Directional and stabilizing selection shaped morphological, reproductive, and physiological traits of the invader Solidago canadensis

Abstract Trait evolution in invasive plant species is important because it can impact demographic parameters key to invasion success. Invasive plant species often show phenotypic clines along geographic and climatic gradients. However, the relative contributions of natural selection and neutral evolutionary processes to phenotypic trait variation among populations of invasive plants remain unclear. A common method to assess whether a trait has been shaped by natural selection or neutral evolutionary processes is to compare the geographical pattern for the trait of interest to the divergence in neutral genetic loci (i.e., Q ST –F ST comparisons). Subsequently, a redundancy analysis (RDA) can facilitate identification of putative agents of natural selection on the trait. Here, we employed both a Q ST –F ST comparisons approach and RDA to infer whether natural selection shaped traits of invasive populations of Solidago canadensis in China and identify the potential environmental drivers of natural selection. We addressed two questions: (1) Did natural selection drive phenotypic trait variation among S. canadensis populations? (2) Did climatic, latitudinal, longitudinal, and altitudinal gradients drive patterns of genetic variation among S. canadensis populations? We found significant directional selection for several morphological and reproductive traits (i.e., Q ST  > F ST) and stabilizing selection for physiological traits (i.e., Q ST  < F ST). The RDA showed that stem biomass of S. canadensis was strongly positively correlated with longitude, while leaf width ratio and specific leaf area were significantly positively correlated with the mean diurnal range. Stem biomass had a strong negative correlation with annual precipitation. Moreover, height of S. canadensis individuals was strongly positively correlated with altitude and precipitation of the wettest quarter. A longitudinal shift in precipitation seasonality likely selected for larger stem biomass in S. canadensis. Overall, these results suggest that longitudinal and altitudinal clines in climate exerted strong selection pressures that shaped the phenotypic traits of S. canadensis.

assess whether a trait has been shaped by natural selection or neutral evolutionary processes is to compare the geographical pattern for the trait of interest to the divergence in neutral genetic loci (i.e., Q ST -F ST comparisons).Subsequently, a redundancy analysis (RDA) can facilitate identification of putative agents of natural selection on the trait.Here, we employed both a Q ST -F ST comparisons approach and RDA to infer whether natural selection shaped traits of invasive populations of Solidago canadensis in China and identify the potential environmental drivers of natural selection.We addressed two questions: (1) Did natural selection drive phenotypic trait variation among S. canadensis populations?(2) Did climatic, latitudinal, longitudinal, and altitudinal gradients drive patterns of genetic variation among S. canadensis populations?
We found significant directional selection for several morphological and reproductive traits (i.e., Q ST > F ST ) and stabilizing selection for physiological traits (i.e., Q ST < F ST ).
The RDA showed that stem biomass of S. canadensis was strongly positively correlated with longitude, while leaf width ratio and specific leaf area were significantly positively correlated with the mean diurnal range.Stem biomass had a strong negative correlation with annual precipitation.Moreover, height of S. canadensis individuals was strongly positively correlated with altitude and precipitation of the wettest quarter.A longitudinal shift in precipitation seasonality likely selected for larger stem biomass in S. canadensis.Overall, these results suggest that longitudinal and altitudinal clines in climate exerted strong selection pressures that shaped the phenotypic traits of S. canadensis.

| INTRODUC TI ON
Biological invasions by exotic plant species often reduce native biodiversity and disrupt ecosystem processes (Vila et al., 2011;Vitousek et al., 1997).Therefore, understanding the ecological and evolutionary processes that underlie phenotypic and genetic variation among invasive plant populations and the capacity of such populations to colonize a broad range of environments is a major goal in ecology.Phenotypic and genetic divergence among natural populations may result from natural selection processes and thus be adaptive (Linhart & Grant, 1996), or alternatively be caused by neutral (i.e., nonadaptive) evolutionary forces such as random colonization, spatially restricted gene flow, and genetic drift in peripheral populations (Campitelli & Stinchcombe, 2013;Endler, 1986).Populations can evolve adaptation to local environmental conditions if they possess sufficient heritable variation in phenotypic traits and natural selection is sufficiently strong (Hall & Willis, 2006).Indeed several invasive plant species have been shown to be locally adapted under a wide range of conditions (Oduor et al., 2016).However, the specific plant traits that are under natural selection and the traits that differ among populations due to neutral evolutionary forces have seldom been investigated (Exposito-Alonso et al., 2018).Trait evolution in invasive plant species is important because it can impact demographic parameters key to invasion success (Hodgins et al., 2018).
A common method to assess whether a phenotypic trait has been shaped by natural selection or neutral evolutionary processes is to compare the geographical pattern for the trait of interest to the divergence in putatively neutral genetic loci (Chun et al., 2009;Leinonen et al., 2008;Sun & Roderick, 2019;Xu et al., 2010).The level of genetic divergence in quantitative traits indicated by Q ST may be influenced by natural selection, as well as by neutral evolutionary forces (Spitze, 1993).In contrast, the level of population divergence in neutral molecular markers indicated by F ST is affected by neutral evolutionary forces and is often assumed to be selectively neutral (Wright, 1951).Thus, assuming no dominance or epistasis in a particular quantitative trait, comparisons of F ST and Q ST for any number of population pairs can yield one of three possible outcomes for the trait (Leinonen et al., 2008;Merilä & Crnokrak, 2001).First, if Q ST of the trait is greater than F ST (i.e., Q ST -F ST > 0), then the trait is under directional selection for different local optima across populations.Second, if Q ST = F ST (i.e., Q ST -F ST = 0), then the effects of natural selection and neutral evolutionary forces on the trait are not distinguishable and the trait is considered selectively neutral.Third, Q ST < F ST (i.e., Q ST -F ST < 0) indicates that a trait is uniform across populations as a result of stabilizing selection (Leinonen et al., 2008, Merilä & Crnokrak, 2001).Ideally, Q ST should be measured in a common garden setting to minimize or exclude differences in quantitative traits between plant populations that may arise from variation in ecological conditions between populations (Sun & Roderick, 2019).
The putative agents of natural selection causing adaptive differentiation among populations can be identified using ordination methods, including a redundancy analysis (RDA), which find significant associations between genetic polymorphism, phenotypic variation, and environmental variables (Capblancq et al., 2018;Capblancq & Forester, 2021).For instance, RDA revealed that partitioning of genetic variation in Helichrysum italicum (Asteraceae) was mainly associated with adaptation to temperature oscillations (Ninčević et al., 2021).In Arabidopsis thaliana (Brassicaceae), loci involved in adaptation to climate were identified using RDA (Lasky et al., 2012).
The RDA can also be used to derive an adaptive index that predicts the performance of individuals in different environmental conditions (Capblancq et al., 2018).Therefore, RDA can complement Q ST -F ST comparisons in the detection of natural selection and the environmental variables driving natural selection.
Low or absence of genetic variation in phenotypic traits that result from founder events and genetic bottlenecks may constrain the capacity of invasive plants to respond evolutionarily to novel selection pressures in the exotic range (Sakai et al., 2001).However, multiple introduction events followed by genetic admixture of populations introduced from different sources in the native range (Dlugosch & Parker, 2008;Lavergne & Molofsky, 2007;Oduor et al., 2015;Rius & Darling, 2014) and/or single introduction events from highly genetically diverse native-range source populations (Novak & Mack, 1993) may facilitate rapid adaptive evolution in invasive plants.
Solidago canadensis (Asteraceae) is a native of North America that is presently invasive in China.In China, the perennial forb was first introduced into the southeastern region for use in ornamental horticulture in 1935 and subsequently escaped to the wild (Dong et al., 2006;Du et al., 2017).Following escape from cultivation, it spread to colonize natural and agricultural ecosystems across southeastern China (Jin et al., 2020;Wan et al., 2020).
A single mother plant produces numerous small wind-dispersed seeds for long-distance dispersal although the species also reproduces clonally (Dong et al., 2006).Previous studies found that S. canadensis exhibited latitudinal and longitudinal clines in quantitative traits in China (Du et al., 2017;Li et al., 2016Li et al., , 2017)).
However, it remains unclear whether the clines in quantitative traits were an outcome of past natural selection or nonadaptive evolutionary forces.
Here, we employed both a Q ST -F ST comparisons approach and

| Field sampling
We sampled 14 S. canadensis populations (Figure 1; Table A1) in ruderal vegetation of China in October 2012.Twelve ramets that represented 12 families were selected in each of the 14 populations, for a total of 168 ramets.To minimize the chance of sampling individuals from the same maternal family more than once, within each population, we sampled ramets from sites that were separated from each other by at least 10 m.Shoots of the individual ramets were removed and the shoot bases with their attached rhizomes were dug up (see Li et al., 2016).The individual rhizomes were then wrapped in wet roll papers to keep moist and transferred to the laboratory until use in the common garden experiment as described below.To test for a geographical cline in phenotypic traits of S. canadensis among the 14 populations, we recorded the longitude, latitude, and altitude of the populations.

| Pre-experimental preparation of plantlets
To reduce environmental carry-over effects, the field-collected S. canadensis individuals were propagated vegetatively in a greenhouse for 4 months under uniform conditions.To achieve this, field-sampled rhizomes were individually grown in plastic pots (30 cm × 30 cm; diameter × depth) that contained a mixture of field soil, sand, and peat in the ratio of 6:3:1, respectively.The field soil was collected from Linhai City, Zhejiang Province, China.The soil mixture F I G U R E 1 Map of 14 Solidago canadensis populations in China that were studied.The population abbreviations and details are presented in Table A1.
had the following properties: pH 6.8, organic matter 27.66 g/kg, total nitrogen 361.0 mg/kg, available phosphorus 8.0 mg/kg, and available potassium 12.0 mg/kg.

| Common garden experiment
In April 2013, we established a common garden experiment on Linhai campus of Taizhou University (121°17′ E, 28°87′ N) to assess quantitative trait variation among S. canadensis individuals from the 14 populations.We obtained three similar-sized (ca.15 cm) plantlets for each of the 12 families in the 14 populations from the preexperiment cultivation of plants described above.The plantlets were grown in a common garden in the field in three blocks.Thus, the total number of experimental plants was 504: 14 populations × 12 individuals (representing 12 maternal families) per population × 3 blocks.Within a block, the 12 individuals were planted in 12 separate plots that each measured 1.5 m × 1.5 m.The individual plants were grown 30 cm apart from each other.Throughout, the experimental plants received only rain-fed water and were not fertilized.

| Phenotypic trait measurement in the common garden
One month after transplant (21-27 May 2013), we took in situ measurements of photosynthesis on the third fully expanded leaf from the top using a portable photosynthesis meter (LI-6400 XT, Li-COR, Inc., Lincoln, NE, USA).The measurements were taken between 9:00 and 11:00 AM under the following conditions: a photosynthetically active radiation of 1400 μmol m −2 s −1 , leaf temperature of 25°C, CO 2 concentration of 400 ppm, and relative humidity of 70%.For each plant, we recorded net photosynthetic rate (P n ), intercellular CO 2 concentration (C i ), stomatal conductance (G s ), and transpiration rate (T r ).These measurements were taken six times on different dates, and the average of the six rounds of measurements was used in the statistical analyses described below.Seven months after transplant (November 2013), we measured various morphological, physiological, and reproductive traits of the S. canadensis individuals.Individual plant height, leaf length (L), and leaf width (L) for the third leaf per plant were measured to a precision of 0.1 cm.We used the L and W dimensions to compute the L/W ratio.Chlorophyll content of the third leaf was measured using a portable chlorophyll meter.We also measured basal stem diameter to an accuracy of 0.01 cm.After 97% of the plants had flowered, we counted the total number of inflorescences per individual plant.Then after the plants had matured, we obtained a total seed count and 1000-seed weight per plant.We also measured the total leaf area per plant using a WinFOLIA computer image analysis system (Regent Instruments Inc, Quebec, Canada) for use in the computation of specific leaf area (SLA).The above-ground plant parts were then harvested and separated into shoots, leaves, and seeds.Biomass of the stem, leaf, and seed was measured after the samples had been dried at 80°C for 72 h.We measured biomass using an electronic balance to an accuracy of 0.01 g.To obtain the total vegetative biomass per S. canadensis individual, we summed up the oven-dried stem and leaf biomass of the individuals.We computed SLA as the ratio of leaf area to leaf dry biomass.
A year later (November 2014), we repeated measurements of the same traits above from the same S. canadensis individuals that were not harvested.After the measurements had been taken, the whole root system for each S. canadensis individual was dug up and washed free of soil particles under running water.The roots and shoots were then dried at 80°C for 72 h and weighed individually to the nearest 0.01 g.

| Leaf sampling
In October 2012, we collected leaf tissues from 30 S. canadensis individuals per population in the same 14 populations that we sampled for rhizomes as described above (Figure 1; Table A1).The sampled leaves were immediately immersed in self-sealing plastic bags that contained silica gel and then stored in the laboratory at room temperature until use in DNA extraction.

| Microsatellite analysis
In November 2012, genomic DNA was extracted from 0.1 g of each leaf sample using a modified sodium dodecyl sulfate protocol on a FastPrep-24 Automated Lysis and Homogenization System (MP Biomedicals, Santa Ana, CA, USA).Total DNA concentration was determined with a NanoDrop 2000 Lite Spectrophotometers (ThermoFisher Scientific, Inc. Rockford, IL, USA).The DNA samples were then diluted to 10 ng/μL and stored at −20°C until use in simple sequence repeats (SSR) analyses.Five primer pairs (synthesized by Boshang Biotechonology Co., Ltd. in China) were used in SSR amplifications (Table A2) (Wieczorek & Geber, 2002).We ran PCR in a 20 μL volume that was made up of 1 × PCR reaction buffer, 1.5 mM Mg 2+ , 4 ng template DNA, 0.2 μM each of forward and reverse primers, 0.2 mM 4 × dNTP mixture, and 1 U Taq polymerase (Promega Cooperation, Madison, WI, USA).We performed PCR amplifications using a PTC 220 Peltier Thermal Cycler (Bio-Rad Laboratories, Hercules, CA, USA) with the following settings: denaturation at 95°C for 5 min, followed by 34 cycles of 30 s at 95°C, 30 s at 58°C (−1°C per cycle), 45 s at 72°C, with a final elongation of 7 min at 72°C.We analyzed the PCR products using a Fragment Analyzer™ Automated CE System (Advanced Analytical Technologies, Inc, Ankeny, IA, USA) with an 80 cm-long capillary column.We used a DNF-900 35-500 bp ds DNA Reagent Kit in the analysis.We genotyped DNA fragments | 5 of 16 DU et al.
using PROSize® 2.0 Data Analysis Software based on the elution time compared with a size standard.

| A test for phenotypic trait variation among the S. canadensis populations
We ran general linear models to test whether the different S. canadensis populations exhibited significant variation in phenotypic trait expressions in the common garden experiment.In the models, identity of S. canadensis population was included as a fixed term, while family and plot were treated as random terms.
We specified morphological, physiological, and reproductive traits of S. canadensis individuals as response variables.Statistical significance of the factors was tested using F-statistic.The models were run separately for measurements that we took in year 1 and year 2 of the experiment to assess whether the plant traits would be expressed consistently during the lifetime of S. canadensis.The statistical analyses were performed in SPSS v. 19.0.

| Analysis of molecular data
We used POPGENE v. 1.31 to compute the number of loci (N), percentage of polymorphic loci (P %), number of different alleles (N a ), number of effective alleles (N e ), Nei's gene diversity per locus (H S ), and Shannon's information index (I).We performed an analysis of molecular variance (AMOVA) (Excoffier et al., 1992) to assess how genetic variation was distributed among and within the populations.The AMOVA was performed in GenAlEx v. 6.501 (Peakall & Smouse, 2005).The significance of genetic differentiation was tested using F-statistic with 1000 random permutations.

| Comparisons between Q ST and F ST
We compared quantitative genetic (Q ST ) and neutral molecular marker (F ST ) differentiation among the 14 populations to infer the relative contributions of natural selection and random genetic drift to clinal variation in quantitative traits of S. canadensis.We computed the among-population coefficient of gene differentiation (F st ) based on microsatellite DNA using GenAlEx v. 6.501 (Peakall & Smouse, 2005).We computed the Q ST index based on the physiological, growth, and reproductive traits of the different S. canadensis populations by applying the formula wherein V pop and V ind represent variation in the quantitative traits among and within populations, respectively (Spitze, 1993).We then computed the difference between the resultant Q ST value for each trait and the common F ST of 0.078 (i.e., Q ST -0.078) (see results below on population genetic structure) to infer the effects of natural selection vs random genetic drift.et al., 2005).A redundancy analysis is an expansion of multiple linear regression, in that multiple explanatory variables are used to explain multiple response variables (Legendre & Legendre, 2012).Collinear variables (BIO1, BIO4-BIO11, BIO13-BIO17, BIO18 & BIO19) were removed to minimize collinearity, as assessed with variable inflation factor (VIF max <10).Moreover, we did not use molecular data in the RDA to test for associations between genetic polymorphism, phenotypic variation, and environmental variables because the VIF factor for all the five loci was greater than the threshold value of 10.Thus, altitude, latitude, longitude, BIO2, BIO3, BIO12, and BIO16 were retained in the subsequent RDA.We applied forward selection to the data using the ordiR2step vegan function with the following stopping criteria: variable significance of p < .01 using 1000 permutations, and the adjusted R 2 of the global model.We standardized the predictors (i.e., subtracted the mean and divided by the standard deviation) to ensure that the variable units were comparable (Legendre & Legendre, 2012).Tests of significance of the global RDA model (containing all the significant variables), individual RDA axes, and individual explanatory variables were performed using a permutational analysis of variance with 999 permutations.The RDA was performed using the Vegan package v. 2.6-4 (Oksanen et al., 2022) in R v. 4.2.3 (R Core Team, 2023).

| Phenotypic trait variation among S. canadensis populations
In the first and second years of the experiment, the S. canadensis individuals exhibited a significant among-population variation in the growth and reproductive traits (Table A3).However, the S. canadensis individuals did not show significant among-population variation in physiological traits except for relative chlorophyll content (Table A3).Moreover, height and stem diameter of the S. canadensis individuals did not vary among the populations in the second year (Table A3)

| Genetic diversity and structure in S. canadensis populations
We scored a total of 49 bands from the five SSR primers.The populations exhibited variability in the genetic diversity indices P, Na, Ne, H, and I (Table 1).The highest genetic diversity was found in the WH population (P = 100%, N a = 2.000, N e = 1.347,H S = 0.242, and I = 0.395), while the lowest genetic diversity (P = 91.84%,N a = 1.918,N e = 1.296,H S = 0.205, and I = 0.340) occurred within the WZ population (Table 1).The AMOVA revealed that most of the genetic variation occurred within the S. canadensis populations (92.186%) (Table 2).There was low genetic differentiation among the S. canadensis populations (F ST = 0.078).Estimated gene flow (N m ) among the populations was 2.955.

| Phenotypic versus genetic differentiation
In the first year of the experiment, Q ST values of all the morphological and reproductive traits and two physiological traits (i.e., chlorophyll content and net photosynthetic rate) were larger than the common F ST value of 0.078 (i.e., Q ST -F ST > 0; Figure 2).However, Q ST values of the physiological traits stomatal conductance, intercellular CO 2 concentration, and transpiration rate were lower than the common F ST value (i.e., Q ST -F ST < 0; Figure 2).In the second year of the experiment, Q ST values of leaf length, leaf width, L/W ratio, SLA, number of inflorescences, seed number, 1000-seed weight, and relative chlorophyll content were larger than the common F ST value (i.e., Q ST -F ST > 0), while Q ST values of the other quantitative traits were lower than F ST (i.e., Q ST -F ST < 0; Figure 2).

| Environmental predictors of phenotypic trait variation in S. canadensis populations
In the RDA for traits that were measured in the first year of the experiment, the global RDA model was significant (F = 7.60; p = .001)and each of the three environmental variables (longitude, BIO2, and BIO12) that were included in the model were significant (p < .05).
The three environmental variables explained 4.5% of the variation in phenotypic traits of S. canadensis.The first axis (RDA 1) was significant (F = 18.66; p = .001)and primarily associated with annual precipitation (BIO12), while longitude and mean diurnal range (BIO2) similarly drove variation along the axis but in the opposite directions (Table 3).The second axis (RDA 2) was only marginally TA B L E 1 Genetic diversity indices of Solidago canadensis at the population and species level.significant (F = 4.14; p = .083)and driven by the mean diurnal range (BIO2) and annual precipitation (BIO12) (Table 3).The third axis was not significant (F = 0.016; p = .998).The stem biomass of S. canadensis was positively correlated with longitude, while the L/W ratio and SLA were positively correlated with mean diurnal range (BIO2) (Figure 3a).However, stem biomass had a negative correlation with annual precipitation (BIO12) (Figure 3a).The three traits had relatively higher loadings on RDA 1 compared to the other traits (Table A4).For the S. canadensis traits that were measured in the second year of the experiment, the overall model was significant (F = 18.18; p = .001)and each of the three environmental variables, including altitude, BIO3, and BIO16, that were included in the model was significant (p < .05).The three environmental variables explained 10.9% of the variation in phenotypic traits of S. canadensis.
The first axis (RDA 1) was significant (F = 54.48;p = .001)and primarily driven by altitude and precipitation of the wettest quarter (BIO16) (Table 4).Isothermality (BIO3) had a relatively weak correlation (0.24) with RDA 1 (Table 4).Height of S. canadensis at maturity was positively correlated with altitude and precipitation of the wettest quarter (BIO16) (Figure 3b).Plant height had a relatively higher loading on RDA 1 compared to the other traits (Table A5).The third axis was not significant (F = 0.002; p = 1.00) (Table 4).However, none of the physiological traits of S. canadensis were significantly correlated with environmental variables (results not shown).

| DISCUSS ION
Population differentiation in growth and reproductive traits as well as chlorophyll content was evident in our range-wide comparison of the invader S. canadensis in China (Table A3).Because S. canadensis individuals from the various populations were grown under uniform conditions in a common garden, intraspecific variation is likely due to genetic variation in the traits.The Q ST -F ST comparisons found evidence of directional selection for morphological and reproductive traits and stabilizing selection for physiological traits (Figure 2), while the RDA found significant associations between phenotypic variation and environmental variables (Figure 3).Because clines in phenotypic traits with an underlying genetic basis potentially indicate natural selection (Alberto et al., 2013;Campitelli & Stinchcombe, 2013), it is plausible that natural selection is the principal force that shaped the genetic architecture of S. canadensis populations.

| Q ST -F ST comparisons suggest that directional and stabilizing selection shaped phenotypic traits of invasive populations of S. canadensis
For all the reproductive traits and some morphological and physiological traits, Q ST values were greater than F ST values in both years TA B L E 2 Results of analysis of molecular variance (AMOVA) that tested for genetic variation among and within 14 populations of Solidago canadensis.  of the experiment (i.e., Q ST -F ST > 0) (Figure 2), which suggests that directional natural selection, rather than neutral evolutionary processes, favored those traits.In contrast, for some physiological traits, Q ST values were lower than F ST values (i.e., Q ST -F ST < 0) in both years of the experiment (Figure 2), which indicates that mean values of those traits were shaped by stabilizing selection.Molecular phylogeography data suggest that populations of S. canadensis in China were founded through multiple introduction events from the North American native range (Wan et al., 2020;Zhao et al., 2015).
The multiple introduction events likely caused high-standing genetic diversity within the introduced S. canadensis populations, which then enabled them to respond evolutionarily to selection imposed by the novel environmental conditions.Moreover, S. canadensis has undergone demographic and range expansion in China (Wan et al., 2020).
Demographic and range expansions can cause an accumulation of rare alleles and low-frequency mutations (Fu, 1997;Tajima, 1989).As S. canadensis has invaded China for ca.80 years, which is equivalent to several overlapping generations, it is also likely that over time, demographic and range expansions increased standing genetic diversity that natural selection acted upon.
One striking result in the present study was low genetic differentiation (F ST = 0.078) among S. canadensis populations despite divergence in growth and reproductive traits among the populations.A possible explanation for this finding is that there was a high degree of gene flow, which led to admixed populations.Solidago canadensis reproduces both clonally and sexually (Li et al., 2017).It is likely that gene flow occurred among widely spaced S. canadensis populations through long-distance pollen flow and seed dispersal (the estimated gene flow was high at 2.955).Studies on other plant species have found evidence of diversifying selection and local adaptation in spite of strong gene flow among populations of those species (Chun et al., 2011;Saether et al., 2007;Sanou et al., 2005), which may be an outcome of positive covariance of allele frequencies among populations (Latta, 2003).It is also possible that the relatively weak population genetic structure that we detected based on neutral genetic markers in the present study may not be strongly indicative of population differentiation in adaptive genes and quantitative traits (sensu Leinonen et al., 2008).Microsatellites have been used to obtain an unbiased estimate of genomic diversity based on the assumption that they are selectively neutral (Väli et al., 2008).However, microsatellites can grossly underestimate significant differences in genetic diversity and structure among populations (Väli et al., 2008).
Therefore, future studies may employ high-throughput DNA sequencing techniques that can identify and score genetic markers that are randomly distributed across the entire genome to more vigorously assess genetic diversity and structure among the invasive populations of S. canadensis.
The Q ST -F ST values for some morphological and physiological traits of S. canadensis were positive in the first year but negative in the second year of the experiment (Figure 2), which suggests that those traits displayed phenotypic plasticity.Between-year variation in phenotypes may be an adaptation strategy to changing environmental conditions (Ledig et al., 2015).Although we did not measure environmental variables in the experimental site over the 2 years of study, it is likely that there was between-year variation in climatic TA B L E 3 Results of a redundancy analysis that tested for correlations between the environmental variables latitude, longitude, altitude, and 19 bioclimatic variables and phenotypic traits of Solidago canadensis from 14 populations that were measured in the first year of the experiment.elements and soil conditions at the study site to which some traits of S. canadensis displayed plastic responses.Phenotypic plasticity is thought to play an important role in biological invasions by allowing species to survive and reproduce under a wide range of environmental conditions (Richards et al., 2006).For some exotic plant species, both local adaptation and phenotypic plasticity may jointly facilitate invasion.For instance, both local adaptation and phenotypic plasticity were linked to successful invasion by Wedelia trilobata (Asteraceae) (Si et al., 2014), A. artemisiifolia (Xiong et al., 2023), and Prunella vulgaris (Lamiaceae) (Godoy et al., 2010).It is likely that phenotypic plasticity and local adaptation both underlie the invasiveness of S. canadensis.
Similar to the present findings, other studies found that natural selection-shaped traits of invasive plant species.For instance, diversifying selection caused divergence in reproductive traits among invasive populations of A. artemisiifolia (Chun et al., 2011).In L. salicaria, time to first flower and size at reproduction were influenced by stabilizing selection (Colautti & Barrett, 2010).Directional selection for flowering time caused adaptive evolution in the invader of agroecosystems Raphanus raphanistrum (Brassicaceae) (Ashworth et al., 2016).These findings broadly support the idea that contemporary evolution of local adaptation can determine invasive species' success (Hodgins et al., 2018).

| Putative agents of natural selection inferred from a redundancy analysis
The results of a RDA showing a positive correlation between stem biomass of S. canadensis and longitude and a negative association between stem biomass and annual precipitation (BIO12) (Figure 3a) indicate that longitudinal cline in precipitation pattern selected for variable stem biomass.Generally, precipitation varies considerably with changes in longitude in many parts of the world, including China (Wang et al., 2020).We found that mean historical annual precipitation (BIO2) for the period 1950-2000 had a negative linear relationship with longitude in the study area (Figure A1).Therefore, it is likely that the relatively drier soils at higher longitudes selected for S. canadensis genotypes with larger stems.A previous study found that S. canadensis developed larger roots under drought conditions (Du et al., 2017), likely as a strategy to extract more water from a drier soil.
Another study found a longitudinal cline in relative biomass allocation to roots and shoots and efficiency in water utilization in S. canadensis (Li et al., 2016).Similarly, Capparis spinosa (Capparaceae) developed larger stems when grown under drought conditions (Gan et al., 2013).
Thus, S. canadensis individuals with larger stems may become adapted and more successful invaders at higher longitudes where there is reduced precipitation.Such individuals may also have stronger negative ecological impacts given that size of invasive plants can be positively associated with their impacts (Parker et al., 1999;Sun et al., 2013).
The present finding of a longitudinal cline in stem biomass is in accord with those of other studies that observed longitudinal clines in plant phenotypes and phenology (Wang et al., 2020).For instance, the mean leaf size and the internode length of Cynodon dactylon (Poaceae) had a significant negative relationship with longitude (Wang et al., 2020).East-to-west longitudinal genetic clines in morphological traits and tolerance against environmental conditions were observed in Abies sachalinensis (Pinaceae) (Kitamura et al., 2020).In another study, North American populations of A. thaliana showed longitudinal clines in flowering time (Samis et al., 2012).
Overall, these studies suggest that different combinations of environmental factors along a longitudinal gradient may impose variable selection pressures on plants leading to longitudinal differentiation among plant populations.
That the leaf traits L/W ratio and SLA were significantly positively correlated with the mean diurnal range (BIO2) (Figure 3a) supports the suggestion that spatial variation in SLA often reflects plant responses to variation in temperature (Rosbakh et al., 2015).It is generally accepted that SLA is correlated with the temperature conditions of a habitat (Poorter et al., 2009).In particular, plant species with low SLA will mainly be found in the colder part of a temperature gradient, whereas species with relatively high SLA values should be largely restricted to sites exposed to higher temperatures (Poorter et al., 2009).Due to the positive linear relationship that SLA has with the relative growth rate of plants (Poorter et al., 2009), it is plausible that S. canadensis populations that occur in warmer regions will have higher SLA and consequently higher relative growth rates and fecundity than S. canadensis populations that occur in colder regions.
Our results showing that height of S. canadensis at maturity was strongly positively correlated with precipitation of the wettest quarter (BIO16) (Figure 3b) are in agreement with those of a global synthesis, which found that plant height was positively correlated with precipitation in the wettest month (Moles et al., 2009).These results suggest that precipitation in the wettest quarter might be a strong selective agent on S. canadensis height.However, the present finding that height of S. canadensis at maturity was positively correlated with altitude (Figure 3b) contradicts results of the global synthesis, which found that altitude was a poor predictor of plant height (Moles et al., 2009).As plant height can influence plant competitive ability above-ground (Gioria & Osborne, 2014), the present results suggest that S. canadensis populations that occur at higher altitudes may have faced strong above-ground competition from the resident flora, which may have led to the evolution of taller stature in S. canadensis.

| Perspectives for future study
The RDA models explained only 4.5% and 11.6% of the variation in phenotypic traits in the first and second year of the experiment, respectively, which indicates that the environmental variables that were included in the models were not the only likely forces of natural selection in the populations under study.It is possible that other environmental variables that we did not measure also imposed selection on S. canadensis traits.Apart from climatic factors, soil biota can drive natural selection on plant traits (Lau & Lennon, 2011).A recent study suggests that older populations of S. canadensis interact with more beneficial and fewer pathogenic soil microorganisms than younger populations of S. canadensis in China (Oduor et al., 2022).
Hence, future studies may explore whether differences in microbial community structure influence selection on S. canadensis traits.
Assessments of geographic clines in the traits of invasive plant populations are often used to infer adaptive evolution in invasive plants although this approach may sometimes overestimate the full extent of local adaptation (Frenne et al., 2013).Geographical clines can in principle also be explained by neutral evolutionary processes (Endler, 1977).Moreover, separate introductions from native populations to similar biogeographic regions in the introduced range can result in parallel clines in the absence of selection.Therefore, approaches that include careful sampling with respect to environmental gradients, genome-wide analysis of neutral genetic markers, and reciprocal transplants in multiple locations among different climatic regions may be employed in future studies to more rigorously test for adaptive genetic variation in fitness-related traits among S. canadensis populations.TA B L E A 3 Results of general linear models that tested the effects of Solidago canadensis population and family identities on phenotypic trait expression by S. canadensis individuals that were grown in a common garden over 2 years (Year-1 and Year-2)..17

Year
Note: In the models, S. canadensis population identity was included as a fixed effect-independent variable, while family identity and plots were included as random effect variables.Statistical significance of the factors was set at p < .05.Values of some traits were measured in the first or second year only.
RDA to infer whether natural selection caused trait divergence among invasive populations of S. canadensis in China and identify the potential environmental drivers of natural selection.We addressed two questions: (1) Did natural selection drive phenotypic trait variation among S. canadensis populations?(2) Did climatic, latitudinal, longitudinal, and altitudinal gradients drive patterns of genetic variation among S. canadensis populations?

F
Comparisons of quantitative and genetic differentiation (Q ST -F ST values) among Solidago canadensis individuals from 14 populations that were grown in a common garden experiment over 2 years (Year 1 & Year 2).
Redundancy analysis (RDA) ordinate plots showing relationships between longitude, altitude, and the bioclimatic variables BIO2, BIO3, BIO12, and BIO16 and traits (stem biomass, leaf/width (L/W) ratio, specific leaf area (SLA), and height) of Solidago canadensis individuals from 14 populations that were measured in year 1 (a) and year 2 (b) in a common garden experiment.Environmental variables are represented by blue arrows.Length of arrows represents the relative importance of that variable.Loadings of S. canadensis traits on the RDA axes showing the direction of correlations (positive or negative) between the traits and environmental variables are shown in Appendices A4 and A5.

Source of variation d.f. SS MS Variance components Percentage of variation (%) p
Abbreviations: d.f., degree of freedom; SS, sum of squares; MS, expected mean squares.
Sample locations, climatic and geographical information of 14 Solidago canadensis populations in China.Simple sequence repeat (SSR) primer sequences that were used in the present study.
Our complementary use of Q ST -F ST comparisons and RDA revealed that both directional selection and stabilizing selection shaped phenotypic traits of S. canadensis populations, and that climatic factors, H., & Li, J.-M. (2023).Directional and stabilizing selection shaped morphological, reproductive, and physiological traits of the invader Solidago canadensis.Ecology and Evolution, 13, e10410.https://doi.org/10.1002/ece3.10410| 13 of 16 DU et al.A PPE N D I X A TA B L E A 1