Moving towards improved surveillance and earlier diagnosis of aquatic pathogens: From traditional methods to emerging technologies

Abstract Early and accurate diagnosis is key to mitigating the impact of infectious diseases, along with efficient surveillance. This however is particularly challenging in aquatic environments due to hidden biodiversity and physical constraints. Traditional diagnostics, such as visual diagnosis and histopathology, are still widely used, but increasingly technological advances such as portable next generation sequencing (NGS) and artificial intelligence (AI) are being tested for early diagnosis. The most straightforward methodologies, based on visual diagnosis, rely on specialist knowledge and experience but provide a foundation for surveillance. Future computational remote sensing methods, such as AI image diagnosis and drone surveillance, will ultimately reduce labour costs whilst not compromising on sensitivity, but they require capital and infrastructural investment. Molecular techniques have advanced rapidly in the last 30 years, from standard PCR through loop‐mediated isothermal amplification (LAMP) to NGS approaches, providing a range of technologies that support the currently popular eDNA diagnosis. There is now vast potential for transformative change driven by developments in human diagnostics. Here we compare current surveillance and diagnostic technologies with those that could be used or developed for use in the aquatic environment, against three gold standard ideals of high sensitivity, specificity, rapid diagnosis, and cost‐effectiveness.

exacerbated by environmental disturbance (habitat loss or destruction, pollution, urbanisation, ocean acidification, climate shift; reviewed by Cable et al. 6 ), population density, diet and intrinsic host factors (immune status, genetics, life-stage and reproductive status 7,8 ).
). The old adage 'prevention is better than cure' still applies with regards to control of infectious disease, but the wider impacts need to be considered if prevention, for example, contributes to antimicrobial resistance or other environmental impacts. Nonchemical interventions, good husbandry, stress reduction, environmental enrichment, dietary supplements, water quality maintenance, stock movement restrictions, quarantine measures, genetically resistant stocks, and regular surveillance all contribute to prevention, 9 but complete harmony is difficult to achieve. 10 Even the best management strategies cannot guarantee protection from disease outbreaks and effective mitigation requires early detection diagnostics: identifying the pathogens, and if possible, quantifying them.
Typically, fish health is first assessed visually through general indicators such as behaviour and appearance. Routine monitoring of fish health is more challenging than for terrestrial livestock due to variable and fluctuating water conditions. Turbidity, sediment type, turbulence and the weather can all affect visibility and obscure detection of clinical signs. 11,12 Like any infectious disease, early diagnosis of aquatic pathogens is vital to minimise morbidity and mortality; once a pathogen or group of pathogens is identified, early intervention can reduce the chances of mass mortalities. For parasites such as Saprolegnia parasitica which cause rapid host death (24-48 h) with no effective cure, early diagnosis is key to reduce population-level losses. 13 The goals for early diagnosis can be categorised under four pillars: sensitivity, specificity, speed and cost (infrastructure, consumables and labour).
This review assesses the range of early diagnostic techniques currently used in aquaculture, the ornamental trade, wild fisheries and aquatic research, and considers future developments. As novel diagnostic techniques are brought to the forefront for human health, greatly accelerated by the SARS-CoV-19 pandemic, this provides potential for translation to animal health methods. Early detection and identification of problem pathogens will allow for effective implementation of control strategies minimising losses and the spread of infection.

| CONSIDERATIONS WHEN SELECTING AQUATIC DIAGNOSTICS
As Emerging (and re-emerging) Infectious Diseases become more common, we must consider technologies utilised in other fields or currently in development for use in aquatic systems, bearing in mind the Technology Readiness Level (TRL; scaled 1-7). This metric defines the maturity of a technology in relation to development, with one reporting the research backing the technology and seven representing the operational testing stage. 14 Diagnostic techniques showing promise with a TRL 1-3 are in their infancy and will require further development before implementation. Although the TRL is primarily applied to terrestrial technologies, it does flag technologies that could be transferred to aquatic systems but doing so is not simple as there are significant challenges regarding the variable and dynamic aquatic environment.
The natural aquatic environment is constantly in flux and resident fish are subject to variations in water quality, oxygen concentrations, light levels, enrichment, competitors and predators, all potentially influencing disease susceptibility. These factors also impede disease surveillance, for example, through difficulty in observation and sample obtainment. Many fish, especially those in the ornamental trade, are transferred long distances to reach the end user and this movement also increases susceptibility and disease risk through mechanical disturbances 15 and reduced water quality from increased CO 2 and buildup of other toxic compounds. 16 Within intensive aquaculture systems, water quality including dissolved oxygen levels are controlled, but stocking density is often pushed to its limit, which can also affect disease susceptibility. 17,18 For many species, high densities increase stress, as is the case with Atlantic salmon (Salmo salar) resulting in increased disease susceptibility. 18 For territorial species, such as Nile tilapia (Oreochromis niloticus), high densities can lower stress, as social aggression is reduced 19 and consequently so too is disease susceptibility. 20 So, disease mitigation is critically dependent on the system and species. The number of aquatic species cultured greatly outnumbers those in terrestrial environments, with around 600 aquatic species farmed commercially. 1 This means there is no "one-size-fitsall" solution for aquatic diagnostics and each method must be tailored towards the culturing system and species.
Resources for aquatic disease diagnosis arise from academic, governmental, and independent organisations. They vary greatly across sectors and geographic regions, and all rely heavily on local specialist knowledge. Within intensive aquaculture, commercial diagnosis routinely utilises off-site or company veterinarians and scientific laboratories, particularly when the pathogens are cryptic. 21 For aquafarmers with limited or no technology including internet access, alternative diagnostic technologies such as tele-diagnosis systems can be employed. 22,23 With growing consciousness of the effects of overfishing on global aquatic ecosystems, funding is being put in place to aid transitions to sustainable fishing and the development of aquatic and coastal jobs. Ensuring sustainability is a concern and efforts vary globally. The European union put in place the European maritime and fisheries fund (EMFF) to support sustainability, 24 with funding split between fisheries and aquaculture, monitoring and enforcement of rules, data collection to improve future knowledge, and to the blue economy through creation and growth of marine jobs. In Asia, the fisheries refugia approach was implemented with the goal of bringing together the fisheries and environmental sectors of the South China Sea, aiming to reduce fishing pressures and aid in habitat management. 25 With the outcome of the fisheries refugia concept resulting in local sustainability of target species, such as lobsters (Panulirus spp. and Thenus orientalis) and tiger prawns (Penaeus monodon) by implementing seasonal closing so that the populations can recover. 26 Projects such as the fisheries refugia allocate areas, however, one key issue with aquaculture is site occupation, with farms requiring large areas for enclosures and associated infrastructure. Open water systems pose additional problems for disease, with spillover/spillback effects between natural and farmed populations. 27 One approach to combat this is the development of inland 'mega-farms', self-contained units, which prevent disease transmission between wild and farmed fish, allowing treatments to be more targeted thereby reducing pollution. 28 For recreational angling, city centre fisheries provide those with limited countryside access an 'authentic' fishing experience from within the city limits. Indoor angling prevents fish from being impacted by weather conditions, inflowing pathogens, invasive nonnative species and predators, but requires large setup and maintenance costs. Similar small inner-city venues for small scale locally produced food are appearing with tilapia, for example grown alongside salad crops in aquaponic systems. 29 All these onshore/inland facilities face optimisation challenges, with husbandry and housing conditions (e.g., lighting, enrichment and flow rate) varying between species and facility, in addition to very strict biosecurity, which is why diseases in these facilities have not been eliminated. 9 As productivity of these indoor aquatic industries is still limited by infectious disease, the development of novel diagnostic techniques is vital for continued growth.
The health of farmed fish and responsible usage of aquatic resources is managed across different scales; from local/regional to trans-national and global efforts. On a regional or national level, fish health may be managed by governmental agencies, such as the UK Centre for Environment Fisheries and Aquaculture Science 30 and the National Oceanic and Atmospheric Administration (NOAA). At an international or transnational level, the Asia-Pacific Fishery Commission (APFIC) 31  INFOFISH and GLOBEFISH, provide information to fisheries worldwide. Aquaculture and the ornamental trade may also benefit from the advice of nutrition companies. Food additives are increasingly included in fish diets to boost the immune system to reduce disease susceptibility. 33,34 If farmers are experiencing problems with specific pathogens, then specialist vets can provide targeted advice to combat the infection. However, there is an increasing number of emerging diseases, such as puffy skin disease or red-mark syndrome, for which the causal agents are unknown so relying on treatments/interventions by vets is problematic. 35 All fish stocks need to be regularly surveyed for pathogens, but progressive budget cuts over recent decades have reduced routine surveillance, such that now surveys only tend to be conducted for research or in response to a disease outbreak. 36 This is a global problem, especially in Europe, Asia, Africa and South America, with survey results suffering bias through false or inaccurate reporting, which further complicates risk assessments. 37 Without regular surveys of fish health, prevention (and indeed early warning of wider ecosystem problems) becomes increasingly difficult, but early diagnostics can at least help maintain fish health of current stocks. The next three Sections (3)(4)(5) cover the three main categories of diagnosis, visual, cellular and molecular, whilst providing details on specific techniques and example pathogens to highlight how such techniques have been applied.

| VISUAL DIAGNOSIS
Visual diagnosis can range from traditional methods of noting changes in behaviour and condition to remote sensing through drones and AI diagnosis ( Figure 1 and Table 1).

| Visual observation for clinical signs and diagnosis
In situ, aberrant behaviour of fish, often followed or accompanied by altered physiology or morphology, are typically early indicators of ill health, often observed via manual surveillance. Common clinical signs include increased opercular rate, gasping at the surface, loss of equilibrium, lesions or abrasions, and string-like faeces. 38  Microscopy is often the next step in visual diagnosis, accuracy of which is again dependent on the expertise of the observer. For microscopic diagnostics, mucus scrapes or tissue sections of the fish are commonly utilised. For example, Chilodonella hexasticha, a ciliated protozoan fish parasite, can be visualised from skin/mucous scrapes without the need for staining, 42 likewise for larger pathogens such as Diplostomum or Trichodina species. Microscopic diagnosis relies on the pathogen being morphologically distinct, which within the cacophony of aquatic pathogens, is a rarity. For gyrodactylids, with >400 Gyrodactylus species described, the majority are morphologically cryptic, requiring sequencing, or electron microscopy, to differentiate species. 43 For the many thousands of Gyrodactylus species, and other fish pathogens, as yet undescribed, sequencing alone is problematic without a morphological reference description, so a combined approach is required. 43 Other than equipment and labour costs, light microscopy is relatively cheap, but the main caveat is user error, which affects the specificity of diagnosis and means low level infections can be overlooked. Diagnosis of fish disease through these traditional methods is highly skill dependent, with variation occurring between the individual carrying out the diagnosis. 44 Microscopy can generate quantified data, but again is dependent on the accuracy of the diagnostician and the representative samples. Many aquatic pathogens, including viruses, are undetectable through light microscopy and require electron microscopy, which is costly, 45

| Histopathology
Histology can be a valuable diagnostic tool if host and or pathogen tissue is available. It can be useful for routine monitoring or once infection has been established, but internal examination requires sacrifice of the target species. Sample processing involves the use of chemical preservatives such as 10% formalin (or even Bouin's fluid, potentially explosive when dry) for tissue fixation, embedding (in paraffin or resin), sectioning, affixing onto a slide and staining 38 using generic (such as Haematoxylin and Eosin) or more specific (e.g., Periodic Acid-Schiff ) stains. 50,51 Slides are then examined for tissue abnormalities or direct pathogen identification ( Figure 3).
Histology is a valuable diagnostic method for many diseases, such as furunculosis and syncytial hepatitis of tilapia, and the cryptic sal-

| Remote sensing
Fish suffering infection will often remain at the surface, in a moribund state and can be picked up by farmers, workers or environmental officers patrolling the water body, but surveying of wild stocks is challenging. This is time-consuming and limited to accessible sites. Drones can be implemented to refine this process, by applying an appropriate resolution to the camera, being able to survey the entire water body from the air, and potentially providing images for immediate diagnosis. 65 Advances  75 Although no current data is available on the efficacy of "Stingray", field tests and feedback from industry are positive, with drone deployment throughout Norwegian and Scottish salmon farms. 75 Technologies such as the "Stingray" combat infections in real time, allowing detection as soon as a louse infects a host, and represents a middle ground between early detection and detection after infection has been established. Remote sensing for pathogen detection and diagnosis is still in its infancy but it presents significant potential for remote detection and quantification of pathogens in an elusive and difficult environment.

| Artificial-intelligence and diagnostic software
Gaining sufficient experience to accurately assess and diagnose fish diseases takes years, hence interest in Artificial Intelligence (AI) to automate diagnosis through digital image processing. 76  the program was quite complex and inaccessible to many, and some farms lacked the necessary resources (e.g., microscopes, water quality equipment) to gather the required information. 82 Clearly, we are in the early stages of remote diagnosis but automating the process through the application of AI and machine learning approaches has the potential to establish a robust high-throughput process with the potential for quantification. They do, however, rely heavily on reference databases and further technology development.
Misdiagnosis still may occur due to the generic nature of clinical symptoms of many fish diseases and difficulty controlling for secondary infection.

| Microbiology
Fish microbial diseases are highly prevalent, as both primary and secondary infections, driven by stress (water quality, poor nutrition and temperature) or other infections. 83 Diagnosis has historically involved isolation and culturing of the causative agent. Direct placement or swabbing of diseased tissue or mucus onto agar is a common method for aquatic bacterial diagnosis, and for some aquatic fungal-like pathogens, followed by analysis of biochemical and morphological traits. 84 Such methods are selective and susceptible to contamination, requiring serial subculturing to obtain a pure strain of the causative agent.
The causative agent of bacterial kidney disease (Renibacterium salmoninarum) is particularly fastidious and grows slowly on regular agar, requiring a specialised agar for rapid growth with a 'nurse' microbe. 85 It also takes time to isolate colonies and observe definitive growth, with reports from 2 weeks 86 up to 19 weeks for subclinical level infections. 87 In contrast, the oomycete pathogen S. parasitica is regularly cultured on potato dextrose agar (PDA) by obtaining small tufts of mycelia from infected fish and embedding them within the agar, producing growth within 2-4 days. 88 Culture dependent methods are limited to pathogens with known nutrient requirements, subject to contamination even with antibiotics in the media, and, for long-term culturing, can be labour intensive. Culturing as a means of diagnosis is unreliable when trying to verify causal agents of polymicrobial infections. 89 In addition, genetic alteration of microbes may occur over time resulting in strains unrepresentative of natural communities.
Culture-independent methods have been instrumental in not only identifying pathogenic microbes but revealing the key role of microbiomes (all microbes within an organism) for fitness, immunity and life span of fish. 90

| Biochemistry
Biochemical methods for diagnostics encompass a variety of techniques all of which utilise some form of biochemical signal to conduct the diagnosis. These techniques vary from those which detect chemical signals (volatile organic compounds, or VOCs) released during infection (e.g., Pawluk et al. 100 who identified chemical cues from infected and uninfected fish), to biosensors that use biochemical reactions to detect (optical, volatile, electrochemical or mass-sensitive) chemical compounds. When considering their application to aquatic diagnostics, the information gained from these health parameters is currently too general for diagnostics, especially in a preventative context, and the benefits would not outweigh the costs.

| Serology
While commonly used in terrestrial veterinary practices, serology is used less in aquatic diagnostics due to insufficient development of methodologies. 101

| MOLECULAR TECHNIQUES
The rapid development of our ability to amplify and sequence genetic material has revolutionised every aspect of biological sciences, from behavioural and evolutionary fields to medical and veterinary sciences.
Molecular diagnosis ranges from standard PCR to next-generation sequencing and environmental DNA techniques ( Figure 4 and Table 2). The cycling procedures for qPCR are the same as those for standard PCR, but the products are typically shorter (<200 bp). After each cycle, the intensity of fluorescence is measured, which indicates the quantity of DNA amplicons in the sample at the given time. 130 qPCR can potentially be utilised to diagnose any pathogen of interest, dependent on the assay design with the ability to detect specific genes and alleles. qPCR is widely used as it is high throughput, highly sensitive, reproducible, and rapid 131 with reduced potential for cross-contamination. 130 Wide success has been achieved using qPCR for aquatic pathogen detection, including Anisakis, 132 Ichtyobodo, 133 viruses (viral haemorrhagic septicaemia) 134  allowing for rapid visualisation. 143 Colorimetric dyes, such as hydroxynaphthol blue and SYBR Green I, have high sensitivity for detecting pathogens, and can be more rapid than LAMP-LFD. 144 This combination of methods facilitated amplification of Taura syndrome virus in shrimp along with removing the need to use a DNA staining agent. 145 Detection of red seabream iridovirus (RSIV) was 10 times more sensitive by LAMP than standard PCR. 146 There is the potential for contamination of target DNA in the final stages due to the high amplification, sensitivity is highly dependent on the designed primers, and the limit of detection may differ for LAMP compared with PCR. 147 By removing the need for expensive (and typically nonportable) thermocyclers and thermally sensitive reagents, LAMP-based detection methods hold great promise for rapid aquatic pathogen diagnosis in the field and low-income regions.
LAMP is one of a growing number of isothermal amplification methodologies, each with their own benefits and detriments. 148 Recombinase polymerase amplification (RPA) substitutes the heat denaturation step of traditional PCR with two proteins (Escherichia coli RecA recombinase and single-strand DNA binding protein) and is carried out over a consistent temperature (often 37 C). This amplification is even more rapid than LAMP, occurring within 5 to 20 min. For aquatic infections, RPA has successfully detected Flavobacterium columnare, 149 Vibrio parahaemolyticus 150 and Tetracapsuloides bryosalmonae 151 to name a few significant aquatic pathogens. RPA is cost-effective, highly specific and sensitive and is a rapid methodology for diagnosis, especially when combined with LFD. 152

| eDNA
Environmental DNA (eDNA) methods have the potential to greatly improve our ability to detect and monitor pathogens in aquatic environments, be that as whole cells or free-floating DNA. eDNA can follow a targeted or passive method; targeted following standard PCR, qPCR or LAMP methodologies to determine presence/absence or abundance of a target species, whilst the passive approach uses primers sharing conserved binding sites to sequence communities of organisms. 153  eDNA is most effective in shallow waters where the benefits of eDNA outweigh regular trapping methods. 159 Most experimental studies utilise water samples when targeting DNA, but sediment is a viable alternative. 160 Asian carp (Hypophthalmichthys spp.) DNA was more concentrated (8-1800 times) in sediment compared with water, 161 but sedimentary eDNA is more likely to present pastspecies occupancy due to resuspension and transport. 162 The relative benefit of sediments compared with water for eDNA sampling is debatable and will depend on the target and the habitat. Drones may be deployed to collect water samples once the desired volume or sampling period has been achieved, or drones could collect smaller water samples ad hoc. 163,164 Methods such as these can be adjusted depending on the target, with buoys collecting water column samples or coring for benthic demersal layer sampling. False positives may arise due to the introduction or transportation of DNA into the water body, whilst certain species release DNA at a sub-detection threshold, leading to false negatives. 162 Water quality also impacts eDNA success, with acidity of water increasing degradation of environmental DNA. 165 166 Targeting eRNA can direct users towards the infective stage of a pathogen. Utilising eRNA poses additional challenges as RNA is less stable than DNA, degrading rapidly, and current costs are high. 167 The greatest benefit of RNA is targeting specific genes only expressed at certain life stages, providing high specificity, but the origins of environmental RNA are poorly understood. 167 The choice of targeting RNA or DNA is highly dependent on the target pathogen. To date, eDNA has been successfully applied to a range of pathogens from iridovirus in red sea bream, 168 ranavirus' in amphibians 169 to chytrid fungus in bullfrogs. 170 The aquatic host range for eDNA applicability ranges from fish and amphibians 171 to crustaceans. 156 eDNA has great potential to predict disease outbreaks.
One study assessed Batrachochytrium dendrobatidis presence before amphibian die-off events, where detection was successful before the mass mortality events. 170  Applying these tests to aquaculture and fisheries would never match this scale but would require significant monetary input. 45 But as with all novel technologies, costs rapidly decrease with time. Also, quality of data and portability will improve with the potential to revolutionise diagnostics of emerging diseases and cryptic pathogens.

| RECOMMENDATIONS AND CONCLUSIONS
The lack of transference of terrestrial techniques to the aquatic environments is due to issues of translation, changing something suited for terrestrial applications to the aquatic environment is not easily done, and requires significant interest and/or funding. The recent thrust in diagnostic development will result in progress not only for human medicine, but diagnostics across disciplines Advances in early pathogen diagnosis have typically been driven by infections of terrestrial hosts, highlighted by the current COVID-19 crisis. One benefit of this pandemic has been the rapid increase in efficient and rapid diagnostic techniques, such as lateral flow immunochromatographic assays providing results within 90 min or adapted LAMP technology. Such advances will hopefully boost the entire diagnostic field, including aquatic pathogens but as previously stated, will require a significant driver to bring in financial support. Lateral flow tests have always had potential for disease diagnosis but were relegated primarily to pregnancy tests due to the lack of sufficient drivers to develop the technology for other users. 181 The COVID-19 crisis demanded utilisation of every tool available, and thus the potential of lateral flow tests was harnessed for rapid diagnostics of the virus and informs how we can turn the retrospective into a reactive approach. 182 The diagnostic potential of many terrestrial diagnostic methods will not be translated for aquaculture without sufficient ecological or monetary drivers.
Indeed, even human neglected diseases are facing the same hurdles. 183 Nevertheless, here we evaluated a variety of diagnostic methods in light of the three pillars for a gold standard diagnostic technique: high sensitivity, low cost, and speed. Going forward, emphasis should be put on two main techniques to advance aquatic diagnostics: AI for visual diagnosis and eDNA for molecular diagnostics. AI has the potential to drastically reduce the time required to survey fish for disease whilst simultaneously allowing for higher throughput but requires significant input in "teaching" the AI to detect specific diseases. eDNA enables detection and quantification both on-site and in the laboratory, making it one of the most versatile diagnostic techniques once sampling methods have been optimised. As our knowledge of these pathogens increases so do our technological advances, where preventing pathogen outbreaks from occurring is the end-goal and these techniques aid this. Human medicine receives more monetary support for research on novel diagnostic methods, but there is always potential for these methods to be transferred to the aquatic environment should the industry or researchers take the time to adapt them.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no new data were created or analyzed in this study.