Meta-analysis of induced anti-herbivore defence traits in plants from 647 manipulative experiments with natural and simulated herbivory

Induced anti-herbivore defences in plants can shape ecosystem structure and functions. Since these are known to be highly variable, quantifying their sources of variation remains important to understand their eco-evolutionary roles. We conducted meta-analysis of 647 experiments (paired control and treatment) from 192 studies to address sources of variation in induced anti-herbivore defensive traits. This covered different agents of herbivory (insect, mammal and clipping), studies in greenhouse and in field settings, and covered 24 types of defence traits in 163 species from 50 families across 24 angiosperm orders. We used meta-regression models-multi-level random-effects (MLREs) and robust variance estimators (RVEs)-to quantify variation due to study setting, herbivore type, chemical identity, angiosperm order, volatility, fertilization, and publication year as moderator variables. There were strong indications for publication bias in favour of large effect sizes, and studies with adequate sample size and precision were rare. Strength of induction was distributed uniformly over the phylogeny of 24 angiosperm orders. Stronger effects were reported from greenhouses than in field conditions, and induction was uninfluenced by herbivore type. Volatiles showed stronger response than non-volatiles in greenhouses for insects. Biosynthetic precursors (e.g. jasmonates) and many defensive chemicals were induced in the greenhouse, but when measured in field conditions, they were unresponsive to herbivory and appear to be constitutive. Similar to constitutive response, induced responses were also unaffected by fertilization. Direction and magnitude of induction differ greatly from previous assessments as these have also changed over time, representing natural epistemological growth. A third of the variability was explained by moderators (marginal R-2: chemical identity, angiosperm order, study setting, herbivore type and publication year); another third (conditional R-2) was attributed to the identity of individual studies and observations. Synthesis. Manipulative experiments reveal many intrinsic differences among defensive chemicals and traits. They suggest plants in greenhouses may respond very differently to herbivores from those in field conditions. Although biosynthetic pathways of chemical responses, their modes of action and their effects on herbivores are well-understood, studies with greater statistical power under ecologically relevant settings are needed to discern induced and constitutive defences in how plants respond to their natural enemies