The Interaction between Root Herbivory and Competitive Ability of Native and Invasive-Range Populations of Brassica nigra

The evolution of increased competitive ability (EICA) hypothesis predicts that escape from intense herbivore damage may enable invasive plants to evolve higher competitive ability in the invasive range. Below-ground root herbivory can have a strong impact on plant performance, and invasive plants often compete with multiple species simultaneously, but experimental approaches in which EICA predictions are tested with root herbivores and in a community setting are rare. Here, we used Brassica nigra plants from eight invasive- and seven native-range populations to test whether the invasive-range plants have evolved increased competitive ability when competing with Achillea millefolium and with a community (both with and without A. millefolium). Further, we tested whether competitive interactions depend on root herbivory on B. nigra by the specialist Delia radicum. Without the community, competition with A. millefolium reduced biomass of invasive- but not of native-range B. nigra. With the community, invasive-range B. nigra suffered less than native-range B. nigra. Although the overall effect of root herbivory was not significant, it reduced the negative effect of the presence of the community. The community produced significantly less biomass when competing with B. nigra, irrespective of the range of origin, and independent of the presence of A. millefolium. Taken together, these results offer no clear support for the EICA hypothesis. While native-range B. nigra plants appear to be better in dealing with a single competitor, the invasive-range plants appear to be better in dealing with a more realistic multi-species community. Possibly, this ability of tolerating multiple competitors simultaneously has contributed to the invasion success of B. nigra in North America.


Introduction
The evolution of increased competitive ability (EICA) hypothesis predicts that escape from intense herbivory and subsequent genetically based re-allocation of resources to increased growth and reproductive output may help invasive plant species to colonize novel habitats in the invasive range [1]. The EICA hypothesis is derived from the optimal defense hypothesis assuming that defenses are costly because they divert resources from growth and reproduction [1,2]. Accordingly, the EICA hypothesis predicts that plants from the invasive range that evolved lower anti-herbivore defenses should evolve increased competitive ability. Hence, when grown under common conditions in the absence of herbivores, invasive-range plants should exhibit significantly higher growth and reproductive output than conspecific plants from the native range [1]. In the presence of herbivores, however, the competitive advantage of plants from the invasive range should be reduced or reversed [1,2]. In most of the experiments testing predictions of the EICA hypothesis, invasive-range and conspecific native-range plants were grown under common conditions in the absence of competitors [3]. In the few studies where competition was explicitly included, only intraspecific competition [4] or pairwise interspecific competitive interactions between plants from invasive and native ranges versus their neighbors were investigated [3,[5][6][7]. However, invasive plants typically compete in multi-species plant communities that may vary with regard to the presence of strong competitor species [8][9][10][11][12]. Therefore, tests for evolution of increased competitive ability in invasive plants require manipulative experiments that compare the performance of invasive-range and conspecific native-range plants under competition in a community setting.
Because plant roots perform such vital functions as uptake of water and nutrients necessary for plant growth [13], herbivore damage on plant roots can cause significant reductions in individual plant growth and fitness, and can influence overall plant community structure [14,15]. Hence, escape from root herbivory may play an important role in explaining increased competitive ability in invasive-range plants. However, of the few studies that have tested the effect of herbivory on competitive ability of invasive-range and conspecific native-range plants [3,5,16], only one tested for the effects of root herbivory. The study found that invasive-range Chromolaena odorata plants outcompeted native-range C. odorata plants, both in the presence and absence of root herbivory [16]. However, as this study tested the effect of root herbivory only in the presence of intraspecific competition, the effect of root herbivory on competitive ability of invasive-range and conspecific native-range plants in a community setting remains unknown.
Within natural plant communities, competitive interactions between plants have two components: competitive effect (i.e., the degree to which a focal plant suppresses the growth of its neighbors) and competitive response (i.e., the degree to which a focal plant can tolerate the impact of its neighbors) [17][18][19][20]. In plant invasions, these two components of competition may characterize different invasion stages; competitive response would be an important determinant of establishment success of the invader within a recipient community, whereas the competitive effect would be important for impact of the invader to the community [6]. However, the few experiments that have investigated both competitive effects and responses of invasive-range and conspecific native-range plants have focused only on pairwise competitive interactions [6,7,21,22], and found mixed results.
In this study, we tested for the separate and joint effects of specialist root herbivory on B. nigra and the presence of a community of competitors on the growth of B. nigra plants from the invasive and native ranges, and the competitive effect of B. nigra on the community. We specifically addressed the following questions: i) In the absence of root herbivory, do B. nigra plants from the invasive range have a higher competitive ability (i.e., a weaker biomass reduction due to growth in a community of competitors and a stronger negative effect on the community) than B. nigra plants from the native range?; ii) Do B. nigra plants from the invasive range suffer more from root herbivory than B. nigra plants from the native range?; and iii) Does root herbivory change the competitive responses and effects of B. nigra plants from the invasive and native ranges differently?

Study species and seed sources
Brassica nigra (Brassicaceae) (L.) W. D. J. Koch is a self-incompatible annual herb native to Europe, Asia and North Africa, and was introduced to North America approximately 200 years ago [23][24][25]. Seeds of B. nigra have long been used in southern Europe, Asia and North Africa for cooking oil, condiment mustard and medicine [25], and most likely for this reason the species has been introduced to other continents. Presently, B. nigra is invasive in certain regions of North America, where it can form thick monospecific stands [26]. Recent studies have found significantly higher resistance to generalist herbivory (in line with the shifting defence hypothesis) [27,28] and higher reproductive output in the invasive-range populations of B. nigra relative to their native-range conspecifics, suggesting rapid post-introduction evolution of the invasive-range populations of B. nigra [5,29].
As competitors of B. nigra, we selected five plant species (two grasses and three forbs). To ensure that native-and invasive-range B. nigra plants had a comparable history of co-evolution with the competitor species, we selected competitor species that co-occur with B. nigra in large parts of both its native and invasive range (The Jepson Interchange, information: http://ucjeps. berkeley.edu/interchange/). The grasses were Elymus glaucus and Nasella pulchra, and the forbs were Medicago lupulina, Sonchus oleraceus and Achillea millefolium. While we did not have a priori information on the competitive effects of the first four species, A. millefolium is known to exert strong competitive effects on other plant species that are invasive in the same range as B. nigra [30][31][32]. Bulked samples of seeds collected from several maternal plants in eight invasive-range and seven native-range populations of B. nigra as well as the five competitor species were obtained directly from the field or from seed germplasm collections (see S1 and S2 Tables).

Ethics statement
Permission to import seeds from North America was granted by the German ministry of agriculture and food (permit number: SAG-2013-018-AG-7, dated 1 st July 2013). The imported seeds had been collected from field sites where the plants grow naturally, and no specific permission was required to access those sites as they were unprotected public land. Furthermore, none of the species used in the present study is endangered or under official protection. As the present study was conducted in the greenhouse, no specific permission was required for that purpose.

Pre-cultivation and experimental design
To test whether B. nigra plants from the invasive and native ranges differed in their responses to growth in a community setting, and whether these responses depended on damage by a specialist root herbivore of B. nigra, we performed a greenhouse experiment in the botanical garden of the University of Konstanz (Germany) between September 2013 and January 2014. In September 2013, seeds of B. nigra and the five competitor species were sown individually in plastic plug-trays filled with a commercial potting soil (Standard soil, Gebr. Patzer GmbH & Co. KG, Sinntal, Germany; organic matter content: 40-50%, pH (H 2 O): 4.5-7.0, electrical conductivity: 200-900μS/cm). The trays were kept in a phytochamber (12h day/night cycle at 21°C/17°C and 90% relative humidity). In a previous germination trial, B. nigra seeds had germinated five days earlier than the five competitor species, and hence for the experiment, the competitors were sown a week earlier (on 23 rd September 2013) than B. nigra seeds. After three weeks, we transplanted the emerged seedlings to 2.5-L round plastic pots filled with sand and vermiculite mixed in a ratio of 1:1. In each pot, we applied 10 g of a slow-release fertilizer (Osmocote Classic 14% N, 14% P 2 O 5 , 14% K 2 O; Scotts, Geldermalsen, The Netherlands).
Employing a full factorial design, we grew individual B. nigra plants from invasive-and native-range populations in the presence or absence of a community of the four plant species without a priori information on the strength of their competitive effects (E. glaucus, N. pulchra, M. lupulina and S. oleraceus). This factor was crossed with the presence or absence of a species that a priori was known to exert strong competitive effects (A. millefolium) [30][31][32]. In treatments where B. nigra was grown with four or five competitors (five or six plants per pot, respectively), an individual B. nigra plant always occupied a central position in the pot while the other competitor species were distributed randomly to one of four or five positions at equal distances around the B. nigra plant. In the treatment where an individual B. nigra plant was grown in pairwise competition with an individual A. millefolium (two plants per pot), the plants were planted on opposite ends of a pot. In the no competition treatment, a single B. nigra plant was grown in the center of the pot. Although an effect of adding A. millefolium might mainly be due to a further increase in density of the competitors, this experimental design nevertheless allows us to test whether the effect of the increase in density differs for native-and invasive-range B. nigra plants. The factorial combination of the four competition treatments (no competition, competition with the single strong competitor A. millefolium, competition with the community in the absence of a strong competitor A. millefolium and competition with the community in the presence of a strong competitor A. millefolium) were further crossed with a root herbivory treatment (presence vs. absence of root herbivory on B. nigra plants) . We imposed the root herbivory treatments on B. nigra plants three and a half weeks after transplanting. As a herbivore, we took Delia radicum (L.) (Diptera: Anthomyiidae), a specialist herbivore exclusively feeding on Brassicaceae and native to Europe [33]. Although D. radicum has been introduced to the north-eastern coast of North America (Newfoundland, Canada) in the 19 th century, where it has been reported as a serious pest of Brassicaceous crops [34,35], no report exists of the insect attacking wild populations of invasive-range B. nigra. The invasiverange populations of B. nigra that we used in the present study were, therefore, considered to have escaped damage by D. radicum. One half of all the B. nigra plants in the experiment were infested with D. radicum by placing five eggs around the root collar of the B. nigra plant using a fine-tipped paint brush. The root collar is also the typical location where females of D. radicum lay their eggs [33]. The eggs were slightly covered with moist sand to prevent desiccation. Emerged larvae burrow into the soil to feed on roots until they pupate [33]. To confirm that the larvae indeed fed on the roots in our experiment, we grew an additional set of 12 B. nigra As the number of populations constitutes the effective number of replicates for testing differences between the native and invasive ranges, we maximized the number of populations (15) over the number of replicates (4 per treatment) per population [36].Thus, for each of the 15 B. nigra populations (eight from the invasive range and seven from the native range), we had eight treatment combinations (2 x 2 x 2: presence/absence of community x presence/ absence of A. millefolium x presence/absence of root herbivory on B. nigra), which resulted in 480 (i.e., 4 x 8 x 15) experimental pots. In addition, we grew eight replicates of: i) the community (E. glaucus, N. pulchra, M. lupulina, and S. oleraceus) without competition from B. nigra and A. millefolium, and ii) the community in competition with A. millefolium. These additional treatments served as controls for testing the suppressive effects of B. nigra and A. millefolium on the community. This resulted in an additional 16 pots. The 496 experimental pots were assigned to four separate blocks in two greenhouse compartments (two blocks per compartment) employing a complete randomized block design whereby each treatment and population combination appeared in every block. The pots were spaced 0.35 m apart. The greenhouse conditions were maintained at a temperature regime of 24 ± 5°C, a light cycle of 16 h: 8 h (Day/ Night), and 50-70% relative humidity. The plants were watered once a day by filling plastic plates placed beneath each pot with tap water.

Measurement of biomass yield
We used aboveground biomass as proxy for fitness since aboveground biomass and seed yield of annual plant species are often positively correlated [37]. Previous experiments had confirmed this for B. nigra [5,29]. We did not harvest below-ground biomass, because it was impossible to separate roots of different species in the competition treatments. All experimental plants were harvested after three months of growth. This was done by cutting the individual plants at the root collar and then placing all the above-ground material belonging to an individual plant in separate paper bags. The individual plants were then dried at 70°C for 72 hours, and then weighed.

Statistical analysis
To test whether B. nigra plants from the invasive and native ranges differed in their biomass responses to the different competition treatments, root herbivory and their interactions, we used linear mixed-effects models fitted with the lme function in the R package nlme [38]. The fixed part of the model included four effects, each with two levels: Achillea (presence vs. absence of A. millefolium), community (presence vs. absence), herbivory treatment (presence vs. absence of root herbivory on B. nigra plants), B. nigra range (invasive vs. native) and all possible interactions, reflecting the full factorial design of our experiment. Maternal resource provisioning to developing seeds can influence early acting plant traits such as seedling growth [39]. To avoid potential bias due to such maternal effects, we used initial height and total leaf count of fourweek old individual plants as co-variates in the models. The random part of the model included population (nested in B. nigra range) and block. The model accounted for heteroscedasticity among populations using the varIdent function available within the lme function [40].
A principal components analysis indicated that changes in community biomass were driven by unidirectional responses of all the four community members (i.e., the first principal component that explained 41% variance had positive loadings for each of the individual community members, S3 Table). Therefore, our analyses of the competitive effect of B. nigra on the community focused on the total community biomass (rather than separately analyzing the biomass of individual species). Because we obviously could not grow the community with the B. nigra specialist herbivore when B. nigra was absent, we did not have a full factorial design for community biomass. Therefore, we did separate analyses to test for effects of B. nigra and herbivory on community biomass. First, to test the general effects of presence/absence of B. nigra and of A. millefolium on community productivity, we selected the subset of cases without herbivory (n = 136), and then constructed linear mixed-effects models. The model fixed parts included Brassica (presence vs. absence of B. nigra), B. nigra range (invasive vs. native; fitted sequentially after B. nigra presence), Achillea (presence vs. absence of A. millefolium) and all possible interactions; the random part included block. Second, to test whether root herbivory on B. nigra mediated the competitive effect of B. nigra on the community productivity, we selected the subset of cases in which the community was grown with B. nigra plants (n = 240), and constructed linear mixed-effects models. The model fixed part included B. nigra range (invasive vs. native), root herbivory on B. nigra (herbivory vs. no herbivory), Achillea (presence vs. absence of A. millefolium) and all possible two and three-way interactions; the random part included block and B. nigra population (nested in range).
For all models, we tested the significance of the interactions and main effects by removing first, the highest order interactions, and then the lower order interactions and finally the main effects, and performing model comparisons using likelihood-ratio tests (see notes below Tables 1-3 for the exact comparisons). All analyses were performed in R v3.0.3 [41].

Brassica nigra biomass
On average, competition with the community significantly reduced aboveground biomass of B. nigra by 33.0% ( Table 1). The presence of A. millefolium had only a small negative effect on biomass of B. nigra, and only so in the absence of the community (-5.99% in the absence of a community vs + 0.73% in presence of the community; significant Achillea x Community (A x C) interaction: Table 1; Fig 1A). In the absence of the community, biomass of invasive-range B. nigra plants was reduced by the presence of A. millefolium (-13.9%; Fig 1C), while this was not the case for native-range B. nigra plants (+ 0.7%; Fig 1C). On the other hand, invasive-range B. nigra plants suffered less from the presence of the community (-24.6%) than native-range B. nigra plants (-40.6%), regardless of the presence/absence of A. millefolium. This was reflected by a significant Range x Achillea x Community (R x A x C) three-way interaction (Table 1; Fig 1C).
Although the overall effect of herbivory on biomass of B. nigra was not significant, it reduced the negative effect of the presence of the community (-42.7% vs -22.7%; significant Community x Herbivory (C x H) interaction: Table 1; Fig 1B). Four-way and all the other three-way and two-way interactions were not significant (

Community biomass
The community produced significantly less aboveground biomass when grown in the presence of B. nigra than in the absence of B. nigra (Table 2; Fig 2A). Range of B. nigra, root herbivory on B. nigra, presence of A. millefolium or any interaction between these factors did not significantly affect community biomass (Table 3; Fig 2B).

Discussion
The evolution of increased competitive ability (EICA) hypothesis predicts that in the absence of herbivore damage, plants from the invasive range should exhibit higher competitive ability than conspecific plants from the native range [1]. Despite intensive research on the EICA hypothesis [42][43][44], the effect of herbivory on post-introduction evolution of competitive ability in invasive plants remains little studied to date [3]. In this paper, we show that native-range B. nigra plants competed better than invasive-range B. nigra plants in the presence of a single strong competitor A. millefolium, while invasive-range B. nigra plants were better competitors than native-range B. nigra plants in a more realistic multi-species community setting. Irrespective of the presence of competitors, invasive-range B. nigra plants tended to suffer from root herbivory, whereas native-range B. nigra plants benefitted from it, although there was only a marginally significant interaction between B. nigra range and herbivory. Overall, these results offer no clear support for the EICA hypothesis. While native-range B. nigra plants appear to be better in dealing with a single competitor, the invasive-range plants appear to be better in dealing with a more realistic multi-species community. Possibly, this ability of tolerating multiple competitors simultaneously has contributed to the invasion success of B. nigra in North America. Competitive ability of B. nigra plants from the invasive and native ranges The significant three-way interaction between range of B. nigra, presence of A. millefolium and presence of community (R x A x C in Table 1) partly supports and partly contradicts the EICA hypothesis. On the one hand, regardless of whether A. millefolium was part of the community, invasive-range B. nigra plants suffered less from the presence of the community than did native-range B. nigra plants (Fig 1C), which supports a prediction of the EICA hypothesis. On the other hand, native-range B. nigra plants suffered less when grown in pairwise competition with A. millefolium in the absence of community than did the invasive-range B. nigra plants (Fig 1C), which is contrary to a prediction of the EICA hypothesis. The finding that invasiverange and native-range B. nigra plants had similar competitive effects on the community regardless of root herbivory treatment on B. nigra is also contrary to a prediction of the EICA hypothesis ( Fig 2B). Most support for the EICA hypothesis stems from studies that have confirmed the prediction of significantly higher growth in invasive-relative to native-range conspecifics, for example in Barbarea vulgaris, Cardaria draba, Rorippa austriaca and Jacobaea vulgaris [45,46]. However, this patterns is not universal, as similar vegetative growth and reproduction was reported for invasive-and native-range plants of Mimulus guttatus [47] and Lythrum salicaria [48]. Only few other studies have examined the prediction of increased competitive ability by manipulating competitive environments, and these also produced mixed results. Similar competitive responses were reported for invasive-and native-range plants of Eschscholzia californica [49], Silene latifolia [50], and Lepidium draba [51]. In contrast, higher competitive response (i.e., higher biomass production) was reported for invasive-range Sapium sebiferum plants than their native-range conspecifics [52]. On the other hand, invasive-range Alliaria petiolata had a significantly lower competitive response (i.e., lower biomass production) than their native-range conspecifics [4]. However, because all these previous studies reported only the competitive responses of conspecific plants from the native and invasiveranges, and no results on the competitive effects, they provide incomplete picture regarding the possible post-introduction evolution of overall competitive ability in invasive plant species.
Our study adds to the few other experiments that have investigated both competitive effects and responses of invasive-range and native-range plants on other plant species [6,7,21,22]. Invasive-range Lythrum salicaria plants exhibited significantly stronger competitive effects and responses than L. salicaria plants from the native range when both groups of plants were grown in pairwise intraspecific and interspecific competition [6]. Invasive-range plants of Centaurea maculosa exhibited stronger competitive effects and responses than native-range C. maculosa plants when grown in pairwise competitive interactions with Pseudoroegneria spicata or Festuca idahoensis [21]. In another study, invasive-and native-range plants of C. maculosa had similar competitive effects and responses [22]. In contrast, grown in pairwise interspecific competition with Urtica dioica, plants of Impatiens glandulifera from the invasive range exhibited weaker competitive effects and were more suppressed by their neighbors relative to nativerange I. glandulifera plants [7]. Clearly, there is no universal pattern emerging from the studies focusing only on pairwise competitive interactions, and we, therefore, assessed competitive effects and responses in a community setting.
Our experimental design provided some indication that, although native-range B. nigra performed better in a pairwise interaction with A. millefolium, invasive-range B. nigra plants may be better competitors than native-range B. nigra plants in a multi-species community setting. This suggests that invasive-range B. nigra plants evolved mechanisms that allow them to tolerate strong competition. Numerous comparative studies have shown that invasive plant species exhibit higher mean values of traits that contribute to competitive ability than native and non-invasive plant species, including a higher capacity to acquire and retain growth resources and/or to exploit resources better (e.g., through early growth and plastic morphological responses such as root-foraging responses) than co-occurring native species [53,54]. It remains to be tested whether such traits explain why invasive-range B. nigra plants appear to be better in dealing with a more realistic multi-species community.
Effects of root herbivory on competitive ability of B. nigra One of our main findings is that root herbivory by a Brassicaceae specialist reduced the negative effects of competition on B. nigra grown in a community setting (Fig 1B). This seems counterintuitive as the specialist root herbivore only targets Brassicaceae, and not the community. A possible explanation is that attack by the root herbivore induced compensatory growth in B. nigra. Indeed, plants can exhibit increased competitive ability when damaged by herbivores through compensatory growth [21,[55][56][57]. Our finding that the beneficial effect of root herbivory tended to be stronger (although only marginally so) in B. nigra from the native range (Table 1 & S2C Fig) suggests that the native-range B. nigra plants have higher compensatory growth (i.e., higher tolerance of herbivory) than invasive-range B. nigra. This confirms previous field experiments with B. nigra that found that the native-range populations expressed significantly higher levels of compensatory growth than the invasive-range populations following damage by a community of above-ground feeding herbivores [5,29].
This pattern may be the result of differential herbivore selection pressures in the respective ranges, and thus have resulted from post-introduction selection imposed by herbivores. Invasive B. nigra, like many other invasive species [28,58], has not escaped herbivory, but has experienced a change in the level of herbivory and the community of herbivores it interacts with. Specifically, invasive-range B. nigra interacts more with generalist and less with specialist herbivore species than native-range counterparts [5]. Phylogeographic evidence indicates that invasive-range populations of B. nigra were introduced from multiple sources in the native range [59], which not only may have facilitated the invasion due to admixture boosting fitness [60], but also have provided standing genetic variation that natural selection can act upon [61,62]. This suggests that invasive-range populations of B. nigra should have sufficient standing genetic variation for selection to operate on.
Shifts in interactions with herbivores may thus have selected for the lower tolerance in response to root herbivory that we observed in invasive-range B. nigra plants compared to native-range B. nigra plants (marginally significant R x H interaction, Table 1; S2C Fig), which weakly supports the shifting defense hypothesis. The shifting defence hypothesis predicts that biogeographical differences in the levels of herbivory and composition of the herbivore community may select for high resistance (expression of high concentrations of less costly qualitative defence compounds like glucosinolates that are most effective against generalist herbivores) and low tolerance (compensatory growth that is most beneficial to plants that are attacked by a high diversity and density of both generalist and specialist herbivores) [27,63]. In a study similar to ours, genotypes of the North American invader C. maculosa that expressed higher compensatory growth in response to herbivore damage (i.e., higher tolerance) also demonstrated stronger competitive responses than genotypes of C. maculosa with lower compensatory growth [21]. However, more studies with similar setups are needed to more conclusively test the shifting defense hypothesis, and the prediction that selection for traits associated with increased tolerance (increased rates of photosynthesis and resource acquisition, and ultimately growth) may indirectly select for increased competitive ability [21,[55][56][57].
Supporting Information S1 Fig. Roots of individual Brassica nigra plants damaged (a) or undamaged (b) by larvae of a specialist root herbivore (Delia radicum). Note the black lesions caused by larval feeding on the damaged root. (TIF) S2 Fig. Mean (± 1SE) Brassica nigra biomass for a) R x A: the 2-way interaction between B.nigra range (R: Inv = invasive; Nat = native) and Achillea millefolium (A: absence = Ach-; presence = Ach+); b) R x C: the 2-way interaction between R and Community (C: absence = Comm-; presence = Comm+ of Elymus glaucus, Nasella pulchra, Medicago lupulina and Sonchus oleraceus); c) R x H: the 2-way interaction between R and herbivory on B. nigra (H: absence = Herb-; presence = Herb+); d) A x H: the 2-way interaction between A and H. The means and standard error (SE) were calculated as follows: 1) for each combination of factor levels, we calculated the mean and standard deviation of population means; 2) for each interaction plot, we calculated the mean of the factor level means that were not involved in the plotted interaction, and standard errors based on the mean standard deviations and the sample size (number of populations) of the smallest group (n = 7). (TIF) S1 File. Data used to test whether B. nigra plants from the invasive and native ranges differed in their biomass responses to the different competition treatments, root herbivory on B. nigra and their interactions.