INTRODUCTION
Marine ecosystems deliver many important services to society such as climate regulation, coastal protection, food provision and recreation (Barbier, 2017; Duncan, Thompson, & Pettorelli, 2015). To maintain ecosystem services, good ecosystem health and proper ecosystem functioning are crucial (Barbier, 2017) and both are inextricably linked to marine biodiversity (Goodwin et al., 2017; Solan et al., 2004). Several legal and policy frameworks have been developed with conservation of marine ecosystem health as main objective (Goodwin et al., 2017; Halpern et al., 2015) and which include monitoring of benthic diversity. The characterization of benthic diversity relies mostly on morphological taxonomy, which is slow and expensive because of manual sorting and visual identification of taxa in the samples. The development of innovative, cost-effective monitoring tools that allow a quick screening of benthic diversity in large marine areas is therefore needed (Borja et al., 2016; Elliott et al., 2018).
DNA metabarcoding is a promising tool for monitoring benthic environments (Aylagas, Borja, Irigoien, & Rodriguez-Ezpeleta, 2016; Aylagas, Borja, Muxika, & Rodriguez-Ezpeleta, 2018; Elbrecht, Vamos, Meissner, Aroviita, & Leese, 2017; Leray & Knowlton, 2015). Instead of identifying all specimens morphologically, DNA is extracted from the total community, a short fragment of the genome is amplified through PCR, the resulting library is sequenced using high throughput sequencing and finally, the resulting sequences are bioinformatically processed (Pawlowski et al., 2018). Each of these steps can introduce bias and errors (Alberdi, Aizpurua, Gilbert, & Bohmann, 2018), and optimization of the methodological aspects of the protocol are needed to achieve reliable, comparable and repeatable data across studies. Different marker genes can yield very different diversity estimates of invertebrate communities and combining different markers increases the detection of all species (Alberdi et al., 2018; Marquina, Esparza-Salas, Roslin, & Ronquist, 2019). For metabarcoding metazoan communities, the mitochondrial COI gene is the preferred marker gene because of the availability of large reference databases and the presence of sufficient genetic variation to allow species level identification (Andujar, Arribas, Yu, Vogler, & Emerson, 2018). The final detection of a particular species in the sample greatly depends on its biomass and on the primers used in the DNA metabarcoding protocol (Elbrecht, Peinert, & Leese, 2017). The PCR step is one of the most influential steps for the detection of species using DNA metabarcoding, where primer mismatches, GC content, template switching, stochasticity and polymerase errors can affect the number of sequence variants found in the dataset (Kebschull & Zador, 2015). Different primer sets are available for the COI gene which also leads to differences in diversity (Braukmann et al., 2019; Elbrecht & Leese, 2017; Lobo, Shokralla, Costa, Hajibabaei, & Costa, 2017). Consequently, it is important to investigate primer performance for (bio)geographic areas and taxonomic groups that have not been studied before to obtain accurate results with DNA metabarcoding.
Next to primer choice, the DNA source used in metabarcoding studies can affect whether or not a species is detected. DNA can be extracted from bulk specimens that have been separated from the sediment by sieving, decantation and manual sorting (Aylagas, Mendibil, Borja, & Rodriguez-Ezpeleta, 2016) or from the ethanol preservative in which the sieved macrobenthos sample was preserved (Zizka, Leese, Peinert, & Geiger, 2019). Bulk DNA samples, can further be sorted in different size fractions to enhance detection of smaller sized animals (Elbrecht, Peinert, et al., 2017; Leray & Knowlton, 2015) which comes at an extra cost of time to process samples. Extracting DNA from the ethanol fixative avoids the time consuming step of having to sort specimens and has the added advantage that voucher specimens are still present after the analyses. However, non-target species such as bacteria and fungi can be encountered more when analyzing the ethanol preservative compared to bulk samples (Gauthier et al., 2020). A comparison between bulk and ethanol samples for freshwater invertebrates showed that the two approaches yield very different community compositions (Marquina, Esparza-Salas, et al., 2019; Zizka, Leese, et al., 2019). Nevertheless, DNA metabarcoding of the ethanol fraction has also shown considerable overlap with morphology based analyses of freshwater communities illustrating its potential for monitoring studies (Martins et al., 2019). In view of the above mentioned advantages when using the ethanol preservative, understanding the different results between bulk DNA and eDNA from the ethanol preservative is essential to determine their applicability for monitoring studies. For instance, it has been shown that small and weakly sclerotized freshwater insects species are overrepresented in the ethanol preservative while large and strongly sclerotized species are overrepresented in the bulk DNA (Marquina, Esparza-Salas, et al., 2019). Consequently, morphological traits of species may be important to explain their detection in DNA metabarcoding studies but this link has hitherto not been investigated for marine macrobenthos.
In this study, we sampled four distinct macrobenthos communities in the North Sea and identified them using traditional morpho-taxonomy before molecular processing. We generated 104 COI reference sequences from macrobenthos species in the Belgian part of the North Sea to allow taxonomic assignment of the metabarcode datasets. Our first aim was to find the best primer set to characterise marine macrobenthos communities through DNA metabarcoding. We tested five primer sets from the literature and defined the best primer set as the set that detected the highest number of morphological species and that was able to differentiate macrobenthic communities. Second, we investigated whether morphological and metabarcode analyses detected the same species. We expected to detect more diversity in the metabarcode datasets compared to the morphological dataset as specimens are not lost during the sorting process. We also expected to find a high proportion of species shared between the two approaches. Third, we evaluated whether eDNA from the ethanol preservative could be used for monitoring by comparing its alpha and beta diversity patterns with that of bulk DNA. Based on the results in freshwater communities, we expected to find beta diversity patterns that differentiate the four communities in the bulk DNA and eDNA from the ethanol preservative and to find additional non-macrobenthic taxa in the eDNA from the ethanol preservative. Fourth, we investigated whether morphological and ecological traits of the macrobenthos species could explain their probability of detection in the bulk DNA and eDNA from the ethanol preservative. Based on the results of insects in freshwater environments (Marquina, Esparza-Salas, et al., 2019), we expected to find a higher probability of detecting small and weakly sclerotized species in the ethanol preservative.