Experimental tissue mixture samples
Across the portfolio of four top-performing markers, sequencing read
depth did not provide a consistent quantitative proxy for the relative
amount of tissue added to the fishmeal mixture. While there was a
tendency for species with lower input amounts to receive fewer reads,
the proportion of reads varied substantially among species that were
added in the same amounts (Fig. 4, Fig. 6A). The mean proportion of
reads for the five taxa added in the same amount showed a positive
relationship for tissue inputs <1%, with the lowest tissue
input (0.01%) receiving an order-of-magnitude smaller proportion of
reads across all four markers (1.7 x 10-4 of reads)
than the next lowest tissue input (0.1%; 5.3 x 10-3of reads; SI, Fig. S6). However, the relationship between read
proportion and tissue input was inconsistent because the 1.91% tissue
input level received, on average, a smaller proportion of reads than
taxa that comprised 1.0% of the tissue input (5.6 x
10-3 and 2.7 x 10-2 of reads,
respectively) and for three of the four markers, the highest tissue
input level (13.32% of fishmeal) received a smaller proportion of reads
than taxa comprising 3.65% of the fishmeal mixture (Fig. 4 and SI, Fig.
S6).
The ratio of fishmeal-to-filler in the experimental feeds had little
effect on the proportion of reads per taxon (Fig. 4). Taxa that
comprised 0.01% and 0.1% of the fishmeal showed more variation than
taxa present in higher proportions, with inconsistent read depth across
fishmeal-to-filler ratios (2%, 10% or 25% fishmeal) and across the
four markers (Fig. 4). Although read depths were inconsistent in terms
of presence/absence detection, fewer of the low-input taxa were
recovered, and this pattern of taxon drop-out increased as the ratio of
fishmeal-to-filler decreased (SI, Fig. S7).
Matrix composition (filler) of experimental feeds did not impact the
recovery of reference taxa or the proportion of reads attributed to
those taxa for constituents that make up >1% of fishmeal.
Patterns in proportion of reads were dominated by variation among taxa
rather than between filler types, and the proportion of reads did not
change between the soy filler and the animal/plant filler for taxa added
at >1% fishmeal (Fig. 5). However, low-input taxa, those
that comprised 0.01% or 0.1% of fishmeal, displayed more variable read
depth between fillers, but not in a consistent way within or across
markers. At the lowest tissue input (0.01%), two markers (12S; MiFish
and one COI primer set; nsCOIFo) consistently showed a higher proportion
of taxon drop-out (SI, Fig. S7).
Although two feeds consisted entirely of filler with no added fishmeal,
sequencing results showed that 80% of fishmeal taxa were present in
these filler-only feeds, indicating that simultaneous preparation of
filler-only feeds and fishmeal feeds resulted in contamination that was
sufficient to be detected by our PCR assay (SI, Fig. S8). Sequencing
read counts for filler 1 (soy flour) were lower than for filler 2, which
contained bloodmeal (from Sus sp. ) and feathermeal (fromGallus sp. ). Both taxa were detected by all four of the markers
designed to amplify fishes, and in the absence of added fishmeal tissue,Sus sp. and Gallus sp. received an average of
~25% of the sequencing reads per replicate, whereas in
the experimental feeds that contained 75% filler 2 and 25% fishmeal,Sus sp. and Gallus sp. received only ~1%
of sequencing reads, highlighting the influence of template competition
during library preparation.
Comparing taxon recovery from the 100% fishmeal and pooled DNA extracts
(MV) showed that reference taxa were detected (presence/absence) equally
well regardless of relative representation when tissue input/DNA
concentration was >1% (Fig. 6A, 6B). For input categories
below 1% (0.01% and 0.1%), more taxa were identified from the 100%
fishmeal sample than from the MV (that contained DNA extracted
separately from each experimental feed fish constituent and pooled in
proportions equivalent to relative tissue inputs to the feeds). In
comparison, when the DNA extracts from each experimental feed
constituent was pooled in equal concentration (the ME pool), we
recovered all taxa with at least one locus with just one exception
(Ophisternon sp. ), indicating that taxon drop-out in the MV pool
was not a result of primer mismatch but rather a result of low input
concentration (all species except for Loliolus sp. were
consistently identified at concentrations above 2.24 x
10-4 ng/ul; Fig. 6B, 6C).