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

is an extremely low cost compared to that of sampling specimens over decades, and which has enabled us to increase the species inventory by 11.4% in a watershed that has been inventoried for over a century [1]. Some of these newly detected species could be cryptic or very scarce, or new species for science but not yet described because they are morphologically similar to already described species: for example, the species of the genus Prochilodus noted affinis in this paper, and probably already identified by metabarcoding of bulks of larvae further down- stream from our study area [11]. A limited detectability of eDNA in the Amazon region. Our study showed that some species that were detected at a sampling site were no longer detected a few kilometres down- stream (around 3 km) at the next sampling site. However, it should be noted that the sites where these species were identified with their DNA do not necessarily correspond to their locations, so this distance may be underestimated. Although it is difficult to compare diffusion distances between different studies, which depends on detectability and therefore on the exper- imental set-up (volume of water filtered, fraction of DNA analysed, etc.), Civade et al. (2016) [60] estimated diffusion distances in French Guiana at between 2 and 3 km, similar to those observed here. However, in the Rhoˆne in France, eDNA was found between a few km to 130 km from its emitting source, but with a detection rate below 0.4% for the latter [61]. A hypothetical explanation could be that the detectability of fish species in the Amazon

basin could be lower than in other regions. As the specific diversity in the Amazon is much greater than that observed at temperate latitudes, if the biomasses are similar, the eDNA con- centration for each Amazonian species would therefore be lower, reducing their detectability. In our case, Orestias ascotanensis, present at the extreme upstream Pelechuco site (a creek of less than 1 m width), was not detected at the Virjen site located 61.7 km downstream (a river over 100 m wide) even though eDNA could drift down this distance in less than 24 hours given that the speed of current at Virjen was around 0.7 m/s. It is therefore difficult to attribute this lack of detection to a total degradation of DNA, given that the half-life of DNA molecules in the aquatic environment is generally considered to be several days, even in tropical conditions [62, 63]. An additional explanation could be the increase in flow rate which, by increasing eDNA dilution, would decrease detectability as highlighted by [64] who showed the influence of both temperature and stream flow level on Corbicula fluminea (clam) detectability. In our case study, both the high water temperature (mean over sites = 25˚C, SE = 4.9) and the large and rapid increase in river discharge, are likely to limit the detectability of eDNA a few kilometres downstream of its emission. We can also assume that the length of our barcodes (185–285 bp), which is greater than those sometimes used, such as 12S (65 bp), would allow us to target less

degraded DNA, which has been recently released into the environment and has not diffused very much [65]. The fact that 73 taxa were only detected by the MK marker (185 bp) compared to the 17 detected only by MK2 (285 bp), tend to lend support to this hypothesis. The use of long barcodes might thus provide a more local image of the diversity detected. The only exception to this trend is the rainbow trout, Onchorhynchus mykiss, a freshwater Salmoniformes, that was unexpectedly detected in several sites of the Tuichi and Beni basins (S7 Table), usually with high identity (>98%). Given what was just explained about the limited diffusion of eDNA, this species, which lives in cold waters at high altitude should not be encountered in the lower portion of the Tuichi River, nor in the Beni River. A likely explana- tion, however, is the presence of many trout aquaculture facilities in the Andean portion of the Tuichi and in several of its tributaries. The presence of trout in very large, unnatural, concen- tration, as is the case in intensive aquaculture, could release large amounts of DNA in the waters that could disperse downstream in spite of the above-mentioned factors limiting eDNA diffusion. PLOS ONE Fish diversity and distribution in the Tuichi basin PLOS ONE | https://doi.org/10.1371/journal.pone.0262357 January 4, 2022 16 / 25