Increasing planting density increases fruit mass and reduces the dispersal ability of a range‐expanding invasive plant, Mikania micrantha

Invasive plants may evolve a suite of distinctive traits during spread in the new range. Among these traits, dispersal ability is an important trait determining the invasion speed of exotic plants. There is evidence that higher dispersal ability is favoured at the invasion front, where population density may be low. However, no study has explicitly tested how planting density in a common garden affects the dispersal ability of invasive plants.


| INTRODUC TI ON
After introduction to the new range, invasive plants often spread rapidly (Horvitz et al., 2017).During range expansion, biotic and abiotic environmental conditions often change, and this may select a suite of new traits (Burton et al., 2010;Chuang & Peterson, 2016).
For example, at invasion fronts, population density and frequency often decrease, and suitable sites that may be occupied by invasive plants often increase (Huang et al., 2015;Robinson et al., 2023).
Under these conditions, traits such as high growth rate (Kilkenny & Galloway, 2013;Tabassum & Leishman, 2020), fast seed germination speed (Tabassum & Leishman, 2018), high reproductive allocation (Burton et al., 2010), and high seed dispersal ability (Huang et al., 2015;Monty & Mahy, 2010;Robinson et al., 2023) may be favoured.These traits may allow invasive plants to rapidly occupy suitable sites at invasion fronts.Among these traits, dispersal ability is an important trait determining the expansion rate and invasion speed of exotic plants (Burton et al., 2010).
Several studies found that diaspore dispersal ability increased toward range edges (Darling et al., 2008;Huang et al., 2015;Monty & Mahy, 2010;Robinson et al., 2023), but there was also evidence that dispersal ability did not increase toward range edges (Bartle et al., 2013).Part of these inconsistencies may be caused by phenotypic plasticity.Some studies measured dispersal ability using diaspores collected in the field (Bartle et al., 2013;Robinson et al., 2023;Tabassum & Leishman, 2018), and others measured dispersal ability using diaspores collected both in the field and in a common garden (Huang et al., 2015;Liu et al., 2021;Monty & Mahy, 2010).Patterns of dispersal ability in the natural field conditions may be influenced by multiple biotic and abiotic factors, which may differ among studies and populations, while patterns of dispersal ability in a common garden may indicate genetic differentiation among populations if populations perform differently under the same growth conditions.For example, for populations of Mikania micrantha in the Guangdong province of China, there was no relationship between dispersal ability and population cover in the natural environment, but there was a humped relationship between these two factors in a common garden, indicating that the genetic variation in dispersal ability was not apparent in natural conditions, probably because dispersal ability of different populations was affected by local environmental conditions differently (Huang et al., 2015).Thus, it is important to examine the patterns of dispersal ability both in the field and in a common garden.
The dispersal ability of wind-dispersed diaspores may be affected by factors such as fruit mass, pappus radius, and diaspore release height (Matlack, 1987;Wender et al., 2005;Zhu, Lukić, et al., 2023).Although it has been hypothesized that low population density at the invasion front may select for high dispersal ability (Burton et al., 2010), no study has explicitly examined if density can affect dispersal traits of invasive plants in a common garden.
A few studies on density-dependent dispersal found that increasing density can either increase or decrease diaspore dispersal.
Negative density dependence of dispersal may arise because high density decreased plant height (Wender et al., 2005) and the proportion of wind-dispersed achenes per seed head (Ruiz de Clavijo & Jiménez, 1998), and increased competition for dispersers (Jansen et al., 2014) and seed mass (Miao et al., 1991).Positive density dependence of dispersal may arise because high density increased seed release height and turbulence (Schurr et al., 2008) and reduced seed mass (Larios & Venable, 2015).In invasion ecology, previous studies exploring patterns of dispersal ability in a common garden grew one plant per pot (Huang et al., 2015;Liu et al., 2021;Monty & Mahy, 2010), but in the field, invasive plants probably grow under intermediate to high-density conditions, and it is still unknown how density affects dispersal ability of range-expanding invasive plants.
Although the relation between field population density and dispersal ability has been tested (Huang et al., 2015), this relation may be masked by factors other than density because in the field, competition (Turnbull et al., 1999), soil conditions (Bergholz et al., 2015) and pollinator availability (Huang et al., 2017) may all affect seed mass.
For example, low density may select for smaller seeds (Turnbull et al., 1999), but on the other hand, low density may increase the extent of pollen limitation, which may select for larger seeds (Huang et al., 2017).Thus, the effect of density on dispersal traits may not be detected in the field, and explicitly examining how planting density affects dispersal traits is needed to further our understanding of the factors driving the evolution of dispersal ability in range-expanding invasive plants.
In this study, we use a range-expanding invasive plant on Hainan island, M. micrantha, to explore how dispersal traits change with distance from the invasion centre and field population density both in the field and in a common garden, and how planting density affects dispersal traits.In another province (Guangdong) of China, the dispersal ability of M. micrantha increased toward the range edge in the natural environment but not in a common garden, and field density affected the dispersal ability in a common garden but not in the natural environment (Huang et al., 2015).This study did not examine if planting density affects dispersal ability in a common garden.On Hainan island with a smaller region and a shorter period since M. micrantha was first recorded, stand level biomass production of M. micrantha decreased toward the range edge under some soil conditions (Zhu, Huang, et al., 2023), but it is still unknown how dispersal ability changes with distance from invasion centre and population density.In the present study, we test the following hypotheses: (1) diaspore dispersal ability of M. micrantha increases toward the range edge because of either an increase in pappus radius or a decrease in fruit mass.(2) Increasing population density may reduce diaspore dispersal ability by increasing fruit mass.(3) Dispersal traits of M. micrantha in natural conditions may correlate more with those in a common garden when planting density is high than when planting density is low.The diaspore dispersal ability of edge plants may increase because edge plants less frequently encounters intraspecific and there are more suitable sites at the range edge, and previous studies found that invasive plants had greater dispersal ability toward range edges because of either an increase in pappus radius (Robinson et al., 2023) or a decrease in seed mass (Tabassum & Leishman, 2018).At the range centre, population density is often high and diaspore dispersal ability may be low (Burton et al., 2010), and thus increasing density may reduce diaspore dispersal ability by increasing fruit mass, which can also increase seedling competitive ability.The correlation between dispersal traits in natural conditions and those in a common garden when planting density is high may be because M. micrantha often grows under intermediate to high density conditions in the field, which may resemble the growth conditions in a common garden when planting density is high.

| Plant species and fruit collection
Mikania micrantha (Asteraceae) is a perennial herbaceous vine that originated from tropical Central and South America (Lowe et al., 2000).It was introduced into China in the late nineteenth century and became invasive in South China in the 1980s (Wang et al., 2003).Its first record on Hainan island was in 2003 in Haikou and Wenchang, and in 2010, it was mainly distributed in Haikou, Wenchang, Lingao, Chengmai, Dingan, Tunchang and Qionghai in the north of Hainan (Fan et al., 2010).Now it can be found in most counties and cities except those in the southwest.The weed has a rapid growth capacity, a high seed germination rate, and a strong fragment regeneration capacity, and can cover the entire habitats through stolon elongation and ramification after a short period of establishment (Liu et al., 2020;Zhang et al., 2004).M. micrantha dominates many roadsides, plantations, farmlands, orchards, and wastelands, and poses great threats to agricultural production and the environment (Fan et al., 2010).
In January and February 2021, we collected fruits of M. micrantha from 27 populations on Hainan island of China (Figure 1).Fruits were collected from 15 populations along the west highway from Haikou to Dongfang (western populations), and no plants of M. micrantha could be found in the south of Dongfang.Fruits were collected from 12 populations along the east highway from Wenchang to Wanning (eastern populations), and no plants of M. micrantha could be found in the south of Wanning.Each population was at least 6 km away from the others.Populations were mostly located in open areas exposed to full sunlight along roads, with some on trees and hillsides.
We collected fruits from multiple inflorescences within each population to reduce potential maternal effects.Fruits were stored at room temperature until usage.
We estimated the distance from the invasion centre, field population density, and frequency of occurrence of the 27 populations of M. micrantha.Because M. micrantha was first discovered in Haikou and Wenchang, and the weed also seriously outbroke in these cities and adjacent regions before 2010, we designated the sites of two large populations in Haikou and Wenchang as the invasion centre (Figure 1).The nearest road distance from each of the remaining 25 populations to the Haikou (for western pop- M. micrantha occurrence.Along two roads where M. micrantha has the potential to spread out, the frequency of occurrence was estimated as the percent of M. micrantha occurrence every 20 (±2) m in a total of 20 observations.The details of these estimations can be found in Zhu et al. (2022).Distance from the invasion centre was negatively correlated with frequency of occurrence but not population cover, probably because edge plants of M. micrantha could reach a high local density through horizontal growth after a short period of establishment (Zhu, Huang, et al., 2023).

| Experimental design
In early August 2021, we sowed the fruits into separate compartments in trays containing potting soil.Several fruits were sowed into a compartment, and the germinated seedlings were thinned into one plant per compartment in 3 weeks.After a month, in early September, 32 fairly strong and uniform seedlings per population were selected and transplanted into 3 L pots containing a 7:3 (v/v) mixed substrate of topsoil and sand.The topsoil was collected from local arable agricultural land, and it was sieved through a 6 mm mesh screen to remove large particles, roots, and stones before usage.No plants of M. micrantha were found in the soil collected.The substrate had a pH value of 5.07, an organic matter content of 10.39 g kg −1 , a total nitrogen content of 0.45 g kg −1 , a total phosphorus content of 0.49 g kg −1 , and a total potassium content of 4.9 g kg −1 .
The experiment was a randomized complete block design with three density treatments and four replicates.In one-third of the pots, a single seedling of M. micrantha was transplanted into a pot.
In another third of the pots, two seedlings of M. micrantha from the same population were transplanted into a pot.In the remaining third of the pots, five seedlings of M. micrantha from the same population were transplanted into a pot.There were a total of 324 pots, con-  S1).It was unknown why plants in nearly half of the pots did not set fruits.Possibly, there were genetic differences in allocation to reproduction among plants, and some plants were less visited by pollinators, particularly those setting fewer flowers.Fruits collected were stored at room temperature until measurement.

| Measurement of diaspore traits
In wind-dispersed plant species, the dispersal ability was determined by a combination of fruit mass and dispersal structure (Matlack, 1987).Previous studies used plume loading, which was determined by the ratio of fruit mass to pappus area, to indicate the dispersal ability of diaspores (Huang et al., 2015;Liu et al., 2021;Monty & Mahy, 2010;Robinson et al., 2023).We used area-mass ratio (AMR) to indicate diaspore dispersal ability: AMR = πR 2 /m, where R is the pappus radius and m is the fruit mass.A higher AMR value indicates a higher diaspore dispersal ability.A previous study used the ratio of pappus volume to achene volume (Cody & Overton, 1996), similar to our measurement.
For the fruits collected in the field in 2021, we measured the traits of 50 diaspores for each population.For each diaspore, the pappus diameter was measured to the nearest 0.01 mm.The pappus radius is one-half of the pappus diameter.Because M. micrantha diaspores were too light, the 50 diaspores were randomly divided into five groups, with each group having 10 diaspores.Ten fruits in each group were weighted to the nearest 1 μg after removing the pappus.The average fruit weight and pappus radius of 10 diaspores in a group were used to calculate AMR.For the fruits collected in each pot in the common garden experiment in 2022, we measured the traits of 20 diaspores.After measuring the pappus radius and removing the pappus, 20 fruits were weighted together to the nearest 1 μg.The average fruit weight and pappus radius of 20 diaspores were used to calculate AMR.

| Statistical analysis
The field data were analysed to examine if dispersal traits changed with distance from the invasion centre and population cover.A linear mixed-effect model was performed with diaspore dispersal ability as the response variable, distance from the invasion centre and population cover as the fixed factors, and population as a random factor.Similar models were performed on fruit mass and pappus radius.
The common garden data were analysed to examine if planting density affected diaspore traits, and if diaspore traits changed with distance from the invasion centre and population cover.A linear mixed-effect model was performed with diaspore dispersal ability as the response variable, planting density, distance from the invasion centre, population cover, the interaction between planting density and distance from the invasion centre, and the interaction between planting density and population cover as the fixed factors, and population and block as random factors.Similar models were performed on fruit mass and pappus radius to analyse which component underlay the changes in diaspore dispersal ability.When a significant effect of planting density was found, post-hoc Tukey tests were used to examine how the three planting density treatments differed from each other.
We further examined how the dispersal traits in the natural environment correlated with those in the common garden.A linear mixedeffect model was performed with diaspore dispersal ability in the common garden as the response variable, planting density, diaspore dispersal ability in the natural environment and their interaction as the fixed factors, and population as a random factor.Similar models were performed on fruit mass and pappus radius.An interaction between planting density and diaspore dispersal traits in the natural environment indicates that the correlation between the dispersal traits in the natural environment and those in the common garden was more evident under a particular planting density treatment.When a significant or marginally significant effect of dispersal traits in the natural environment was found, Pearson correlation analyses were performed to examine how the predictor correlated with diaspore traits in the common garden under each planting density treatment.
Because multiple tests may increase false discovery rates, for each P value obtained in the linear mixed-effect models, we applied a Benjamini-Hochberg (BH) correction to account for multiple hypothesis testing (Benjamini & Hochberg, 1995), and the BH adjusted p values were shown.The analyses were performed in JMP 13 (SAS Inst.).

| RE SULTS
Our first hypothesis is that the dispersal ability of M. micrantha increases toward the range edge because of either an increase in pappus radius or a decrease in fruit mass.We found that in natural conditions, distance from the invasion centre was unrelated to diaspore dispersal ability (AMR), fruit mass, and pappus radius (Table 1), indicating that dispersal traits did not change with distance from the invasion centre.In the common garden, distance from the invasion centre was also unrelated to diaspore dispersal ability, fruit mass and pappus radius (Table 1).
Our second hypothesis is that increasing population density may reduce diaspore dispersal ability by increasing fruit mass.We found that either in natural conditions or in the common garden, population cover was unrelated to dispersal ability, fruit mass and pappus radius (Table 1).In the common garden, planting density (i.e. the number of plants per pot) significantly affected dispersal ability and fruit mass but not pappus radius (Table 1).Post-hoc Tukey tests indicated that dispersal ability was lower and fruit mass was higher under the five plants per pot treatment than under the one plant per pot treatment (Figure 2).All of the interactive effects were nonsignificant (Table 1).
Our third hypothesis is that dispersal traits of M. micrantha in natural conditions may correlate more with those in the common garden when planting density is high than when planting density is low.We found that there was a marginally significant correlation between diaspore dispersal ability in the natural environment and that in the common garden (p = .054;Table 2).Although the interactive effect between planting density and dispersal ability in the natural environment was non-significant (Table 2), further Pearson correlation analyses found that diaspore dispersal ability in the natural environment correlated significantly with that in the common garden under the five plants per pot treatment (Figure 3), but not under the one and two plants per pot treatments (Figure 3).The effects of fruit mass in the natural environment, pappus radius in the natural environment, and their interaction with planting density were nonsignificant (Table 2).

| DISCUSS ION
We found that either in the natural environment or in the common garden, the dispersal ability of M. micrantha diaspores did TA B L E 1 Linear mixed-effect models for the effect of distance from invasion centre, population cover, planting density, the interaction between planting density and distance from invasion centre, and the interaction between planting density and population cover on dispersal-related traits (dispersal ability, fruit mass, and pappus radius) of Mikania micrantha.Benjamini not change with distance from the invasion centre.Previous studies found that in the field, dispersal ability increased (Darling et al., 2008;Huang et al., 2015;Robinson et al., 2023;Tabassum & Leishman, 2018) or did not change (Bartle et al., 2013;Monty & Mahy, 2010) with distance from invasion centre, and in common gardens, fewer but similar results were found (Huang et al., 2015;Monty & Mahy, 2010).The reason may be that many factors can affect dispersal traits, particularly in the field.For example, seed or fruit mass may be affected by plant density and interspecific competitors (Bergholz et al., 2015;Germain et al., 2019;Turnbull et al., 1999), abiotic conditions (Bergholz et al., 2015), and pollinator availability (Huang et al., 2017).As an outcrossing plant (Hong et al., 2007), if pollen limitation is more severe at the range edge on Hainan island, the fruit mass of M. micrantha may increase (Huang et al., 2017), but edge plants may also decrease fruit mass because of low population frequency, resulting in a non-significant relation.Overall, the relation between these biotic and abiotic factors and distance from invasion centre may depend the specific contexts of species and regions of study, and thus the relation between dispersal traits and distance from invasion centre differs among studies.
We found that in the common garden, fruit mass increased and diaspore dispersal ability decreased with increasing planting density.Previous studies found that larger seeds promoted competitive ability under high-density conditions (Tremayne & Richards, 2000;Turnbull et al., 1999), but there is also evidence that plants experiencing more competition produced smaller seeds that dispersed farther from their mother plant (Larios & Venable, 2015).Our results indicate that plants of M. micrantha under high-density conditions may produce larger fruits to promote competitive ability.
However, in the field, population cover did not correlate with fruit mass.This might also be because many other factors not examined affected fruit mass in the field, such as pollinator availability (Huang et al., 2017), interspecific competitors (Germain et al., 2019), and abiotic conditions (Larios & Venable, 2018), which might act in the opposite direction of population cover.In the common garden, population cover also did not correlate with fruit mass, indicating that plants from populations with high field density might not have evolved to produce larger seeds to increase their seedling competitive ability.However, in the Guangdong province of China, population cover correlated with fruit mass of M. micrantha in a common garden but not in the field, and fruit mass was highest and dispersal ability was lowest when field population cover was intermediate, possibly because when density was very high, high dispersal ability was favoured for the avoidance of competition (Huang et al., 2015).
Thus, the effect of field population density on dispersal traits differed among regions, even for the same species.
We found that dispersal ability in the natural environment correlated positively with that in the common garden.Although the interaction between planting density and diaspore dispersal ability in the natural environment was non-significant, the correlations were positive but significant only under the five plants per pot treatment.Thus, we only have tentative support for our hypothesis that  dispersal traits of M. micrantha in natural conditions may correlate more with those in the common garden when planting density is high than when planting density is low.The non-significant differences among planting density treatments might be because the sampled populations were not enough, the area of study is small and the inherent differences among populations were also small and difficult to detect.If more populations were sampled in a larger area of study (e.g.sampling populations in both Hainan and Guangdong of China), we might detect significant differences among planting density treatments.
There may be several biases in our experimental approach.First, there might be differences in environmental conditions (e.g.soil and climate conditions) among the field populations, and plants from populations with similar growth conditions in the field as compared to those in the common garden might grow better.However, the area of study is small, and the experiments were conducted at a site near the middle of the range centre and edge.Thus, this possibility was minimized.Second, maternal effects might play a role.However, the effects of the maternal environment may disappear in about 2 months (Van Kleunen & Fischer, 2003).In the present study, plants were grown for nearly 5 months before setting fruits.Also, all of the populations were located in habitats that received full sunlight, with typical laterite soils on Hainan island and away from the seashore and salinity stress, which likely weakened any environmental maternal effects.Third, even though fruit provisioning is likely maternaldetermined (Larios & Venable, 2015), free pollination might be a problem if paternal genes affect fruit mass.Fourth, the area of study (Hainan island) is small, which made it difficult to detect significant differences.Nevertheless, we found some significant differences.
One of the patterns we found was weak and tentative (i.e. the correlation between dispersal ability in the natural environment and that in the common garden differed among planting density treatments), and this may be viewed as a hypothesis that needs to be tested in invasive plants spreading in a larger region.Finally, we explained the effect of fruit mass on dispersibility through wing loading, but it was also possible that a lower fruit mass increased fruit number (Henery & Westoby, 2001), and the more fruits an individual produced, the more likely it was to have an extremely long-distance dispersal event.
The evolution of dispersal ability during range expansion in invasive plants has received much attention (Bartle et al., 2013;Huang et al., 2015;Monty & Mahy, 2010;Robinson et al., 2023;Tabassum & Leishman, 2018), but these previous studies did not examine how planting density in a common garden affects dispersal ability.We found explicit evidence that increasing density increased fruit mass and reduced dispersal ability in the common garden, a pattern that was not detected in the field.This indicates that the effect of population density on dispersal traits may be masked by other factors in the field.We also have a tentative pattern that diaspore dispersal ability in the natural environment correlated more with that in the common garden when planting density was high than when planting density was low.This may be because invasive plants often grow under intermediate to high density conditions in the field, and dispersal traits are selected under intraspecific competition.Taken together, we suggest that further studies exploring the patterns of dispersal traits in range-expanding invasive plants in a common garden should consider intraspecific competition.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors have no conflict of interest to declare.

DATA AVA
ulations) or Wenchang (for eastern populations) population was considered as the distance from the invasion centre.Since stolons of M. micrantha often intertwined with each other, it was difficult to determine plant numbers.Therefore, we used average percent cover with five 2 m 2 plots as a surrogate for population density.Because population cover may only indicate the number of empty sites at a local scale, we also estimated the frequency of F I G U R E 1 The distribution of the 27 populations of Mikania micrantha under study on Hainan island.The two black arrows indicate the range centre, and the two white arrows indicate the southwest and south edges.
sisting of 27 populations, 3 planting density treatments, and 4 replicates (blocks).The experiment was conducted in an open field at the Chinese Academy of Tropical Agricultural Sciences in Danzhou (19°51′ N, 109°49′ E).After nearly 3 months of growth, in late November when plants of M. micrantha began to flower but probably depleted the soil resources, we added 2 g slow-release fertilizer pellets (N:P:K 16:9:12, Osmocote, Scotts Miracle-Gro, Marysville, OH) into each pot.During the experiment, pots were watered when necessary, and within a block, pot positions were randomly changed every 2 weeks.A 2-m long stick was set up in each pot for M. micrantha to climb, which could also control for diaspore release height.A saucer was placed under each pot to prevent the penetration of M. micrantha roots into field soil.Free-pollinated flowers began to set fruits in late December.From January to March 2022, we collected mature fruits of M. micrantha twice a week.The experiment ended in March when all plants withered and yielded no more fruits.Plants in only 176 of the 324 (54.3%) pots set fruits (Table

F I G U R E 2
Effect of planting density on dispersal ability (A) and fruit mass (B) in the common garden.The mean values + SE are shown.TA B L E 2 Linear mixed-effect models for the effect of planting density, dispersal traits (dispersal ability, fruit mass, and pappus radius) of Mikania micrantha in the natural environment, and their interaction on dispersal traits in the common garden experiment.Benjamini-Hochberg adjusted p values are shown.Significant effects (p < .05)are shown in bold type.
This study was supported by the National Key R&D Program of China (2021YFC2600400), the National Natural Science Foundation of China (32071662, 32071520), Chinese Academy of Tropical Agricultural Sciences for Science and Technology Innovation Team of National Tropical Agricultural Science Center (CATASCXTD202311), and the Central Public-interest Scientific Institution Basal Research Fund (1630042023001).
I L A B I L I T Y S TAT E M E N T Data supporting the results are accessible in the Dryad Digital Repository (https:// datad ryad.org/ stash/ share/ j5JeZ HpYhi a4BKW 6t09b7_ AFC4_ bpx2g hAlxC 8mQ -eM) (DOI: 10.5061/dryad.fqz612jzj).F I G U R E 3 The correlation of dispersal ability between the natural environment and the common garden experiment when one (a), two (b), or five (c) plants were grown in a pot.Filled circles represent western populations, and open circles represent eastern populations.The mean values ± SE are shown.
-Hochberg adjusted p values are shown.Significant effects (p < .05)are shown in bold type.