Tropical marine biodiversity through a molecular lens: eDNA metabarcoding across oceans, habitats, and time

Environmental DNA (eDNA) metabarcoding is a relatively new tool in the field of marine ecology, yet in the last decade its use has grown exponentially. Encouraging metabarcoding efforts that support and/or augment conventional survey results have paved the way for implementation of molecular based monitoring, yet it is unclear whether the tools should be directly compared or validated with e.g., visual methods. Afterall, the sensitivity of molecular methods goes far beyond what can be seen, even with a microscope. Furthermore, the tools are also still held back by the paucity of reference sequences in online databases. In this thesis, I examined the use of eDNA metabarcoding in the context of highly diverse tropical marine habitats. In chapter two, the ability of metabarcoding to reflect currently accepted biodiversity patterns was assessed. I used standardised methods across two vastly different yet characteristically similar regions, namely the Caribbean Sea and the Western Indian Ocean, to be in a position to directly compare how metabarcoding performs in different environments. Despite quite a range of habitats being studied with metabarcoding to date, to my knowledge no one has done a comparison of this scale to evaluate metabarcoding performance. The results of this chapter show clearly how an understanding of underlying biodiversity levels is crucial for effective experimental and monitoring design, as well as bioinformatics processing when using metabarcoding approaches. Chapter three takes a closer look at one of the sites studied in chapter two, by analysing temporal data from a reef and a mangrove habitat in Tela, Honduras. Metabarcoding efforts often suffer from a lack of resolution as a result of gaps in reference databases, and therefore this chapter examined the differences in spatio-temporal patterns that could be found whether using taxonomically higher resolution data or data that purely passed through bioinformatics processes and was thus assumed to be of good quality. The taxonomically unassignable data was found to be more informative in terms of finding temporal changes, yet it is clear that taxonomic information of some level is required to draw conclusions and make predictions from the observed changes. With this in mind, chapter four presents a new tool for investigating the so-called “dark matter” of metabarcoding; the conventionally unassignable molecular operational taxonomic units (mOTUs) or actual sequence variants (ASVs). Instead of strict taxonomic assignments, this tool uses a phylogenetic placement approach of your query reads to a large reference tree of COI and COI-like (i.e., of bacterial and archaeal origin) sequences currently available in online databases. By applying this tool to metabarcoding data from various projects, we found further evidence that a large proportion of the resulting mOTUs/ASVs from marine eDNA samples originate from bacterial sources. In chapter five, to address this issue, I developed a capture probe protocol with the aim to isolate target metazoan DNA templates from eDNA samples, or alternatively remove the abundant bacterial DNA templates, prior to the amplification step of a conventional metabarcoding effort. In order to fully understand the effects of the protocol, all of the steps were evaluated for changes in the abundance and richness of COI and bacterial 16S, by amplifying and sequencing for both. The results clearly demonstrated the nonspecificity of popular COI primers, but also revealed the range of DNA that capture probes are able to pull out from eDNA extracts, providing a basis for suggesting an exciting new approach for isolation of target DNA.