3.2 Plant diversity or soil legacy effects on herbivore-induced responses
Both in the PPI (Fig 3a-c ) and the PSI (Fig 3d-f ) experiment, we discovered significant differences in the foliar metabolome composition across all plant diversity levels and soil legacies between control and herbivore-induced plants in all plant species. When we tested for the regulation of metabolites between control and induced plants, we found that the total number of up-regulated metabolites was higher than the total number of down-regulated metabolites across all species (Appendix Fig A2 ). Furthermore, we observed that the absolute number of regulated metabolites was highest when plants had grown in different soil legacies in the PSI experiment. This effect was strongest for L. vulgare , while R. acris showed the overall strongest response in numbers of regulated metabolites in both the PPI and PSI experiment (Appendix Fig A2 ).
In contrast, we found no significant effect of plant diversity in the PPI experiment and of soil legacy in the PSI experiment on the induced metabolome in either species (Table 1 ). However, when we compared foliar metabolomes of herbivore-induced plants grown in monocultures with conspecifics growing in mixtures, we discovered a total of 141 significantly up- or down-regulated metabolites (Fig. 4 ). Both heterospecific plant-plant and plant-soil interactions affected the induction of metabolites compared to conspecific plant-plant or plant-soil interactions. Overall, heterospecific plant-plant interactions regulated more induced metabolites than plant-soil interactions in leaves ofL. vulgare (26 vs. 14) and R. acris (40 vs.24). In comparison, heterospecific plant-soil interactions had a stronger effect on the regulation of herbivore-induced metabolites in leaves of G. pratense than heterospecific plant-plant interactions (21 vs. 16; Fig. 4 ). In R. acris , we discovered that heterospecific plant-plant and plant-soil interactions had contrasting effects on the regulation of induced metabolites. Heterospecific plant-plant interactions strongly down-regulated the induction of metabolites, while plant-soil interactions strongly up-regulated the induction of metabolites (Fig. 4 ). In contrast, these modulating effects of heterospecific plant-plant or plant-soil interactions on the induction of metabolites were mostly similar or less pronounced in herbivore-induced plants of G. pratense or L. vulgare (Fig. 4 ). Across all species and both experiments, we found no de-novo regulated metabolites in herbivore-induced plants (Appendix Fig. A3 ); all up- and down-regulated metabolites were present in control plants as well. Similar to the analysis of regulated metabolites in leaves and roots, the tentatively assigned metabolites in herbivore-induced plants mostly belonged to the family of phenolics, in particular flavonoids, their precursors, and derivatives. Besides, we tentatively assigned two metabolites in L. vulgare as an iridoid and an alkaloid glycoside (Table 2 ).