Revealing the potential and limitation of eDNA metabarcoding inventories
Fishes are the most diverse group of vertebrates on Earth with varying body sizes, environmental niches and diets. Monitoring fish assemblages in marine biodiversity hotspots like the Coral Triangle is a great challenge particularly for small, rare, cryptobenthic or elusive species. Here we show that the percentage of sequenced species is highly variable among families preventing any robust estimation of species richness. Instead Operational Taxonomic Units have the potential to reveal the presence of a broad range of fish species, i.e. from different lineages and with contrasted life-history traits. For instance, cryptobenthic families have been poorly documented and are often ignored in traditional visual censuses ((7) Brandl et al 2018) while they strongly influence ecosystem functioning ((13) Brandl et al 2019). Similarly, traditional visual censuses often miss highly mobile and elusive species such as sharks ((9) Boussarie et al 2018).
Among the 310 assigned fish species, we detected the presence of small cryptobenthic species such as Gobiodon histrio orOstorhinchus selas , a goby and a cardinalfish with a maximum length below 40 mm, respectively. We also detected large pelagic fish such as the dogtooth tuna (Gymnosarda unicolor ) or the thresher shark (Alopias pelagicus ) reaching over 2 m and 4 m long, respectively. Flagship species for conservation were also present in our DNA samples such as the over-exploited Napoleon wrasse (Cheilinus undulatus , Endangered, IUCN redlist,www.iucnredlist.org), the Scalloped hammerhead shark (Sphyrna lewini , Endangered) and several shark species being classified as Near Threatened (NT) (C. brevipinna , C. Leucas , C. sorrah , C. melanopterus ,T. obesus ).
Even if not assigned at species-level, OTUs can be defined as distinct entities for which their distribution and temporal variability can be assessed and monitored ((41) Cordier et al 2017). Moreover, the OTUs and their associated sequences can remain in public repositories until they are assigned to a species, subspecies or complex as databases improve ((42) Wangensteen et al 2018). However, the major caveat of using OTUs for diversity inventories is that they cannot be directly considered as species with complete certainty. Species with intra-specific genetic variability can produce two separate OTUs, overestimating species diversity. Conversely, two species phylogenetically close to each other with low genetic variability can be grouped into a single OTU, thus underestimating species diversity. The accuracy of diversity inventories using eDNA metabarcoding is thus directly based on the taxonomic resolution of the barcode used and genetic variability among families but also the number of samples.
Here we also reveal the gap of biodiversity that remains to be detected using OTU accumulation curves. The effort can be massive for some families (Fig. 6) and more ambitious eDNA sampling campaigns should be on the agenda in species-rich regions like the Coral Triangle. OTU accumulation curves can also serve to evaluate the efficiency of a sampling method (e.g. punctual filtration, transect filtration), the sampled area or the diversity of habitats that are required (e.g. depth, complexity, distance from the seafloor) and their location (e.g. proximity of reefs, hotspots) especially when targeting rare, elusive, highly mobile or cryptobenthic families of fish.
The contrasts between assigned taxa diversity, OTU diversity and OTU asymptote diversity show that the detectability varies strongly among fish families. These contrasts can be related to the ecology of the species but also to the state of the retrieved DNA fragments (intra or extracellular), their sources (e.g. gametes, larvae, feces), their release rate (different and their diffusion in the water column (limited or wide). For instance, a benthic fish species such as gobies with a small movement range would release DNA fragments through skin and feces on a small area. However, such species could release a massive number of gametes carried through the water column ((13) Brandl et al. 2019) so may appear highly detectable during breeding season. Further comparative works are urgently needed between visual, camera and eDNA metabarcoding surveys to better estimate the level of detectability of each species or family in order to provide reliable biodiversity assessments. For instance, coupling eDNA metabarcoding and video surveillance allows the detection of eighty-two fish genera from 13 orders on reefs and seagrass with only 24 genera in common ((43) Stat et al 2019). Investigating biodiversity should also consider its multiple components including functional and phylogenetic diversity that are key for reef ecosystem functioning ((44) Duffy et al 2016). Associating OTUs to species might allow to fill this gap but it will require massive sampling and sequencing efforts.