We used high-throughput qPCR to screen for 58 infective agents in four Atlantic salmon farm cohorts from British Columbia throughout their production cycles. We measured presence and copy number for target genetic sequences, characteristic of specific viral, bacterial, and eukaryotic agents, including several recently discovered viruses30,31, known or suspected to cause disease in salmon. These agents displayed various temporal patterns of prevalence and intensity, with several displaying elevated levels in dead and dying fish.
The data and analyses we have presented provide a unique look into the epidemiology of farmed salmon populations, and wildlife/livestock diseases generally. No past studies have had access to multiple farmed-salmon cohorts, throughout their production cycles, with the capacity to molecularly screen for a large suite of infectious agents. Other work has reported agent data for dead-sampled fish collected in BC as part of Fisheries and Oceans Canada’s farm audit program40, but such analyses lack the time-series nature of the results we have presented. To our knowledge, no other studies have provided such detailed, comprehensive information for infective agents in domestic or wild populations over time. This study, therefore, presents a substantial step toward effectively monitoring shared wildlife/livestock diseases, made possible by cutting-edge technology, as predicted previously22.
While our findings offer specific insight to salmon farmers, aquaculture managers, and those concerned with the disease ecology of sympatric wild salmon, we caution that results remain correlative, and relevant patterns (e.g. apparent mortality signatures) require further investigation. Unfortunately, a lack of regular testing of dead and dying fish (collection was opportunistic, at farms’ discretion) resulted in potential for associated patterns to be obscured. Due to this potential bias in the sampling design, we are unable to draw conclusions related to farm-level mortality rates, but several agents showed patterns of note, including elevated levels in dead and dying fish.
Agent patterns
Perhaps the clearest single-agent pattern—the elevated load of T. maritimum in dead and dying fish (Fig. 3B)—matches generally accepted patterns in aquaculture. Induced tenacibaculosis can be responsible for substantial on-farm mortality worldwide41, and mouthrot resulting from T. maritimum in the east Pacific causes substantial losses42. In our study, mouthrot was noted during veterinarians’ sample processing for cohorts one, three, and four in the months after ocean entry. We note that elevated levels in dead and dying fish could represent the bacterium’s acknowledged role as an opportunistic pathogen41, rather than a direct cause of mortality. We also note the positive correlations between T. maritimum load and that of a number of different agents (Fig. 5), consistent with the view that T. maritimum may facilitate co-infections43. Given its high overall prevalence in fish (Table 3), secondary factors—such as co-infections—might exacerbate infection with T. maritimum, playing a role in mortality.
K. thyrsites intensity was elevated in dead and dying fish for cohorts three and four, around the time they were transferred to their final marine locations (Fig. S8). In both cases, the cohorts finished their production cycles in farms on the central east coast of Vancouver Island (Fig. 1), a region in which the risk of K. thyrsites infection is acknowledged to be high44. This myxozoan parasite is economically important due to post-mortem myoliquefaction seen in infected fish, but it is not considered a pathogen45, and it is unclear why higher gene-copy levels would be observed in dead/dying fish. K. thyrsites may merely replicate faster in stressed fish (in this case due to transport). Our surveillance of pathogens did not include skeletal muscle tissue, where K. thyrsites spores develop, but it did include heart, which can be infected by the parasite46. We note that K. thyrsites was correlated with PRV, with both agents known to infect muscle tissue (although red blood cells are the primary infective tissue for PRV). Follow-up histopathological investigation may provide some insight into K. thyrsites distribution and any associations with pathology or patterns of co-infection.
PRV, which is the causative agent of Heart and Skeletal Muscle Inflammation (HSMI)47 and has recently generated controversy in BC28,29,48, shows several patterns of note. PRV prevalence increased to near ubiquity over time (Fig. 2D), concurrent with an increase, peak, slight decline, and then stabilisation of intensity (Fig. 3B). Although our perspective is limited to sampled fish, with a noted potential for bias, the observed PRV trends were consistent across all four cohorts, and the intensity patterns are consistent with previously reported dissemination, peak replication, and long-term persistence phases of the virus within hosts29,48. Past findings suggest that PRV may induce an antiviral response in hosts that can protect them against certain co-infections49,50. Perhaps counter to the generality of this claim, PRV and ASCV exhibited the strongest load correlation out of any we observed across our data set (Fig. 4). ASCV was originally isolated from salmon with HSMI, and was initially thought to play a role in the disease51. Other work has found no relationship between ASCV and PRV infections or HSMI52. In the case of a related baitfish calicivirus, however, there is evidence that viral co-infection is linked to disease manifestation53, so further work is needed to tease these relationships apart. In general, dead and dying Atlantic salmon in our study did not show elevated prevalence or intensity of PRV, except shortly after ocean entry in cohort one (Figs. 2D, 3B). This mortality signature corresponds to the onset of lesions diagnostic of HSMI in cohort one, which subsequently spread to affect the majority of that farm population for most of a year29.
The gill chlamydia bacterium, C. Syngnamydia salmonis, showed a consistent trend towards elevated prevalence in dead and dying salmon (Fig. 2B). Observed intensity was low, however, often averaging approximately a single copy (Fig. S15). Sequencing has validated past detections of this agent on the Fluidigm BioMark™, and has also revealed SNP diversity within the primer-binding region, resulting in potential underdetection (Miller et al. unpublished). Given a putative mortality signature and the lack of prior epidemiological study of this recently discovered agent54,55, we would suggest further work on C. Syngnamydia salmonis.
Ephemeral mortality signatures appeared for several agents. F. psychrophilum was clearly elevated in dead and dying fish in-hatchery, although we only had access to two hatchery cohorts and cannot draw general conclusions. Intensity of both the ASCV and CTV-2 were elevated in sampled dead and dying fish from at least one cohort shortly after ocean entry (Figs. 3A, S2). Both viruses were also present in-hatchery. The previously reported Cutthroat trout virus appears to be apathogenic56 in trout, and has been detected in Atlantic salmon57. Little is known about the novel variant for which we screened, although in situ hybridisation has revealed that infection can be systemic and extensive in the brain (Mordecai et al. 2020). As for the ASCV, associated pathology was found to be likely due to PRV contamination51. Extremely limited information about these two viruses, paired with our findings, warrants further investigation (e.g. with histopathology and in situ hybridisation) to determine if either virus is linked with pathology. As both these viruses were detected in Chinook salmon (Mordecai et al. 2020), and considering their high prevalence in Atlantic salmon farms, the potential risk they pose to wild Pacific salmon populations should be a priority for future research.
Infectious agent levels overall, as measured by relative infectious burden, showed a clear trend in smolts coming out of freshwater hatcheries for cohorts three and four. There, infectious burden was much higher (in one case 10 000 times higher) in dead and dying fish than in live-sampled fish. While the effect dissipated once fish entered the marine environment, it is clear that hatchery fish are dying with—or of—elevated levels of infection. The patterns we observed likely reflect the transition from freshwater to saltwater, with a coincident shift in infective-agent communities58. Smoltification has also been associated with immune depression59, and elevated infectious burden around the time of ocean entry may reflect this. Where we had dead/dying hatchery samples, however, infectious burden was elevated weeks before ocean entry, hinting at the potential for problems in-hatchery.
Across all agents, we observed apparent coinfection signals that clearly differed from random chance (Figs. 5, S21). We point out, however, that due to the longitudinal nature of our study and cursory investigation of agent correlations, the correlation results almost certainly indicate shared temporal trends that may or may not indicate underlying interactions. For example, Candidatus Branchiomonas cysticola and Flavobacterium psychrophilum were positively correlated with each other, but negatively correlated with a number of other agents. This could be due to both agents being common in-hatchery but not in the marine environment (Figs. S3, S6), counter to many other agents’ patterns. Alternatively, the correlation could be due to a biological relationship, perhaps in relation to gill health. We intend to follow up on a number of correlational patterns.
Agent idiosyncrasies
Several agents showed unexpected patterns, or patterns that may be connected to their particular biology.
The putative Narna-like virus, a recently discovered agent31, showed elevated prevalence in dead and dying fish (Fig. S9). This pattern was mainly due to over-representation in dead-sampled fish, as we detected the agent in 13.2% of dead fish, 1.6% of moribund fish, and 0.4% of live-sampled fish in saltwater. Given that Narnaviridae, of which the putative Narna-like virus is a member, is thought mainly to infect fungi60, this virus may be associated with a fungal decomposer. This is speculative, however, and recent genomic evidence from across taxa suggests that the Narnaviridae may be much more widespread than previously thought61.
Counter to the common trend, P. pseudobranchicola tended to be less common in dead and dying fish than in live-sampled fish (Fig. 2C), with dead fish, in particular, tending to exhibit the lowest levels (results not shown). P. pseudobranchicola primarily infects the pseudobranch62, a structure near the gills involved in oxygenating blood in the eye. Infection also occurs in tissue collected for this study, especially gill63, and we speculate that loads in dead fish could be reduced due to myxospore release or degradation of delicate gill tissue after host death. Given that we did not sample the pseudobranch, it is likely that our data underestimates the load of this organism.
The sampling environments (freshwater or marine) of several detections were unexpected. In particular, we detected K. thyrsites and T. maritimum (Fig. 3C) in freshwater hatcheries, although these agents are considered marine species64,65. It is possible that these hatcheries introduced saltwater in the weeks before ocean transfer, to prepare smolts for release. We also detected F. psychrophilum, considered a freshwater bacterium66, in marine net pens (Fig. 2A). The bacterium is known to survive in brackish water67, however, and this is not the first time it has been detected in a marine setting40,68.
Broader connections
Not all infective agents cause disease, and even agents that do can be present long before—or long after—clinical symptoms. Our work presents only a piece of the puzzle in what is a multifaceted, complex scenario of shared wildlife/livestock disease in salmon aquaculture. The controversy surrounding PRV in BC, as an example, illustrates this complexity. While conventional lab challenges using PRV from BC sources have failed to reproduce in BC fish the extent of HSMI lesions observed on Norwegian farms48,69, work related to our study has been able to identify and shed light on HSMI, and related jaundice/anemia in Chinook salmon, in BC salmon farms28,29. While we saw a putative mortality signature in one cohort during this study, the normal course of PRV infection was not always associated with mortality (e.g. Figs. 2D, 3B). More work will be required to elucidate the nuances of PRV infection, factors that induce associated disease, and possible resultant mortality. A fruitful place to start would be to carry out sampling and diagnostics of dead and dying fish in farms and pens experiencing elevated mortality.
Although we have shown putative mortality signatures for several infective agents in farmed Atlantic salmon, these are not necessarily the agents that pose the greatest risk to wild salmon. For one thing, a given agent need not produce the same effects in different species28,70. For another, contact between populations may not coincide with infection maxima. Depending on when farm smolts enter the marine environment, for example, PRV could be at low prevalence in the spring, when a number of wild Pacific salmon species migrate as juveniles15. Other times of year would be more relevant for interactions with other wild species, and there is much scope for transfer between farmed and wild environments. In addressing shared wildlife/livestock disease, we need to consider both wildlife and livestock as populations that serve as potential reservoirs of disease agents, and are susceptible to outbreaks71. In this context, surveillance and monitoring are essential facets of disease management23, providing raw material to develop understanding of disease and build effective management strategies. Parallel work is monitoring wild populations for the same agents we have investigated here, with the prospect of cross-referencing patterns and impacts72,73,74.
The ubiquity of infectious agents on the farms leads naturally to discussion of potential control strategies, which present a variety of challenges in aquaculture. Vaccination has proven successful at times, but the salmon aquaculture industry has a somewhat chequered history with uptake, since vaccination can affect host growth, and thus the bottom line13. In addition, vaccines have only been developed for a handful of agents. Reducing translocations can be an effective control strategy on land20,78, but transmissive properties of the marine environment and highly mobile marine carrier hosts pose challenges to isolating host populations geographically (Krkosek 2015). Our findings provide circumstantial evidence that some agents (e.g. K. thyrsites) respond to translocations. The fact that two of our four focal cohorts moved substantial distances throughout their respective marine production may be cause for concern, considering the infective-agent populations we have shown those cohorts to have harboured. In general, aquaculture-associated disease and related management decisions have a history of generating political controversy75. Infective-agent monitoring and analyses are critical for designing, implementing, and evaluating effective disease-control measures, and for bridging divides in debate surrounding aquaculture.
With respect to the aquaculture industry, the tools we employed in this study may prove useful for disease management and fish health. We have shown that for many agents, patterns of infection in dead and dying fish mirror those in live fish. By integrating high-throughput infectious-agent screening with existing monitoring of dead fish, farm vets and managers could access a wealth of otherwise unavailable or costly information. Combining such results with strategic sacrificial sampling of live fish during mortality peaks could allow additional insight into which agents may be driving mortality. Protecting the ‘herd’ (and its wild neighbours) may justify such mortal sampling. Furthermore, in other ongoing work, we have seen that much infective-agent information is accessible via nonlethal gill biopsy, which also enables high-throughput screening for gene expression patterns associated with various patterns of stress76 and disease77. Used appropriately, such a combination of tools could be very powerful.
Disease monitoring is never complete, and detection always lags behind pathogen spread78, but new technologies—such as those we employed here – can facilitate efficient, lower-cost surveillance and monitoring. Surveillance for existing pathogens and identification of previously unknown pathogens is part of the integrative approach required to understand and control existing and emerging infectious diseases22. Here, we have further demonstrated the utility of high-throughput, modern genetic techniques for monitoring known infective agents and for generating information about previously under-studied agents26,29,30,31,37,38. Further work will target the risk of transfer between wild and farmed hosts and prioritize threats to salmon, farmed and wild.
Source: Ecology - nature.com
