No evolutionary change in the mating system of Solanum rostratum (Solanaceae) during its invasion in China

The mating system of flowering plants plays a key role during the process of invasion. Evolution from outcrossing to selfing can allow rapid regeneration of a population after long-distance dispersal by providing reproductive assurance. Solanum rostratum is a self-compatible annual herb that exhibits a high level of outcrossing in its native populations. However, the mating system of invasive populations of S. rostratum has never been assessed. Here, we investigated the mating system based on 11 microsatellite loci and explored ecological factors that may influence the outcrossing rate among 10 invasive populations of S. rostratum in China. We found that the mean outcrossing rate was 0.69 ± 0.12 (range 0.49 to 0.83) with multiple paternity within progeny arrays (average effective number of sires = 7.86), which suggests a mixed mating system dominated by outcrossing. Combined with the uniformly high outcrossing rate (0.70 ± 0.03) previously reported in its native range, these results indicate that there has been no evolutionary shift in mating system during the invasion in China by S. rostratum. There were no relationships between outcrossing and population size, population density, altitude, latitude or longitude. Furthermore, high outcrossing of S. rostratum in China may be facilitated by enantiostyly and heteranthery.

The mating system of plant is the pattern in which gametes combine in plant populations 1 , and can influence the frequency of genes within the population 1,2 . As a result, the mating system of invasive plant species plays a fundamental role in determining the genetic diversity and genetic structure of a population and is closely related to adaptive evolution and invasion 1,2 . When alien flowering species are introduced to a new region, these species often occur at low density and experience increased pollen limitation because of the reduction of potential mates and suitable pollinators [3][4][5] . Such negative conditions can persist for a long time after the initial founder event resulting in Allee effects 6,7 , which can limit the spread of invaders. Therefore, natural selection may favor an association between selfing and colonizing ability 1,8 , as selfing can allow a rapid build-up of a population after long-distance dispersal by providing reproductive assurance according to Baker's Law [9][10][11] . Efforts to explore the mating system of invasive species constitute an important task to better understand the biology of plant invasions and to predict microevolutionary changes in various environments.
Recent studies have used molecular techniques to evaluate selection on mating systems during biological invasion by comparing outcrossing rates between invasive and natural populations 1 . The evolutionary transition from outcrossing to selfing has occurred in some invasive plants, such as Eichhornia paniculata and Eichhornia crassipes [12][13][14][15] , owing to the stochastic loss of mating types in these species. There are also some cases that have no evolutionary shift in the mating system during invasion, such as those involving Ambrosia artemisiifolia and Senecio inaequidens 11,16 , owing to self-incompatibility. However, the evolution of mixed mating strategies, a mixture of self-fertilization and cross-fertilization, may be more complex 2,17 . And studies on the maintenance of mixed mating systems in hermaphroditic invasive plants have only been conducted relatively recently.
For animal-pollinated plants, there are evidences that the mating system of them is influenced not only by biotic factors, such as floral display, pollinator type and population size [17][18][19] , but also by environmental conditions, including geographic location and habitat fragmentation 20,21 . However, to the best of our knowledge, only a handful of studies 18 have assessed the impact of biotic factors and environmental conditions on mating patterns of invasive species. Thus, it is necessary to investigate the factors that influence the mating system of invasive plants.
Solanum rostratum (Solanaceae) is a buzz-pollinated, self-compatible annual herb with hermaphroditic flowers. The species is believed to be native to Mexico and the USA 22 , but it has spread to China, Europe, Russia, Canada, South Korea, and Australia [22][23][24] . S. rostratum was found for the first time in Liaoning province in 1981, and has spread across a large area in North China during the last 36 years 24 . Populations of S. rostratum occur in disturbed habitats such as roadsides, riversides, sides of railways and abandoned fields in China, which is as same as the habitats in its origin area. The main pollinators of S. rostratum are bees in the genera Bombus and Xylocopa in both invasive and native populations. The bright yellow flowers of S. rostratum in China exhibit monomorphic enantiostyly and strong herkogamy 25,26 as in its native region. The species has two mirror-image floral morphs alternating along the inflorescence, with flowers that present style and pollinating anther opposite each other, situated either right or left of the floral axis. The stamens of S. rostratum are divided into four small yellow feeding stamens and a single tinged brown large pollinating stamen according to their different functions 27 . The four centrally located feeding stamens serve to attract and reward bees, and the single pollinating stamen, which is deflected to either the right-or left-hand side of the flower opposite the style, contributes disproportionately to pollen reaching the stigmas of other flowers 27 . These special floral traits promote outcrossing by reducing selfing between the same flower morph 28,29 ; in its native Mexican range, S. rostratum is highly outcrossed 30 . However, the mating system of invasive populations of S. rostratum has never been assessed.
Here, we investigated the mating system of S. rostratum among ten invasive populations in China and the effect of population size and other ecological factors on mating in S. rostratum. The present study had three objectives concerning the reproductive and invasive biology of S. rostratum. First, we assessed the mating system of S. rostratum by quantifying its mating system parameters using microsatellite markers. Second, we compared the outcrossing rate of Chinese populations with that of native populations in Mexico. Finally, we evaluated relationships between outcrossing rates and population size and other ecological factors such as altitude and latitude. Our study provides the first detailed evidence of the mating system of S. rostratum in its invasive Asian range.

Results
Mating system. The Fig. 1). The average multilocus outcrossing rate (t m ) across Chinese populations was 0.69 ± 0.12 (mean ± SD), ranging from 0.492 ± 0.225 in population YG to 0.834 ± 0.064 in population CY. The independent samples t-test showed that the outcrossing rates of Chinese populations had no significant difference with the previously reported outcrossing rates of Mexican populations 30 (P > 0.05; Fig. 2). These results indicate that there has been no evolutionary shift in the mating system during the invasion of China by S. rostratum.
Estimates for the outcrossing rate between related individuals (t m −t s ) were significantly greater than zero (range 0.009 ± 0.027 to 0.137 ± 0.023), except for population NZ (−0.002 ± 0.05), suggesting some degree of biparental inbreeding in most Chinese populations.
Estimates for the outcrossing rate between related individuals (t m −t s ) were significantly greater than zero in both Chinese populations (range 0.009 ± 0.027 to 0.137 ± 0.023, except for population NZ −0.002 ± 0.05) and Mexican populations (range 0.034 ± 0.024 to 0.072 ± 0.028) 30 , suggesting some degree of biparental inbreeding in most populations. The correlation of paternity within sibships for multilocus cases (r pm ) was low in all populations (0.13 ± 0.09), suggesting that approximately one in ten maternal siblings were expected to have the same pollen donor parent. Moreover, the effective number of male parents per progeny arrays was 7.86 among the ten studied populations (N ep = 7.86), which was lower than that of native populations in Mexica (N ep = 8.97) 30 .
There were no significant relationships between multilocus outcrossing rates and population size, population density, altitude, latitude or longitude (P > 0.05). However, the smallest population YQ (population size = 60) had a high outcrossing rate (0.783 ± 0.07). There was no evidence of a relationship between biparental inbreeding and Chinese populations showed relatively lower genetic diversity (Table 2)

Discussion
Evaluating the mating system of invasive species is an important task to better understand the conditions in which self-or cross-fertilization or a mixed mating strategy is favored in invasive populations 1,20 . In many cases, we expect an evolutionary shift towards a higher frequency of self-fertilization, because selfing provides reproductive assurance under the initial founder event, resulting in Allee Effects. Solanum rostratum is a self-compatible annual herb that relies on insect visitation for cross-pollination. Based on our field experiments, in which we excluded pollinators using mesh bags, the capacity for self-fertilization in S. rostratum is very low in North China. Our estimates of mating patterns in S. rostratum indicated a mixed mating system dominated by outcrossing, and the mating system of S. rostratum did not experience an evolutionary transition during its invasion of China. Recent analyses of the accumulated data have highlighted the high frequency of species with mixed mating systems 17,31 . However, the evolution of mixed mating strategies of invasive species is complex. On one hand, selfing can produce seeds under conditions of low density or during frequent episodes of colonization with severe inbreeding depression. On the other hand, outcrossing can maintain high genetic diversity within populations. There may be a balance between selfing and outcrossing for S. rostratum under different conditions, and efforts to explore the reproductive success of different populations can help us better understand the maintenance of mixed mating systems of S. rostratum during invasion.
There are many ecological factors that can influence outcrossing rates. Variation in flower design and display has been shown to influence pollinator visitation rates 32,33 , and the rates of pollinator visitation can directly affect the amount of outcrossing rates in self-compatible plants 19,20 . The flowers of S. rostratum exhibit monomorphic enantiostyly and strong herkogamy. In addition, it has been hypothesized that the high outcrossing rate in this species is probably facilitated by its floral morphology 29,30 . Moreover, S. rostratum is a buzz-pollinated plant, and insect size is the main determinant of whether a visitor acts as a pollinator or a pollen thief 34 . During the visit of effective pollinators, heteranthery results in the deposition of pollen on one side of pollinator's body against the style. Pollen from one floral morph will be deposited on the stigma of the opposite floral morph, thus reducing pollinator-assisted selfing within flowers. However, geitonogamy is unavoidable in large individuals, in which left-and right-styled flowers occur at the same time.
The mating system of flowering plants is influenced not only by its flower design and display but also by population size, geographic location and habitat fragmentation 20,35 . In our preliminary study, in which we performed pollen addition experiments, we reported that pollen addition significantly increases fruit set and seed production and that smaller populations suffer from more severe pollen limitation 36 . The ecological context of pollen limitation can negatively impact plant-pollinator interactions and generate diverse selective forces on plant mating systems. In this study, we did not find a significant relationship between multilocus outcrossing rates and population size, altitude, latitude or longitude. However, the smallest population, YQ, had a high outcrossing rate (0.783 ± 0.07), whereas the largest population, CJ, had a relatively low outcrossing rate (0.543 ± 0.069). There is evidence that the outcrossing rate was associated with the floral display 17 . In our field investigations, plants in the YQ population were small; some individuals only opened one or two flowers per day, and the flowering season lasted for more than four months, which may reduce pollen transfer between the same plants. By contrast, plants in the CJ population were relatively large; and most individuals opened more than five flowers per day during the flowering season, which may promote geitonogamous selfing. Therefore, the floral display may have an impact on the outcrossing rate in this species; however, this phenomenon needs further study.
In this study, population CY had the highest outcrossing rate (0.834 ± 0.064) in the ten studied populations in China, and population CY also has the highest outcrossing rate in the world based on the present studies. There are reasons explaining why the highest outcrossing rate occurred in population CY. First, population CY has a relatively long invasion history, and it is the first introduced population reported in China 24 . This may have given the species enough time to adapt to the environment and co-evolve with local pollinators. Second, population   CY is located in riversides without habitat fragmentation, which provides relatively favorable conditions for S. rostratum. Third, the genetic diversity of population CY is very high (PIC = 0.310), which may guarantee the ability to maintain high outcrossing rates 1,4,13 . By contrast, population YG, which had the lowest outcrossing rate (0.492 ± 0.225), is the last population that we found in this study and is located on the side of a railway. The genetic diversity of population YG is also the lowest (PIC = 0.043). The difference in the two populations may indicate that the invasion history, living habitat and genetic background could influence the mating system of S. rostratum. And further studies should be conducted to explain the maintenance of mixed mating system in S. rostratum. Moreover, the offspring used for mating system analysis were seedlings, and this may overestimate the outcrossing rate of S. rostratum. The mean germination rate of the ten populations was high (0.94 ± 0.05), and the germination rate was not correlated with the outcrossing rate. However, in many outcrossing species, the germination rate of self-fertilized seeds is lower than that of outcrossed seeds because of inbreeding depression 37 . Therefore, inbreeding depression functioning in the offspring of S. rostratum may lead to a higher estimate of the outcrossing rate. Further studies to determine the extent of inbreeding depression in S. rostratum will provide additional reproductive information on this highly outcrossing species.
In conclusion, our results suggest that a mixed mating system dominated by outcrossing of S. rostratum occurs in its invasive Asian range. The average multilocus outcrossing rate was 0.69 ± 0.12 (ranging from 0.49 to 0.83), with an average effective number of 7.86 male parents in Chinese populations. Compared with the high level of outcrossing rates (0.70 ± 0.03) in native populations in Mexico 30 , these results indicate that there was no evolutionary transition from outcrossing to selfing during S. rostratum's invasion of China. Moreover, there were no significant differences between the outcrossing rates and population size, altitude, latitude or longitude. We speculate that high outcrossing rates in invasive populations of S. rostratum may be facilitated by precise pollination owing to herkogamy and monomorphic enantiostyly.

Methods
Sampling design. To estimate mating parameters in S. rostratum, ten Chinese populations were sampled in a variety of habitats. The ten populations cover a large area of the invasive distribution in China and exhibit a wide range of population size (Table 3; Fig. 3). We randomly collected fruits from 30 individuals per population in October and November 2010. We then extracted seeds from the fruits and placed them in airtight containers at 4 °C.
For each population, we detected the genotypes of all seedlings randomly sampled from 12 mother plants and 12 young seedlings per mother plant. We did not use seeds because they were rich in endosperm, and DNA extraction yields poor-quality material 30 . To maximize the germination rate, the seeds were pretreated with a 1000-ppm solution of gibberellic acid for 24 h. We sowed 50 seeds per family (5 seeds per fruit and 10 fruits per family) in plastic seedling-raising dishes in March 2011 and placed the dishes in a greenhouse at Beijing Normal University. Finally, we randomly sampled a total of 1435 seedlings belonging to 120 maternal families (Table 3).
DNA extraction and genotyping. We extracted genomic DNA from leaves dried in silica gel of the 1435 offspring using a plant genomic DNA kit (Tiangen, Beijing, China). We scored the genotypes of the DNA samples using 11 pairs of microsatellite primers, which were previously developed for S. rostratum 38 . The 11 loci were amplified using multiple polymerase chain reactions (PCR) with a Microsatellite PCR kit (Qiagen, Shanghai, China). The PCR amplifications were performed using a Bio-Rad thermocycler as follows: one cycle of 95 °C for 5 min; 30 cycles of 95 °C for 30 s, 58 °C for 180 s, and 72 °C for 30 s; and a final extension step at 60 °C for 30 min 38 . The PCR products were sequenced using an ABI 3730xl capillary sequencer with GeneScan 500 LIZ as the internal size standard. The results were analyzed with GeneMapper Software version 3.7 (Applied Biosystems, Foster City, CA, USA).

Data analysis.
We analyzed the number of alleles (N a ), expected heterozygosity (H e ), observed heterozygosity (H o ), polymorphism information content (PIC) and the combined probability of exclusion of a single parent (PE sp ) for each population using the computer programs GENALEX version 6.4 39   F is was tested using 11000 randomizations, and the threshold was adjusted for multiple comparisons using the Bonferroni correction. To test differences in genetic variation between the ten populations in China, we estimated the population differentiation (F st ) among populations using 1000 permutations in FSTAT version 2.9.3 42 . The presence of null alleles was calculated using Micro-Checker software 43 .
Mating system parameters for each population were estimated using the expectation-maximization method of the computer program MLTR version 3.2 44 . We estimated the multilocus outcrossing rate (t m ), single-locus outcrossing rate (t s ), and outcrossing rate between related individuals (t m −t s ). The outcrossing rate between related individuals (t m −t s ) can be used to estimate biparental inbreeding. We also calculated the correlation of outcrossing rate (r t ), the correlation of paternity within sibships for both multilocus cases and single cases (r pm and r ps , respectively), and the extent of outcrossed paternity by related male parents (r ps −r pm ). The correlations of paternity (r pm and r ps ) represent the fraction of siblings that share the same male parents 44 . We calculated the effective number of male parents per progeny arrays among the ten populations as N ep = 1/r pm 45 . Standard deviations were computed based on 1000 bootstrap values, with the resampling of entire maternal families. For tests of statistical significance, we examined the distribution of 1000 bootstrap values to determine whether outcrossing rates differed significantly from 1.0 46,47 . When values were significantly less than 1.0, the species exhibited a mixed mating system. Comparison of outcrossing rates between Chinese populations and Mexican populations was done with independent samples t-test 11 . The relationships between multilocus estimates of outcrossing rate and population size, altitude, latitude and longitude were analyzed by Spearman rank correlation tests. In addition, the relationship between biparental inbreeding and population size was also analyzed by Spearman rank correlation tests.  Table 1. This figure was produced using ArcGIS software version 10.1 (http://www.esrichina.com.cn/softwareproduct/ ArcGIS/).