Environmental Conditions Influence eDNA ... - ACS Publications

Article pubs.acs.org/est

Environmental Conditions Influence eDNA Persistence in Aquatic Systems Matthew A. Barnes,*,† Cameron R. Turner,† Christopher L. Jerde,† Mark A. Renshaw,† W. Lindsay Chadderton,‡ and David M. Lodge† †

Department of Biological Sciences and Environmental Change Initiative, University of Notre Dame, Notre Dame, Indiana 46556, United States ‡ The Nature Conservancy, c/o Notre Dame Environmental Change Initiative, Unit 117, 1400 East Angela Boulevard, South Bend, Indiana 46617, United States S Supporting Information *

ABSTRACT: Environmental DNA (eDNA) surveillance holds great promise for improving species conservation and management. However, few studies have investigated eDNA dynamics under natural conditions, and interpretations of eDNA surveillance results are clouded by uncertainties about eDNA degradation. We conducted a literature review to assess current understanding of eDNA degradation in aquatic systems and an experiment exploring how environmental conditions can influence eDNA degradation. Previous studies have reported macrobial eDNA persistence ranging from less than 1 day to over 2 weeks, with no attempts to quantify factors affecting degradation. Using a SYBR Green quantitative PCR assay to observe Common Carp (Cyprinus carpio) eDNA degradation in laboratory mesocosms, our rate of Common Carp eDNA detection decreased over time. Common Carp eDNA concentration followed a pattern of exponential decay, and observed decay rates exceeded previously published values for aquatic macrobial eDNA. Contrary to our expectations, eDNA degradation rate declined as biochemical oxygen demand, chlorophyll, and total eDNA (i.e., from any organism) concentration increased. Our results help explain the widely divergent, previously published estimates for eDNA degradation. Measurements of local environmental conditions, consideration of environmental influence on eDNA detection, and quantification of local eDNA degradation rates will help interpret future eDNA surveillance results.

■

INTRODUCTION Recent advances in genetic surveillance techniques, such as the collection and identification of shed tissues or feces, have improved environmental management by permitting species detection even when collection of whole organisms is impractical.1 The term environmental DNA, or eDNA, has emerged to describe recent surveillance efforts in which extraorganismal genetic materials (e.g., sloughed cells, feces, gametes, and other particles) are obtained through methods designed to sample the environment (e.g., air, water, soil) rather than capture whole organisms.2 For example, recent applications of eDNA surveillance in aquatic ecosystems have employed grab samples of water for the detection of amphibians in lentic3,4 and lotic5,6 ecosystems, fish in freshwater7−10 and marine11 environments, and marine mammals.12 The promise of eDNA to detect aquatic species was identified as one of the 15 global conservation horizons for 2013 in an annual report from Trends in Ecology and Evolution.13 However, ensuring the success of future applications of eDNA detection depends on increased understanding of the production and persistence of genetic materials in the environment as well as technical details of sample collection © 2014 American Chemical Society

and analysis. Despite recent examples (cited above) that demonstrate the efficacy of eDNA surveillance, few studies have examined the rates of eDNA production or degradation as well as the environmental factors that may affect those rates. Yet without understanding how environmental factors influence eDNA degradation, it is difficult to conclude whether a positive eDNA detection represents fresh genetic material from an organism recently occupying the sample space or relics from an organism that occurred in the more distant past. Recent eDNA degradation experiments have involved removal of target organisms from controlled environments followed by monitoring the persistence of target eDNA over time. eDNA degradation rates observed in this manner have differed across species and experimental setups. For example, larval amphibian eDNA maintained a detection probability of >5% for 25 days in glass beakers, and fish eDNA maintained >5% detection probability for 17 days in small experimental ponds.14 However, eDNA decayed exponentially and became Received: Revised: Accepted: Published: 1819

October 23, 2013 January 13, 2014 January 14, 2014 January 14, 2014 dx.doi.org/10.1021/es404734p | Environ. Sci. Technol. 2014, 48, 1819−1827

Environmental Science & Technology

Article

example, manipulations of fish species cause changes in food webs and physicochemical variables.19 Accordingly, to create laboratory mesocosm treatments, we randomly assigned 0, 1, or 3 Goldfish (Carassius auratus; each ∼5 cm in length) to each mesocosm. To avoid introduction of allochthonous DNA, Goldfish were not fed during the experiment. Because fish excretion promotes nutrient availability,20 we expected (and observed) Goldfish to have a fertilization effect on their mesocosms, promoting algal and heterotrophic microbial activity, as well as altering chemical conditions such as pH through ammonia excretion.21 Thus we used a gradient of Goldfish abundance to create a gradient of environmental conditions typical of natural covariation in factors hypothesized to affect eDNA degradation. One appealing aspect of eDNA as a surveillance tool is the relative ease with which samples can be collected. 17 Accordingly, of the many environmental variables potentially influenced by Goldfish presence, we selected four easily measured covariates as indices of physicochemical variation that were identified in our literature review as (i) potentially influential on eDNA degradation and (ii) expected to differ with Goldfish abundance. Using ethanol extraction and a TD700 fluorometer (Turner Designs), we measured chlorophyll a concentration (three replicates, each 250 mL volume) as an indicator of primary production.22 Biochemical oxygen demand (BOD) served as an indicator of community metabolism and was measured using the BOD5 method (250 mL volume).23 As an indicator of microbial density, we quantified total eDNA concentration (i.e., from any organism) in total genomic eDNA extractions of three samples from each mesocosm using a Nanodrop 2000 (Thermo Fisher Scientific). We also measured pH with a HACH HQ40d multi probe (Hach Company, Loveland, Colorado, USA). To lower the risk of water movement (and eDNA contamination) between mesocosms, we measured pH within three 250 mL water samples removed from each mesocosm rather than inserting the probe into the mesocosm itself. Furthermore, all environmental measurements were collected at only one time during the experiment, 168 h (7 days) following setup. We used principal component analysis to combine all measured covariates into a single environmental index (PC1). All statistical analyses in this experiment were performed in R 2.12.1 (R Foundation for Statistical Computing). To determine the influence on eDNA degradation of the Goldfish-mediated environmental variation, we introduced eDNA from a separate fish species, Common Carp (Cyprinus carpio) to each experimental mesocosm and measured Common Carp eDNA concentration as it decreased over time. To provide a highly concentrated eDNA source, we collected water from a Common Carp pond (0.06 ha) at the Potawatomi Zoo, South Bend, Indiana, USA. At the zoo, we collected surface water with a peristaltic pump and stored water in 20 L carboys for transport to the solarium. In the solarium, carboys were shaken vigorously to homogenize the water sample, and 2 L Common Carp eDNA water was distributed to five experimental mesocosms per Goldfish treatment. Two mesocosms per Goldfish treatment served as negative controls, receiving 2 L well water rather than zoo water. Following the addition of 2 L water from the Common Carp habitat (for experimental units) or 2 L additional well water (controls), total volume of each mesocosm was 4 L (N = 5 zoo aliquot + 2 controls = 7 mesocosms per Goldfish level).

reduced beyond the limit of qPCR detection after 0.9 day and 6.7 days for two marine fish species in aquaria.11 In another laboratory experiment, New Zealand mudsnail eDNA remained detectable for 21 days following snail removal.15 In an outdoor mesocosm experiment with two amphibians, eDNA persisted 7−14 days following removal of organisms.8 Across these studies, researchers have speculated that environmental factors such as temperature, pH, conductivity, and microbial community composition may influence eDNA degradation. However, we are not aware of any study that has measured environmental covariates in an attempt to elucidate these relationships. Although eDNA research has enjoyed a recent surge in popularity due to improved genetic methodology and potential conservation applications,16,17 microbiologists and researchers from other diverse fields have been extracting DNA from environmental samples for decades,2 and existing data may inform current understanding of eDNA degradation. To that end, we conducted a literature review to summarize what is already known about the potential drivers of eDNA degradation in aquatic systems. We also performed an eDNA degradation experiment including concurrent measurement of several environmental conditions in an effort to evaluate their influence on eDNA degradation rate.

■

MATERIALS AND METHODS Literature Review. On 31 January 2013, we conducted a Thomson Reuters Web of Science search using the following search terms: eDNA OR ″environmental DNA″ OR ″extracellular DNA″ AND degradation OR decay OR decomposition. From the search results, we identified publications that attempted to analyze extraorganismal DNA degradation in response to environmental variation, including both laboratory manipulations and in situ analyses across environmental gradients. We also considered other literature reviews describing the influence of environmental factors on DNA as well as additional relevant publications appearing in our search results’ cited literature. Based on the reviewed publications, we assembled a list of factors that impact eDNA degradation. Impacts included changes in the degradation rate, or the quantity or behavior of DNA in the environment in response to environmental conditions. Experimental Manipulation. To complement our literature review, we conducted a laboratory mesocosm experiment to quantify the effect of a gradient of environmental characteristics on eDNA degradation. We prepared 21 individually aerated 20 L mesocosms, each filled with 2 L well water (initial pH ≈ 7.5). Mesocosms were evenly distributed in space (approximately 0.2 m between mesocosms) in a climate-controlled solarium where they remained at 25 °C (±1 °C standard deviation) and received natural light that decreased from 14 h 50 min to 14 h 10 min over the course of the 21 day experiment. Many of the studies encountered during our literature review highlighted the importance of the activity of microbes and their enzymes in eDNA degradation (see Results and Discussion); therefore, in our experiment we sought to manipulate the microbial community while concurrently manipulating multiple environmental factors to mimic the correlations between factors that typically occur in the surface waters of natural freshwater ecosystems. Among aquatic ecosystems, changes in diversity and abundance of local biota produce far-reaching effects on water quality and ecological processes.18 For 1820

dx.doi.org/10.1021/es404734p | Environ. Sci. Technol. 2014, 48, 1819−1827


Article

Figure 1. (A) Cumulative number of eDNA degradation studies over time. Macrobial detection studies considered eDNA for the purpose of detecting multicellular organisms. Microbiology studies examined genetic detection of single-celled organisms and extracellular DNA. Water quality studies included eDNA detection of fecal pollution. (B) Factors demonstrated to influence eDNA degradation and number of studies providing support. Note that individual studies could implicate multiple factors as drivers of eDNA degradation and appear in multiple columns.

NaCl, and we incubated at −20 °C overnight. Next, we centrifuged at 16 873 RCF for 10 min to pellet DNA, discarded the supernatant, and rinsed the pellet twice with 150 μL 70% ethanol. Finally, pellets were resuspended in 100 μL TE buffer. Common Carp eDNA in each sample was quantified via qPCR using primers designed to be species-specific and produce a 146 bp amplicon (forward sequence: GAGTGCAGGCTCAAATGTTAAA; reverse sequence: GTAAGGATAAGTTGAACTAGAGACAG); pilot experiments confirmed that these primers would not amplify tissue-derived Goldfish DNA (Supporting Information). Individual reactions (N = 3 technical replicates per sample) consisted of 2 μL each primer at 3 μM concentration, 2 μL pure water containing 4 μg/μL BSA, 10 μL Power SYBR Green Master Mix (Life Technologies), and 4 μL DNA extract. The qPCR reaction consisted of an initial activation step of 95 °C for 10 min

To monitor eDNA degradation, we collected 250 mL water from each mesocosm at the time of setup and then at 8, 16, 24, 72, 96, 120, 168, 336, and 504 h (0.3, 0.7, 1, 3, 4, 5, 7, 14, and 21 days, respectively) elapsed time. To prevent contamination across mesocosms, water samples were collected in autoclaved bottles while wearing a new set of nitrile laboratory gloves for each collection. Water was immediately filtered through 1.2 μm Isopore membrane filters (EMD Millipore), and filters were stored in individual sterile microcentrifuge tubes at −20 °C until DNA extraction. We extracted eDNA following a modified version of the procedure described by Coyne et al.24 We combined filters with 700 μL cetrimonium bromide (CTAB) cell lysis buffer, then added 700 μL 24:1 chloroform:isoamylalcohol. Following 5 min of gentle agitation, we centrifuged at 14 000 rpm for 15 min and transferred 500 μL supernatant to a new tube. We then added 500 μL isopropanol and 250 μL 5 M 1821



Article

followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. The reaction was followed by a melting curve analysis transitioning from 60 to 95 °C over 20 min, and we excluded technical replicates with melting curves differing from positive controls.25,26 Positive controls were utilized for absolute quantification of Common Carp DNA via comparison to a standard curve (Supporting Information).27 A new standard curve and duplicate negative control reactions were prepared with each plate of qPCR reactions. We analyzed qPCR reactions using the default Noiseband setting in Realplex 2.2 (Eppendorf). We confirmed that qPCR product matched the target amplicon via Sanger sequencing (ABI 3730xl, Applied Biosystems) of a subset of positive reactions. Technical replicates that failed to amplify were assigned a value of zero rather than omitted from the analysis.28 This convention is based on the assumption that many samples actually were absent of Common Carp DNA, especially as degradation occurred over time. Statistical analyses used a single mean concentration (copies/reaction) of Common Carp eDNA for each mesocosm at each time. We used a Kruskal−Wallis test to confirm that initial Common Carp eDNA concentration did not differ between the three Goldfish treatment levels. Because previous research has analyzed eDNA degradation in aquatic systems with presence− absence detection14 and absolute quantification,8,11 we used both approaches. Using maximum likelihood estimation, we fit a mixed-effects logistic model to Common Carp eDNA presence−absence data over time, treating replicate as a random effect within the model while testing whether Goldfish treatment level influenced the model. Because we observed no influence of Goldfish treatment on logistic model fits (see Results and Discussion), we used the R package pROC29 to assess the fit of a single logistic model based on presence− absence data across all mesocosms using area under the receiver operating characteristic curve (AUC), where AUC = 0.5 indicates the model predicts outcomes no better than random, and AUC ≥ 0.7 indicates strong predictive ability.30 We also used linear regression to determine whether initial Common Carp eDNA concentration predicted time until detection failure. For analyses of eDNA concentration, we ln+1 transformed all Common Carp eDNA concentration data and fit a general linear mixed effects model to these data over time, treating Goldfish treatment level as a fixed effect and mesocosm as a random effect. We also used Pearson correlation to ask whether a relationship existed between our index of environmental variability (PC1) and Common Carp eDNA degradation rate because we observed no influence of Goldfish treatment level on eDNA degradation rate (see Results and Discussion) and Goldfish treatments resulted in a gradient of environmental conditions rather than three distinct treatment levels. For this analysis, we used decay rates estimated by fitting an exponential decay model, D(t) = D0e−rt, where D(t) is the concentration of Common Carp eDNA present at time t, D0 is the initial concentration of Common Carp eDNA, and r is the decay rate, to raw data from each mesocosm using the R function nls.

Most publications related to one of three general categories: microbial metabolism and transformation, fecal pollution and water quality assessment, or macrobial species detection via eDNA surveillance (Figure 1). Citations indicating exchange of ideas across these three general categories were infrequent, suggesting that stronger collaboration across these research communities could benefit all three lines of research. One lesson that emerged from our literature review is that eDNA represents a heterogeneous mixture of genetic materials ranging from chromosomes and plasmids encased within intact cells and cellular debris to extracellular DNA fragments freely floating in the environment or adsorbed onto other particles within the environment.31 Although microbiologists have made considerable efforts to distinguish between intra- and extracellular DNA,31−33 investigators applying eDNA to detect macrobiota have not. The condition of target eDNA in the environment may have important implications for comparison between efforts that have collected eDNA via precipitation3,8,12 and size-selective filtration.5,7,11 Therefore, more explicit consideration of the method used to capture eDNA is essential for future eDNA surveillance efforts. Whether eDNA is extracellular, intracellular, free, or adsorbed, the reviewed literature demonstrated that there are multiple possible fates of eDNA in the environment, with many factors potentially interacting to preserve or degrade eDNA.34−38 We organized the factors that potentially influence eDNA persistence into three broad categories: characteristics of the DNA molecule, abiotic environmental characteristics, and biotic environmental characteristics. DNA can exist in extremely different lengths, sequences, and conformations, and each of these characteristics influence how DNA interacts with its environment and degrades over time. In particular, the length and conformation of DNA fragments influence the binding of DNA to other particles39−42 and its interaction with the microbial community.43 Containment by membranes influences eDNA degradation, as DNA within dead or dying cells is subject to breakdown during programmed cell death as well as unregulated degradation by intracellular enzymes.44−46 On the other hand, cellular and organellar membranes can provide protection from abundant external degrading forces.47−51 Abiotic environmental characteristics influence eDNA degradation through a variety of mechanisms. Higher temperatures directly increase DNA degradation by denaturing DNA molecules and indirectly degrade eDNA by increasing enzyme kinetics and microbial metabolism.52−58 Hypersaline and anoxic environments can influence eDNA conformation and stability and restrict exonuclease activity that may limit degradation.53,59 Other abiotic factors such as ultraviolet radiation are expected to degrade eDNA based on experiences in the laboratory.60 However, in some experimental studies, solar radiation had no effect on the rate of eDNA degradation,50,61,62 while in other studies light exposure increased eDNA degradation rate.57,63 Many studies have documented the important role that sediments play in eDNA preservation, by both binding and protecting DNA molecules from other forces and by binding and inactivating extracellular nucleases.36,39,41,64−71 DNA binds more strongly to clay particles than to sand or silt,72 and rates of adsorption of DNA to clay differ among types of clays,42,73,74 suggesting that the composition of sediments (suspended or benthic) may influence eDNA degradation. Additionally, local environmental conditions such as salinity75 and pH76 mediate DNA−sediment interactions.

■

RESULTS AND DISCUSSION Literature Review. Our initial literature search returned 97 potentially relevant publications, which a review of abstracts allowed us to narrow to 37 relevant publications. Additional searching within the literature cited sections of these papers increased the total to 68 publications for more careful analysis. 1822



Article

oxygen demand (loading = 0.51), total eDNA (0.50), chlorophyll concentration (0.49), and pH (0.49) all demonstrated similar influence on PC1. Given our experimental design, it is difficult to distinguish whether loadings indicate that all measured environmental variables similarly influenced PC1 (and, in turn, eDNA degradation) or whether the similar loading scores are the result of correlations between variables. Thus, we interpret the effect in terms of this composite index of combined treatments, which mimics natural gradients in surface waters. At the onset of the experiment, Common Carp eDNA concentration did not differ significantly among Goldfish treatments (Kruskal−Wallis p = 0.230), whereas control treatments (i.e., mesocosms not spiked with water containing Common Carp eDNA) tested negative. However, detection of Common Carp eDNA occurred within control mesocosms intermittently from 8 to 72 h (0.3−3 days). Such detections were sporadic (mean = 8 of 10 time points per control mesocosm with no Common Carp detection; max = 4 time points per control mesocosm with Common Carp detection) and at low levels (mean = 1 copy/reaction; max = 12 copies/ reaction). Because controls did not contain Common Carp eDNA at the beginning of the experiment, and because Common Carp eDNA in the control mesocosms tapered off as eDNA concentrations in experimental mesocosms decreased, these data suggest that minor cross-contamination among mesocosms occurred, probably as a result of the small water droplets produced by aeration. In hindsight, it is clear that such droplets could easily traverse the 0.2 m between mesocosms. Cross-contamination was undesirable within the experimental design, but likely exerted little influence on the overall trends in persistence time observed throughout the experiment because contamination concentrations were very low and appeared to be distributed randomly across mesocosms. Furthermore, we sequenced 86% of qPCR products from the entire experiment, and all sequences were confirmed as 100% match to our target Common Carp amplicon, indicating that Goldfish eDNA was not the source of contamination within control mesocosms or Common Carp eDNA detection in experimental mesocosms. Across all treatments, Common Carp eDNA detection decreased over time. Logistic regression did not identify a significant influence of our Goldfish treatment levels on model fit (p = 0.7). Considering data pooled over three goldfish treatments, detection probability was predicted significantly and accurately (logistic regression p < 0.001, AUC = 0.968; Figure 2). The model estimated that 47.2 h (∼2 days) represented the point at which qPCR became more likely to fail to detect Common Carp eDNA than to positively detect it, and 101.1 h (4.2 days) passed before failure to detect Common Carp eDNA became 95% probable. In general, the logistic model suggested a more rapid decline in eDNA detection probability than previous studies. However, extreme cases in our study (i.e., 2 technical replicates from 1 mesocosm at 168 h [7 days] elapsed time and 1 technical replicate from each of 2 mesocosms at 336 h [14 days]) in the zero Goldfish treatment were consistent with some longer published values. For example, eDNA remained detectable for 7−14 days8 and 25 days14 in studies of freshwater amphibians, and in an aquarium study of marine fish, eDNA was detectable for up to 7 days following fish removal.11 It is difficult to speculate about what factors contributed to eDNA longevity because few environmental parameters were reported in previous studies. In our experiment, the pair of zero Goldfish mesocosms with the persistent

Several experiments have illustrated the importance of the biotic community for eDNA degradation. For example, DNA degrades more rapidly in untreated water than in autoclaved or otherwise sterilized controls.52,57,77−84 Extracellular enzymes also play a large role in eDNA degradation.85−88 Observations in filtered but unsterilized water samples indicate that the influence of extracellular enzymes can outlast the presence of microorganisms themselves.70,80,81,84 Indeed, many of the abiotic influences on eDNA degradation that we have identified are most notable due to their effects on extracellular enzyme activity. Overall, our literature review revealed an extensive list of factors that influence macrobial eDNA degradation, but no experiments directly testing their impact. We have identified several key issues to consider as the eDNA research agenda continues to expand. First, a more thorough understanding is needed of the origin and characteristics of the particles containing eDNA (e.g., are the particles bound by membranes?). Promising lines of research include investigating which types of genetic materials (e.g., feces, reproductive fluids, sloughed cells) organisms most commonly shed into their environment. Second, better quantification of rates of production and degradation of eDNA are required. Third, research is needed to identify the most effective field collection methods for eDNA. Fourth, research comparing the eDNA composition and shedding rates between different taxonomic groups will inform the extent to which conclusions are transferrable across species. In addition to acting directly upon eDNA, many of the identified characteristics work synergistically or antagonistically to preserve or degrade eDNA. Experimental manipulations and surveys of degradation rates across natural environmental gradients will help elucidate the relative importance of different factors on eDNA degradation and increase the interpretability of eDNA surveillance results. Experimental Manipulation. Many of the studies encountered during our literature review highlighted especially the importance of the activity of microbes and their enzymes in eDNA degradation; therefore, in our laboratory experiment we monitored indicators of the community of microbes and other microorganisms such as BOD, chlorophyll a concentration, and total eDNA (i.e., from any organism) as well as environmental pH, an important factor in extracellular enzyme function. Indeed, BOD, chlorophyll a concentration, pH, and total eDNA differed among Goldfish density treatments (Table 1). Table 1. Characteristics of Different Goldfish Treatment Levelsa Goldfish per mesocosm

BOD (mg/L)

Chlorophyll (μg/L)

pH

Total DNA (μg/L)

0 (N = 5) 1 (N = 5) 3 (N = 5)

1.6 ± 0.1 4.8 ± 1.0 9.4 ± 0.6

2.3 ± 1.0 257.0 ± 62.6 484.5 ± 185.3

9.2 ± 0.0 9.7 ± 0.1 9.9 ± 0.1

7.7 ± 1.7 96.8 ± 9.4 276.4 ± 27.9

a

All errors presented represent ±1 standard error.

Increasing Goldfish density resulted in increased BOD, chlorophyll a concentration, pH, and total eDNA, suggesting that Goldfish excreta and egesta fertilized both autotrophic and heterotrophic microbes. Principal components analysis condensed these highly correlated characteristics into a single index of environmental variability. The first principal component (PC1) explained 90.64% of the overall variation. Biochemical 1823



Article

marine fish species11 and r ≈ 1−2 (data only presented graphically) for two larval amphibian species.8 A general linear mixed model of ln+1 transformed Common Carp eDNA concentration over time did not implicate our three Goldfish treatment levels as a significant driver of degradation rate (p = 0.1398). However, we observed that Goldfish treatments resulted in a gradient of environmental conditions across mesocosms. A significant negative correlation (Pearson’s r = −0.715, p = 0.003) existed between Common Carp eDNA degradation rate and our calculated index of environmental variability, PC1 (Figure 4). The direction of the effect was surprising: degradation rate decreased as biochemical oxygen demand, chlorophyll a concentration, pH, and total eDNA concentration increased. Figure 2. Logistic regression analysis of Common Carp eDNA detection probability (pooled over all three Goldfish treatments). Observation data equal 1 (detection) or 0 (no detection). Data are jittered along the y-axis, and the x-axis is displayed on a log scale to increase visibility of individual measurements. The line represents the best fit logistic function describing Common Carp eDNA detection probability (P = [(f(t))/(1 + f(t))], f(t) = 2.498 − 0.053t)).

eDNA had an average of 26 and 33 copies/reaction Common Carp eDNA concentrations at the onset of the experiment, in line with the mean across all mesocosms (33 ± 4 copies/ reaction ± 1 SE); furthermore, linear regression across all mesocosms indicated that no relationship existed between initial Common Carp eDNA concentration and time until final detection (p = 0.223). Thus, eDNA remained detectable longest in the treatments in which overall biological activity was lowest as indexed by BOD, chlorophyll, and total DNA (Table 1). The concentration of Common Carp eDNA followed a pattern of exponential decay (Figure 3). Considering data across all mesocosms, our estimated degradation rate (±1 SE) of r = 0.105 ± 0.014 was slower than previously reported values: r = 0.322 and 0.701 (SE not reported) for a pair of

Figure 4. Estimated Common Carp eDNA decay rate decreased as environmental principal component 1 (PC1) increased.

In presence−absence analysis (Figure 2), zero Goldfish mesocosms had late isolated occurrences of Common Carp eDNA; however, when we used eDNA concentration over time to calculate degradation rate, zero Goldfish mesocosms tended to demonstrate the highest rates of eDNA degradation (Figure 4). This contrast in results between different analyses within our single experiment constitute an important warning to future eDNA surveillance programs, regardless of environmental conditions. These observations highlight the potential for sampling goals and strategy (i.e., presence−absence of eDNA vs quantification of the concentrations of eDNA) to dramatically change interpretations of eDNA surveillance results, especially in situations where eDNA is at low concentrations or heterogeneously distributed throughout the environment. Given the importance of microbial metabolism as a driver of eDNA degradation documented in our literature review, we predicted that increased community metabolism resulting from the fertilization effect produced by Goldfish would increase eDNA degradation rates. We offer three possible explanations for the divergence between our prediction and the results. First, although some previous studies found no effect of sunlight exposure on eDNA degradation,50,61,61 our results implicate a

Figure 3. Quantification of Common Carp eDNA followed exponential decay over time (pooled over all three Goldfish treatments). Each datum indicates the average of three technical replicates for one mesocosm at one time, and data are jittered along the x-axis to improve visibility. Line represents the best fit exponential decay function describing Common Carp eDNA concentration (D(t) = 32.164 e−0.105t). Note that a qPCR reaction included 4 μL extracted DNA. 1824



■

ACKNOWLEDGMENTS This project was funded primarily by US EPA’s Great Lakes Restoration Initiative. M.A.B. also received funding from Great Lakes Fisheries Trust, Great Lakes Protection Fund, and the National Science Foundation IGERT #0504495. Thank you to L Arriaga and staff at the Potawatomi Zoo for water access and support. C. C. Y. Xu assisted in the development of eDNA capture, extraction, and qPCR techniques. A. Baldridge, J. Corush, and A. Deines helped with experimental setup and maintenance. M. Suckow and the Freimann Life Sciences Center oversaw animal care under University of Notre Dame Institutional Animal Care and Use Committee protocol #13027. This manuscript benefitted from conversations with J. Livermore and W. West. A. Baldridge, A. Deines, C. Gantz, E. Grey, L. Sargent, M. Wittmann, and three anonymous reviewers provided comments that improved an earlier draft of this manuscript. This is a publication of the Notre Dame Environmental Change Initiative.

decrease in light penetration by increased algal density as a potential driver of Common Carp eDNA degradation. Any increase in degradation rate that may have resulted from increased community metabolism in the treatments with Goldfish may have been outweighed by a decrease in degradation rate caused by reduced light. Second, Goldfish presence could have also reduced heterotrophic microbial degradation of Common Carp eDNA by providing a more preferred nutrient source through excretion. Goldfish-derived materials may have been a more nutritious substrate for microbial metabolism compared to Common Carp derived material, which would have been partly metabolized by the time of collection. Third, variations in abiotic environmental conditions may have also influenced metabolic activity of the heterotrophic microbial community, although this seems less likely than the explanations above. Previous studies have shown that deviations from neutral pH can reduce degradation rates, especially for substances that must be degraded with the aid of extracellular enzymes.89 For example, several studies documenting genetic detection of degrading leaf litter found decay rates negatively associated with increasing acidity (