Evaluation of fish biodiversity in estuaries using environmental DNA metabarcoding

sequence identity >98.5% to the reference sequences. Above procedures were described in S1 Text in detail. All negative controls in sampling stations and PCR blanks were also analyzed using this pipeline. The number of reads corresponding to every fish detected in the negative control were deleted (S1 Table) and flathead grey mullet Mugil cephalus is removed from Tama River L2 station after this process. Species verification The species obtained by pipeline still needed to be verified because sequencing results com- prised only a short region (170 bp) of 12S rRNA [24], and similar sequences might correspond to different species. Also, multiple species could be incorporated into a single species, and vice versa. We checked all species on the list with the original aligned sequences using the NCBI Basic Local Alignment Search Tool (http://blast.ncbi.nlm.nih.gov/Blast.cgi), and applied MEGA7 [28] to construct a neighbor-joining tree for all stations characterized by occurrence of the same species. When several species shared the same or similar (>99%) aligned sequence, we confirmed the species identity by referring to species distribution reported by the IUCN (https://www.iucnredlist.org), FishBase (http://www.fishbase.de), illustrated books of Japanese fishes [29–31], and personal communications. For example, the Japanese black porgy (Acanthopagrus schlegelii) and the Okinawa seabream (Acanthopagrus sivicolus) have the same aligned sequence, but the Okinawa seabream cannot exist in the waters of any station from the

present study. On the contrary, we combined two or more species that were considered to be local variations, even if their sequences differed substantially. Species whose reads number amounted to <0.05% of total reads of library were deleted because they were potentially caused by contamination, as indicated by Andruszkiewicz et al. [32] with some modifications. If species that were obviously not expected in this area were detected, but represented commonly consumed food items (e.g. Allaska pollock, Tuna), they were regarded as contamination and removed as well. Estimates of biodiversity Even if fish biomass could be reportedly determined by eDNA [33], eDNA has been limited to certain species. Moreover, it has not been applied to metabarcoding because of species-specific amplification rates [34], environment-dependent degradation rates [16, 35], and PCR inhibi- tion by environmental factors [10, 12]. Therefore, the estimate of biomass requires a complex model and the possible use of eDNA for this purpose needs to be verified. Biodiversity is some- times calculated by functions such as ‘number of species’ and ‘biomass;’ however, as biomass information was not available in the present study, we considered ‘species richness’ as a proxy for ‘biodiversity.’ Environmental data set Data regarding nutrients were obtained from the Ministry of the Environment of Japan (http://water-repo.env.go.jp/water-repo/). We used the annual mean value of nutrient concen- tration combining total nitrogen (TN) and total phosphorus (TP) published in the Measure-

ment Results of Water Quality in Public Waters in FY 2016 (Ministry of the Environment) as a water quality index of the river. The annual mean value is based on 6–12 measurements a year at each monitoring point. The monitoring points corresponding to the target watersheds (points using the TN and TP values) were the most downward points of each river. The revetment rate was calculated by measuring the distance of artificially protected areas, such as concrete-sealed piers or concrete tetrapods, within a distance of 3 km on both sides of PLOS ONE Biodiversity of five estuaries PLOS ONE | https://doi.org/10.1371/journal.pone.0231127 October 6, 2020 5 / 15