Multi-locus DNA metabarcoding of zooplankton communities and scat reveal trophic interactions of a generalist predator

To understand the ecosystem dynamics that underpin the year-round presence of a large generalist consumer, the Bryde’s whale (Balaenoptera edeni brydei), we use a DNA metabarcoding approach and systematic zooplankton surveys to investigate seasonal and regional changes in zooplankton communities and if whale diet reflects such changes. Twenty-four zooplankton community samples were collected from three regions throughout the Hauraki Gulf, New Zealand, over two temperature regimes (warm and cool seasons), as well as 20 samples of opportunistically collected Bryde’s whale scat. Multi-locus DNA barcode libraries were constructed from 18S and COI gene fragments, representing a trade-off between identification and resolution of metazoan taxa. Zooplankton community OTU occurrence and relative read abundance showed regional and seasonal differences based on permutational analyses of variance in both DNA barcodes, with significant changes in biodiversity indices linked to season in COI only. In contrast, we did not find evidence that Bryde’s whale diet shows seasonal or regional trends, but instead indicated clear prey preferences for krill-like crustaceans, copepods, salps and ray-finned fishes independent of prey availability. The year-round presence of Bryde’s whales in the Hauraki Gulf is likely associated with the patterns of distribution and abundance of these key prey items.

Given the degraded DNA template present in the whale scat samples, we undertook several precautions. As previously mentioned, DNA was extracted in a laminar flow UV hood to minimise contamination. All PCR mixtures were prepared in a UV laminar flow hood in a separate room. DNA template was added to the PCR mixture in a separate UV flow hood.
Multiple negative controls were used per PCR and results were only used from PCRs were there were no visible bands in the negative controls. Furthermore, sample types were extracted and amplified, and amplicons cleaned, in separate experiments, to limit the chance of cross-contamination.

Selecting clustering threshold
To select the clustering threshold for use in the analyses, we assessed the change in measures of biodiversity with different clustering levels, by calculating the α, β and γ diversity indices (effective and Shannon) of Jost [3,4] using the R package vegetarian (Charney and Record 2012) for each DNA barcode. Several series of OTUs were created by clustering at iterative levels of sequence similarity, incrementing by 1% between 90% and 100% [6]. This revealed that clustering at 98-100% similarity resulted in high levels of diversity across all three DNA barcodes ( Supplementary Fig. 1). The diversity indices started to plateau when clustering was at 97%. This threshold was used for further analyses.  Single-celled eukaryotes Myzozoa Single-celled eukaryotes Centroheliozoa Single-celled eukaryotes Choanoflagellida Single-celled eukaryotes Euglenozoa Single-celled eukaryotes Ciliophora Single-celled eukaryotes Rhizaria Primarily single-celled eukaryotes   Warm  15258  3  18207  34  S06  Inner  Cool  11477  8  2075  37  S07  Inner  Cool  76565  3  14836  81  S08  Inner  Cool  92697  6  15136  29  S09  Inner  Cool  64078  4  4495  66  S10  Outer  Cool  --8630  39  S11  Outer  Cool  --4099  9  S12  Outer  Cool  --26070  34  S13  Outer  Warm  --73508  29  S14  Outer  Cool  7127  5  7763  52  S15  Outer  Cool  21132  5  17563  46  S16  Inner  Cool  227  -6235  14  S17  N/A  Warm  6022  3  13606  25  S18  Inner  Warm  46235  2  --S19  Inner  Warm  21677  11  --S20  Inner  Cool  513 ---  Both unconstrained PCO and constrained CAP analyses were conducted on the zooplankton data for both DNA barcodes, using four different transformations reflecting OTU composition and relative read abundance, using OTU identity (Jaccard and Bray Curtis distances) and genetic similarity (UniFrac distances). PCO analyses are an unconstrained ordination that was conducted to visualise the data and show any emergent patterns. CAP analyses are constrained ordination that are conducted to test specific hypotheses [7].

Collection
Matched water samples were collected, using the same method, from water adjacent to the whale scat immediately after sample collection.

Characterising composition of matched water control samples
This was done in the same way as described for the plankton and whale scat samples in the main manuscript.

Investigating the impact of environmental contamination on whale scat samples
To determine environmental contamination on the scat samples we used the dataset of matched water samples collected concurrently to a subset of the scat samples. For these paired samples, we first identified the OTUs at high abundance (arbitrary >1% or >5%) in the matched water as potential environmental contaminants in the scat sample. These OTUs were removed from the scat samples to check potential environmental contamination, giving a new sample type -'control' whale scat sample. We constructed PCO graphs to visualise the clustering of scat and matched water samples and then control scat and matched water samples.
If the control scat sample differs significantly from its paired matched water where the standard scat samples are not, this suggests that environmental contamination is a factor in the composition of the scat samples. Alternatively, similarity could exist because matched water samples were taken predominantly in areas where whales were foraging. Therefore, the matched water sample could reflect the taxa in a potential prey patch. To test this hypothesis, we constructed similarity matrices as described above, for both composition and relative abundance of taxa, for both the COI and 18S DNA barcodes. We then repeated the PERMANOVA with sample type, region and season as fixed factors for all matched water and scat samples (raw and rarefied data), then paired matched water and control whale scat samples (raw and rarefied data). Due to the multiple comparisons, we applied the Bonferroni correction to the results within each DNA barcode (α = 0.05; α/8 = 0.006).

Characterising composition of matched water samples
All matched water samples produced reads for either COI and/or 18S DNA barcodes (Stable 3-1), although two matched water samples (M22 and M23) in the 18S DNA barcode were excluded from rarefaction-based analyses due to low number of reads. Chordata and Arthropoda were the most common Phyla observed across matched water samples and DNA barcodes (SFig. 3-1). We restricted our analyses to the 1,101 COI DNA barcode OTUs and the 267 18S DNA barcode OTUs identified to Class.

No evidence for substantial environmental contamination of whale scat samples
Principal coordinate analyses showed that the scat samples did not cluster distinctly from the matched water controls for the COI and 18S DNA barcodes (Figs S3-2 and S3-3). Multifactorial PERMANOVAs, incorporating season and region, showed sample type was not a significant factor for any dataset analysed (Table S3-2).
To examine whether this lack of differentiation was due to potential environmental contaminants in the scat, we removed OTUs abundant in the matched water samples from the paired scat samples to construct 'control' whale scat samples. Results were similar when excluded OTUs represented either 1% or 5% of matched water samples, therefore, we present results at the conservative 1% threshold. The control scat and matched water samples scattered throughout the PCO visualisation, with no obvious clustering by sample type (SFig 3-5 and 3-6). PERMANOVA analyses suggested that sample type was not a significant driver of differences in sample compositions, based on relative abundance, composition or phylogenetic similarity, for either the COI or 18S DNA barcodes. Results were nonsignificant using rarefied and full datasets, and with a limited dataset comparing the paired scat and adjacent water samples (STable 3-2).