Riparian avian insectivores not only depend on terrestrial insect prey but also benefit from the inclusion of aquatic prey during critical life history periods. Diets identified herein show that Prothonotary Warbler (Protonotaria citrea) nestlings were provisioned with aquatic prey throughout the breeding season across their range, but with variation in prey frequency of occurrence and taxonomy. Anthropogenic activity and climate change may impact the trophic link especially between aquatic and riparian habitats by altering the presence, abundance, and timing of prey availability. Thus, we used DNA metabarcoding of fecal samples to quantify the frequency of occurrence of nestling diet items at nine sites across their breeding range that differed in expected aquatic prey consumption. We analyzed spatial and temporal differences in the occurrence and multivariate diet assemblages of each prey source. Lepidoptera was the predominant terrestrial prey occurring in diets across space and time, whereas emergent aquatic insects and freshwater mollusks in aquatic diet exhibited greater variation. The frequency of emergent aquatic prey occurrence in nestling diets ranged from 61-100% across sites and was greater for early-season nestlings. The seasonal decrease in aquatic prey consumption indicates a potential temporal shift in the nutritional landscape from aquatic to terrestrial prey sources and a possible nutritional phenological mismatch for early nestlings as climate change advances the timing of insect emergence. Our findings also suggest that Prothonotary Warblers respond to environmental variability by consuming alternative prey and argue for future research investigating the extent to which shifting diets have nutritional consequences for riparian nestlings.

During the 2018 and 2019 breeding seasons, we sampled 9 Prothonotary Warbler populations at sites monitored annually throughout their breeding range. Local nest monitors collected fecal sacs from nestlings aged 6-11 days during routine nest visits. Nestlings readily defecate when handled during banding, enabling collection of fresh fecal samples. We stored fecal sacs in glass vials with 96% ethanol at room temperature to preserve DNA until extraction. 

We extracted DNA from samples using the Qiagen Fast DNA Mini Stool kit, modifying the manufacturer’s instructions to increase yield of degraded DNA (Trevelline et al. 2016). A two-stage targeted amplicon sequencing protocol was used to amplify prey DNA from our DNA extractions (Naqib et al. 2018). The first PCR stage amplified sequences with the primers LCO1490 and CO1-CFMRa (hereafter, ANML; Jusino et al. 2019). The second PCR stage added unique 10-bp barcodes to each sample, allowing samples to be pooled during sequencing while retaining sample information for downstream analysis. Sequencing was performed using a combination of Illumina MiniSeq and Illumina MiSeq (2x150 paired-end reads). Additional details regarding laboratory methods are provided in the main article.

We completed all steps for trimming and quality filtering to obtain amplicon sequence variants (ASVs) from demultiplexed sequences using the dada2 package (1.18; Callahan et al. 2016) in R (4.0.3; R Core Team 2020). Sequences were trimmed at the 5′ end to remove primers and at the 3′ end to remove lower quality segments (Phred Scores <30) as much as possible while also retaining a 12-bp overlap between forward and reverse sequences for merging paired sequences (Callahan et al. 2016).

We retrieved matching reference sequences for representative ASVs and taxonomy from NCBI BLAST (National Center for Biotechnology Information Basic Local Alignment Search Tool; Sayers et al. 2022) using Biopython (1.79; Cock et al. 2009), then performed sequence identification consensus with the statistics provided by BLAST. Consensus details are described in the main article. 

We attempted to account for read variation by rarefying our dataset using the phyloseq R package (rarefy_even_depth; McMurdie and Holmes 2013). We performed separate rarefaction simulations for sequence depths between 100-2,000 reads/sample, at 100-read intervals. The rarefied sequence depth of 1,000 reads/sample was chosen as the best tradeoff between an increase in diet identifications and a decrease in the number of samples meeting the required depth. Above 1000 reads, we did not detect new diet taxa at the levels of analysis (i.e., taxonomic order or family). All downstream data analyses were performed with the rarefied dataset.

Most data processing and analyses were performed in R (4.0.3; R Core Team 2020). R scripts are available from the corresponding author upon reasonable request.

Biopython was used to perform the NCBI BLAST and sequences identification consensus. Python scripts are available from the corresponding author upon reasonable request.