https://orcid.org/0000-0001-7393-3889
Figures
Abstract
Invasive plants have the potential to interfere with native species’ reproductive success through a number of mechanisms, including heterospecific pollination and hybridization. This study investigated reproductive interactions between a native North American woody vine (American bittersweet, Celastrus scandens) and an introduced congener (oriental bittersweet, C. orbiculatus). The decline of C. scandens in the eastern portion of its range is coincident with the introduction and spread of C. orbiculatus, and the two species are known to hybridize. The relationship between proximity and floral production of conspecific and heterospecific males on fertilization and hybridization rates was measured at a field site in northwestern Indiana, USA where both species occur and reproduce. We found that the invasive vine had an extreme advantage in both male and female floral production, producing nearly 200 times more flowers per staminate plant and 65 times more flowers per pistillate plant than the native. Using nuclear microsatellite DNA markers we found that hybridization rates were asymmetric; 39% of the C. scandens seeds tested were hybrids, compared to only 1.6% of C. orbiculatus seeds. The asymmetric hybridization rates were likely not solely due to greater abundance of C. orbiculatus pollen because experimental hand crosses revealed that C. scandens had a higher rate (41%) of heterospecific fertilization than C. orbiculatus (2.4%). We previously reported that few hybrids were observed in the wild, and hybrids had greatly reduced fecundity. Thus, in our system, the threat posed by heterospecific pollen is not replacement by hybrids or introgression, but rather asymmetric reproductive interference. Reproductive interference extended to distances as great as 100 meters, thus, efforts to conserve the native species must reduce its exposure to C. orbiculatus over a relatively large spatial scale.
Citation: Zaya DN, Leicht-Young SA, Pavlovic NB, Ashley MV (2021) Heterospecific pollination by an invasive congener threatens the native American bittersweet, Celastrus scandens. PLoS ONE 16(3): e0248635. https://doi.org/10.1371/journal.pone.0248635
Editor: Juliana Hipólito, Instituto Nacional de Pesquisas da Amazonia, BRAZIL
Received: June 7, 2020; Accepted: March 2, 2021; Published: March 23, 2021
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: All data files are available from the Illinois Data Bank. DOI: 10.13012/B2IDB-6925722_V1 URL: https://doi.org/10.13012/B2IDB-6925722_V1.
Funding: DNZ was awarded funding from the University of Illinois at Chicago. NBP and SALY were awarded funding from the United State Geological Survey. DNZ and MVA were awarded funding from Chicago Wilderness. DNZ was awarded funding from the Illinois State Academy of Science. DNZ was awarded funding from the National Park Service, through the Great Lakes Research and Education Center. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The commercial entity RK&K provided support in the form of salaries for author SLY, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of this author is articulated in the ‘author contributions’ section.
Competing interests: One author (SLY) is currently affiliated with a commercial entity, though she was not at the time that the study was conceived and executed. This affiliation does not alter our adherence to PLOS ONE policies on sharing data and materials. The authors have declared that no competing interests exist.
Introduction
Threats posed by invasive plant species to native ecosystems are a growing concern, with mounting evidence of increasing rates of invasion [1–3], disruptions of ecosystem function [1, 4–6], and reductions in native community diversity [5, 7–9]. Recent work has highlighted the potential of invasive plants to disrupt pollination systems and negatively impact reproductive success of native plants [10–12]. While much attention has focused on pollinator networks and pre-pollination dynamics in invaded systems [10, 11, 13], invasive species can also alter post-pollination dynamics. Threats posed by increased heterospecific pollination (also referred to as ‘interspecific’ pollen by some authors) and reproductive interference during invasions has been documented in several systems [14–17]. Two recent meta-analyses demonstrate that the threat posed by invasive plants through reproductive interference is potentially widespread. Arceo-Gómez and Ashman [12] found that invasive plants are more detrimental as heterospecific pollen donors than are native plants, and that the negative effects are greater when the invasive pollen donor and recipient are closely related. Burns et al. [18] found that invasives produce more than 250 times more flowers per plant compared to their native relatives. While Burns et al. [18] emphasize that this allows invasives to compensate for or avoid pollen limitation, an additional consequence could be increased heterospecific pollination of co-occurring natives that share their pollinators.
Reproductive interference by invasives can occur even without disruption of pollinator networks and visitation rates through heterospecific pollination. Pollen from an invasive species may reduce successful fertilization of natives by conspecific pollen [16]. This could occur through chemical or physical interference in the stigma by invasive pollen [19] or because the invasive pollen is simply much more abundant than native pollen [20]. Rejection of invasive pollen is especially difficult when the native and invasive species are closely related, such as congeners [12]. If heterospecific fertilization occurs, female reproductive effort may be wasted on fertilized ovules that rarely develop into viable seeds [21]. If viable hybrids survive, this may lead to decline of the native due to seed discounting and wasted reproductive effort [22]. Additionally, hybrids may be especially numerous, fit, or fecund, leading to “hybrid swarms” that exclude the native species [6, 23, 24]. Finally, continued asymmetric introgression may lead to genetic dilution of the native species’ gene pool [20, 25, 26].
In this study we investigated reproductive interactions between an invasive vine, Celastrus orbiculatus Thunb. and a native congener, C. scandens L. Celastrus orbiculatus was introduced to North America from Asia [27] and has subsequently spread across much of the native range of C. scandens. The two species are known to hybridize in controlled settings [28] and in the wild [29]. Hybrid individuals are uncommon in the wild, but a previous study found that all genetically identified hybrids had a C. scandens maternal lineage [29]. It has not been demonstrated that this asymmetric hybridization is linked to the decline of C. scandens, but declines are especially severe in parts of the eastern United States where the invasion is the oldest [30, 31]. While large-scale plant surveys can provide some insight into the dynamics of invasion and hybridization, fine-scale studies of invasive heterospecific pollen transfer are needed to directly assess impacts on native plants [12, 19].
We hypothesized that reproductive interference of C. scandens occurs in the presence of C. orbiculatus. We collected data at the Indiana Dunes National Park (Indiana, USA) at a site where C. scandens and C. orbiculatus occur in similar abundances, which provided an opportunity to test this hypothesis (Fig 1). Because both species are dioecious, we could specifically assess the relationship between male floral production and reproductive interference. We used experimental hand-crosses to compare the responses of C. scandens and C. orbiculatus to heterospecific pollination. For open-pollinated plants in the field, we evaluated the relationship between the availability of conspecific or heterospecific pollen and rates of fertilization and hybridization. Hybrids were identified using previously published genetic markers. Our study was designed to test our hypothesis of reproductive interference between C. orbiculatus and C. scandens, understand the causes of asymmetrical pollen flow, evaluate the threat that heterospecific pollination poses to C. scandens, and advise conservation strategies for an increasingly rare native plant.
Aerial imagery was accessed through the U.S. Geological Survey’s National Map Viewer, and collected by the National Agriculture Imagery Program (U.S. Department of Agriculture) in the year of the study, 2008.
Materials and methods
Study species
Celastrus scandens L. (Celastraceae), commonly known as American bittersweet or American staff vine, is the only member of the genus native to North America [32]. It is a liana (woody vine) usually found in open habitats. Its range extends from southern Quebec to South Dakota, south to western Texas through Georgia. The native range of C. orbiculatus Thunb., commonly known as oriental or Asiatic bittersweet, is in Korea, Japan, and China [32] where it is found in thickets and lowland slopes. In North America, it can also thrive in shaded habitats that would likely exclude C. scandens [33]. The North American range of C. orbiculatus overlaps with much of the range of C. scandens east of the Mississippi River, with fewer reports west of the Mississippi River. Both species are usually dioecious, although rare individuals and populations displaying monoecy or bisexual flowers are known. In both species, staminate flowers have five stamens. Pistillate flowers have a single pistil, a superior ovary, and three carpels with two ovules each. A single fruit can have up to six seeds. Insects, particularly native bees, pollinate both species [34], with some wind-pollination observed for C. orbiculatus [35]. Birds are largely credited as the dispersers of the seeds that sit inside the three bright red fleshy arils surrounded by an orange (C. scandens) or yellow (C. orbiculatus) capsule that breaks open with first frost [34].
Celastrus orbiculatus was introduced as an ornamental vine to the eastern USA in the mid- to late-nineteenth century. The first reports of it escaping into the wild came in the early twentieth century; by the middle of the twentieth century it was recognized as a pest species rapidly spreading in the eastern USA [27, 36]. The species continues to spread, having recently (c. 2010) been observed for the first time in Minnesota. Land managers and foresters consider C. orbiculatus a troublesome species because it is a strong competitor that crowds out native vegetation [37], negatively affects forestry operation [38], and can alter natural successional trajectories [39]. Notably, C. scandens has declined in the eastern United States where invasion by C. orbiculatus is oldest and most extreme [30, 31, 40], although a causal link between the two patterns has not been established. Celastrus scandens has been listed as a species of special concern in several states; it may be extirpated in Delaware, while it is listed as a vulnerable species in New York, a species of special concern in Connecticut, a threatened species in Massachusetts, and an endangered species in North Carolina and Rhode Island.
Study site, sampling, species identification
Experimental crosses and field observations took place at the Portage Lakefront within the Indiana Dunes National Park (near 41.63°N, 87.18°W). The study site is composed of southern Lake Michigan sand dunes and wooded edges, bounded by the town of Ogden Dunes, Indiana, USA to the west and the Port of Indiana to the east (Fig 1). The site was chosen because reproductive individuals of both species are abundant, which is uncommon. We attempted to include all flowering individuals in the observational studies, but some flowering individuals that may have been in nearby residential areas were not included in the study. In total, the study included 39 C. orbiculatus individuals (22 staminate, 14 pistillate, and 3 putatively monoecious) and 40 C. scandens individuals (17 staminate, 21 pistillate, and 2 putatively monoecious). This work was permitted by the National Park Service, permit number NH08.5216.
Field identification using reproductive structures [41] was verified by genetic analysis of leaf tissue. Genomic DNA was extracted from 20–25 mg of ground leaf material using the DNeasy Plant Mini Kit (Qiagen) and each individual was scored at five microsatellite loci that are able to distinguish the two species and their hybrids [29]. Microsatellite loci, PCR conditions, and genotyping methods are described in Zaya et al. [29]. Genetic identities were tested using two Bayesian clustering approaches, STRUCTURE version 2.3 and NewHybrids version 1.1 beta [42, 43] using options previously used to identify C. scandens X C. orbiculatus hybrids [29]. We used the posterior probability of an individual genotype belonging to a single genetic cluster, q, to categorize hybrid seedlings. Seedlings with q > 0.85 were categorized as the result of conspecific pollination, while those with q < 0.65 were categorized as hybrids. No seedlings had q values between these cutoffs. Spatial coordinates were measured with GPS for each individual included in the study. The coordinates were used to calculate the distance to each staminate plant for every pistillate individual.
Hand pollinations
Experimental hand cross-pollinations were carried out in 2006 to test differences between the species in accepting heterospecific pollen, a possible mechanism behind asymmetric hybridization. Inflorescences on each of 10 pistillate plants, five pistillate plants of each species, were bagged with bridal mesh while still in bud, to exclude pollinators. For each pistillate plant, 25 flowers were hand-pollinated with freshly dehisced anthers from five staminate plants of the other species (five flowers were devoted to each staminate plant). In total, 250 flowers were hand-pollinated. The flowers were observed approximately 10 days later for signs of fertilization, in the form of an ovary that had swelled and not senesced. We compared the rate of successful heterospecific fertilization in each species.
Floral production, phenology and open pollination patterns
We recorded the total number of open staminate or pistillate flowers for each individual on 15 dates, between 20 May and 12 June 2008. All open flowers on an individual were counted and recorded when possible. Some individuals were too large or climbed too high to count all flowers, in which case a proportional sampling technique was used. For plants covering a large area, we estimated the number of flowers from approximately three square meters, and extrapolated to the ground area covered by the plants as estimated from on-the-ground measurements.
We collected data on fertilization, fruit set, and hybridization rates from individually marked flowers. On each date that plants were observed, we marked up to 10 newly opened flowers on each pistillate plant with colored string tied to the pedicel. Flowers were chosen haphazardly, as attempts to randomize flower selection were prohibitively slow. We did make an effort to select flowers from different places on the pistillate plant, covering its full width and height. We checked each flower for evidence of fertilization 9 to 17 days after it was marked and recorded fertilization success (persistent and swollen ovary) and followed the development of the same flowers through fruit maturation. The flowers on the cyme of C. orbiculatus are all of equal rank, while the inflorescence for C. scandens is a panicle with a hierarchical structure. The position on the inflorescence was recorded for each marked C. scandens flower in order to investigate potential differences in maternal investment and fertilization rate [see 44]. On 5 October and 11 October 2008, we recorded whether the marked flowers had successfully developed into mature fruits and collected those fruits. Resulting seeds (three seeds per ovary) were prepared for germination according to the method described in Young and Young [45] and sowed in the greenhouse at the University of Illinois at Chicago after 90 days of cold stratification. Greenhouse lighting relied on natural sunlight with an approximately 12-hour photoperiod. Greenhouse temperature ranged from approximately 15°C to 25°C. In total, 104 seedlings resulted from the original 716 marked flowers (43 seedlings from 317 C. scandens flowers, and 61 seedlings from 399 C. orbiculatus flowers). We genetically tested the seedlings to determine the species identity of the pollen donor using the nuclear microsatellite DNA markers and statistical methods described briefly above and detailed in Zaya et al. [29].
For both fertilization rate and fruit set, we investigated the relationship with several predictors. Analyses were conducted separately for each species. We were most interested in how fertilization and fruit set were related to the availability of conspecific and heterospecific pollen, but we also tested for a relationship with total floral production of the maternal plant and, in C. scandens, inflorescence position. The sample size of seedlings was too small to test for correlations between hybridization rate and multiple predictors, but the difference in hybridization rate between species was statistically compared. Details of the data analyses are described below.
Fruits collected during the fertilization study were supplemented by additional fruit collections from C. scandens, with the goal of increasing sample size and statistical power. Emphasis was placed on fruits from the native species because a) no wild hybrids found had a C. orbiculatus maternal lineage [29], and b) there is more urgency in understanding hybridization in C. scandens, as it is species of concern in much of its native range. Approximately 250 fruits were collected from 18 C. scandens pistillate plants. We extracted seeds from these fruits, applied cold stratification, and conducted genetic analyses using leaf tissue from the resulting seedlings [29]. We tested the relationship between hybridization rate and predictors relating to the availability of heterospecific and conspecific pollen (detailed below).
Data analysis
To test the hypotheses that fertilization rate, fruit set, and hybridization rate differed for open-pollinated C. scandens and C. orbiculatus, we used generalized linear models (GLM) with binomial responses. For each test, the explanatory variable was species identity of maternal plant, and the subjects (i.e., units of replication) were the individual maternal plants (fertilization rate and fruit set, n = 40 [23 C. scandens and 17 C. orbiculatus]; hybridization rate, n = 19 [nine C. scandens and 10 C. orbiculatus]). Monoecious plants were included in these analyses. For analyses of hybridization rates only plants with flowers resulting in at least one seedling were included.
We were interested in the influence of conspecific and heterospecific pollen on the fate of individual flowers (at fertilization and fruit set), and how it differed between species. To explore these relationships, we used generalized linear mixed-effects models (GLMM), with maternal identity as a random factor. The two species were analyzed separately. The subjects of the analysis were individual flowers (for C. scandens, n = 317 in 23 maternal groups; for C. orbiculatus, n = 399 in 17 maternal groups). The response variable tested was the binary fate of a flower at the fertilization stage (fertilized or unfertilized) or fruit stage (developed or not developed). Analyses of fruit development only included fertilized flowers. Four explanatory variables were tested as fixed effects: total floral production during the reproductive period, weighted conspecific mating potential, weighted heterospecific mating potential, and inflorescence position (in C. scandens, only). We constructed candidate models with a maximum of one fixed effect because of significant collinearity among potential explanatory variables. The “weighted mating potential” is a measure that considers both the distance to and floral production of available mates. Greater values indicate higher potential for pollen transfer. It is constructed to be inversely related to the distance to a potential mate, and positively associated with floral production of a potential mate. The weighted mating potential was calculated after Wagenius et al. [46], with some modification:
(1)
We removed an incompatibility term, cij, and instead excluded other pistillate plants in the calculation (those which would have cij = 0). We weighted the mating potential by log10 male floral production, Fj, for each j-th father on the particular day that a flower was marked. Distance from the i-th mother (which carried the flower) to each j-th father is dij. We used the same value for γ, the inverse of the mean pollination distance (13.3 m-1), as Wagenius et al. [46], who studied an insect-pollinated self-incompatible plant (Celastrus spp. are dioecious).
To test the relationship between hybridization rate in C. scandens and the availability of conspecific and heterospecific pollen, we used a GLM with binomial response. The subjects of the analysis were individual maternal plants from which seeds were collected and germinated (n = 18, with 259 total seedlings). The response variable was the proportion of successfully germinated seedlings that were hybrids. Four fixed effects were tested: distance to the nearest conspecific male (staminate C. scandens), distance to the nearest heterospecific male (staminate C. orbiculatus), log10 floral production of the nearest co-flowering conspecific male, and log10 floral production on the nearest co-flowering heterospecific male. All fixed effects were included in separate models. Additionally, we tested three two-way interactions (conspecific distance X heterospecific distance, conspecific distance X conspecific floral production, heterospecific distance X heterospecific floral production). A staminate plant was considered as “co-flowering” if it had open staminate flowers on the mean flowering date of the pistillate subject.
To test for species differences in the experimental hand crosses, we used a GLM to compare the rate of successful heterospecific pollination. The GLM was constructed so that fertilization success in heterospecific crosses was the binomial response and species identity of maternal plant was the explanatory variable (n = 10 pistillate plants, five of each species).
All statistical tests were completed using R version 3.6.2 [47]. The GLMM tests were implemented using the lme4 package [48], and in all cases random intercept models were constructed with maternal identity as the random effect. For each test, we used the Akaike Information Criterion corrected for small sample size (AICc) to select the best model from a group of candidate models. Akaike weights were calculated for each model, which can be interpreted as the probability that a given model is the one that best approximates the data assuming one exists in the candidate set. In all cases a null model, with no fixed effects, was included as a candidate model. The coefficient of determination (R2) was estimated for GLM using the rsq package [49, 50] and for GLMM using the pseudo-R2 (R2GLMM) [51, 52] from the MuMIn package [53].
Results
Hand pollinations
In the experimental hand pollinations, 51 of the 125 (41% [s.e. = 4.4%]) C. scandens pistillate flowers receiving heterospecific pollen were fertilized compared to only three (2.4% [s.e. = 1.4%]) of the C. orbiculatus flowers. Statistical comparison confirmed the difference between species. The model that included pistillate plant species identity greatly outperformed the null model (ΔAICc = 60.4, Akaike weight > 0.9999).
Floral production, phenology, and open pollination patterns
Observations of floral production show that the invasive C. orbiculatus has a significant advantage in male and female fecundity compared to the native C. scandens (Table 1). Pistillate flowering in C. orbiculatus began on 21 May and ended on 6 June, while in C. scandens it lasted from 28 May to 12 June. In pistillate flowers, there was a 65-fold difference in output between C. orbiculatus (mean of 13,100 flowers per plant, s.e. = 4,700) and C. scandens (202 flowers per plant, s.e. = 50). Staminate flowering in C. orbiculatus lasted 21 days, from 20 May to 9 June, while in C. scandens staminate flowering lasted 16 days, from 28 May to 12 June (Fig 2). The peak flowering dates were offset by 5 days—31 May for C. orbiculatus, 5 June for C. scandens. Across 15 dates of measurement, we observed 723 flowers per staminate individual in C. scandens (s.e. = 282) and 142,500 per individual in C. orbiculatus (s.e. = 43,800), a 197-fold difference. When comparing the respective peak flowering dates, C. orbiculatus had 277 flowers per staminate individual and C. orbiculatus had 28,000, more than a 100-fold difference (Fig 2). Even on the peak flowering date for C. scandens (5 June) there were 77 times more C. orbiculatus flowers per staminate individual (277 in C. scandens, 21,200 in C. orbiculatus). Not surprisingly, these differences in floral production resulted in large differences in conspecific and heterospecific mating potential for each species. For example, conspecific mating potential never exceeded 5.9 for C. scandens pistillate plants but ranged up to 29.0 for C. orbiculatus (Fig 2).
Observations made across 19 staminate individuals in C. scandens (red squares) and 25 staminate individuals in C. orbiculatus (black triangles). Vertical dotted lines represent the dates of peak flowering in C. orbiculatus (31 May, black line) and C. scandens (5 June, red line).
Three C. orbiculatus and two C. scandens individuals were monoecious, with both pistillate and staminate flowers. One type of flower dominated in each individual. In two cases (one of each species) only two pistillate flowers were observed compared to dozens of staminate flowers. In two plants (one of each species) staminate flowers outnumbered pistillate flowers by approximately 3-to-1. In one large C. orbiculatus plant, there was a 3.5-to-1 ratio in favor of pistillate flowers. Monoecious individuals were included in calculations of explanatory variables (e.g., conspecific mating potential) for other pistillate individuals, but the staminate flowers from the same individual were not considered in such calculations.
Observations on the fate of open-pollinated pistillate flowers are shown in Table 2. Fertilization rate was 27.4% for C. scandens and 44.1% for C. orbiculatus. We found statistical support for a difference between species with respect to fertilization rate (GLM that included species as a fixed effect outperformed the null model, ΔAICc = 19.2, Akaike weight > 0.9999, adjusted R2 = 0.130). Conspecific mating potential was associated with increased fertilization in both species (Fig 3) and was the best-supported explanatory variable in both species (Tables 3 and 4). However, for heterospecific mating potential, the patterns differed between the two species. In C. scandens, heterospecific mating potential was positively correlated with fertilization, a relationship that was strongly supported (ΔAICc = 10.1 when compared to null model, Table 3). In C. orbiculatus, the relationship was negative and weakly supported (ΔAICc = 0.2 when compared to null model, Table 4).
Results from 317 individually marked C. scandens (A, B) and 399 individually marked C. orbiculatus (C, D) flowers are shown. Numbers above the symbols represent the number of flowers included in each binned category. In C. scandens, fertilization rate was statistically associated with conspecific mating potential (A; Table 3) and heterospecific mating potential (B; Table 3). In C. orbiculatus, fertilization rate was only associated with conspecific mating potential (C: Table 4). See Eq 1 for the calculation of mating potentials. Error bars represent the standard error.
When following the fate of marked pistillate flowers, we found that 8.4% of flowers in C. scandens developed into mature fruits (Table 2). In C. orbiculatus, 10.8% of flowers developed into mature fruits (Table 2). When considering only fertilized flowers, the fruit set rates were 32.1% in C. scandens and 23.9% in C. orbiculatus. We did not find statistical support for a difference between species when considering all flowers (GLM with no fixed effect was preferred, ΔAICc = 1.0, Akaike weight = 0.627) or only fertilized flowers (ΔAICc = 0.41, Akaike weight = 0.551).
In C. scandens, the proportion of fertilized flowers that developed to fruit was not associated with any of the explanatory variables we tested, as the null model outperformed all candidate models with fixed effects (Table 5). In C. orbiculatus, fruit development was negatively associated with total floral production; fewer fertilized flowers successfully developed on pistillate plants with more flowers (Table 6). The model with heterospecific mating potential had statistical support similar to the null, and the model with conspecific mating potential had substantially less statistical support (Table 6).
Genetic analyses of germinated seedlings that arose from marked flowers showed a large difference in the number of hybrid seedlings between the two species (Table 2). In C. scandens, 22 of 43 (51.2%) seedlings that germinated were hybrids, while only one of the 61 (1.6%) seedlings from C. orbiculatus were hybrids. We found strong statistical support for a difference between species with regard to hybridization rate. The GLM that included species as a predictor of hybridization rate performed much better than the null model (ΔAICc = 37.6, Akaike weight > 0.9999, adjusted R2 = 0.419).
When an additional 216 C. scandens seedlings were included, from 18 pistillate plants, supplementing the data from the individually marked flowers, we found an overall hybridization rate of 0.39 (Table 2). For this expanded data set, the model for best predicting hybridization rate included the distance to the nearest staminate C. orbiculatus, log floral production of the nearest co-flowering staminate C. orbiculatus, and the interaction of these two predictors (Akaike weight > 0.9999; Table 7). This model vastly outperformed the next best model, which included only the log floral production of the nearest heterospecific male (ΔAICc = 34.4, Akaike weight < 0.0001). Hybridization rate increased as the distance to the nearest C. orbiculatus staminate plant decreased (Fig 4). This relationship was more pronounced when the nearest C. orbiculatus male had greater floral production, as demonstrated by the statistical interaction (Fig 4; Table 7). Models including fixed effects associated with the nearest conspecific male were not well-supported (ΔAICc > 70 compared to top model, Akaike weight < 0.0001), though they performed better than the null model (Table 7).
Data from open-pollinated pistillate plants. Each point represents the proportion hybrids from a single pistillate C. scandens (n = 18 pistillate plants, with 259 total seedlings). Red triangles represent the six pistillate individuals for which the nearest C. orbiculatus had the fewest flowers, blue squares represent the six pistillate individuals for which the nearest C. orbiculatus had the most flowers. Black circles are pistillate plants for which the nearest C. orbiculatus had an intermediate number of flowers. Curves represent the predicted values from the best performing GLM, and colors correspond to the points.
Discussion
Our study was designed to investigate the reproductive interactions between a declining native plant, C. scandens, and an aggressive invasive congener, C. orbiculatus. Celastrus orbiculatus thrives across the range of conditions where C. scandens occurs [37], creating the potential for interactions with the invasive vine in most places where C. scandens is found. We chose a study site where both species occur in roughly equal abundances, and we investigated fertilization and hybridization patterns in both species. Our findings confirmed our hypothesis that reproductive interference of C. scandens occurs in the presence of C. orbiculatus.
Despite similar abundances of individuals carrying staminate flowers (n = 19 for C. scandens and n = 25 for C. orbiculatus), the invasive vine had an extreme advantage in male floral production, producing hundreds more flowers per staminate plant than the native. In C. orbiculatus, the axillary positioning of inflorescences allows flowers to form along an entire stem—as opposed to the terminal inflorescences in C. scandens that are limited to the end of a stem [30]. An extreme advantage in flower production by invasive species compared to their native relatives seems to be a common trait of invasive plants [18]. The 200-fold advantage for C. orbiculatus we found in terms of male floral production (Fig 2) is similar to the average reported in a recent meta-analysis [18]. These authors suggest that high floral production of invasive plants allow them to avoid pollen limitation by attracting more pollinators than native species. The invasive species also had an advantage in female floral production, with 65 times more flowers per pistillate plant. Both species had similar rates of fertilization and fruit development in this study, and C. orbiculatus was found to have greater seed germination rates in other studies [31, 40]. Thus, we also have evidence of a large advantage in female fecundity for the invasive vine. For C. orbiculatus, fruit set decreased in pistillate plants with greater total output, where the strongest predictor of fruit development was a negative association with total floral production (Table 4). This may indicate that C. orbiculatus experiences resource limitation at our study site.
The high male fecundity of C. orbiculatus appears to explain the increased fertilization rate in nearby pistillate plants of both species (Fig 2). The best predictors of fertilization rate differed between the two species. For the native species, both increased conspecific and heterospecific mating potential were correlated with higher fertilization rates. For the invasive species, only conspecific mating potential was significantly correlated with fertilization. Thus, proximity to flowering staminate C. orbiculatus increases fertilization of both the native and invasive species. These results support an emerging pattern that underscores the threat that heterospecific pollen from invasive plants may pose to native plant reproductive success [12, 17]. In this case, it is likely that high rates of heterospecific pollination lead to usurpation of C. scandens ovules through production of hybrids. Our previous study found that hybrids show reduced seed set and small, likely inviable pollen [29], so the reproductive effort of C. scandens is probably wasted on production of hybrid offspring. Early-flowering C. scandens may be more susceptible to hybridization because of greater exposure to heterospecific pollen as peak flowering in C. orbiculatus occurs at about the same time as the start of flowering in C. scandens.
Once a flower is fertilized, none of our potential explanatory variables predicted development to fruit in C. scandens (Table 5). With regards to the variables we measured, all fertilized flowers have the same likelihood of developing into fruit. Most importantly, we did not find statistical evidence that flowers exposed to large amounts of heterospecific pollen were less likely to develop into fruit (i.e., were more likely to be aborted). This pattern is consistent with the high rates of hybridization we measured in C. scandens seedlings. Genetic tests of seedlings showed that in C. scandens, 39% of seedlings tested were hybrids, while hybrids comprised only 1.6% of seedlings from C. orbiculatus. These values are notable because of the extreme asymmetry (nearly unidirectional) and because such a high proportion of C. scandens seedlings were hybrids. The proportion of hybrid seedlings in C. scandens is large compared to some other studies that measured hybridization rates between native and introduced species, such as the 16% and less reported for Eucalyptus nitens and E. ovata in Tasmania [54] or the 30% and less for Eucalyptus benthamii and E. viminalis in New South Wales [55], but it is not unprecedented. In unmanipulated plots, Burgess et al. [56] found 77% of seedlings from Morus rubra were hybrids, and Zalapa et al. [57] reported that 55% of elm trees sampled in Wisconsin were hybrids between native Ulmus rubra and the invasive U. pumila.
Interestingly, neither the proximity nor floral production of conspecific males were associated with hybridization rate in C. scandens. On the other hand, hybridization in C. scandens was higher when staminate C. orbiculatus were nearby (Fig 4), and the spatial scale for the impact of heterospecific pollination was relatively large. The best model predicts a hybridization rate of 43% (s.e. = 4.0%) for C. scandens that were 50 meters from the nearest C. orbiculatus staminate plant of mean size, and 11% (s.e. = 4.0%) for individuals that were 100 meters away (Fig 4). Native bees thought to be the primary pollinators for Celastrus can have flight distances much greater than 100 meters [58]. Our results are comparable to previous work on Morus [22, 56]. After removing all introduced individuals and hybrids within 50 meters of the native individuals, Burgess et al. [22] found 63% of seedlings collected from M. rubra were fertilized by M. alba or hybrids. We previously reported only 20 hybrids in a widespread sample of 239 (8.4%) non-native Celastrus [29]. In our system, the threat posed by heterospecific pollination is not replacement by hybrids or introgression, but rather asymmetric reproductive interference.
Results from the experimental hand crosses showed that C. scandens readily accepts C. orbiculatus pollen, but C. orbiculatus largely rejects pollen from C. scandens, providing direct evidence that the two species react differently when heterospecific pollen arrives to a stigma. Such differences in prezygotic barriers can be responsible for the creation of asymmetric hybridization [19], and indeed the few Celastrus hybrids identified in the field all had a C. scandens seed parent [29]. The difference in the ability to recognize and reject heterospecific pollen might be expected in this system; the native range of C. orbiculatus overlaps with several congeners, while C. scandens is not naturally sympatric with any closely related species [32]. Heterospecific tolerance or avoidance is likely to evolve in regions with higher floral diversification [59] and the degree of tolerance may be related to the recipient’s history of exposure to heterospecific pollen [60]. Pollen germination and pollen tube growth may be difficult to prevent between closely related species because they share similar stigma morphology and pollen-stigma recognition systems [12, 19, 61]. However, invasive species may be especially tolerant of heterospecific pollen. In a coastal dune system investigated by Suárez‐Mariño [17], the invasive species (Bidens pilosa) received more heterospecific pollen than the natives (Cakile edentula and Scaevola plumieri), but heterospecific pollen receipt only had negative effects on the native species. In our system, it is unlikely that C. orbiculatus received more heterospecific pollen than C. scandens, but the hand pollination results indicate that the invasive species would be able to reject almost all it received. Thus, differences in male fecundity as well as different responses to heterospecific pollen both contribute to asymmetric hybridization in this system.
Overall, the results of our study strongly suggest that the decline of C. scandens in its historical range is due at least in part to reproductive interference by C. orbiculatus. Although C. orbiculatus has been shown to be a superior competitor in several respects and can escape the negative effects of high density [33, 40, 62], the targeted nature of reproductive interference and the spatial scale of its action make it a more likely explanation than resource competition alone. Celastrus orbiculatus is a recent invader of North America that continues to spread and proliferate. The density-dependence associated with reproductive interference through heterospecific pollination suggests that the decline of C. scandens will accelerate and spread over a larger area, unless large and broad efforts to reverse the spread of C. orbiculatus are undertaken. A feedback between decreased abundance and increased heterospecific pollination in C. scandens may occur [63], potentially leading to extirpation from much of its native range. Efforts to conserve C. scandens can derive guidance from our results, such as reducing exposure to C. orbiculatus pollen and providing substantial isolation from C. orbiculatus, as reproductive interference may span tens to hundreds of meters. Previous work showed that hybrids have reduced fecundity [29], so introgression does not threaten the genetic identity of C. scandens at this time. Thus, C. scandens populations that were previously exposed to heterospecific pollen may still represent pure lines that are suitable for conservation or re-establishment.
Acknowledgments
We thank Ralph Grundel, Krystalynn Frohnapple, Jennifer Ison, Weihua Gao, and the UIC Center for Clinical and Translational Science for assistance in planning this study. Permission to conduct the study was granted by the National Park Service and aided by Joy Marburger. Field, laboratory, and greenhouse data collection was aided by Krystalynn Frohnapple, James Daley, Sarah Dean, Karen Zaya, Saji Abraham, Kevin Feldheim, Kirsten Dittel, Dominick Fosco, Iryna Shak, Jim Scios, Noranne Magee, John Podejko, and James Rollins. Stuart Wagenius helped immensely with data analysis. Boris Igic, Jeremie Fant, Henry Howe, Janet Rizner Backs, Eun Sun Kim, John Wilk, Emi Kuroiwa, Jason Palagi, Jenny Zambrano, Wesley Bickford, David W. Inouye, and two anonymous reviewers provided feedback that improved the manuscript. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. This manuscript was completed in partial fulfillment of the doctoral degree from the Graduate College at the University of Illinois at Chicago to DNZ.
References
- 1. Van Drunen WE, van Kleunen M, Dorken ME. Consequences of clonality for sexual fitness: Clonal expansion enhances fitness under spatially restricted dispersal. Proceedings of the National Academy of Sciences. 2015;112. pmid:26195748
- 2. Seebens H, Blackburn TM, Dyer EE, Genovesi P, Hulme PE, Jeschke JM, et al. Global rise in emerging alien species results from increased accessibility of new source pools. Proceedings of the National Academy of Sciences. 2018;115: E2264–E2273. pmid:29432147
- 3. Seebens H, Blackburn TM, Dyer EE, Genovesi P, Hulme PE, Jeschke JM, et al. No saturation in the accumulation of alien species worldwide. Nature Communications. 2017;8: 14435. pmid:28198420
- 4. Levine JM, Vilà M, Antonio CMD, Dukes JS, Grigulis K, Lavorel S. Mechanisms underlying the impacts of exotic plant invasions. Proceedings of the Royal Society of London Series B: Biological Sciences. 2003;270: 775–781. pmid:12737654
- 5. Vilà M, Espinar JL, Hejda M, Hulme PE, Jarošík V, Maron JL, et al. Ecological impacts of invasive alien plants: a meta‐analysis of their effects on species, communities and ecosystems. Ecology Letters. 2011;14: 702–708. pmid:21592274
- 6. Johnson MG, Lang K, Manos P, Golet GH, Schierenbeck KA. Evidence for genetic erosion of a California native tree, Platanus racemosa, ongoing introgressive hybridization with an introduced ornamental species. Conservation Genetics. 2016;17: 593–602.
- 7. Powell KI, Chase JM, Knight TM. Invasive plants have scale-dependent effects on diversity by altering species-area relationships. Science. 2013;339: 316–8. pmid:23329045
- 8. Didham RK, Tylianakis JM, Gemmell NJ, Rand TA, Ewers RM. Interactive effects of habitat modification and species invasion on native species decline. Trends in Ecology & Evolution. 2007;22: 489–496. pmid:17673330
- 9. Pearse IS, Sofaer HR, Zaya DN, Spyreas G. Non‐native plants have greater impacts because of differing per‐capita effects and nonlinear abundance-impact curves. Ecology Letters. 2019;22: 1214–1220. pmid:31112014
- 10. Morales CL, Traveset A. Interspecific pollen transfer: Magnitude, prevalence and consequences for plant fitness. Critical Reviews in Plant Sciences. 2008;27: 221–238.
- 11. Albrecht M, Padron B, Bartomeus I, Traveset A. Consequences of plant invasions on compartmentalization and species’ roles in plant-pollinator networks. Proceedings of the Royal Society B: Biological Sciences. 2014;281: 20140773–20140773. pmid:24943368
- 12. Arceo-Gómez G, Ashman T-L. Invasion status and phylogenetic relatedness predict cost of heterospecific pollen receipt: implications for native biodiversity decline. Journal of Ecology. 2016;104: 1003–1008.
- 13. Aizen MA, Morales CL, Morales JM. Invasive mutualists erode native pollination webs. PLoS Biology. 2008;6: e31. pmid:18271628
- 14. Lopezaraiza–Mikel ME, Hayes RB, Whalley MR, Memmott J. The impact of an alien plant on a native plant–pollinator network: an experimental approach. Ecology Letters. 2007;10: 539–550. pmid:17542933
- 15. Dietzsch AC, Stanley DA, Stout JC. Relative abundance of an invasive alien plant affects native pollination processes. Oecologia. 2011;167: 469–479. pmid:21484398
- 16. Matsumoto T, Takakura K-II, Nishida T. Alien pollen grains interfere with the reproductive success of native congener. Biological Invasions. 2010;12: 1617–1626.
- 17. Suárez‐Mariño A, Arceo‐Gómez G, Sosenski P, Parra‐Tabla V. Patterns and effects of heterospecific pollen transfer between an invasive and two native plant species: the importance of pollen arrival time to the stigma. American Journal of Botany. 2019;106: 1308–1315. pmid:31553505
- 18. Burns JH, Bennett JM, Li J, Xia J, Arceo‐Gómez G, Burd M, et al. Plant traits moderate pollen limitation of introduced and native plants: a phylogenetic meta‐analysis of global scale. New Phytologist. 2019;223. pmid:31116447
- 19. Ashman T-L, Arceo-Gómez G. Toward a predictive understanding of the fitness costs of heterospecific pollen receipt and its importance in co-flowering communities. American Journal of Botany. 2013;100: 1061–1070. pmid:23624924
- 20. Anttila CK, Daehler CC, Rank NE, Strong DR. Greater male fitness of a rare invader (Spartina alterniflora, Poaceae) threatens a common native (Spartina foliosa) with hybridization. American Journal of Botany. 1998;85: 1597–1601. pmid:21680319
- 21. Prentis PJ, White EM, Radford IJ, Lowe AJ, Clarke AR. Can hybridization cause local extinction: a case for demographic swamping of the Australian native Senecio pinnatifolius by the invasive Senecio madagascariensis? New Phytologist. 2007;176: 902–912. pmid:17850249
- 22. Burgess KS, Morgan M, Husband BC. Interspecific seed discounting and the fertility cost of hybridization in an endangered species. New Phytologist. 2008;177: 276–283. pmid:17944826
- 23. Barnes WJ, Cottam G. Some Autecological Studies of the Lonicera X bella Complex. Ecology. 1974;55: 40–50.
- 24. Vilà M, D’Antonio CM. Hybrid vigor for clonal growth in Carpobrotus (Aizoaceae) in Coastal California. Ecological Applications. 1998;8: 1196.
- 25. Ayres DR, Zaremba K, Sloop CM, Strong DR. Sexual reproduction of cordgrass hybrids (Spartina foliosa X alterniflora) invading tidal marshes in San Francisco Bay. Diversity and Distributions. 2007;14: 187–195.
- 26. Balao F, Casimiro-Soriguer R, García-Castaño JL, Terrab A, Talavera S. Big thistle eats the little thistle: does unidirectional introgressive hybridization endanger the conservation of Onopordum hinojense? New Phytologist. 2015;206: 448–458. pmid:25401776
- 27. Del Tredici P. Untangling the twisted tale of oriental bittersweet. Arnoldia. 2014;71: 2–18.
- 28. Pooler MR, Dix RL, Feely J. Interspecific hybridizations between the native bittersweet, Celastrus scandens, and the introduced invasive species, C. orbiculatus. Southeastern Naturalist. 2002;1: 69–76.
- 29. Zaya DN, Leicht-Young SA, Pavlovic NB, Feldheim KA, Ashley MV. Genetic characterization of hybridization between native and invasive bittersweet vines (Celastrus spp.). Biological Invasions. 2015;17: 2975–2988.
- 30. Steward AM, Clemants SE, Moore G. The concurrent decline of the native Celastrus scandens and spread of the non-native Celastrus orbiculatus in the New York City metropolitan area. Journal of the Torrey Botanical Society. 2003;130: 143.
- 31.
Leicht SA. The Comparative Ecology of an Invasive Bittersweet Species (Celastrus orbiculatus) and its Native Congener (C. scandens). Doctoral dissertation, University of Connecticut Storrs. 2005.
- 32. Hou D. A revision of the genus Celastrus. Annals of the Missouri Botanical Garden. 1955;42: 215–302.
- 33. Pavlovic NB, Leicht-Young SA. Are temperate mature forests buffered from invasive lianas? Journal of the Torrey Botanical Society. 2011;138: 85–92.
- 34. Brizicky GC. The genera of Celastrales in the southeastern United States. Journal of the Arnold Arboretum. 1964;45: 206–234.
- 35. Wyman D. Fruiting habits of certain ornamental plants. Arnoldia. 1950;10: 81–85.
- 36.
Patterson DT. The ecology of oriental bittersweet, Celastrus orbiculatus, a weedy introduced ornamental vine. Doctoral dissertation, Duke University. 1974.
- 37. Leicht-Young SA, Silander JA, Latimer AM. Comparative performance of invasive and native Celastrus species across environmental gradients. Oecologia. 2007;154: 273–282. pmid:17724616
- 38. McNab WH, Meeker M. Oriental bittersweet: A growing threat to hardwood silviculture in the Appalachians. Northern Journal of Applied Forestry. 1987;4: 174–177.
- 39. Fike J, Niering WA. Four decades of old field vegetation development and the role of Celastrus orbiculatus in the northeastern United States. Journal of Vegetation Science. 1999;10: 483–492.
- 40. Dreyer GD, Baird LM, Fickler C. Celastrus scandens and Celastrus orbiculatus—Comparisons of reproductive potential between a native and introduced woody vine. Bulletin of the Torrey Botanical Club. 1987;114: 260–264.
- 41. Leicht-Young SA, Pavlovic NB, Grundel R, Frohnapple KJ. Distinguishing native (Celastrus scandens L.) and invasive (C. orbiculatus Thunb.) bittersweet species using morphological characteristics. Journal of the Torrey Botanical Society. 2007;134: 441–450.
- 42. Anderson EC, Thompson EA. A model-based method for identifying species hybrids using multilocus genetic data. Genetics. 2002;160: 1217–1229. pmid:11901135
- 43. Falush D, Stephens M, Pritchard JK. Inference of population structure using multilocus genotype data: dominant markers and null alleles. Molecular Ecology Notes. 2007;7: 574–578. pmid:18784791
- 44. Wesselingh RA. Pollen limitation meets resource allocation: towards a comprehensive methodology. New Phytologist. 2007;174: 26–37. pmid:17335494
- 45.
Young JA, Young CG. Seeds of woody plants in North America. Portland, Oregon: Dioscorides Press; 1992.
- 46. Wagenius S, Lonsdorf E, Neuhauser C. Patch aging and the S-allee effect: Breeding system effects on the demographic response of plants to habitat fragmentation. American Naturalist. 2007;169: 383–397. pmid:17230399
- 47.
R Core Team. R: a language and environment for statistical computing, 3.6.2. Document freely available on the internet at: http://www.r-project.org. Vienna, Austria: R Foundation for Statistical Computing; 2019
- 48. Bates D, Mächler M, Bolker B, Walker S. Fitting Linear Mixed-Effects Models using lme4. Journal of Statistical Software. 2015;67:1–48.
- 49. Zhang D. rsq: R-Squared and related measures. R package version 1.1. 2018.
- 50. Zhang D. A Coefficient of Determination for Generalized Linear Models. The American Statistician. 2017;71: 310–316.
- 51. Nakagawa S, Johnson PCD, Schielzeth H. The coefficient of determination R2 and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded. Journal of the Royal Society Interface. 2017;14. pmid:28904005
- 52. Nakagawa S, Schielzeth H. A general and simple method for obtaining R2 from generalized linear mixed‐effects models. Methods in Ecology and Evolution. 2013;4: 133–142.
- 53. Barton K. MuMIn: Multi-Model inference. R package version 1.42.1. 2018.
- 54. Barbour RC, Potts BM, Vaillancourt RE, Tibbits WN, Wiltshire RJEE. Gene flow between introduced and native Eucalyptus species. New Forests. 2002;23: 177–191.
- 55. Butcher PA, Skinner AK, Gardiner CA. Increased inbreeding and inter-species gene flow in remnant populations of the rare Eucalyptus benthamii. Conservation Genetics. 2005;6: 213–226.
- 56. Burgess KS, Morgan M, Deverno L, Husband BC. Asymmetrical introgression between two Morus species (M. alba, M. rubra) that differ in abundance. Molecular Ecology. 2005;14: 3471–3483. pmid:16156816
- 57. Zalapa JE, Brunet J, Guries RP. Patterns of hybridization and introgression between invasive Ulmus pumila (Ulmaceae) and native U. rubra. American Journal of Botany. 2009;96: 1116–1128. pmid:21628262
- 58. Greenleaf SS, Williams NM, Winfree R, Kremen C. Bee foraging ranges and their relationship to body size. Oecologia. 2007;153: 589–596. pmid:17483965
- 59. Arceo-Gómez G, Schroeder A, Albor C, Ashman T-L, Knight TM, Bennett JM, et al. Global geographic patterns of heterospecific pollen receipt help uncover potential ecological and evolutionary impacts across plant communities worldwide. Scientific Reports. 2019;9: 8086. pmid:31147606
- 60. Arceo-Gómez G, Raguso RA, Geber MA. Can plants evolve tolerance mechanisms to heterospecific pollen effects? An experimental test of the adaptive potential in Clarkia species. Oikos. 2016;125: 718–725.
- 61. Martin FW. Pollen Germination on Foreign Stigmas. Bulletin of the Torrey Botanical Club. 1970;97: 1.
- 62. Ashton IW, Lerdau MT. Tolerance to herbivory, and not resistance, may explain differential success of invasive, naturalized, and native North American temperate vines. Diversity and Distributions. 2008;14: 169–178.
- 63. Levin DA, Francisco-Ortega J, Jansen RK. Hybridization and the extinction of rare plant species. Conservation Biology. 1996;10: 10–16.