Discussion
Benthic impacts of marine finfish farms have been studied for decades (e.g. Hargrave, Duplisea, Pfeiffer, & Wildish, 1993). To improve benthic monitoring of fish farm impacts, novel approaches are being developed including new biological indices (Keeley et al., 2012) and more accurate measurements of pore-water sulphide concentration (Cranford, Brager, & Wong, 2017). Recently, eDNA metabarcoding based methods (e.g. including machine learning and using metabarcoding data in biotic indices) for benthic assessment have been proposed (Cordier et al., 2017, 2018; Keeley, Wood, & Pochon, 2018; Lejzerowicz et al., 2015). In this study, we further determined eDNA metabarcoding to be a useful complement to traditional macrofaunal benthic assessments of fish farms by analyzing biotic signals associated with benthic impacts of salmon farming in macrofaunal polychaetes data and eDNA metabarcoding data.
Polychaetes are an abundant taxonomic group in benthic macrofauna (Dean, 2008; Tomassetti & Porrello, 2005) and opportunistic polychaete complex (OPC) are considered good indicator species for environmental monitoring as they respond well to organic enrichment gradients (Pocklington & Wells, 1992). In this study, alpha diversity (richness) calculated using genetically identified polychaete data was consistently reduced near salmon cages compared to further out. This was consistent with previous findings that most polychaetes are sensitive to organic enrichment (Tomassetti & Porrello, 2005), and species richness of benthic fauna was reduced by marine finfish farming (Holmer, Wildish, & Hargrave, 2005; Neofitou, Vafidis, & Klaoudatos, 2010). The presence of Capitella at three stations close to cage edge (i.e. 0 m, 15 m, and 30 m from cage edge) and its absence at reference stations at all farm sites further supports that this genus is a good indicator for organic enrichment caused by finfish farming (Pearson & Rosenberg, 1978).
Consistent with results obtained using macrofaunal polychaete data, evidence of benthic biodiversity being impacted by fish farming across the six farms was also apparent using eDNA metabarcoding: there was a negative correlation between alpha diversity parameters and sediment pore-water sulphide concentration, and an increase in alpha diversity at distant stations. This indicates that alpha diversity parameters estimated from eDNA data can be used to analyze the spread (or ’footprint’) of benthic impacts at marine finfish farm sites. While other studies have shown eDNA metabarcoding generally provides similar conclusions for environmental assessments to those provided by morpho-taxonomy (Lanzén et al., 2016; Lejzerowicz et al., 2015; Pawlowski et al., 2014, 2016; Pochon et al., 2015), the community composition as revealed by these two methods can be very different (Cowart et al., 2015; Kelly et al., 2017). This was indeed the case in this study, as we detected a much higher number of Polychaeta species using morpho-taxonomy combined with DNA barcoding (145 taxa) than using eDNA metabarcoding (38 OTUs). The major reason causing this could be the difference in amount of sediments used: eDNA metabarcoding was conducted using ca. 1.5 g sediments per station (six 0.25 g samples) while macrofaunal Polychaeta were collected from 24 liters of sediments per station. Thus, it is not overly surprising that more species were detected by picking specimen from sieved sediments, despite that eDNA metabarcoding can detect much smaller species including meiofauna and microfauna. Another important reason could be the primers biases (Zinger et al., 2019): primers used for eDNA metabarcoding may not be very efficient for Polychaeta and/or DNA from other taxonomic groups (e.g. diatoms) were more competitive for PCR amplification. In addition, COI evolves more quickly than 18S and thus is more effective at differentiating species; this also likely contributed to the difference (Tang et al., 2012). Last but not least, public databases not having enough reference sequences for metazoans may have led to a low number of Polychaeta taxa detected in our eDNA data; there were 148 OTUs assigned as Eukaryota and not any further.
Both nematodes and polychaetes are known to be sensitive to organic enrichment caused by fish farming (Dean, 2008; Mirto et al., 2014; Mirto, La, Gambi, Danovaro, & Mazzola, 2002; Tomassetti & Porrello, 2005). In this study, nematode OTU number correlated much more strongly with pore-water sulphide concentration than polychaete OTU number did, and there were also many more nematode OTUs amplified than polychaete OTUs. This is presumably due to the small size of nematodes (Heip, Vincx, & Vranken, 1985) and that they spend their entire life cycle within sediments, whereas polychaetes are relatively larger-bodied, as well as varying sulphide tolerances and habitat strategies (Fauchald & Rouse, 1997; Pocklington & Wells, 1992). However, the proportion of benthic metazoan reads in our metabarcoding data was not high, which may have impacted the diversity of metazoan species detected. This is a common problem when using a general (i.e. all eukaryotes) 18S rRNA marker regardless of whether samples are collected in polluted areas (e.g. Lanzén et al., 2016; Lejzerowicz et al., 2015) or non-polluted areas (e.g. Sinniger et al., 2016). Increasing the proportion of metazoan reads should improve the accuracy of eDNA-based methods. Possibilities for achieving this include: treating sediments prior to DNA extraction, such as elutriation by decantation (Brannock & Halanych, 2015) and meiofauna extraction (Fonseca et al., 2011); the use of more metazoan-specific primers; and/or the use of a blocking primer to inhibit amplification of non-target species during PCR (Vestheim & Jarman, 2008). Extracting DNA from a higher amount of sediments also may increase number of species detected.
As already described, we assessed biotic responses to organic enrichment using alpha diversity richness parameters derived from all benthic metazoan reads and, more specifically, from two particular groups: nematodes and polychaetes. In addition, we used eDNA data to investigate responses to organic enrichment at the individual OTU level. Two OTUs showed higher relative abundances in oxic sediments and both of them were assigned to Sabatieria punctata . A previous study found theSabatieria genus had higher abundance near the fish cage than at the reference station (Mirto et al., 2002). This discrepancy in results between their findings and this study may be due to a different sample collection period. We collected samples during peak fish production while their monthly sample collection started 15 days after fish were transferred to the cage; populations of benthic fauna can increase when organic enrichment is slight and then decrease when organic enrichment is severe (Pearson & Rosenberg, 1978). We found OTUs assigned to chicken and Atlantic salmon in eDNA data. Chicken feather is used in fish feed diets and sediments were collected from Atlantic salmon farms. We believe they are true reads, although their numbers (9,959 for chicken and 6,371for salmon) were low compared to total number of reads (42.3 million) obtained. Given that these taxa were generally detected close to the net-pens, we feel that these occurrences were directly associated with farming activities.
We did not calculate biotic indices for the sediment samples analyzed here as none is available for this purpose in Canada. While environmental assessments based on geochemistry are very fast to perform, without biological information there is no way for benthic monitoring programs to assess cumulative impacts of fish farming on benthic communities (Lear, Dopheide, Ancion, & Lewis, 2011). Given that eDNA metabarcoding can provide a rapid estimation of species composition and diversity and has high potential to complement current monitoring programs, a few studies have proposed the use of eDNA metabarcoding for benthic environmental health status assessment. For example, Aylagas et al. (2014) used presence-absence of species based on metabarcoding data to calculate a genetics-based AZTI’s Marine Biotic Index (gAMBI). However, gAMBI requires OTUs to be accurately assigned to species level, and it uses only species in the AMBI database to calculate the biotic index, potentially rendering most OTUs in a given dataset unusable. Lejzerowicz et al. (2015) used eDNA and eRNA metabarcoding to calculate the Infaunal Trophic Index and three AZTI Marine Biotic Indices, and found that metabarcoding-based indices provided similar ecological assessments of sediments to morphology-based assessments, although only 19.5% of generated sequences were metazoan. Keeley et al. (2018) assigned OTUs from three metabarcoding markers to different Eco-Group categories, and then used OTU presence-absence data to calculate a few biotic indices for benthic impact assessments of fish farms. Those indices were shown to have strong linear relationships with the Enrichment Stage index used for benthic assessments of fish farms in New Zealand (Keeley et al., 2012, 2018). In our study, we categorized sediments into four organic enrichment statuses based on sediment pore-water sulphide concentration following the Canadian Aquaculture Regulation (AAR, 2016) and used three methods to identify OTUs related to organic enrichment. These OTUs are thus ideal candidates to test for the development of biotic indices to aid benthic impact assessments in British Columbia (and potentially similar environments elsewhere). Alternatively, if further study can confirm the consistency of the responses of these OTUs to organic enrichment status across a higher number of study sites, one or several of them may be amenable for use as a threshold-based bioindicator.