Metabarcoding
We followed the methodological approach described in Scholz and Voigt (2022). In brief, DNA was extracted from the frozen fish gut content by applying NucleoSpin© Food and NucleoSpin© Soil Kit (Macherey-Nagel GmbH & KG, Düren, Germany) as outlined in the manufacturer’s instructions. We performed two DNA extractions for each gut content sample. The concentration of the extracts was determined by fluorometric quantification in a Qubit Fluorometer (Qubit fluorometric quantification dsDNA High Sensitivity Kit, ThermoFisher Scientific, Walham, USA). Some of the DNAs had to be cleaned and concentrated using a DNA Clean and Concentrator Kit (Zymo Research, 17062 Murphy Ave, Irvine, CA 92614, USA) to get rid of PCR-inhibitors. Throughout the laboratory work, we strictly applied protocols to prevent contaminations by alien DNA or PCR products. The presence of contaminations was checked through all laboratory steps using different negative controls.
We performed a double-PCR strategy with dual indexing. The first PCR amplified the target region CO1 (Cytochrome oxidase subunit 1) region (Galan et al. 2018), the second PCR added the indices to the target region. Products were checked with an agarose gel and cleaned twice with magnetic beads (CleanNGS, GC biotech, Waddinxveen, Niederlande). All products were quantified by fluorometric quantification in the plate reader (Quant-iT™ dsDNA Assay Kit, high sensitivity, ThermoFisher Scientific, Walham, USA) and pooled in equimolar concentration. If necessary, the final library was purified and concentrated by using CleanNGS beads. The quality and integrity of the library were confirmed using the Agilent 2200 TapeStation with D1000 ScreenTapes (Agilent Technologies, Santa Clara, California, USA).
Sequences were generated at the Berlin Centre for Genomics in Biodiversity Research (BeGenDiv) in two runs on the Illumina MiSeq platform (Illumina, San Diego, California, USA) using v3 chemistry with 600 cycles. The quality of the generated reads was evaluated using FastQC v.0.11.9 and multiqc. The remaining adapter sequences were removed using cutadapt (version 2.8).
Sequencing reads processing from quality control to taxonomic assignment was performed using the R package “dada2” (Callahan et al. 2016). We assigned taxonomy to the inferred Amplified Sequence Variants (ASVs) up to species level based on the reference database provided by the BeGenDiv (Heller et al. 2018). Taxonomy was assigned based on the single best hit or a last common ancestor (in case of multiple best hits) with 50 out of 100 bootstrap replicates as minimum bootstrap confidence for assigning a taxonomic level. For post-sequencing removal of reads caused by contamination, we used the R package “microDecon” (McKnight et al. 2019) which uses the proportions of ASVs in blank samples (negative controls) to systematically identify and remove contaminant reads from the metabarcoding data set. Afterwards, we summed up reads for pseudo-biological replicates and removed reads which were only present in one of two technical replicates to further increase the power and quality of our data set.
We restricted our dataset to results of sequences on the species level and deleted the finding of bat sequences (Myotis daubentoniid andPipistrellus pygmaeus ) in the gut content of seven fish individuals, as we assumed that this was not the results of selective feeding on the bats, but instead the incorporation of bat feces. All species identified in the gut content were classified into the categories “aquatic” and “terrestrial”, corresponding to their dominant life phases and we counted the number of ingested terrestrial species for each fish individual.