Estimating species richness using environmental DNA

Abstract

The foundation for any ecological study and for the effective management of biodiversity in natural systems requires knowing what species are present in an ecosystem. We assessed fish communities in a stream using two methods, depletion-based electrofishing and environmental DNA metabarcoding (eDNA) from water samples, to test the hypothesis that eDNA provides an alternative means of determining species richness and species identities for a natural ecosystem. In a northern Indiana stream, electrofishing yielded a direct estimate of 12 species and a mean estimated richness (Chao II estimator) of 16.6 species with a 95% confidence interval from 12.8 to 42.2. eDNA sampling detected an additional four species, congruent with the mean Chao II estimate from electrofishing. This increased detection rate for fish species between methods suggests that eDNA sampling can enhance estimation of fish fauna in flowing waters while having minimal sampling impacts on fish and their habitat. Modern genetic approaches therefore have the potential to transform our ability to build a more complete list of species for ecological investigations and inform management of aquatic ecosystems.

Keywords: Chao estimator; electrofishing; freshwater community; metabarcoding; species identity.

Figure 1

Sample of fish species inhabiting Juday Creek (clockwise from top left; Rock bass, Creek chub, Brown trout, Steelhead, Blacknose dace, Mottled sculpin, White sucker, and Green sunfish).

Figure 2

Flowchart describing bioinformatics steps taken to analyze MiSeq data. (A) Quality filtering is the process of removing low‐quality reads from further analysis. (B) Demultiplexing is the process of separating out individual samples that were run together on the same MiSeq run. (C) Merging is the process of combining overlapping reads into a single read. (D) Additional quality filtering step. (E) Dereplication is the process of eliminating duplicate reads. (F) Clustering is the process of grouping similar unique reads into OTUs. (G) HMMer is described in the manuscript. (H) SAP is described in the manuscript.

Figure 3

Decision‐making flowchart of species assignment utilizing OTU sequences inputted into SAP and USEARCH. For example, in the case of path 1 (P1): if SAP provides a species‐level assignment with posterior probability >=95%, USEARCH has a global alignment with identity >=97%, and the species assignments from the two approaches are identical, we use the species assignment. If the two assignments from each program are different (P2), we manually check the assignment against GenBank reference and alignment and make the decision as to the appropriate species assignment.

Figure 4

The number of PhiX control sequences that were incorrectly assigned to a sample as a function of the number of reads assigned to the sample. PhiX is an Illumina control library that is spiked into the pooled libraries prior to loading on the MiSeq. There are no laboratory steps or possibilities for contamination during preparation; therefore, any misassignment of this control is due to incorrect demultiplexing or index assignment due to cross‐contamination between clusters on the flow cell. This is therefore an instrument artifact and not likely due to any laboratory contamination, but demonstrates the plausibility that a low number of reads can be misassigned to any sample.

Figure 5

Species richness for Juday Creek from 1997 to 2013 for electrofishing for all four reaches combined, including species captured (black dots) and bias‐corrected Chao II species richness estimates (open circles) with 95% confidence intervals (vertical bars).

Figure 6

Species richness rarefaction curves (incidence based) for Juday Creek in 2013 for electroshocking (circles) and eDNA metabarcoding (triangles) samples. Vertical bars are 95% confidence intervals. Dashed lines represent observed values for the two methods. Absence of confidence intervals indicates that the estimated species richness is the same as the observed species richness.