Biodiversity assessment of marine benthic communities with COI metabarcoding: methods and applications

Ecosystem biomonitoring is crucial for proper management of natural communities during the Anthropocene era. With the advent of new sequencing technologies, DNA metabarcoding has been proposed as a game-changing tool for biomonitoring. In this Thesis we plead for the use of metabarcoding of a highly variable marker to infer not only the interspecies but also the intraspecies variability to assess both biogeographic, at the species level, and metaphylogeographic patterns, at the haplotype level. We focused on highly complex hard-substratum benthic littoral communities. The term "Metaphylogeography", coined in this Thesis, refers to the study of phylogeographic patterns of many species at the same time using metabarcoding data. However, as of the start of this Thesis, only a few studies had tested the metabarcoding method to directly characterize the whole eukaryotic community in highly diverse benthic ecosystems. This required to set up and calibrate methods for these communities as a prior step. We first evaluated both the sampling methods and the bioinformatic pipelines. We assessed the viability of detecting the environmental DNA released from the benthic community into the adjacent water layer using metabarcoding of COI with highly degenerated primers targeting the whole eukaryotic community. We sampled water from 0 to 20m from shallow rocky benthic communities and compared the DNA signal with the results obtained from metabarcoding directly the benthic communities by traditional quadrat sampling. We also designed a pipeline combining clustering and denoising methods to treat metabarcoding data of COI. We considered the entropy of each codon position of this coding fragment both to improve the detection of spurious sequences and to calibrate the best performing parameters of the software used. In addition, we created our own denoising program, DnoisE, to incorporate information on the codon position. This new code and parameter calibration were required as the co