Phenotypic responses to light, water, and nutrient conditions in the allopolyploid Arabidopsis suecica and its parent species A. thaliana and A. arenosa: Does the allopolyploid outrange its parents?

Abstract Polyploid species possess more than two sets of chromosomes and may show high gene redundancy, hybrid vigor, and masking of deleterious alleles compared to their parent species. Following this, it is hypothesized that this makes them better at adapting to novel environments than their parent species, possibly due to phenotypic plasticity. The allopolyploid Arabidopsis suecica and its parent species A. arenosa and A. thaliana were chosen as a model system to investigate relationships between phenotypic plasticity, fitness, and genetic variation. Particularly, we test if A. suecica is more plastic, show higher genetic diversity, and/or have higher fitness than its parent species. Wild Norwegian populations of each species were analyzed for phenotypic responses to differences in availability of nutrient, water, and light, while genetic diversity was assessed through analysis of AFLP markers. Arabidopsis arenosa showed a higher level of phenotypic plasticity and higher levels of genetic diversity than the two other species, probably related to its outbreeding reproduction strategy. Furthermore, a general positive relationship between genetic diversity and phenotypic plasticity was found. Low genetic diversity was found in the inbreeding A. thaliana. Geographic spacing of populations might explain the clear genetic structure in A. arenosa, while the lack of structure in A. suecica could be due to coherent populations. Fitness measured as allocation of resources to reproduction, pointed toward A. arenosa having lower fitness under poor environmental conditions. Arabidopsis suecica, on the other hand, showed tendencies toward keeping up fitness under different environmental conditions.


| INTRODUC TI ON
Polyploidization is recognized as a driving force for angiosperm diversification and speciation (Wendel, 2015). Polyploid species possess more than two sets of chromosomes, acquired either by intraspecific genome doubling (autopolyploidy) or by merging of genomes of different species through hybridization (allopolyploidy). A newly formed polyploid combines genes from two individuals, and this opens for hybrid vigor and masking of deleterious, recessive alleles (te Beest et al., 2012). Neo-or subfunctionalization may lead to genetic innovation (Hegarty & Hiscock, 2008;Lynch & Force, 2000), and a high gene redundancy suggests that polyploids could withstand inbreeding and population bottlenecks better than their diploid counterparts (Song et al., 2012;te Beest et al., 2012). Following this, polyploids may harbor high levels of genetic diversity, especially if there are multiple origins of the polyploid species. The genomic changes and increased genetic diversity may lead to altered morphology, physiology, and ecology (Parisod et al., 2010). Generation of new expressional patterns and novel epigenetic variation could also contribute to this (Chen, 2007;Comai, 2005). Following this, polyploids could have an adaptive advantage in new or changing environments, giving the polyploid hybrid species a higher fitness than either of the parental species. The effects are expected to be more pronounced for allopolyploid species. At the same time, there are genetic forces associated with polyploidization that may be detrimental. For example, polyploidization is a process that changes the genome abruptly in just one generation, and this may lead to a notoriously unstable genome. A result may be unstable mitosis and meiosis, giving aneuploid cells, and problems with gene expression due to development of uneven relationships between genes and regulatory factors (Comai, 2005). Epigenetic re-modeling could also cause instability in newly formed polyploids (Comai et al., 2003).
The physiological effects of polyploidization are relatively little explored, but Soltis et al. (2016) pinpoint cases that are relatively well described: genome doubling within each cell lead to larger cells, again leading to larger stomata and vascular cells, higher photosynthetic rate and gas exchange due to the larger stomata, and increased susceptibility to drought due to larger xylem vessels, leading to differences in stress resistance between polyploids and their parents (Soltis et al., 2016). Furthermore, it is assumed that higher genetic diversity constitutes a foundation for higher fitness, hence if a polyploid species has a higher level of genetic diversity than its parent species this may be advantageous (Reed & Frankham, 2003).
The adaptive advantage of an allopolyploid may result from phenotypic plasticity and/or fitness homeostasis (Godfree et al., 2017;Scheiner, 1993;Stevens et al., 2020). Phenotypic plasticity is the ability to exhibit a wide range of phenotypes across varying environmental conditions (Bradshaw, 1965;Schlichting, 1986). However, this may not necessarily imply higher fitness. Fitness homeostasis is the ability to keep fitness as equal as possible between varying environmental conditions (Hulme, 2008;Richards et al., 2006). It is proposed that high phenotypic plasticity provides wider possibilities to adapt to new environments (Davidson et al., 2011;Sultan, 2000), while high fitness homeostasis could imply better abilities at coping with and adapting to stressful environments (Godfree et al., 2017;Hulme, 2008;Richards et al., 2006;Stevens et al., 2020). Summed up, a theoretical framework for a possible positive relationship between polyploidy and abilities to adapt is established (Flagel & Wendel, 2009). Nevertheless, conclusive results for polyploid species outcompeting their parent species in their ecological niche or expansion to niches unavailable to the parent species have been difficult to establish (Soltis et al., 2016). Note that in this work, we use phenotypic plasticity on species and population level, whereas phenotypic plasticity in its most strict sense refers to the ability of a single genotype to respond differently to various environments.
To investigate physiological effects of polyploidization and its possible role in giving adaptive advantages compared to parent species, we use the hybrid complex consisting of the allopolyploid species Arabidopsis suecica (Fr.) Norrl. ex O.E. Schulz and its two parent species, A. thaliana (L.) Heynh. and A. arenosa (L.) Lawalrée. Arabidopsis suecica originates from a hybridization between the mostly diploid A. thaliana and the mostly autotetraploid A. arenosa (Jakobsson et al., 2006;O'Kane et al., 1996), possibly within the eastern parts of A.
thaliana's native range (Beck et al., 2008;Novikova et al., 2017). The formation of the species probably occurred through the fertilization of a female, unreduced A. thaliana gamete with a normal, male A. arenosa gamete from a tetraploid A. arenosa (Jakobsson et al., 2006;Novikova et al., 2018;Säll et al., 2003). It is believed to have originated between 12,000 and 300,000 years ago, somewhere south of its present native distribution in Sweden and Finland (Jakobsson et al., 2006;Novikova et al., 2017;Säll et al., 2003). Specifically, Novikova et al. (2017) suggest multiple origins, likely after the last glaciation maximum in Eastern Europe or central Eurasia. Burns et al. (2021) conclude that the process leading to the species A. suecica has been gradual, and they find no evidence of genome shock.
All three species are winter annuals, forming an overwintering basal rosette of leaves in the autumn and a flowering stem in the following spring (Baskin & Baskin, 1983). While A. arenosa is a strictly outcrossing species, A. thaliana and A. suecica are self-fertilizing species, which set seeds regardless of whether they are pollinated or not (Säll et al., 2004). All three species prefer dry habitats. In Norway, A. thaliana often grows in dry meadows, rock crevices and on ledges, while the other two typically are found on sandy soils-often close to road verges and along railways (Elven, 2005).
In this paper, we investigate variation in phenotypic variables in response to different nutrient, light, and water treatments as well as genetic diversity in the diploid A. thaliana and A. arenosa and its polyploid daughter species A. suecica. Specifically, we ask the following research questions: 1. Does the allopolyploid A. suecica show more genetic diversity and larger phenotypic plasticity and/or fitness homeostasis than its parent species? 2. Are there any relationships between genetic diversity, fitness homeostasis, and phenotypic plasticity in the study species? 2 | MATERIAL S AND ME THODS 2.1 | Study area Seeds of A. thaliana, A. suecica, and A. arenosa were sampled from 10 wild populations in three different geographic areas in SE Norway ( Figure 1). The number of sampled populations per species was three for A. thaliana and A. suecica, and four for A. arenosa. All species were sampled in each geographical area ( Table 1).
For each population, 20 randomly chosen individuals were sampled. The life stage of the collected plants was not standardized. If a population consisted of less than 20 individuals, as many individuals as possible were sampled. The lowest number of individuals sampled per population was 6. The plants were dried, and the seeds extracted and transferred to 2-ml tubes (Eppendorf, Hamburg, Germany).

| Measurements of ploidy level and chromosomal numbers
In order to ensure that all populations of the study species had the expected chromosomal numbers and ploidy levels, DNA content was measured with flow cytometry. Seeds from each of the populations grown in the experiment were sown in pots and grown to a size where harvesting was permissible. For each population, three individuals were selected for harvesting. Leaves corresponding to a total area of 1-2 cm 2 were harvested. Leaf harvesting was not standardized. Flow cytometry was performed, and DNA ratios were obtained by G. Geenen, Plant Cytometry Services (Schjindel, The Netherlands). Diploid A. thaliana from the "Columbia" line was acquired from the University of Tromsø and provided as a control sample along with the experimental samples. For internal control Ilex crenata "Fastigiata" was used.

| Analysis of phenotypic responses to different treatments
Seeds from the 10 sampled populations were grown under controlled environmental conditions in a growth chamber. To assess whether different species react differently to varying environmental conditions, eight different treatments were applied in a 2 3 factorial design. These treatments consisted of all different combinations of dry and wet water conditions, rich and poor nutrient conditions, and high and low light conditions. Five replicates were grown per treatment combination. This adds up to 10 populations × 8 treatment combinations × 5 replicates = 400 plants grown in total. Pictures of the experimental design are shown in Figure 2.
Eight trolleys with a size of 100 × 60 cm were covered first with plastic, and then with felt mats to transport the water evenly over the whole trolley. 50 circular 8C-101 flowerpots with a diameter of 8 cm (Billund Potter, Billund, Denmark) were placed on each trolley. 400 flowerpots were prepared overall. Each flowerpot was filled with Gartnerjord soil (Tjerbo Torvfabrikk, Rakkestad, Norway) consisting of 86% Sphagnum peat, 10% sand and 4% granule clay. One trolley was assigned to each treatment combination. For each population, seeds from all sampled individuals were mixed on a white paper sheet, then several seeds were drawn randomly and sown in each pot. The different populations were distributed randomly within each trolley. 9 l of water were applied to each trolley after sowing.
The seeds were stratified for four days in 4xC and 24 h darkness.
Then, conditions were changed to 20°C/17°C day/night tempera- When the seedlings had reached the stadium where primary leaves started to become visible, they were thinned so that one plant remained in each flowerpot. For some populations, transplantations between pots were done. The plants were allowed one week of optimal growth conditions before treatments were applied. Nutrient treatment was applied by giving nutrient solution made from 1.25 ml Superba NPK 14-4-21 + mikro (Nordic Garden AS, Stokke, Norway) and 1 l water to each of the rich nutrient trolleys once per week, while no nutrients were applied to the poor nutrient trolleys. The water used for making the nutrient solution was included in the total amount of water given to the plants, as described below. that passes through. The amount of light below the fabric was measured to be 80-90 µmol m −2 s −1 , equivalent to a reduction of 60-70%.
Water treatment was initially performed by applying 2 l of water three times a week to the wet condition trolleys, and 1 l of water three times a week to the dry condition trolleys. The light-shading fabric was found to heavily reduce evaporation from the low-light trolleys, so to obtain similarity in water conditions between the lowlight and the high-light trolleys, the low-light trolleys were watered once a week, applying 2 l of water to the wet condition trolleys and 1 l of water to the dry condition trolleys.
Vernalization was initiated 39 days after sowing (35 days after germination conditions were initiated). Growth conditions were changed to 4°C constant temperature and an 8/16 h light/dark cycle.
Since growth was low during vernalization, nutrients were applied on average every third week, in the same doses as described above.
The amount of light in the growth chamber was reduced to avoid the plants dying from light stress. The amount of light was measured to be 125-135 and 27-32 µmol m −2 s −1 for the high light and low light treatments, respectively. For the high-light trolleys, watering was done by applying 2 l of water once a week to the wet condition trolleys, and 1 l of water once a week to the dry condition trolleys.
For the low-light trolleys, watering was done by applying 2 l of water once every third week to the wet condition trolleys, and 1 l of water once every third week to the dry condition trolleys.
Based on findings in Lewandowska-Sabat et al. (2012), vernalization conditions were kept for 9 weeks. At the end of vernalization, 102 days after sowing, growth conditions were changed to 23°C/20°C day/night temperature and 16/8 h light/dark cycle to allow flowering. Nutrients, light and water treatments were the same as before vernalization. These conditions were kept for 33 days, when the growth experiment was ended.
During the whole experiment, the trolleys were moved around within the chamber, and pots were moved around on the trolleys periodically to avoid edge effects. This was performed haphazardly.

| Measurements of phenotypic variables
Phenotypic variables were measured at different times. At the initiation of vernalization, three different variables were measured: TA B L E 1 List of populations where seeds were sampled, specifying locality codes, locality names, what geographical areas the different localities belong to, species, collection date, latitude in degrees north (Lat (°N)), and longitude in degrees east (Long (°E))    from T-SFRO3. The remaining variables were superimposed onto a biplot of the first two NMDS axes.

| Data analysis
To assess the effect of treatment and species on the different variables, linear mixed effects models or generalized linear mixed effects models were run, using the R packages nlme (Pinheiro et al., 2017) and lme4 (Bates et al., 2007). Water was initially regarded as giving no effect, but an assessment of model selection criterions (data not shown) found that including water gave slightly better models. For the final models, a single factor was constructed, where each level corresponded to a specific combination of light, nutrients, water, and species for a total of 2 3 × 3 = 24 levels. Population was added as a random effect.
To run the models, the response variables biomass and number of flowers were log-transformed. Days to flowering and number of leaves were considered count data, and Poisson models were used for assessing them. Number of flowers could also be considered count data, but as the numbers were so large, we concluded that the variable should be considered as continuous instead of discrete. Table 2 gives an overview of transformation of variables, and which models that were run for each response variable. Both linear mixed effects models and generalized linear models with Poisson family were fit using maximum likelihood. All models were checked for assumptions of normality and equality of variance between groups by conferring Q-Q and residual plots. Poisson models were checked for over-and underdispersion.
Post hoc testing of the models was done by applying general linear hypothesis methods from the R package multcomp (Hothorn et al., 2008). These methods give a generalization of the Tukey post hoc test that can be used on unbalanced designs. To model reaction norms for each species to the applied treatments, common letter displays based on multiple comparisons between all pairs were constructed.
To check whether the species reacted differently to environmental stress and showed differences in phenotypic plasticity, confidence intervals for estimated differences between high and low levels of treatments were constructed. Nutrient effects were assessed separately for each combination within the two other treatments. A corresponding approach was used for light and water effects. To adjust for multiple comparisons, a confidence level of 99% were used in the post hoc analyses.
Coefficients of variation were calculated for each response variable. The measurement gives an indication on the amount of phenotypic plasticity (Schlichting & Levin, 1984;Sultan, 2001  A linear mixed model, and general linear hypothesis post hoc methods as described above, were applied to the comparison model.
A confidence level of 99% was used for the post hoc analyses. The theory was that a higher value of C means relatively more allocation of resources to fitness, and vice versa. A smaller difference in C between good and poor environmental conditions could be interpreted as a higher degree of fitness homeostasis.

| Population structure and genetic diversity
The dataset was examined for population structure using the software Structure, a software that can allocate individuals to genetic groups based on AFLP data (Pritchard et al., 2000). Analyses were run using Structure ver. 2.3.4 at the Lifeportal, University of Oslo (https://www.uio.no/engli sh/servi ces/it/resea rch/hpc/porta ls/lifep ortal Paleo ntolo gía Elect rónica), with 10 6 iterations and a burn-in of 10 5 iterations. An admixture model was used; meaning that for each individual different parts of the genome are allowed to descend from different groups. Linkage between markers was not considered. A minimum of one population (K = 1) and a maximum of 9 populations (K = 9) was allowed per analysis. For each value of K, 10 independent runs were done. The results were assessed using the R functions in Structure-sum (Ehrich, 2011). The number of clusters was chosen after an evaluation based on the following criteria: (1) all runs gave similar results, (2) similarity coefficient close to 1.0, (3) highest possible ln P (data) and (4) highest possible ΔK (Evanno et al., 2005;Pritchard et al., 2000). Structure analysis was run for each species. In addition, an analysis incorporating all individuals was run in order to see whether the different species clustered separately.
To visualize the clusters in a multidimensional space, principal coordinate (PCO) analysis was run on a distance matrix calculated with Dice's coefficient of similarity (Dice, 1945). The PCO analyses were run in PAST ver. 2.17c (Hammer et al., 2001), and scores for the two first components were extracted and plotted in R. PCO analysis was run for all species together, and separately for each species.
To assess and compare the diversity of the sampled populations and species, 95% confidence intervals for Nei's Genetic Diversity (Nei, 1987) was constructed using bootstrapping over 1000 replicates with the R functions in AFLPdat (Ehrich, 2006). Analyses of molecular variance (AMOVA) (Excoffier et al., 1992) were performed in Arlequin ver. 3.5 (Excoffier & Lischer, 2010). This was done for each species based on groups inferred from the original populations. If the number of clusters inferred from Structure came out differently from the original populations, an additional AMOVA was run based on the inferred clusters (unless the inferred number of clusters was one).

| Comparison of genetic diversity and phenotypic plasticity
To assess possible relationships between genetic diversity and phenotypic plasticity, a Mantel test was run to compare Euclidean distance matrices calculated from Nei's Genetic Diversity and Coefficients of variation for all phenotypic variables. The test was run on the eight populations where results from both growth experiments and genetic analyses were available. A corresponding test was also done with a phenotypic distance matrix calculated from coefficients of variation where each variable was scaled to unity. The scaling was done by dividing all values in the variables by the highest value in the variable.

| Measurements of ploidy level and chromosomal numbers
The populations used in the experiment mainly showed the expected chromosomal numbers and ploidy levels: 10 chromosomes/ diploid for A. thaliana, 32 chromosomes/tetraploid for A. arenosa and 26 chromosomes/tetraploid for A. suecica. There were two exceptions: One individual showed a lower chromosomal number than expected in A. arenosa. This might be due to aneuploidy, but it might also be due to errors in the measurement. One alleged individual A.
thaliana showed a chromosome number that one would expect for A.
arenosa. This is probably due to a confusion of samples. The samples used for flow cytometry were not used in the other experiments, hence this likely did not affect the rest of the results.

| Multivariate analysis of phenotypic variables
Clustering of phenotypic variables in response to light and nutrients show that A. thaliana was separated from the other two species (Figure 3). There was a weak trend that A. suecica occupied the space between A. arenosa and A. thaliana. There was not a clear clustering between treatments, although rich nutrients and high light tended to cluster on the top left side of the plot. This indicates that rich nutrients and high light were associated with taller plants, higher biomass and more flowers. For water, no clustering tendencies were observed. For all variables included in the NMDS, R 2 were >0.50 and p-values were <0.001.

| Analyses of phenotypical variables
The general trend was an increase in the measurements of phenotypic variables from low light, poor nutrients via low light, rich nutrients/high light, poor nutrients to high light, rich nutrients (Figure 4). Overall, very few significant differences were observed between dry and wet conditions (Figure 3), meaning that the water treatment provoked few effects in the experiment.
Arabidopsis arenosa seemed to exhibit larger phenotypic plasticity when it comes to height and number of flowers, while A. suecica differed from the other species when it comes to leaves at bolting ( Figure 4; Table 3). The variation was large on the population level, but the general trends from the species level were reflected in the populations.
In the analysis of the comparison variable C (which compares response variables connected to fitness with other response variables connected to phenotypic plasticity), the species had similar values within the high light treatment (Figure 6). Within the low light treatment, A. arenosa seemed to exhibit lower values than the other two species, although this trend was not significant. Still, this could indicate that A. arenosa allocates fewer resources to keep up fitness under low light treatments than the other two species. Responses in the C variable between high and low treatment levels were not significantly different between species within any of the treatments (data not shown). in A. thaliana. The genotyping error was calculated to be 3.30%.

| Population structure
The results from Structure showed a clear clustering of the different species (Figure 7a). This was confirmed by the PCO (Figure 8a), where we also see that A. suecica was placed in the middle of the first axis between its parent species. On the population level, A.
arenosa showed a clear population clustering both in Structure and PCO (Figures 7b and 8b). In A. suecica no clear population structure was found (Figure 7c Figure 9 shows the graphs that underlie the decisions on optimal numbers of clusters.

| Genetic diversity
The A. arenosa populations exhibited significantly higher genetic diversity than the A. suecica and A. thaliana populations (Figure 10), and this was confirmed on the species level ( Figure 11). Two of the A.

| Analysis of molecular variance (AMOVA)
The AMOVA showed that the between-populations percentage of variation in the AFLP markers was 27.5% in A. arenosa, 34.5% in A.
suecica and 58.8% in A. thaliana when considering the original populations (Table 4). When considering K = 2 clusters in A. thaliana, the between-population percentage of variation was still quite high (48.6%).

| Comparison of genetic diversity and phenotypic plasticity
The Mantel test showed a significant positive correlation between phenotypic plasticity measured as coefficients of variation and the measurements of Nei's Genetic Diversity (Table 5). This indicates that there is a relationship between higher genetic diversity and larger phenotypic plasticity on the population level among the study species. The corresponding test done with a distance matrix created from coefficients of variation scaled to unity also yielded a significant positive correlation (Table 5).

| Arabidopsis suecica is intermediate to its parent species in both pheno-and genotype
The species are clearly separated based on genetic analyses and there is no sign of hybridization between them (Figures 7a and   8a). Arabidopsis suecica is intermediate between A. thaliana and A. arenosa (Figure 8a), reflecting its status as an allopolyploid offspring species. This is also seen in the phenotypic analysis (Figure 4), although the tendency is not as clear as it is for the genotypic.
No clear population structure could be found in A. suecica, even though inbreeding species are expected to exhibit more genetic structure among populations than outcrossing species (Loveless & Hamrick, 1984). All investigated populations of A. suecica were found along railway lines, indicating that most of the Norwegian A.
suecica consists of a large, coherent population. Arabidopsis thaliana is at least partly indigenous in Norway, whereas A. arenosa and A. suecica are introduced. It is plausible that the railway populations of A. suecica in Norway have a single, recent common ancestor, and that there has not been enough time for a clear population structure to develop. The AMOVA results show that the between-population variation was much higher in A. thaliana than in the two other species (Table 4), probably due to A. thaliana having developed genetic differentiation over a longer time period than the other species.
Most of the phenotypic variables show common trends among the species (Figures 4 and 5). However, A. arenosa differed from the other species in several response variables when it came to response to light treatment. The reason for this might be found in the species' life histories. Arabidopsis arenosa requires insect pollination, while the two other species are selfers (Säll et al., 2004). The amount of light could thus have less impact on the selfing species' ability to reproduce successfully. Nutrient availability gives similar responses in all three species. In the wild, the species tend to grow in sandy, nutrient-poor soil (Elven, 2005). The similar response patterns suggest that they thrive under poor conditions but have the capability to behave opportunistically when nutrient availability improves. Few responses were observed between differing watering regimes, possibly because the applied treatments did not concur with what could be classified as high and low levels of water for Arabidopsis species.

| Arabidopsis arenosa responds most strongly to changing environments and shows the highest level of genetic diversity
We hypothesized that the allopolyploid A. suecica would show larger phenotypic plasticity in changing environments than its parent species.
Except for number of leaves at bolting in response to nutrients, this was not the case, and it rather seems that A. arenosa shows the greatest phenotypic plasticity (Figures 4 and 5; Table 3). As an allopolyploid

| The relationship between phenotypic plasticity, genetic diversity, and fitness
We identified a positive relationship between genetic diversity and phenotypic plasticity in response to environmental variation (Table 5). AFLP markers are often presumed to be neutral and mostly within non-coding regions (Hufbauer, 2004), and thus high diversity in AFLP markers should not necessarily confer a higher expressional diversity. However, Caballero et al. (2013) investigated distribution of AFLP markers in the genome for several eukaryotic species.
They found that for the EcoRI/MseI system up to 87% of the markers were within coding regions, indicating that AFLP markers are not necessarily neutral. The positive relationship between phenotypic plasticity and genetic diversity might be explained by the species' life histories. Both the larger plasticity, mainly as the result of more extreme responses to the light treatment, and higher genetic diversity could be due to A. arenosa's outcrossing, insect-pollinated nature (Kilkenny & Galloway, 2008;Schoen & Brown, 1991).
It has been postulated that plasticity is favored if the environment is variable (Callaway et al., 2003;Charmantier et al., 2008;Lande, 2009;Valladares et al., 2014), and following this phenotypic plasticity could play an important role in a species' ability to expand and adapt to novel environments (Davidson et al., 2011;Via et al., 1995). For instance, phenotypic plasticity increased a species' tolerance to herbivore attacks (Agrawal, 2000). This may suggest that species with higher plasticity should show higher fitness, but it has proven difficult to establish a relationship between those two (Davidson et al., 2011;Hulme, 2008). Further on, fitness homeostasis (the ability to keep up reproduction when conditions get worse) is not necessarily favored by a high degree of plasticity in traits directly connected to fitness (Hulme, 2008 It should be noted that fitness is difficult to measure directly. Ideally, an experiment would run over several generations to quantify fitness, but this was not possible within the timeframe of this project. This study was restricted to measuring certain variables that  to fitness or not connected to fitness. The choice of variables connected to fitness in this experiment (flowers and biomass) was chosen based on methods used by Davidson et al. (2011). More flowers confer possibilities for higher offspring production. When it comes to biomass, Weiner et al. (2009) advocates an allometric relationship between biomass and reproduction. In that perspective, total biomass could be viewed as a good fitness proxy. Fitness is also a question of viability. Seed production and viability was not assessed in the experiment, but it should be included in future experiments.

| CON CLUS IONS
In an environment that is changing faster than ever, understanding the mechanisms involved in range expansion and adaptation is important. Both phenotypic plasticity, genetic diversity, and polyploidy has been suggested as driving forces for introduction and adaptations to novel environments. This led us to investigate if the allopolyploid species A. suecica had more genetic diversity, a higher degree of phenotypic plasticity and a higher degree of fitness than its parent species A. thaliana and A. arenosa. We were unable to reveal any advantages for the allopolyploid A. suecica in our experiment in neither genetic diversity nor phenotypic plasticity. On the contrary, we found that A. arenosa had the highest level of diversity and phenotypic plasticity, probably due to its outcrossing nature. Across all species, we did find a positive relationship between phenotypic plasticity and genetic diversity, but this was not related to ploidy.
A. arenosa showed tendencies towards having the lowest degree

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data, that is, AFLP markers and phenotypic measurements, can be found in the DRYAD repository at https://doi.org/10.5061/ dryad.dv41n s216.