Impacts of DNA extraction and PCR on DNA metabarcoding estimates of soil biodiversity

Soils are ubiquitous and important components of terrestrial ecosystems, richly populated with diverse organisms with important ecological roles. DNA metabarcoding is a promising tool for efficient and taxonomically comprehensive analyses of soil biodiversity, but the outcomes of these analyses are likely affected by basic methodological factors such as sampling and laboratory protocols. We investigated the impacts of DNA extraction methodology and multiple PCRs on DNA metabarcoding biodiversity estimates for multiple taxonomic groups in soil. We applied four DNA extraction methods with different size constraints in parallel to fixed volumes of soil from a pair of forest sites, followed by sets of ten individually indexed PCRs and Illumina sequencing of prokaryote (16S) and eukaryote (18S, fungal 26S, and metazoan cytochrome c oxidase subunit I) barcodes. Methods utilizing larger soil volumes resulted in higher biodiversity estimates for arthropods but not for prokaryotes or microeukaryotes, and improved spatial discrimination of metazoan communities but not prokaryotes. Each DNA extraction method resulted in a biased community composition, and these biases were consistent across sites and amplicons. Some 36%-41% of operational taxonomic units (OTUs) were shared between all four DNA extraction methods and OTUs from many phyla had contrasting abundances from different methods. From 0.2% to 19% of OTUs, accounting for 9.4% to 97% of sequence reads, were amplified in all 10 PCRs from each DNA extraction method and sample. OTUs detected in only a single PCR accounted for 26% to 41% of OTUs but only 0.3% to 2.8% of the sequence reads from each extraction method and sample. The inferred species richness increased strongly with the number of PCRs carried out. We conclude that a consistent DNA extraction method should be used for comparisons between samples, confirm that larger samples of soil (>= 15 g) should be used for analyses of metazoan biodiversity than for microbes, and that the number of PCR replicates undertaken strongly affects species richness estimates.