Species-Specific Assay Development

We designed a primer set and a dual-labeled hydrolysis probe for use in qPCR detection of M. temminckii eDNA (Table 1) using 11 mitochondrial control region sequences representing populations from across the known distribution (Roman et al. 1999; Echelle et al. 2010). Species-specific primers were designed using Primer-BLAST software (Ye et al. 2012), and were screened for compatibility and secondary structure using OligoAnalyzer (Owczarzy et al. 2008). The dual-labeled probe was designed in OligoArchitect Online (Sigma-Aldrich). Because the sequence of interest was rich in adenine and thymine nucleotide bases, we implemented locked nucleic acid technology in the probe design to meet melting temperature requirements.

Table 1. Macrochelys temminckii quantitative polymerase chain reaction assay targeting an 84-base-pair section of the mitochondrial DNA control region. The probe is dual-labeled with a reporter dye and a quencher, and uses locked nucleic acid (LNA) technology. Capital letters in the probe sequence indicate LNA bases, while lowercase letters indicate standard DNA bases.

View Fullscreen

A BLAST search (Clark et al. 2016) indicated that the common snapping turtle (Chelydra serpentina) is the only potentially sympatric species that exhibits substantial genetic similarity to M. temminckii. Thus, we used Geneious v.8.1.9 (Kearse et al. 2012) to align and visually inspect sequences representing C. serpentina against our designed assay and verified the presence of 5, 3, and 3 nucleotide mismatches between C. serpentina and our forward primer, reverse primer, and probe sequences, respectively. The observed number of nucleotide mismatches should prevent amplification of nontarget DNA (Wilcox et al. 2013). However, to increase confidence in assay specificity, we obtained tissue-derived DNA (using either muscle from the tail or rear-foot webbing) from C. serpentina, and 2 additional sympatric species, the red-eared slider (Trachemys scripta) and the painted turtle (Chrysemys picta), and subjected the DNA to qPCR using the designed assay. For these reactions, we diluted the nontarget DNA to 2 concentrations (0.1 ng/μl and 0.01 ng/μl) and used the optimized qPCR protocol detailed below.

Assay Evaluation

To evaluate both the efficiency and sensitivity of our assay, we generated a qPCR standard curve using tissue-derived (from webbing of rear foot) M. temminckii DNA of known concentration (measured with Nanodrop fluorospectrometer, ThermoScientific) which had been serially diluted at 6 levels from 0.4 to 4 × 10⁻⁶ ng/μl. We used tissue-derived DNA, rather than diluted PCR product, because we wanted to mimic concentrations likely to be found in positive eDNA samples and avoid introducing a source of potential contamination.

Each qPCR was prepared in an ultraviolet (UV)–sterilized hood within a clean room never exposed to amplified DNA. Reactions were optimized for ideal reagent concentrations, and incorporated Taqman Environmental Master Mix—a reagent effective in combating potential issues with inhibition which are known to occur in eDNA samples collected from systems with high levels of suspended organic particulate matter (Cao et al. 2012; Jane et al. 2015). Total volume equaled 20 μl, and consisted of 10 μl Taqman Environmental Master Mix, 0.8 μl of each 10 μM primer (forward and reverse), 0.4 μl of 10 μM probe, 4 μl PCR grade H₂O, and 4 μl DNA/eDNA extract. The cycling protocol began with 10 min at 95°C followed by 55 cycles of 95°C for 15 sec and 60°C for 60 sec. Each dilution was run in triplicate on a 24-well PikoReal real-time PCR system (ThermoScientific), with amplification (i.e., fluorescent signal) observed using PikoReal software 2.2 (ThermoScientific). Using the automatic threshold option, the same software was used to calculate cycle threshold (C_T) values, or the number of qPCR cycles at which a sample crosses a fluorescence threshold and is considered positive for amplification. In testing the sensitivity of our assay, we examined C_T values at each DNA concentration and assessed the limit of detection (LOD, the minimum concentration of eDNA that can be reliably detected in a sample).

Aquaria Trials

We tested our assay for use in eDNA detection by collecting water samples from aquaria at the Mississippi Museum of Natural Science, Jackson, Mississippi. One tank (approximately 1893 l) housed an adult M. temminckii. A second tank (approximately 757 l) contained a number of potentially co-occurring turtle species representing 5 genera: Apalone, Graptemys, Pseudemys, Sternotherus, and Trachemys, but no M. temminckii. One liter of water was collected from just below the surface of each tank using new, sterile plastic bottles, which had been rinsed 3 times in tank water prior to sample collection. Collected water was immediately filtered using 1.5-μm microfiber glass (MFG) filters (Whatman 934-AH, no. 1827-047) and a vacuum pump. This filter type and pore size were chosen to facilitate time-efficient filtering of future field samples, which were expected to be turbid and contain organic materials (e.g., algae). Promptly after filtering, DNA was extracted from one-half of each filter using an adjusted DNeasy (Qiagen) method developed by Goldberg et al. (2011). All aspects of the Goldberg method were adhered to except we omitted the overnight drying step and excluded the use of costly Qiashredder columns (Qiagen). Filter remnants not used during extraction were preserved for future use, if needed, in absolute ethanol. qPCR protocols were identical to those outlined above for assay evaluation. Here, we incorporated a positive control (tissue-derived M. temminckii DNA at 0.0004 ng/μl) and 3 negative controls (filtration, extraction, and PCR).

Field Trials

We tested, optimized, and validated our laboratory and field methods in situ using water samples collected from both a lentic system (surface area = 1.98 ha) and multiple lotic systems (as described below) in which M. temminckii presence was either known or suspected. Water samples were collected April–September when turtles were expected to be active. Because evidence suggests increased water levels and increased flow rates may result in decreased eDNA detection (Jane et al. 2015; Klymus et al. 2015; Cannon et al. 2016; Gingera et al. 2016; Wilcox et al. 2016), no sampling occurred after significant rainfall events when water levels and discharge were abnormally high. Following filtration and extraction, each water sample was subjected to qPCR using either triplicate or sextuplicate replicate reactions, with both positive and negative controls included for quality assurance. A water sample was considered positive for M. temminckii eDNA if 1) the sample amplified in at least 1 of the qPCR replicates, and 2) associated negative controls (filter, extraction, PCR) exhibited no amplification. Samples that failed to amplify across the total number of qPCR replicates (0 positives) were considered to either contain no M. temminckii eDNA, or to contain eDNA concentrations below the LOD.

During the first phase of field trials, we followed the sampling, filtering, and qPCR protocols described above in aquaria trials. Although these initial attempts at eDNA detection were successful in the lentic system, our methods did not produce a positive eDNA signal in the presence-known lotic system. Thus, we optimized methods for use in flowing systems using an experimental approach (Table 2) based on methods and recommendations from previous eDNA studies (Santas et al. 2013; Baldigo et al. 2017) where we varied sampling (volume and location within stream channel), filtering (volume and number of filters used), extraction (number and portion of filters used), and qPCR techniques (combining/not combining DNA from different filters/extractions). During optimization, we worked within the Yockanookany River, a fifth-order stream approximately 18 m wide at the point of collection with an average real-time discharge for the date of collection measured at 36.6 m³/sec via US Geological Survey (USGS) gage 02484500 located < 70 m upstream of the collection site. We made an effort to collect samples downstream of what appeared to be potential M. temminckii microhabitat (i.e., submerged woody debris piles; Riedle et al. 2006; Howey and Dinkelacker 2009; Lescher et al. 2013). In addition, we began incorporating an exogenous internal positive control reagent (Exo IPC; Applied Biosystems) into all reactions to test for PCR inhibition. Partial or complete inhibition, if present, was detected as in McKee et al. (2015).

Table 2. Experimental approach used to determine optimal techniques for detection of alligator snapping turtle (Macrochelys temminckii) environmental DNA (eDNA) in lotic systems. Water samples were collected within the Yockanookany River. Bold values indicate detection success as observed across 3 quantitative polymerase chain reaction (PCR) replicates per sample. C_T = cycle threshold of detection.

View Fullscreen

Results from the experimental approach (Table 2) indicated that 2-l water volumes provided the simplest method for satisfactory eDNA results. Thus, we combined this optimal eDNA sampling method with a traditional M. temminckii trapping survey to validate our eDNA methods and to provide a comparison of survey techniques. This was done in the Chickasawhay River, a sixth-order stream with an average across-collection-sites width of 44.1 m and an average real-time discharge for the date of collection measured at 414.5 m³/sec via the nearest USGS gage (gage 02477500), located approximately 32 river-km upstream of the most upstream collection site. Here, 2-l water volumes were collected approximately 30 m downstream of baited 90–120-cm-diameter hoop nets (n = 16 traps). Nets were set the previous day and, to facilitate a blind-sampling effort and prevent potential contamination, were not checked for M. temminckii occupancy until all eDNA water samples (n = 16) had been collected. Water depths were sufficient to require sample collection by boat. Samples were collected from the bow of the boat and downstream of traps as the boat traveled upstream. All samples were kept on ice until filtration. During filtration, filters clogged (< 2 drips/5 sec) quickly, often before one-third of the sample could be filtered, resulting in impractically long filtration times. Thus, filtration was modified such that each water sample was filtered through 2 MFG filters until each clogged, regardless of total volume (Table 3). In accordance with results from our experimental optimization process (Table 2), the 2 filters resulting from each sample were combined into a single extraction, using one-fourth of each filter (for a combined halffilter). Because we wanted to test whether increasing the number of qPCR replicates would increase detection, each eDNA elution underwent qPCR in both triplicate and sextuplicate reactions. Subsequent eDNA detection results from our blind samples were compared with occupancy results from the associated trapping effort.

Table 3. Macrochelys temminckii detection results within the Chickasawhay River where eDNA and traditional trapping methods were combined and compared. Bold values indicate detection success. eDNA detection refers to the number of positive amplifications observed across qPCR replicates per sample. Samples that were positive for the first round of triplicate reactions did not undergo the second round of reactions (see Table 2 for definitions of abbreviations).

View Fullscreen

To test the practicality of an eDNA survey, and to demonstrate utility of optimized methods (“proof of concept”), we emulated an actual eDNA survey within what is likely to be a typical field setting. Here, 7 woody debris piles were chosen randomly from a 5.26-km stretch of the Pearl River, a sixth-order stream with an average across-collection sites width of 74.6 m and average real-time discharge for the date of collection measured at 73.2 m³/sec via the nearest USGS gage (gage 02486000) located 6.6 river-km downstream of the most downstream collection site. As before, 2 l of water were collected at each site, just below the surface, and from the bow of a boat while moving upstream. In order to maximize the likelihood of M. temminckii eDNA detection, we modified sampling such that water was collected within close proximity (≤ 1.5 m downstream) of these suspected microhabitats. Following the optimized protocol, samples were again filtered through 2 MFG filters until each clogged, regardless of total volume (Table 4). Extractions were modified to include Qiashredder columns, as originally outlined in the Goldberg extraction method (Goldberg et al. 2011), but which had been previously excluded in this study due to cost. Because our goal was to maximize the likelihood of detection, and because Qiashredder columns were shown to increase eDNA detection in Goldberg et al. (2011), we considered elevated costs to be an acceptable tradeoff when encountering turbid waters. Once again, each eDNA elution underwent qPCR in both triplicate and sextuplicate reactions.

Table 4. Macrochelys temminckii eDNA results using targeted sampling to mimic an actual eDNA survey in which optimized methods were used. Here, 2-l water samples were collected downstream of suspected microhabitat (woody debris piles) in the Pearl River. Bold values indicate detection success. eDNA detection refers to the number of positive amplifications across qPCR replicates per sample (see Table 2 for definitions of abbreviations).

View Fullscreen

Species-Specific Assay Development

The species-specific assay, MacTem_84eDNA (Table 1), amplifies a short, 84-base pair segment of the mitochondrial control region belonging to M. temminckii and does not amplify in potentially sympatric, nontarget turtle species, including the genetically similar C. serpentina. The primer set is a perfect match to all M. temminckii haplotypes screened, except Haplotype K, which represents the Suwanee River lineage in Florida (Roman et al. 1999; Echelle et al. 2010; Thomas et al. 2014). When compared with Haplotype K, a single nucleotide mismatch exists within the reverse primer sequence and within the probe region, which may prevent efficient use in that lineage. A pilot study is recommended before use in the Suwanee River region and before use in any region not represented within Roman et al. (1999) and Echelle et al. (2010).

Assay Evaluation

The assay was determined to be reliable and consistent with an efficiency of 92.29% (y = −3.522x + 23.448; R² = 0.97). Observations from the standard curve indicated that DNA concentrations between 4.0 ng/μl and 4 × 10⁻⁶ ng/μl were associated with C_T values between 21.22 and 42.60, with C_T values increasing as expected with decreasing eDNA concentrations. Positive detections were observed across all replicates within the standard curve representing each DNA concentration, except in 4 × 10⁻⁶ ng/μl, where amplification was observed in only 2 of the 3 replicates (C_T: 40.67 and 40.58). Furthermore, 4 × 10⁻⁵ ng/μl exhibited a wide spread of C_T values, with amplification observed at 38.15, 41.17, and 42.6 cycles (4.45-cycle difference between the largest and smallest C_T), whereas minimal variation (between 0.09- and 0.51-cycle difference) was observed among C_T values representing replicates at all other DNA concentrations. This suggests that eDNA concentrations at 4 × 10⁻⁶ ng/μl may be approaching the LOD. Thus, eDNA concentrations < 4 × 10⁻⁶ ng/μl in future field samples may amplify less consistently (or not at all) across qPCR replicates, with a potential for false-negative results.

Trials

The positive and negative controls in all qPCR reactions performed as expected with positives resulting in observable M. temminckii amplification and negatives resulting in no observable amplification. No inhibition was detected in trials incorporating Exo IPC. During validation of the assay, we successfully recovered M. temminckii eDNA from aquarium water, with positive amplification observed across all wells in the triplicate reaction using a 1-l volume water sample (C_T: 33.52, 33.55, 34.12). As expected, our species-specific assay did not amplify eDNA in aquarium samples from the tank that housed several turtle species, but no M. temminckii.

Positive eDNA detection was accomplished in the lentic system using 1 l of water (positive across 2 of 3 qPCR replicates, C_T: 42.30 and 49.27). Lotic systems, however, required further optimization of sampling, filtering, and qPCR techniques before satisfactory results were obtained. Using a combination of techniques outlined in Tables 2–4, positive eDNA amplification was observed across lotic systems representing streams of various size and flow/discharge. When lotic field samples were positive for M. temminckii eDNA (Tables 2–4), C_T values ranged from 39.71 to 51.19 (average C_T: 41.79) and resulted from filtered water volumes ranging between 850 and 2000 ml (average volume = 1200 ml). No sample, when positive, amplified across all of its associated qPCR replicates; when M. temminckii eDNA detection occurred, the most commonly observed number of positive amplifications across replicates was 1 (i.e., 1 positive detection across either a triplicate or sextuplicate reaction). Increasing the number of qPCR replicates resulted in improved detection, with at least a 100% increase in the observed number of eDNA-positive samples (Tables 3 and 4). During the comparison of survey methods (Table 3), both triplicate and sextuplicate qPCR reactions (eDNA) outperformed traditional trapping methods in detecting M. temminckii, with positive eDNA detections occurring even when traps were unoccupied. However, in one instance, sextuplicate qPCR reactions were required before agreement between eDNA methods and trapping was achieved (i.e., an occupied trap resulted in a positive eDNA detection), suggesting that fewer qPCR replicates may lead to false-negative results.