Foraging specialization of flower visitors in a grassland community: insights from metabarcoding versus traditional methods.

Abstract

Ecological networks of plant-pollinator interactions at the community level were traditionally studied using natural history-type direct observations as well as microscope-based palynology. Over the last decade, these methods have been complemented by emerging molecular technologies such as metabarcoding of pollen loads on insects which holds the promise to reveal interactions that would be difficult to observe directly. We compared traditional methods of studying plant-pollinator interactions (observational data, palynology) with metabarcoding data to assess the degree of complementarity of the different approaches and whether the molecularbased approach revealed previously unknown interactions. This was done for a community on Mount Gilboa, KwaZulu-Natal, South Africa. In an initial exploratory study, I demonstrated successful amplification of DNA from fuchsin-fixed pollen which opens the possibility of using this method of pollen barcoding from specific sites on pollinator bodies, including fragile archival specimen. My results showed that networks constructed from metabarcoding data were more complex with a higher resolution of ecological interactions than those obtained using the traditional methods of microscopy of pollen-loads and visitor records. A complete community approach combining both the nocturnal and diurnal components showed a complex network comprising 66 pollinator species and 172 plant species, corresponding to a network with low modularity, high nestedness and high linkage density, and which is likely to be ecologically robust owing to high generality within the network. However, because many of the plants and animals in the community have not yet been barcoded and were not available in the DNA barcode reference library, it was difficult to obtain species-specific names for 72% of the putative plant species and 45% of the putative animal species in these networks. These were assigned to the lowest taxonomic level possible after cross referencing against known occurrence records. Nocturnal pollination networks revealed a 2.5-fold increase in interaction diversity, and 4.5-fold increase in pollinator generality when using metabarcoding data in comparison to conventional microscopy. Hyper diverse countries such as South Africa, however, still require significant resources to build comprehensive reference libraries allowing for sequence data to accurately assign species names. I conclude that the greater resolution and throughput obtained through metabarcoding can increase our understanding of complex ecological interactions and networks.