Validating a multi-locus metabarcoding approach for characterizing mixed-pollen samples

Abstract

Background The mutualistic interaction between entomophilous plants and pollinators is fundamental to the struc- ture of most terrestrial ecosystems. The sensitive nature of this relationship has been disrupted by anthropogenic modifications to natural landscapes, warranting development of new methods for exploring this trophic interaction. Characterizing the composition of pollen collected by pollinators, e.g. Apis mellifera, is a common means of explor- ing this relationship, but traditional methods of microscopic pollen assessment are laborious and limited in their scope. The development of pollen metabarcoding as a method of rapidly characterizing the abundance and diversity of pollen within mixed samples presents a new frontier for this type of work, but metabarcoding may have limita- tions, and validation is warranted before any suite of primers can be confidently used in a research program. We set out to evaluate the utility of an integrative approach, using a set of established primers (ITS2 and rbcL) versus melisso- palynological analysis for characterizing 27 mixed-pollen samples from agricultural sites across Canada. Results Both individual markers performed well relative to melissopalynology at the family level with decreases in the strength of correlation and linear model fits at the genus level. Integrating data from both markers together via a multi-locus approach provided the best rank-based correlation between metagenetic and melissopalynologi- cal data at both the genus (ρ = 0.659; p < 0.001) and family level (ρ = 0.830; p < 0.001). Species accumulation curves indicated that, after controlling for sampling effort, melissopalynological characterization provides similar or higher species richness estimates than either marker. The higher number of plant species discovered via the metabarcoding approach simply reflects the vastly greater sampling effort in comparison to melissopalynology. Conclusions Pollen metabarcoding performed well at characterizing the composition of mixed pollen samples rela- tive to a traditional melissopalynological approach. Limitations to the quantitative application of this method can be addressed by adopting a multi-locus approach that integrates information from multiple markers.