Environmental DNA, or eDNA, metabarcoding has transformed biodiversity monitoring over the past decade. Instead of catching and killing insects to identify them, scientists can now collect traces of genetic material that organisms shed in their environment and identify the species from those fragments. In pollinator research, the pioneering work of Thomsen and Sigsgaard in 2019 showed that flowers themselves can serve as a source of arthropod eDNA, allowing researchers to reconstruct pollination networks non-invasively. Since then, a growing number of studies have confirmed that eDNA collected from flowers can effectively reveal insect-plant interactions with high sensitivity, taxonomic breadth, and temporal resolution.

However, the method has faced persistent technical obstacles. One major problem is that flowers are not only visited by pollinators; they are also home to small insects such as thrips, aphids, and pollen beetles that may live on the flower and enter the sample as whole individuals. These flower-living organisms contribute orders of magnitude more DNA than the trace material left by transient pollinators. Because the polymerase chain reaction, or PCR, preferentially amplifies the most abundant DNA in a sample, the genetic signal of bees can be outcompeted by the overwhelming DNA of these abundant residents, producing false negatives. A second challenge lies in the universal arthropod primers themselves, which exhibit well-documented amplification bias toward certain taxonomic groups and often underrepresent bees in mixed floral community samples.

These biases are especially problematic for studies aiming to reconstruct plant-pollinator networks, because such inference requires extensive within-group taxonomic coverage rather than a focus on the most common arthropod taxa in general. The role of certain bee species may be underestimated, skewing network structure and leading to misleading conclusions for biodiversity assessments and conservation efforts. Despite bees being one of the most ecologically and economically important groups in pollination networks, no bee-specific metabarcoding primer had been available until now.

To close this gap, the research team targeted the mitochondrial cytochrome c oxidase subunit I gene, or COI, the most widely used DNA barcode for the animal kingdom. COI combines conserved primer-binding sites with sufficient sequence variation to discriminate between species, and it is well represented in public sequence databases such as BOLD and NCBI. BOLD currently contains approximately 112,000 publicly available bee sequences, of which roughly 108,000 correspond to the standard COI barcode region. The researchers downloaded COI sequences from flower-visiting arthropod families previously detected in flower wash samples, aligned them, and removed low-quality or suspect records, yielding a reference set representing 173,830 unique arthropod species, of which 2,394 were bees covering 39 percent of Austrian bee diversity.

From this alignment, the team identified conserved regions unique within bees, paying particular attention to avoiding shared consensus with non-target groups such as thrips and pollen beetles, which had previously dominated PCR output from flower wash samples. Using manual expert judgment alongside established primer design principles, they generated two forward and three reverse primer candidates with lengths of 20 to 30 base pairs, melting temperatures of 50 to 60 degrees Celsius, and expected amplicon sizes of roughly 200 to 400 base pairs. Degenerate bases were incorporated to account for sequence variation among bee taxa, and all candidates were checked for secondary structures such as hairpins and dimers. The selected pair, BeePrimeF and BeePrimeR, amplifies a short COI fragment of 198 base pairs, a length suitable for species-level resolution in most organismal groups and compatible with Illumina sequencing platforms.

Validation proceeded in two stages. In silico, the primers were screened against an alignment of 67,352 centroid sequences clustered from a total of 316,254 sequences retrieved from NCBI, using the R package PrimerMiner to calculate weighted penalty scores that account for mismatch position and type. Both primers showed high binding complementarity to bee sequences, while their affinity to non-target arthropod groups was substantially lower, with mismatches concentrated at the critical 3-prime ends where they strongly reduce amplification efficiency. In vitro, the primers were tested against DNA extracts from 48 bee specimens representing 38 species and 23 genera spanning all six bee families of Central Europe, alongside 34 non-target extracts including spiders, thrips, beetles, flies, bugs, butterflies, wasps, and ants. Thirty-three of the 38 bee extracts were successfully amplified, an 86.84 percent success rate, while only nine of the 34 non-target extracts yielded positive results, indicating a non-target exclusion rate of approximately 73.52 percent. Notably, no amplification occurred in groups known to be problematic in flower metabarcoding, such as thrips, pollen beetles, or aphids.

The decisive test came with field samples. The researchers applied BeePrime to eDNA extracted from 48 flower-wash samples collected in Austrian flower strips in 2024 and compared the results directly with the universal arthropod primer pair fwhF2/fwhR2n, one of the most widely recognized primers available. BeePrime detected bee DNA in 35 of the 48 samples, compared with only 17 for the universal primer, raising per-sample detection from 35.4 percent to 72.9 percent. At the species level, the universal primer identified 14 bee-positive samples, while BeePrime detected 87, a sixfold increase. Individual floral samples typically yielded multiple bee species rather than single detections, with a mean of 1.8 species-level detections per sample and a maximum of 6. Across the entire dataset, the universal primer recovered just six distinct bee species, whereas BeePrime detected 27.

The improvement in sequencing efficiency was equally striking. Amplifications with the universal primer produced only very low proportions of bee reads, with a median of 0.03 percent, in many cases approaching zero. BeePrime yielded a median of 15 percent and a mean of approximately 30 percent bee reads per sample, a difference confirmed as highly significant by a Wilcoxon rank-sum test. Thrips, which dominated the genus-level dataset with almost 85 percent of all reads under the universal primer, were reduced to roughly 30 percent under BeePrime, while wild bees rose from only a few percent to nearly 50 percent of reads. Together with honeybees, bees represented more than 60 percent of the total relative read proportion with the new primer pair.

Every bee taxon detected with BeePrime had previously been recorded at the same sites using traditional hand-net sampling and morphological identification, corroborating the molecular assignments. The authors caution, however, that empirical performance has so far been demonstrated only for common bee taxa in Central European agricultural landscapes, and that local testing is advised before applying the primer pair in other biogeographic regions. Several Apidae species, including some carpenter bees and long-horned bees, did not amplify successfully despite acceptable primer fit, so underdetected lineages should not automatically be interpreted as absent in ecological network analyses. The researchers also emphasize that read numbers do not translate directly into bee abundance, since DNA deposition depends on body size, body hairiness, and behavior. Nevertheless, by overcoming a key methodological bottleneck in pollinator eDNA metabarcoding, BeePrime substantially strengthens the molecular toolkit available for monitoring bee diversity and resolving plant-pollinator networks, marking a major step toward establishing floral eDNA as a routine approach in ecological monitoring and conservation.

Subject of Research: Development and validation of a bee-specific COI metabarcoding primer pair for non-invasive monitoring of plant-pollinator interactions

Article Title: A bee-specific metabarcoding primer pair for non-invasive assessments of plant–pollinator interactions

Article References: Gurten, S., Sint, D., Traugott, M., & Rennstam Rubbmark, O. R. (2026). A bee-specific metabarcoding primer pair for non-invasive assessments of plant–pollinator interactions. Metabarcoding and Metagenomics, 10, Article e183708. https://doi.org/10.3897/mbmg.10.183708

Image Credits: AI Generated

DOI: 10.3897/mbmg.10.183708

Keywords: bees, environmental DNA, metabarcoding, pollinators, primer design, COI barcode, plant-pollinator networks, biodiversity monitoring, bee-specific, primer, pair, non-invasive