Freshwater biodiversity provides important ecosystem services and is at the core of water quality monitoring worldwide. To assess freshwater biodiversity, genetic methods such as metabarcoding are increasingly used as they are faster and allow better taxonomic resolution than manual identification methods. Either sampled organisms are used directly for "bulk metabarcoding," or water is filtered and the extracted environmental DNA serves as a proxy for biodiversity via "eDNA metabarcoding." Despite the advantages of both methods, questions remain regarding their comparability and applicability for routine biomonitoring and stressor impact assessment. Therefore, we compared metabarcoding results from bulk and eDNA samples taken from 19 streams spanning a wide gradient of farming intensities in New Zealand. We performed PCR with highly degenerate cytochrome c oxidase I primers and sequenced libraries on an Illumina MiSeq. The inferred community composition differed strongly between the two methods. More taxa were captured by eDNA than bulk-sample metabarcoding (5,819 vs. 1,483), but more of the commonly used invertebrate bioindicator taxa (mayflies, stoneflies and caddisflies) were found in bulk (47) than eDNA samples (37). Catchment-wide and local land use impacts on communities were detected better by eDNA metabarcoding, especially for non-metazoan taxa. Our findings imply that bulk-sample metabarcoding resembles classical freshwater biomonitoring approaches better, as more indicator macroinvertebrate taxa are captured. However, eDNA metabarcoding might be better suited to infer the impact of stressors on stream ecosystems at larger scales, as many new and potentially more informative taxa are registered. We therefore suggest exploring both methods in future assessments of stream biodiversity.

Metabarcoding is increasingly used to assess species diversity by high‐throughput sequencing where millions of sequences can be generated in parallel and multiple samples can be analysed in one sequencing run. Generating amplified fragments with a unique sequence identifier ('tag') for each sample is crucial, as it allows assigning sequences to the original samples. The tagging through so‐called fusion primers is a fast and cheap alternative to commercially produced ligation‐based kits. However, little is known about potential bias and inconsistencies introduced by the long nucleotide tail attached to those primers, which could lead to deficient reports of community composition in metabarcoding studies. We therefore tested the consistency and taxa detection efficiency of fusion primers in (1) a one‐step and (2) two‐step PCR protocol as well as (3) a commercially manufactured Illumina kit using mock communities of known composition. The Illumina kit delivered the most consistent results and detected the highest number of taxa. However, success of the two‐step PCR approach was only marginally lower compared to the kit with the additional advantage of a much more competitive price per library. While most taxa were also detected with the one‐step PCR approach, the consistency between replicates including read abundance was substantially lower. Our results highlight that method choice depends on the precision needed for analysis as well as on economic considerations and recommend the Illumina kit to obtain most accurate results and the two‐step PCR approach as a much cheaper yet very robust alternative.

Environmental DNA metabarcoding has become a popular tool for the assessment of freshwater biodiversity, but it is largely unclear how sampling time and location influence the assessment of communities. Abiotic factors in rivers can change on small spatial and temporal scale and might greatly influence eDNA metabarcoding results. In this study, we sampled three German rivers at four locations per sampling site: 1. Left river side, surface water 2. Right river side, surface water, 3. Left side, close to the riverbed, 4. Right side, close to the riverbed. For the rivers Ruhr and Möhne, sampling was conducted three times in spring, each sampling one week apart. The Ruhr was again sampled in autumn and the Gillbach was sampled in winter. Sequencing on an Illumina MiSeq with COI primers Bf2/BR2 revealed diverse communities (6493 Operational taxonomic units, OTUs), which largely differed between rivers. Communities changed significantly over time in the Ruhr, but not in the Möhne. Sampling location influenced recovered communities in the Möhne and in the Ruhr in autumn. Our results have important implications for future eDNA studies, which should take into account that not all eDNA in rivers is everywhere and not at all times.

The hyporheic zone, i.e. the ecotone between surface water and the groundwater, is a rarely studied freshwater ecosystem. Hyporheic taxa are often meiofaunal (<1 mm) in size and difficult to identify based on morphology. Metabarcoding approaches are promising for the study of these environments and taxa, but it is yet unclear if commonly applied metabarcoding primers and replication strategies can be used. In this study, we took sediment cores from two near natural upstream (NNU) and two ecologically improved downstream (EID) sites in the Boye catchment (Emscher River, Germany), metabarcoding their meiofaunal communities. We evaluated the usability of a commonly used, highly degenerate COI primer pair (BF2/BR2) and tested how sequencing three PCR replicates per sample and removing MOTUs present in only one out of three replicates impacts the inferred community composition. A total of 22,514 MOTUs were detected, of which only 263 were identified as Metazoa. Our results highlight the gaps in reference databases for meiofaunal taxa and the potential problems of using highly degenerate primers for studying samples containing a high number of non-metazoan taxa. Alpha diversity was higher in EID sites and showed higher community similarity when compared to NNU sites. Beta diversity analyses showed that removing MOTUs detected in only one out of three replicates per site greatly increased community similarity in samples. Sequencing three sample replicates and removing rare MOTUs is seen as a good compromise between retaining too many false-positives and introducing too many false-negatives. We conclude that metabarcoding hyporheic communities using highly degenerate COI primers can provide valuable first insights into the diversity of these ecosystems and highlight some potential application scenarios.

Traditional morphological methods for species identification are highly time consuming, especially for small organisms, such as Foraminifera, a group of shell-building microbial eukaryotes. To analyze large amounts of samples more efficiently, species identification methods have extended to molecular tools in the last few decades. Although a wide range of phyla have good markers available, for Foraminifera only one hypervariable marker from the ribosomal region (18S) is widely used. Recently a new mitochondrial marker cytochrome oxidase subunit 1 (COI) has been sequenced. Here we investigate whether this marker has a higher potential for species identification compared to the ribosomal marker. We explore the genetic variability of both the 18S and COI markers in 22 benthic foraminiferal morphospecies (orders Miliolida and Rotaliida). Using single-cell DNA, the genetic variability within specimens (intra) and between specimens (inter) of each species was assessed using next-generation sequencing. Amplification success rate was twice as high for COI (151/200 specimens) than for 18S (73/200 specimens). The COI marker showed greatly decreased intra- and inter-specimen variability compared to 18S in six out of seven selected species. The 18S phylogenetic reconstruction fails to adequately cluster multiple species together in contrast to COI. Additionally, the COI marker helped recognize misclassified specimens difficult to morphologically identify to the species level. Integrative taxonomy, combining morphological and molecular characteristics, provides a robust picture of the foraminiferal species diversity. Finally, we suggest the use of a set of sequences (two or more) to describe species showing intra-genomic variability additionally to using multiple markers. Our findings highlight the potential of the newly discovered mitochondrial marker for molecular species identification and metabarcoding purposes.

Foraminifera are a species-rich phylum of rhizarian protists that are highly abundant in many marine environments and play a major role in global carbon cycling. Species recognition in Foraminifera is mainly based on morphological characters and nuclear 18S ribosomal RNA barcoding. The 18S rRNA contains variable sequence regions that allow for the identification of most foraminiferal species. Still, some species show limited variability, while others contain high levels of intragenomic polymorphisms, thereby complicating species identification. The use of additional, easily obtainable molecular markers other than 18S rRNA will enable more detailed investigation of evolutionary history, population genetics and speciation in Foraminifera. Here we present the first mitochondrial cytochrome c oxidase subunit 1 (COI) gene sequences (“barcodes”) of Foraminifera. We applied shotgun sequencing to single foraminiferal specimens, assembled COI, and developed primers that allow amplification of COI in a wide range of foraminiferal species. We obtained COI sequences of 49 specimens from 17 species from the orders Rotaliida and Miliolida. Phylogenetic analysis showed that the COI tree is largely congruent with previously published 18S rRNA phylogenies. Furthermore, species delimitation with ASAP and ABGD algorithms showed that foraminiferal species can be identified based on COI barcodes.

Repeated Quaternary glaciations have significantly shaped the present distribution and diversity of several European species in aquatic and terrestrial habitats. To study the phylogeography of freshwater invertebrates, patterns of intraspecific variation have been examined primarily using mitochondrial DNA markers that may yield results unrepresentative of the true species history. Here, population genetic parameters were inferred for a montane aquatic caddisfly, Thremma gallicum, by sequencing a 658-bp fragment of the mitochondrial CO1 gene, and 12,514 nuclear RAD loci. T. gallicum has a highly disjunct distribution in southern and central Europe, with known populations in the Cantabrian Mountains, Pyrenees, Massif Central, and Black Forest. Both datasets represented rangewide sampling of T. gallicum. For the CO1 dataset, this included 352 specimens from 26 populations, and for the RAD dataset, 17 specimens from eight populations. We tested 20 competing phylogeographic scenarios using approximate Bayesian computation (ABC) and estimated genetic diversity patterns. Support for phylogeographic scenarios and diversity estimates differed between datasets with the RAD data favouring a southern origin of extant populations and indicating the Cantabrian Mountains and Massif Central populations to represent highly diverse populations as compared with the Pyrenees and Black Forest populations. The CO1 data supported a vicariance scenario (north–south) and yielded inconsistent diversity estimates. Permutation tests suggest that a few hundred polymorphic RAD SNPs are necessary for reliable parameter estimates. Our results highlight the potential of RAD and ABC-based hypothesis testing to complement phylogeographic studies on non-model species.

No abstract