The review considered publications applying DNA metabarcoding to terrestrial arthropods, including whole-organism community samples, tissue, pollen, gut contents, faeces and environmental DNA. It covered research on species-richness assessments, pollen analysis and food-web studies, excluding individual DNA barcoding, PCR-free methods and long-read sequencing. The literature search was conducted in Google Scholar and EBSCO Discovery Service on 17 January 2022 using terms combining quantitative searches, insects and metabarcoding, and additional eligible studies were included based on the authors’ expertise.

The authors extracted information about study type, sample type, species group, methods and parameters, then assessed whether each approach could support numbers of individuals, relative abundance or biomass while considering additional equipment, cost and labor. This scope means the conclusions concern methodological options identified in the reviewed literature rather than a new experiment or a universal validation of terrestrial arthropod monitoring. The need for faster and broader monitoring is linked to well-documented declines in arthropods and the limits of traditional identification.

Sorting trap catches and identifying specimens by morphology are costly, time-consuming and dependent on taxonomic expertise that is in short supply. DNA metabarcoding can raise sample throughput by automating laboratory and bioinformatic workflows, and it may resolve cryptic species, reveal intraspecific genetic diversity and permit the study of non-invasively collected material such as faeces. It can also detect pollen and prey or predator DNA, making it useful for examining pollination and food-web interactions.

Yet implementation in policy monitoring has been constrained by skepticism among taxonomists, missing expertise and infrastructure, non-standard molecular protocols, incomplete reference databases and the destruction of specimens during DNA extraction. The quantitative interpretation of sequence data remains a central obstacle. A read count is not an individual count.

PCR then introduces taxon-specific distortion because primers bind with different efficiencies, with mismatches near the 3′ end of a primer potentially having a strong effect. Other influences include variation in target-marker GC content, amplicon length and marker-gene copy number, as well as stochastic processes during amplification and sequencing and filtering decisions during bioinformatic processing. Sample complexity can also affect how much sequence is recovered.

A meta-analysis of 22 metabarcoding studies found a weak relationship between biomass and read counts, accompanied by large uncertainty. Because those studies used different protocols, organisms and sample types, their results are not directly comparable, but the analysis supports the central conclusion that raw read counts are unsuitable as stand-alone abundance estimates. Several semi-quantitative metrics are widely used despite their limitations.

Frequency or percentage of occurrence reduces sequence data to presence and absence across samples. Bipartite network analyses use the presence of taxa to represent interactions, while relative read abundance, rarefaction, logarithmic or rank transformation, and related transformations retain different aspects of the read-count table. These measures can reveal community composition and are often informative for pollen, faecal, gut-content and whole-organism samples.

Their calculation is straightforward, but factors from extraction through PCR and sequencing mean that they generally cannot support defensible estimates of the number or weight of organisms. The review therefore distinguishes the absolute number of individuals belonging to a species from its proportional representation in a sample and treats biomass, the weight of those individuals, as a separate quantity. Correction factors offer a more direct attempt to reduce read-abundance bias.

One method processes artificial mock communities of known composition alongside environmental samples, measures how each taxon is represented and uses the result to adjust reads from samples of unknown composition. This can improve abundance estimates where taxa in the mock community also occur in the field samples, a restriction that is substantial for hyperdiverse arthropod catches containing many taxa. Correction factors may also be extended to close relatives when similar primer-related skews are assumed, and one beetle study found strong phylogenetic signals in primer-template mismatches.

An alternative iterative approach begins with randomized correction factors, calculates predicted specimen counts from sequence data, compares those predictions with counts from morphologically identified specimens and repeatedly revises the factors until the two sets converge. This method has produced predicted individual counts that correlated highly with actual counts, but it requires high-quality reference material, concordant morphological and molecular identifications, and substantially more time and expense. Correction factors can also fluctuate when PCR conditions change, and community evenness, copy-number variation and differences in DNA quality can remain uncorrected.

Spike-ins, also called internal standards, offer a comparatively simple way to improve comparisons among samples. A defined amount of reference DNA is added to every sample, preferably after tissue lysis but before extraction, so that it is co-extracted, co-amplified and co-sequenced with the sample. The reference can be tissue, genomic DNA, pre-amplified DNA, a plasmid or synthetic DNA, provided that it contains suitable primer-binding sites and is represented in the reference database.

Dividing the reads assigned to each amplicon sequence variant by the reads assigned to the spike-in can correct for differences in sequencing depth between samples. Studies reviewed by the authors reported improved within-species abundance comparisons, and the method has potential for biomass estimation. It requires little additional laboratory effort and can be used across sample types, but it does not correct biases among species within an individual sample.

The literature reviewed did not systematically resolve the likely roles of sample complexity and taxon-specific primer efficiency in that limitation. Primer design and the use of multiple loci can improve taxon recovery and lessen some distortions, but neither provides a general solution to counting. Primer sets must be universal enough to cover the organisms of interest, generate fragments suitable for the sequencing platform and permit species-level identification.

Their performance should therefore be validated with complex mock communities, because over- and under-amplification of entire arthropod lineages or dominant plant taxa in pollen can distort read abundance. Multi-locus metabarcoding uses markers with different taxonomic biases, including COI and ribosomal or chloroplast regions, and has been applied to whole-organism, tissue, pollen, faecal, gut-content, soil and environmental-DNA samples. Summarizing rank order or median proportional abundance across loci can reduce the influence of any one marker.

Separate PCRs and analyses, however, increase cost and labor, while multiplexing may introduce competition among primer sets. Loci often recover discordant species lists because of different amplification efficiencies and gaps in reference databases. Such lists can broaden coverage, but abundance estimates are generally limited to taxa detected by more than one marker.

The review’s most promising current combination for whole-organism community samples is to photograph and process specimens individually, then recombine selected material for mixed-sample sequencing. Photography records the number of organisms and permits body-size measurements that can support biomass estimates, while metabarcoding supplies species-level identifications that may be difficult to obtain from images alone. Handling specimens separately also makes it possible to remove or sample body parts for DNA extraction rather than necessarily destroying an entire voucher, and the images provide a permanent reference for later checks.

The approach is exceptionally labor-intensive, but automated imaging and machine-learning-based recognition are emerging as ways to improve efficiency. Researchers could combine individually processed specimens into pseudo-communities to retain cost-efficient metabarcoding while linking sequence results to counts and measurements. A related strategy weighs whole-organism samples, uses flow cytometry for pollen or applies quantitative or digital droplet PCR, although each additional method has restrictions on the samples with which it can be combined.

A comprehensive workflow would pair direct specimen information with calibrated molecular data rather than treat any one correction as sufficient. Specimens can be photographed, counted, measured and weighed, with a standardized spike-in added before extraction. The resulting amplicon sequence-variant table provides taxonomic assignments, and image-derived counts, body measurements and biomass estimates can then be assembled with the molecular results.

Positive controls such as mock communities, negative controls, biological replicates and technical replicates are recommended alongside evaluation of primer performance. The additional costs and labor may be justified by improved data, although limited monitoring budgets can make technical replication difficult. Environmental DNA is more demanding because DNA persistence and transport depend on environmental conditions, sampling and the cellular or molecular state of the material.

For that application, the review regards presence and occurrence data, occupancy models and experimental designs that support indirect counts as more defensible than individual abundance estimates. Future approaches include unique molecular identifiers, Bayesian and occupancy models, tracer approaches, decay corrections and PCR-free mitochondrial or nuclear sequencing. Unique molecular identifiers label DNA fragments before amplification, allowing the number of original template copies represented by a sequence variant to be estimated.

Occupancy models can account for false negatives, while Bayesian models can incorporate primer choice and other workflow parameters. PCR-free approaches avoid amplification-related bias, but incomplete reference genomes, extraction bias, variable gene copy number, computational demands and data-storage requirements constrain their use. The review’s broader conclusion is that DNA metabarcoding should complement rather than replace morphological monitoring.

It can efficiently expand the species coverage, detect sensitive trends and support pollen and food-web analyses, but reliable counts and biomass currently require calibrated controls, suitable correction methods and direct specimen-level measurements rather than interpretation of raw reads alone.

Subject of Research: Methods for estimating terrestrial arthropod abundance and biomass with DNA metabarcoding

Article Title: Abundance estimation with DNA metabarcoding – recent advancements for terrestrial arthropods

Article References: Sickel, W., Zizka, V., Scherges, A., Bourlat, S. J., & Dieker, P. (2023). Abundance estimation with DNA metabarcoding – recent advancements for terrestrial arthropods. Metabarcoding and Metagenomics, 7, Article e112290. https://doi.org/10.3897/mbmg.7.112290

Image Credits: AI Generated

DOI: 10.3897/mbmg.7.112290

Keywords: DNA metabarcoding, terrestrial arthropods, abundance estimation, biomass estimation, arthropod monitoring, spike-ins, mock communities, primer bias, environmental DNA, ecological monitoring, Abundance, estimation