Abstract
DNA metabarcoding from the ethanol used to store macroinvertebrate bulk samples is a convenient methodological option in molecular biodiversity assessment and biomonitoring of aquatic ecosystems, as it preserves specimens and reduces problems associated with sample sorting. However, this method may be affected by errors and biases, which need to be thoroughly quantified before it can be mainstreamed into biomonitoring programmes. Here, we used 80 unsorted macroinvertebrate samples collected in Portugal under a Water Framework Directive monitoring programme, to compare community diversity and taxonomic composition metrics estimated through morphotaxonomy versus metabarcoding from storage ethanol using three markers (COI-M19BR2, 16S-Inse01 and 18S-Euka02) and a multimarker approach. A preliminary in silico analysis showed that the three markers were adequate for the target taxa, with detection failures related primarily to the lack of adequate barcodes in public databases. Metabarcoding of ethanol samples retrieved far less taxa per site (alpha diversity) than morphotaxonomy, albeit with smaller differences for COI-M19BR2 and the multimarker approach, while estimates of taxa turnover (beta diversity) among sites were similar across methods. Using generalized linear mixed models, we found that after controlling for differences in read coverage across samples, the probability of detection of a taxon was positively related to its proportional abundance, and negatively so to the presence of heavily sclerotized exoskeleton (e.g., Coleoptera). Overall, using our experimental protocol with different template dilutions, the COI marker showed the best performance, but we recommend the use of a multimarker approach to detect a wider range of taxa in freshwater macroinvertebrate samples. Further methodological development and optimization efforts are needed to reduce biases associated with body armouring and rarity in some macroinvertebrate taxa.