A metabarcoding framework for wild bee assessment in Luxembourg

Abstract

Wild bees are crucial organisms for terrestrial environments. Their ongoing decline could cause irreparable damage to ecosystem services vital to plant reproduction and human food production. The importance of taking swift action to prevent further declines is widely acknowledged, but the current deficit of reliable taxonomic information complicates the development of efficient conservation strategies targeting wild bees. DNA metabarcoding can help to improve this situation by providing rapid and standardized mass identification. This technique allows the analysis of large numbers of specimens without the need for specialized taxonomic knowledge by matching high-throughput sequencing reads against public DNA barcode reference libraries. However, the validation of this approach for wild bees requires the evaluation of potential error sources on a regional scale. Here we analyzed the effects of three potential error sources on a metabarcoding pipeline customized for the wild bee fauna of Luxembourg. In an in silico study, we checked the completeness of the BOLD reference library for 349 species found in the country, the correspondence between molecular and morphological species delimitation for these taxa, and the amplification efficiency of three commonly used metabarcoding primer pairs (mlCOlintF/HCO2198, LepF1/MLepF1-Rev and BF2/BR2). The detection power of the pipeline was evaluated based on the species recovery rates from mock communities of known composition under variable DNA concentration treatments. The reference barcode library evaluation results show that 97% of the species have at least a single barcode in BOLD Systems (minimal length 196 bp) and that 85% of species have ≥ 5 barcodes in the public domain. The mlCOlintF/HCO2198 target fragment presented the highest coverage (77.94% of the species with full barcode sequences), followed by the target fragments of LepF1/MLepF1-Rev (77.65%) and BF2/BR2 (68.48%). Only 60% of the morphospecies presented a complete coverage of the prominent Folmer region (658 bp). The in silico amplification efficiency analysis shows that the BF2/BR2 primer pair has the best-predicted amplification performance, but none of the primer combinations evaluated can be expected to efficiently amplify all local wild bee genera. Finally, all species detection rates in the mock communities, except for the sample with the most discrepant DNA concentrations, were above 97%, with no significant differences found among treatments. These results indicate that the detection capacity of the pipeline is robust enough to be used for the reliable assessment of local wild bee biodiversity, even if species from various size categories are pooled together. Primer bias has a major effect on species detection, which can be acknowledged with a preliminary assessment of primer-template mismatch and sophisticated methodological designs (e.g. mock community controls, replicates). Overall, the metabarcoding pipeline here described provides a suitable tool for quick and reliable taxonomic identification of the regional wild bee fauna to aid conservation initiatives in Luxembourg – and beyond.