Experimental tissue mixture samples
Metabarcoding is typically used to detect both rare and abundant constituents in mixtures, and most application cases (diets, larval assemblages, net tows, etc.) include species in unequal proportions along with varying amounts and types of non-target material. To evaluate detection power in this scenario, we used fishmeal mixed with non-target tissue composed of different fillers. The purpose of the non-target tissue (filler) was to test whether metabarcoding markers are negatively impacted by fillers, either because of a loss of on-target sequencing reads or because of potential PCR inhibition. Our experimental mixtures emulated aquaculture feeds; but the experimental feeds have similarities to other mixed sample types, for example, stomach contents of an omnivore where fish prey items may be mixed with other (non-fish) constituents. In our test case, we freeze-dried muscle tissue from 30 of the unvouchered fish species in the FR pool, coarsely homogenized each sample in a coffee grinder, and then freezer milled samples flash-frozen with liquid nitrogen into a powder. Powdered samples were then assigned to one of six abundance levels and added to make up either 13.33%, 3.65%, 1.91%, 1%, 0.1%, or 0.01% of the total mixture (by weight), thus spanning four orders of magnitude in relative representation from abundant-to-rare (Table 3). Each abundance level was represented by five species. This experimental design allowed us to assess how dominant and rare taxa added at discrete proportions to a heterogenous mixture relate to the proportion of sequencing reads attributed to each taxon and to compare amplification biases across multiple taxa added in the same amount to the fishmeal.
To test how nontarget material or mixture matrix could affect metabarcoding efficiency and therefore, species recovery, the multi-species fishmeal mixtures were combined with two unique filler types to make seven individual experimental feeds with low (2%), medium (10%), and high (25%) ratios of fishmeal-to-filler (Table 3). Fillers for experimental feeds included plant-derived materials – grain and grass flours – and animal byproducts – bloodmeal and feathermeal – to emulate mixture constituents used in fish production (i.e., aquaculture feeds), but are also representative of non-fish diet components. Fishmeal proportions also mimicked potential levels of fish tissue added to aquaculture feeds, from low (0%-2%) to high (25%) proportions of fish in the feed mixture. By multiplying the ratio of fishmeal-to-filler by the fishmeal tissue in the experimental feed, we could test the detection threshold for individual taxa in the overall tissue mixture down to 0.0002% of the experimental feed by mass (i.e., the smallest tissue input – 0.01% of fishmeal – at 2% fishmeal-to-filler).