Plankton trophic interactions play a crucial role in ecosystem processes. Diet analysis using high-throughput sequencing methods such as metabarcoding can provide new insight where traditional methods have been limited. We used 16S ribosomal RNA genemetabarcoding to identify phytoplankton fromthe guts of the copepod Pseudodiaptomus forbesi and in seston from the Cache Slough Complex, a tidal freshwater reach of the San Francisco Estuary. Cyanobacteria, assumed to have low nutritional value for copepods, were detected in all copepod samples and comprised the highest relative read abundance in metabarcoding results. Di erential abundance analysis, used to compare representation of operational taxonomic units between copepod and seston samples, showed that two filamentous taxa (a streptophyte and a cyanobacterium) were most overrepresented in copepod samples, whereas cryptophytes and most ochropytes (diatoms and related taxa) were underrepresented in copepod samples. These findings could reflect unexpected feeding patterns or trophic upgrading. Understanding the capabilities and limitations of DNA metabarcoding is key to its use in diet analysis and integration with traditional approaches.

Copepods and seston (available prey) were collected by boat during eight sampling events at five locations (Fig. 1) in June and October 2015, a drought year, as part of a broader study (Kimmerer et al., 2018). Of 23 sampling events throughout the CSC (not shown), eight sampling events had sufficient P. forbesi adults for diet analysis. Groups of 5 copepods per sampling event were processed whole because their small size precluded dissecting guts intact. Seston was collected by Van Dorn sampler from ~1 m depth and prefiltered through a sterile 150 µm nylon mesh sieve to remove larger pieces of detritus and minimize filter clogging. Samples were filtered using syringe or vacuum filtration (<100 mm Hg) onto a 1 µm pore, 47mm diameter polycarbonate filter (Whatman) which was transferred to a sterile tube using sterile forceps (n =2 to 3 filters per sampling event). Copepods maintained without food were used to as negative controls to determine if non-ingested organisms were detected, and laboratory-fed copepods were used to confirm successful detection of known prey. 

DNA was extracted from copepods and whole filters using E.Z.N.A. Tissue DNA Kit (Omega Biotek) and amplified using16S rRNA gene V3-V4 region primers (5'-CCTACGGGNGGCWGCAG-3' and 5'-GACTACHVGGGTATCTAATCC-3'). Library preparation followed Illumina guidelines. Libraries were sequenced on two Illumina MiSeq runs at loading concentrations from 6 to 12 pM using 2 x 250 V2 or 2 x 300 V3 chemistry with PhiX genomic library loaded at 20%. Samples were demultiplexed using Illumina MiSeq Control Software v2.2 MiSeq Reporter v2.6 and analyzed in MacQIIME v1.9.1 (Caporaso et al., 2010). Quality-filtered sequences (Phred score at least 20) were aligned in MacQIIME using PyNAST (Caporaso et al., 2010) and clustered into Operational Taxonomic Units (OTUs) at 97% genetic similarity using UCLUST (Edgar, 2010). OTUs with fewer than 10 total reads were removed from the data. Taxonomy was assigned using Greengenes v13_8 (DeSantis et al., 2006) and PhytoREF (Decelle et al., 2015). 

Pilot data from 18S rRNA gene sequencing from a subset of samples is included in this dataset. Sequencing and analysis followed 16S rDNA methods except library preparation followed Stoeck et al. 2010 and OTUs were identified using the SILVA reference database v. 11https://doi.org/10.1093/bioinformatics/btq4611 (Pruesse et al., 2007; Quast et al., 2013).