Developing a non-destructive metabarcoding protocol for detection of pest insects in bulk trap catches

Abstract

Metabarcoding has the potential to revolutionise insect surveillance by providing high-throughput and cost-effective species identification of all specimens within mixed trap catches. Nevertheless, incorporation of metabarcoding into insect diagnostic laboratories will first require the development and evaluation of protocols that adhere to the specialised regulatory requirements of invasive species surveillance. In this study, we develop a multi-locus non-destructive metabarcoding protocol that allows sensitive detection of agricultural pests, and subsequent confirmation using traditional diagnostic techniques. We validate this protocol for the detection of tomato potato psyllid (Bactericera cockerelli) and Russian wheat aphid (Diuraphis noxia) within mock communities and field survey traps. We find that metabarcoding can reliably detect target insects within mixed community samples, including specimens that morphological identification did not initially detect, but sensitivity appears inversely related to community size and is impacted by primer biases, target loci, and sample indexing strategy. While our multi-locus approach allowed independent validation of target detection, lack of reference sequences for 18S and 12S restricted its usefulness for estimating diversity in field samples. The non-destructive DNA extraction proved invaluable for resolving inconsistencies between morphological and metabarcoding identification results, and post-extraction specimens were suitable for both morphological re-examination and DNA re-extraction for confirmatory barcoding.

Figure 1

(A) Non-destructive DNA extraction protocol newly developed in this study. (B) Specimens from a trap sample before (left) and after (right) non-destructive DNA extraction using QuickExtract. Both Hemiptera (above) and other insect (below) specimens are preserved.

Figure 2

(A) Taxonomic composition of the reference database, displaying number of unique species for each order within Insecta, as well as for other classes of Arthropods. Number of amplicon sequence variants (ASVs) from (B) mock communities and (C) field trap samples successfully assigned to taxonomic ranks for each locus using the Ribosomal Database Project (RDP) naïve Bayesian classifier and exact matching with reference sequences. (D) Taxonomic overlap between loci at the family, genus, and species level for all samples.

Figure 3

The expected (based on number of individuals) and observed (based on sequencing reads) relative abundance of each species in each mock community pool. Acizzia alternata and A. solanicola are aggregated for display purposes as these species could not be differentiated by the 18S loci. Observed relative abundance data is shown for the mean across the three loci, and for COI, 18S, and 12S separately. False positive and negative genera are indicated in each pool based on a detection threshold of 0.01%. Asterisks (*) indicate pools that had all species correctly identified.

Figure 4

The Diuraphis noxia (Russian wheat aphid) nymph specimen responsible for the apparent “false positive” reads detected in the 1000 Pool 1 mock community, following non-destructive DNA extraction.

Figure 5

Heat map displaying the relative abundance of different Arthropoda taxa in both mock community and field trapped insect samples based on sequencing reads. Taxa designated ‘spp.’ denote genus-level classification. The loci contributing to detection of each taxa is indicated by a red (12S), yellow (18S), or blue (COI) dot. The size and number inside the dot indicate if more than one amplicon sequence variant (ASV) was assigned to the taxonomic rank. Mean relative abundances across all three amplicons are displayed on a log10 scale. Only ASVs that could be reliably classified to genus with a relative abundance above 0.01% are reported.