The effect of alien species on plant-pollinator network structure across a gradient of plant species invasion

The interactions between pairs of native and alien plants via shared use of pollinators have been widely studied. Studies of invasive species effects at the community level on the other hand are still scarce. Few community level studies, however, have considered how differences in the intensity of invasion, and degree of floral trait similarity between native and invasive species, can mediated effects on native plant-pollinator communities. Here, we evaluated the effect of alien species on overall plant-pollinator network structure, and species-level network parameters, across nine coastal communities distributed along 205 km at Yucatán, México that vary in alien species richness and flower abundance. We further assessed the effect of alien plant species on plant-pollinator network structure and robustness via computational simulation of native and invasive plant extinction scenarios. We did not find significant differences between native and alien species in functional floral phenotypes, the visitation rate and species composition of the pollinator community. Variation in the proportion of alien plant species and flower abundance across sites did not affect plant-pollinator networks structure. Species-level network parameters (i.e., normalized degree and nestedness contribution) did not differ between native and alien species. Furthermore, our simulation analyses revealed that alien species are functionally equivalent to native species and contribute equally to network structure and robustness. Overall, our results suggest that alien species are well integrated into native coastal plant-pollinator networks which may be facilitated by high levels of floral trait similarity and pollinator use overlap. As a result, alien species may play a similar role than that of natives in the structure and stability of native plant and pollinator communities in the studied coastal sand dune ecosystem.

Introduction Fig). The sand dune ecosystem is continuous along the entire coast, which extends over approximately 320km but is interrupted in a few areas by mangrove and lagoon systems [35]. Thus, the studied area 129 encompasses the full distribution of the sand dune ecosystem in the north coast of the Peninsula. We selected 130 nine sites (i.e. co-flowering communities) with different "levels" of invasiveness previously identified along

249
For the latter two scenarios we removed as many plant species from the network as the number alien species 250 at each site. We compared the structure of these networks with that of (d) "intact" (observed) networks that 251 included all plant species at a site (i.e., natives + aliens). By comparing the structure of "intact" networks to 252 those were only aliens were removed (i.e. "aliens removed") we evaluated the potential effect of alien 254 vs. "native removed " and "random removal" extinction scenarios, we evaluated if the effect of removing

262
To statistically evaluate the differences in network structure among the 'extinction scenarios', we 263 conducted mixed GLMM models with extinction scenario ('intact', 'aliens removed', 'natives removed' and 264 'random') as a fixed effect and site as random effect. We evaluated differences on the following network-

288
The average percentage of alien flowers was high (75.7 ± 30.32), although we observed high among-site 289 variation in the proportion of alien flowers (Table 1; range from 11% to 99%). A total of 14255 floral visits 290 were recorded. The lowest number of total flower visits was observed at Charcas (504) and the highest at 291 Sisal (3335) (Fig 1). Mean pollinator richness per plant species was highly variable among sites, being the 292 highest at Playa Maya and the lowest at Sisal (Fig 2a). However, no significant differences were observed 293 among sites or between native and alien plants (F ≤ 2.05, p > 0.05 in both cases). Although we observed 294 significant differences among sites in pollinator species composition (PERMANOVA, pseudo-F 8, 26 = 1.83, p 295 < 0.05), no differences were observed between native and alien plant species (PERMANOVA, pseudo-F 1,26 = 296 0.7, p > 0.05). Furthermore, we did not find significant differences in pollinator visitation rate among sites 297 (Fig 2b), or between alien and native species (F ≤ 0.83, p > 0.4 in both cases). However, pollinator species

313 314
Plant-pollinator networks 315 We recorded a total of 30 insect-pollinated plant species belonging to 19 families (S2 Table). A total of 73 316 insect species were recorded belonging to three orders: Diptera (26 species), Hymenoptera (27 species) and between eight and 17 plant species and 22 and 38 pollinator species (Table 2; Fig 3).  Fig 3). Overall, the most visited native plants were Cakile 324 edentula and Scaevola plumieri (Fig 3) and the most visited invasive plants were Bidens pilosa and 325 Alternanthera microcephala (Fig 3). Overall specialization (H2  . the 'degree of invasiveness' did not affect plant-pollinator network structure. normalized degree and nestedness contribution (Table 3). In contrast, we observed a significant effect of 343 plant origin on interaction strength (Table 3). However, this difference occurred only at one site (Chapo 2),

344
where alien plants showed higher interaction strength than natives (t 47 = 3.32, p = 0.002). Species-level 347 Effect of alien species on plant-pollinator network structure via 352 simulation of extinction scenarios 353 We observed significant differences in network specialization and nestedness among the different 354 extinction scenarios (F 3,24 ≥ 3.69, p < 0.05; Fig 4), but this was not the case for modularity (F 3,24 = 0.55, p > 355 0.05). Overall network specialization in the 'natives removed' scenario was significantly higher compared to 356 all other extinction scenarios and to the 'intact' network (t ≥ 3.5, p < 0.05, in all cases, Fig 4a) suggesting that 357 the loss of native species may increase network specialization. The effect size of network specialization 358 showed an 18% increase when native plants were removed compared to 'intact' communities (Table 4; S2 359 Figure).

365
We observed a significant decrease in network nestedness in the 'intact' network compared to the other 366 extinction scenarios (t ≥ 3.1, p > 0.05 in all cases, Fig 4b), with the exception of the 'natives removed' scenario 367 (Fig 4b). The effect size for nestedness showed a decrease of 19% in the 'intact' network compared to all 368 extinction scenarios (Table 4; S2 Figure). However, the decrease in nestedness in the 'aliens removed' scenario 369 was almost twice as high as in the 'native removed' scenario (13% and 6% respectively), suggesting that alien 371 'intact' network vs. 'native removal' and 'random removal' scenarios showed that removal of plant species, 372 regardless of their origin, significantly reduces network robustness (Table 4; Fig 4c). The effect size for 373 modularity did not change significantly between any scenario (Table 4; S2 Figure).
374 375 Table 4. Mean (± SE) plant-pollinator interaction network metrics and average rate of change (∆ %, according 376 to Hedge´s size effects) under different extinction scenarios in nine sites along the north coast of Yucatan,
Network topology estimator Random models Intact networks Effects of alien species at species level network 382 We found a significant effect of plant origin (nested in site) on normalized degree and strength (F ≥ 2.05, 383 p<0.01 in both cases), but not for nestedness contribution (F 17,190 = 1.45, p=0.11). However, for normalized 393 0.059). Furthermore, for none of the species-level parameters we found significant differences between native 394 or alien 'removal' scenarios (t ≤ 0.7, p ≥ 0.4 in all cases).

396
Our results show that none of the evaluated network parameters is affected by increasing 'intensity' of plant 397 invasion (alien species richness and flower abundance), suggesting that even low alien species richness or 398 flower abundance can have significant impacts on native plant and pollinator communities. Our results also 399 suggest that alien species are well integrated in native plant-pollinator networks (Fig 3), and that pollinator 400 use overlap with natives is high, which is likely mediated by high levels of floral trait similarity between 401 alien and native species. Consistent with these results our simulated extinction scenarios suggest that alien 402 species play an equivalent role to natives in network structure and stability in our studied coastal plant 403 communities. These and other results are discussed in detail below.

404
Contrary to our expectations, and in spite of extensive among-site variation in alien species richness 405 and floral availability, we did not observe among-site differences in any of the network structural parameters 406 suggesting that the plant-pollinator network structure is not affected by the 'intensity' of species invasion. It 407 is possible that this lack of effect may be mediated by the high abundance of the introduced honey bee A.
408 mellifera at our study sites, which contributes disproportionally to flower visitation at all sites (ca. 60% of 409 visits; see Fig 3). It has been shown that 'super generalist' pollinators such as honeybees can facilitate the 410 integration of alien plants into native pollination networks and support the structure of the network in the 411 presence of alien species [11,16]. Thus, the high incidence of A. mellifera at our study sites may be a key 412 factor mediating plant-pollinator network structure regardless of alien species richness and flower abundance 413 at a site (i.e. intensity of invasion). Interestingly, in our study, the only community that did not show a 414 significant nested structure (Charcas site ; Fig 3) was also the one with the lowest proportion of A. mellifera 415 visits (5.7%), lending support to the prediction that this 'super generalist' pollinator plays an important role hampered our ability to observed significant effects (