Assessing distribution of Didemnum vexillum in Scotland using environmental DNA

Discussion

Need for standardization of DNA-based tools and robust sampling strategies for use in marine biomonitoring

This pilot study was designed to explore use of eDNA to assess the distribution of D. vexillum, a significant marine invader, in Scotland following its introduction in 2009 (Beveridge et al., 2011). Like with any other novel monitoring tools used in the regulatory context, the need for standardized protocols is essential to ensure robustness of analysis and to allow for benchmarking and comparisons between different data sets. For eDNA-based surveys, some early guidelines, describing general principles related to study design and challenges associated with inferring presence of a species from eDNA data exist (Goldberg et al., 2016). More recently, new guidelines describing the options and choices which can be made at each step of the process, from sample collection and through to laboratory workflow, have been published (Bruce et al., 2021), and together with principles of design and validation of targeted single-species real time PCR assays (Thalinger et al., 2021), these documents represent a valuable resource for the development of standardized eDNA protocols.

Despite the fact this study was designed and carried out before most of these guideline documents were available, the majority choices and decisions made here comply with current recommendations. The qPCR assay used here (Matejusova et al., 2021) could be classified as a “substantially” validated assay (level 4 out of possible 5) according to the proposed validation scale (Thalinger et al., 2021). This means that if this assay produces a positive amplification of the target species in water samples, the target is very likely to be present at the given location or its immediate vicinity. However, if the assay does not generate an amplification, accurate interpretation of the target absence is more complicated and sufficient replication, both biological (number of samples collected) and technical (number of PCR reactions carried out for each sample), is necessary to account for imperfect detection.

It is essential to understand the false negative and false positive error rates and the level of biological and technical replication needed to mitigate these errors to confidently infer presence of organisms at studied sites (Ficetola et at., 2016). Having well-designed sampling strategies and laboratory processing pipelines, with reduced rates of imperfect detections, lead to greater confidence in eDNA-derived data and ultimately in increased uptake of this approach for statutory monitoring and decision making. A review published by Burian et al. (2021) summarizes the sources of both types of error and where in the process (field sampling or laboratory and data analysis) they occur. Sources of false positives are most likely linked to poorly designed assays amplifying non-species targets or as a result of in field or laboratory contamination with the target species DNA. The present study mitigated the rate of false positives by robust validation of assay specificity, collection and processing of negative controls at each step of the process, and by stringent laboratory cleaning and separation of pre- and post-PCR processes, as recommended in past and current guidelines (Goldberg et al., 2016; Thalinger et al., 2021; Bruce et al., 2021). It is worth mentioning that one blank field negative control, filtered alongside the samples collected in the yacht marina in Largs, showed inconsistent amplification of D. vexillum eDNA. However, as the other sampling sites visited on the same day were either consistently D. vexillum eDNA negative or positive, for all samples collected and technical replicates analysed, coupled with the fact that only disposable equipment was used to filter water samples, it was concluded that this level of contamination did not represent a significant risk for the outcome of this study. As the blank control was processed alongside water samples collected at Largs Yacht Haven Marine, it is most likely that the blank field control was contaminated at that site, either during filtration or sample transportation.

The occurrence of false negatives can be associated with the design of laboratory tests used and PCR inhibition, but more likely it is a consequence of the availability and quality of eDNA in the environment sampled (see review by Burian et al., 2021). Most commonly, the availability of eDNA is shown to be linked to shedding rates (for example, Allan et al., 2021), heterogeneity of eDNA sources in water, and eDNA degradation exacerbated by various environmental factors that reduce its detectability (for example, Stoeckle et al., 2017; Seymour et al., 2018; Harrison et al., 2019).

Statistical approaches, such as site occupancy models, designed to account for imperfect detection when inferring the presence of a species, already exist and have been demonstrated as directly applicable for use in eDNA-based surveys. Site occupancy modelling can be used effectively to estimate the number of samples and technical replicates necessary to detect a species of interest with given confidence. However, to accurately infer occupancy, data from repeated sampling during a known period of constant presence of a species is needed to estimate the detection probabilities of a target while accounting for imperfect detection (Schmidt et al., 2013). The uptake of statistical modelling during the sampling design stage has been historically low due to the highly specialised analytical skills needed but in recent years a number of studies clearly demonstrate the benefits of site occupancy models (Da Silva Neto et al., 2020; Buxton et al., 2021; 2022; Tingley et al., 2021). Improved occupancy models, now also taking into account false positives and the effect of environment covariants, have been developed (Griffin et al., 2019) and user-friendly software package has been published (Diana et al., 2021).

In the present study, a substantive sample replication, with 10 water samples collected at each site, split across two sampling points, was carried out to reduce the rate of potential false negative errors at the sample collection stage (Mauvisseau et al., 2019; Buxton et al., 2021). At the laboratory analytical stage, three technical qPCR replicates were carried out, which is comparable to choices made in a large number of the most recent eDNA related papers (for example, Roux et al., 2019; Wood et al., 2019). It should be noted that other papers suggest that increasing the number of technical replicates to four or five might be beneficial in reducing biases from imperfect detections (Buxton et al., 2021). However, Buxton et al. (2021) also concluded that greater biological replication at the water sampling stage has a greater impact on the overall prediction of species occupancy than increased technical replication at the qPCR level. As number of biological and technical replicates needed to account for imperfect detection is species-dependent, a detailed analysis of the generated presence and absence D. vexillum eDNA data using statistical occupancy models (Griffin et al., 2019; Diana et al., 2021) will be carried as future next steps towards designing a sensitive and cost-effective D. vexillum eDNA sampling strategy.

Potential introduction and spread of D. vexillum in Loch Creran and wider Lynn of Lorn systems

For the Loch Creran and wider Lynn of Lorn areas, the number of sites where D. vexillum eDNA was detected was low, suggesting the rate of introductions and subsequent spread might so far be limited. The present study shows the presence of D. vexillum eDNA directly at the Pacific oyster farm in Loch Creran where colonies of D. vexillum have been identified (Cottier-Cook et al., 2019) but also at the Rubha Garbh and potentially South Shian shorelines, located in a proximity to the farm. It is likely that D. vexillum eDNA could have been transported to these locations by the water currents, explaining the observed dis-concordance between the detection of eDNA and RAS assessments (Brown et al., 2018b; Begg et al., 2020). However, it would be advisable to carry out a subtidal rapid assessment at the pontoon in South Shian and its vicinity to survey for potential colonies of D. vexillum, especially as this is an operational pontoon servicing the loch and the ongoing finfish aquaculture activities.

Conversely, it is less likely that eDNA could have been transported by water current from the Pacific oyster farm to Loch Creran marina in Rubha Dearg and unlikely for Dunstaffnage Marina situated outside of Loch Creran. The subtidal and intertidal RASs carried out in Dunstaffnage Marine in 2010 (Beveridge et al., 2011) and in Loch Creran Marina in 2018 and 2020, respectively (Brown et al., 2018; Begg et al., 2020), and found no evidence of D. vexillum colonies. However, both marinas are expected to be well connected with other marinas in the Clyde area (data from www.rya.org.uk), for example Largs Yacht Haven, where D. vexillum is known to be present (Beveridge et al., 2011), so accidental, two-directional transfer of NIS between these sites, as boat’s biofouling, cannot be excluded. In September 2022, Marine Directorate carried out a brief rapid assessment of a pontoon in Dunstaffnage Marina, in a close proximity to where D. vexillum eDNA samples were collected, and found the D. vexillum colonies growing on a pontoon and pontoon-associated seaweed (Figure 2).