Abstract
Environmental impact assessment and regular environmental monitoring of marine life are prerequisites for the construction, operation, and decommissioning of offshore wind farms (OWFs). Molecular methods such as metagenomics, quantitative PCR or metabarcoding, are increasingly being considered as a possible complement or alternative to currently used marine baseline and monitoring methods, both for pelagic (open water) and benthic (seafloor) organism studies. Metabarcoding is one such molecular method that uses DNA sequence differences between species in specific so-called marker genes to identify organism biodiversity in a bulk specimen sample containing multiple animals, or environmental sample from water or sediment. While a simplification, a list of unique gene sequence variants (variously referred to as ASVs or OTUs in the literature) in a metabarcoding dataset can be conceptually considered similar to different species in a classical species list.
The following report is an assessment of the performance of such metabarcoding data – directly from DNA in sediment samples – in sediment impact monitoring as a complement to the current standard using species lists based on identification of >1 mm softbottom macrofauna. While recent eDNA studies have become more numerous, cumulative experience in particular habitats is still low: Thus this report should be considered an exploratory pilot study.
In metabarcoding, there are many gene markers available, which target different parts of the organism community in the sample (bacteria, protists, multicellular organisms etc.). Here, we sequenced two gene markers: the metabarcoding 18S rDNA ribosomal marker, regions V1-V2 (targeting eukaryotes in general) and the metazoan (animal) cytochrome oxidase subunit I (COI) marker. The 18S marker was expected to provide data on sequences from several protist groups, meiofauna (animals 1 mm); the COI marker is a well-established macrofaunal marker employed as the main animal marker in the Barcode of Life initiative, and it has the highest taxonomic coverage in online databases, such as GenBank, which allows more sequence variants to be identified to known species.
The aims were to (i) study 18S V1-V2 eukaryote and COI metazoan benthic community composition; (ii) assess the performance of the chosen eDNA sampling design, sample type, replicate number and molecular markers in detecting environmental impact; and (iii) compare how the metabarcoding results compare to morphological taxonomy of 1 mm sieved macrofauna from the same stations. The metabarcoding sediment samples were collected together with standard macrofauna samples at the Hywind OWF during spring 2022.
The study comprises 15 sampling stations at 110-120 m depth from two Hywind turbines (12 samples) and three reference stations: Sediment was collected at the same time as the standard monitoring parameters (processed separately by DNV) using van Veen grab sampling, frozen and transported to the NORCE lab in Bergen, where after DNA extraction the 18S rDNA V1-V2 and COI markers were amplified and subject to high-throughput metabarcode sequencing. Resultant sequence data was denoised, filtered and clustered using a custom dada2 and swarm pipeline to produce relevant ASV and OTU tables of community composition. These data are presented here together with the 1 mm sieved macrofauna monitoring data processed by DNV.
The DNV morphological dataset comprised 10 010 specimens from 212 species, with major phyla including annelids, mollusks, arthropods, and echinoderms. Biotic index values showed very good to good conditions for all stations and no environmental impact was detected. nMDS analysis showed that station replicates mostly separated into distinct clusters based on the separate sampling areas.
From a 25 million read dataset, the post-filtering 18S rDNA V1-V2 metabarcoding dataset produced 1 473 OTUs following a 2*10-5 abundance cutoff, mostly from the SAR supergroup (Stramenopiles, Alveolata, Rhizaria), but with a 16% 380 OTU metazoan fraction. High-level taxonomic groups were mostly consistent across stations. Shannon index biodiversity values were consistent in the total dataset and for single-celled organisms, intermediately consistent for meiofauna, and least consistent for macrofauna, indicating incomplete sampling for this fraction. Area nMDS separation was less clear than in the morphological data.
From an initial 37 million raw sequences, 1.6 million sequences could be assigned as metazoans in the COI dataset comprising 118 OTUs, the rest representing non-target bacteria or unidentifiable sequences. Metazoan sequences included nine phyla, unevenly distributed over the stations in the dataset. Shannon and sensitivity index values were variable across replicates and stations in the dataset, indicating insufficient sampling effort relative to the targeted organism group, and no clear pattern emerged from the nMDS analysis.
In the absence of detected environmental impact in any of these three datasets, we focused our analysis of the comparative performance of the datasets on sampling comprehensiveness and variability of the data relative to replicates and stations used in the study sampling design. Based on our results, we found that the morphological dataset was better able to show fine-scale differences in organism composition in the data based on the three areas sampled, while metabarcoding data was a bit noisier in this regard. The COI dataset exhibited the most variability between sites, followed by the 18S macrofaunal data. This shows that the sediment volumes used here could be insufficient for these “larger” animals. Single-cell protist and <1 mm meiofaunal 18S data were more similar between samples, but even so could not group the samples by area as well as the morphological data.
We conclude that i) while sediment eDNA has previously showed robust performance in detecting environmental impact on benthic communities, the sampling coverage used in this study did not provide equal-quality data to morphology in detecting small changes in species composition between closely related unaffected sites; ii) taxonomic coverage in online identification databases is still unable to identify large parts of metabarcoding datasets to low level (e.g., genus/species); iii) to lessen the effects of sample variability and increase the resolution of the metabarcoding data, a combination of concentrating on small/single-cell organisms and increasing sample effort, finetuning bioinformatic parameters, and further development of metabarcoding-specific biotic indices, is expected to raise metabarcoding data quality.