DNA metabarcoding is a molecular technique frequently used to characterize diet composition of insectivorous birds. However, results are sensitive to methodological decisions made during sample processing, with primer selection being one of the most critical. The most frequently used DNA metabarcoding primer set for avian insectivores is ZBJ. However, recent studies have found that ZBJ produces significant biases in prey classification that likely influence our understanding of foraging ecology. A new primer set, ANML, has shown promise for characterizing insectivorous bat diets with fewer taxonomic biases than ZBJ, but ANML is not yet widely used to study insectivorous birds. Here, we evaluate the ANML primer set for use in metabarcoding of avian insectivore diets through comparison with the more commonly used ZBJ primer set. Fecal samples were collected from both adult and nestling Purple Martins (Progne subis subis) at two sites in the USA and one site in Canada to maximize variation in diet composition and to determine if primer selection impacts our understanding of diet variation among sites. In total, we detected 71 arthropod prey species, 39 families, and 10 orders. Of these, 40 species were uniquely detected by ANML, whereas only 11 were uniquely detected by ZBJ. We were able to classify 54.8% of exact sequence variants from ANML libraries to species compared to 33.3% from ZBJ libraries. We found that ANML outperformed ZBJ for PCR efficacy, taxonomic coverage, and specificity of classification, but that using both primer sets together produced the most comprehensive characterizations of diet composition. Significant variation in both alpha- and beta-diversity between sites was found using each primer set separately and in combination. To our knowledge, this is the first published metabarcoding study to directly compare avian diet characterizations produced with both ANML and ZBJ primer sets.

Data represent DNA metabarcoding data produced with two COI primer sets (ZBJ and ANML) applied to DNA extracts from purple martin (Progne subis subis) fecal samples. Whole fecal samples were collected opportunistically from nestling Purple Martins during nest checks and banding activities in Orlando (FL, USA n = 68), Erie (PA, USA n = 19), and Winnipeg (Manitoba, Canada, n = 36) during the 2019 breeding season.

Whole fecal samples or sub-samples of feces collected with sterile swabs were collected from adults at the Orlando site only (n = 11) during this same time period. Genomic DNA was extracted from a homogenized subsample of each fecal sample using the DNeasy PowerSoil kit.

Library preparation consisted of (a) amplicon PCR (50 cycles for ZBJ, 40 cycles for ANML), and (b) indexing PCR (8 cycles) using the Illumina XT kit. PCR reactions were cleaned up with 1.1X Sera-Mag SpeedBeads, and PCR2 products (i.e., libraries) were quantified spectrophotometrically before equimolar pooling. The final library pooled was QC checked by TapeStation and quantified by qPCR.

The library pool was sequenced using a paired-end approach (2x300bp) with v3 sequencing reagents (Illumina, MS-102-3003) following manufacturer’s instructions, with the addition of a final 2-minute denaturing step at 96°C. The library pool was spiked with 20% PhiX (Illumina, FC-110-3001) and loaded onto the flow cell at 15 nM.

DNA sequencing reads were demultiplexed on the MiSeq instrument into separate FastQ files for each sample amplified with either ZBJ or ANML barcoding primer sets. DNA sequencing data are published in raw form as fastq files; no read processing has been performed on these files.