Constitutive and induced chemical defense
To test the contribution of secondary metabolites to constitutive and induced defenses, we measured chemicals in leaves that were harvested from previously healthy and herbivore-damaged plants. Terpenoids and phenols are important secondary metabolites for defense against a wide range of herbivores (Mithöfer & Boland, 2012). Thus, we used the same non-native species and their native congeners used in the bioassay to measure total phenolics and total triterpenoids in leaves harvested from healthy and herbivore-damaged plants.
Leaves were flash frozen in liquid nitrogen and stored at -80°C for analysis. Total phenolic concentration was calculated as described in Supplementary Methods 2.
Growth rates
To evaluate the relationship between the strength of induced defenses and plant growth, we measured the relative growth rate of non-native and native plants. Two weeks after transplanting, we harvested ten similar-sized plants for each species and measured total dry biomass (weight1). After six more weeks, we harvested conspecifics and determined total dry biomass (weight2). We calculated relative growth rate (RGR) of each species as: RGR = (weight2 - weight1) / 42 days. There were 12 replicates for each combination of weight1 plants and weight2 plants for each non-native species and native congeners.
Statistical analysis
Herbivore pressure - To test for difference in herbivore pressure (binary data of undamaged vs. damaged leaves in a cbind matrix) among the non-native plant species in the field survey, we used a Wald Chi-square test applied on a Generalized Linear Mixed Model (GLMM) with a binomial distribution. Site was a random effect. We used a bootstrap method to test whether results differed when the number of native plant species was equal to the number of non-native species. We tested for difference in herbivore pressure (logit-transformed percentage of damaged leaf area) among the non-native plant species in the common garden experiment using a Wald Chi-square test applied on a Linear Mixed Model (LMM) with blocks as random effects. We conducted same analyses for native species. To test for differences in herbivore pressure (binary data of undamagedvs. damaged leaves in a cbind matrix) between the non-natives and natives in the field survey, we used a Wald Chi-square test applied on a GLMM with a binomial distribution. Site and species nested in origin as random effects. We also tested for differences in herbivore pressure (logit-transformed percentage of damaged leaf area) in the common garden experiment using LMM with blocks and species nested in origin as random effects. Furthermore, we used GLMM with binomial distribution that included random terms for sites and species to test whether percentage of damaged leaves (binary data of undamaged vs. damaged leaves in a cbind matrix) depended on herbivore biomass in the field survey and used LMM that included random terms for block and species to test whether percentage of damaged leaf area (logit-transformed) depended on herbivore biomass in the common garden experiment. Finally, for non-native species and their native congeners, in both the field survey and common garden experiment, we used LM to examine the relationship between leaf damage and herbivore biomass for the two experiments using mean values for species.
Constitutive and induced defenses - We assessed constitutive and induced defenses using larval weight gain and chemical contents for each species. For constitutive defense, we used larval weight gain on the leaves of plants that had not been previously attacked. For induced defense, we calculated the larval weight gain on the leaves of previously attacked plants minus the mean of larval weight gain on the leaves of un-attacked plants. Constitutive and induced defenses expressed by chemicals were evaluated using the same methods. We used the percentage of herbivore-damaged leaf area for each species in the common garden experiment as herbivore pressure. To evaluate relationships among constitutive defense, induced defense and herbivore pressure, we carried out Pearson correlations in which we multiplied larval weight with -1 since higher larval weight gain indicates lower defense. Mean values per species were used for above analyses and non-native and native species were analyzed separately. To test constitutive and induced defense between non-native and native species we used LMM with species as a random effect. Finally, to test whether changes in chemicals might underly changes in herbivore growth, we conducted Pearson correlations across both herbivory treatments and all species to examine the dependence of larval weight gain on phenolics or triterpenoids using mean values per species.
Strength of growth and induced defense - To test for differences in relative growth rate between non-native plant species and native congeners, we used a LMM with species nested within origin as random effects. Furthermore, we calculated difference in induced defense in term of larval weight gain and difference in plant growth rate between non-native species and corresponding native congeners. We then used Pearson correlation to evaluate relationship between difference in induced defense and difference in plant growth rate.
Homogeneity of variances and normality of distributions of data were checked before data analysis and P-values were corrected by False Discovery Rate (FDR) (Benjamini & Hochberg, 1995). All statistics were carried out using R (version 4.0.5) with the ‘car’, ‘lme4’, and ‘RVAideMemoire’ packages (Bates, 2014).