1. Meta-analyses of eDNA literature
Our meta-analyses showed that filter pore size, water temperature, target gene, and water source could influence eDNA degradation, not as individual parameters but in conjunction. We focused on three substantial interactions that were included in almost all supported models. Firstly, the interaction between filter pore size and water temperature influenced eDNA decay rates (Figure 1). Considering that a larger pore size filter can selectively collect eDNA particles in larger size fractions, our result implied that higher water temperature could accelerate the degradation of eDNA in larger size fractions by a greater degree than that in smaller size fractions. However, it is unlikely that smaller-sized eDNA itself is less affected by higher temperature-mediated degradation, and its apparent persistence can be increased by the inflow of eDNA from larger to smaller size fractions, as described in Jo et al. (2019b). Organic matter in water, including eDNA, is degraded by microbes and extra-cellular enzymes in the environment for uptake, and their activities are promoted by moderately high temperatures (less than 50°C) (Price & Sowers, 2004; Nielsen et al., 2007; Arnosti, 2014; Strickler et al., 2015). During the degradation processes, aqueous eDNA in larger size fractions, such as intra-cellular DNA, is believed to flow into smaller size fractions, such as extra-cellular DNA. This suggests that water temperature does not uniformly influence the apparent degradation of eDNA among the different size fractions, and the effect of temperature on eDNA degradation might be buffered in smaller-sized eDNA particles. Thus, the effect of temperature on eDNA degradation would be smaller when using a smaller pore size filter and collecting eDNA particles at various size fractions.
Secondly, the interaction between the target gene (nuclear or mitochondrial) and water temperature influenced the eDNA decay rates; higher water temperature could accelerate the degradation of nuclear eDNA by a greater extent when compared with mitochondrial DNA (Figure 2). This may be attributed to the difference in the protection conferred to the DNA molecules against the attack of extra-cellular enzymes in the environment by the outer nuclear and mitochondrial membranes. In contrast to mitochondrial DNA, which is surrounded by a non-porous outer membrane (Ernster & Schatz, 1981), nuclear DNA is enclosed in a porous membrane (45-50 nm in diameter; Fahrenkrog & Aebi, 2003), rendering it more susceptible to environmental extra-cellular enzymes, and thus, more likely be degraded by a greater degree at higher temperatures (Price & Sowers, 2004; Strickler et al., 2015). However, these results should be interpreted with caution, because the number of nuclear eDNA decay rate constants (n = 17) included was considerably lower than that of mitochondrial eDNA decay rate constants (n = 89). It is necessary to estimate nuclear eDNA decay rates in various environmental and experimental conditions in the future, which would enable a more robust comparison of eDNA degradation between nuclear and mitochondria DNA.
Thirdly, the interaction between the target gene and water source influenced the eDNA decay rates (Figure 3). Although the effects of water source on eDNA degradation differed between nuclear and mitochondrial DNA, it was evident that eDNA degradation was suppressed in artificial waters, such as tap water and DW, when compared to that in natural waters. Eichmiller et al. (2016) compared the degradation of common carp (Cyprinus carpio ) eDNA in natural waters with different trophic states, and found that eDNA decay rates in well water were lower than those in eutrophic and oligotrophic waters, which could be attributed to the lower microbial activity in the former. Our results were generally consistent with those of Eichmiller et al. (2016). Using tap water and DW as water sources can lead to underestimation of eDNA persistence in the natural environment. Moreover, no significant difference could be observed in the eDNA decay rates between freshwater and seawater. The difference in eDNA persistence between freshwater and seawater has previously been reported; some studies indicated faster eDNA degradation in seawater than in freshwater (Thomsen et al., 2012; Sassoubre et al., 2016), whereas Collins et al. (2018) showed that eDNA degradation was higher in terrestrially-influenced inshore waters than in ocean-influenced offshore environments. Marine systems are generally characterized by higher salinity and ionic content, higher pH, and more stable temperatures when compared with freshwater systems, which can promote DNA preservation in water (Okabe & Shimazu, 2007; Schulz & Childers, 2011; Collins et al., 2018). However, the direct effects of microbial abundance and composition and other physicochemical parameters of water were not included in our meta-analyses. Thus, greater variations in eDNA decay rates in seawater when compared with artificial water and freshwater observed in our meta-analyses might partly be explained by such microbial and physicochemical conditions. The effects of various nutrient salts and microbial activities on eDNA persistence and differences in the eDNA degradation processes between freshwater and seawater systems require further investigation.
The interaction between filter pore size and water source influenced the eDNA decay rates in some supported models; however, its effect was relatively smaller when compared with those of the interactions discussed above (Table 2). The water source might affect the apparently longer persistence of smaller-sized eDNA described previously. Although no linear regressions were statistically significant (Figure S1;P > 0.05), the increase in eDNA decay rates with larger filter pore sizes appeared to be greater in seawater than in artificial water, which might also be attributed to the differences in microbial activities among the different water sources.
Contrary to these four factors, model selection in the present study did not strongly support the effects of DNA fragment size and its interactions with other variables on the eDNA decay rate (Table 2), which may be due to the potential bias of DNA fragment sizes in the eDNA studies included in the meta-analysis. Only three studies have previously estimated eDNA decay rates in water targeting longer DNA fragments (>200 bp) (aqueous eDNA; Jo et al. 2017; Weltz et al., 2017; Bylemans et al., 2018), and there was no consensus on the relationship between eDNA degradation and DNA fragment size among these studies. Although our additional meta-analysis, which targeted only shorter DNA fragments (70 to 190 bp), supported rapid eDNA degradation in longer DNA fragments, as suggested by Jo et al. (2017) and Wei et al. (2018), the analysis might be considered slightly arbitrary, and thus, the validity of the result would need to be tested in the future. Interactions between DNA fragment size and other factors may become evident when more information is available on eDNA persistence and degradation at different fragment sizes.