As the performance of primer/probe assays can be inconsistent between study systems, formal validation, preferentially testing on local natural eDNA samples, should be the first step in developing large-scale eDNA-based surveillance protocols to ensure that the generated findings are reliable for the resulting decisions to be pertinent29,30. One of the most important aspects that has to be considered when evaluating the performance of a primer/probe assay is its specificity, i.e. the degree to which it can specifically detect the target sequence without interacting with genetic material of closely related sympatric non-target species. Specificity evaluation is especially important in the context of aquatic eDNA, as its highly degraded nature necessitates the use of shorter sequences as barcodes, which inherently harbors fewer species-specific regions19,27. Although primer specificity of the Goldberg and the Lin assay were previously excessively tested in silico for potential cross-amplification with co-occuring non-target species in the Western US and the Beijing area, respectively29,30,50, it is imperative to repeat the analyses for amphibian species inhabiting a particular study area. The in silico tests performed here confirmed that both primer/probe assays specifically detected the bullfrog sequence without interfering with sympatric amphibian species indigenous to Western Europe (Supplementary Fig. S1).
Nevertheless, in vitro and in situ tests should complement in silico specificity tests, since primers can interact with non-target sequences regardless of nucleotide mismatches26,51 and the outcome of amplification can also significantly depend on sample and system specific PCR conditions52. Although primer performance was already comprehensively evaluated in vitro29,30,50, we conducted additional in situ tests on natural eDNA samples containing DNA originating from a mixture of sympatric indigenous amphibian species under variable environmental conditions (Supplementary Table S1), which provides a more pragmatic indication of an assay’s specificity in the geographic area of interest relative to an in vitro approach28. When both assays were run on water samples collected from bullfrog-free water bodies harboring most of the co-occurring amphibian species in Western Europe, no bullfrog eDNA amplification could be observed for either of the two assays under study (Fig. 2), while the IPC and the bullfrog positive reference samples were always successfully amplified. Therefore, it can be safely stated that both assays specifically target and detect bullfrog eDNA in Western European water bodies, independent of sympatric non-target amphibian species.
A second aspect that determines primer/probe assay performance is its sensitivity, i.e. its ability to detect even the slightest traces of bullfrog eDNA. When the two primer/probe assays under study were subjected to a range of eDNA samples collected from water bodies that are currently invaded by bullfrogs with varying intensities, the two assays did not significantly differ in eDNA concentration quantified per sample (Fig. 2, Fig. 3a). However, in terms of detection resolution of both target and IPC (i.e. the separation of the target-positive and target-negative droplets), the Goldberg assay outperformed the Lin assay (Fig. 3b,c). Detection resolution is an important feature for ddPCR analyses, especially when water samples are supplemented with inhibitory compounds that can obstruct the differentiation of positive and negative droplets31,53,54. The higher detection resolution of the IPC when ran in duplex with the Goldberg assay was, at least partly, expected, as the optimal annealing temperature of the primers and probes targeting the IPC is equal to that of the Goldberg assay (i.e. 60 °C). Nevertheless, this implies that the correction factor based on IPC measurements, and thus final eDNA concentrations, can be more accurate when adopting the Goldberg rather than the Lin assay when combined with this particular IPC used in this work. Altogether, our results thus indicate that for the ddPCR approach applied, the Goldberg assay is more robust in detecting bullfrog eDNA in Western Europe compared to the Lin assay, which led us to select this assay for further analyses.
The controlled mesocosm experiment showed that the Goldberg assay was able to quantify eDNA signals according to bullfrog abundance and biomass. Especially at the tadpole life stage, the obtained eDNA concentrations predicted the tadpole abundance remarkably well (with 99% of the variance explained), whereas this relation was somewhat less pronounced at the juvenile life stage (Fig. 4a,b). This strong correlation observed at the tadpole life stage was well above the in vitro average of 0.82 as reported in a meta-analysis32, suggesting relatively homogenous DNA discharge rates among bullfrog larvae and densities compared to other species. Besides, DNA quantification via ddPCR has been shown to be more precise than measurements via qPCR31,54, the latter being overrepresented in the meta-analysis, and can additionally explain the higher predictive power observed, especially in combination with the inclusion of IPC’s in our workflow. Juvenile bullfrogs, on the other hand, are not permanently submerged in the water column and often reside on land near the water body or on aquatic vegetation as larval gills are functionally exchanged for lungs upon metamorphosis. Since our mesocosms were equipped with terrestrial islands for juveniles to dwell out of the water, this might have introduced more variation in the association between eDNA concentration and juvenile bullfrog abundance. Nevertheless, at a per individual basis, mean eDNA emission rates did not differ significantly between both life stages studied (Fig. 5a), indicating that, at least under controlled conditions, the Goldberg ddPCR primer/probe assay provides a relative accurate prediction of bullfrog abundance in the water column irrespective of life stage. The per gram eDNA emission rate, on the other hand, was significantly higher for the lighter juveniles relative to the heavier tadpoles (Fig. 5b), a pattern that also emerged in other studies such as with bluegill sunfish (Lepomis macrochirus), where heavier adults on average had a higher per individual eDNA release rate, whereas their per biomass eDNA release rate was lower relative to the lighter juveniles55. This finding, in conjunction with the highly variable biomass distribution among individuals in bullfrog populations, makes it more appropriate to predict bullfrog abundance rather than biomass from eDNA signals picked up in blind systems.
Under field conditions, however, relationships between abundance and eDNA concentration can be affected by other factors that introduce additional variation. A recent meta-analysis32 showed that species abundance under natural conditions explained a substantially lower proportion of the variation in eDNA concentrations (0.57 on average). This can be attributed to the complex interplay of several factors affecting eDNA production and degradation rates, such as temperature, UV exposure, pH, microbial activity50,56,57, animal behavior37, and season- and age-dependent shredding rates38,55. In addition, the accumulation of particulate and dissolved substances (such as calcium ions, humic and tannic acids, polymers, etc.) and eDNA from non-target species can further hamper target DNA amplification58, resulting in underestimations of the actual abundances of target organisms. However, the amplification in ddPCR appears to be less susceptible to inhibitory substances than qPCR, and is therefore expected to be more robust, especially when the target DNA template is scarce33,35,59. Besides, the inclusion of IPCs in our workflow may contribute to the standardization of variation in eDNA concentrations resulting from sample-specific suboptimal DNA extraction or ddPCR inhibition28,60 (see35 for more details). Altogether, these findings allow the in vitro established relation to be extrapolated to natural lentic systems in order to acquire rough indications of bullfrog abundance independent of their life stage, given an integrated and standardized sampling design. Nevertheless, it would be useful to apply this assay in parallel with conventional monitoring programs and eradication campaigns to further validate its ability to provide an approximation of local population density in natural ponds.
Since eDNA patterns in the field are expected to be influenced by seasonal alterations in both the ecology of the target species and its environment37,56,57, we assessed year-round temporal variation in bullfrog eDNA concentrations for two subsequent years in three natural ponds. Over this eighteen month study period, the three ponds showed similar patterns in seasonal variation in eDNA concentrations: a consistently low concentration from December to July preceded one peak in late summer (between August and October), followed by a decrease during winter time (Fig. 6). These temporal eDNA patterns largely reflect the seasonal phenology of bullfrog populations in Western Europe, where mating peaks during late spring and early summer and a new generation tadpoles most commonly emerges between July and September61. Reproduction in amphibian species has been documented to result in two peaks in eDNA concentrations in natural conditions, one representing the reproductive behavior of the adults and the subsequent mass release of gametes, followed by a second peak representing the emergence of a new generation of tadpoles38,62,63. Since we sampled our study ponds on a monthly basis whereas the timing between bullfrog breeding and larval emergence ranges between a few days up to one week64,65, it is plausible that we could not differentiate these two peaks. Moreover, given that the breeding season of bullfrogs spans a few months, and that multiple breeding events can occur in the same water body64,65, mating and emergence of bullfrog larvae presumably cannot be differentiated in time in terms of eDNA concentrations. On the other hand, it can also be assumed that the studied ponds served as refuges for first-year juveniles escaping competition and predation from congener adults in nearby breeding ponds, or as stepping stones or foraging sites44,48,65,66. Juveniles were indeed observed in large numbers in and around each of the three study ponds during summer in both years. A minimal influence of bullfrog adult presence on the observed peaks in eDNA concentrations was expected, since adults mostly reside out of the water on the shorelines of the water bodies during the breeding season65. It should also be noted that the observed peaks in eDNA concentrations could have been intensified by the extreme dry summers in 2019 and 2020, which resulted in strong declines of the water table.
As temperatures decrease in autumn and cross the threshold of 15 °C, which generally occurs from October onwards, bullfrogs enter winter torpor. They mostly hibernate in the water at the bottom of a pond, semi-immersed in mud, but winter lethargy or adopting terrestrial hibernacula instead is not exceptional49,61. Therefore, the decreased metabolic rate, and hence eDNA release, during hibernation36, or the decreased abundance could explain the observed decrease in measured eDNA concentrations following the peak in late summer. Although previous research was unsuccessful in detecting eDNA of two hibernating endangered frog species in headwater streams during the winter season39, we detected bullfrog eDNA all year round in the studied ponds. This suggests that our intense sampling strategy and efficient primer/probe assay could pick up even minute quantities of eDNA during the period bullfrogs are inactive or scarcely present. Altogether, these findings confirm previous research that has shown that eDNA concentrations closely track the seasonal phenology of the target species, and indicate that eDNA signals obtained from summer sampling most accurately predict bullfrog abundances present in these systems. The patterns observed here could thus provide crucial information on the timing when large-scale eDNA detection campaigns and eradication programs targeting the bullfrog would be most fruitful 36,39,63,67.
With this work, we report the validation process and exploratory research required for appropriately implementing species-specific eDNA approaches for reliable detection and quantification in large-scale monitoring campaigns. We tested two primer/probe assays specifically designed for bullfrog, and showed that one of them offered the highest detection resolution, which is an imperative feature for the early detection of this invasive species to be able to rapidly respond with the necessary eradication measures. The mesocosm experiments showed that the most robust ddPCR primer/probe assay provides a pragmatic approximation of bullfrog abundance independent of life stage, whereas under natural conditions remarkable seasonal eDNA patterns could be revealed in some permanently infested natural ponds. The outcome of our validation process suggests that this protocol is ready to be implemented in large-scale monitoring campaigns, in order to coordinate, evaluate, and eventually fine-tune such eradication programs67.