic differentiation or homogenize populations and may swamp adaptation to local conditions (Ellstrand, 2014). However, the assignment test by FLOCK showed that the gene flow among populations had been inhibited at certain degrees. Some genetic types were almost limited in one population, for example, C4, C5, and C10, suggesting these clusters were locally adapted, and gene flow was limited between populations even on small spatial distance. Especially, JD population is not far from other populations; however, only one individual from other populations was found. This observation indicates some barrier to gene flow although under strong dispersal potential. Gene flow follows migration, but not necessarily, and can only occur after successful establishments of migrants and reproduction (Tigano & Friesen, 2016). Heterogeneous habitats cannot only limit dispersal of seeds or other propagules by mis-adaptation, but also may inhibit gene flow by natural selection (Sexton et al., 2014).
The environment of both the CRE and the QRE is highly heterogeneous not only at different sites of each estuary in salinity, tide, sediment charge, etc., but also at different altimetric positions of the same intertidal zone (Bu, 2013). Thus, we propose that the heterogeneity of environments at different scales limit successful establishment of migrants among some populations and also among clusters within population, and therefore constrain gene flow between different environments, which will result in a relatively high genetic differentiation.
F ST values support this suggestion. In our study, a relatively high genetic differentiation value (F ST = 0.1857) was detected although all locations hydrologic linked. The greatest interregional differentiation was found between TW and QD1 population (F ST = 0.3325), which was responded to the very long distance between them (~800 km), but when TW was excluded, the overall F ST was 0.1263, and also exhibited a relatively high level of genetic differentiation even in the absence of physical barriers at this regional scale.
Population differentiation represents a historic gene flow rather than current value (Ellstrand, 2014;Ouborg, Piquot, & Groenendael, 1999). In this vein, the coexistence of a relatively high genetic differentiation and frequent ongoing dispersal suggests that some processes, for example, selective pressure and genetic drift may restrict successful establishment of dispersal via seeds or tubers, thus reduce effective gene flow and also increase the probability of local adaptation which contribute to differentiation (Bolnick & Otto, 2013;Wang & Bradburd, 2014). Empirical studies have shown that most geographical peripheries are also ecologically marginal (Abeli, Gentili, Mondoni, Orsenigo, & Rossi, 2014), meaning more disadvantageous environmental conditions and relatively less migrants or/and gene flow. In the periphery of a range, rather than in central populations, due to higher pressure F I G U R E 5 IBE analyses for 13 populations in CRE and QRE when excluding TW population considering different dataset: all loci dataset (a), outlier loci dataset (b), and neutral dataset (c). The mantel test scatter plot shows the relationship between the pairwise genetic differentiation (F ST ) and the ecological distance between their locations. The ecological distance is measured by the salinity difference between sample locations, and the average surface salinity in the growing season (summer) of S. mariqueter was used (showed in Table 1) from genetic drift, reduced effective population sizes, found effects, and restricted gene flow (Dennenmoser, Nolte, Vamosi, & Rogers, 2013;Pandey & Rajora, 2012;De Ryck et al., 2016;Sexton et al., 2014), it tends to create genetic distinct clusters and promote genetic differentiation between populations. Besides, analysis of population structure in STRUCTURE considering only outliers showed a stronger differentiation pattern compared to those obtained with the entire dataset or neutral dataset, and the most differentiation was occurred between the peripheries (QD1, QD2) and central populations. These results co-contribute to the conclusion that local adaptation is an important driven force for genetic distinct clusters in peripheral populations.
This doubtless reflects barriers to dispersal, such as geographic distance and divergent selective regimes (Heydel et al., 2017;Neiva, Pearson, Valero, & Serrao, 2012), but also, in some case, population colonization history. The most common mode of isolations is IBD and IBE. The former results in gene drift, which is controlled by mutation and gene flow, and the latter causes different natural selections due to habitat heterogeneity. IBD effect was significant at large spatial scale, such as interregional scale (near 800 km between the two estuaries and Taiwan coastal areas) but nonsignificant at regional scale (approximate 20-300 km) when TW was excluded based on entire dataset. This result indicated that the S. mariqueter populations at the CRE or the QRE were deviate from the migration-drift equilibrium, which may be disrupted by the divergent selection process (Bradburd, Ralph, & Coop, 2013;Sexton et al., 2014;Shikano, Jarvinen, Marjamaki, Kahilainen, & Merila, 2015), the long-distance dispersal or stochastic colonization. However, for these thirteen populations, compared with nonsignificant relationship between genetic differentiation and geographic distance with entire or neutral dataset, a significant strong IBD effect was detected with 21 outlier dataset, reflecting the effect of spatial distance on isolation seems to be promoted by divergent selection. Our findings are consistent with Jones et al. (2013), based on the AFLP analysis; under weak selection, the strength of IBD was lower than under strong selection as a result of that strong IBD can confound landscape.
When natural selection mainly drives genetic differentiation, the correlation between genetic differentiation and some environment factors often can be found (Sexton et al., 2014;Wang & Bradburd, 2014). If isolation by environment plays a key role, we expect environmentally similar locations will also be genetically similar.
However, in this study, the genetic differentiation and the difference of one important environment factor, salinity, did not show a significant line-relationship without (mantel test, r = 0.018, p = 0.379) or with outlier dataset (mantel test, r = −0.049, p = 0.374), but showed a concave curve (r 2 = 0.11, p = 0.012 and r 2 = 0.04, p = 0.215, respectively). However, this result does not mean the habitat heterogeneity of S. mariqueter cannot result in the genetic differentiation of this species, because a significantly positive correlation with the salinity difference >4.0 ppt (r = 0.33, p = 0.021) was found. However, we cannot explain why the genetic differentiation was significantly and negatively correlated with the salinity difference when the latter was equal to or less than 4.0 ppt (r = −0.44, p = 0.015). A potential reason is that when the salinity difference was low, the effect of natural selection on the genetic differentiation might be diluted by other factors, such as gene flow and phenotypic plasticity (the ability to tolerate salt stress under lower salinity). Some population pairs (e.g., QD2 and HZ2, QD1 and YY) with similar salinity (lower salinity difference) but were located at different estuaries, respectively, also existed a high level of genetic differentiation (Table 2). In contrast, some neighbor populations (e.g., QD1 and QD2, HZ1 and HZ2, YY and BL) had a lower genetic differentiation though with a larger salinity difference. These results indicate the effects of salinity are overwhelmed by spatial distance or other ecological process in certain areas and also suggest that salinity is not the only selective factor driving population differentiation. Only with a large salinity difference, its effect on genetic differentiation can be observed.
Genetic structure in S. mariqueter also could reflect "isolation by population history" which has been evidenced by recently studies (Maas et al., 2018). This mode of isolation emphasize the importance of colonizers in shaping subsequent population genetic structure, also termed "historical priority effects" (Maas et al., 2018 founded the population could originate from multiple source populations rather than from a single source population. However, the successful colonizers are almost the first colonizers due to densitydependent ecological priority effects which may further mediated by evolution via adaptation to local conditions (Maas et al., 2018).
This observation has been evidenced, and the high plasticity would benefit colonization of new locations. Given that S. mariqueter populations were suffered repeated reclamation/colonization in certain areas, past colonization history could be a major factor influencing its population genetic structure. The hierarchical AMOVA revealed that the proportion of variation among eight groups with different colonization history was very low with all loci dataset, but the value arises to 43.60% considering only outliers. This result suggests that the selected loci are closely related with past colonization history; thus, the population differentiation scenario involving the colonization history is reinforced. For example, Hengsha Island (HS) and Jiuduan shoal (JD) are alluvial islands in the CRE only with a short history (~160 years for the former and <100 years for the latter, Yang, 1998;Hu, Cheng, Hu, & Hu , 2004). These two alluvial islands are very close, but their hydrologic conditions are distinctively different from each other (e.g., salinity in Figure 1). Thus, according to the result that C5 was limited at Jiuduan shoal, we have reason to believe C5 is an adaptive genetic type to Jiuduan shoal. This fact indicates that migrants from other populations have undergone a rapid adaptive evolution at this new alluvial island. We speculate that the similar evolutionary events are likely to happen at Hengsha Island before, an elder alluvial island, since C4 was also limited at Hengsha Island.
In such cases, early differentiation among sympatric colonization history populations could have been initiated through a reduction in gene flow among locally adapted groups occupying discrete environments. And also, the genetic clusters C10 and C11 were limited at QD1 and QD2, respectively, providing an explicit evidence that local adaptation is more likely to occur in peripheral populations, most likely as a consequence of historical arrival of founders with subsequent inbreeding and dispersal limitation due to the heterogeneous environment in combination with genetic drift effects.
To date, the patterns of genetic structure of estuarine populations with molecular markers cover a wide range of possible outcomes among and within species. Such observations may reflect varying degrees of isolation driven by selective and neutral processes. In this study, though S. mariqueter has a small distribution range with no geographical barrier, a relatively high level of genetic differentiation was found among populations of this species, and this differentiation was proved to be affected by the interaction between geographic distance and environmental variability, as well as population colonization history. The results may help to disentangle the relative contributions of the underlying processes in the formation of genetic structure among estuarine populations, especially for high plants. Besides, this study provides a signal of local adaptation occurred in peripheral populations although high migration in these estuaries, suggesting that effective conservation of S. mariqueter should include maintaining all populations cover its distribution range regardless of population size, thus promoting preservation.

ACK N OWLED G M ENTS
The authors would like to thank the two anonymous reviewers for the help provided in improving this manuscript. We thank Shimin Tang for the sampling work in field and thank Ziyi Chen for assistance with drawing Figure 1. This research was supported by the National Natural Science Foundation of China (31670223 and 31270407).

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interests. Y. and W. Z. wrote the manuscript.
Available data consists of "ece-Sm-AFLP data" including 0-1 matrix of entire loci, outliers, and neutral dataset in different sheets for all individuals, "ece-Sm-Mcheza" including the list of potential loci under selection with MCHEZA, DFDIST, and BAYESCAN software, respectively, and "ece-Sm-summary of FLOCK" including the assignment test results from FLOCK analysis.