Phytochemical drivers of insect herbivory: a functional toolbox to support agroecological diversification

Abstract

Plant metabolism is a key feature of biodiversity that remains underexploited in functional frameworks used in agroecology. Here, we study how phytochemical diversity considered at three organizational levels can promote pest control. In a factorial field experiment, we manipulated plant diversity in three monocultures and three mixed crops of oilseed rape to explore how intra- and interspecific phytochemical diversity affects pest infestation. We combined recent progress in metabolomics with classic metrics used in ecology to test a box of hypotheses grounded in plant defence theory. According to the hypothesis of 'phytochemically mediated coevolution', our study stresses the relationships between herbivore infestation and particular classes of specialized metabolites like glucosinolates. Among 178 significant relationships between metabolites and herbivory rates, only 20% were negative. At the plant level, phytochemical abundance and richness had poor predictive power on pest regulation. This challenges the hypothesis of 'synergistic effects'. At the crop cover level, in line with the hypothesis of 'associational resistance', the phytochemical dissimilarity between neighbouring plants limited pest infestation. We discuss the intricate links between associational resistance and bottom-up pest control. Bridging different levels of organization in agroecosystems helps to dissect the multi-scale relationships between phytochemistry and insect herbivory.

Keywords: Brassica napus; associational resistance; bottom-up; mixed crop; plant defence; plant metabolism.

Figure 1.

Field experimental setup. The field was surrounded by 5 ha of oilseed rape. In mixed crops, the proportions of trap plants are 14%, 7% and 7% of radish, ‘Escape’ and ‘Lidea’, respectively.

Figure 2.

Phytochemical diversity detected in oilseed rape. Metabolomic network depicting the 259 features with chemical assignation among the 1074 features detected. Grey segments between features indicate structural similarities (cosine score >0.7). Comparisons of MS/MS fragments with GNPS libraries enabled us to assign 104 features to specific compounds (six-branched stars) and 155 features to chemical classes when they were analogous to specific compounds (four-branched stars) or when they belong to metabolomic clusters (circles).

Figure 3.

Relationships between pest infestation and natural compounds. Black lines indicate significant mixed linear models (‘block’ as random factor; p ≤ 0.05). Grey areas depict prediction intervals (80%) based on simulated models (n = 1000). Correlations are given as Pearson coefficients.

Figure 4.

Variations of plant metabolome across cultivars and species. Hierarchical clustering indicates that sampled plants (columns) aggregate differentially according to the production of 47 features (rows) retrieved from random forest analyses.

Figure 5.

Relationships between pest infestation and phytochemical makeups of individual plants. Phytochemical abundance is the quantitative sum of all features detected in a single plant. Phytochemical richness is the number of different features detected in a single plant. Black lines indicate significant mixed linear models (‘block’ as random factor; p < 0.05). Grey areas depict prediction intervals (80%) based on simulated models (n = 1000).

Figure 6.

Relationships between pest occurrence and phytochemistry at the plant community level. Each grey dot depicts a bootstrapped plant assemblage (n = 1000). Phytochemical abundance is the quantitative sum of all features detected in each plant community. Phytochemical diversity is estimated with β-diversity indicators (average Jaccard dissimilarity index). Black lines and grey areas indicate coefficients of significant linear models (p ≤ 0.05) and confidence intervals (80%), respectively.