Challenges of eDNA to track invasive species
Environmental DNA capture
Because traditional surveys detected the species only occasionally in only one or two locations in the Delta, it is clear from our results that eDNA reported higher sensibility with a low catch effort. eDNA-based methods to monitor invasive species in aquatic ecosystems have been designed and applied successfully to detect these species even at very low densities (Dejean et al. 2012, Smart et al. 2015, Baudry et al. 2021). In our study, this methodology allowed us to infer a more complete picture of the extent of the invasion process at the first stage, as has been reported in other previous studies (Ficetola et al. 2008a, Dejean et al. 2012, Everts et al. 2021). Moreover, the logistical requirement for eDNA sampling and the persistence of eDNA in the environment beyond the presence of the species, are also arguments strengthening the use of eDNA methods. However, similarly to other monitoring methods, eDNA methodology has also several critical points along the whole process, from sampling to the interpretation of the data (Darling et al. 2011).
A first challenge of eDNA assays is water filtration with cellulose nitrate filters. Filters with small pore size are strongly recommended for eDNA sampling as they optimize eDNA capture at low concentrations (Majaneva et al. 2018; Takahashi et al. 2020). However, in turbid water bodies with a lot of organic matter or suspended sediment, filters clog quickly and filtering rate is so slow that it is impossible to filter an optimal volume (at least 500 ml), specially required when eDNA is scarce. Several options have been suggested: increasing pore size, pre-filtering samples or reducing water volume of samples. All of them lead to lower yields of target DNA (Majaneva et al. 2018; Hunter et al. 2019). Alternatively, Hunter et al. (2019) increased the filtered volume and got higher DNA yield by combining several filters in a single Phenol-Chloroform-Isoamyl DNA extraction. We adapted this alternative solution using the DNeasy Blood & Tissue Kit of Qiagen, which is recommended for eDNA (Jeunen et al. 2019). The total desired volume was filtered using as many filters as necessary (two in our case) which were then preserved and processed separately until the eDNA was transferred to the spin columns of the kit. Thus, the digestion volume of the two filters belonging to the same location was collected in a single DNeasy Mini spin column. This modification in the DNA extraction protocol allowed us to recover the eDNA from a total 500 ml volume avoiding problems of filter clogging.
PCR and false positive/negative results
The design of primers may be related to false positive and false negative results of the eDNA protocols. In this sense, the length of the fragment might be a critical issue. On the one hand, if the fragment is too short, the risk of amplification artefacts and sporadic contamination (both causing false positive results) is higher. On the other hand, too long DNA sequences are prone to false negatives as long templates do not persist in the environment (Deagle et al. 2006). Therefore most published papers with water sampling have PCR fragment sizes shorter than 150 bp (Dejean et al. 2012; Takahara et al. 2013; Wilcox et al. 2013).
Although we took precautions to reduce the risk of contamination (DNA extraction in a separate room, PCR in UV-chamber, amplification of negative and positive controls), we detected cross-contamination in some negative controls that make us exclude some ‘positive’ results and use alternative primer sets for specific PCR.
False positive results can also be due to the persistence of eDNA when the species has already disappeared from the water body. However, Dejean et al. (2012) proved that bullfrog eDNA persisted in freshwater ecosystems a maximum of two weeks after animal removal.
Nevertheless, the main problem to face is to avoid false negatives by the presence of PCR inhibitors. This is particularly concerning when sampling turbid waters such as those from wetlands in Ebre Delta. Several protocols have been proposed to improve eDNA yield, such as adding chemical compounds or performing mechanical processes to remove inhibitors during the DNA extraction (McKee et al. 2015; Hunter et al. 2019). An alternative solution is the dilution of eDNA extractions to reduce inhibitors. This is an easy method that does not have the economic cost of removing potential PCR inhibitors. However, this approach can be problematic when DNA concentrations are low, because diluting the extractions also reduces DNA concentration and hence the sensitivity of the PCR assay (Goldberg et al. 2013; McKee et al. 2015). The negative effect of inhibitors can also be assessed using a second PCR with universal primers. In our case, a set of unspecific amphibian primers (16SA-L and 16SB-H) was used to amplify all eDNA extractions. As other species of amphibians are expected in all the sampled locations, negative results were indicative of inhibition. Our positive results from universal PCR show that dilution avoids inhibitor effects in all samples. Previous results from McKee et al. (2015) show that a 10-fold dilution is enough to reduce qPCR inhibition effectively. In most of our samples a 1:100 dilution was necessary to avoid the effects of inhibitory compounds. As we will discuss just below, this 1:100 dilution did not compromise the overall sensitivity of the PCR assay because several replicates were simultaneously amplified (at least 2 for each dilution). Therefore, the possibility of false negatives due to the random and unequal distribution of the very few DNA molecules in the dilutions was avoided.
Replicates and threshold of positivity
The last critical point when using eDNA is the suitable number of replicates and the threshold of positive tests to consider the presence of the species to be certain. The detection of alien invasive species relying solely on DNA-based methods has been controversial, especially when detection can result in costly management implications (Darling and Mahon 2011). In this context, performing an optimal number of replicates to avoid missed detections and setting the minimum number of positive replicates to avoid false positives are both strictly necessary. To assess these parameters, previous studies calculate detection probabilities of eDNA analyses. However, this is only possible when eDNA results can be compared with traditional methods of detection outcomes or with experiments in controlled conditions (Ficetola et al. 2008a; Dejean et al. 2012). In our case, since the eDNA studies started, visual and acoustic surveys detected only a couple of specimens in a very localised area. Therefore, the detection probability of our PCR assay could be compared to traditional methods but did not be able to be tested empirically.
In general, the detection probability is not high when the density of the species is low. Ficetola et al. (2015) recommended at least eight PCR replicates to avoid false negatives when the detection probability is lower than 0.5. Goldberg et al. (2013) conducted controlled experiments with different densities of the invasive New Zealandmudsnail (Potamopyrgus antipodarum) and used three replicates for each density treatment to reach the detection of even 1 individual in 1.5 L of water. In these scenarios with several replicates per sample, it is important to consider the possibility of crossed or sporadic contamination. Thus, to avoid false positives, Taberlet et al. (1996) recorded an allele only if it was observed at least in two out of 10 replicates when they analysed samples with very few DNA. More recently, Ficetola et al. (2015) suggested the same strategy in eDNA metabarcoding studies, remarking that a sufficient number of replicates was necessary to avoid false negatives with low detection probabilities.
According to the pipeline described in methods to validate results and considering all the replicates together, 14-16 specific PCRs per location were performed in 2020 survey. Of these 16 amplifications, negative results in most of the 1:10 dilutions suggest the presence of PCR inhibitors at this dilution. Accordingly, replicates of 1:10 dilutions were abandoned and the analysis was performed with 12 replicates instead. Following the recommendations from previous studies (Ficetola et al. 2015), positive identification was called when the specific PCR amplified at least two of the 12 probes. Under these criteria, there were two positive locations in the June 2020 survey and seven in the July survey.
Moreover, Sepulveda et al. (2020) suggested using different PCRs targeting different genomic locations. This would increase the reliability of positive detections. In our case, we used an alternative primer set of the specific PCR (cytbF2 + cytbR1) to confirm positive amplifications of American bullfrog eDNA in samples of the most recent survey (July 2020). This approach confirmed the detection of American bullfrog in five locations (Table 3).
Early invasion of the American bullfrog in the Ebre Delta revealed by eDNA
According to the arguments so far discussed, several restrictions should be and have been applied in the interpretation of the undertaken American bullfrog eDNA assay in the Ebre Delta. Even under the most conservative scenario, we can confirm the presence of this species in at least five locations in the last survey (July 2020, Fig.1). The detection of the species in several locations through eDNA analyses contrasts with a very local detection of only two individuals through visual or calling surveys. This suggests an early first stage of the invasive process (Jerde et al. 2011), and it shows once again that eDNA assays improved detection sensibility with respect to the traditional methods, with a much lower sampling effort (Dejean et al. 2012) .
Interestingly, the higher number of positive replicates was found in the Ground zero area, where the first bullfrog tadpoles were observed in June 2018 (Table 1 and Fig. 1). However, the eDNA sampling of the 2019 year was negative, and the presence of this species in the Delta was not reported again until June 2020, in a place seven kilometres from Ground zero. In 2020, a single deceased specimen was found in the DNA10 location, and the eDNA analyses confirmed this incipient introduction of this species again. Curiously, the eDNA survey of June 2020 failed to detect the species in the DNA10 location, but it was detected in DNA9, where waters from the DNA10 region are collected. Within the Ground zero area, a single deceased specimen was found and another individual was heard in July 2020, after the last eDNA survey (Table 1). These history records and the eDNA results might suggest two alternative hypotheses regarding the American bullfrog invasion in the Ebre Delta. First, it is possible that the eradication plan carried out in 2018 in Ground zero area was sufficient to eliminate tadpoles, but some post-metamorphic terrestrial individuals survived and escaped from this area before the construction of the metallic fence was completed (end of July 2018). Then, if few individuals survived but they did not establish, bullfrog density in 2019 could have been not enough to be detected even by eDNA analyses (either because the concentration of eDNA was too low or because sampling was not extensive enough). Detections in 2020 should then be attributed to these survivor individuals. The concentration of positive detections in the Ground zero area in 2020 survey,could be suggesting that two years after the first introduction, the invasive American bullfrog persists in the first site of detection or probably it comes back to the original site of introduction to reproduce. The fact that 29 post-metamorphic juveniles (> 200 g weight) were captured in autumn 2018 outside the fence surrounding the Ground zero area, support this first hypothesis.
Alternatively, it is possible that the 2018 eradication plan was successful and American bullfrog was eradicated. Therefore, the two specimens found and the positive eDNA results in 2020 would correspond to a new introduction process. In other countries, reiterated intentional releases of the American bullfrog have been documented (Ficetola et al. 2007a; Kamath et al. 2016), and the same could be taking place in Ebre Delta. European legislation prohibits introductions of American bullfrog and its commercial farming is completely forbidden in the Iberian Peninsula (Royal Decree 630/2013, 2 August). However, this legislation could change if the species was already established. For instance, the Autonomous Government of Catalonia (SRM/1/2019, 17 May 2019) has changed its law and recently allowed the commercialization of the Blue crab Callinectes sapidus, another invasive species in the Ebre Delta (Farré et al. 2021). Therefore, a hypothetical premeditated release of bullfrogs could be related to the gastronomic and economic potential of the American bullfrog commercialization and the possibility of a law change if the species was already introduced.
The failed attempts of American bullfrog establishment in the Delta should not lower the guard. Blackburn et al. (2015) link the success of the three stages of the invasion process (introduction, establishment and spread) to several critical aspects. Specifically, the number of invasive specimens plays an important role at the first stage of the invasion, because a larger number of transported individuals will increase the probability of success. The number of arrived individuals is also important at the second stage, as a small number of individuals leads to a reduced genetic diversity that can compromise the process of adaptation to the new habitat. For the American bullfrog, Ficetola et al. (2008b) suggested that an extremely low number of founder specimens can be enough for a successful invasion process. For instance, it is estimated that the Italian invasive population descended only from two females and one male introduced in 1930. The success at the second stage also requires that the new habitat has similar conditions to those of the native one (Suarez and Tsutui 2008). Ficetola et al. (2007b) suggests that certain environmental factors (mainly related to climate) are critical to determine the probability of the establishment of the American bullfrog. The projection for the environmental suitability for bullfrog made by these authors (see Fig. 3 in Ficetola et al. 2007b) indicates the south and the west of the Iberian Peninsula as less suitable by bullfrog establishment. This could explain that despite some individuals have been reported in this region, they have never become invasive (Urioste and Bethencourt 2001; Pleguezuelos et al. 2002; Ficetola et al. 2007b). However, the situation is quite different in the northeast (including Ebre Delta region), where environmental suitability for the bullfrog establishment reaches up to 50/100 according to Ficetola et al. (2007b).
As a conclusion, eDNA allowed us to delimitate the extent of the invasion of the American bullfrog in the Delta, yielding a higher sensitivity with lower sampling effort than traditional methods. In this context, eDNA assays are essential tools to facilitate detection, control and eradication of the species at the first stage of the invasion process in the Ebre Delta. Even at low population density, the American bullfrog may represent a high level of risk for the Ebre Delta ecosystem, a fragile ecosystem already endangered by climate change and the establishment of other invasive species (Gilioli et al. 2017; Zografos 2017; Farré et al. 2021). In such a situation, Darling and Mahon (2011) stated that, despite controversial arguments, DNA-based methods might be the only tool to promote management actions prior to assumption of unacceptable invasion risks.