Tag Jumping Produces Major Distortion on Metabarcoding‐Based Reconstructions of Past and Present Environments

ABSTRACT Analysis of environmental DNA (eDNA) through DNA metabarcoding has become an important technique for environmental science as it allows precise reconstructions of species communities in a fast, cheap and non‐invasive way. In this study, we scrutinize how environmental reconstructions derived from metabarcoding data may be affected by a process in which sample specific labels (tags), added to sequences for identification of individual samples, are changed unintentionally during adapter ligation causing translocation of sequences between samples (‘tag jumping’). We compare animal and plant communities reconstructed using sedimentary eDNA records processed according two different protocols: (i) a twin‐tagging approach (control) where all amplicons received the same tag on both sides ( N = 102); and (ii) a combinatorial tagging protocol (affected by tag jumps) where each amplicon received a unique combination, but where some tags on each side were reused to form new combinations ( N = 102). We analyzed six different sediment matrices and observed higher average number of taxa in the combinatorial tagging dataset in comparison to our twin‐tagged dataset serving as a reference for results unaffected by tag jumps. In the control dataset with twin tagged amplicons, reconstructed animal communities were statistically different in 14 out of 15 pairwise comparisons, while only 8 out of 15 of the comparisons were different when samples were analyzed using the combinatorial tagging protocol. All of the inferred plant communities were statistically different when analyzed with a twin‐tagging approach, while 20% of these plant communities were not different in our combinatorial tagged dataset. Our results clearly show that tag jumps added species to samples where they were not originally present and affects interpretations of species diversity and time‐trends for whole communities. We conclude that tag jumping, being rarely discussed in metabarcoding studies, constitutes a concern in parity with direct sample contamination.