Materials and method

Environmental conditions

Our experimental design is summarized in Table 1 and encompassed eight treatments A-H. We filled each of 28 aquariums with 3.7 L water obtained from a natural water system in The Netherlands (GPS: N 52 10.056, E 4 28.086). We kept the aquariums under controlled conditions in the laboratory facility of Naturalis Biodiversity Center (Leiden, The Netherlands). There was no gravel or substrate inside and the aquariums were not aerated. The aquariums were placed on a laboratory bench and the treatments were equally distributed over the space. We varied the pH in the aquariums to either ‘high’ (above 8) or ‘low’ (below 5.7) and the amount of OM to either 10 grams added, or none at all, resulting in four treatments (see Table 1). OM content of the water was increased in the following way: five grams of decaying leaf material of locally growing plane trees (Platanus hispanica ) and five grams of leaf material of locally growing European beech (Fagus sylvatica ) was added to the water after sterilizing the leaves for 1 hour at 120 ºC to prevent introduction of microorganisms that degrade eDNA. To lower the pH we acidified the water using 3.7% HCl to a pH of 5. During the experiments we monitored the pH of the water (17 measurements, Supplemental Table 1) and we added additional HCl if the pH exceeded 5.7. We refilled the aquariums to the original level, each time that samples were collected for eDNA sampling of pH monitoring. The water in the aquariums was kept at room temperature.

Inoculation of living shrimps

Prior to inoculation with DNA sources, we took samples from all 28 aquariums to estimate the level of background DNA of Gammarus pulex present. Four aquariums were not inoculated with any DNA source and served as control.
In twelve aquariums, we added eight live shrimps (G. pulex ) in the final stages of their development. Last instars were chosen to avoid differences in molting and propagation between the aquariums. All individuals used in this study were collected in Wageningen, the Netherlands, from a single population in the wild (GPS: N 51 58.500, E 5 38.820). We removed dead shrimps and replaced them with live ones, and we also removed newborn shrimps (for details see supplemental material: STable2).

Spiking DNA

On the date that we removed the shrimps from the aquariums, we spiked another twelve aquariums with 4.99 μg tissue-derived extracellular genomic DNA of G. pulex . We measured DNA degradation in these aquariums from two hours after spiking, measuring every 60 minutes. The DNA used for spiking was extracted from tissue of G. pulex using the Qiagen DNeasy Blood & Tissue following the Spin-column protocol. We measured DNA concentration in the extracts using a Qubit 2.0 fluorometer (Life Technologies).

Real-time quantitative PCR

eDNA degradation was monitored using a CFX96TMReal-Time PCR System. We developed a species-specific qPCR primer set using Geneious (PulexF1: ACGTAGACCTGGTATATCTATAGACC & PulexR1 CCGGCTAAAACAGGTAAGGA) to amplify a 98bp fragment of COI; we developed another primer set using primer-BLAST of NCBI ((Ye et al., 2012) (PulexF2: GGAGCTTGGGCTAGTGTTGT and PulexR2: CGTGAGCGGTGACTAATGACG) to amplify an 118bp fragment of COI. Both primer combinations worked well, but we selected primer pair PulexF1 & PulexR1 to do the experiment. We checked the specificity of both primers in silico using primer-BLAST (2013/02/28) with the setting that unintended targets should have at least two mismatches within the last five base pairs at the 3’ end for one of the primers. Primer-BLAST only showed hits of indigenous organisms except for Gammarus duebeni . However, in the case of the primer pair PulexF1 & PulexR1 a total of seven mismatches was found. Furthermore, G. duebeni does not occur in the region and occurs in a habitat type different from that at the location where we obtained aquarium water.
eDNA extraction For extracting eDNA, we added 15 mL of water samples to 1.5 mL of 3M sodium acetate and 33mL absolute ethanol and stored it at -20 ºC (following Ficetola et al., (2008)). We centrifuged the mixture (9400g, 35 min, 6ºC) and discarded the supernatant. To extract DNA from the pellets, we used the Qiagen DNeasy Blood & Tissue kit (spin-column protocol) after Thomsen et al., (2012a); Thomsen et al.,(2012b). Quantitative real-time PCR (qPCR) was performed in a total volume of 20 µL using 10 µL GoTaq PCR Master Mix 2X (Promega), 0.4 µL of both primers, 5.2 µL nuclease-free water and 4 µL template. We performed PCRs in 96-well plates and included in each plate at least one negative and one positive PCR control reaction (both in triplicate).
eDNA samplingWe sampled eDNA in the aquariums 28 days after they had been inoculated with live shrimps to estimate the amount of eDNA that had been accumulated. Thereafter, the shrimps were removed. To estimate the survival of eDNA, samples were collected after 12, 24, 36, 48, 60, 72, 96, 168, 288, 504, 1008 and 1680 hours. We stopped sampling when the average Ct-value of a sample exceeded 47 (see below).

Avoiding false positives

In this study, we took several measures to avoid false positives (i.e. detecting eDNA when no animals were around). For detection of invertebrates in field samples, the use of specific-binding probes is paramount for reliably detecting target organisms. Even when the concentration is extremely low, this approach can result in more sensitive and specific detection of target DNA (Schultz and Lanze, 2015; Goldberg et al., 2016). However, because the concentration of eDNA in our controlled aquariums was relatively high we were able to use a less sensitive, low-cost approach including GoTaq qPCR 2X Master Mix in a real-time quantitative PCR assay, which contained BRYT Green, a fluorescent dye that binds to double-stranded DNA. Since BRYT Green dye binds to all double-stranded DNA, the presence of double stranded non-target DNA, such as primer dimers, can also result in a fluorescent signal. Ct-values were converted to numbers of molecules based on the principle that 2[CtStandard –CtSample] is the fold difference in concentration of sample and standard used. Standards (i.e. series of increasing known concentrations) were made for each PCR plate and resulting Ct-values plotted against the 10log(number of molecules). Linear regression analysis of the average across plates then enabled calibrating the standards and calculating numbers of molecules in the aquarium samples. The detection limit was thereafter determined based on sample concentrations collected from control aquariums, and the from all other aquariums prior to inoculation (see also supplementary figure S1).
Defining the amount of detectable eDNAEach aquarium was sampled twice at each sampling time. Three water samples were collected from the aquariums with live shrimps just before they were removed from the aquariums, to be able to accurately determine the accumulation of eDNA in de aquariums. In 12 samples the DNA pellet did not form properly during extraction, in which cases only one sample was analyzed.

Quantifying qPCR inhibition

We quantified the amount of PCR inhibition in the samples (N=36) that were collected from the aquariums containing shrimps just before they were removed from aquariums. We did this by performing an inhibition qPCR test (see details below). We repeated this in the samples obtained from the spiked aquariums just after spiking (N=23) and in the samples collected from the control aquariums that were obtained at the same time (N=8). The qPCR reactions were spiked with an artificial fragment of DNA (CGGAGGTGCACTTACAGATAGAGTCACATGTCGTGTCTAACGCGCAGCAGTAGTGTCTGAACACGAGTCCTTCC) cloned into an pUC57 plasmid. The primers ART3-F (CGGAGGTGCACTTACAGATAGAG) and ART3-R (GGAAGGACTCGTGTTCAGACA) were used to amplify the fragment. For each sample, three qPCR reactions were performed containing 33, 333 and 3333 molecules of the artificial DNA fragment.
We performed the inhibition qPCRs in a total volume of 20 µL using 10 µL GoTaq PCR Master Mix 2X (Promega), 0.4 µL of both primers, 4.2 µL nuclease-free water, 4 µL template (either aquarium water or nuclease-free distilled water) and 1 µL containing the artificial DNA molecules. The cycling conditions were identical to those used for detection of the shrimp DNA. A standard curve was generated using each DNA concentration in triplicate, in which nuclease-free distilled water was added instead of aquarium sample. We assessed response variable CT-values of the control, shrimp and spiked data and explanatory variable CT-values of the standard deviated from slope 1 in order to assess the presence of inhibition. Therefore, linear mixed effect models were used (see below under Controls and limit of detection ). The R2 and efficiency of the qPCR assay were calculated based on a standard containing 10, 100, 1000, 10000 target-molecules (results not shown).

Statistics

In general, best-fitting models were selected with Akaike’s Information Criterion corrected for small sample sizes (AICc, see eqn. 1):
AICc=2k −2Log(L )+(2k (k +1))/(nk −1) (eqn.1)
with Log denoting the natural logarithm, L the likelihood of the model, k the number of estimated parameters in the model, andn the sample size (Bolker 2008). The minimum AICc value indicates the best-fitting model. Model fits are evaluated with respect to the AICc-difference (ΔAICc) between the considered model and the best model. Models within the interval ΔAICc < 2 are considered equivalent (Bolker 2008). In this set of models, Ockham’s razor (parsimony criterion) was used to choose the best model, containing the smallest number of parameters. We used Fisher’s least square difference (LSD) test with Bonferroni correction from the agricolae R library (Mendiburu, 2016) to test for differences in the amount of eDNA accumulated in the aquariums with live shrimps.