Rudolf Meier

Bioassessment of freshwater quality with eDNA is a powerful alternative to traditional methods involving collecting, sorting, and identifying metazoan taxa with morphology (e.g., macroinvertebrates). Particularly attractive for routine monitoring would be an eDNA method that uses a remote‐controlled boat for collecting small volumes of water without filtration in order to simplify sampling. Such a method would not likely capture eDNA signals for all metazoan species, but may nevertheless allow for cost‐effective, and frequent monitoring of water quality, which is important for tropical waterbodies that require year‐round surveillance. We here optimize a molecular protocol for capturing metazoan signatures based on eDNA obtained from 15 ml of water. To test the robustness of the method, we used samples from two tropical reservoirs with known differences in water quality to optimize molecular procedures so that they yield repeatable results. Each reservoir was sampled at three sites (“biological replicates”) and each water sample was subsampled twice before extracting the eDNA using ethanol precipitation for both subsamples (“technical replicates”). We then tested how much DNA (0.1 ng to 15 ng) and how many PCR cycles (25 or 35) minimized the variance of the metazoan eDNA signal between biological and technical replicates. For this purpose, we amplified a 313 bp COI minibarcode using a universal metazoan primer pair. We found that regardless of template amounts or PCR cycle numbers, the eDNA signatures for both reservoirs were distinct because only 17 of 59 mOTUs (mainly planktonic crustaceans and rotifers) were shared. We also found that template amounts between 0.5 and 15 ng yielded overall similar results, but the use of 35 PCR cycles significantly depressed the number of detected species ( p ‐value < 0.05). Fortunately, the differences between the detected metazoan community of the reservoirs were so strong that all biological and technical replicates could be assigned unambiguously to their source reservoir although the variance between technical replicates remained high for all PCR experiments (Bray‐Curtis dissimilarity: 5%–20%; Jaccard distance: 10%–40%).