Results
An average of 2848 reads per sample were assigned to the reference ofC. nozawae (Table 1), all with sequence identity of >99.4%. A total of 66 haplotypes were detected from eDNA, whereas 58 haplotypes were detected from tissue samples. Among the haplotypes detected from eDNA, 35 haplotypes matched with 100% identity to haplotypes detected from tissue samples, and the other haplotypes matched with 99.4%–99.7% identity (representing differences of 1–2 bases). The number of haplotypes detected in each population was 2–17 (mean: 8.62) in eDNA and 1–11 (mean: 6.43) in tissue samples, and the haplotype diversity h S ranged from 0.01–0.83 (mean: 0.57) in eDNA and 0.00–0.89 (mean: 0.64) in tissue samples. Global G ST was 0.33 in eDNA and 0.29 in tissue samples, and the average D PS was 0.86 in both eDNA and tissue samples. Pairwise G ST ranged from 0.01–0.90 (eDNA) or 0.01–0.89 (tissue), and D PSranged from 0.17–1.00 (eDNA) or 0.25–1.00 (tissue) (Table S3).
The spatial distribution of haplotypes was generally consistent between eDNA and tissue samples (Figure 2). Some major patterns of the spatial structure could be commonly found, such as the dominance of a unique haplotype in Pop 3, the presence of haplotypes commonly found at many sites in Pop 1–11, and Pop 12 being composed almost entirely of one haplotype also found in the downstream sites (Pop 13–15). Compared with the results of the STRUCTURE analysis using SNP data, some patterns such as the presence of spatial structure or the uniqueness of Pop 3 were common, but there were some differences from D-loop region results, such as the absence of a clear differentiation between Pop 1–11 and Pop 12–13 in SNP data.
The statistics of genetic diversity and differentiation calculated from eDNA were positively correlated with those calculated from tissue samples (Figures 3 and S2). For genetic differentiation, the correlation coefficient was r = 0.73 (p < 0.001) forG ST and r = 0.83 (p < 0.001) forD PS. From the spatial autocorrelation analysis by the Mantel correlogram, r values were consistently positive up to approximately 30 km and significant up to at least 15 km in all genetic markers, indicating that gene flow can be considered particularly strong in this range. Correlations of genetic differentiation calculated between eDNA and SNP data were low when calculated for all 21 populations, but there was some correlation when calculated only for population pairs less than 15 km apart (Figure 3, Table 2).
Additional data treatment of the eDNA dataset did not increase the correlation of its statistics with tissue-based statistics (Tables 2 and S4). Treatment 2, which involved bold data removal, significantly worsened the correlation in most datasets (95% CI not overlapping). In terms of the correlation coefficient alone, treatment 3 (semi-quantitative approach) sometimes outperformed the no treatment dataset, although the differences were not significant.
From the basic analysis including spatial and water temperature data, the detection patterns from our eDNA data were congruent with those of previous studies, with a significant correlation between genetic differentiation and waterway distance (r = 0.50, p < 0.001 for 19 populations; r = 0.52, p < 0.01 for upstream populations; Figures 4 and S3) and no significant correlation between genetic differentiation and water temperature differences (r = 0.20, p = 0.16 for 19 populations; r = 0.32, p = 0.06 for upstream populations).