Introduction Plant–animal mutualisms are highly asymmetric, such that if a plant species depends strongly on an animal species, the animal typically depends weakly on the plant, and vice versa [1,2]. Thus, the resulting mutualistic webs have a nested structure—a robust property of this type of networks [3] —whereby specialists interact preferentially with generalists, rather than with other specialists, and interactions between generalist partners form the network core [2,4]. This limited reciprocal dependence or mutualism strength might increase web stability, buffering plant and animal species against the extinction of any of their partners [1,5–8]. Additionally, a decrease in mutualism strength may indicate changes in network architecture whereby some components of an interaction network become more weakly connected, or disconnected, whereas others become more central. Here we show that integration of invasive mutualists into native plant–pollinator networks, while it does not alter overall web connectivity, decreases mutualism strength by increasing the concentration of interaction links in a few alien species. Given the arrival of propagules of alien organisms in a new locality, invasion is usually triggered by different types of mostly human-related disturbances and/or promoted by an enemy-free space that creates appropriate conditions for establishment [9,10]. Once established, aliens can increase in abundance and even dominate an entire community through a series of direct and indirect facilitative and self-perpetuating mechanisms [9–11], which can cause displacement of native competitors and disruption of their interactions [12–15]. In particular, the fate of alien flowering plants and flower-visiting animals in a novel environment may depend largely on how well they integrate into existing pollination webs [16]. If they integrate poorly—due, for instance, to a lack of coevolutionary history with their native counterparts—then their success, in terms of seed production for plants and nectar and pollen acquisition for pollinators, may be conditioned by the presence of other alien partners. However, absolute failure of aliens to integrate into native pollination webs seems unlikely, because many plant–pollinator interactions are rather unspecific and diversified, and alien mutualists have a high chance of interacting with native generalists [17,18]. In any event, preferential interaction between alien partners might create a separate network compartment with little effect on the structure of the original native web. On the other hand, because many invasive plants and pollinators are themselves highly generalist, their interactions with other alien and native species could become central in the structure of modified plant–pollinator webs. Alien integration need not to alter the architecture of former pollination networks in terms of its connectivity, or who interacts with whom. For instance, alien plants may compete with native plants for space and resources (light and water) but not necessarily for pollinators, whereas alien and native flower-visiting animals may present different activity periods or exploit different floral resources [19–21]. However, as invasion progresses, some alien mutualists may become increasingly abundant and/or change their per capita interaction strength, elevating their chance of interacting with a large number of partners [10,22,23] and, as long as network connectivity remains constant, of “sequestering” interaction frequency and links from the original web. The transfer of interactions from native to alien generalists might create a positive feedback that fosters invasiveness and subjects native species to novel ecological and evolutionary dynamics [11]. We explored the effects of invasive species on the structure of pollination networks by compiling information provided by 10 paired, quantitative, plant–pollinator webs, eight from the temperate forests of the southern Andes and two from oceanic islands, which differ in the functional incidence of alien species. First, we evaluated the degree of mutual dependence between interacting partners (i.e., mutualism strength), characterizing networks with different levels of invasion. Second, we assessed whether a decrease in mutualism strength found in the most invaded networks was accompanied by a shift in the identity, from native to alien, of the generalist partners participating in the most asymmetric interactions and also by differences in the ecological role played by native and alien mutualists. Finally, we investigated whether an increase in alien dominance could result in a loss of interactions and decrease in connectivity between native partners. We report that although alien species could behave as mostly unnoticed commensals during initial stages of invasion, during later stages they monopolize interactions, including those that previously formed the core of native pollination webs. Results The Pollination Webs The 10 pollination webs included in this study varied in the total number of interacting species (i.e., their sizes) and number of alien species recorded (Table 1). The sizes of the study webs (21–69 species) and number of interactions (23–145 links) were smaller than those of many tropical pollination webs [24] or webs assembled from observations collected over large areas or over long time periods [18], but typical of local webs from other temperate regions and isolated islands [24–26]. In addition, these are the few webs studied, so far, that include both alien mutualists and some estimate of interaction frequency, a measure that relates strongly to plant reproductive success and is presumably associated with the amount of floral resources gathered by a given pollinator species [27]. Table 1 Characteristics of the Ten Plant–Pollinators Webs Analyzed The functional importance of invading plants and animals in each pollination web was estimated from the proportions of the sum of all visitation frequencies recorded at the flowers of alien plants and for alien flower visitors, respectively (Table 1). These estimates are determined by both the number of alien species present in each web and the total interaction frequency of each individual alien species. We should acknowledge, however, that the per visit efficiency in pollen delivery or in resource uptake could differ between native and alien species and might accentuate any expected effect of invasive species on either plant or animal fitness based on changes in interaction frequency alone [15]. For all five pairs of webs, the four pairs from the southern Andes and the pair from oceanic islands, the web with the highest incidence of visits to alien plants also had the highest incidence of visits from alien pollinators (Table 1). We considered the average between these two proportions as an index of the degree of invasion of a pollination web, which ranges from 0 for a web with no interacting alien species to 1 for a web characterized exclusively by interactions between alien species. Despite extensive variation, the four southern Andean webs from mostly undisturbed habitats and the web from Flores island represent the five lowest values of this index ( 0.32). Invasion and Mutualism Strength We calculated the mutual dependency for all pairs of interacting species in each web based on estimates of interaction frequency, and we consider the mean of all pairwise nonzero products of mutual dependencies as a measure of mutualism strength [1,28]. Each estimate of mutualism strength was compared with a distribution of expected values generated by a randomization procedure, and observed values were standardized by their respective expected means to lessen the influence of web size and total number of links (Table 2). Table 2 Correlations between Total Species Number, Number of Links, Invasion Index, and Mutualism Strength The studied pollination webs exhibited generally smaller mean products of mutual dependences between interacting species than expected by chance alone. All but two webs, both from the lightly-invaded group, had standardized mutualism strengths below the limit set by the 2.5 percentile of their respective randomly generated distributions (Figure 1A). The highly-invaded web of each of the five network pairs had consistently lower standardized mutualism strength than its lightly invaded counterpart (binomial test, p = 0.03125). More generally, mutualism strength—either standardized or not—varied inversely with the extent of invasion (Table 2 and Figure 1A). This declining trend can be attributed directly to the effects of interacting alien species, as the mutualism strength of the sub-web formed by the native species showed a weak positive association with invasion index (Figure 1B). As a consequence, the (standardized) mutualism strength of the native sub-web was relatively similar to the mutualism strength exhibited by the whole network for the lightly invaded webs (mean ± standard error [SE] = −0.31 ± 0.043 versus −0.25 ± 0.047; paired t-test, t = −1.40, degrees of freedom [df] = 4, p = 0.23), whereas it was much larger for the highly invaded group (mean ± SE = −0.17 ± 0.042 versus −0.43 ± 0.047; t = −5.05, df = 4, p 0.8) associated with interacting aliens was larger than those characterizing interactions between native species (60.6% versus 42.7%; χ2 = 13.53, df = 1, p 8), aliens exhibited a higher slope than natives (F = 3.45, df = 1, 67, p = 0.06, and F = 42.44, df = 1, 177, p 0. Based on estimates of interaction frequency, we calculated the mutual dependency for all pairs of interacting species in each web. The frequency of an interaction relative to its row total (i.e., the fraction of all animal visits to a plant species by a particular animal species) represents the dependence of plant species i on pollinator species j, whereas the frequency of a given interaction relative to its column total (i.e., the fraction of all visits by an animal species to a particular plant species) represents the dependence of pollinator species j on plant species i. We defined the strength of the mutualistic interaction between plant species i and pollinator species j as the product of their respective dependences [1] and the mutualism strength for an entire web as the mean of all pairwise nonzero mutualistic strengths. A web dominated by either mutually weak or highly asymmetric interactions will exhibit low mutualism strength [1,28]. The observed mutualism strength was compared with a distribution of expected values generated from 10,000 randomized datasets, where we shuffled the observed interaction frequencies within each matrix with the restriction that each species had at least one interaction [2,23] (Protocol S1). Because of the sensitivity of mutualism strength to the number of species and links, we standardized the observed mutualism strength (O) for each web as (O – E )/ E , where E is the expected mean mutualism strength of its corresponding simulated distribution. This relative measure of mutualism strength was not influenced significantly by network size (Table 2). For each web, we estimated both the mutualism strength of the whole network and that of the native sub-web (i.e., after excluding interactions with and between alien species). To identify the causes of changes in mutualism strength with increasing invasion, we compared the mean and distribution of the asymmetry between interacting pairs of native plant–pollinator species versus interacting pairs that included at least one alien species. Asymmetry between species i and j was characterized by AS(i, j) = max , which ranges between 0 and 1 [1]. Supporting Information Dataset S1 The Ten Pollination Networks (159 KB XLS) Click here for additional data file. Protocol S1 Matlab Code for Generating Randomized Distributions of Mutualism Strength (2 KB RTF) Click here for additional data file.
