• Open Access

“Sight-unseen” detection of rare aquatic species using environmental DNA


  • Editor
    Corey Bradshaw

Christopher L. Jerde, Center for Aquatic Conservation, University of Notre Dame, P.O. Box 369, Notre Dame, IN 46556-0369. Tel: (574) 631-2665; fax: (574) 631-7413.
E-mail: cjerde@nd.edu


Effective management of rare species, including endangered native species and recently introduced nonindigenous species, requires the detection of populations at low density. For endangered species, detecting the localized distribution makes it possible to identify and protect critical habitat to enhance survival or reproductive success. Similarly, early detection of an incipient invasion by a harmful species increases the feasibility of rapid responses to eradicate the species or contain its spread. Here we demonstrate the efficacy of environmental DNA (eDNA) as a detection tool in freshwater environments. Specifically, we delimit the invasion fronts of two species of Asian carps in Chicago, Illinois, USA area canals and waterways. Quantitative comparisons with traditional fisheries surveillance tools illustrate the greater sensitivity of eDNA and reveal that the risk of invasion to the Laurentian Great Lakes is imminent.


In the absence of tools to detect rare species, uncertainty about when, where, and for how long a management action should be implemented often results in inaction or ineffective use of management resources (Thompson 2004; Lodge et al. 2006). Detection probabilities for rare species are typically low in all environments, but particularly in aquatic environments where organisms are hidden beneath the water's surface. Fish surveillance programs, for example, traditionally employ nets or electrofishing gear. Because these tools usually have low capture probabilities per target organism, they are reliable indicators of occurrence only for species present at moderate-to-high abundance (Magnuson et al. 1994). In contrast, for rare species, the low detection probability of these tools often leads to an error in inference—concluding a species is absent when it is actually present (Gu & Swihart 2004). In the case of rare species, the only solutions are to increase sampling effort or change to a detection tool with greater detection probability (McDonald 2004). However with traditional surveillance tools, increases in sampling effort sufficient to achieve a usefully high detection probability for rare species are often infeasible.

Here we describe the application of environmental DNA (eDNA) as an effective surveillance method for rare fishes in a large river and canal complex. We show that it is more sensitive than traditional tools, has no risk of harming the species under study (Beja-Pereira et al. 2009), and effort can feasibly be increased for species management. The eDNA surveillance method exploits an advantage of aquatic environments: the aqueous environment suspends sloughed tissues, making it easy to sample and detect DNA from even rare organisms that are present but invisible to traditional tools. This conservation application of the eDNA method was inspired by previous identification of whales (Megaptera novaeangliae, Physeter macrocephalus, and Eubalaena glacialis) using DNA in the backwash and feces of diving individuals (Amos et al. 1992), and the use of DNA to detect bullfrogs (Rana catesbeiana) in ponds in France (Ficetola et al. 2008).

We demonstrate the usefulness of eDNA surveillance with a case study involving two species of Asian carps, silver carp (Hypophthalmichthys molitrix) and bighead carp (H. nobilis), which have caused harm to fisheries, recreational use of waterways, and human safety as they have invaded much of North America's Mississippi River basin (Kolar et al. 2007). Both carp species now threaten to invade the Laurentian Great Lakes through a set of human-built waterways in and near the city of Chicago, Illinois, which connects the Mississippi River basin and the Great Lakes-St. Lawrence River basin (Garvey et al. 2010). Specifically, we target eDNA surveillance at locations on the putative leading edge of invasion, compare the detection capabilities between standard fisheries and eDNA surveillance methods, and report far more advanced invasion fronts for each species. Invasion fronts detected with eDNA surveillance suggest that both species of carps are north of electric barriers installed to prevent fish passage, making clear that management actions to prevent invasions of Lake Michigan and the other Great Lakes are much more urgent than suggested by traditional fisheries methods. The application of eDNA monitoring methods will have broad research and management applicability in freshwater, estuarine, and marine ecosystems for threatened and endangered species and for invasive species.


The study area was the Chicago area waterway, a large river and canal complex that has been an important invasion pathway for multiple species including zebra mussels (Dreissena polymorpha) and round goby (Neogobius melanostomus)(Brammeier et al. 2008) (Figure 1). The flow of the Chicago River and other connected waterways was reversed in 1900 to redirect Chicago's wastewater from Lake Michigan; throughout the system, water now flows from Lake Michigan into Mississippi River tributaries (Cooley 1913). In the early 1970s, bighead and silver carp escaped fish farms in Arkansas in the southern Mississippi River basin (Kolar et al. 2007), with the first recorded observation of bighead carp in the Illinois River, a northern tributary of the Mississippi River, occurring in 1986 (USGS, http://nas.er.usgs.gov). Upstream range expansion and population growth of bighead and silver carp in the Illinois River and its tributaries is ongoing (Williamson & Garvey 2005; Sass et al. 2010).

Figure 1.

Connection of the Great Lakes-St. Lawrence and Mississippi River basins through the Chicago area waterway system. The system provides a direct hydrological connection between the Great Lakes and Mississippi River watersheds (the Illinois River is a tributary of the Mississippi River). Water flows from Lake Michigan into each part of the waterway but volumes are regulated and limited by control structures at Wilmette Pump Station, Chicago Lock, and O’Brien Lock. Over 70% of the annual flow is driven by storm water and dilute sewerage discharges from four main outfalls that combine with natural tributary inflows to produce a mean annual (2005) discharge at Lockport of approximately 77 m3/s (2725 ft3/s) (Brammeier et al. 2008). The electric barriers (star) are in the CSSC (not in the Des Plaines River).

Marker development

Molecular markers (Table 1) were designed for H. nobilis and H. molitrix using publically available sequence information (GenBank, http://www.ncbi.nlm.nih.gov). To ensure species specificity, we targeted short fragments of the mitochondrial d-loop region (Taberlet 1996), and used BLAST (Basic Local Alignment Search Tool; Genbank, http://www.ncbi.nlm.nih.gov/blast) to compare the markers to all available sequence data, including those of closely related species and nontarget species common to the Chicago area water system. We also tested the markers against tissue-derived DNA for a subset of species (Table S1, Supporting Information).

Table 1.  Species-specific PCR primers used for amplification of targeted mitochondrial d-loop sequences of bighead and silver carp
TargetPrimer namePrimer sequence

Sample collection, filtration, and extraction

From June 29, 2009 to May 20, 2010, we collected approximately 1,000 2-l surface water samples during 15 sampling trips. Water was collected in autoclaved 2-l Nalgene bottles, and sample locations measured with GPS (Garmin eTrex Venture HC (s.e., <10 m)). Water samples were vacuum-filtered within 24 hours of collection onto 1.5-μm pore size glass fiber filters. Filters containing sample filtrates were placed in 50-mL tubes and stored at -20 °C until further processing. We extracted DNA from the filters using the PowerWater DNA Isolation kit (MOBio Laboratories Inc., Carlsbad, CA) following manufacturer's recommendations with one exception: the final DNA elution was done using deionized water instead of the provided buffer solution.

PCR amplification and evaluation

Asian carp DNA presence was detected using a minimum of eight polymerase chain reactions (PCR) per water sample (amplification cocktail described in Supporting Information). PCR results were visualized and photodocumented under UV light on a 1% agarose gel stained with ethidium bromide. Positive reactions were identified for bighead and silver carp by a single, distinct band at 312 bp and 191 bp, respectively. Using these methods, the lowest concentrations we could detect of genomic DNA from pure DNA extracts from bighead and silver carp were 3.30 × 10−8 ng/μL and 7.25 × 10−11 ng/μL, respectively.

Quality assurance and control

During each sampling trip, water samples were collected from upstream to downstream. All equipment, including boats, used in the sampling effort was sterilized using a 10-minute exposure to 10% bleach solution (Prince & Andrus 1992). Two 2-l bottles, filled with deionized water (cooler blanks), were placed in each cooler and taken into the field to test for contamination. Prior to filtering each sample from the field, 1−l of deionized water was passed through each sterilized filter apparatus (equipment control). All cooler blanks were screened for contamination, and all positive detections from field samples had the corresponding equipment control processed. The water for the cooler blanks and equipment controls was sourced from the same tap. Negative and positive controls were run on each gel. For approximately 5% of all positive samples above the electric barrier, positive bands were extracted from the agarose gel using a Qiagen QIAquick gel extraction kit (Qiagen, Inc., Valencia, CA) and bidirectionally sequenced on an Applied Biosystems ABI 3730xl DNA Analyzer. Sequence data were screened using the BLAST as further confirmation of marker specificity to each of the two target species.

Negative control reach comparisons

As an additional layer of negative control, we sampled from two adjacent river reaches where both carp species are absent. The Des Plaines River flows into the Brandon Road Pool of the Chicago area waterway system but Hoffman Dam (41° 49’ 15.17” N, 87° 49’ 20.46” W) provides a barrier to upstream range expansion by Asian carp (Figure 1). This system has similar species and water chemistry as found in lower sections of the Des Plaines River where we detected carp DNA. Second, we sampled the St. Joseph River, a tributary of Lake Michigan with similar species as found in the upper Chicago area waterway system.

Catch per unit effort comparisons

During our study, the U.S. Army Corps of Engineers, Illinois Department of Natural Resources, Illinois Natural History Survey, and U.S. Fish and Wildlife Service conducted 62 days of electrofishing surveys downstream of the electric barriers from June through November of 2009. To estimate catch per unit effort for electrofishing, we relied on the most detailed information recorded by the management agencies, which consisted of the number of days spent electrofishing. We used a low estimate of effort for electrofishing, with each shock boat consisting of a three-person crew, resulting in 24 person hours needed for an 8-hour electrofishing workday; this results in a catch per unit effort for electrofishing that is upwardly biased because time for travel, preparation, and fieldwork was probably greater than 8 hours and crews may have exceeded three people. Every 100 eDNA samples required 16 person hours for field work and approximately 24 person hours to filter the samples. We averaged 44 samples processed per week. Therefore it took 0.06 person days to collect and 0.114 person days to process a sample, for a total of 0.174 person days per eDNA sample. For electrofishing a catch was defined as a captured or observed Asian carp. For eDNA surveillance, a catch was defined as at least one positive PCR reaction in a sample. We compare the catch per unit effort for electrofishing and eDNA surveillance in three adjacent pools in which both methods were used and which encompassed the presumptive edge of the invasion front in summer 2009. We focused on electrofishing for comparison with eDNA because it has been the most frequently used tool in past fisheries surveillance (Sass et al. 2010).


Molecular markers (Table 1) are species specific. Using the BLAST approach and tests against fish tissue samples in the laboratory, there was no binding sufficient to provide a false positive (Supporting Information). The markers designed to detect bighead carp eDNA amplify a 312 bp section of the mitochondrial d-loop (control region). The genetic distance (as uncorrected p genetic distance calculated using MEGA 4.0; Kumar et al. 2008) between this fragment of DNA for bighead carp and other species is: silver carp, 13.7%; and common carp, 25.2%. The genetic distance between the 191 bp silver carp amplicon and other species is: bighead carp, 8.4%; and common carp, 22.8%.

All sequences (N= 20) from positive samples collected above the electric barrier for bighead or silver carp matched the target species. No detections resulted from samples collected from the negative control reaches (Des Plaines River above Hoffman Dam, n= 15; St. Joseph River, n= 60). An independent, USEPA-led audit, which evaluated the marker specificity, similarly concluded the markers are reliable indicators of bighead or silver carp presence (Blume et al. 2010). No cooler blanks tested positive, and no equipment controls tested positive.

We detected DNA from both silver and bighead carps on multiple occasions from multiple reaches of the waterway (Table 2), well upstream of the putative invasion front that had been determined with standard fishing methods (Figure 2). Prior to eDNA surveillance, the management agencies placed the invasion fronts at about river miles 274 and 281.5 for silver and bighead carps, respectively, approximately 10–15 miles south of the electric barriers. Our eDNA surveillance placed the invasion front for silver carp in the Calumet Harbor of Lake Michigan (RM 333.5, Figure 2D) and at least as far north as river mile 325 for bighead carps in the Calumet River, within 13 km of Lake Michigan (Figure 2C). Silver carp DNA was also detected at river mile 341 in the Chicago River's North Shore Channel (Figure 2E) and at river mile 326.5 in the main channel of the CSSC in downtown Chicago (Figure 2B). Both locations are less than 1 km from Lake Michigan.

Table 2.  Gradient of evidence for species presence Thumbnail image of
Figure 2.

Detection proportions of eDNA for bighead (A,C,E) and silver carp (B,D,F) in the CSSC (A,B), the Calumet Sag Channel (C,D), and the North Shore Channel of the Chicago River (E,F). Locks, barriers, and endpoints of channels are marked with vertical dashed lines. In the reach of the waterway downstream of Brandon Road Lock, both carp species had been collected previously using traditional electrofishing methods (Figure 3). See Supporting Information for the number of 2-l water samples collected within each 2.5 river mile reach.

The catch per unit effort for both electrofishing and eDNA surveillance declined upstream, consistent with approaching the invasion front (Figure 3), but eDNA always had a higher catch per unit effort than electrofishing, despite our bias in calculations that favored electrofishing. For electrofishing, only one silver carp was detected in the Brandon Road Pool after 93 person days of effort that was motivated and targeted by our discovery of Asian carp eDNA. At the lowest Asian carp density (Lockport Pool), both bighead and silver carps were detected only with the eDNA method (Figure 3) until a rotenone application on December 3, 2009 confirmed bighead presence.

Figure 3.

Catch per unit effort for electrofishing (bars) and eDNA surveillance for silver (gray dots) and bighead (black dots) carps from south (downstream) (Marseilles pool) to north (upstream) (Lockport pool); Lockport Lock & Dam is downstream of the electric barriers. Lines are 90% confidence intervals. For electrofishing, silver and bighead carp were not consistently distinguished. A single silver carp was observed with electrofishing in Brandon Road pool (CPUE = 0.01) whereas electrofishing detected no Asian carps in Lockport Pool (CPUE = 0).


The eDNA evidence reported here was the first indication that both bighead and silver carps had not only reached the electric barriers but were also present north of the barriers, which are intended to prevent their dispersal into Lake Michigan. On June 22, 2010, commercial fisherman caught an adult bighead carp within 13 km of Lake Michigan, only 4 km upstream of the nearest positive eDNA detection, further supporting what the eDNA evidence suggested 8 months earlier. In the Calumet River, silver carp and bighead carp appear to have no impediment to access to Lake Michigan. Regardless of the source of the carps north of the electric barrier (see below), eDNA surveillance has revealed much more imminent risks of invasion of Lake Michigan by both silver and bighead carps than had been indicated by standard surveillance methods.

Both bighead and silver carps pose a potentially major threat to the fisheries, recreational values, and other ecosystem services provided by the Great Lakes. For example, bighead carp alter trophic structures, resulting in declines of bait, commercial, and game species (Sampson et al. 2009), and the presence of leaping silver carp has curtailed recreational uses of heavily infested waterways (Kolar et al. 2007). If numbers sufficient to establish a self-sustaining population access Lake Michigan, both species are likely to reproduce, spread, and have substantial negative impacts in portions of the Great Lakes and many of their tributaries (Kolar et al. 2007; Herborg et al. 2007; Cooke & Hill 2010). However, Allee effects, demographic stochasticity, and environmental stochasticity pose major hurdles to population establishment (Jerde et al. 2009); most incipient invasions fail multiple times before they succeed (Drake & Lodge 2006). It is not therefore inevitable that either bighead or silver carps will establish populations in Lake Michigan, especially if the numbers of individuals with access to Lake Michigan can be reduced (Lockwood et al. 2005). Thus the urgent management challenge for the state and national agencies responsible for stewardship of Great Lakes resources is to reduce the numbers of Asian carps upstream of the electric barriers and/or reduce access by those fishes to Lake Michigan.

In addition, because eDNA evidence indicates that at least silver carp have entered Lake Michigan, surveillance is warranted within Great Lakes rivers that may be colonized and could support successful spawning (Kolar et al. 2007). The eDNA method appears well suited to rapid surveys across the large spatial scale that will be required in the Great Lakes.

Finally, the largest potential source of additional individuals is the abundant populations of both species south of the barriers. Given that the effectiveness of the electric barriers are uncertain (Conover et al. 2007; USACE 2010), there is also an urgent need to reduce populations south of the electric barrier to lower the probability of additional upstream dispersal.

Assessing and controlling errors is critical for inferring species presence (Gu & Swihart 2004), especially for an indirect surveillance method like eDNA (Beja-Pereira et al. 2009). We controlled false positives by assessing the specificity of the molecular marker, contamination of the samples during transport and processing, and the correct operation of laboratory equipment. No evidence exists of false positives from contamination, molecular marker misspecification, or equipment failure. In addition, waters known to be absent of Asian carp yielded negative results.

On the other hand, the occurrence of false negatives is affected to an unknown degree by many factors including the density of target fish upstream, the mixing of the water column, the flow rate of the river, and the degradation rate of DNA. In reaches where no Asian carp were detected, uncertainty remains about whether the fish are absent or if we failed to detect fish that were present. Only calibration studies that manipulate the factors potentially leading to false negatives will elucidate the false negative rate. What will remain certain, however, is that the false negative rate will be inversely related to the abundance of the target species (Gu & Swihart 2004; McDonald 2004).

Inferences from a positive eDNA result about fish abundance are currently limited, even where a fish has been captured (Table 2). A positive result indicates only presence, not abundance, of the target species. Because it is not yet known how the abundance, size, behavior, or other individual or population characteristics affect the strength of the eDNA signal under different environmental conditions, the only inference that can be drawn with confidence from a positive result is that at least one individual is or has been present in the recent past (England et al. 2005). Depending on many variables in any particular site, a positive result could indicate one or many fish. However, under similar environmental conditions (e.g., water temperature), such as those represented during our sampling across the different reaches in the Chicago area waterway system (Figure 2), the strength of the eDNA signal likely correlates positively with abundance of the target species. However, because flow rates and turbulence will undoubtedly affect dilution of the eDNA signal, we believe it is not advisable to make detailed inferences about relative abundance of carps on the basis of comparisons of the strength of the eDNA signal between the low flow upstream reaches of the waterway (e.g., the Calumet Sag Channel and the North Shore Channel) and the high flow, turbulent downstream reaches (lower river miles) of the waterway (Table 2 and Figure 2).

It is possible, but not likely, that Asian carp DNA could enter the upper Chicago area waterway (Figure 1) by sources other than living fish, such as sewage and wastewater, bilge water discharge, excrement from predatory fish or waterfowl, or from dead fish carried on barges and boats from downstream. None of these alternate sources explains the overall spatial patterns or repeated detections from independent sampling trips spread across a 12-month period (Table 2). For example, barges and towboats do not typically pump ballast because they are not equipped to exchange large quantities of water; furthermore discharging water from below the electric barriers above the electric barriers is prohibited through federal regulations. Additionally, these vessels do not use some sections of the Chicago area waterway system where we detected eDNA (e.g., North Shore Channel, Figure 2F). To date, no alternative pathway has been demonstrated to transport eDNA in quantities and occurrences sufficient to maintain the pattern of detection reported here (Figure 2 and Table 2). The most plausible explanation for positive detections is the presence of live fish.

Without effective surveillance methods, successful management of rare species is impossible (McDonald 2004). The pattern of greater catch per unit effort using molecular genetic detection compared to more traditional sampling approaches has also been documented in marine systems where invaders are at low abundance (Hayes et al. 2005). We believe that improvements to the eDNA method, and applications to additional species and ecosystems, will increase the ability of natural resource managers to foster endangered species and reduce the damages of invasive species in aquatic environments. Targeting management earlier and in more appropriate locations will also improve the cost effectiveness of management (Leung et al. 2002).


Funding provided by U.S. Army Corps of Engineers (#W912HZ-08-2-0014), Great Lakes Protection Fund (#867), and NOAA CSCOR (#NA09NOS4780192) to the University of Notre Dame, and the Joyce Foundation (#07-30723) to The Nature Conservancy. K. Baerwaldt contributed the electrofishing data, insightful advice on Asian carp habitats, and comments on the manuscript. A. Rohly, S. Mysorekar, M. Brandner, E. Elgin, M. Budny, M. Galaska, M. Barnes, M. Brueseke, C. Turner, and J. McNulty assisted with the research. D. Chapman, J. Hoover, and J. Killgore provided advice about Asian carp biology. The U.S. Fish and Wildlife Service, the Illinois Department of Natural Resources, and the U.S. Army Corps of Engineers contributed logistical support for the water sample collection.