Tele02 primer leads five-way test of eDNA fish metabarcoding primers

When researchers mixed DNA from 45 Central European freshwater fish species into artificial communities and screened the mixture with five commonly used environmental DNA primer pairs, one candidate clearly pulled ahead. The tele02 primer pair, a modified version of the widely used MiFish-U primer, recovered the highest number of true-positive species while producing the fewest false-positive and false-negative detections of any primer tested. The finding, published as a systematic mock community evaluation, offers the most direct experimental guidance to date for laboratories that need to choose a primer pair for fish monitoring in Central European rivers and lakes.

Environmental DNA metabarcoding has become a mainstay of aquatic biodiversity assessment because it is fast, sensitive, non-invasive and cost-efficient compared with traditional methods such as gillnetting or electrofishing. Yet the technique depends on a chain of decisions, from sampling strategy to wet-lab processing, reference database completeness and, critically, the choice of primers. A poor primer choice can generate false-negative detections that erase species from a monitoring record, or false positives that inflate taxon lists with species that were never present. Although dozens of fish-targeting primer pairs have been published, systematic head-to-head tests on European freshwater communities were largely missing. Earlier comparisons had focused on Asian and Australian fish faunas, and European evaluations had been limited to estuarine and coastal samples or a smaller-scale study of UK lake fish.

To close that gap, the team built mock communities from mucus swabs taken from 66 specimens representing 45 species, collected by fish bioassessment experts during electrofishing campaigns in autumn 2020 at five sites across Germany spanning the Rhine and Danube catchments. Each fish was handled briefly outside the water while a sterile swab was drawn along its flank, and samples were stored cold until DNA extraction at the University of Duisburg-Essen. Extraction yielded an average DNA concentration of 5.73 nanograms per microlitre, with individual samples ranging from 0.05 to 26.3 nanograms per microlitre. According to the FishBase database, 123 freshwater fish species are reported from Germany, and the researchers manually added the round goby and rainbow trout, two non-native species absent from that list. Their 45-species mock community therefore represents about 36.6 percent of the fish fauna reported from Germany, including 50 percent of native, 26 percent of introduced, 22.2 percent of questionable and 8 percent of not-established species.

Two versions of the mock community were prepared. The first, MC1, was normalized to an equimolar 2 nanograms of DNA per species, while the second, MC2, pooled 1 microlitre of each extract, creating a community with unequal DNA concentrations per species. Both were amplified with five primer pairs targeting three mitochondrial markers: tele02 and MiFish-U on 12S ribosomal RNA, 12SV5 on 12S, SeaDNA-mid on the cytochrome c oxidase subunit I gene, and L2513/H2714, here called LH16S, on 16S ribosomal RNA. A two-step PCR approach tagged the amplicons with barcodes and Illumina sequencing adapters, and the pooled library was sequenced on a MiSeq 250 bp paired-end platform. In total, 60 first-step PCR reactions were run, including five replicates per mock community and two negative controls per primer pair.

Sequencing produced 8,254,293 raw reads, of which 7,745,593 quality-filtered reads clustered into 140 OTUs for tele02, 105 for MiFish-U, 120 for 12SV5, 111 for SeaDNA-mid and 142 for LH16S. Negative controls contained 4,379 reads in total, almost all of them, 4,338, in the 12SV5 reactions, and these were subtracted before analysis. Nearly all primer pairs amplified little non-fish DNA, with fish accounting for 96 to 98 percent of OTUs, but SeaDNA-mid stood out: half of its OTUs were non-fish taxa, including annelids, arthropods, diatoms, green algae, oomycetes and rotifers. In read proportions, however, non-target OTUs stayed marginal for every primer, at 98.2 to 100 percent fish reads. Species-level assignments supported by unambiguous taxonomy were highest for MiFish-U and tele02, each at 60 percent of species-level OTUs, followed by SeaDNA-mid at 49 percent, 12SV5 at 46 percent and LH16S at 41 percent.

After removing bycatch taxa and curating ambiguous assignments, the number of species detected per primer ranged from 36 for SeaDNA-mid to 45 for 12SV5, but raw detection counts told only part of the story. Measured against the known mock community composition, tele02 achieved the highest congruence, with 37 true-positive species, 2 false positives and 8 false negatives. MiFish-U followed with 35 true positives, 2 false positives and 10 false negatives, and SeaDNA-mid recorded 33 true positives, 3 false positives and 12 false negatives. The two broad vertebrate primers fared far worse: 12SV5 yielded 27 true positives against 18 false positives and 18 false negatives, while LH16S detected 23 true positives with 17 false positives and 22 false negatives. Only the 12SV5 and LH16S datasets contained marine fish taxa that were never part of the mock community, including surgeonfishes, sea chubs, cusk-eels, armoured sea robins, gunnels and eelpouts. Because no marine samples had been processed in the laboratory, cross-contamination could be ruled out; the researchers attribute these phantom detections to the conserved genomic regions those primers target, which sacrifice taxonomic resolution for breadth.

Primer bias also left a measurable imprint on read proportions. Although all five markers are mitochondrial and thus present at equal copy numbers, several species showed dramatically overrepresented reads under one primer: perch reached 12.65 percent of reads with 12SV5 against an average of 1.04 percent with the other primers, Danube salmon hit 17.36 percent with SeaDNA-mid versus 2.5 percent elsewhere, and nine-spined stickleback reached 17.49 percent with SeaDNA-mid against 0.65 percent. These differences point to variable primer binding efficiencies and caution against treating read abundance as a straightforward proxy for fish biomass. Reproducibility, by contrast, was excellent across the board: mean Jaccard similarity among five technical replicates ranged from 0.95 for MiFish-U to 0.99 for 12SV5. Input DNA concentration showed no correlation with read numbers for any primer, with Pearson coefficients between 0.12 and 0.16 and p-values at or above 0.05, yet read counts per species were significantly correlated between the equimolar and unnormalized mock communities, with coefficients from 0.68 to 0.81.

Not every failure traced back to the primers themselves. Six mock species, the bullhead Cottus gobio, striped ruffe, river lamprey, Danube roach, European mudminnow and zingel, were detected by no primer pair at all. For Cottus gobio, a common Central European species with 34 reference sequences across all three markers in the Midori2 v249 database, the researchers suspect the swab sample itself, since the species is routinely recovered in other eDNA datasets. The striped ruffe lacks any 12S reference sequence, preventing species-level assignment with the 12S primers, and the zingel’s absence despite available references most likely reflects sampling or laboratory error. The two lamprey ecotypes Lampetra fluviatilis and L. planeri, meanwhile, share mitochondrial haplotypes and cannot be distinguished with the short fragments eDNA metabarcoding employs, a limitation the authors expect to persist for the foreseeable future. The Danube roach case hinged on missing 12S references in v249, which the newer v255 release partially closes, while the European mudminnow, listed as vulnerable on the IUCN Red List in 2010, was confounded with the invasive Eastern mudminnow, raising the possibility that the swabbed specimen itself was misidentified.

Oversplitting, in which a single species fragments into multiple OTUs, added a further caveat. Forty-eight cases were recorded overall, with tele02 accounting for the most at 12, and the highest ratios reaching seven OTUs per species for ruffe under tele02 and tench under 12SV5. The authors note that this inflates alpha diversity estimates but can be mitigated with post-clustering curation tools such as LULU filtering, and it does not affect species-level analyses where OTUs sharing an assignment are merged. On balance, the study concludes that tele02 is arguably the best currently available primer pair for eDNA metabarcoding of European freshwater fish, combining the highest detection ability with strong reproducibility and minimal false detections, with MiFish-U close behind. SeaDNA-mid performed comparably on true positives but its heavy co-amplification of non-fish DNA could demand deeper, costlier sequencing on real environmental samples. The broader lesson for programmes such as monitoring under the European Water Framework Directive is twofold: primer choice fundamentally shapes eDNA results, and closing gaps in curated reference databases remains essential to eliminate avoidable false negatives.

Subject of Research: Evaluation of five eDNA metabarcoding primer pairs using mock communities of Central European freshwater fish

Article Title: Evaluating five primer pairs for environmental DNA metabarcoding of Central European fish species based on mock communities

Article References: Macher, T.-H., Schütz, R., Yildiz, A., Beermann, A. J., & Leese, F. (2023). Evaluating five primer pairs for environmental DNA metabarcoding of Central European fish species based on mock communities. Metabarcoding and Metagenomics, 7, Article e103856. https://doi.org/10.3897/mbmg.7.103856

Image Credits: AI Generated

DOI: 10.3897/mbmg.7.103856

Keywords: environmental DNA, eDNA metabarcoding, fish monitoring, primer pairs, mock communities, tele02, MiFish-U, reference databases, freshwater fish, Central Europe, biodiversity monitoring, 12S rRNA