Unveiling biogeographical patterns of the ichthyofauna in the Tuichi basin, a biodiversity hotspot in the Bolivian Amazon, using environmental DNA

recovered by pipetting and then 1.7 ml of Na acetate 3M pH 5.5 and 11 ml of cold isopropanol were added and mixed by inversion to precipitate DNA, then stored at -20˚C for 48h, centri- fuged 30 min at 4˚C and 6,200 g. After elimination of the supernatant, the DNA pellets are washed with 1.5 ml of 70% ethanol, centrifuged for 15 mins and allowed to dry after elimina- tion of the supernatant. The pellets are finally resuspended in 200 μl of TE. A 25 μl fraction is purified with XP ampure beads (1.8X) before amplification by PCR. Primers design, PCR amplification of barcodes and library preparation. We designed two COI metabarcoding primers targeting Amazonian fishes: one of 185 bp (MK1) and another of 285 bp (MK2) which share the same forward primer and are positioned 5’ from the standard COI-5P barcode. The R Primerminer package [36] was used to extract the COI sequences available of the 10 main freshwater fish Orders present in the Amazon basin. A manual design of the primers was carried out on the 10 consensus alignments carried out for each Order. These two barcodes were amplified separately with the following primers: F1-TCHACHAAYC AYAAAGAYATYGGYACYCT with R1-ACYATRAARAARATYATYACRAADGC and F1 with R2-CARAARCTYATRTTRTTYATTCGNGG. Given the fact that eDNA samples are known to be generally quite fragmented (Harrison et al., 2019 [37] for a review and Deiner et al., 2017b [38] for contrasting results), we chose to use these two barcodes and compare their perfor- mance. The second barcode which completely overlaps the first is supposed to be more resolu-

tive because it is longer but less sensitive given its length. The taxonomic resolution, at the species level, of MK1 and MK2 barcodes for Amazonian fish species was 85 and 89%, respec- tively (see details in S2 Text). The libraries were prepared according to a 2-step PCR [39]. Each library was indexed using unique combinations of double indexing to more effectively exclude chimeric amplicons and index hopping during downstream analysis (unexpected combina- tions were assigned as indeterminate). For each filter cartridge three libraries from three inde- pendent DNA extractions were amplified for both barcodes. Two mock samples made up of 40 taxa (DNA extracted from tissues) for which we know the exact taxonomic composition were included to assess the sensitivity and specificity of the barcodes used in this study. As controls, they also make it possible to validate the bioinformatic treatments implemented, because we expected that the estimation of their specific composition would be close to the known composition. An additional control called "alien" was included to verify cross-contamination during the preparation and sequencing of libraries. In addition to the 22 negative field controls (3 DNA per negative cartridge +1), 14 extraction negative controls and 6 PCR negative controls (ultra- pure water instead of DNA) were also amplified and sequenced in the same run with the sam- ples. Overall including negatives controls, positives controls and samples, 200 libraries for each barcode were pooled, then purified and sequenced (paired end, 150 bp) using Novaseq 6000

Illumina platform at the Novogene Co., Ltd. facilities. Data analysis. Data processing and taxonomic assignment. Amplicon sequences were ana- lysed using the FROGS pipeline [40] with standard operating procedures. Amplicons (MK1 and MK2) were split according to their primer sequences and size (120–280 nucleotides and 200–300) and clustered into operational taxonomic units (OTUs) using Swarm (aggregation distance: d = 1). After chimera removal, OTUs were kept if present in at least one library and with a minimum abundance of three sequences. OTUs alignment against a custom COI fish database was performed by blastn (version 2.8.1). This local database was built with fish COI sequences mined from GenBank and Bold and also implemented with published sequences [41], but not yet available under GenBank (S2 Table). This final database has 160,944 entries representing 3,480 genera and 15,429 fish spe- cies distributed worldwide. For Peru and Bolivia among the 1,605 species currently known [42, 43] only 684 species are represented in the COI database. Although this demonstrates the PLOS ONE Fish diversity and distribution in the Tuichi basin PLOS ONE | https://doi.org/10.1371/journal.pone.0262357 January 4, 2022 6 / 25