Evaluation of bias on the assessment of diet breadth of herbivorous insects using molecular methods

The interactions between herbivores and their host plants play a key role in ecological processes. Understanding the width and nature of these interactions is fundamental to ecology and conservation. Recent research on DNA‐based inference of trophic associations suggests that the host range of phytophagous insects in the tropics may be wider than previously thought based on traditional observation. However, the reliability of molecular inference of ecological associations, still strongly dependent on PCR and thus exposed to the risk of contamination with environmental DNA, is under debate. Here, we explored alternative procedures to reduce the chance of amplification of external, nondiet DNA, including surface decontamination and analysis of mid/hind guts, comparing the results with those obtained using the standard protocol. We studied 261 specimens in eight species of Neotropical Chrysomelidae that yielded 316 psbA‐trnH intergenic spacer sequences (cpDNA marker of putative diets) from unique and multiple‐band PCR results. The taxonomic identity of these sequences was inferred using the automated pipeline BAGpipe, yielding results consistent with 31 plant families. Regardless of the protocol used, a wide taxonomic spectrum of food was inferred for all chrysomelid species. Canonical Correspondence Analysis using these data revealed significant differences attributed mainly to species (expectedly, since they represent different ecologies), but also to treatment (untreated vs. cleaned/gut samples) and PCR results (single vs. multiple bands). Molecular identification of diets is not straightforward and, regardless of the species’ niche breadth, combining approaches that reduce external contamination and studying multiple individuals per species may help increasing confidence in results.