Breeding system and pollination ecology of introduced plants compared to their native relatives†
The authors thank J. Chase, K. Olsen, members of the Knight laboratory at Washington University in St. Louis and the Kremen laboratory at the University of California, Berkeley, Ø. Totland, and one anonymous reviewer for comments that improved the manuscript. K. Moriuchi, S. Zang, and A. Chung helped with pollinator observations and experiments. They also thank T. Morhman for plant identification and location information and J. Chase for logistical support. Funding for this project was provided by Tyson Research Center, Washington University in St. Louis, Howard Hughes Medical Institute, American Association of University Women, Missouri Native Plant Society, and National Research Initiative of the USDA Cooperative State Research, Education and Extension Service, grant no. 05–2290.
Abstract
Identifying how plant–enemy interactions contribute to the success of introduced species has been a subject of much research, while the role of plant–pollinator interactions has received less attention. The ability to reproduce in new environments is essential for the successful establishment and spread of introduced species. Introduced plant species that are not capable of autonomous self-fertilization and are unable to attract resident pollinators may suffer from pollen limitation. Our study quantifies the degree of autogamy and pollination ecology of 10 closely related pairs of native and introduced plant species at a single site near St. Louis, Missouri, USA. Most of these species pairs had similar capacities for autogamy; however, of those that differed, the introduced species were more autogamous than their native congeners. Most introduced plants have pollinator visitation rates similar to those of their native congeners. Of the 20 species studied, only three had significant pollen limitation. We suggest that the success of most introduced plant species is because they are highly autogamous or because their pollinator visitation rates are similar to those of their native relatives. Understanding and identifying traits related to pollination success that are key in successful introductions may allow better understanding and prediction of biological invasions.
Understanding the mechanisms that allow introduced plant species to successfully establish and persist is a primary focus of research in invasion biology (5; 45; 29; 37; 50). One prominent hypothesis is the enemy release hypothesis (25), which posits that introduced species have increased fitness in their introduced range relative to their native range (or relative to native species in their introduced range) as a result of release from their natural enemies such as herbivores and pathogens. However, when introduced species leave their native habitats, they are not only escaping enemies, which inhibit their fitness, but they may also leave behind their natural mutualists (e.g., pollinators), which can be vital to their reproduction and establishment. 11 estimated that 90% of angiosperms are biotically pollinated. For introduced plants that require pollinators, their reproductive success depends on their ability to attract the services of resident pollinators that can provide accurate pollen transfer in their new range. Understanding how plants succeed despite being decoupled from their native pollinators is critical if we are to understand and predict biological invasions and will provide insight into the ecology and evolution of plant–pollinator interactions.
The reproductive success of introduced species will depend on the ability of the plant to (1) produce offspring in the absence of pollinators (autonomous self-fertilization), (2) attract resident pollinators, and (3) have sufficient pollen transferred by pollinators to maximize seed set and prevent pollen limitation. The establishment of introduced species is often initiated from a small founding population. When populations are small, autogamous or vegetative reproduction is expected to be favorable because a single individual is sufficient for colonization (e.g., 7). This theory leads to the hypothesis that introduced plant species are more likely to have autogamous breeding systems compared to native plant species. Consistent with this hypothesis, 15 found that plant families that only contain biotically pollinated species are less likely to contain species that invade natural areas than those that contain at least some abiotically pollinated species. Also, 42 studied the breeding systems of 17 introduced woody species in South Africa and found that all of them were capable of self-pollination and 72% were capable of autogamy (see also 53; 54).
Although capacity for autonomous self-fertilization (“autogamy” throughout) could be beneficial to introduced species colonizing new areas, autogamy is not necessary for successful introductions. Introduced plant species that are not autogamous need to be sufficiently pollinated by resident pollinator species in their introduced range to successfully establish or their specialist pollinator needs to have been introduced (43). Several studies have shown that introduced plant species that require pollinators can successfully attract resident generalist pollinators (43; 10; 31; 38; 19; 20; 35; 9). The ability to attract pollinators, however, does not guarantee sufficient quantity and quality of pollen transfer (18; 24; 44; 2). This information leads to the additional hypothesis that introduced plant species that are not autogamous must have pollinator visitation rates equal to or greater than their native relative or they will suffer from pollen limitation.
Understanding the degree of pollen limitation and the breeding systems of introduced plant species will help identify the role of pollination in invasions; however, few pollen supplementation experiments testing for pollen limitation have been conducted on introduced species. Of over 1000 pollen supplementation experiments synthesized in a recent review, only 10 were of introduced plant species (28), and thus, there was insufficient power to detect difference in the magnitude of pollen limitation between native and introduced plant species.
Relatively few studies of pollination ecology have compared introduced species to closely related native species (10); however, such contrasts might increase the power of comparative tests (30) if some of the variation in breeding system and pollination success is due to phylogeny (17). In some systems, there is a signal of phylogeny on breeding system traits (26; J. H. Burns, University of California–Davis; R. B. Faden, Smithsonian; and S. J. Steppan, Florida State University, unpublished data), and thus we might expect that it would be necessary to control for relatedness in breeding system comparisons. This study is the first experimental manipulation of the pollination ecology of multiple pairs of closely related introduced and native congeners (or confamilials if congeners were not available).
To assess the role of pollination ecology on the reproductive fitness of introduced plants, we evaluated three possible factors affecting reproductive success: autogamy, attraction to pollinators, and pollen limitation. We expected that (1) breeding system would be conserved within a family (null hypothesis), but, when it was not, the introduced plant would have a higher degree of autonomous self-fertilization, (2) introduced species that were not autogamous would have similar pollinator visitation rates as their native relatives (because of similarity in floral traits across plants within a family), and (3) plants that are not autogamous and attract fewer pollinators would show significant pollen limitation. We found support for autogamy, high pollinator visitation, and significant pollen limitation in some of our introduced plant species.
MATERIALS AND METHODS
Study site and species
These studies were conducted over four summers (2004, 2005, 2007, 2008) at Tyson Research Center, an 80-ha field station owned and managed by Washington University in St. Louis and located 40 km southwest of St. Louis, Missouri, USA. Most of the site (85%) consists of oak–hickory forest, and the remainder consists of old fields and open areas with plants.
We compared species pairs, all of which are biotically pollinated, from 10 plant families that occur at the field site (Fig. 1). Species were paired by family, and when possible by genus, to control for differences in traits across lineages. To limit noise caused by temporal differences, all data for both species in each pair were recorded within the same year. Species classifications as native or introduced were determined using USDA categorization (51). Introduced species were all introduced from Eurasia and were classified as either invasive or naturalized based on the USDA Plants Database (51) and Missouri Exotic Pest Plant (32) list (Fig. 1). For most of our native and introduced plant species, multiple populations were present within our study site. In these cases, we choose study populations for which the native and introduced species within a pair occurred at similar population sizes and floral densities.
All species examined in this study. Descriptions are based on observations recorded during the study. Note: All species occurred at the Tyson Research Center. All species are classified as forbs by the USDA PLANTS Database (51) except Lonicera japonica, L. flava, and Vicia villosa, which are vine, vine/shrub, and vine/forb, respectively. Introduced species were all introduced from Eurasia and classified as invasive or naturalized based on the USDA PLANTS Database and Misssouri Exotic Pest Plant List (32). Pollen limitation studies were not conducted on the Asteraceae or Lilliaceae pairs. A hyphen (-) indicates no data were recorded for this species.
Breeding system experiments
To determine whether introduced and native species differed in their ability to autonomously self-fertilize, we conducted pollinator exclusion experiments. We determined the degree to which each species could reproduce in the absence of visits from animal pollinators because autogamy might be relevant for colonization success (7, 8). Flowers (or inflorescences, racemes) of each species (see Fig. 1 for sample sizes) were individually bagged with thin, small-mesh (<1 mm) netting prior to blooming to prevent pollinator access but allow wind access. These flowers were paired with nearby flowers on different plants that were hand pollinated with outcross pollen (see pollen supplementation experiment methods below) and left unbagged. The size of the bags used and the number of flowers inside the bag differed among the species pairs in this study because flower size and the spatial clustering of flowers within a plant differed among these pairs. For most species, we bagged and hand pollinated one flower. However, for some pairs, such as the Brassicaceae, a raceme of flowers was bagged, but only one flower, which was marked on the pedicel, was included in the experiment. For the Asteraceae, the whole inflorescence was manipulated. Bags were large enough to give flowers space to open but not so large that they weighed down the plant. Further, care was taken to ensure that the bags were large enough that they did not rub against the stigma and anthers of the flower, inadvertently pollinating the flower(s) within the bag. Flowers were chosen for this experiment when they were in the bud phase, and plants in the hand-pollinated treatment were hand pollinated as they opened (see pollen supplementation experiment methods later). We documented whether the flower formed a fruit, and if a fruit formed, we counted the number of seeds per flower when possible. For species pairs for which we had data on number of seeds per flower, the degree of autogamy is the ratio of the bagged to hand-pollinated seed set per flower. For species pairs for which we did not count seeds per flower (i.e., because of the difficulty of accurately counting seeds for some species pairs), fruit set (number of fruits/number of flowers) was used as a measure of autogamy. We chose to measure the ratio of bagged to hand-pollinated rather than bagged to open-pollinated so that we would not overestimate the degree of autogamy for species with high pollen limitation. For each plant family, we used a one-way ANOVA to determine whether the native and introduced species differed in their degree of autogamy. We also calculated 95% confidence intervals around each estimate of autogamy to determine whether each species exhibited significant (different from zero) autogamy.
Pollinator observations
To determine whether introduced and native species differed in pollinator visitation rates, we conducted observations of pollinator visits to each species. Each species was observed at least five times for 20 min, and observations were spaced evenly throughout the day, excluding the hottest portions when pollinators are not active. A visit was counted if the visitor was observed on the flower's sexual organs. Observations were shifted to accommodate the blooming time of Potentilla recta, which only blooms in the afternoon, and Commelina and Convolvulus species, which only bloom in the morning. No pollinator observations were conducted at night, although there is some evidence that Silene stellata is partially nocturnally pollinated, and therefore pollinator visitation rates might be underestimated for this species.
Often several flowers on different plants were observed during a single observation. In these cases, we pooled the visitation rate across these flowers to get a single data point for visitation rate. We calculated visitation rate as the number of visits per flower per 20-min observation. Visitation rate was continuously distributed and was treated as a continuous variable. For each plant family, we used one-way ANOVA to determine whether native and introduced flowers differed in their visitation rate. The visitation rate was square-root transformed for species in the Caprifoliaceae, Commelinaceae, and Liliaceae families, and assumptions of ANOVA were checked and met for all tests presented.
Pollen supplementation experiments
We conducted pollen supplementation experiments to quantify pollen limitation for each species. Approximately 20 pairs of flowering individuals (40 individuals total) of each species were marked with neutrally colored (brown or black) yarn and randomly divided into two treatment groups: control or supplement. Pairs were similar in size and close to each other to control for microsite variation. One of the individuals from each pair was chosen as the control treatment, marked, and left exposed to pollinators without manipulation. The other individual was supplemented with pollen collected from a single donor at least 1 m outside the experimental patch and applied using a small paint brush or tweezers. Each flower chosen for the experiment was on a separate plant. The flower-level, as opposed to whole-plant-level, manipulation was chosen to maximize the amount of effort spent in quantifying pollen limitation for multiple pairs, maximizing the power to generalize. Flower-level manipulations may result in overestimation of pollen limitation if plants reallocate resources to flowers with supplemented pollen within a plant (27).
We documented whether the flower formed a fruit, and if a fruit formed, it was collected and seeds from each fruit were counted when possible. Flowers that were damaged were not included in the analyses. For each pair of flowers, we calculated the magnitude of pollen limitation as ln(number of seeds per flower in supplement treatment + 1) − ln(number of seeds per flower in control treatment +1) using standard methods for ecological meta-analyses (e.g., 22). If seed set was identical in the control and supplement treatment, then the magnitude of pollen limitation equals 0. When seed set in the supplement treatment exceeded seed set in the control treatment, pollen limitation was determined to be greater than zero which signifies pollen limitation. For each plant family, we used one-way ANOVA to determine if the native and introduced species differed in their magnitude of pollen limitation. We also calculated 95% confidence intervals around each estimate of pollen limitation to determine whether each species had significant (different from zero) pollen limitation.
RESULTS
Breeding system was not significantly different in six of 10 species pairs (Fig. 2A). Autogamy levels for five of 10 introduced species and four of the 10 native species differed significantly from zero (Appendix S1, see Supplemental Data with online version of this article). The Fabaceae, Scrophulariaceae, and Liliaceae pairs failed to set any seed in the bagged treatment (i.e., no autogamy; Appendix S1). The Asteraceae species were both fully autogamous. The introduced Commelinaceae, Caryophyllaceae, Convolvulaceae, and Rosaceae had significantly higher fruit set when bagged than did their native relative (F1,25 = 11.68, P < 0.001; F1,7 = 52.76, P < 0.001; F1,22 = 44.11, P < 0.001; F1,13 = 9.66, P < 0.01, respectively).
The degree of autogamy (ratio of reproductive success [seeds per flower or fruit set] in the bagged : pollen supplement treatment) (A), pollinator visitation rate (visits/flower/20 min) (B) and magnitude of pollen limitation (ln [supplement reproductive success] – ln [control reproductive success]) (C) for introduced and native species pairs. Introduced species and native species that differ for these response variables are indicated by *** P < 0.001, ** P < 0.01 and * P < 0.05.
Pollinator visitation rates of introduced species did not differ significantly from or were greater than their native species pair in eight of nine cases (Fig. 2B). In the Rosaceae pair, the introduced species had a significantly higher visitation rate than its native counterpart (F1,8 = 20.9, P < 0.002). However, flowers of the introduced Caprifoliaceae were visited less frequently by pollinators than flowers of the native species (F1,8 = 8.46, P < 0.02).
Three species (the introduced Fabaceae, the introduced Caprifoliaceae, and the native Convolvulaceae) had significant pollen limitation (i.e., different from zero) (Appendix S2, see Supplemental Data with online article). Despite having the same pollinator visitation as its native species pair, the introduced Fabaceae species had a pollen limitation magnitude of 0.797 ± 0.206, while the native had a pollen limitation magnitude of −0.128 ± 0.163. The introduced Caprifoliaceae received significantly fewer visits than its native congener and had significantly higher pollen limitation (Fig. 2B and 2C). The native Convolvulus sepium was significantly pollen limited, in spite of relatively high visitation rates (Fig. 2B). Pollen limitation differed between the native and introduced species in the Brassicaceae, Fabaceae, Convolvulaceae, and Caprifoliaceae families (F1,35 = 5.608, P = 0.02, F1,26 = 11.96, P = 0.001; F1,18 = 5.652, P = 0.03; and F1,62 = 7.93, P = 0.01, respectively), where the introduced species is more pollen limited for three of these cases, and only the native Convolvulus is more pollen limited than the introduced Convolvulus. Note that in the case of Convolvulus, the native was less autogamous than the introduced species. Pollen limitation for the Liliaceae pair was not calculable due to low population numbers and primarily vegetative reproduction.
DISCUSSION
The reproductive success of a plant is a complex function of breeding system and pollination ecology (21; 56). Understanding reproductive success has profound implications for species invasions and fitness. We present the first study to contrast autogamy, pollinator visitation, and pollen limitation of multiple pairs of related introduced and native species, to examine the role of these in the reproductive success of introduced species. We predicted that established introduced plants will either (1) have the capacity for autonomous self-fertilization and thus not require pollinators, (2) be able to attract resident pollinators, or (3) be significantly pollen limited. Our results show support for each of these possible outcomes. Thus, introduced plants are diverse in their pollination ecology: some are successful because of their autogamous breeding system, some are able to successfully attract resident pollinations, and others manage to establish and spread despite significant pollen limitation.
We expected that the degree of autonomous self-fertilization would be conserved for most species pairs, and when it differed, the introduced species should have greater autogamy. The similar breeding system of most of our species pairs suggests that there is a strong signal of phylogeny on breeding system traits, which has been shown in other studies (26; J. H. Burns, R. B. Faden, and S. J. Steppan, unpublished data). However, four of our species pairs differed significantly in their degree of autogamy, and in three of these cases, the introduced species was more autogamous than its native relative. Our results may indicate that autogamy is an important factor in the establishment of introduced plant species, as suggested by 8, and are consistent with other synthetic studies that show that autogamous plant lineages have high representation in introduced flora (e.g., 42; 53; 54). However, our study and these reviews can only show a pattern of higher autogamy in introduced plants. A true test of Baker's law would consider the global invasiveness of both native and introduced species in each pair, for which we do not account. Further studies of plants during their establishment phase are necessary to conclude if Baker's (1974) mechanism is correct.
Our second hypothesis was that introduced species would have greater pollen limitation than native relatives, if they had lower pollinator visitation. Most pairs did not differ significantly in their pollinator visitation rate. There were significant differences in visitation rate for the Caprifoliaceae and Rosaceae pairs: the Caprifoliaceae had greater visitation for the native species and the Rosaceae had lower pollinator visitation for the native species. It is unclear why the introduced Caprifoliaceae, Lonicera japonica, which has a white flower that is similar in size and morphology to its native congener, L. flava (a yellow-flowered species), is less attractive to pollinators. We note that the native L. flava was primarily pollinated by carpenter bees, which appeared to have a high rate of nectar robbing (fruit set was only 22% for this species). The higher visitation rate of the introduced Rosaceae might be due to its larger petal size and taller, more erect stems. While we were careful to choose locations in which the native and introduced pair occurred at similar densities and population sizes, it is possible that small differences in population size and density caused the differences in visitation rate observed for the Caprifoliaceae and Rosaceae species pairs. In addition, we did not measure other floral traits that pollinators are known to respond to, such as floral scent and nectar quantity and quality. In general, our results are consistent with observations in other studies that introduced plants can successfully attract resident generalist pollinators (43; 10; 31; 38; 19; 20; 35).
We predicted that those introduced species without significant autogamy or high visitation rates should be more pollen limited than their native relatives. We found limited support for this prediction (in the Caprifoliaceae pair): the only introduced species that was not autogamous and had significantly lower visitation rates than its native pair also had a higher magnitude of pollen limitation relative to its native pair. However, while the breeding system and pollinator visitation rates of the introduced Fabaceae were similar to those of its native relative, it was still more pollen limited. Our native Fabaceae species is known to be a host plant for the northern cloudywing (Thorybes pylades), which was also the most frequent visitor, while the introduced species was visited primarily by generalist bees. Thus, while visitation rates are similar between these species, the quality of each visit might be much higher for the native species. The native Convolvulaceae was more pollen limited than its introduced pair, which is not surprising because it was also less autogamous. In most cases, large effect sizes corresponded well with significance, but in a few cases we did not detect significant autogamy or pollen limitation in some species where the magnitudes of the effects were fairly high (e.g., Verbascum blattaria has pollen limitation of 1.23 but does not have statistically significant pollen limitation) but power was low (online Appendix S2). Therefore, the lack of significant autogamy or pollen limitation in cases with small sample sizes should be interpreted cautiously.
Pollen limitation does not necessarily limit a plant's ability to establish and spread because both Vicia villosa (introduced Fabaceae) and Lonicera japonica (introduced Caprifoliaceae) are pollen limited and have also been highly successful in their introduced range. Both species are widespread in the USA and are considered invasive by the U. S. Department of Agriculture (51). One possible reason for their success in spite of significant pollen limitation is that both species are capable of reproducing asexually, and their local population growth may not be seed dependent. Each of these species also produces many flowers, so, while individual flowers may be pollen limited, the plant will still produce many seeds and thus may not suffer demographic consequences. Further, both of these species are successful in disturbed habitats. Vicia villosa is able to fix nitrogen and may be able to readily establish on disturbed, nitrogen-poor soils. Significant pollen limitation has been shown in other successful invasive plants. For example, 40, 41) demonstrated that Cytisus scoparius was highly pollen limited, and yet, its population growth rate was still high.
In our study, we only found significant pollen limitation in one of the native species. This result is in contrast with recent reviews that show that >60% of plant populations are pollen limited (12; 28). The lack of pollen limitation in most of our native plant species might be due to our choice of study species. We chose native plant species that were closely related to introduced plant species, and because breeding systems and floral traits are often conserved in lineages, we may have preferentially chosen native species that are highly generalized in their pollination or autogamous and thus less likely to suffer from pollen limitation. Indeed, ecologists choose to study native plants that have interesting pollination ecology (e.g., orchid and lily species are overrepresented in the pollination literature), and therefore the studies available for review are a biased sample of angiosperms (28, T. M. Knight, unpublished results). However, it could also be argued that most plants are generalized in their pollination (55) or to some extent autogamous, and we cannot directly compare our plant species to those reviewed in the pollen limitation meta-analysis by 28 because most of the studies in the meta-analysis do not evaluate degree of autogamy.
This study focuses on already established introduced plants; a broader understanding of the role of pollination ecology in biological invasions will require considering other phases in the invasion process for several reasons. First, the fitness consequences of different breeding systems may shift at different stages of invasion (e.g., autogamy may be favored at early stages of invasion, but breeding systems that promote outcrossing might be favored after establishment). Second, Allee effects, a positive relationship between fitness and either numbers or density of conspecifics (3, 4; 49), may decrease the probability that an introduced species will establish and/or slow its spread (e.g., 1, but see 16; 13; 52, 53). Third, self-compatibility could also play a significant role in the establishment of introduced species. Self-compatible species might be less restricted by propagule supply in the early stages of an introduction if geitonogamy provides reproductive assurance (47; 46; 39; 36), which could enhance establishment success even in small founding populations (e.g., 7; 42).
The successful establishment and spread of introduced species is often discussed in the context of the enemy release hypothesis (reviewed by 34; 14; 23; 50). We suggest that a broader view of plant–animal interactions to include mutualists might provide a greater understanding for why some plants establish and become invasive while others fail (see also 43; 33). We provide one of the first empirical examinations of the role of pollination ecology for multiple pairs of related introduced and native species.
We suggest that future work examining the pollination ecology of native and introduced species should consider: (1) the interactive effects of enemies (herbivores, florivores) and pollinators (e.g., 48). If introduced plants receive less damage because of a release from enemies, then they may have more resources to allocate to floral attraction, allowing them to better attract resident pollinators. (2) Future work should consider the pollination success of species that are distantly related to the native community (phylogenetically novel introduced species). Novelty may provide a pollination advantage if the introduced plant possesses novel traits that pollinators prefer, such as greater nectar production or larger floral displays. Alternatively, novel species may have novel traits that the resident pollinators are unfamiliar with, resulting in fewer pollinator visits or less accurate pollen transfer (see also 31). (3) Future work should also consider pollen quality and pollinator effectiveness. Because introduced plants are not coevolved with the resident pollinators, the quality of pollen arriving and the effectiveness by which it is deposited may be lower per pollinator visit than that of native plant species (e.g., 6). Future studies such as these would help provide a more comprehensive understanding of the role of pollination ecology and breeding system in introductions, which may be key in understanding and predicting current and future invasions.
