Use of Environmental DNA for Early Detection of Invasive Elodea Species in Interior Alaska

Elodea spp. are aquatic submerged plants native to much of North America. However, two species, Canadian waterweed E. canadensis and western waterweed E. nuttallii, are invasive to Alaska (Carey et al. 2016). These two species and their hybrids (hereafter collectively referred to as Elodea) are an immediate threat in Alaska, because dense stands degrade juvenile salmon Onchorhynchus spp. habitats, salmon spawning habitats, and impede boat and float plane movement on lakes (Carey et al. 2016, 2023; Schwoerer et al. 2020). Elodea can be spread via boats and floatplanes; because it can reproduce vegetatively, a single fragment can start a new infestation (Carey et al. 2016). We define an Elodea infestation as a density that can cause ecological damage and require human intervention to mitigate further spread. Many ecologically and commercially important regions are at high risk for future Elodea introductions via floatplane traffic (Carey et al. 2016; Schwoerer et al. 2022), including the Bristol Bay, Alaska, USA, sockeye salmon run, which is the largest in the world (Cunningham et al. 2019). Removal of Elodea is expensive; use of fluridone (an organic compound often used as an aquatic herbicide) appears to be the most effective method of eradication (Sethi et al. 2017).

Hybrids of E. canadensis and E. nuttallii produce viable stands (Cook and Urmi-Konig 1985). Their identification, along with species-specific designations, can be helpful to understand the spread from existing stands or introduction of new stands. However, identification is challenging both within the Elodea genera and within the family Hydrocharitaceae (Bowmer et al. 1995; Simpson 1984). New infestations in Alaska can be morphologically challenging to distinguish interspecifically; therefore, developing both genus-selective and species-selective markers allows for a robust approach to identify Elodea. Next-generation sequencing was used to verify a hybrid at Stormy Lake on the Kenai Peninsula, E. canadensis in several lakes in Cordova and Anchorage, and E. nuttallii in Fairbanks (K. Mohatt, United States Forest Service, unpublished data).

In Alaska, the most common detection methods for Elodea are visual searches of water bodies from boats and airplanes. Visual searches are aided by the retrieval of a rake thrown from a boat or shoreline. While these methods are effective for identifying visible outbreaks, it would be advantageous to have an early-detection method before an outbreak becomes visible. Environmental DNA (eDNA; i.e., genetic material extracted from an environmental sample such as water), can be an effective tool to determine presence of a plant species (Fujiwara et al. 2016; Matsuhashi et al. 2016; Gantz et al. 2018). Several factors can affect the eDNA detection process. For example, Kuehne et al. (2020) found that detection of Eurasian milfoil Myriophyllum spicatum and Brazillian elodea Egeria densa was affected by plant growth, senescence, and abundance. Additionally, in a heavy infestation of E. canadensis in Lake Steinsfjorden in Norway, eDNA concentration varied seasonally within the lake and as a function of distance from the source in an outlet stream (Angles d’Auriac et al. 2019).

Angles d’Auriac et al. (2019) found E. canadensis eDNA was detected at a visible outbreak of the plant; however, we found no information about the efficacy of using eDNA as an early-detection method for a few plant fragments. Reliable detection of Elodea eDNA originating from less readily observed stands, before population eradication becomes difficult or infeasible, would be an important advance for early detection and rapid response. Gantz et al. (2018) found that eDNA from small amounts of Elodea (0.25–25 g) could be detected with certainty in an aquarium, yet it is unclear whether it could be detected in a natural environment. To determine if eDNA monitoring of Alaska water bodies for incipient, early-stage Elodea invasions is feasible, an estimate of distances at which eDNA can be detected from recently introduced fragments is needed. Our objectives were to 1) evaluate the effectiveness of each of the four quantitative real-time Polymerase Chain Reaction (qPCR) assays for identifying Elodea from eDNA extracted from natural water samples, and 2) estimate the distance at which eDNA could be detected from Elodea plants in 19 L containers with a minimum of 80% probability of site occupancy and 95% confidence.

The United States Army Engineer Research and Development Center developed the four Elodea eDNA assays utilized in this study using standard approaches (ERDC 2017, Guan et al. 2019). We sequenced chloroplast DNA from a single specimen of E. canadensis from New York, one specimen of E. nuttallii from Washington, one specimen of an apparent E. canadensis x E. nutallii hybrid lineage from Missouri that provided a chloroplast sequence closely related to E. nuttallii, and one specimen of E. bifoliata from Montana. We then extracted whole cpDNA molecules from plant tissues following Mariac et al. (2000). Solutions of cpDNA from each sample were prepared for paired-end sequencing-by-synthesis (SBS) using Nextera XT Index Kits (illumina®, San Diego, CA), with unique nucleotide indexes attached to the sequencing libraries for each sample. We prepared duplicate libraries, each with a unique index, for each sample. We performed SBS on all 10 libraries using the illumina MiSeq system and 150 base-pair (bp) paired-end reads. We used MiSeq Reporter Software (illumina) to sort the resulting pool of sequences by index into separate data sets. For each sample (duplicates treated individually), we identified and merged paired DNA sequence reads in Geneious R9 (Biomatters Ltd., Auckland, New Zealand). We assembled chloroplast DNA genomes by aligning the reads to a complete cpDNA sequence for ELCA7 (NC_018541) found in the National Center for Biotechnology Information (NCBI) online genetic data repository GenBank (Benson et al. 2013), using the medium sensitivity/fast settings in Geneious R9. Aligned sequences were visually scanned for highly conserved regions for genus-level marker development and highly variable regions for species-level marker development. We designed forward and reverse primers, and associated probes, using Primer3 version 2.3.4 as embedded in Geneious R9 (Rozen and Skaletsky 1999). We then tested for potential nontarget cross-amplification issues using Primer-BLAST as provided online by NCBI (Ye et al. 2012).

We selected and tested an initial batch of 21 draft markers for relative sensitivity (to 1 ng/µl of cpDNA) from each of E. canadensis, E. nuttallii, and E. bifoliata. We measured the DNA concentrations of Elodea cpDNA extracts using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA) and used these concentrations to create 1 ng/μL cpDNA solutions in sterile, DNAse-free water. The qPCRs contained1X TaqMan® Environmental Master Mix 2.0 (Applied Biosystems, Foster City, CA), 0.5 μM of each primer, 0.125 μM of probe, 1 μL of the 1 ng/μl cpDNA template (final approximate mass of template DNA = 0.05 ng), and DNAse free/sterile water to achieve a total volume of 20 μL. We ran the qPCR thermal cycle program on a ViiA™ 7 Real-Time PCR System (Applied Biosystems) and included an initial 10-min denaturation step at 95° C, followed by 40 cycles of 95° C for 15 s and 60° C for 60 s. We ran each combination for the draft marker and Elodea cpDNA template as three replicate qPCRs. The mean cycle number at which fluorescence crossed the detection threshold (Ct) was used as a basis for comparisons of performance among markers. At the end of these trials, we eliminated 15 markers from further testing and carried six potential markers into specificity testing. We initially tested markers with 1 ng/μL concentrations of E. canadensis, E. nuttallii, and E. bifoliata to assess whether the markers would amplify the target Elodea species and not amplify nontarget Elodea. Marker-species combinations were tested with triplicate qPCR replicates, and the mean cycle number at which fluorescence crossed detection threshold (Ct) was used as a comparative metric for marker sensitivity. The 15 markers that were eliminated either failed to amplify the target Elodea species or amplified in both target and nontarget Elodea species. We conducted specificity testing with specimens collected from 15 nontarget co-occurring aquatic plants and one multicellular algae (Chara) across multiple locations with Elodea infestations representing Fairbanks Borough regions and Valdez-Cordova regions. We isolated whole genomic DNA (gDNA) from each plant sample using modified cetyltrimethyl ammonium bromide (CTAB) methodology (Doyle and Doyle 1987).

Following methods described in Farrington et al. (2015), we created standard test gDNA solutions of 1 ng/μL in sterile, DNAse-free water for each sample following DNA quantitation with the Nanodrop 1000. We then tested each sample against the remaining six potential markers using qPCR, 1 μl DNA template, and reaction protocols described above. Sample identifications were further validated by staff at the Missouri Botanical Garden and archived there as voucher specimens. Following this target testing of non-Elodea, we tested the six markers for intrageneric amplification using both E. canadensis and E. nuttallii plant tissue samples from multiple regions in Alaska (Cordova-Valdez, Fairbanks Borough, Yukon-Koyukuk Borough). We isolated whole gDNA from plant samples and tested using the qPCR conditions.

Our initial sensitivity and specificity testing narrowed the marker suite to four optimal performing markers, with two markers targeting Elodea plants in general, one marker targeting E. canadensis, and one targeting E. nuttallii (Table 1). These four markers were designated Elod-1, Elod-2, ELCA7-1, and ELNU2-1, respectively. We evaluated the final suite of four markers for limits of detection (LOD) using custom designed synthetic double-stranded DNA fragments (gBlocks®; Integrated DNA Technologies, Coralville, IA) and the discrete threshold protocol described by Klymus et al. (2020).

We collected water samples to field validate assays from two visible Elodea infestations at Chena Lakes near Fairbanks, Alaska (64.775°N, 147.232°W) on 28 July 2017, and Potter Marsh, Anchorage, Alaska (61.056°N, 149.797°W) on 6 February 2018. We introduced Elodea plants, sourced from Chena Lakes with an approved permit, to the 5.8-ha Small Arms Complex Pond (Test Pond; Figure 1) located on Fort Wainwright, Alaska, approximately 16 km from Chena Lakes, on 14 August 2018. The Test Pond is within the perimeter of a military installation with no unregulated public access. We contained the Elodea from Chena Lakes in two 19 L permeable containers; each were filled to 25% bucket capacity and transported to the Test Pond within 1 h of collection. Both Elodea containers were constructed with mesh sides and top (mesh openings were 2.5 mm by 2.5 mm) that allowed water to flow through the container and contained all plant fragments (Figure 2). One container was anchored at the northern side of the pond and the second container was anchored at the southern part of the pond (Figure 1). The tops of the containers were approximately 25 cm from the surface. The containers were opened and examined biweekly during summer to monitor plant growth and scrub algae off mesh to allow water to flow through the container.

We estimated sample sizes required to estimate probability of site occupancy as a function of distance from known Elodea sources using a multilevel occupancy model (Nichols et al. 2008; Mordecai et al. 2011; Schmidt et al. 2013). We simulated presence-absence data with an average 80% probability of site occupancy (site occupancy was simulated to decrease as a function of distance from Elodea source), 50–70% availability probability, and 70–90% detection probability. The 80% site occupancy was our statistical objective. We expected high eDNA occupancy near the Elodea sources based on the high occupancy rate of eDNA at known Elodea infestations. We selected these availability probabilities based on the minimum required to develop an Elodea eDNA monitoring program across the Alaska region, using our knowledge of inventory costs in roadless areas. The detection probabilities were selected based on our preliminary lab work from known infestations. We estimated the parameter estimates from each simulation, with varying sample sizes, results were stored, then repeated simulations 1,000 times. Confidence intervals were estimated for parameter estimates using the stored results. These simulation methods were similar to those used by Erickson et al. (2019), who provide a good overview of required sample sizes for eDNA inventories. Our results indicated we needed a minimum of 25 sampled grid cells, two water samples collected at each grid cell, with each water sample analyzed in triplicate for the PCR analysis step in the lab.

We used a 25 m x 25-m grid to define our sample frame on 27 September 2018 (Figure 1). Our sample unit was defined as a grid cell, where two 1 L samples were collected (hereafter, a sample is 1 L in volume). We used a stratified random sampling design. We sampled 21 grid cells within our shoreline strata (grid cells that included shoreline) and 9 grid cells in the non-shoreline strata (grid cells that did not include shoreline). This resulted in our closest sample being 8 m from an Elodea plant and our farthest sample being 135 m from a plant.

In 2019, we reduced the size of the grid to 12.5 m x 12.5 m, increased our sample size to 41 grid cells (Figure 1, Table 2), and sampled closely to the plants because Elodea eDNA was not detected in 2018. We continued with a stratified random sample, sampling 31 shoreline grid cells and 10 non-shoreline grid eDNA (i.e., cells). We also collected several other samples (n = 33), which we referred to as nonstatistical sample, to provide extra coverage in 2019. We collected two samples at 0 m (within 10 cm of the anchored container), 1 m, 3 m, and 5 m from the containers during the 28 August 2019 sampling event. Additionally, in 2019, 20 grid cells were sampled with a Smith-Root eDNA Sampler™ (formerly ANDe™) backpack filtration system. This backpack sampler allowed transect sampling; 1 L of water was collected by moving the transect sampler across the entire grid cell. Finally, we collected water samples on 3 March 2019, using an ice auger, to sample below the ice at 0 m, 1 m, and 3 m from the overwintered Elodea containers, which were anchored at a depth of 3 m to prevent housing from being crushed by ice.

We followed standard eDNA sampling protocols (Carim et al. 2016; Dunker et al. 2016; Evans and Lamberti 2017). We collected shoreline samples without entering the water and sampled non-shoreline locations from a canoe or rowboat. The point of entry for water into the collection device was approximately 20 cm below the surface (Newton et al. 2016). We tested 1 L samples of municipal water (blanks) periodically throughout each sampling day to ensure there were no false positive detections (n = 6 in 2018 and n = 8 in 2019 at the Test Pond, and n = 1 at Potter Marsh and Chena Lakes; Table 2). We filtered each 1 L water sample through a Whatman 1.2 µm glass fiber filter. After collection of each sample, we removed the collection device from the water while running the pump for approximately 30 sec to dry the filter. We did not allow the filter assembly and collection tube to touch anything out of context with the site (e.g., worker skin, clothing, or other anthropogenic items). We then used individually packaged, sterile forceps to remove and fold the filter (filter surface inward) prior to placement into a 50 ml sterile sample tube and sealed shut. We labeled the sample tube with date, site identification number, GPS coordinates, number of filters sampled, and collector initials and placed each tube in a separate, sealed sample bag. Additionally, new latex gloves were worn for each sample collected. We stored the bagged sample tubes in a cooler with an ice pack until taken back to a –18°C freezer (within 24 hours) at the United States Fish and Wildlife Service (USFWS) Fairbanks laboratory for temporary storage before being shipped to the USFWS Conservation Genetics Laboratory (CGL) in Anchorage, Alaska, for storage at –20°C.

We tested each sample for Elodea eDNA at the CGL, where we extracted DNA from each sample using The Qiagen DNeasy® Blood & Tissue Kit and Investigator® Lyse & Spin Kits (Qiagen GmbH, Hilden, Germany). We modified the standard DNeasy protocol to utilize the Lyse&Spin tubes for the filter digest stage. The entire filter was digested in adjusted volumes of 370 µL of ATL buffer and 25 µL of proteinase K for a final volume of 395 µL per sample. We added a total of 400 µL each of AL and ethanol to the supernatant following digestion and discarding of field filters. We followed the DNeasy handbook for volumes of Buffers AW1 and AW2. We adjusted the final dilution to 120 µL of Buffer AE at 55°C. All qPCR reactions were performed individually for each assay in 20µL volumes consisting of the following at final concentrations: 1X TaqMan Environmental Master Mix 2.0 (Thermo Scientific Inc.), 1X Exogenous Internal Positive Control (Exo-IPC), 1X Exogenous IPC DNA, 0.5 µM of each assay forward and reverse primer, 0.125 µM of each TaqMan MGB probe (6-FAM), 4 µl of DNA template and ddH2O for the remainder. The qPCR thermal profile was as follows: 50°C for 2 min and 95°C for 10 min, followed by 50 cycles of 95°C for 15 sec and 60°C for 1 min. All qPCR reactions were run in triplicate on a QuantStudio 12K Flex Real-Time PCR System (Applied Biosystems). A total of 12 non-template controls (4 μl diH20 in place of template) were randomly placed on each 96-well PCR plate for contamination detection.

We collected 10 water samples next to patches of visible Elodea and used qPCR assays to successfully identify E. nuttallii, E. canadensis, and their hybrids. At Chena Lakes, we detected Elodea eDNA using both genera markers and the E. nuttallii marker in five of five samples. At Potter Marsh, we detected Elodea eDNA from both Elodea genera markers and our E. nuttallii marker in five of five samples, and in three of five samples eDNA E. canadensis was also detected. These eDNA samples supported the morphological identification of species and hybrids at these locations.

We collected 186 water samples at the Test Pond during 2018 and 2019 (Figure 1, Table 2). The Elodea plants in the containers appeared to flourish at the Test Pond during both summers, with a bright green color and approximately doubling their volume. In 2018, Elodea eDNA was not detected in any of the 60 water samples. In 2019, Elodea eDNA was not detected in 124 samples and detected in only 2 samples collected within 10 cm of the containers. The additional, nonstatistical samples were also undetectable for Elodea eDNA on 3 March and on 28 August 2019, within 5 m of the containers, nor was Elodea eDNA detected in the 20 grid samples resampled using a transect method. Finally, there was no indication of cross-contamination among samples as Elodea eDNA was not detected in any of the field-blank samples.

We successfully used Elodea qPCR assays to detect eDNA from E. canadensis, E. nuttallii, and their hybrids from water samples. These new markers allow biologists a quick and relatively inexpensive method to verify species-level identification and flag putative hybrids for Next Generation Sequencing (NGS) determination. Our markers cannot be used to identify hybrids; however, our species-specific markers could be used as a surveillance tool to determine whether new outbreaks are exhibiting signal from one species exclusively or both, which may indicate hybridization. We could then use NGS to verify hybridization.

Although our results support the use of eDNA to identify Elodea species in visible infestations in Alaska, we found no evidence to support the use of an eDNA collection protocol for the early detection of Elodea in a small, cold-water lake in interior Alaska. Despite healthy plant growth, Elodea eDNA was only detected within 10 cm of the containers. Similarly, Kuehne et al. (2020) found low levels of detection and concluded there were substantial hurdles to use eDNA for early detection of aquatic invasive plants.

The lack of Elodea eDNA detections in our Test Pond was unlikely to arise from issues with PCR inhibition, which can occur in samples (Lance and Guan 2020), because Elodea eDNA was detected in samples taken from infestations at Chena Lakes and Potter Marsh. The lack of Elodea eDNA detections was not likely a function of assay sensitivity, as all the assays were demonstrated, during development and validation, to exhibit qPCR limits of detection (LOD) below two copies (i.e., molecules) of the target locus, well within the LOD range characteristic of effective eDNA assays (ERDC 2017; Klymus et al. 2020). These assays may be useful for surveys in systems with more established Elodea stands, in detection of this species in bulk samples from plankton net tows, or as a tool for identification of Elodea from tissue samples that cannot otherwise be identified.

We found that Elodea eDNA could be detected with a water sample within 10 cm of the contained Elodea in agreement with Gantz et al. (2018). However, beyond detections at the source, we did not detect Elodea eDNA elsewhere in the Test Pond. Although plants do not release constant amounts of DNA (Matsuhashi et al. 2016), both of our sampling events at the Test Pond occurred during autumn, which has been shown to be the peak of eDNA concentrations for E. candadensis in Lake Steinsfjorden in Norway (Angles d’Auriac et al. 2019). We hypothesized that detection of eDNA would decrease as distance increased from our contained plants. Had this been the case, there would have been many factors that could have been examined to understand their role in eDNA distribution within the Test Pond (e.g., depth, temperature, seasonal variation, etc.).

Our sampling was limited to inference at this single pond. Still, this study was a realistic application of using water samples to detect Elodea eDNA. In interior Alaska, lakes are reliably ice-free for no more than five months, and August sampling is likely the best temporal window to collect Elodea eDNA. Sampling in August also allows time for eDNA spread during the summer growing period but is before leaf fall and lake turnover, which typically occurs in September. Additionally, it would be difficult to conduct a more spatially robust experiment on Elodea-free lakes in an extreme cold-water environment like interior Alaska, because it was challenging to obtain permits to introduce an invasive species, despite security assurances to prevent spread. However, our results indicate that detection of Elodea eDNA from small stands is unlikely at any significant distance from the stand—a scenario that would be typical in early-stage establishment from an accidental release. Our study supports the assertion that eDNA surveys for early-stage Elodea invasions are unlikely to outperform current observation-based studies, such as conventional rake pulls and transect surveys.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content of functionality of any supplemental material. Queries should be directed to the corresponding author.

Reference S1. Carim KJ, McKelvey KS, Young MK, Wilcox TM, Schwartz MK. 2016. A protocol for collecting environmental DNA samples from streams. Gen. Tech. Rep. RMRS-GTR-355. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 18 p.

Reference S2.ERDC 2017. Monitoring invasive species on Joint Base Elmendorf-Richardson, Alaska, using eDNA technology. Prepared by the U.S. Army Engineer Research Development Center (ERDC) Environmental Laboratory, Vicksberg, MS. 75 p.

Data S1. Data collected for an investigation of Elodea eDNA detection in Alaska, from July 28, 2017 to August 28, 2019.

This project was funded by the United States Fish and Wildlife Service; the Conservation Genetics Laboratory and the Northern Alaska Field Office helped with sampling and supported the project. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the United States Fish and Wildlife Service. Joshua Buzby, Fort Wainwright Alaska, was instrumental in supporting field operations at the Test Pond. We would also like to thank several sample collectors: Peter Rice, University of Montana; Joel Swift, Missouri Botanical Garden; Gary Dick, US Army Research and Development Center; Jennifer Parsons, State of Washington Department of Ecology as well as the Journal Reviewers and Editors for their thoughtful and constructive review that improved this paper. This study was conducted on the traditional lands of the Dena’ina Athabascan and Tanana people who have been, and will continue to be, stewards of these lands since time immemorial.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.