Abundance estimation depending on eDNA particle size and size fraction
The aquarium experiment using zebrafish revealed that regardless of DNA marker length, eDNA particles collected by a 10 µm pore size filter could not explain the fish abundance more accurately than those managed by other smaller pore size filters. Notably, we observed higherR2 values between the shortest eDNA fragments (132 bp) and fish abundance by increasing the filter pore sizes. However, the value radically decreased when using the 10 µm pore size filter, suggesting that the relationship between eDNA concentration and species abundance can be worsened when such “huge” eDNA particles and extremely larger size fractions are targeted. As expected in the Introduction, massive eDNA particles, such as tissue clumps, reflect the species abundance less precisely because of their potential heterogeneous dispersion and distribution in water (Turner et al., 2014; Furlan et al., 2016; Jo et al., 2020b). Moreover, mainly in natural environments, eDNA particle sizes can be increased by adsorbing suspended organic matter, making DNA molecules physiochemically stable and able to survive longer in water (Levy-Booth et al., 2007; Torti et al., 2015). Therefore, detecting such eDNA is likely to reflect past biological information and obscures current species distribution and abundance (Barnes & Turner, 2016; Jo et al., 2021).
On the contrary, this study indicates that eDNA particles at the 3–10 µm size fractions could be important for better performance of abundance estimation. First, the zebrafish eDNA particles collected by a 3 µm pore size filter reflected the fish abundance significantly better than those managed by a 10 µm pore size filter, corresponding to the result from the experiment targeting the upside-cumulative eDNA from T. japonicus . Second, for the downside-cumulative eDNA from Japanese jack mackerel, the R2 values between the target eDNA concentration and fish biomass were significantly lower at the 0.4–3 µm size fraction (i.e., excluding the >3 µm size fraction) than at the >0.4 µm size fraction, whereas little difference in the R2 values between the >0.4 and 0.4–10 µm size fractions existed (i.e., even excluding the >10 µm size fraction). Third, in the zebrafish experiment, the slopes of the linear regressions (the sensitivity of the eDNA concentration in response to changes in species abundance) were significantly higher for the eDNA particles collected by a 3 µm pore size filter than those managed by a 10 µm pore size filter. Summarily, these results suggest that an “appropriately” larger eDNA particle is suitable for accurate and sensitive estimation of the species abundance via eDNA.
The 3–10 µm size fraction is the fraction in which macrobial eDNA in water is most abundant (Turner et al., 2014; Jo et al., 2019; Zhao et al., 2021). It is possible that the size fraction mainly includes a single cell and intact nuclei and mitochondria of known sizes. Therefore, eDNA sampling focused on the specific size fraction where target eDNA is abundant may lessen false-negative eDNA detection and variance of eDNA quantification, consequently estimating species abundance more accurately and sensitively. Alternatively, different physiological origins and physiochemical structures of eDNA might result in its different persistence and spatial dispersion, hence influencing the relationship between eDNA concentration and species abundance. Additionally, the >3 µm pore size filter generally reduces filter clogging, increasing the filtration volume of the water samples considerably compared with smaller pore size filters (e.g., Kumar et al., 2021). Moreover, Takasaki et al. (2021) recently compared the pre-filtration performance on eDNA detection using various pore size filters (10–840 µm pore sizes). The result showed that 10 µm pre-filtration effectively reduced PCR inhibitors (e.g., humic substances) in river water samples without decreasing the detectability of target eDNA. From the above, it would be promising to collect and analyze target eDNA focusing on the 3–10 µm size fraction to better estimate species abundance.
Conversely, the performance of abundance estimation was poor for eDNA particles at the <3 µm size fraction. According to the experiment targeting size-fractionated eDNA from T. japonicus , the selective collection of eDNA particles at the 0.8–3 µm size fraction tremendously worsened its relationship with fish biomass and sensitivity in response to changes in fish biomass (i.e., theR2 value and slope). This trend could partly contribute to the result that the slopes of linear regression tended to be lower for the Japanese jack mackerel’s eDNA particles at the >0.8 µm size fraction than that at the >10 µm size fraction. Furthermore, variations of the abundance estimation accuracy via eDNA across filter pore sizes in some previous studies are likely attributed to eDNA particles at the 0.8–3 µm size fraction. Takahara et al. (2012) assessed the relationships between common carp (Cyprinus carpio ) eDNA concentration and fish biomass using different filtration methods, reporting a betterR 2 value for the 3 µm pore size filter (i.e., >3 µm size fraction; R2 = 0.93) than the 0.8 µm pore size filter following a 12-µm pre-filtering (i.e., 0.8–12 µm size fraction; R2 = 0.85). Eichmiller et al. (2016) also found that the R2values between common carp eDNA concentration and fish biomass were higher for 5 µm pore size filters (R2 = 0.86) than for 1 µm pore size filters (R2 = 0.71). Moreover, with the frequent non-detection (zero copies) of T. japonicus eDNA at the 0.4–0.8 µm size fraction, theR2 values and slopes in the linear regressions at the 0.4–0.8 µm size fraction were biased and overestimated by a zero-inflation of the target eDNA concentration. Sequel to this, it agrees to a lower proportion of fish eDNA at the <0.8 µm size fraction (20% at most; Turner et al., 2014).