Re-analyses of relationships between size-fractioned eDNA and fish abundance
We reanalyzed the dataset from a previous study to further examine the effects of particle size and size fraction on eDNA-based abundance estimation (Jo et al., 2019). The aforementioned study measured the size distribution of T. japonicus eDNA derived from mitochondrial and nuclear DNA. Briefly, rearing water was sequentially filtered using different pore size filters; the particle size distributions of target eDNA were compared among rearing temperature, fish biomass in a tank, and target genes (mitochondrial or nuclear). Detailed information on the experimental design, water sampling, and molecular analyses can be accessed in Jo et al. (2019). Here, we performed linear regressions using the dataset and assessed the variation in the relationship between eDNA concentration and fish biomass according to eDNA size fraction. The eDNA samples that were collected a day before the fish removal and passed through sequential filters with 10, 3, 0.8, and 0.4 μm pore sizes (i.e., >10, 3–10, 0.8–3, and 0.4–0.8 μm size fractions, respectively) were compiled. The fish biomass in 200-L tanks ranged from 5.8 to 460.1 g. We then calculated two kinds of cumulative eDNA concentrations from each size-fractioned eDNA concentration as follows:
(i) Upside-cumulative eDNA (UC) meaning the cumulative eDNA concentration at all size fractions that can be collected by a given pore size filter:
・UC>10 µm = C>10 µm
・UC>3 µm = C>10 µm + C3–10 µm
・UC>0.8 µm = C>10 µm + C3–10 µm + C0.8–3 µm
・UC>0.4 µm = C>10 µm + C3–10 µm + C0.8–3 µm + C0.4–0.8 µm
(ii) Downside-cumulative eDNA (DC) meaning the cumulative eDNA concentration at all size fractions that can be collected by the smallest pore size filter (0.4 µm) following pre-filtration with a given pore size filter:
・DC>0.4 µm = C0.4–0.8 µm+ C0.8–3 µm + C3–10 µm + C>10 µm
・DC0.4–10 µm = C0.4–0.8 µm + C0.8–3 µm + C3–10 µm
・DC0.4–3 µm = C0.4–0.8 µm + C0.8–3 µm
・DC0.4–0.8 µm = C0.4–0.8 µm
where CX means the concentration of target eDNA at X µm size fraction (i.e., size-fractioned eDNA). As described above, we performed linear regressions of size-fractioned (CX) and both cumulative eDNA (UCX or DCX) concentrations (log10-transformed + 1) against fish biomass (log10-transformed) and estimated their R2values and regression slopes with 95% CIs. Additionally, the relativeR2 values were calculated to simplify the effect of size fraction on the relationships, where theR2 values at a given size fraction were divided by the R 2 value at >10 µm size fraction for the size-fractioned (CX) and upside-cumulative (UCX) eDNA and at >0.4 µm size fraction for the downside-cumulative eDNA (DCX) corresponding to a similar temperature level and marker type.
Moreover, we assessed the effect of eDNA size fraction on theR2 values and slopes of linear regressions, using the three types of eDNA concentrations and linear mixed models. In the former models, the R2 values (Fisher’s z-transformed) were included as the dependent variable, each size fraction (µm) was included as the fixed effect, and temperature levels (13, 18, 23, and 28°C) and marker types were included as the random effects. On the other hand, in the latter models, every kind of eDNA concentration (CX, UCX, or DCX; log10-transformed + 1) was included as the dependent variable, fish biomass (log10-transformed), each size fraction, and their interaction were included as the fixed effects, and tank replicates, temperature levels, and marker types as the random effects. In these analyses, water temperature levels and marker types were not included as the fixed effects because these factors were not of interest in this study.