False positives and negatives
Long-term fish surveys of the studied reservoirs provided robust background data on the expected distribution and abundance of individual species that allows these reservoirs to serve as models to study the mechanisms underlying spatial fish distribution. Gillnets are most commonly used to sample fish despite their invasiveness, size- and species-selectivity (Blabolil et al., 2016). Plentiful long-term (Znachor et al., 2020) and detailed spatially-resolved seasonal data (Rychtecký & Znachor, 2011) as well as gillnet data dating back to 2004 (Blabolil, et al., 2017b; Říha et al., 2015) are available for Římov. Klíčava was intensively studied for decades after dam building (Straškrabová & Pivnička, 2001), and together with Žlutice, is studied for current ichthyological research interests (e.g. Vejřík et al., 2019). Three and five gillnet campaigns in Klíčava and Žlutice respectively are supported by data from other sampling methods (i.e. electrofishing, fyke-nets, seining, hook lines and trawling) as well as expert knowledge about fish-habitat relationships (Blabolil et al., 2017b) All of these reservoirs are also routinely monitored by the Vltava River authorities for water chemistry and biological parameters (www.pvl.cz).
Based on expert knowledge and long-term monitoring of these reservoirs and their fish fauna, only a single taxon was identified as a clear false positive. Capelin, a marine fish of the North Atlantic, is likely environmental contamination, possibly through the use of Northern pike bait. No target fish DNA was found in the PCR positive and negative controls or the field/filtration, evaporation, and extraction blanks, which indicates that the field and laboratory protocols used were effective in avoiding contamination. Only one species, burbot, was previously observed in Římov (last caught in spring 2011) and not detected by eDNA metabarcoding. Burbot is the only member of the family Gadidae in Czech waters, therefore misclassification is highly improbable. However, rarer eDNA templates are more likely to be missed during water sampling or lost during molecular workflow processing (Kelly et al., 2019). Burbot is active in cold water, thus eDNA sampling in winter or early spring may improve detection (Blabolil et al., 2018). We performed three PCR replicates per sample in order to reduce stochastic variation, but greater replication may have increased detection probability of rare species within individual samples (Harper et al., 2019b).
Some fishes, such as asp and rudd or European perch and pikeperch, belong to different genera, but sequences amplified using the 12S-V5 primers were identical. These species pairs were excluded from downstream analyses because of assignment uncertainty and differing ecologies, i.e. piscivorous asp and pikeperch preferring open water, and invertivorous European perch and omnivorous rudd preferring structured habitats (Blabolil et al., 2016, 2017a, 2017b). Taxonomic coverage, resolution, and accuracy could be improved for these and other genetically similar taxa (e.g. sturgeons and coregonids) by using different or additional primers that target bony fish, such as MiFish (Miya et al., 2015) and Teleo (Valentini et al., 2016). For taxonomic assignment, we used BLAST at 99% identity against a reference database for UK fishes (Hänfling et al., 2016) that was supplemented with sequences for missing Czech fishes. Reference databases should ideally comprise sequences from specimens occurring in the study area for all species to account for potential phylogeographic variation as species can be locally adapted (Delrieu-Trottin et al., 2019). With a more comprehensive reference database, it may be possible to use 100% BLAST identity for taxonomic assignment.