Optimizing target-to-total DNA ratio in eDNA studies: effects of sampling, preservation, and extraction methods on...

There are many decisions to be made when sampling for environmental DNA (eDNA) analysis, whether using a targeted, single-species assay or community-based metabarcoding. Of the entire workflow from sampling water to bioinformatic analyses, the first steps in the process of collecting water, filtering it, and preserving the filter membranes represent major decision points upon which the success of downstream processes depend. Though many previous studies have compared water volume filtered, filter pore size, and preservation and extraction methods, the conclusions are often that they produce different results, but it is unclear which is the optimal approach for a given purpose. Here, rather than provide yet another methods comparison paper, we provide a framework for how to make informed decisions from a methods comparison and, importantly, how to combine data collected via different methodological choices. We investigate (1) the volume of water filtered and the filter pore size and (2) the preservation method and extraction method of samples with a specific lens on how these choices impact the detection of a single targeted species (Atlantic bottlenose dolphin, Tursiops truncatus, via quantitative PCR (qPCR)), although in principle these findings apply to single-species assays more generally. We find that larger pore size filters (5 µm vs. 1 µm) and larger volumes of water (3 L vs. 1 L) maximize the ratio of amplifiable target DNA to total DNA without compromising the absolute detection of target. We also find that maximizing total DNA yield during extraction (phenol chloroform vs. two commercial kits) does not always increase target detection likely due to the concentration of inhibitors and co-extraction of off-target DNA. We also comment on variation including technical and biological variability between replicates, finding that by homogenizing source water before filtering removes much of the biological variation. Finally, we present a statistical model that allows for inclusion of data from samples collected and processed in different ways, enabling researchers to change protocols or include data from other field sampling efforts, thereby opening up more possibilities to extend datasets and analyses.