2.1 Mock eDNA preparation

We sampled 10 mg of tissue from the gills of frozen yellow perch (n = 5), P. flavescens, which were caught 2 km north of the Trent University campus in the Otonabee river (Ontario, Canada), and are a ubiquitous and common species in our studied lakes. Whole DNA was extracted using the modified Qiagen DNEasy extraction protocol specific for high molecular weight DNA extraction for fish. The extracted DNA was quantified using the Qubit fluorometer to ensure that the amount of DNA was above the 30 ng/μL threshold we required to set up the microcosms and bioassays. Extracted DNA was pooled for easier experiment setup and re-quantified before use. DNA was kept frozen at −20°C.

To generate stock eDNA solution of a desired copy number, we used the COI gene in P. flavescens with primers (forward: CAGGGGTTTCCTCAATTCTAGGT, reverse: CCAGCGGCAAGAACAGGTAGT) obtained from Hernandez et al. (2020) that designed these oligos specifically for eDNA surveys. To ensure primer binding efficiency and amplification limits for the SYBR Green quantitative PCR, a temperature gradient PCR was performed with the primer pair, ranging from 50 to 65°C. After determining the optimal annealing temperature, a dilution series of the extracted DNA- ranging from a total 100,000 copies to 10 copies—was performed to determine the detection limit and quantification limit following the methods outlined in Klymus et al. (2020). Copy number of stock DNA was calculated assuming the mitochondrial genome of yellow perch is 16,537 base pairs (Bélanger-Deschênes et al., 2013), and the number of mitochondria per cell in the gill tissue was assumed to be 500 (Brown, 2008; Grasset et al., 2016).

2.2 Water collection for microcosms and bioassay

All lakes sampled in this experiment are in the Kawartha Highlands region to the north of Peterborough, Ontario, Canada (Figure 1a). The selected lakes were Anstruther, Big Cedar, Beaver, Bottle, Catchacoma, Chemong, Gold, Long, Lower Stoney, Upper Stoney, Wolf, Pigeon, and Raccoon (Figure 1b). All lakes were sampled over the course of 2 days during the afternoon, both days were sunny days following a week of no rain and consistent temperatures (average 19.2°C). At each of the sampling sites, water was collected approximately 4 meters from shore by wading in and collecting 4 L of water in acid-washed and MilliQ rinsed 4 L carboys. Water was transported back to Trent University (Peterborough, Ontario) in coolers, processed for chemical analyses, and deployed in the experimental setup (below) within 6 h of collection. For each lake, a subsample of the water was reserved and run through the qPCR protocol to ensure a baseline measurement of P. flavescens DNA was below the detection limit of the assay, to ensure no cross-contamination from the equipment, and to establish there was no baseline perch eDNA levels in the source water.

Water sub-samples were filtered and processed for total dissolved phosphorus (TDP) and total dissolved nitrogen (TDN) as described by Soballe and Fischer (2004). Particulate carbon in the form of size-fractionated carbon (SF-C) was collected by filtering the water through a 25 mm GFF filter and then analyzed on the Elementar vario EL cube CHNS elemental analyzer (Elementar Analyse Systeme GmbH, Langenselbold, Germany). Chlorophyll-a levels were measured with fluorometry using methods described by Larson et al. (2016). pH of the lakes was established using averages from samples collected across years (Table S2).

2.3 eDNA microcosm experiment

Microcosms were set up in 20 mL glass scintillation vials with airtight plastic lids. 20 mL of unfiltered, unprocessed lake water and the appropriate amount of DNA to reach a final total concentration of 1000 copies/mL of extracted perch DNA was added. Two vials were set up as experimental replicates for each lake. The vials were loosely capped to ensure air flow was allowed, and they were placed in a vial tray. This tray was then covered with aluminum foil to block out any light, and the tray was placed in a chamber that was temperature controlled to 20°C for the entirety of the sampling period. At 0, 3, 6, 12, 24, 48, 72, 96, and 120 h, the microcosms were sampled. When sampling, vials were lightly inverted two times to ensure an even distribution of DNA through the water to get a representative sample. A 20 μL sample was then taken and placed in a 1.5 mL tube and stored at −20°C.

2.4 eDNA bioassay experiment

A total of 50 mL of lake water from three lakes (Anstruther, Chemong, and Lower Stoney) was added to 200 mL Whirl-Pak bags. Each lake was subjected to 3 different treatments, with 3 replicates per treatment. The treatments involved: (1) adding P to a nominal concentration of 50 μg/L (high P); (2) filtering the water through a 0.22 μm mixed cellulose esters membrane filter prior to its addition to the bag (filtered); and (3) a control treatment with no filtration or additions (not filtered). Perch DNA was added to a total concentration of 1000 copies/mL into each bag regardless of treatment. After being tightly sealed to avoid any contamination, bags were placed in a floating basket and submerged in the storm water pond on the Trent University campus. Bioassay bags were sampled at 0, 3, 6, 12, 24, 36, and 48 h, with a 100 μL sample collected at each time point from each bag and stored in 1.5 mL tubes, the tubes were then kept in the −20°C freezer. Water temperature throughout was measured using a HOBO logger.

2.5 eDNA quantification

qPCR analysis was conducted directly on water samples using SYBR Green qPCR. The master mix involved the following reagent volumes per reaction: 5 μL of SYBR Green Master Mix, 0.45 μL of each forward and reverse primer (10 μM) and 1.6 μL H₂O. As each reaction was a total of 10 μL, each well contained 7.5 μL of master mix and 2.5 μL of the individual time series water samples. Each water sample was run in triplicate, and each plate contained standardized samples consisting of 10,000, 1000, and 100 copies/μL to calibrate the machine and be able to convert C_t values to copy numbers. A negative control was included on every plate. The qPCR cycling was conducted on the QuantStudio 5 with the following parameters, 1 stage at 50°C for 2 min, followed by 95°C for 10 min before 50 cycles of 95°C for 10 sec, followed by 58°C for 30 sec; a melt curve was not used. C_T values were converted to copy numbers by using the C_T values associated with the known standards. Average hourly copy numbers were normalized by transforming them into the ratio of copies remaining in the system relative to the starting concentration to control for pipetting error and stochasticity. This procedure was completed for each lake and for each treatment for the bioassay data.

2.6 Statistical analysis

All statistical analyses were performed using R v. 4.2.0. To determine the decay rate, k, for each associated lake, the nls.expodecay function from the aomisc package (Onofri, 2023) was used with the ratio data and hours; this provides a model that fits a first-order exponential decay curve: C_t = C₀ e^−kt. Here the first order exponential decay equation, which is the commonly applied model in eDNA degradation studies, C_t denotes the normalized data in the form of ratio of remaining copies at time (t), C₀ denotes the initial number of copies as a ratio (=1) in the system (given as a by the aomisc function), and k represents the decay rate constant. C_t is dependent on the interaction between k and t. We ran a linear model between the k values (decay rate) and the measures of water quality to determine the correlation between chemical measurements and the exhibited rate of decay. Finally, a correlation matrix was conducted to illustrate the relationship between variables.

3.1 Measures of water quality

Water quality measurements (TDP, TDN, SF-C, chl-a, and pH) were typical for oligotrophic lakes in this region with relatively low nutrient and chlorophyll concentrations. TDP measures ranged from 4.31 to 9.92 μg/L, TDN measurements ranged from 151.40 to 329.95 μg/L, SF-C levels ranged from 83.99 to 306.93 μg/L, chl-a concentrations ranged from 2.97 to 8.02 μg/L, and pH ranged from 7.21 to 8.125 (Table 2). There was no significant (<0.7) level of correlation between any measured variable (Table S3). All sampled lakes showed no baseline levels of perch DNA in the water prior to experimental setup.

3.2 Microcosm decay rates

Ten lakes fit the first-order exponential decay models and thus produced k values. For three lakes (Bottle, Beaver, Pigeon), we were not able to fit a first-order exponential decay curve and thus we excluded these lakes from further statistical analyses, though all showed a decline in eDNA (Figure S1). We excluded these lakes from analysis as they exhibited linear rates of decay as opposed to the expected first order exponential rate of decay; rates of linear decay would not be on the same level of magnitude as those exhibited by exponential decay and thus were not comparable. Lakes exhibited notably different exponential decay curves (Figure 2) as reflected by varying rates of decay (Table 2). The rates of decay ranged from 0.27 to 0.06, with a larger k value indicating a greater rate of eDNA degradation. Raccoon Lake exhibited the fastest rate of decay (k = 0.27), and Long Lake exhibited the slowest rate (k = 0.06). There was a positive relationship between decay rate and the level of chl-a in each lake (Table 3); the other water quality parameters showed no significant effect on rates of decay (p > 0.05).

3.3 Bioassay decay rates

In our bioassay experiments, rates of decay were on the lower end of the range that we found in our first experiment. Nonetheless, rates of decay were lowest in the filtered treatment (mean k = 0.06), followed by the non-filtered treatment (mean k = 0.08), and the highest rate of decay was exhibited with the added phosphorus treatment (median k = 0.09; Table 1). The lake water used in this bioassay had differing water quality depending on which lake the water was collected from (Table S1). Chemong Lake, which had the highest level of chl-a, exhibited the fastest rate of eDNA degradation, consistent with the microcosm results. The non-filtered treatments of Anstruther Lake and Lower Stoney Lake also matched the trend seen in the microcosms with Anstruther Lake having the lowest chl-a concentration and the slowest rate of degradation. During the bioassay deployment water temperature fluctuated between 26.68 and 18.43°C, with a median water temperature of 22.24°C.

4.1 Considerations for eDNA sampling and interpretations

Water quality and nutrient load is highly variable between ecosystems, and even lakes that are a part of the same watershed can exhibit extremely different water chemistry (Wagner et al., 2011). Variable water quality could affect eDNA presence and absence by affecting the abundance of organisms and the decay rate of the shed eDNA. Whereas previous studies have investigated the impact of nutrient limitation on eDNA degradation in marine ecosystems (Salter, 2018), showing that variability in eDNA levels can be correlated to the microbial metabolic response to limitations in nutrient availability, this phenomenon has not been thoroughly investigated in freshwater systems. The chemical composition of a lake is indicative of the water quality and nutrient availability, both can vary widely between lakes in the same area (MacLeod et al., 2017) and among regions at broader scales (Wagner et al., 2011). Environmental conditions could thus produce inaccurate estimates of species presence, absence, and community composition especially if sampling strategies do not account environmental effects on eDNA degradation. Beyond more robust sampling with individualized sampling plans, having an understanding on the factors that are impacting the rate of eDNA degradation will enable sampling designs to account for such differences in data analysis. Most of the measured environmental conditions did not affect eDNA decay; this likely stems from there being little variability in these parameters between lakes (Table 2). It is also possible that some variables do not strongly affect decay rates; for example, TDP concentrations are generally low in the Kawartha Highland lakes and are not strongly connected to bacterial activity. Further exploration on a larger scale (i.e. greater range of lake environments) would be necessary to determine whether the other measures of water quality can affect eDNA degradation.

DNA concentrations of 1000 copies/mL are representative of concentrations previously reported for a densely inhabited environment (Saito & Doi, 2021). Thus, the degradation results presented here are representative of the extracellular eDNA fraction in natural environments. Based on our microcosm experiments, there is a possibility that target DNA would be missed if the same sampling strategy were applied for all lakes. For example, the half-life, meaning the number of hours it took to reach 50% [C_t = 0.5; C_0.5] of the starting copies, of Long Lake was 13 h, whereas the half-life of the eDNA in Raccoon Lake is 2.5 h (Table 2), highlighting stark differences in the system impacting the rate of eDNA degradation. A shorter half-life for example would implicate the need for more frequent sampling in areas of lower fish density as there is a risk that if sampling is done on a longer time scale the eDNA would be not detectable in the system. Without establishing the limnological parameters of aquatic ecosystems that directly influence the rate of eDNA degradation, such as chl-a, there is a risk of false negatives, particularly if the results were to be compared to lakes of varying water quality. Interpretations should also rely on the known spatiotemporal distribution of the target species, as the distribution of eDNA is highly dependent on the species range and environmental temperature (Lawson Handley et al., 2019). Therefore, the collective spatiotemporal species-species effects, combined with water quality measurements, should be considered in tandem with the known chemical impacts on eDNA detection and interpretations.

4.2 Chlorophyll as an indicator and bacterial degradation

Parameters assayed for their effects on the degradation of eDNA typically include UV-B exposure, pH, and temperature (Barnes et al., 2014; Eichmiller et al., 2016; Strickler et al., 2015). Bacterial abundance is not often measured as many of the counting techniques require specialized equipment, such as a flow cytometer (Nebe-von-Caron et al., 2000), or require a large amount of time for microbial counts with a microscope (Bratbak, 1993). Chlorophyll-a exhibited a clear relationship with the rate of eDNA decay (Table 3). As chl-a is the primary form of chlorophyll found in phytoplankton (French et al., 1972), it lends itself to be a measurement of algal biomass in the system, and bacterial abundance is known to be positively correlated with algal biomass in lake ecosystems (White et al., 1991). Due to this well characterized relationship, the concentration of chl-a in a lake can be used as an indirect measure of microbial abundance. Algae likely increases the activity of bacteria and thereby increases eDNA degradation (Otten & Gons, 1991; but see Barnes et al., 2014). For example, P-limited algae are known to exude considerable amounts of dissolved organic carbon, which can increase bacterial biomass and microbial P-limitation (Berman-Frank & Dubinsky, 1999). Based on this understanding of the relationship between algal abundance and subsequently released nutrients, it can be hypothesized that elevated chl-a levels correlate to an increased metabolic activity of microbes, and thus would increase the rate at which microbes are able to dissolve eDNA as a source of extra nutrients. As the relationship between chl-a abundance and bacterial abundance is well established, and the protocol to measure chl-a level in water is quite accessible (Aminot & Rey, 2002), chl-a levels have a potential to act as a key parameter in the development of a lake profile for considerations involved in eDNA sampling plan designs.

Aquatic bacteria can secrete a class of ectoenzymes known as phosphohydrolytic enzymes, which include 5′-nucleotidase, endo- and exonucleases, phytase and phosphatases (Grzyb & Fraczek, 2012). These enzymes are secreted by bacteria to degrade organic sources of phosphorus into forms that are soluble and small enough to diffuse through their cellular membranes (Grzyb & Fraczek, 2012). Activity level of these enzymes, as well as the rate of their secretion, can be reliant on the pH of the water (Lance et al., 2017), however the lakes surveyed in this study had relatively neutral pH levels, and did not impact the rate of eDNA degradation (Table 3). It is important to note that at a lower level of pH (pH <6), pH can increase the rate of eDNA degradation (Seymour et al., 2018). One major reservoir of organic phosphorus is dissolved DNA, which is one form of eDNA (Grzyb & Fraczek, 2012), and was the form of DNA used in this study. eDNA that is dissolved in the water can be targeted by the 5′-nucleotidase enzymes that are secreted by many bacteria and cyanobacteria found throughout the water column (Zakataeva, 2021). In our filtered treatment in the bioassays, we used a 0.22 μm pore size filter, which would remove aquatic bacteria (which are 0.4–1.0 mm in diameter (Button & Robertson, 1999)). This filtered treatment showed the lowest mean rate of eDNA decay in the bioassay experiment, which is consistent with the hypothesized role that bacteria have on the degradation of eDNA as in the filtered treatment the bacteria were fully removed from the system. However, comparing the bioassay treatments suggests that the bacteria are using the added inorganic phosphorus instead of targeting and degrading the organic phosphorus in the free-floating eDNA. Phosphohydrolytic enzymes serve the purpose of hydrolyzing different bonds that maintain the DNA structure to make the phosphorus in the backbone soluble and available for absorption. The bacteria prioritizing the use of inorganic P instead of phosphorus in the eDNA is consistent with the idea that inorganic phosphorus is more easily acquired than organic phosphorus (Cotner & Wetzel, 1992). Elevated levels of inorganic forms of phosphorus could therefore influence the rate of eDNA degradation in the system, with a higher level of phosphate forms correlating to a lower rate of eDNA degradation.

4.3 Further directions and conclusions

This study acts as a proof of concept of the role that microbes play in the mechanisms of eDNA degradation. The conclusions drawn from this study were from microcosm and small volume bioassays, and it would be valuable to replicate the study but in larger tank-like environments with in situ conditions to help elucidate eDNA degradation mechanisms. It is also important to highlight that the findings associated with the different rate of decay between lakes is mostly representative of what would be considered the dissolved fraction of whole eDNA in a natural system. eDNA is composed of a variety of sources and in different states, with dissolved mitochondrial DNA only representing a fraction of the total picture, therefore further studies should be done to investigate the ways in which measures of water quality and microbial abundance impact the rate of wide spectrum eDNA to better influence sample design.

The results of this study highlight the important role that lake water quality has on the rate at which eDNA degrades, but more specifically show the role that bacteria play in the eDNA degradation. eDNA sampling guidelines currently do not suggest the modification of sampling strategies based on the lake composition but rather adjust the number of samples collected solely based upon the rarity of the target species (Abbott et al., 2021; Carim et al., 2016). Without incorporating sampling considerations based on the chemical and physical makeup of a lake, or accounting for these differences when making inferences, there is a higher risk of inaccurate quantifications and interpretations of the data. Beyond the well-studied parameters like UV-B penetration, temperature and pH, chl-a presents itself as an informative and accessible measure to add to act as an indicator of bacterial abundance, and thus expected decay rate. Further research needs to be conducted to determine the exact role that bacteria play in the degradation of eDNA and whether this role changes with variable supplies of inorganic phosphorus.