Fig 1.
Scheme of steps for an eDNA protocol with tested methods for each step.
Cost and processing times (in minutes) of each method are shown next to the method name. Number of samples for the measured times varies as the number of samples that can be run in parallel varies between steps. Costs are estimated per sample.
Table 1.
List of DNA oligonucleotides used in this study.
Fig 2.
Relationship between total DNA yield and filtration and extraction protocols.
(A) No crossing between lines indicates that on average, DNA yield ranking for the filters is independent of the extraction protocol. Error bars represent the 95% confidence intervals. The combination of Whatman filter and NaOH extraction wasn’t able to amplify the target Chinook DNA. (B) No crossing between lines indicates that on average, DNA yield ranking for the extraction method is independent of the filter type. The NaOH extraction protocol is the only case where the ranking order is not maintained and can be explained by the added effects of carry-on inhibitors.
Fig 3.
Distribution of DNA capture ratio for each filter type from the Automatic Differentiation Variational Inference model.
The broadness of the curve shows the variability of the ratio of the input DNA that binds to the filter. The peak of each distribution is the mean yield ratio of DNA recovery for that filter type. The nitrocellulose filter yielded the highest recovery ratio with little efficiency overlap compared to glass and Whatman filters.
Fig 4.
Percentiles and median filtration time in order to filter 1 L of estuarine water for each filtration method.
The glass filter outperforms nitrocellulose and Whatman by a significant margin in terms of average filtering time and consistency in the filtering time. Dots are filtration events while the black line represents the median value filtering time. Boxes indicate 10% quantiles.
Fig 5.
Modelled distribution of percentage yield for each extraction protocol.
The width of each of the curves shows the variability in modelled yield. The peak of the distribution is the mean yield per extraction type. Qiagen DNeasy yields the best yield with little overlap with other methods. Meanwhile magnetic beads and NaOH have shown similar distributions with significant overlap, while both dipstick methods underperform the other methods. It is important to note that this plot does not take into account PCR inhibitor carryover, which might vary significantly between methods.
Fig 6.
Influence of each eDNA protocol segment to total eDNA yield ratio estimate.
The extraction method was the factor that had the highest influence on the total eDNA yield of the total protocol, while inhibitors didn’t have a significant impact compared to the other components of the total protocol.
Fig 7.
Effects of adding a secondary inhibitor removal step to the eDNA estimation protocol.
(A-B) DNA yield variation from using a OneStep PCR Inhibitor Removal©. The nitrocellulose filter produced the highest inhibitor carryover levels at the same time it captured the highest percentage of free eDNA. This suggested that the nitrocellulose filter captured particulates indiscriminately with high efficiency (C-D) Estimated distributions for inhibitor carryover for filter and extraction method. Aside from NaOH extraction, other methods had similar distributions of carryover PCR inhibitors with high overlap. Therefore, NaOH extraction, even if it has an elevated eDNA yield, doesn’t properly address the high levels of PCR inhibitors commonly encountered in environmental samples.
Fig 8.
Decision tree for choosing the protocol which will yield the most information given research constraints.
A glass filter is recommended in most cases, as long as the focus isn’t maximizing DNA yield with no time or cost constraints. Magnetic beads also are advised in general for its balance between DNA yield and time to process the samples, while cost can be mitigated by producing magnetic beads solution in-house (~$0.55/mL) instead of buying Ampure XP ($15–$70/mL) [31].
Fig 9.
Comparison between eDNA protocols for DNA yield, cost and time to process 96 samples.
Methods were sorted by yield and shown in log10 scale.