Estimation accuracy and eDNA state
Our statistical modeling showed that the use of larger pore size filters could improve the accuracy of eDNA-based abundance estimation. According to previous studies, the cellular/molecular structure of larger eDNA particles derived from intra-cellular DNA, such as cells and tissues, is degraded into smaller eDNA particles over time (Jo et al., 2019b). Apparent persistence of such smaller-sized eDNA could thus be longer than that of larger-sized eDNA, and larger eDNA particles collected using larger pore size filters are more likely to be recently released and less degraded than smaller eDNA particles. Such larger-sized ‘fresher’ eDNA is expected to reflect species presence and abundance at a spatiotemporally finer scale, consequently improving the accuracy of eDNA-based abundance estimation. Although individual eDNA studies could not clearly infer the relationship between accuracy and filter pore size (Takahara et al., 2012; Eichmiller et al., 2016), our meta-analysis supports the applicability of larger pore size filters for improved abundance estimation via eDNA analysis for the first time.
In contrast, some datasets reported high R2 values using smaller pore size filters (Figure 3), which can collect both larger-sized fresher eDNA and smaller-sized older eDNA. Thus, these studies may have collected a higher proportion of larger-sized eDNA while using smaller pore size filters; in particular, laboratory experiments with excessively high abundances (e.g., Takahara et al., 2012; Doi et al., 2015) may have collected large quantities of large-sized fresh eDNA. Although studies using larger pore size filters (especially >3 µm) were limited in our dataset, further empirical studies targeting larger-sized eDNA particles would conceivably contribute to the robustness of our results and potentially provide a new approach to improve the accuracy of eDNA-based abundance estimation in the field.
Moreover, the particle size distribution of target eDNA will dictate whether the selective collection of larger-sized eDNA using a larger pore size filter is effective. eDNA particles from fish are generally concentrated at 1 to 10 µm size fractions (Turner et al., 2014; Jo et al., 2019b), whereas eDNA particle size distributions for other taxa remain largely unknown except for Moushomi et al. (2019) targetingDaphnia magna . Estimating eDNA particle size distribution provides information on its cellular/molecular state in water (Jo et al., 2019b) and helps determine a suitable filter pore size for its efficient collection (Turner et al., 2014). In particular, research on the particle size distributions of eDNA from crustaceans and mussels, which show weaker correlations between eDNA concentration and abundance, would support the applicability of larger pore size filters for improving the accuracy of their eDNA-based abundance estimation.
In contrast, PCR amplicon size (i.e., DNA fragment length of target eDNA) was not significantly correlated with R2 values. However, almost all the studies in our meta-analysis targeted shorter eDNA fragments (<200 bp); thus, the effect of PCR amplicon size on the accuracy of eDNA-based abundance estimation may be underestimated. Owing to higher decay rates, detection of longer eDNA fragments may mitigate the effect of degraded eDNA and improve species abundance estimation accuracy (Jo et al., 2017). Nonetheless, this study was conducted in a situation where false-positive inferences of target individuals could be obvious (i.e., the effect of fish markets and dead individuals); thus, the performance of longer eDNA fragments for species abundance estimation in more ‘ordinal’ situations is unknown. Future empirical studies targeting longer eDNA fragments (>300 bp) are needed to elucidate the importance of PCR amplicon size on eDNA-based abundance estimation.