Figure
1: Biological signatures in the form of eDNA or eRNA can be detected
from plants non-invasively to trace out complex interactions.
Illustration presents hypothetical examples of PAI (e.g., pollination,
herbivory, frugivory, and mutualism) including representative examples
in the literature.
4. Current advancements in eDNA for the study of PAI
Pollination is one of the most well studied PAI since it brings about gene recombination (Faegri & Pijl, 1979), and exemplifies myriad central ecological and evolutionary principles and theories. In pollinator PAI, the loss of even a singular plant species can trigger rapid extinction of specialist pollinators, which is also of serious ecological and economical concern (Klein et al., 2007; NEA, 2011). To date, researchers have taken advantage of eDNA-based analysis to detect and monitor pollinators, their feeding preferences, species-specificity, niche separation, and coevolution
(Table 1). In particular, eDNA metabarcoding of honey samples has been demonstrated to detect more taxa than conventional methods, where species-specificity (i.e., identification of generalists and specialists), foraging activity, and complex interactions have been analysed rapidly and cost-effectively (Hawkins et al., 2015; De Vere et al., 2017). Interestingly, eDNA from honey samples can also help to identify other entomological signatures within forests or agricultural fields, such as those from plant-sucking insects whose “honeydew” droplets are incorporated in honey reserves (Utzeri et al., 2018). Bovo et al. (2018) further demonstrated the utility of eDNA tools to understand the micro-ecosystem within honey bee colonies by detecting the eDNA signals from five distinct groups (i.e., arthropods, plants, fungi, bacteria, viruses). Although not strictly PAI, this study further exemplifies eDNA-based methods as a potential avenue for information regarding wildlife diseases and epidemics.
While complex pollinator networks are typically difficult to identify and discriminate using conventional sampling, eDNA collections taken directly from flowers or leaves have further shown promise to gain an in-depth understanding of dynamic pollinator and herbivore interactions (Thomsen & Sigsgaard., 2019; Kudoh et al., 2020). For example, Thomsen & Sigsgaard (2019) detected eDNA signatures from 135 arthropod species originating from diverse ecological groups deposited on wildflowers (e.g., pollinators, parasitoids, gall inducers, predators and phytophagous species), and suggested potential use of eDNA approaches for estimating interactive species compositions, deducing the effects of environmental change, and monitoring endangered, cryptic and invasive species (Thomsen & Sigsgaard, 2019).
Understanding the complex interactions between frugivores and plants also remains a challenge, but recent strides using eDNA traces to detect specific interactions of fruit-eaters have now made this prospect more convenient. For example, Monge et al. (2020) successfully amplified salivary eDNA of frugivorous birds (Ara macao) from tropical almond (Terminalia catappa) fruit remains. Albeit with limited success, this study further provided proof-of-concept for the use of eDNA in non-destructive sex identification, potentially ushering in a new frontier for studying sex-specific differences in PAI.
Herbivores often prefer a certain plant or group of species, which may cause shifts in plant composition. Thus, it would be beneficial to identify the diversity of plant taxa eaten by particular herbivores and the number of herbivores visiting focal plants. For herbivory, eDNA-based methods have been shown to detect large numbers of taxa more efficiently than other sampling methodologies (e.g., microscopic analysis of fecal sample, bulk DNA metabarcoding; Tournayre et al., 2021). In fact, eDNA metabarcoding has also been applied to understand the dietary overlap and competition among domestic and wild herbivores (ter Schure et al., 2021). Notably, sampling matter may be a restricted application to large organisms with detectable faecal deposits. To overcome this limitation, salivary samples can be collected to identify herbivores that have fed upon specific plants, even from small taxa (e.g., from browsed twigs, Valentin et al., 2020; or leaves, Nichols et al., 2012). For example, Nichols et al. (2015) applied eDNA analysis across a large forest landscape, proving the utility of this method for studying cryptic browsing behaviour. Salivary eDNA signatures can also be used to assess foraging preferences and niche separation among species (e.g., Van Beeck Calkoen et al. 2019). Impressively, salivary eDNA signals from insect herbivores within mesocosms have also shown a positive correlation between rim length (i.e., total outer edge) of feeding marks and eDNA concentration, implying eDNA signatures may be able to quantitatively delineate the amount of herbivory (Kudoh et al., 2020).
Detecting plant-pathogen/parasite interactions through eDNA has also recently become possible. Derocles et al. (2015) for example, successfully amplified trace DNA from plants-leaf miners-parasitoid interactions and Thomsen & Sigsgaard (2019) detected numerous phytophagous species, parasitoids, gall inducers, and predator insects through the metabarcoding of flowers. Although in these studies non-target taxa were used to isolate eDNA, this also can be done in a complete non-destructive manner using newly developed surface-sampling methods (see Valentin et al. 2020). Cumulatively, these studies provide a foundation for detecting antagonistic and cryptic plant-arthropod interactions with applications for disease monitoring and pest management.
Mutualistic relationships between plants and animals (e.g., insects and nematodes) assist plant growth and development, and these relationships can also be studied effectively through eDNA analysis (Ladin et al., 2021). For example, Rasmussen et al. (2021) used eDNA metabarcoding to explore how the diversity of fungi and arthropods were affected by different agricultural management practices. For a more historical perspective of mutualistic relationships, Gous et al. (2019) applied eDNA methods to investigate pollinator interactions that had occurred over a century ago via ancient honey samples, highlighting eDNA’s potential to reveal a time series of species interactions.
Certainly, eDNA methods have advanced our ability to accurately detect the occupancy of species (Deiner et al., 2021), and are highly cost and time-efficient (Qu & Stewart 2019). Indeed, they have even outperformed conventional methods of biodiversity sampling in several comparisons (McElroy et al. 2020; Fediajevaite et al. 2021), including their ability to capture increased taxonomic diversity compared to conventional methods which can be applicable for large-scale monitoring (Macher et al., 2018). Thus, eDNA-based methods have gradually overcome some limitations associated with conventional monitoring techniques (e.g., field identifications). Perhaps most importantly for the assessment of ecological integrity and functionality, eDNA has the ability to detect entire communities rapidly.
5. Current limitations
There remains a need to understand current limitations of eDNA analysis, especially when it pertains to PAI detection and interpretation. Limitations are spread out among each step of the collection-analysis-interpretation process (Figure 2). The existing limitations of this method are:
(I) The complex, and often idiosyncratic, ecology of eDNA. In effect, practitioners may sample different sources of eDNA (cellular, extracellular, extra-organismal, etc.) (Stewart, 2019; Rodriguez‐Ezpeleta et al., 2021), which may lead to different PAI interpretations. For example, pollen and spores (extra organismal DNA) are ubiquitous in the atmosphere, travel long distances (through wind or water), and contain adaptations to persist dormant stages for long periods of time. These, when settled and collected on non-targeted and non-interacting organisms, can lead to misinterpretation. Alternatively, extracellular DNA and cellular DNA are generally specific to places where
organisms recently moved and are subject to easy degradation. Thus, clear differentiation of their behavior may help to draw more precise conclusions.
(II) The production and release of eDNA into the environment can also occur at different rates, where eDNA concentration can depend on many variables such as life stage, metabolic activity, or breeding season (Stewart, 2019). What’s more, production rate of eDNA is most likely influenced by species interactions themselves (e.g., competition between/among species) (Stewart, 2019). In fact, mixed-species populations have been shown to increase eDNA production rates when housed together compared to single species populations (Sassoubre et al., 2016). Beside the aforementioned characteristics, the persistence of eDNA (Barnes & Turner, 2016; Deiner et al., 2017; Kudoh et al., 2020), and its transport in and between environmental media (air, water, soil) should also be considered (Barnes & Turner, 2016; Lacoursière‐Roussel & Deiner, 2021), especially given that these parameters have yet to be standardized for many taxa.
(III) Translating eDNA quantification metrics to organismal abundance has been controversial (Marshall et al., 2021), although recent research has advanced the possibility of absolute quantification (Tillotson et al. 2018; Hoshino et al., 2021) and even predicting dispersion time of eDNA within the environment (Marshall et al., 2021).
(IV) A universal limitation to any genetic-based species identification reliant on databases, is certainly missing species sequences, sequencing error, cloning vector contamination, and the redundancy of data (Singh, 2015). These issues may cause species misidentification which may also lead to the failure in decrypting accurate PAI (Sheppard et al., 2005; Roslin & Majaneva, 2016).
(V) As eDNA methods sometimes struggle with low detection rate, more comparisons are needed between eDNA and conventional surveys (e.g., camera, malaise traps) for translation of results and inferences between these different methods.
(VI) The detection of niche partitioning using eDNA-based methods is only just beginning (ter Schure et al., 2021) and fine-scale partitioning (e.g, different herbivory behaviour on the same plant) may be difficult with current eDNA analysis techniques.
(VII) Unsurprisingly, and similar to conventional approaches, eDNA methods also encounter some technical field and laboratory challenges. This is often because eDNA samples frequently contain PCR inhibitors thereby further reducing already low DNA concentrations (McKee et al., 2015). Besides this, false-positive and false-negative detections are also a matter of concern.
(VIII) Laboratory protocols, including the method of standardization, is directly dependent on sampling procedures, sample quality, environmental factors, and molecular markers design. Although recent studies show evidence of overcoming some technical limitations, such as primer development; chloroplast and nuclear primer for plants (rbcL, matK, trnH-psbA, ITS2, etc.), group specific primers for animals (MiFish, MiBird, etc), protocol standardization, and removing the barrier of inhibitors (Burian et al., 2021), collection and analysis optimization may still be required. Mitochondrial COI is the most common universal barcode for animals demonstrating good species discrimination (Che et al., 2012), but in plants, no single universal barcode provides suitable taxonomic resolution (Jones et al., 2021). For plants, multiple primers from two primary plastid regions in the chloroplast (e.g., rbcl and matK), frequently combined with nuclear regions (e.g., ITS2), have been used for barcoding but none of these have been found to be suitable across all species due to rampant introgressive hybridization and polyploidy (Jones et al., 2021).