Multivariate statistics coupled to generalized linear models reveal complex use of chemical cues by a parasitoid

Abstract

Understanding how animals integrate multiple cues, in particular complex mixtures of volatile chemicals, is a subject of current interest. Insect behavioural responses to volatile blends have traditionally been analysed separately to the changes in blends themselves, making it difficult to link behaviour directly to volatile cues. We coupled principal components analysis (PCA) and generalized linear models (GLMs) to link volatiles released by plants in response to pea aphid, Acyrthosiphon pisum, feeding directly to the behaviour of the aphid parasitoid Aphidius ervi. We used choice bioassays to compare parasitoid response to odours from aphid-damaged and undamaged plants, then collected volatiles from the same plants used in bioassays and analysed volatile profiles with PCA. The principal components explaining 99% of the variation in the data set were used as explanatory variables in a GLM to analyse the behavioural response. This was done for two plant species: alfalfa, Medicago sativa, and broad bean, Vicia faba. For each species, a single principal component was important in explaining insect attraction. In both cases it explained a small amount of variability in the volatile data set (7.0% and 0.5% in M. sativa and V. faba, respectively). For both plants, the analysis revealed the presence of volatiles that supported or inhibited parasitoid attraction. Compositional analysis of the blend revealed no major changes in either plant, highlighting that A. ervi can detect minor changes, ignoring the major variability in the blend. The approach could be valuable for behavioural studies on multisensory orientation by foraging animals.