4.3 Universality and resolution of primers
Our analysis did not identify one single outperforming metabarcode. The universality of primers was variable among taxa, with Euka02 showing the highest performance for some phyla (platyhelminthes, molluscs, annelids and even some arthropods), and Inse01 showing a generally good performance for insects. However, each of these metabarcodes has some drawbacks. For instance, Euka02 amplifies very long sequences for some taxa of crustaceans (Isopoda and Amphipoda; Guardiola et al.2015; Taberlet et al. 2018) thus their eDNA metabarcoding with this marker is problematic. Conversely, Inse01 is a metabarcode developed specifically for insects, and fails to amplify key freshwater taxa such as Turbellaria and molluscs (Fig. 1). In our in vitroanalysis, BF1_BR2-COI showed a moderate amplification rate, still for insects a relevant proportion of specimens were not amplified (Fig. 1). This is in contrast with previous analyses, that successfully amplified 100% of tested insects using BF1_BR2-COI (Elbrecht & Leese 2017). Differences might be due to DNA quality, as this primer amplifies relatively long metabarcodes (>300 bp). Some of our >1500 specimens were old, and this can cause DNA degradation, while the starting material of Elbrecht and Leese (2017) was probably of better quality. Furthermore, in several cases BF1_BR2-COI did not amplify the DNA of our target organisms, but amplified the DNA of contaminants, i.e. other organisms for which small body fragments were probably present in the tube, and that perhaps showed excellent match with the primers. Unfortunately, these conditions (degraded DNA, and contemporaneous presence of many organisms) are typical of eDNA metabarcoding studies, stressing the complexity of finding appropriate primers.
Differences in performance were also strong when considering the resolution of the metabarcodes. BF1_BR2-COI clearly showed the best resolution while Euka02 showed a very poor performance, as in many cases it failed even at the family level (Fig. 3; see also Tang et al.2012). COI is a highly variable region, and this has promoted its use as standard barcode for animals (Hebert et al. 2003; Hebert et al. 2004; Andújar et al. 2018). The excellent performance of BF1_BR2-COI can also be explained by the relatively long amplified region. Inse01 showed an intermediate performance, as its resolution was insufficient for species-level identification, while genus level identification was good for most of taxa (Fig. 3). It must be remarked that these are optimistic values of resolution, given that our database was far from complete, particularly at the genus-level and for some taxa, therefore a more complete database could yield poorer resolution values.