Infectious agent release and Pacific salmon exposure at Atlantic salmon farms revealed by environmental DNA

Abstract

The potential risk posed by infectious agents (IAs) associated with netpen aquaculture to wild fishes is determined based on the "release" of IAs from netpens into the environment, the "exposure" of the wild fish to those released agents, and the "consequence" for wild fish experiencing infection by those agents. Information available to characterize these three factors is often lacking, and the occurrence of transmission from aquaculture to wild fish as well as potential consequences of such transmission are difficult to observe. In this study, we utilized environmental DNA (eDNA) to characterize the release of dozens of IAs from, and exposure of Pacific salmon to, Atlantic salmon aquaculture. We combined these factors with the consequence of infection, as determined by the literature, to identify IAs that may pose a risk to wild salmon exposed to aquaculture in British Columbia, Canada. Over an 18-month period, eDNA samples were collected from seven active and four inactive netpen aquaculture sites in the Broughton Archipelago, BC. A meta-analytical mean across 22 IAs showed that the odds of IA detection at active sites was 4.3 (95% confidence interval = 2.3:8.1) times higher than at inactive sites, with 11 IAs in particular demonstrating a pattern consistent with elevated release. Oncorhynchus tshawytscha was the only Pacific salmon species presenting eDNA detections more likely to occur around and within active netpens relative to inactive sites. After considering the evidence of negative consequences of infection (from previous literature) in tandem with release model results, we determined that Tenacibaculum maritimum, Tenacibaculum finnmarkense, Ichthyobodo spp., and Piscine orthoreovirus are potential risks to Pacific salmon exposed to marine netpen aquaculture. These IAs, and others demonstrating patterns consistent with release but with insufficient prior research to evaluate the consequences of infection, require further studies that identify the factors influencing the intensity of release, the spatial extent of release around netpens, and the prevalence of infection in wild fish within known distances from netpens.

Fig. 1

Risk to wild fish caused by infectious agent transmission from netpen aquaculture requires the intersection of release, exposure, and consequence. These three factors of risk analyses are represented in the three circles. Presence of two but not three of these factors does not indicate risk, and the outcome of two factors in the absence of the third is described in their overlaps in the figure.

Fig. 2

Raw data from the Broughton Archipelago reveal that T. maritimum DNA (orange points) was more likely to be detected in water samples from active farms compared to inactive or fallow sites, T. maritimum was frequently detected in S. salar tissues (green points), and O. tshawytscha eDNA (blue points) was more likely to be detected in water samples from active farms and was commonly detected in the presence of T. maritimum. Points exceeding 40 (max Ct) indicate samples were collected but the target nucleic acid was not detected. Black vertical dashed lines indicate S. salar stocking dates and red vertical dashed lines indicate harvest dates (thus farms are fallow after red lines). Note that Cypress Harbour was a broodstock facility so that occasional, but never complete, harvest occurred and Doctor Islets was the only farm to be harvested and then restocked during the study period (fallow from November 2021 to March 2022).

Fig. 3

Plots A–C depict odds ratios for sample location (inactive site versus active farm) in generalized linear mixed models of eDNA presence. Points indicate the odds ratios and horizontal lines represent 95% confidence intervals around the odds ratios. An odds ratio with a 95% confidence interval exceeding 1 (red vertical line) indicates that the eDNA of that taxa is more likely to be detected at an active farm than an inactive site.

Fig. 4

Consequence scores (left panel), and accompanying uncertainty scores (right panel), were created for each modeled infectious agent based upon published literature. The consequence score was a composite of the weight of evidence from three categories: challenge studies, histological and/or clinical signs, and field or epidemiological studies (scoring rationale and references provided in Table S1). For each category where no information was available, or only pertained to S. salar, a point was added to the uncertainty score.

Fig. 5

Infectious agents are positioned to demonstrate the interplay between their likelihood of release associated with aquaculture (x-axis (log-scale)—odds ratios from models in this study and a reanalysis of data from Shea et al.) and the consequence of infection (y-axis—consequence scores determined from the literature; see Figure 4, Table S1). The consequence axis is not linear, IAs are ordered by consequence score (for visibility) and horizontal dotted lines indicate unit divisions between those scores. For the consequence axis, overall scores are a composite of negative impacts demonstrated from challenge studies, clinical signs and/or histopathology, and field or epidemiological studies (Table S1). The font size of the infectious agent names indicates when literature is lacking for Pacific salmon (i.e., uncertainty)—the smallest text indicates no publications featuring Pacific salmon exist for any of the aforementioned topics, and the largest text indicates at least a single study featuring Pacific salmon exists for all components of the consequence score. Error bars indicate 95% confidence intervals around the estimate. The vertical dashed lines and surrounding ribbons represent the means and 95% confidence intervals from a meta-analysis (Figure S29) of the multiple models from each study, corresponding by color.

Fig. 6

Schematic outlining the study design. Environmental DNA collections were used to characterize infectious agent release at active farms and exposure of wild fish. Evidence from the literature was used to develop agent-specific consequence scores.

Fig. 7

Environmental DNA samples were collected from Atlantic salmon netpen aquaculture sites in the Broughton Archipelago (A), located on Canada’s West Coast (B), on the north end of Vancouver Island in British Columbia (C). In panel A, red circles and text identify active farms and blue squares and text identify decommissioned farms. Panel B shows the Broughton Archipelago (red star) on a map of North America. Panel C shows all salmon aquaculture netpens on the British Columbia coast (at the time of the study) and around Vancouver Island (VI) as well as the Broughton Archipelago (red rectangle). This map was created by the authors in R version 4.3.0 (https://www.r-project.org/) using shoreline data from the Global Self-consistent, Hierarchical, High-resolution Geography Database Version 2.3.7 (https://www.soest.hawaii.edu/pwessel/gshhg/).