Using environmental DNA to census marine fishes in a large mesocosm

Abstract

The ocean is a soup of its resident species' genetic material, cast off in the forms of metabolic waste, shed skin cells, or damaged tissue. Sampling this environmental DNA (eDNA) is a potentially powerful means of assessing whole biological communities, a significant advance over the manual methods of environmental sampling that have historically dominated marine ecology and related fields. Here, we estimate the vertebrate fauna in a 4.5-million-liter mesocosm aquarium tank at the Monterey Bay Aquarium of known species composition by sequencing the eDNA from its constituent seawater. We find that it is generally possible to detect mitochondrial DNA of bony fishes sufficient to identify organisms to taxonomic family- or genus-level using a 106 bp fragment of the 12S ribosomal gene. Within bony fishes, we observe a low false-negative detection rate, although we did not detect the cartilaginous fishes or sea turtles present with this fragment. We find that the rank abundance of recovered eDNA sequences correlates with the abundance of corresponding species' biomass in the mesocosm, but the data in hand do not allow us to develop a quantitative relationship between biomass and eDNA abundance. Finally, we find a low false-positive rate for detection of exogenous eDNA, and we were able to diagnose non-native species' tissue in the food used to maintain the mesocosm, underscoring the sensitivity of eDNA as a technique for community-level ecological surveys. We conclude that eDNA has substantial potential to become a core tool for environmental monitoring, but that a variety of challenges remain before reliable quantitative assessments of ecological communities in the field become possible.

Figure 1. Rarefaction curve for 10⁵ subsamples of OTU clusters recovered from the mesocosm Aquarium tank.

The best-fit curve is shown for clusters of ≥99% identity. 95% confidence interval shown as shaded area, derived from a sliding window analysis of the subsampled data, using a window size of 20,000 reads and 20% overlap between windows. The frequency distribution of assigned reads recovered is shown in the inset figure.

Figure 2. Density distributions for sampled parameter values in the mixing model.

Shaded areas represent 5×10⁴ model results constituting the 99th percentile of goodness-of-fit. Dashed red lines indicate the parameter value for the overall best-fit model described in the main text. Note that the optimal set of parameter values does not necessarily coincide with the point of maximum density for any given parameter, as the parameters are not independent of one another. Parameters in the left-hand column represent proportions of each source of DNA mixed into the sampled aquarium tank; parameters in the right-hand column represent proportions of DNA generated by each genus whose 12S mtDNA was detected within the tank.

Figure 3. Comparison of the proportion of eDNA sequences recovered to estimated species biomass in the 1-L tank sample.

(A) Relationship between the proportion of eDNA sequences and proportion of biomass in the tank (Best fit line = y = 0.0759*In(x)+0.5257) and (B) the rank abundances of these proportions for the four tank exhibit genera detected. The error bars represent the standard deviation of the three individual PCR replicates for the 1-L tank sample.

Figure 4. Relative abundances and sources of the taxa detected in the 1-L tank sample.

Abundances are based on the weighted average of the three sequencing replicates of 1-L tank samples. The size of the circles is proportional to the number of sequence reads, with the largest circle equivalent to 425,178 reads (50.6% of the tank sample). Tank (blue) refers to DNA that was generated within the tank. Feed (purple) includes both gel and pellet diets. Unknown (gray) refers to DNA that was detected in the tank but was not generated in the tank and did not appear in the intake or feed. The tree diagram is derived from NCBI taxonomic groupings, rather than from an evolutionary phylogenetic analysis.