Results
In this study, the abundance of eDNA (i.e. the fish mitochondrial 12S rRNA gene) in the extracted DNA from the aquarium experiments using two mock fish communities was quantified with dPCR and qSeq. The eDNA of only three fish species, H. neglectus , M. anguillicaudatus , and O. latipes , were quantified with dPCR, whereas qSeq quantified all five fish species used in this study. We also calculated the relative abundances of the sequences of each fish species in the sequence library obtained by iTag-sequencing using Mifish primers for comparison with dPCR and qSeq.
We found that, for MC1 the abundance of eDNA generally decreased with time, although the variation in quantified abundance was large (Figure 2, left). Specifically, dPCR showed that the abundance of O. latipe s eDNA was 2.1-7.1 × 105 (mean: 4.3 × 105) copies/L on day 0 and dropped to 5.0 × 103–3.1 × 105 (mean: 1.2 × 105) copies/L on day 1. The abundances of H. neglectus and M. anguillicaudatus on day 0 were lower thanO. latipes , with 3.7 × 104–1.6 × 105 (mean: 7.9 × 104) copies/L and 1.1 × 104–1.2 × 105 (mean: 6.3 × 104) copies/L, respectively. On the last experimental day (day 4), eDNA abundances were lowest, with mean abundances of 4.3 × 105, 8.8 × 104, and 6.3 × 104 copies/L for O. latipes , H. neglectus , and M. anguillicaudatus , respectively.
Abundances obtained from qSeq were generally consistent with the results from dPCR, except for in MC1AR2. More specifically, qSeq quantified the abundance of eDNA of MC1 (Figure 2, left, middle row) on day 0 to be 1.1-5.4 × 105 (mean: 2.6 × 105) copies/L, 3.8 × 104–5.8 × 105(mean: 2.1 × 105) copies/L, and 1.7 × 104–1.4 × 105 (mean: 6.7 × 104) copies/L for O. latipes , H. neglectus , and M. anguillicaudatus , respectively. On the last experimental day (day 4), the eDNA had decreased to 9.8 × 102–2.9 × 104 (mean: 8.8 × 103) copies/L, 8.2 × 102–6.6 × 104 (mean: 1.2 × 104) copies/L, and 2.2 × 103–2.5 × 104 (mean: 9.5 × 103) copies/L. In addition to the three fish species, which could be quantified using previously established assays (i.e., designing specific primers and probes), two fish species, C. temminckii  and R. flumineus,  were quantified by qSeq without establishing an assay for each species. The decreasing trend in eDNA with time was similar for these two species compared with the other three species, and the abundances were within the range of the other three species.
Mock fish community 2 had three individuals of H. neglectuscompared with one in MC1. Compared with MC1, a different trend was observed in MC2, where the highest abundance of eDNA was consistently observed on day 1. After day 1, as in MC1, the abundance tended to decrease over time. The abundances of eDNA from individual species in MC2 was lower than MC1. For example, according to dPCR quantification of MC2 on day 0, there were 5.0 × 103–8.6 × 104 (mean: 5.2 × 104) copies/L, 3.0 × 103–9.7 × 104 (mean: 5.0 × 104) copies/L, and 1.0 × 103–1.7 × 104 (mean: 8.3 × 103) copies/L forO. latipes , H. neglectus , and M. anguillicaudatus , respectively. Similarly, qSeq resulted in fewer copy numbers of eDNA from MC1 compared with MC2, with 3.6 × 103–5.6 × 104 (mean: 2.0 × 104) copies/L, 1.6 × 103–4.9 × 104 (mean: 2.6 × 104) copies/L, and 1.3 × 103–1.7 × 104 (mean: 8.2 × 103) copies/L forO. latipes , H. neglectus , and M. anguillicaudatus , respectively on day 0 in MC2. The other two species (C. temminckii  and R. flumineus ), which were only quantified using qSeq, showed a similar trend, with eDNA abundances peaking on day 2 and decreasing with time. Across all observations, eDNA abundances obtained from dPCR and qSeq were generally in agreement.
Although the relative proportions obtained in MiFish are not comparable with the absolute quantitative values obtained by dPCR or qSeq, the relative proportions in the MiFish sequencing library (Figure 2, bottom row, presented in log-scale for the comparison) followed different trends to the results from the two quantitative methods. In MC1, the MiFish results of from AR1 and AR2 were similar to the two quantitative methods (i.e., dPCR and qSeq). However, BR1 and BR2 generated completely different results; the relative abundance of eDNA of only three fish increased with time. More specifically, in MC1BR2, the relative abundance of H. neglectus eDNA on day 0 was 1%, and it increased to 4-11% by day 4. The relative abundances of O. latipes  and M. anguillicaudatus  in the sequence library also increased over time. However, these results were concurrent with a decrease in the relative abundance of eDNA from C. temminckii , and therefore might give a false impression that those fish species increased over time. Similarly, relative abundances in the sequence library were not consistent with the quantification results of the other two methods. For instance, in MC2AR2, the trend in relative abundance ofO. latipes was opposite to the other two methods, with the lowest relative abundance observed in day 2. These results confirm that relative abundances alone cannot be used to quantitatively discuss the behaviour of eDNA.
Previous studies have demonstrated a correlation between biomass and the density of eDNA in natural or laboratory environments (Yates, Fraser, and Derry, 2019; Doi et al., 2015). In this study, however, no significant correlation between biomass and abundance of eDNA was observed (Figure 3). The mean weight of M. anguillicaudatus  was 4.2 g, which was approximately 10-fold higher than that of O. latipes  at 0.47 g (Table 2). On the other hand, the abundance of eDNA from M. anguillicaudatus was generally lower than that fromO. latipes , regardless of the quantification method (Figure 2). This discrepancy might be attributed to the difference in the discharge rate of eDNA between fish species. MC2 contained three individuals ofH. neglectus compared with one in MC1; however, the abundance ofH. neglectus eDNA was lower in MC2 than MC1 (Figure 3).
The abundance of eDNA for each fish species quantified with dPCR was strongly correlated with that quantified with qSeq (Figure 4). The correlations were significant (p < 0001) and had R2 values of 0.643, 0.859, and 0.786 for H. neglactus , M. anguillicaudatus , and O. latipes , respectively. Relationships between qSeq and dPCR had slopes of ~1 and were not significantly different to 95% CIs. However, dPCR resulted in higher values than qSeq for most of the eDNA samples of O. latipes . The clear linear significant correlation between quantified abundances obtained by dPCR and qSeq indicates that using qSeq instead of the normal HTS can add quantitative information to species composition data based on obtained sequences without establishing a specific assay for each fish species of interest.