Feeding intensity of insect herbivores is associated more closely with key metabolite profiles than phylogenetic relatedness of their potential hosts

Determinants of the host ranges of insect herbivores are important from an evolutionary perspective and also have implications for applications such as biological control. Although insect herbivore host ranges typically are phylogenetically constrained, herbivore preference and performance ultimately are determined by plant traits, including plant secondary metabolites. Where such traits are phylogenetically labile, insect hervivore host ranges are expected to be phylogenetically disjunct, reflecting phenotypic similarities rather than genetic relatedness among potential hosts. We tested this hypothesis in the laboratory with a Brassicaceae-specialized weevil, Ceutorhynchus cardariae Korotyaev (Coleoptera: Curculionidae), on 13 test plant species differing in their suitability as hosts for the weevil. We compared the associations between feeding by C. cardariae and either phenotypic similarity (secondary chemistry—glucosinolate profile) or genetic similarity (sequence of the chloroplast gene ndhF) using two methods—simple correlations or strengths of association between feeding by each species, and dendrograms based on either glucosinolates or ndhF sequence (i.e., a phylogram). For comparison, we performed a similar test with the oligophagous Plutella xylostella (L.) (Lepidoptera: Plutellidae) using the same plant species. We found using either method that phenotypic similarity was more strongly associated with feeding intensity by C. cardariae than genetic similarity. In contrast, neither genetic nor phenotypic similarity was significantly associated with feeding intensity on the test species by P. xylostella. The result indicates that phenotypic traits can be more reliable indicators of the feeding preference of a specialist than phylogenetic relatedness of its potential hosts. This has implications for the evolution and maintenance of host ranges and host specialization in phytophagous insects. It also has implications for identifying plant species at risk of nontarget attack by potential weed biological control agents and hence the approach to prerelease testing.