Non-invasive naturalized alien plants were not more pollen-limited than invasive aliens and natives in a common garden

Study species and location

To test whether pollen limitation is a barrier to invasion success, we did a breeding-system experiment in the research garden of the University of Konstanz, Germany (47°41′31·49″N; 9°10′46·09″E). In this experiment, we assessed the degrees of pollen limitation and autofertility of a total of 24 alien and native plant species (Table S1 in Supporting Information). To avoid bias due to different degrees of relatedness among species, we selected triplets of confamilial (three triplets) or congeneric (five triplets) species, with each triplet comprising a native, an invasive alien and a non-invasive alien plant species (Table S2). We included as many species as possible, but this was limited by the availability of three closely related species representing each plant group within a triplet. We selected herbaceous species occurring in Germany, and we classified them as alien (neophytes) or native based on their status in the BiolFlor data base of the German flora (Kühn, Durka & Klotz 2004). The alien species were classified as invasive or non-invasive based on their frequency of occurrence in Germany, according to the Floramap website (http://www.floraweb.de/pflanzenarten/cg_floramap/). We classified aliens as ‘invasive’ if they have been recorded in more than 600 grid cells (20% of the total number of grid cells in Germany) since 1980 and as ‘non-invasive’ when they have been recorded in fewer than 200 grid cells. All invasive aliens in our study have been reported from 633 to 2331 (median: 1733) grids cells, while all non-invasive aliens have been reported from 5 to 181 (median: 94·5) grid cells. All native species we used have been reported from 64 to 2656 (median: 2207) grid cells. Although the distribution size of species at larger scales does not necessarily correlate strongly with abundance at smaller scales (e.g. Dawson et al. 2013), we think that grid-cell frequency gives a good indication of the spread of species in the landscape, which is in accord with the definition of invasion by Richardson et al. (2000a). For 15 of the 16 alien species, we found information on the year of first record in Germany (BiolFlor data base) or the Czech Republic (Pyšek, Sádlo & Mandák 2002), or on the year of first cultivation (Table S1). All of these alien species have a residence time of at least 100 years in central Europe, and the time since introduction was comparable for invasive and non-invasive species, suggesting that differences in invasiveness are not simply due to differences in residence time. Moreover, there was no significant correlation between grid-cell frequency and the frequency of introduction of species, measured as the number of botanical gardens where a species is grown according to SysTax (http://www.biologie.uni-ulm.de/systax/; Kendall's tau = 0·301; P = 0·123). So, high (or low) grid-cell frequency does not necessarily indicate high (or low) introduction frequency. The geographic origin of the alien species in our study includes temperate, subtropical and tropical regions (Table S1).

The research garden is situated outside the city of Konstanz and surrounded by forests and grasslands, typical for a Central European semi-natural landscape. We did the study in a common garden for four major reasons. First, because many invasions have started from gardens (Hulme 2011), studying pollen limitation in such an environment is highly relevant. Secondly, a common-garden experiment avoids potentially confounding effects due to site variation (e.g. variation in population size and dissimilarities in pollinator communities). Thirdly, the plants were grown under common environmental conditions with the same watering and fertilization regime to avoid variation in seed production due to resource limitation. Thus, differences in seed production are not confounded with the effect of resource fluctuations occurring in the wild. Fourthly, it would have been difficult to find non-invasive alien species in the wild because they are less frequent.

Plant materials

To simulate the relatively small size of a population at an early stage of invasion, we initially grew 60 individuals (c. 12 individuals per m²) of each species outside in the common garden. To this aim, we obtained seeds from commercial seed suppliers, botanical gardens, and our own seed-germplasm collection at the University of Konstanz (Table S2). We sowed c. 100–200 seeds of each species in a 0·6-L tray. For most species, we sowed the seeds in June and July 2012. For Geranium macrorrhizum and Lepidium densiflorum, we sowed the seeds in November 2012. For Eschscholzia californica, Papaver somniferum and additional Bidens connata, we sowed the seeds in March 2013. We individually repotted the seedlings into pots filled with commercial potting soil (Einheitserde, Sinntal-Altengronau, Germany). For a few species that we could not grow from seeds, we dug out seedlings from a forest and a grassland site near the University of Konstanz (Heracleum mantegazzianum and Impatiens glandulifera) and from the botanical garden of Bern, Switzerland (Impatiens balfourii). We then let the plants grow in a greenhouse until July 2012 (Bidens and Impatiens triplets) and until May 2013 for most species, after which we positioned them outside in the garden. According to the sizes of the species, we used pots of different sizes (Table S2). Overall, 12–60 plants per species were flowering and therefore suitable for the experiment. Timing of flowering varied among species and, in many cases, within triplets, but the order of flowering was not confounded with status of the species (see Table S2). Because the species did not flower at the same time and because of the large workload, we spread the experiment over three consecutive years (2012–2014, Table S2). We always treated all the species within a triplet in the same year, except for Bidens connata for which not enough data could be collected in the first year. Species that were sown in the first year but did not flower in the same year were kept over winter in a non-heated greenhouse. All 60 potted individuals of the native Impatiens noli-tangere died before we applied the treatments, probably due to an infestation by aphids. Therefore, treatments imposed on this species were undertaken in two nearby natural populations of approximately 50 flowering individuals c. 100 m away from the research garden, occupying each c. 10 m². Our results were robust as the results remained qualitatively the same when we excluded this species from the analysis (Tables S3 and S4).

Breeding-system experiment

To assess the degree of pollen limitation and the degree of autofertility of each of the 24 species, our breeding-system experiment included three treatments: pollen supplementation, open pollination and pollinator exclusion. Each treatment was done on a separate plant. The number of individual plants per treatment varied from 2 to 25 per species (15 in most cases), on a total of 12–60 flowering plants per species (Table S2). The treatments were applied to 1–28 (10 in most cases) flower units per plant (Table S2). For convenience, we use the term ‘flower unit’ to refer to individual flower for most species and capitulum for the Asteraceae.

In the pollen-supplementation treatment, we took pollen from a conspecific pollen donor and used tweezers or toothpicks to place the pollen on the stigmas of recipient plants. We marked the hand-pollinated flowers. To avoid potential bias due to the quality of pollen of a single donor, we used, whenever possible, different plants as pollen donors for each hand-pollinated flower of a recipient plant. For the Asteraceae, to ensure that each floret within a capitulum was pollinated, we repeated the treatment several times. In the open-pollination treatment, we marked unmanipulated flowers and left them open to natural pollination. In the pollinator-exclusion treatment, we used white nylon-mesh bags to exclude pollinators. We marked and bagged flower buds before anthesis. To avoid self-pollination due to contact with the bags, the exclusion bags were supported by metal wires. In total, we treated 16 824 flower units (pollen supplementation: 5173; open pollination: 6124; pollinator exclusion: 5527) on a total of 893 individual plants.

Two to 6 weeks after we did the treatments, we scored the number of flower units that produced fruits. We collected the developed fruits and stored them in individual paper bags. We randomly subsampled up to five fruits per plant (up to 25 for Lepidium and Solidago species) and then counted the number of seeds they contained. This enabled us to compare fruit set and seed production per fruit among the three pollination treatments (Fig. S1). Because many plants of the invasive Heracleum mantegazzianum died after we applied the treatment, we did not obtain data from the pollinator-exclusion treatment for this species.

Data analysis

Indices of pollen limitation and autofertility of native, invasive alien and non-invasive alien plant species

In addition to the analyses above on fruit set and seed production per fruit, to compare the degrees of pollen limitation and autofertility among native, invasive alien and non-invasive alien plant species, we also calculated overall pollen-limitation and autofertility indices per species. These indices combine fruit and seed production as they are based on the number of seeds produced per flower unit treated. We calculated indices of pollen limitation (PL) using three different formulae (Larson & Barrett 2000; Vamosi et al. 2006; Harmon-Threatt et al. 2009; Eckert et al. 2010). Because we obtained similar values of those indices (Table S6), we only show results for indices calculated as

Positive values of PL indicate pollen limitation, and PL = 0 indicates no pollen limitation (Eckert et al. 2010). Negative values of PL could occur when the number of seeds under pollen supplementation was lower than under open pollination (Vamosi et al. 2006). As it was done in previous studies (e.g. Larson & Barrett 2000), and to facilitate future syntheses, we additionally calculated an adjusted index by setting all negative values (33% of the observations) to zero.

To assess the degree of autofertility of each species, we calculated an index of autofertility (AF) as

AF = 0 indicates that the species is not autofertile (Eckert et al. 2010). Values of AF larger than one could result when the number of seeds under pollinator exclusion was larger than under pollen supplementation. We also calculated an adjusted index by setting all values larger than one (9% of the observations) to one.

To compare the degrees of pollen limitation and the degrees of autofertility among native, invasive alien and non-invasive alien plant species, we built linear models (Zuur et al. 2009) with PL and adjusted PL, and AF and adjusted AF as dependent variables, and plant group as a fixed-effect explanatory variable. To account for phylogenetic non-independence of our species in the analyses (Felsenstein 1985), we constructed a phylogeny of the 24 species using the program Phylomatic online version 3 (Webb & Donoghue 2005). We further resolved polytomies within the Phylomatic phylogenetic tree using published phylogenies based on DNA studies (Table S5, Fig. S2). We adjusted the branch lengths of the resulting tree using the bladj function in Phylocom (Webb, Ackerly & Kembel 2008), which considers the node ages (Wikström, Savolainen & Chase 2001). We included the phylogenetic correlation structure corGrafen (Grafen 1989) using the ape package (Paradis, Claude & Strimmer 2004) in our models. All the analyses above were performed using the R software, version 3.1.1 (R Core Team 2012).

Fruit set and seed production of native, invasive alien and non-invasive alien plant species under three pollination treatments

Fruit set in the pollen-supplementation and in the open-pollination treatments was close to 100% across the three plant groups (i.e. native, invasive and non-invasive alien, Fig. 1a). All three plant groups and most species in these groups (Fig. S1) set fruit in the pollinator-exclusion treatment, but fruit set was on average lower in the pollinator-exclusion than in the pollen-supplementation and open-pollination treatments (Fig. 1a). These differences in the proportion of flower units that produced fruits were indicated by a significant main effect of pollination treatment (Table 1).

Seed production per fruit was also not higher in the pollen-supplementation than in the open-pollination treatment across the three plant groups (Fig. 1b), and for most species in these groups (Fig. S1). Overall, seed production was reduced in the pollinator-exclusion compared to both the pollen-supplementation and the open-pollination treatments (Fig. 1b). This reduction was more pronounced for invasive aliens, as species in this group produced, on average, 11 seeds more per fruit in the pollen-supplementation and the open-pollination treatments compared to the pollinator-exclusion treatment, whereas non-invasive alien species produced four seeds more, and native species produced two seeds more (Fig. 1b). These differences in the number of seeds produced per fruit were indicated by a significant main effect of pollination treatment and a significant interaction between plant status and pollination treatment (Table 1). There were no significant effects of year of treatment on fruit set and seed production per fruit (Table 1).

Indices of pollen limitation and autofertility of native, invasive alien and non-invasive alien plant species

In line with the results above, the index of pollen limitation was on average low across the three plant groups (<0·2; Fig. 2a). All three plant groups were autofertile to some degree as their AF indices were >0 (Fig. 2b). Our phylogenetically informed analysis showed that the three groups did not differ significantly in their indices of pollen limitation and autofertility (Table 2). Similar results were obtained when the analyses were performed using adjusted indices of pollen limitation and autofertility, in which negative pollen-limitation values were set to zero and autofertility values larger than one were set to one (Fig. 2; Table 2). Low values of Grafen's rho indicated a weak phylogenetic signal of pollen limitation and autofertility in our set of species.

Native, invasive alien and non-invasive alien plants had low pollen limitation

Irrespective of plant group, pollen limitation was low in our common-garden experiment (Fig. 2). This finding diverges from that of several reviews of pollen supplementation experiments, which have shown that pollen limitation is widespread and that its magnitude is relatively high in natural populations (Burd 1994; Knight et al. 2005). Most species in our study had actinomorphic flowers, which are usually indicative of a generalized pollination strategy (Ollerton et al. 2007), although we included one triplet with zygomorphic flowers (the Impatiens triplet). This could explain the overall low degrees of pollen limitation we found, as generalized plants are often less pollen-limited (Martén-Rodríguez & Fenster 2010). On the other hand, Vamosi et al. (2006) showed that plants have relatively low degrees of pollen limitation in Central Europe compared to other parts of the globe, probably due to low degrees of competition for pollinators in less diverse temperate floras compared to more diverse tropical floras (e.g. Wolowski, Ashman & Freitas 2013). Therefore, the observed low levels of pollen limitation might not be specific to our common garden but might be common in other parts of Central Europe as well. Nevertheless, future studies should also compare pollen limitation of plants in gardens to plants in natural sites.

Pollen limitation was low in our relatively small artificial founder populations (up to 60 plants per species). This finding also diverges from that of previous studies, which have shown that population size is correlated to the degree of pollen limitation, with small populations likely to be more pollen-limited than large ones (e.g. Ågren 1996; Knight 2003; Ward & Johnson 2005). A high degree of pollen limitation in small populations is likely due to insufficient pollen transfer as a result of reduced flower visits, reduced conspecific and increased heterospecific pollen deposition, and increased within-plant visits (Ågren 1996; Ward & Johnson 2005). It could be that our common garden may have created a high diversity of plants with a positive feedback on overall flower visitation and thereby have resulted in low pollen-limitation values (Ghazoul 2006; Scherber et al. 2010; Bennett & Gratton 2013). This would, however, apply to most gardens, from which many invasive species have escaped (Hulme 2011). Moreover, a recent study in Switzerland showed that flower visitation to native, invasive alien and non-invasive alien plants was actually lower in garden environments than in (semi-)natural environments (Chrobock et al. 2013). Nevertheless, as the levels of pollinator and pollen limitation may vary between different situations (Knight et al. 2005), future studies should also compare pollen limitation among native, invasive and non-invasive alien species in environments with different pollinator communities.

In our study, 33% of the species had negative values of the pollen-limitation index. A previous global review of pollen limitation also showed many instances of negative pollen-limitation indices (Vamosi et al. 2006). Such negative values could result by chance due to variation in flower characteristics, such as flower size and ovule number, introduced by the use of randomly chosen plants and flowers for the pollen-supplementation and the open pollination treatments. Such chance events are more likely when there are few replicates. Indeed, the relatively low sample sizes for Heracleum manteggazianum and Papaver somniferum (Table S2) might by chance have returned negative indices for these species. Negative indices might also result from a manipulation error (Larson & Barrett 2000) and might therefore be zero or close to zero instead of negative. When we excluded the triplets with many occurrences of negative values of the pollen-limitation index (Apiaceae, Balsaminaceae and Papaveraceae) from the analysis, we obtained qualitatively similar results (Table S7). Therefore, the occurrence of negative values should not impair our conclusion that pollen limitation was overall low in our study.

Native, invasive alien and non-invasive alien plants were autofertile but still dependent on pollinators

Native, invasive alien and non-invasive alien species in our study were almost all autofertile to some degree (Figs 1 and 2b). Interestingly, some of the study species such as Aster laevis, Eschscholzia californica, Solidago canadensis, S. graminifolia, S. virgaurea are listed as self-incompatible in the BiolFlor data base (Table S1; Kühn, Durka & Klotz 2004). This may partly reflect that species with low levels of autofertility may have been classified as self-incompatible in the BiolFlor data base. This illustrates that autofertility is not an absolute yes/no characteristic of a species but a continuous one that may vary (Fenster & Martén-Rodríguez 2007; Jia & Tan 2012). An early study by Darwin (1876) showed that Eschscholzia californica is plastic in its breeding system, as offspring of self-sterile plants from Brazil were self-fertile in England, and offspring of these plants sent back to Brazil became self-sterile again after a few years. Indeed, self-incompatibility is frequently not strict (Raduski, Haney & Igić 2012), which indicates that we need more quantitative estimates of the actual degree of self-(in)compatibility and variation therein.

Because autofertile plants are able to set fruit and produce seeds without compatible mates and pollinators, this might have helped them prevent pollen limitation (Martén-Rodríguez & Fenster 2010; Hargreaves, Weiner & Eckert 2015). However, while fruit set and seed production for some species in our study were as high in the pollinator-exclusion treatment as in the pollen-supplementation treatment (Fig. S1), the degree of autofertility was not correlated with the degree of pollen limitation (Pearson's r = 0·1404; P = 0·523; n = 23; Fig. S3). This suggests that the low degrees of pollen limitation in our study were not driven by autofertility but most likely by effective pollinator visitation (e.g. Schuster et al. 1993). Future studies, therefore, should try to directly link variation in flower visitation to alien and native species to the degree of pollen limitation.

Non-invasive alien plants were not more pollen-limited than invasive alien and native plants

Our results suggest that native, invasive alien and non-invasive alien plants in our study are able to attract sufficient pollinators. Because we used confamilial triplets, our set of study species might overall, at least within each triplet, have similar traits related to attractiveness to pollinators, such as floral rewards (Ornelas et al. 2007). A recent theoretical study demonstrated that closely related plants are more likely to attract similar numbers of pollinators because visitation is related to plant size, flower colour and symmetry (Rafferty and Ives 2013). Gibson, Richardson & Pauw (2012) showed, in a field survey, that an overlap of flower visitors between native and invasive plants was associated with similarity of traits between the two plant groups. Therefore, each species within the triplets of closely related insect-pollinated plants in our study might have been visited by a similar number and set of pollinators. On the other hand, a recent study showed that alien plants that have managed to establish naturalized populations in Switzerland were better than non-naturalized plants in attracting abundant and diverse flower visitors in a botanical garden (Razanajatovo et al. 2015). This suggests that pollinator limitation might prevent naturalization under certain conditions.

The role reproductive characteristics play in invasion success may be context and stage dependent. Indeed, the invasion process is complex and involves multiple stages (Dietz & Edwards 2006; Richardson & Pyšek 2012; van Kleunen, Dawson & Maurel 2015). As we were interested in the role of pollen limitation and autofertility for invasiveness, we selected non-invasive plants that have already managed to establish a few naturalized populations as controls and have similar residence times as the invasive ones (Table S1). Our results suggest that pollen limitation and autofertility might not be important at this stage of the invasion process. It could be, however, that invasive and non-invasive aliens in our study have already overcome the pollen-limitation barrier at the naturalization stage. Indeed, the difference in pollinator visitation reported by Razanajatovo et al. (2015) was between naturalized and non-naturalized alien species. Therefore, pollinator limitation, in contrast with its importance for the naturalization process, may be less important for invasion success once the alien plants have become naturalized. Future studies should test this directly by also including non-naturalized alien species in addition to naturalized and invasive alien species.

Non-invasive alien plants were not less autofertile than invasive alien and native plants

We showed that non-invasive aliens were not less autofertile than the invasive ones (Fig. 2b). According to Baker's Law, invasive plants are more autofertile than non-invasive plants because they can establish and spread in new environments without their usual pollinators, and without compatible mates when populations are still small and most individuals are likely to be closely related (Baker 1967; Pannell & Barrett 1998). Baker's Law may apply at both the naturalization stage (van Kleunen et al. 2008) and the invasion stage (Rambuda & Johnson 2004; van Kleunen & Johnson 2007; Hao et al. 2011; Pyšek et al. 2011; Ward, Johnson & Zalucki 2012). Although there is evidence for Baker's Law as a rule, it does not mean that self-incompatible species cannot become invasive. Indeed, a few widely invasive plants require the presence of compatible mates and pollinators to produce seeds. For example, Cirsium arvense is dioecious and therefore obligately outcrossing; nevertheless, it is one of the most widely distributed invasive plants globally (Randall 2012). Possibly, traits other than breeding system may make the difference between invasive and non-invasive aliens in Central Europe.

Invasive alien plants in our study produced more seeds per fruit than non-invasive alien and native plants in both the open-pollination and the pollen-supplementation treatments (Fig 1b, Table 1). Although this pattern appears to be mostly driven by the three invasive species Papaver somniferum, Senecio inaequidens and Heracleum mantegazzianum (Fig. S1), it could be that invasive aliens can take more advantage of the presence of pollinators (i.e. attract more pollinators) compared to non-invasive aliens and natives, because they overall invest more in reproduction. Indeed, it has been shown that invasive plants have higher investment in reproduction (van Kleunen, Weber & Fischer 2010; Moravcova et al. 2010). Using 17 invasive and non-invasive alien species, Burns et al. (2013) demonstrated that higher sexual reproduction of invasive plants resulted in higher population growth rates compared to their non-invasive relatives. Therefore, the overall investment in reproduction might drive the invasion success.