Conclusions, limitations, and perspectives
To our knowledge, the present meta-analysis is the first to report variations in eDNA-based estimation accuracy of species abundance among different target taxa and filter pore sizes (reflecting eDNA particle size distribution). The significance is that the relationship between eDNA quantity and species abundance can be driven not only by eDNA transport and degradation processes but also by the characteristics of eDNA (its production source and state). Some recent studies have suggested the possibility of improving the accuracy of eDNA-based abundance estimation by statistically accounting for the processes of eDNA production, transport, and degradation (Carraro et al., 2018; Cerco et al., 2018; Fukaya et al., 2021). In contrast, our meta-analyses shed a new light on the importance of what characteristics of eDNA should be targeted for more accurate estimation of species abundance. In particular, our findings on the effects of eDNA state imply that ‘more recently released’ eDNA, existing as larger eDNA particles and potentially longer eDNA fragments, more precisely reflect species abundance in the field. This knowledge will complement abundance estimation approaches that consider eDNA spatiotemporal dynamics; that is, understanding eDNA characteristics, including production source, particle size, and fragment length, as well as eDNA production, transport, and degradation processes, will enable us to further enhance the potential of eDNA analysis as a non-disruptive and cost-efficient tool for species abundance estimation. Therefore, accumulating knowledge of eDNA states and their interactions with the spatiotemporal dynamics (e.g., the processes of production, transport, and degradation) is crucial (Jo & Minamoto, 2021).
It should be noted that there are some potential biases and limitations in our meta-analyses. First, regardless of accounting for random effects and multiple factors, we observed the high degree of heterogeneity across studies and datasets in our meta-analysis. This infers that there remain a number of variables that our study could not consider. For example, environmental parameter such as water qualities and temperature has not considered here, although this point was barely considered as the categorical factor (laboratory/natural, lentic/lotic, freshwater/marine). Degradation of eDNA, which is accelerated by higher temperature (Strickler et al., 2015; Jo et al., 2019a), could affect its persistence time in water and possibly the goodness of relationships between eDNA quantity and species abundance. In addition, the difference in other technical steps, including eDNA storage and extraction, in the analysis should be considered. Some studies filtered water samples on site and transferred the filter samples to the laboratory, while others transported water samples and filtered them in the laboratory (Kumar et al., 2020). There are also substantial variations of DNA extraction protocols across studies (commercial kits or in-house formulations; Kumar et al., 2020). These methodologies might accordingly influence the estimation accuracy of species abundance via eDNA, as well as eDNA detectability and quantification.
Second, we also acknowledge the publication bias in this study. The asymmetry funnel plot infers that there may be some ‘hidden’ studies that both the estimation accuracy and its variance are small. This may partly be attributed to the studies that were not included in this meta-analysis because the indices of abundance estimation accuracy (Pearson’s correlation coefficients or R2 values) were not estimated or presented in the manuscript. Besides, our collected dataset was concentrated toward studies targeting fish species, which might cause biased and over-dispersed estimation for other taxa. Accumulating additional empirical studies for various taxa, assay strategy, and environmental conditions is necessary to validate the findings of our meta-analyses and further elucidate the influence of eDNA characteristics on eDNA-based estimation of species abundances.
Furthermore, although not considered in the present study, nuclear eDNA, particularly targeting multiple copies of ribosomal RNA genes, may be applicable for more accurate eDNA-based species abundance estimations. Relative to mitochondrial eDNA, targeting multi-copy nuclear eDNA can improve detectability and yield (Minamoto et al., 2017; Jo et al., 2020b) and nuclear eDNA may degrade more rapidly due to potential differences in membrane and DNA structures (Bylemans et al., 2018; Jo et al., 2020b). In addition, nuclear eDNA production may also be less biased by individual growth and developmental stages, whereas mitochondrial eDNA production is expected to be suppressed with maturity and aging (Jo et al., 2020b). Understanding both the characteristics and dynamics of eDNA will fill a gap between eDNA concentration and species abundance in the field, and update current eDNA analysis as a more refined tool for biodiversity and ecosystem monitoring and stock assessment.