Variations in terrestrial arthropod DNA metabarcoding methods recovers robust beta diversity but variable richness and site indicators

Abstract

Terrestrial arthropod fauna have been suggested as a key indicator of ecological integrity in forest systems. Because phenotypic identification is expert-limited, a shift towards DNA metabarcoding could improve scalability and democratize the use of forest floor arthropods for biomonitoring applications. The objective of this study was to establish the level of field sampling and DNA extraction replication needed for arthropod biodiversity assessments from soil. Processing 15 individually collected soil samples recovered significantly higher median richness (488-614 sequence variants) than pooling the same number of samples (165-191 sequence variants) prior to DNA extraction, and we found no significant richness differences when using 1 or 3 pooled DNA extractions. Beta diversity was robust to changes in methodological regimes. Though our ability to identify taxa to species rank was limited, we were able to use arthropod COI metabarcodes from forest soil to assess richness, distinguish among sites, and recover site indicators based on unnamed exact sequence variants. Our results highlight the need to continue DNA barcoding local taxa during COI metabarcoding studies to help build reference databases. All together, these sampling considerations support the use of soil arthropod COI metabarcoding as a scalable method for biomonitoring.

Figure 1

Overview of sampling methods. The 1C3E experiment was designed to look at the effect of increasing the volume of field soil sampled. The XC3E experiment was designed to look at the effect of the volume of field soil pooled before DNA extraction. The 1C1E and 1C3E samples were compared to look at the effect of processing 1 or 3 DNA extractions per sample. The 1C3E (1 core sample, 3 DNA extractions) experiment included 216 samples from 2 sites, 36 replicates, and 3 layers. The XC3E (2–15 pooled core samples, 3 DNA extractions) experiment included 120 samples from 2 sites, 5 pooling treatments, 4 replicates, and 3 layers. The 1C1E (1 core sample, 1 DNA extraction) experiment included 48 samples from 2 sites, 8 replicates, and 3 layers.

Figure 2

Arthropod ESV richness increases with increasing field sampling effort but varies little when more DNA extractions are performed. Richness is shown for (a) bioinformatically pooled, individually collected field samples, (b) manually pooled field samples, and (c) the difference between samples processed with 3 pooled DNA extractions and 1 DNA extraction (positive values indicate greater richness from 3 pooled DNA extractions; negative values indicate greater richness from 1 DNA extraction). ‘915’ refers to the largest class of pooled samples that is 15 for all bioinformatically pooled samples but varies for manually pooled samples. At Island Lake the largest class is comprised of 15 pooled samples but at Nimitz, the largest class contains 15 pooled samples except for the bryophyte layer where 9–14 samples were pooled.

Figure 3

Clustering of samples across sites and soil layers are robust to intensity of field sampling and number of DNA extraction replicates. Clustered groups are shown based on (a), the collection of individually processed samples, (b) manual pooling of 1–15 samples with the number of pooled samples indicated by the plotted number, and (c) single samples processed with 1 or 3 pooled DNA extractions with the number of extractions indicated by the plotted number.

Figure 4

Taxonomic distribution of site indicator ESVs for each site. Heat trees comprised of all the site indicator ESVs, pooled across all sampling methods, are shown for each site. In each tree, color indicates the number of samples where each taxon was detected; text and node size indicate the number of site indicator ESVs in each taxon. To improve readability, labels have been added only to nodes present in at least half the samples. Taxa that could not be confidently identified are indicated by an asterisk (*).