Morphological traits and community composition of bulk DNA and eDNA from the ethanol preservative
A detailed comparison between communities recovered in bulk DNA and eDNA from the ethanol preservative was conducted for primer set A. Both DNA datasets detected more species than the morphology based identification (69 species in bulk DNA, 64 in eDNA and 57 in morphology, ESM Fig 7). Although community composition significantly differed between bulk DNA and eDNA of the ethanol preservative (Table 2), 49 species were also shared between the two types of sample (ESM Fig 7). Nevertheless, the differences observed between the two sample types could be explained by morphological traits of the macrobenthos species. The two biological traits larval stage and life span did not result in significant terms and were not included in the final model. The final model included fixed effects of sample type, body size and body skeleton, as well as the two way interaction terms sample type * body size and sample type * body skeleton. The probability of detecting a species in bulk DNA or eDNA from the ethanol preservative was differently affected by body size and body skeleton as shown by the significant sample type * body size and sample type * body skeleton interaction effects (Table 3). The probability of detecting species in the bulk samples was overall high (Fig 5), with no significant differences between body sizes and body skeleton (all pairwise comparisons had p > 0.05). In contrast, the probability of detecting species in the eDNA from the ethanol preservative was significantly higher for soft bodied species compared to species with chitine (p = 0.0001) or CaCO3 in their skeleton (p = 0.0001). The probability of detecting species with small and medium body sizes was significantly higher in bulk DNA samples than in the eDNA samples from the ethanol preservative (body size 0-20: p < 0.0001; body size 21-100: p = 0.003; Fig 5), while the probability of detecting species with large body sizes was not significantly different between the two sample types (p = 0.18). The presence of body skeleton had also a clear impact on the probability of species detection in the two sample types: the probability of detecting species with chitine or CaCO3 in their skeleton was significantly lower in the eDNA of the ethanol preservative compared to the bulk DNA samples (p = 0.0002 and p < 0.0001, respectively, Fig 5), while no significant difference was observed between the two sample types for species with soft body skeleton (p = 0.21).