Skip to main content
European Commission logo print header

The development of tools for tracing and evaluating the genetic impact of fish from aquaculture: “AquaTrace”

Final Report Summary - AQUATRACE (The development of tools for tracing and evaluating the genetic impact of fish from aquaculture: “AquaTrace”)

Executive Summary:
The aim of AquaTrace was to contribute to a prosperous and sustainable future for European aquaculture. This ambitious goal has been pursued through state-of-the art genetic and genomic approaches leading to new valuable knowledge for the target species European sea bass, turbot, gilthead sea bream, Atlantic salmon and brown trout. AquaTrace has collected basic information on the biology, conservation, aquaculture and genetics of the target species, summarized in species leaflets, and has evaluated the general use of breeding programs and seed market in European aquaculture. New genomic resources have been generated in the marine species European sea bass, turbot and gilthead sea bream, which has provided unprecedented high resolution in insights on population structuring in wild and farmed populations. This has allowed the development and evaluation of innovative molecular genetic tools, which will vastly improve the essential ability for tracing marine farmed fish in the wild and for documentation of their potential effects on wild conspecifics. These novel analytical tools are at an advanced stage also allowing instant uptake for stakeholders from the aquaculture industry for aquaculture breeding and broodstock management. Experimental studies using the model organisms Atlantic salmon and brown trout revealed that even when a low proportion of an individual’s genetic variation originate from farm fish, the examined traits (hatch-timing, growth and maturation) deviated from those expressed in genetically pure, wild fish. In addition, we were able to locate some of the genomic regions responsible for these trait differences. In general farmed fish outgrew wild fish under all conditions, with hybrids having intermediate growth rates. Based on the collective results from the marine and model species we have attempted to assess and contain potential risks associated with aquaculture, as escapes or releases of domesticated aquaculture fish can have adverse effects on native fish gene pools. We have produced a risk assessment white paper summarizing the main insights from the project and their implications for management with a series of concrete recommendations and decision aids. Importantly, when properly integrated, these tools, insights and recommendations can be used by the industry, managers and policy makers to assist in assuring sustainability, productivity and welfare across the sector.


Project Context and Objectives:
Project context
Ecosystem-based management (EBM) has, since 2003, been an integral part of the European Common Fisheries Policy, which aims at the integration of a series of new objectives with the established goals of maximizing (i) the sustainable yield of a fishery; (ii) economic returns; and (iii) social utility in terms of employment and income. An important corollary of the approach is the fundamental requirement to conserve the integrity of exploited fish stocks and their natural environment. Any fishing or aquaculture activity that threatens to disrupt such integrity needs to be identified, and measures for mitigation implemented. In addition to overexploitation of capture fisheries, threats to the integrity of wild stocks derive from aquaculture practices, including the potential genetic impact of escapees. The basic principles of EBM have further become entrenched within the Marine Strategy Framework Directive (MSFD) descriptors for monitoring of progress towards Good Environment Status (GES). It is within the principles and policy context of the MSFD, embedded within the Common Fisheries Policy, that the current project AquaTrace, was developed. While global capture fisheries production has levelled off within the past decade, worldwide aquaculture has become the fastest growing animal food producing sector: aquaculture already provides c. 50 percent of world fish supply for human consumption, with a significant potential for further growth. Hitherto, world aquaculture is heavily dominated by the Asia–Pacific region, which accounts for 89 percent of production in terms of quantity and 79 percent in terms of value; in contrast, the average annual growth in aquaculture production in Europe since 2000 has also slowed markedly to 1.7 percent. Prior leaders in aquaculture development such as France and Spain are characterised by falling production in the past decade. In 2009 the European Commission proposed a new impetus to the sustainable development of aquaculture through improved competitiveness, environmentally-friendly production, high standards of animal health and welfare, and improved governance. Thus, aquaculture represents a key component of the solution to meet the escalating demand for fish. As such, there is a priority need to develop an appropriate legislative framework within the European Union aquaculture sector underpinned by cutting-edge research and technology. Outputs of any such activity need to implement breeding programmes and farming technologies which are economically viable, environmentally friendly and perceived as socially acceptable.

Foremost among the concerns in relation to ‘environmentally friendly’ is the issue of aquaculture fish escaping from their production enclosures (‘escapees’), which pose a threat to the integrity and levels of biodiversity both through direct competition for resources and through genetic ‘pollution’ of local populations of conspecifics. These escapees are a feature of aquaculture occurring chronically and acutely. In consequence of the rising concerns about aquaculture impact on the environment, there is an urgent need to identify methods that allow us to assess and monitor any genetic effects of aquaculture escapees on wild populations. Nevertheless, discriminating between marine fish populations is often challenging, primarily owing to standard genetic markers exhibiting relatively low levels of population differentiation. Accordingly, there are yet few coordinated databases that allow collation of information on population boundaries and dynamics of populations beyond local scales, although exceptions exist. Discrimination between wild and farmed strains of marine fishes is further complicated by the fact that each farmed strain has its own history of selection and domestication, sometimes including recurrent backcrosses to wild-origin brood-stock. These breeding processes have typically not been documented, and may mask the frequency and direction of interactions. Hence, new genetically-based tools are required to provide unambiguous proof of current and past interactions, including introgression and the incidence of hybrids. Here, the recent technological advances in development and application of genome-wide markers in fishes offer a highly promising approach that can be specifically aimed at tracing the genetic origin and potential hybrid status of individual fish in a wild/farm context.
To perform a rigid assessment of the potential impact of aquaculture fish on wild populations information is required at two levels: 1) Rigid assessment of rates of introgression, including across spatial and temporal scales (‘How many populations are affected by introgression, and to which extent have their genetic integrity been altered?’), and 2) assessment of effects of introgression on the fitness of local wild populations (‘How severely does introgression affect the fitness of wild populations?’). The first level of information can be addressed using an informed and extensive sampling design combined with development and application of state-of-the-art genomic resources, and is fully attainable in marine species such as European sea bass, gilthead sea bream and turbot. The second level requires deeper insight into genomic structure and functionality of the target species. Here, genomic information should be combined with experimental assessment of heritable components of the fitness and life-history traits suspected to be under threat due to admixture between genomes that have been subject to different types of selection (natural vs. breeding- and/or domestication selection). This has until now only been obtainable for relatively few ‘model species’, for which there is adequate genomic resources and knowledge on functional genomics, and which can be held and manipulated under experimental conditions throughout their life-cycles. Such species include Atlantic salmon, Salmo salar, and brown trout, S. trutta.
The rationale behind AquaTrace was to develop reliable and cost-effective molecular tools for the identification of the genetic origin of both wild and farmed fish (assignment and genetic traceability), as well as for the detection of interbreeding and assessment of genetic introgression between farmed and wild stocks. This work was carried out on three marine fishes of economic significance, the European sea bass (Dicentrarchus labrax), gilthead sea bream (Sparus aurata) and turbot (Scophthalmus maximus). To address quantitative effects of farm introgression, the rationale was to examine links between key fitness and life-history traits and specific functional genetic variation between wild and farmed fish, using Atlantic salmon and brown trout as model species. AquaTrace has been charged with scientific objectives relating to addressing and assessing the potential genetic impact of aquaculture escapees introducing genes that have been undergoing adaptation to farmed conditions through breeding and domestications selection to wild populations. Nonetheless, methods and aims also have implications for general knowledge on local adaptation in wild populations, and thus also apply in a restocking context, e.g. when locally depleted wild populations are stocked with non-native strains that potentially are mal-adapted to local conditions.
However, AquaTrace also acknowledged that for successful implementation of tools outside a purely conservation framework, it is imperative to recognise the interdependence of scientific, economic and social objectives related to sustainable aquaculture and the role of public perception. Advances will be modest if recommendations and investments in technology are perceived as restrictive, punitive and at odds with shared responsibilities and contributions from scientists, policy makers and the industry. It is further essential that developed tools are validated to internationally recognised forensic standards to allow uptake by end-users. The application of tools for monitoring and mitigation must thus be seen as being supportive to the industry, representing one of the many approaches that should be used to secure growth, economic prosperity and social acceptance. Similarly, traceability of products has become a specific request of consumers, sustained by national and European policies. Here, genetic tools offer cost-effective strategies for supporting quality plans, enforceable by law where required, aimed at tracing and monitoring the origin of aquaculture products. By integrating cost-effective, reliable tools with new knowledge on the potential genetic impact of farmed fish on wild stocks, the aims were to strengthen the framework for environmental protection, sustainability, and fair governance within the European aquaculture sector to address several inter-related objectives.


Main objectives

AquaTrace has recognised the central role that population diversity and genetic integrity plays in sustainable utilisation of marine ecosystems in general, and the need to minimize the negative impacts of aquaculture escapees in particular. AquaTrace aimed to develop robust traceability systems that incorporate major spatial and temporal differentiation in three European species, all of which are of importance to aquaculture industry, capture and/or recreational fisheries and ecosystem function. The consortium also acknowledged that any method or technology emanating from AquaTrace has been adapted to end-user needs, and thus have been evaluated against a combination of applicability and economic criteria. Thus, the main objectives were:
1. Integrate information from recent projects related to assessing genetic impacts on wild populations of turbot, Gilthead sea bream, European sea bass, Atlantic salmon and brown trout and to provide an overview of the selective breeding programmes implemented in the target species across European aquaculture in order to design a generic template for a synthesis on potential impact of aquaculture.
2. Provide a comprehensive collection of tissue samples for genetic analysis of wild and farmed populations of European sea bass, Gilthead sea bream and turbot. Samples included archived historical collections representing temporal replicates. The outputs comprised an integrated archive of tissue samples from external parties and the AquaTrace consortium.
3. Develop genomic resources, including Single Nucleotide Polymorphism (SNP) markers, in turbot, Gilthead sea bream, European sea bass that allow for both robust determination of the genetic structure of wild populations, as well as for tracing aquaculture strains. The outputs comprised DNA sequence data and population-informative candidate SNP markers.
4. Establish baseline information on genetic and demographic (life history) traits of natural populations and aquaculture stocks of turbot, Gilthead sea bream, European sea bass, and to perform population genetic analyses to identify genetic stocks and their connectivity, farm/wild hybrids and evidence of local adaptation. The outputs comprised population- and aquaculture strain level signatures associated with fish origins in representative spawning/broodstock groups, including assessment of temporal variation.
5. Determine links between genetic and phenotypic differences between wild and farmed stocks in model species, Atlantic salmon and brown trout. Common garden studies that determine variance in key fitness related traits were employed in combination with genome-wide association mapping and gene expression studies. Outputs comprised estimates of heritable differences in a suite of fitness and gene expression traits between wild and farm strains, together with identification of the specific genome regions associated with the functional adaptations that diverge between fish from wild and farmed conditions.
6. Develop genetic marker tool systems that are reliable, cost-effective and fully transferrable, that allow both robust assessment and monitoring of interactions between wild and aquaculture stocks and as well as traceability of individual farm strains and family groups. The outputs comprised a set of methods which have been optimized for a representative range of specific applications and scenarios based on statistical modeling and testing. Outputs further comprised protocols for fully forensically validated molecular methods with associated Standard Operating Procedures.
7. Develop an AquaTrace website, database and web-based information access platform that allow transfer of knowledge and technology to stakeholders, e.g. in relation to aquaculture breeding programmes, conservation and enforcement policies and associated socio-economic consequences. Outputs included an internal database of AquaTrace information and results, and a web-based data access tool which ensured sustained impact of AquaTrace along and beyond the project’s time-frame.
8. Provide a risk assessment on the impact of introgression by escaped farm fish and in a restocking context (i.e. stocking wild populations with hatchery strains). In addition to a comprehensive risk assessment, the outputs included a list of suitable indicators for assessing and monitoring introgression and its associated effects, potential mitigation strategies under current practices as well as general management recommendations, setting AquaTrace in the context of the Marine Strategy Framework Directive.
Overall, the generated outputs from AquaTrace were expected to improve knowledge on potential genetic impact of aquaculture on native populations of the five target species turbot, European sea bass, Gilthead sea bream, Atlantic salmon and brown trout via the development and application of species-specific genetic tools. AquaTrace activities were expected to enhance understanding of the dynamics, temporal stability, distribution and adaptive traits of natural populations of the target species. AquaTrace thereby has aimed to enhance the MSFD aim to promote sustainability through conservation of genetic resources, as well as in the protection of consumer interests. AquaTrace has provided a forensically validated framework for traceability of farm fish in the five target species, providing end-user tools that can be applied to assessment and monitoring of genetic impact of farm escapees. To enable a comprehensive assessment of potential impact of farm escapees, AquaTrace has combined development of novel tools and generation of novel empirical data with information from existing databases, genomic resources, tools and projects of relevance.

Project Results:
Aquaculture breeding practices

Within AquaTrace we conducted an online survey on the major aquaculture breeding companies operating in Europe. Six fish species were targeted: European sea bass, gilthead sea bream, turbot, rainbow trout, Atlantic salmon and common carp. For brown trout, mostly exploited for recreational fishing, no breeding programs were identified. A total of 31 respondents contributed to the survey, representing 36 active breeding programs and 75% of the European breeding organizations identified. The survey shows that the implementation of selective breeding programs is a continuous process for all target species. Except for carp and turbot, at least one new program has been initiated for each species during the last 3 years. In most cases (68% of the programs) precautions are taken by the breeding companies to monitor the increase of inbreeding at each new generation. This concerns all programs of salmon and turbot, compared to roughly half for the other species. Family-based breeding schemes were predominant, but individual based selection was more frequently applied in marine species. An increasing number of programs use either genomic or marker-assisted selection. Five out of the six programs running a genomic selection program have also implemented marker assisted selection (MAS) and use DNA fingerprinting to assign parentage. The number of programs using molecular tools has more than doubled since the previous survey (2008), mainly for establishing pedigrees. The most frequently selected trait is still growth performance, but the number of selected traits has been increasing over the years through the addition of traits such as disease resistance or product quality. Artificial fertilization is the preferred means of reproduction. This technique is systematically implemented on salmon, trout, carp and turbot programs, while for seabass and seabream mass spawning is often used as a fallback method. When asked to describe the reproduction pattern of their selected broodstock, the majority of the breeding organizations (71%) declare to use more than 200 broodfish per generation. In carp, salmon and trout the number of parents is higher than for other species, with three programs using more than 800 parents, and none less than 100. The production of sterile fish is seen mostly in salmonids, i.e. in rainbow trout (all programs) and in two salmon programs. Seed production estimates (2012) for sea bass, sea bream and salmon have the same order of magnitude, i.e. around half a billion juveniles or eggs. These values are flanked by a much lower number (20 million) of turbot juveniles and a higher figure of more than 2 billion rainbow trout eggs. The main seeds producing countries for turbot, seabream and seabass juveniles are Spain, Greece and Turkey, while Norway and Turkey are the main producers of salmon and rainbow trout eggs respectively. Looking at the number of seed exporting countries, we observe two salmon programs outpacing the others by trading their seeds in 40 countries. They are followed by three programs exporting seeds to 10 countries. At the other end are companies whose breeding programs have been implemented to serve their own production needs first or secondarily for customers located in a limited number of countries. The dispersion of seeds from selected broodstock in the market is characterized by two broad trends. For the turbot and salmon seed the markets are dominated by genetically improved fish (93–100%), and for the other species selected seed have a medium market share (31–52%). These values, which reflect the situation depicted by the surveyed respondents, are likely to underestimate the contribution from selected seeds in the market for the latter species.

Looking at the development of selective breeding across Europe, it is clear that there is a large heterogeneity among species and countries. Most salmon and turbot produced today originate from selected stocks, while over half of the cultivated fish of the other species are in an early domestication process or of wild origin. No single factor was identified that explains the different efforts among the countries in developing genetically improved material. However, the strategic long-term visions of development by larger companies may play a facilitating role for the development of selective breeding at national levels. We show that the build-up of new breeding programs is an ongoing process for all species. Once implemented, programs evolve with new traits being added over time to increase the productivity, but also to answer consumer demands. For most breeding companies, the transnational trade of seed is part of an overall effort to reach new customers and increase their market share. Some producers, like in those of salmon, have become major players in terms of both traded seed and number of export countries, while others are still only producing seed for their own needs or for a limited number of customers; this is particularly the case for the marine fish species where the added value of selected seed is not fully recognized yet. Two models of hatcheries currently exist, differentiated according to species. The first of these is the ‘salmon model,’ where the selected candidates are transferred to multiplier stations to provide sufficient commercial seed for smolt production and final grow-out. Here, the transfer of the genetic progress in production is delayed one by generation. The second model is the ‘marine fish model,’ where hatcheries are part of a horizontal organization. The breeding company is most often an integrated part of a commercial farm, and no intermediate multipliers are required. The main difference between these two models is specifically linked to the fecundity of the species, which varies by a factor of 100 between salmonids (2–3000 eggs/kg body weight) and marine fish species (up to 200–400,000 eggs/kg body weight). Regarding the escapees, we could only report incomplete data due to the limited information available, with important variations according to the species and the geographic area where they occur.


Integration of previous knowledge and resources

We have prepared a standard template of dissemination for the targeted species and based on this we provided overviews and synthesis of available knowledge related to potential genetic impacts on wild populations for Atlantic salmon, trout, turbot, gilthead sea bream and European sea bass. The species leaflets are available as interactive documents and printable PDF files at the AquaTrace web site (https://aquatrace.eu/species_leaflets). The leaflets fulfill an important aim of the project, which is to collate biological information such as life-history traits, genetic and genomic characteristics, ecological knowledge as well as conservation, fisheries and aquaculture specific issues. The comprehensive leaflets compile published information from previous international research projects, but also information from national projects, which would not have been available to a wider community without AquaTrace. The collation of comprehensive information from international and national scientific literature and reports represents a milestone of information for each species. Scanning the literature also made it evident how scarce the biological information is for the targeted marine species. Essential basic biological information such as population size (spawning biomass) age of individuals and growth rates are virtually missing. Thus, any assessment of wild population status and potential impact of farmed fish is severely hampered. In order to assure proper management, conservation of genetic resources and sustainable aquaculture, the generation of this basic information should be a top priority.


MARINE SPECIES

Collection of samples
One of the goals of AquaTrace was to collect and archive the most representative and informative collection of samples for genetic/genomic analysis of wild and farmed populations of the European sea bass, gilthead sea bream, and turbot. For natural populations, the objective was to ensure the most complete geographic coverage and to allow analysis of temporal stability (historical samples). For farmed populations, highly domesticated/selected broodstocks were targeted as well as samples from selected hatcheries based on location, volume of annual production, type of production, and broodstock origin. For all the three target species a large number of specimens were processed. All samples have been properly stored for future reference, while sampling data are available through the AquaTrace database. Extracted DNA was used in downstream genomic activities. In total, more than 4,000 samples were collected for the European sea bass: 3,061 wild samples (novel and archived) and 1,334 farmed samples (novel). In total, more than 3,500 samples were collected for gilthead sea bream: 1,740 wild samples (novel and archived); 1,536 farmed samples (novel) and 300 from controlled crosses (full sib families). In total, around 1,500 turbot samples were collected: 1,240 wild samples (novel and archived) and 382 farmed samples (novel). While sampling coverage was limited by the low availability of wild specimen in a few areas, due to decrease in abundance of target species (often confirmed by catch data) or by environmental conditions that limit the presence of these species in some areas (e.g. turbot in warmer Mediterranean waters and sea bream in the Atlantic), the overall targets are considered fully achieved.

Development of genomic resources
Genome-wide sequencing of sea bream was completed using a combination of double-haploid material and standard sea bream genomic material. The double-haploid sample was used in order to minimize heterozygosity within resulting sequence data to simplify sequence assembly. For turbot, AquaTrace benefitted from access to sequence data produced in a parallel project. Therefore work in AquaTrace focused on additional sequencing for turbot to compliment the sequence data previously generated, using existing samples. Whole genome paired-end (PE) sequencing was performed using Illumina GAIIx from several libraries to facilitate assembly: i) PE libraries with insert sizes of 200bp and 500bp; ii) mate-pair (MP) libraries of 3 and 5kb; and iii) a fosmid library of 140,000 clones. The sequence data generated was subjected to bioinformatic analysis in order to align all of the individual short sequence reads and assemble them into a smaller number of longer sequences. The sequences for the genome are generated in ca. 100 base fragments which are then compiled into ‘contigs’ and these contigs are subsequently assembled into ‘scaffolds’. The larger the contig and scaffold size, the more complete the assembly of the genome. The assembly for turbot was more comprehensive than that for sea bream due to the greater amount of pre-existing sequence data that was incorporated into the final assembly. The existing sea bass assembly was further developed with sequences generated in AquaTrace. While all three reference genomes were sufficiently assembled for subsequent phases of the project, variation in the degree of assembly provided the basis for a comparative study that has recently been submitted for publication.

ddRAD libraries were prepared for wild and aquaculture populations for each of the three species. The preparation of the ddRAD libraries involved DNA extraction of multiple samples (>1000 per species), followed by treatment of the DNA to create batches of 144 individually labeled samples for sequencing. Each sample has been enzymatically digested (twice) and then replicated under PCR in order to provide sequencing targets for around 10,000 regions of DNA in each sample. In total, 42 libraries were successfully produced. These libraries were comprised of 6155 samples (2355 sea bass; 2300 sea bream; 1500 turbot). This exceeded the target of 5750 samples described in the Technical Annex revision. In order to process the ddRAD sequence data it was necessary to develop a novel bioinformatics pipeline. Briefly, the pipeline is a series of analytical procedures used to handle very large volumes of genomic sequence data. The pipeline produced here aimed to combine all of the DNA sequences produced for all samples from the different libraries in the bioinformatics analysis. Following this first step, the sequences are aligned and SNP DNA markers are identified. These SNP markers are then genotyped in each sample, to produce the final species genotype files for each population in each species. The pipeline developed was based on the STACKS software, which was subsequently customized to tailor the analysis to the ddRAD data produced within AquaTrace. Customization involved evaluating multiple parameters within the sequence alignment and SNP calling stages to optimize the accuracy and the number of SNPs discovered while minimizing data loss. Additional bioinformatics scripts were written to enable data handling upstream and downstream of the STACKS analysis stage. Finally, the pipeline included a process for mapping the resulting sequence reads against the available reference genome sequence data. The complete pipeline was used as the basis for ddRAD sequence data analysis
The production and assembly of genome-wide sequence data for sea bream and turbot complimented the reference genome already available for sea bass to create a comprehensive genomic framework for the subsequent population genetic analysis. In addition to facilitating subsequent phases of marine genomic work within AquaTrace, these tasks have already delivered data that have formed the basis of a comparative technical paper. The ddRAD library preparation required the majority of partners to learn new laboratory techniques and to subsequently scale these up to perform analysis of large population-wide sample sets, coordinated across four laboratories in four different countries. The success of this work was vital to the progress of AquaTrace as a whole and should be considered a standout achievement, delivering molecular genetic outputs, young researcher training and international multi-partner collaboration.
Establishment of genetic and demographic baselines
The main aim was to isolate and genotype SNP DNA markers in each of the marine species (turbot, sea bass and sea bream) for the population genomic analysis of natural and aquaculture populations over both spatial and temporal scales. These data were used to create a genetic baseline for traceability on a pan-European scale. Additionally, a database of demographic information for each marine species was constructed, based on literature and public data, to allow the assessment of introgression/hybridization consequences on natural populations through empirical and simulation analyses (see below). After the production of ddRAD genomic libraries and a tailored bioinformatics pipeline, population samples were genotyped at more than 1000 SNP DNA markers, to investigate population structure and enable assignment of fish in both aquaculture and natural populations. The baseline data was analysed using cutting edge statistical tools in order to assess structuring of wild and aquaculture populations. In total 19 sea bass (1925 individuals), 17 sea bream (2165 individuals) and 10 turbot (1059 individuals) libraries were sequenced. Samples with a low number of reads (< 350,000) were filtered out and SNP discovery was carried out using both a de novo (without reference genome) or a reference-based approach, as implemented in the programs Stacks and dDOCENT. Stringent filters were applied to obtain a reliable dataset. Markers that were shared by less than 80% of individuals, had a Minor Allele Frequency (MAF) lower than 0.25%, Linkage Disequilibrium (LD) higher than 0.7 and deviated from Hardy-Weinberg equilibrium were discarded. Several different software were used to calculate standard genetic parameters: observed and expected heterozygosity and private alleles (GenAlEx); Fst (Arlequin); Effective Population Size (NeEstimator); pairwise relatedness (Coancestry). Clustering analysis was mainly based on the software Structure and on Discriminant Analysis of Principal Component (DAPC) as implemented in R’s package adegenet. Markers were screened to unveil signs of adaptation (i.e. markers potentially under selection, called “outliers”), and distinguish genetic differentiation due to adaptation to different environments and genetic differentiation due to demographic processes (described with neutral markers). At least two different approaches were used to detect outlier loci, based on locus specific Fst and heterozygosity (Lositan), unexpected high Fst (BayeScan) or correlation with environmental parameters (BayEnv). Different stringencies were used to define “outliers” (including only markers detected by at least two approaches) and “neutral” (excluding markers detected by at least one approach) datasets. SeaScape genetics analyses were performed using a redundancy analysis (RDA) implemented in the VEGAN R-package.

Sea Bass wild populations: Based on a set of 2709 molecular markers developed in the AquaTrace project, Mediterranean and Atlantic wild sea bass are shown to be highly divergent (Fst = 0.11); they represent two distinct types (named clades). Both clades separate at the Almeria-Oran oceanographic front, a well-known biogeographical barrier. Atlantic sea bass differentiate weakly but distinctly into three groups (Fst = 0.015) with most fish belonging to a single clade, except for fish living off Norway (northern North Sea) and off the Strait of Gibraltar. Mediterranean populations are clearly split into a Western and Eastern group (Fst = 0.026). The Western group is homogeneous and shows evidence for introgression from the Atlantic clade. Within the Eastern Mediterranean there is an Adriatic-Ionian group, a group off Tunisia, an Aegean and Black Sea group and a Levantine group. Re-assignment success to most of the identified groups is high (80-100%). Seascape genetics analyses pointed to the importance of both distance and temperature in separating the three groups. Our results are consistent with other published evidence of population structure in this species and point to the importance of historical factors, contemporary dispersal and local environmental factors acting upon natural populations.

Sea Bass farmed populations: Showed a range of genetic diversities; the mean number of alleles varies from 1.56 to 1.87 while the observed heterozygosity ranges from 0.16 to 0.18. These values reflect the diverse origins of the source material, the breeding practice (size of broodstock, natural or artificial fertilisation and culling) and the selection practice and pressure (from high to low). Genetic differentiation (as measured with Fst) shows small to large differences between farms, depending on their composition and broodstock origin (see later). There is strong evidence of genetic material from the Atlantic Ocean being present in Mediterranean farms and of Mediterranean material in one Atlantic farm. Some farmed stocks show evidence of a homogenous and high selection intensity. While this constitutes an asset for tracing fish back to the source farm, it is with the present genetic dataset not possible to identify a unique genetic profile for each individual farmed sea bass, i.e. it is currently not feasible to assign individual all fish back to specific farms with sufficient confidence.
Sea Bream wild populations: After filtering, a total of 2165 individuals were genotyped at 1240 loci. For sea bream wild populations, clustering analysis suggested a relatively strong subdivision between Atlantic and Mediterranean basins (Fst values 2-3%) and a less strong, though significant subdivision within the Mediterranean in three “sub-basins” (West Mediterranean, Ionian and Aegean) (Fst values from 0 to 1.8%). From a temporal point of view, no significant genetic differentiation was found between samples from the same areas (Ionian and Aegean seas) collected in different years (6 to 9 years of difference). Within group genetic variability was high in all wild population samples (Ne estimates ranging from 500 to >10,000). Analyses based on the “outlier” dataset composed by 15 loci showed a stronger differentiation within the Mediterranean basin. Allele frequencies at six loci showed significant correlation with environmental variables (temperature and salinity). Analyses at neutral loci showed a significant level of differentiation between Atlantic and Mediterranean, while differentiation within Mediterranean regions is weaker and non-significant in most pairwise comparisons for the neutral data.
Sea Bream farmed populations: Farm broodstocks are generally less genetically diverse than wild populations, with estimates of effective population sizes (Ne values) ranging from 30 to 200 (depending on broodstock size). In addition, broodstocks are more differentiated among each other (Fst ranging between1.2% and 5.7%), which suggests strong genetic drift due to many generations of selection or founder effects. When compared to the wild counterpart, most of the farmed populations are markedly different from wild populations, but some broodstocks with a more recent history of hatchery selection show similarities with some wild populations. In some cases, Atlantic vs Mediterranean origin of broodstocks can be identified, which is a much more difficult task for highly selected broodstocks that are highly divergent from all the wild populations.
Turbot wild populations: Genomic libraries were constructed for all 1059 turbot individuals from wild and farmed populations and 7,403 SNPs were initially detected. Duplicated SNPs, loci with no turbot genome matching and presenting low genotype coverage (< 80%), MAF (minimum allele frequency) lower than 0.002 across all samples and significant and systematic FIS values in most locations (p<0.05 using Bonferroni) were removed. Hence, a final set of markers with 755 SNPs was used for subsequent studies. Genetic studies of wild populations were carried out for 672 individuals from the Northeast Atlantic Ocean, Baltic, Mediterranean and Black Seas. Global genetic diversity and effective population size (Ne) of the Mediterranean, Baltic and Black Seas were lower than the Atlantic locations, likely due to historical isolation. Using stringent parameters, 17 distinct candidate outliers for divergent selection were suggested in different comparisons: using all wild samples (12), within the Atlantic region (3), between Atlantic and Baltic Sea (5), between Atlantic and Black Sea (8), between Black and Baltic Seas (2) and for the south and north of the Baltic Sea (4). Balancing selection was mainly detected in the analysis involving the Baltic and Black Seas (7), which share common environmental conditions (i.e. low salinity), but was also observed between the two Baltic sampling sites (2). Four main different population units were suggested by the Bayesian clustering analyses performed with Structure, corresponding to the Atlantic region, Baltic Sea, Mediterranean Sea (Adriatic) and Black Sea. DAPC analyses corroborated these results and also suggested a subtle structure within the Atlantic region, where some locations from the edges of the geographical distribution (Norway to the north and Spain and Biscay Bay to the south) appeared slightly differentiated. Using the putative neutral dataset (513 SNPs), genetic differentiation decreased among all sampling sites, but similar population structure was detected, probably due to geographical isolation and demographic history. On the other hand, genetic structure using six Atlantic and Baltic basin-specific divergent outliers confirmed adaptive divergence between these two areas and also revealed Norway as slightly differentiated within the Atlantic region. Seascape analyses (redundancy analysis or RDA) indicated that SSS (Sea Surface Salinity), SBS (Sea Bottom Salinity) and SST (Sea Surface Temperature) could drive the detected genetic structure in turbot, with the previously defined four groups. Hence, Baltic Sea locations were separated from the rest of the locations by salinity and temperature, the Black Sea was separated by salinity, the Mediterranean (Adriatic) was separated by temperature, and inside the Atlantic group a southern-northern thermal cline from warmer to colder regions was detected.
Turbot farmed populations: When wild and farmed samples were compared, global allelic variation and effective population size were lower for individuals from hatcheries, likely due to genetic drift effects. However, similar levels of heterozygosity were detected for the hatcheries, evidencing the continuous genetic monitoring usually applied in turbot broodstocks. Moreover, most specific analyses on farmed samples revealed two distinct origins for these samples (ORI1 and ORI2, Fst ≈ 0.051) that were selected for subsequent investigations. Considering the presence of farms is basically only in the Atlantic European coast, samples from Atlantic were used as representative wild samples to investigate the possible impact of aquaculture on wild populations (Fst ≈ 0.066 when compared with farmed ORI1; and Fst ≈ 0.030 for ORI2). The use of the three “reference” clusters (Wild Atlantic, ORI1 and ORI2, respectively) in STRUCTURE software allowed detection of 116 individuals presenting some level of farmed ancestry along all Atlantic and Baltic areas. Two fishes from the English Channel and Bay of Biscay presented high evidence for a farmed origin, while other samples presented various levels of farmed-wild genetic admixture. Wild and farmed samples were also used to produce simulated individuals and to infer the power of certain SNP panels to correctly assign individuals to wild stock and hatcheries. Preliminary assignment tests indicated that around 150-200 SNPs were able to classify properly up to 90% of wild or farmed simulated individuals (parental ones) and 80% of hybrids and backcrosses. Finally, 24 putative different divergent outliers were detected between wild Atlantic and farmed stocks (17 for ORI1 and eight for ORI2) indicating divergent selection related to domestication processes. The majority of outliers was associated with growth and resistance to diseases and identified in the turbot genetic map.
In conclusion all three marine species displayed significant wild population structuring across their distributions. The largest differences were found between Mediterranean and Atlantic populations, but also within basin genetic differences were evident, in particular for the Mediterranean. Both demographic processes and environmental selection (distance, temperature and salinity) seem to have driven the structuring of populations. In general farmed populations show less intra-population diversity, but larger diversity among populations and also higher variation in levels of differentiation. There was evidence of transfer og genetic material among farms in different basins. These findings represent a very large addition to the knowledge of the genetic population structure for these species, and define previously unclear genetic patterns of differentiation of wild populations. In addition, this is the first extensive analysis of European broodstocks. The information provided here is an important resource for the conservation and management of these species and for monitoring genetic effects of farmed fish on wild conspecifics.


Development of molecular tools for assessing and monitoring

The identification of large numbers of single nucleotide polymorphism markers is now relatively straightforward for non-model species. The Aquatrace project used RAD sequencing to identify SNP markers in sea bass, sea bream and turbot using samples from across the range of the three species and also a sizeable proportion of populations from aquaculture facilities. This work focused on identifying smaller panels of informative SNPs for each of the three species that would allow traceability of individual fish to source population – be that farm or wild and identifying introgressed individuals in the wild. This was achieved through the development of SNP panels that allowed assignment at various levels, and using simulations to test the power of these panels to assign individuals to populations or hybrid classes, both now and into the future, under a range of scenarios. The specific aims were 1) to design and test traceability assays for the three species, sea bass, sea bream and turbot, 2) test the power of those assays including in the future as a result of genetic drift in farmed populations, 3) investigate the power available to assign individuals back to different hybrid classes, 4) identify introgression in the wild and finally 5) to review SNP genotyping costs for small assays.
SNP panels were designed and tested for different sea bass traceability scenarios. These scenarios demonstrated that a high level of assignment power could be achieved with a small number of markers across a broad scale scenario comparing the wild Atlantic population with that of the Mediterranean. However, the smaller scale scenarios, such as comparing east to west populations in the Mediterranean require larger panels of markers. When using 40-50 SNP loci, rates of mis-assignment were relatively high for populations that were in close geographic proximity and individuals could not be confidently excluded from either source population. It therefore suggests that testable scenarios would need to be clearly defined before the design and application of minimal SNP panels. SNP panels were designed and tested for different sea bream traceability scenarios. These scenarios demonstrated a lower level of assignment power than in sea bass, in particular when all farms and wild populations were included in the panel design. An improved level of correct assignment can be achieved by refining the scenario in question and using an increased number of markers. The data suggests that testable scenarios would need to be clearly defined and limitations acknowledged before the design and application of minimal SNP panels. SNP panels were designed and tested for different turbot traceability scenarios. These scenarios tested demonstrate that a reasonable level of assignment power could be achieved even with a small number of markers across a scenario that compares farmed to wild caught turbot. Farm specific panels appear to perform better at correctly assigning farmed individuals back to their origin and also reduce the number of wild individuals that are mis-assigned. Without knowing the location of stocking history of any one farm, it seems likely that the farm ‘3’ is genetically more dissimilar to any of the wild Atlantic populations that were included in the assignment. This is demonstrated by larger differences in allele frequencies and a lower rate of mis-assignment using a 30-50 SNP panel. Power analysis was conducted across the three species. The power of each panel to assign individuals was investigated using different numbers of SNP markers. For the three species power of assignment increased with the number of markers.
Simulations
Assignment success was simulated using a variety of effective population sizes and over different numbers of generations in both an ‘in silico’ simulation experiment and also simulations based on real allele frequencies and SNP panels. For the in-silico study, drift had a slightly beneficial effect on assignment success across generations. This is expected as in small populations, allele frequencies will drift faster than in large populations and assignment should get stronger into the future. This is particularly the case where allele frequencies are already biased one way or the other – the probability of fixation is directly proportional to the initial allele frequency. When initial allele frequencies started out as 0.5 in both the wild and farmed populations, assignment success clearly increased across generations as a result of drift. It is important to note that allele frequencies of the simulated wild population will have changed very little in comparison to the farmed population which has a much smaller Ne. Secondly, assignment of wild individuals to wild was not expected to be different among the three Ne scenarios tested as the Ne is not modified in the wild population, only the farmed population. The assignment of wild populations investigates whether assignment of wild individual changes under a drift scenario in farmed populations. For the study based on real allele frequencies in sea bass populations, there was no marked increase or decrease in assignment success across the generations in any of the different Ne’s used. Thus, a 7 SNP marker set designed to assign individual sea bass to Atlantic or Mediterranean origin will have utility to assign individuals for at least 20 future generations and probably much longer.

Assignment into introgression classes
Two methods were used to investigate the power of assignment to different introgression classes for a sea bass Atlantic Mediterranean example and sea bream farmed wild example. These methods were assignment to simulated introgression classes and subsequent individual assignment. For sea bass, both hybrid identification methods showed high assignment success for the pure classes, but different success in assigning to introgressed classes. The assignment implement in Genodive was successful in assigning the backcrossed individuals back to the correct class, but was not at all successful in correctly assigning F1 and F2 simulated individuals. However, the majority of individuals in the F1 and F2 classes assigned back to one of the backcrossed categories. Overall, the power to detect introgressed individuals (as opposed to the exact class of introgression) was high. The power analysis conducted in newHybrids was more successful in assigning individuals to hybrid class. Using the full 1044 loci, assignment to each class was over 80% for all of the classes, and over 90% for the pure and backcrossed individuals. The lower success for F1 and F2 was again associated with individual assigning back to backcrossed classes. Consequently, the overall success of identifying introgressed individuals per se was high. For turbot assignment to pure classes was high (>90%) with assignment to hybrid classes lower. Assignment to backcrossed classes was higher than those for F1. This mirrors the findings in sea bass and sea bream, where the assignment of F1 individuals was less successful than for the backcrossed classes. For the sea bream example, the genetic difference between the farmed and wild sample was much lower. Consequently, assignment success was lower than for sea bass. The assignment method was not successful in assigning individuals to the different classes, but was successful in assigning individual to a single ‘introgressed’ category. Assignment success was much lower than that found in the sea bass. Assignment to F1 and F2 classes was very low using a 0.9 probability cut-off. The sea bream example paired a divergent farm population with a typical Western Mediterranean bream population. Most farm populations sampled were far more similar to the wild populations than for the Ne used in the drift simulations. Thus, while there is some power to identify introgressed individuals in this case, in most cases there will be no power to detect introgression in general with the current set of SNP loci. In contrast, estimates of relatedness were used highly successfully to identify escaped individuals. Escapees are expected to have high pairwise relatedness in comparison with wild individuals. Using this approach a single Greek population sea bass (PK) was identified to consisted almost entirely of escaped related individuals.
Overarching the simulations showed that it is generally possible to assign wild samples of the marine species back to their geographical origin. In many cases is possible to develop specific panels of relatively few markers with high assignment power. However, there is significant variability among species and regions. In a large number of scenarios it is also possible to determine wild or farmed ancestry of individuals, but this is complicated by the diverse origin of farmed populations. Accordingly, more specific simulations on a case by case basis are necessary for testing the multitude of possible scenarios of escape events. A powerful way of identifying escaped farmed fish is through relatedness analysis. With the markers at hand it is possible to evaluate overarching levels of introgression for a high number of possible scenarios for all three species (with variability). But it is difficult to assign individuals to specific hybrid classes (F1 or F2 backcrosses).


Forensic validation of tools and SOP development for laboratory and analytical processes

The overarching aim was to take small in silico SNP panels developed above and evaluate their suitability and readiness for application to fish traceability. This evaluation includes validation studies to generate the data necessary to assess the robustness, reliability and reproducibility of tools developed within AquaTrace for monitoring or enforcement purposes. The objectives were as follows: 1) To undertake forensic developmental validation of molecular tools for each species (turbot, sea bream and sea bass) in order to enable their application to monitoring or enforcement under a range of analytical variables; 2) To collate validation data and produce laboratory Standard Operating Procedures (SOPs) and guidelines for internal validation for end-users; 3) To develop and validate SOPs for data analysis to trace the origin of escaped aquaculture fish; 4) To carry out an inter-laboratory comparison exercise to assess the transferability of all methods to end-users.
The laboratory validation of the molecular tools was the largest single aspect and tested several parameters on 45 genotyping assays. Of the 45 assays which started the validation process, only 1 assay failed the validation due to a lack of repeatability. In addition, the validation included a statistical assessment of genotyping differences between the methods used to identify the SNPs and those used to genotype them as a smaller panel. This assessment identified genotyping differences (~3%) between the two genotyping methodologies. While this level is what one would expect between methodologies, it has an impact on individual sample genotypes, and we have shown that this could, in turn, lead to the incorrect assignment of individuals. While incorrect assignment may be rare, this result has an impact on how we would recommend the panels to be used. We would strongly recommend that the panels are not used in a forensic context until the reference dataset (currently produced with ddRAD methodology) has been re-genotyped with the same methodology used for the validation study (KASP™ genotyping chemistry). After this has been done the reference data and test sample data would be fully comparable and reduce the potential for mis-assignments. Larger panels of SNPs were also identified by species-lead partners, including candidates that a) exhibit differences between the populations sampled, b) are selectively neutral, and c) may be useful for an assessment of parentage. These larger panels were developed into Sequenom assays to be offered as a genotyping service to end users. These panels offer a flexible approach to exploiting the most useful AquaTrace-identified SNPs for new samples for a variety of possible uses, extending the legacy of the project. Production and test of laboratory Standard Operating Procedures (SOPs) was developed based on the results of the laboratory validation and inter-laboratory comparison. This SOP includes the details for ordering the assays from the manufacturer, and internal validation guidelines prior to using the method with new samples. Positive control material was produced for all 44 assays, and for as long as current stocks exist, these will be made available to end users to aid in quality control of genotyping.

Standardised data analysis methods
To trace the origin of escaped aquaculture fish standardized data analysis is necessary in order for the small traceability panels to be useful to end-users. Following the inter-laboratory comparison, data analysis SOPs were produced for the traceability panels validated for all three marine species. In all cases, they could be applied to trace the origin of escaped aquaculture fish within defined scenarios. For sea bass, the SNP panel was designed to differentiate wild fish of a Mediterranean or Atlantic origin. However, this panel could also be used for traceability of aquaculture escapes in the Mediterranean in locations where Atlantic broodstock are in use. For sea bream, the panel was selected to differentiate wild and aquaculture-origin fish among Greek populations. For turbot, the panel was designed to distinguish fish of aquaculture and wild origin in the Atlantic. Initially, five Aquatrace partners were keen to be involved in the inter-laboratory comparison. Before the comparison began, all partners were asked to perform a pre-validation test to check that the genotyping chemistry could be used successfully. Following this step, four partners were able to continue to the comparison stage and all reagents, test samples and controls were distributed with draft SOPs for laboratory work and data analysis. Concordant results were achieved across all four partner labs, in both the laboratory method and downstream data analysis.
Overall the validation of the molecular markers for traceability scenarios showed that the developed assays were generally robust to the assessed parameters. Likewise laboratory SOPs could be developed as well as protocols for result analysis and data interpretation. Also the assessment of the robustness of assays on various genotyping platforms in different labs was successful. This demonstrates that, provided the requirements of the SOPs are met in terms of equipment and experience, these methods should be straightforward for any end-user to apply within their facility.


MODEL SPECIES

Common garden experiments

Trout experimental conditions
Trout families were established in Denmark in the winter of 2012, and these were transferred to the Matre hatchery at IMR (Norway) in the late winter (early 2013). Trout families were crossed using broodstock representing: 1) genetically pure wild fish, 2) pure hatchery strain, and 3) admixed wild-produced fish, so that the following types of family crosses were available: I) ‘pure wild’ (wild ♂ x wild♀, three families), II) ‘pure hatchery’ (hatchery♀ x hatchery♂, two families), III) ‘admixed’ (admixed♂ x admixed♀, two families), IV) ‘F1 hybrid’ (wild♀ x hatchery♂, three families; hatchery ♀ x wild ♂, two families), V) ‘backcross to wild’ (admixed ♀ x wild♂, three families; admixed ♂ x wild ♀, two families), and VI) ‘backcross to hatchery’ (admixed ♀ x hatchery♂, three crosses; admixed ♂ x hatchery ♀, two crosses). These 22 families were successfully hatched and grown in a total of 8 experimental tanks (exactly as according to the experimental design see below).

In order to compare growth and survival of trout in common-garden experiments for fish of Danish origin, the families produced were counted, mixed and subjected to the following conditions: High temperature (24 hour light), low temperature (24 hour light), 12 hour day-length treatment (standard temperature), semi-natural tanks, low density (outdoor, natural temperature and light), semi-natural tanks, high density (outdoor, natural temperature and light) and control conditions, fish retained for 2½ years for QTL mapping and maturation experiments. With the exception of the latter conditions the above experiments were all terminated within the first year (as planned), providing DNA samples for analysis to identify pedigree and thereafter compute reaction norm analysis. At this stage, the fish for the maturation experiment were fin clipped and measured, then PIT tagged for continued rearing. These fish were transferred to saltwater and reared until age 2+. Trout families were also established in France for Mediterranean origin fish in winter 2012. A total of six families were established based on broodstock consisting of six unique males and six unique females that exhibited varying levels of introgression from stocked fish of Atlantic genetic origin. French common garden analyses were designed to address variance in hatching time under farm introgression. Here, the timing (hours from fertilization to hatching) was recorded for all brood from the six unique families, followed by sampling of offspring tissue for downstream analyses.
Salmon experimental conditions
Salmon families were produced in the Matre hatchery in Norway as planned in the autumns of 2012 and 2013, and were successfully hatched and reared in the planned experiments. This included rearing salmon in a total of 12 experimental tanks. Salmon tissue samples for DNA analysis was collected for all experiments. In order to compare growth and survival of salmon in common-garden experiments, the families produced were counted and mixed and subjected to the following conditions. Year-class 1: Standard hatchery conditions (control tanks), standard hatchery conditions – high temperature experiment, standard hatchery conditions – low temperature experiment. Year-class 2 experiments :Standard hatchery conditions (control tanks), standard hatchery conditions – extra high fish density, standard hatchery conditions – extra low fish density, semi-natural conditions, high fish density competition selection experiment and semi-natural conditions, low fish density competition selection experiment. These experiments were all terminated in the first year, fish measured and DNA samples taken. These samples and data were thereafter delivered to downstream analytical work, where data was analysed for reaction norms. Samples were also made available for QTL mapping. Salmon RNA samples were taken from fish produced in these experiments consisting of individuals from both year-classes of fish produced, i.e. 2012 and 2013. Overall, the common garden experiments were highly successful as no major upsets were recorded. It is always a huge challenge to undertake experiments with live fish. In this case the experiments included the challenging task of transportation of trout eggs across the North Sea between Denmark and Norway. In general high survival was observed for both trout and salmon throughout the experiment, indicating good rearing conditions and securing that the investigated genetic and environmental parameters were the main sources of variance in growth, gene expression and survival.


Linking adaptive life-history traits and genetic variation

The overarching objective was to identify the key differences between farmed (or stocked) fish and wild conspecifics at the molecular genetic level using Atlantic salmon and brown trout as models. More specifically, objectives were to 1) identify underlying genomic differences between wild and farmed fish and their hybrids under natural and controlled experimental conditions, 2) to identify links between genotype and phenotype for key life-history traits such as growth, maturation and survival under controlled environmental conditions, and 3) to elucidate gene-transcription differences between wild and farmed fish and their hybrids under controlled environmental conditions.
Screening of SNP variation was successfully completed in both salmon and trout, and SNP genotypes have been recorded for a total of 7000 wild, hybrid and farm-origin salmon (Norwegian) and of 2150 trout (1490 Danish and 660 French), representing wild, hatchery and various levels of admixed backgrounds. For the reaction norm analyses we additionally genotyped, using microsatellite markers, a total of 8474 salmon and 5474 trout from common garden experiments to establish parentage. All genetic marker data were hence successfully obtained and fed into the downstream analyses. The work set out to investigate differences between wild, domesticated and F1 hybrid individuals, the latter to explore heritability of gene expression, a factor for consideration when considering the impact of farm escapees. A series of three experiments were performed to examine gene expression at two juvenile life stages; sac fry and early feeding stage fry, in a total of two different farm strains and three different wild populations. Samples from these experiments (1348 in total) were successfully collected and analysed using a custom-designed, oligonucleotide microarray (Agilent platform) with four 44 K probes. Several approaches were used to map genomic variation influencing early stage life-history traits and to determine differences among family crosses with different genetic backgrounds with respect to wild versus farmed origin. For both model species, the main part of work was based on information from the common garden studies carried out as described above. These experiments were combined with analyses of gene expression (salmon only) and genomic markers, enabling an assessment of links between genotype and phenotype across wild, farmed and hybrid fish.
Salmon common garden experiments
Three experiments were conducted to improve knowledge of the genetic impacts of aquaculture on native salmon populations by investigating trait variation among wild, farmed, and first generation (F1) hybrid salmon for key life-history traits, specifically growth and survival across several environmental gradients. The first addressed trait variation under different temperatures, the second addressed trait variation under different levels of density competition and the third addressed trait variation under food competition.
In the first experiment, juvenile growth was compared among nine populations of farmed (two strains), wild (five populations), and F1 hybrid (two populations) salmon at three contrasting temperatures: 7°C (low treatment), 12°C (control), and 16°C (high treatment). On average, both farm populations outgrew wild and hybrid salmon, and the hybrid populations displayed intermediate growth. A significant temperature-by-population effect was found, indicating that the growth differences were population-specific, where some wild populations performed better than others relative to hybrid and farmed populations at certain temperatures. Therefore, the competitive balance between farmed and wild salmon may depend on the thermal profile of the river and the genetic background of the respective populations. While limited to first generation hybridisation (F1), results indicate that risk management of local fish populations could benefit from a more spatially resolved approach. In the second experiment, the relative growth differences between farmed, wild and F1 hybrid salmon were studied at three contrasting densities within a hatchery environment and two contrasting densities within a semi-natural environment. Mortality was low for all groups in the hatchery environment, and was highest for all groups in the low density semi-natural treatment. Farmed salmon significantly outgrew hybrid and wild salmon in all treatments. Within the hatchery environment, growth of all experimental groups decreased with an increase in fish density. Importantly however, the reaction norms for growth were similar across treatments for all groups. Thus, we found no evidence to suggest that the offspring of farmed salmon as a result of domestication have adapted to higher fish densities than wild salmon. Consequently, the substantially higher growth rate of farmed salmon observed in the hatchery compared to wild salmon does not appear to be caused by differences in their ability to grow in high density hatchery scenarios. In the third experiment, growth was compared between farmed, wild and F1 hybrid salmon when reared at three contrasting feeding regimes in order to understand how varying levels of food availability affects relative growth. Groups were reared in single strain tanks and the three treatments consisted of ad lib feeding, access to feed for four hours every day and access to feed for twenty-four hours on three alternate days in a week. Mortality was low in all treatments, and food availability had no effect on survival of all groups. As expected, the offspring of farmed salmon significantly outgrew the wild fish, while hybrids displayed intermediate growth. Furthermore, the relative growth differences between the farmed and the wild salmon did not change across feeding treatments, indicating a similar plasticity in response to feed availability. Although undertaken in a hatchery setting, these results suggest that food availability may not be the sole driver behind the observed reduced growth differences found between farmed and wild fish under wild conditions.
Salmon QTL studies
Following on from these common-garden studies, QTL based studies were performed to gain further understanding of specific genomic regions underlying the observed differences in life-history traits between wild and farmed salmon, and their hybrids. In total, analyses comprised SNP data for a total of ~7000 salmon juveniles from 134 individual families of either wild, farm or hybrid origin. The analyses led to the identification of several genomic regions associated with growth in salmon of farmed, hybrid and wild origin. For example, chromosome 2 was strongly associated with growth, and other chromosome regions linked with growth included 6, 9 and 15. This means that genetic variation at one or more genes located in these genome regions is statistically associated with growth (i.e. cause growth differences). It is however not possible with the applied design to identify specific genes causing the growth differences. DNA is inherited in blocks and on a first-generation cross as in our study the DNA blocks are substantial (due to the lack of male recombination in salmon). Within each of these blocks there may be up to 100 genes or more. The results are however fundamental for understanding what sorts of genetic differences have evolved between farmed and wild salmon as a response of domestication, and constitute an important starting point to look further into these genomic regions, e.g. in application of marker assisted selection.
Salmon gene expression studies
In analyses of gene expression and variance among wild and farmed salmon and their hybrids and backcrosses, the key observation was of high variability in both the number and identity of differentially expressed transcripts across the experiments, perhaps reflecting the different stocks, stages and stressors being examined. The extent of differential expression appeared to be stock combination specific. For example, comparing sac fry of the Mowi farm strain with wild population Figgjo fish showed much higher levels of differential expression compared to either the Salmobreed (farm strain) vs. Arna (wild strain), or Salmobreed (farm strain) vs. Vosso (wild strain) comparisons. Salmon may be expected to be highly variable among wild populations, reflecting longer term adaptation to local environmental conditions, which may explain these observations. Some general trends in gene expression differences were consistent among the experiments, in particular the gene expression ‘behavior’ of hybrids with respect to the pure stocks. It clearly highlights differential expression of some individual transcripts between farm and wild pure crosses. Behavior of hybrids is generally intermediate between parent stocks. However, some show maternal or paternal dominance. While maternal effects are considered to be particularly relevant during the embryonic stage of fish, it was interesting to find maternal dominance in gene expression persisting, at least to the fed-fry stage. Though not as pronounced, there was clear evidence of paternal effects for the expression of some genes. Though often ignored, there is a growing body of evidence for biologically significant paternal effects in fish and these results contribute to the notion that paternal genotypes affect gene-expression. In an effort to find common differentially expressed transcripts that might serve as universal markers for domestication, lists of significant differentially expressed genes from multiple microarrays were compared. These lists comprised data from the three experiments undertaken in the current work package, together with previously published compatible microarray studies. A total of 14 candidate genes were selected, for which primers were successfully designed for validation by qPCR. While the efficiency of all primer sets was found to be high (greater than 90%), indicative of robustly functioning assays, none of the transcripts showed consistently significant differential expression, either in control (constitutive) or stressed conditions. This lack of consistent behavior is likely to be attributable to multiple factors including, low fold change differences, high variability between individuals (both biological and technical in nature), lack of array specificity, co-assay of homeologue transcripts.
Trout common garden experiments:
A French and a Danish population were used to assess effects of farm/wild admixture on trout juvenile traits. Experiments were carried out using common-garden set-ups in France and Norway, respectively. Both populations had been stocked with a domesticated hatchery strain of extant origin, and molecular marker studies had previously shown that both populations were introgressed following hybridization between wild and farm-origin fish. Inference for French and Danish scenarios was used to address potential fitness effects at different juvenile stages (as well as at adult life stages, see section below). A series of common-garden studies were set up to examine effects of admixture on: 1) embryonic hatching rate (using French trout), and 2) juvenile survival and growth rates (using Danish trout). In Danish trout, a reaction norm study was set up to test effects of temperature regime on survival and growth under different combinations of admixture. In parallel, a QTL study was set-up for a replicate of the same experimentally produced families to identify genomic regions associated with growth in trout of farmed, admixed and wild origin.
In analyses using Danish trout as a model, trout families were produced so that they represented a range of hybridisation and introgression levels. In total, nine different categories of admixture, ranging from genetically pure wild, over various degrees of farm/wild admixed to pure farm strain were produced and their expression of life history traits was compared. The first experiment was a temperature reaction-norm study, analysing the relationships between a specific family genetic background (wild, farm or admixed) and the traits survival, size at age, weight and growth. Using linear modeling, relationships were examined across three different temperature regimes to assess whether the performance of fish with different genetic backgrounds varied across temperature. It could e.g. be expected that farm fish (and to various extents their descendants), which are typically genetically adapted to production at higher temperatures, would perform less well under lower temperatures similar to those experienced in the wild. Results showed that there was a positive and additive relationship between the degree of farm admixture and growth (i.e. the more farm genes are represented in a family, the higher their growth rate). Results showed that even families with relatively low levels of farm admixture (~20% of their genes originating from stocked farm fish) exhibited statistically significant deviations in growth patterns, compared to genetically pure wild families. Growth responses differed across temperature regimes and there was a trend for introgression having the relatively largest effect at temperatures closer to wild conditions, than at (higher) temperatures closer to farm conditions. However, the trend was not statistically significant and it is therefore not possible to determine if introgression is more likely to affect population expression of these fitness traits more under current, compared to under increasing temperatures. There was no evidence that survival from embryo to juvenile stage was associated with genetic origin. The temperature reaction norm study demonstrated a heritable effect of farm origin on life history traits size and growth. A QTL study was therefore applied to map the genetic background for juvenile and adult trait differences. Here, 1570 offspring from a total of 22 families allowed to grow and mature and then genotyped. Three main traits were examined: juvenile growth, sub-adult growth, maturation rate, as either one-, two-year old, or older. The last category included all fish that were alive but not maturing at the time of experiment termination in December 2015, and which would presumably mature as three-year old, or later. All offspring were sexed, using either visual inspection of gonads in all individuals at termination of experiment or molecular sexing of fish that died prior to final sampling, in order to determine whether relationships between traits and genetic properties were influenced by gender. A total of 700 fish had matured by the end of the experiment and 771 had not. QTL analyses identified links between juvenile weight, length, growth and maturation rate and genetic variation at multiple linkage groups. Genetic variants affecting growth and maturation were identified in both wild and farm fish. For analyses using French trout as a model, eleven mature males were collected in Les Usses River, in a site harboring a brown trout population with an Atlantic introgression rate formerly estimated to 0.40-0.45. Eleven mature females were obtained from La Puya hatchery farm, representing a local Mediterranean strain formerly showing a 14% Atlantic farm origin introgression rate. Eleven full-sib families were produced in December 2013. Three families were discarded due to low egg quality and low family size led to discard of two additional families just before hatching, and thus six families were used in the hatching experiment. Newly hatched fry in each family were recorded every six hours, and tissue samples were collected for subsequent SNP genotyping. Analyses of SNP data for parents and offspring showed a much lower introgression rate than expected, which probably led to reduced power for identifying QTL associated with hatching timing. Nonetheless, a QTL analysis identified five linkage regions showing a statistically significant association with hatching time (early versus late). However, comparison of allele occurrence in domesticated farm strain and presumably genetically pure Mediterranean samples did not allow determining the lineage origin of the segregating alleles.
Trout maturation experiments
Trout families from Denmark were in the common garden study allowed to mature under experimental conditions, in order to examine links between genetic background and the timing of maturation. As detailed above, a link was demonstrated between maturation and genetic variants observed in both farm and wild fish. There was further a trend that early maturation was associated with domestication in that the more of a fish’s genome originated from the domesticated strain, the higher the likelihood that it matured as one- or two-year old, rather than at an older age, which was the predominant life-history in pure wild fish. Analyses under natural conditions centered on assessment of whether the degree of admixture correlated with maturation timing in the wild. Here, analyses were based on trout from a French Mediterranean population in the Dranse L’Abondance in the Rhone system, exhibiting strong admixture between native fish and stocked fish of Atlantic farm origin. Individuals with different timing of spawning maturation within a single spawning season were genotyped using SNPs to establish whether there was a link between maturation timing and specific genetic variants. Samples of adult fish were collected in a 15 km stretch of the Dranse d’Abondance river (department de Haute-Savoie, France) throughout the spawning season. Age was determined by microscopic scalimetric methods. The stage of sexual maturity was evaluated by direct examination of gonads. A total of 273 fish were selected and genotyped. However, there was no clear evidence for a relationship between admixture degree and maturity, although there was a trend that more mature farm fish were encountered early in the season. Also, there was no relationship between admixture degree and age at maturity, as estimated from whether fish were mature or immature at a given age and year. If fish were assigned to a category by their admixture coefficient, as well as their maturation type (maturing at either 1+, 2+, 3+ or 4+ years), there was a clear separation among admixture categories, but not among maturation types, either across or within specific loci.


Mapping introgression of functional genes

Admixture and introgression result in populations and genomes with genetic ancestry from two or more source populations (typically one wild, and one or more farm strains). Based on analyses of introgression of specific loci within and among introgressed trout and salmon populations, the generality of introgression processes were assessed, giving information about whether introgression can be expected to follow similar patterns across genomic regions (‘genomic clines’), populations and species. The genomic clines approach works by ordering individuals from admixed populations by their hybrid index, which is the proportion of a hybrid individual's genome inherited from one parental population (or species) (in this case, the proportion of loci with pure wild ancestry). The patterns of introgression at each locus are then ascertained by comparing the locus-specific probabilities of ancestry to the hybrid index and fitting a cline model to each locus. Genomic cline analyses were carried out for one French and two Danish trout populations, and for three Norwegian salmon populations. Results showed that introgression of farm/hatchery genes into wild populations does not always follow the same pattern, in that some genes or gene regions appear to mix more easily into the wild gene pool, than do others. Conversely, some gene variants or gene regions appear to be restrained from entering wild gene pools, possibly through selection against individuals carrying them. Although it was not possible to qualify in the present study, there were overall more loci showing reduced introgression than loci showing increased introgression. Results also showed that genetic marker based estimates of admixture are sensitive to the specific loci that are used to assess admixture proportions, thus estimates of introgression may vary strongly depending on the specific markers that are used to assess introgression and whether they show linkage with functional variation. For example, in a sample of wild-caught trout collected in 2012, the average hybrid index (proportion of gene pool originating from farm fish) was estimated at 36%, 10% and 77%, depending on whether, respectively, 3572 genome-spanning SNP loci, 47 SNP loci showing reduced introgression or 14 SNP loci showing increased introgression rate, were included in the analysis. Temporal and geographical sampling showed low or no overlap in loci exhibiting non-random introgression patterns across geographical populations. Thus, only one locus was found to show reduced introgression in multiple populations, in this case all three salmon populations. Due to a combination of low genetic divergence between wild and farm genotypes (e.g. there were no fully diagnostic loci for farm/wild origin) and relatively low sample sizes for temporal replicates, it was not possible to demonstrate change in genomic clines over time. It was thus not possible to establish whether focal loci showed altered allele frequencies across temporal samples of individuals with similar overall admixture levels. Nonetheless, a significant result from temporal sampling was that once release of farm/hatchery fish ended, the genetic profiles of wild-caught fish converged back towards pre-impact profiles. Repeated temporal sampling is required to determine the trajectories of wild genotype profiles and to which extent they will re-establish to pre-stocking status.

Conclusions for common garden studies
The following overarching conclusions were reached from the common garden studies of the model species: 1) In all experiments farmed fish outgrew wild fish, while hybrids and back-crosses displayed mostly intermediate growth. As expected, growth was thus found to be an important life-history trait exhibiting strong difference between farm and wild fish, as well as between genetically pure wild fish and farm/wild hybrids. 2) Population-specific differences in traits (life-history traits as well as gene expression) were observed, suggesting that individual populations are likely be differentially affected by introgression. The genomic cline analyses showed that the population variation in genome-wide introgression patterns could potentially be a result of different types or strength of natural selection against specific gene variants found in farm strains. 3) Even at relatively low levels of introgression (i.e. if only a low proportion of an individual’s genetic variation originate from farm fish) the examined traits deviated from those expressed in genetically pure, wild fish. 4) In all experiments QTLs associated with traits showing farm/wild differentiation were identified, including traits juvenile hatching and growth rate, sub-adult growth rate and maturation timing. The conclusion from salmonid model species (from results generated by AquaTrace and other studies) is that non-native strains, i.e. individuals exhibiting different evolutionary trajectories, irrespective of whether they are of farm breeding or of extant wild origin, show mal-adaptation compared to native populations, and introgression hence lowers fitness in wild populations. It is evident that in marine fishes, different scenarios may apply. In some cases, aquaculture fish appear not to be genetically very different from wild fish, whereas in others they represent strains that have undergone several generations of breeding selection and are more strongly diverged from wild fish. Evidence from salmonids can be transferred to inference for marine fish. Firstly, hybridization between escaped farmed fish and wild fish will lead to genetic changes that can be assessed using genetic markers. Secondly, introgression can be monitored using genetic markers. Genetic monitoring can be applied to determine the degree of genetic change incurred by introgression (as a proxy for how much mal-adaptation is incurred on contemporary populations), and whether introgression leads to a general reduction in genetic variation of wild populations (thus affecting long-term genetic ‘health’ of wild populations). Evidence from salmonids shows that vulnerability to, and tolerance of, introgression can be determined on a case-by-case basis, and whether escapes from specific farms at specific frequencies and magnitudes are likely to incur fitness costs to wild populations. Predictions are most likely to be accurate if information can be obtained about 1) the genetic identity and variability of farmed strains and wild populations, 2) the structure and size of wild populations, and 3) the frequency, volume and age structure of escapes. A synthesis of the implications of model species results in relation to the generality of introgression processes and effects on fitness was produced for integration into the Risk Assessment White paper.

RISK ASSESSMENT AND MANAGEMENT RECOMMENDATIONS
an extensive Risk Assessment tackling the issue of potential genetic introgression of farmed escaped fish into wild populations has been undertaken which forwards a panel of eight recommendations that should help underpin the management of risks inherent to marine Aquaculture Activity. The Risk Asessment contains nine main chapters: [1] EXECUTIVE SUMMARY / [2] INTRODUCTION: FARMED FISH AND WILD CONSPECIFICS / [3] SEA BASS, SEA BREAM AND TURBOT / [4] RISK QUANTIFICATION AND CONSEQUENCES / [5] CONCLUSIONS / [6] RECOMMENDATIONS / [7] RISK ASSESSMENT MANAGEMENT SUPPORT / [8] GLOSSARY / [9] REFERENCES. The white paper starts out by looking at the populations at risk of the three target species. It includes a concise summary of the respective biology and exploitation patterns for the three species, as such knowledge is needed to enhance our understanding of the vulnerability and tolerance towards natural and anthropogenic impacts. The paper continues by describing stock structures followed by a depiction of possible hazards emerging from aquaculture activities. Moreover, based on a literature review, new empirical data, simulations and modelling, it provides an assessment of existing risks emerging from exposure of wild animals to farmed fish due to release and escape events. Finally, based on compiled and reviewed existing knowledge and AquaTrace research results, a set of recommendations directed at risk managers are delineated. For all three target species capture fisheries is small in comparison to aquaculture production. For European sea bass, capture fishing was until recently - apart from a few national rules - not regulated. In 2015 scientific analyses have reinforced previous concerns about the decline of the stock by the International Council for the Exploration of the Sea. For commercial and recreational fishing of wild gilthead sea bream some restrictions are in place, but in general, no management measures are implemented for wild stocks, except for fish, mostly juveniles, entering estuaries and coastal lagoons in the Mediterranean Sea. Turbot is an important by-catch species in demersal fisheries and not considered endangered. Declines in wild catches and some genetic evidence suggest however a historical reduction in population size. In contrast to the sea cage farming of sea bass and sea bream aquaculture, turbot farming is almost exclusively land-based, in the early stages of cultivation in closed recirculation systems and later in open flow systems for on growing. The AquaTrace breeding survey highlighted significant differences in selection efforts in the aquaculture industry, which are summarised in the RA. Despite its rapid development, the impact of breeding programmes on cultured stocks remains limited. Selective breeding will likely become more commonly applied as it creates great opportunities for aquaculture. This implies an enhanced drive for genetic diversification of farmed and wild fish. From the samples covering a large part of the native distribution range of all three target species, genomic markers were developed and used to genotype the sampled individuals, to characterise the wild and farmed populations and to reveal signs of introgression.
Marine species
European sea bass is composed of three main wild groups, an apparently almost panmictic population in the North-eastern Atlantic and two Mediterranean populations, separated at the Siculo-Tunisian Strait. The Eastern Mediterranean sea bass is genetically split in several subgroups while its western counterpart is relatively homogenous and shows closer affinity to the Eastern Mediterranean than to the Atlantic group. In the Western Mediterranean Sea there is some evidence for introgression of the Atlantic lineage. Off the Strait of Gibraltar there is also evidence for natural introgression of the Mediterranean lineage sea bass into the Atlantic population. Farmed populations show a range of genetic diversities, reflecting the broad origin of the source material, breeding practice and selection pressure, and resulting in some cases of pronounced genetic differentiation between farms. AquaTrace found evidence of genetic material from the Atlantic Ocean in Mediterranean farms and of Mediterranean material in one Atlantic farm. It is not feasible to identify a unique genetic profile for each farmed sea bass. However, some genetic characteristics point to unique sources of domesticated genetic material. Some wild samples show evidence of escapees and these show high proportions of individuals that are closely related to one another. Mediterranean samples show higher numbers of related individuals per sample than the Atlantic. In gilthead sea bream the results show the existence of genetic differentiation between Atlantic and the Mediterranean populations and within the Mediterranean populations between Western Mediterranean, Ionian and Aegean Sea. The latter appeared to be the most differentiated population. The population in the North Adriatic Sea emerged to be more similar to those of the Greek Ionian Sea and the population from the Italian Ionian coast looked more similar to West Mediterranean samples. The possibility that genetically close or very different fish may escape or be released and that both, leakage and mass escape events are likely to occur would require a localised case-by-case assessment of introgression and fitness risks. In turbot four main genetically differentiated regions across the distribution range have been identified: the Atlantic area, mostly compatible with a large panmictic population; the Baltic Sea, showing moderate differentiation from the Atlantic area; and the Mediterranean and Black Sea, which are highly differentiated from the Atlantic region. Farmed turbot is capable of surviving and reproducing in the wild leading to introgression in wild populations. This introgression is likely to have occurred over several generations. A set of genetic markers was developed to monitor escapes or releases from farms and to evaluate introgression in wild populations of turbot. Pure farm and introgressed individuals may represent close to 15% of individuals collected. The impact is related mainly to intentional releases, aimed to supplement depleted fisheries in threatened coastal habitats, although a minor risk could be associated with farming.

Model species
Salmonids belong to the most affected fishes when it comes to introgressive hybridisation. This is due to the widespread exposure to genetically divergent farmed fish, mainly from escapes, of which Atlantic salmon is an emblematic example. As a group, salmonids have therefore been intensively studied in attempts to clarify how hybridisation and introgression affect wild populations, serving as a paradigm for less well studied marine fish species. As the model species indicate, non-native strains show mal-adaptation compared to native populations and introgression hence lower fitness in wild populations. Since in the marine target species the farmed fish may be very similar genetically or very different from their wild conspecifics, the fitness effects caused by escapees in wild populations will be population and scenario specific. Some results obtained in modelling with salmonids are also valid for the marine target species. Firstly, hybridisation between escaped farmed fish and wild fish will lead to genetic changes that can be assessed using genetic markers. Secondly, introgression can be monitored using genetic markers. Genetic monitoring can be applied to determine the degree of genetic change incurred by introgression, and whether introgression leads to a general reduction in genetic variation of wild populations. Evidence from salmonids also shows that vulnerability and tolerance can be determined on a case-by-case basis, whether escapes from specific farms at specific frequencies and magnitudes are likely to incur fitness costs to wild populations. Predictions are most likely to be accurate if information can be obtained about 1) the genetic identity and variability of farmed strains and wild populations, 2) the structure and size of wild populations, and 3) the frequency, volume and age structure of escapes.
Factors identified to affect impact of introgression in salmonids can in general be directly transferred to marine fish, and mitigation actions can be planned to address them. Such factors are the genetic differences and the heritable trait differences between wild and farmed fish, the frequency of intrusion and the relative frequency of escaped to wild fish on spawning grounds. Efforts to avoid introgression can also be prioritised, incorporating aspects of both conservation and economic goals. It would for example be possible to conduct a spatially explicit assessment of whether the use and escapes of specific farm strains is more likely to inflict genetic damage in some areas compared to others, e.g. in areas inhabited by vulnerable, genetically unique populations. To that end, based on its research findings and by incorporation additional available information, AquaTrace has developed a Risk Assessment Decision Aid Table which should be applied as to decide on further action. In conclusion, the AquaTrace White Paper on Risk Assessment and Management Recommendations on Marine Aquaculture Escapees tackles an important and very acute issue. It constitutes a crucial deliverable of the project that builds on the entire set of AquaTrace research activities and findings. More detail on the expected implications of the risk assessment is provided in the section on potential impact.

Potential Impact:
Genomic resources
Our ability to develop and utilise genome-wide approaches to DNA analysis has been largely based on the recent, rapid advances in high-throughput DNA sequencing technologies that are delivering increasingly massive amounts of sequence data per unit cost and per unit time. Harnessing the information content of the genome requires methods that can reduce the complexity of the data and allow comparative analysis of informative genetic variation within the whole dataset. The identification, characterisation and selection of SNP DNA markers from throughout the genome currently provides the best approach to developing genetic analysis systems that deliver maximum information content with minimum cost for multiple samples. The rationale for this approach to aquaculture identification and traceability is that within the genome, a small proportion of SNP markers should be associated with, or correlate to, the population of origin of individual fish. The SNP development ‘pipeline’, starting with high-throughput sequencing and ending in SNP genotyping assays, can be designed in several different ways, depending on the available resources and project objectives. In AquaTrace, the five target species differed in their existing levels of SNP marker resources and genome sequence data, with salmon and trout already before the project benefitting, from significant prior research and development of genome-wide SNP panels. Sea bass had a relatively complete reference genome but fewer SNPs, and sea bream and turbot having more fragmented genome and transcriptome sequence data available. The AquaTrace production and assembly of genome-wide sequence data for sea bream and turbot complimented the reference genome already available for sea bass to create comprehensive genome wide sequence data for all three species. This genomic information can be ‘mined’ by other researchers within the field and by the industry. In addition to facilitating subsequent phases of marine genomic work these tasks have already delivered data that have formed the basis of a comparative technical paper. The identified, genotyped and in silico validated SNPs for the three marine species (2709 SNPs for sea bass, 1240 for sea bream and 755 for turbot) provide a significant permanent genomic resource for future research into these economically valuable species. This represents a significant advance in the techniques and resources available to aquaculture genetic management. Thus, they do not only represent a unique resource for genetic traceability of wild and farmed fish for the three species, but also provide a resource for designing SNP panels for parentage assignment or to be used in marker assisted selection by the industry.

Population genetic baselines
Establishing genetic baselines for natural and aquaculture populations of sea bream, sea bass and turbot has been necessary to achieve the downstream objectives of AquaTrace and has significantly advanced our fundamental understanding of the genetic population structure of these species. This, in turn, allows us to predict and manage change within commercial fisheries and aquaculture fish throughout Europe. Knowledge of the basic genetic make-up of farmed and wild fishes is a prerequisite for estimating potential gene flow from farmed escapees to wild populations; similarly the genetic consequences of escapees or restocking practices, which remains largely unknown in marine species, requires the availability of natural population genetic baselines (pre and post release of farmed individuals), to monitor the potential genetic and life-history trait changes in natural populations. Furthermore the ability to measure intra-population changes at genes of adaptive value allow human effects on aquaculture broodstocks to be observed within natural populations after escapement or restocking, providing insights into the neutral/adaptive behavior of selected traits. Briefly, our analyses showed that Mediterranean and Atlantic wild sea bass were highly divergent and differentiated further into intra-basin groups. The farmed populations displayed highly variable levels of genetic uniqueness and evidence of transfer of broodstock between basins. Sea bream wild populations showed weaker genetic structure, however; suggesting subdivision of the samples into four main groups. Farm broodstocks were much divergent from wild population and were characterized by lower amount of within-group genetic variability compared to wild populations. Four main different population units were suggested in turbot and farmed samples represented two distinct genetic origins of aquaculture broodstocks. These findings represent a very large addition to the knowledge of the genetic of the three species, and define previously unclear genetic pattern of differentiation of wild populations. In addition, this is the first extensive analysis of European broodstocks, which is an important resource for the conservation and management of this species. The concept of population management supported by DNA-based traceability methods has until recently only been employed for a number of salmonid species. Such applications have now become available for sea bream, sea bass and turbot. The development and application of population-based traceability methods in the three marine species will advance consumer protection, providing the means to verify fish labeled as originating from sustainable populations/stocks (MSC certified) or aquaculture production (ASC certified), representing a major step forward in the resources available to enforce of fisheries regulations and conserve population biocomplexity. By undertaking high-throughput SNP genotyping in natural and aquaculture populations AquaTrace has built a comprehensive, spatially explicit database of genetic information, which can serve as a baseline for further research, management and support of the industry.

Detection of escapees using traceability tools
The development and demonstration of traceability tools for determining the origin of farm escapees has been expanded by AquaTrace to include sea bream, sea bass and turbot. Secondly, the traceability tools themselves have been standardized, validated and tested across multiple laboratories following quality assurance standards that have provided the technical data and protocols necessary for their application to commercial or enforcement activities throughout Europe. Third, the tools will be able to support a broader range of applications than source farm identification alone. The population genetic baseline developed within AquaTrace will enable additional questions relating to wild geographic provenance, farmed provenance, genetic introgression and restocking success to be evaluated, using different combinations of SNP markers. The end result provides a new type of traceability system for marine aquaculture fish, flexible enough to answer multiple questions, reliable enough to be reproduced throughout Europe, robust enough to deliver high quality data and sufficiently cost effective to allow access by industry and fisheries control authorities alike.

Genetic impact measures
In order to evaluate effects of escaped or restocked domesticated strains on fitness in introgressed wild populations, there is a need to generate information on both 1) the molecular basis for traits associated with natural vs. domestication selection and 2) how specific domesticated genes ‘behave’ under an introgression process. It has therefore been of paramount importance to produce empirical knowledge on introgression processes in order to advance the field. Although methods and results are not likely to be 100% transferrable between salmonids and marine fishes, our approach has provided a framework which can be used to direct marine applications including in a restocking context. We have used salmon and trout as model species to assess potential impact of farm introgression on fitness related traits within a comprehensive research framework using the most advanced common garden facilities available. There, we have applied common-garden experiments to identify phenotypic differences related to fitness. Common-garden experiments may be defined as experiments whereby individuals of different genetic backgrounds (here, farmed, hybrid and wild) are reared under identical conditions. By subjecting all groups to the exact same environment, any resulting phenotypic differences between the genetic groups are a result of their genetic composition. When the same genotypic groups are compared in multiple environments, the different phenotypic response is known as the genotype x environment interaction. The shape of this variation among environments (genetic-based plasticity) is known as the phenotypic response norm. The common-garden approach thus offers a statistically robust, predictable and not least replicable means to accurately identify phenotypic differences between wild and farmed fish and their genetic backgrounds. Based on careful consideration of hypotheses and knowledge about traits potentially under local adaptation in salmon and trout, AquaTrace focused on the following fitness traits: 1) juvenile growth and survival rates, including in response to temperature regime, feeding regime, density and (semi-) natural versus hatchery environment; 2) Age at maturation 3) Rates of reproductive tactics. In combination with access to full-scale experimental common-garden experiments we took advantage of novel genome-scale molecular resources for both model species and state-of-the-art gene expression analysis methods in salmon, extensive information about genetic structure in natural populations and novel developments in statistical approaches to linking quantitative traits with specific gene regions and for establishing the genetic architecture of introgression across populations. We obtained the following general insights. 1) In all experiments farmed fish outgrew wild fish, while hybrids and back-crosses displayed mostly intermediate growth. 2) Population-specific differences in traits (life-history traits as well as gene expression) were observed. 3) Population variation in genome-wide introgression patterns could be a result of different natural selection against specific gene variants found in farm strains. 4) Even at relatively low levels of introgression the examined traits deviated from those expressed in genetically pure, wild fish. 5) In all experiments QTLs (Quantitative Trait Loci) associated with traits showing farm/wild differentiation were identified, including the traits juvenile hatching and growth rate, sub-adult growth rate and maturation timing. The overarching conclusion from salmonid model species generated in AquaTrace is that non-native strains, i.e. individuals exhibiting different evolutionary trajectories, irrespective of whether they are of farm breeding or of extant wild origin, show mal-adaptation compared to native populations, and introgression hence lowers fitness in wild populations. It is evident that in marine fishes, different scenarios may apply. In some cases, aquaculture fish appear not to be genetically very different from wild fish, whereas in others they represent strains that have undergone several generations of breeding selection and are more strongly diverged from wild fish. Evidence from salmonids can be transferred to inference for marine fish. Firstly, hybridisation between escaped farmed fish and wild fish will lead to genetic changes that can be assessed using genetic markers. Secondly, introgression can be monitored using genetic markers. Genetic monitoring can be applied to determine the degree of genetic change incurred by introgression (as a proxy for how much mal-adaptation is incurred on contemporary populations), and whether introgression leads to a general reduction in genetic variation of wild populations (thus affecting long-term genetic ‘health’ of wild populations). Evidence from salmonids shows that vulnerability to, and tolerance of, introgression can be determined on a case-by-case basis, and whether escapes from specific farms at specific frequencies and magnitudes are likely to incur fitness costs to wild populations. Predictions are most likely to be accurate if information can be obtained about 1) the genetic identity and variability of farmed strains and wild populations, 2) the structure and size of wild populations, and 3) the frequency, volume and age structure of escapes. A synthesis of the implications of model species results in relation to the generality of introgression processes and effects on fitness are documented in the Risk Assessment White paper (see below).

Risk evaluation tools
The project has summarized existing knowledge on the population structure in the relevant species, and the implications for genetically based differences in life-history traits among populations. This has in part been from the review of existing knowledge (species pamphlets), and new knowledge generated in the AquaTrace project. This has formed the basis for the risk analysis including suggestions for mitigation with a resulting white paper: ‘AquaTrace White Paper on Risk Assessment and Management Recommendations on Marine Aquaculture Escapees’. The white paper starts out by looking at the populations at risk of the three target species. It includes a concise summary of the respective biology and exploitation patterns for the three species as such knowledge is needed to enhance our understanding on vulnerability and tolerance towards natural and anthropogenic impacts. The paper continues by describing stock structures followed by a depiction of possible hazards emerging from aquaculture activities. Moreover, based on a literature review, new empirical data, simulations and modelling, it provides a preliminary assessment of existing risks emerging from exposure of wild animals to farmed fish due to release and escape events. Finally, based on compiled and reviewed existing knowledge and AquaTrace research results, the following eight recommendations are provided: 1) There is a need to formulate specific management goals linked to potential genetic risks emerging from EU-aquaculture, that set thresholds against which such risks could be evaluated and assessed. 2) An EU wide harmonized compulsory and immediate notification requirement of escape incidents from marine aquaculture installations covering the production from hatcheries to harvest of the market-size fish ought to be established. 3) AquaTrace proposes the implementation of a traffic light system, where established thresholds help Aquaculture and Risk Assessment Managers to take action 4) Stocking actions should be documented and recorded on a mandatory basis. 5) Scientific studies and assessments on populations at risk should be pursued and enhanced. 6) A traceability scheme for the trade and exchange of fish eggs and juveniles should be established. 7) The risk assessment approach should be refined to confined wild stocks and aquaculture areas. To this end regional monitoring programs should be established taking advantage of the traceability toolbox for European sea bass, gilthead sea bream and turbot developed by AquaTrace. 8) The introgression by farmed marine fish observed strongly recommends the evaluation of the impact on wild populations through common garden experiments, similar to those carried out in Atlantic salmon and brown trout by AquaTrace.

In conclusion, the AquaTrace White Paper on Risk Assessment and Management Recommendations on Marine Aquaculture Escapees tackles an important and very acute issue, It is generally acknowledged that aquaculture could be key to meet the escalating demand for fish worldwide, and also a major protein source and source of employment and income for the European Union (EU). Fostering sustainable aquaculture is one of the pillars of the Common Fisheries Policy and an important component of the Blue Economy and Blue Growth. However, the Scientific, Technical and Economic Committee for Fisheries (STECF), the scientific advisory body to the European Commission, confirms in their latest report that aquaculture production is dominated by Asian countries covering almost 90% of the production volume. In contrast, the EU28 contribution to world aquaculture production has been decreasing significantly over time in both volume and value terms, representing only 1.7% and 3.2% of global production in 2014. One impediment to EU aquaculture growth are concerns about the environmental sustainability of Aquaculture activity: There is clearly a need to reconcile aquaculture activity with existing environmental legislation both national as that of the EU. In fact, according to STECF, experts have pointed out that environmental regulations, and inherent difficulties in the licensing process due to multilevel governance and competition for space both on land and in the coastal zones continue to be the most important areas to be addressed to increase growth in the EU aquaculture sector. Foremost among such environmental concerns is fish escaping from their aquaculture production facilities (‘escapees’), which might pose a hazard to the integrity and levels of biodiversity through genetic impact of wild populations of conspecifics. Escapees are a feature of aquaculture that can occur both acutely and chronically, e.g. through leakage. Restocking of wild populations with farmed fish can constitute similar risks. We expect this White Paper to be introduced to aquaculture governance and management on EU-level. Major dissemination efforts have been pursued to that end, such as informing DG MARE and stakeholders in general. Moreover, a scientific publication will emerge from this effort as to inform the scientific community about this Risk Assessment approach and its relevance for a thriving sustainable and profitable aquaculture that is compliant with environmental legislation and enjoys a positive perception and acceptance by the public.

Contribution to the Ecosystem-based approach to fisheries management
The Common Fisheries Policy is moving towards the adoption of an ecosystem-based approach to management, requiring advice on the long-term effects of fisheries activities on the structure and functioning of marine ecosystems. In common with capture fisheries, there has been heightened awareness recently on the necessity to extend such principles to aquaculture. An ecosystem approach for aquaculture is a strategy for the integration of the activity within the wider ecosystem in such a way that it promotes sustainable development, equity and resilience of interlinked social and ecological systems: thus, an ecosystem approach to aquaculture management is not about managing or manipulating ecosystems but is concerned with ensuring aquaculture management decisions do not adversely affect ecosystem function and productivity, and so marine resource use is sustainable in the long term. The ability to identify and monitor the potential genetic impact of escapees and restocked fish into wild populations acquired through AquaTrace provides a baseline for conservation of ecosystem diversity and function as linked to sustainability through risk assessment and mitigation. Moreover, understanding the two-way interactions between aquaculture and the environment is a critical step towards securing an adequate regulatory framework. Forensically validated tools as provided here and ecosystem-based strategies will further promote public perception of aquaculture best-practice: a prerequisite for growth and sustainability because aquaculture relies strongly on a high-quality rearing environment and the demands of the market. Presently, scientific advice on policy making and legislation in relation to aquaculture is mainly channeled into recommendations for food safety aspects (e.g. through the European Food Safety Authority (EFSA)) and in relation to certification-based traceability to establish the identity of processed products. Governance in relation to wider aspects of environmental protection through considering biological interactions between farmed and wild fish (escapees and stocking) lacks a consensus legislative framework, and would profit from scientific knowledge and advice described above. Communication by the European Commission on “New Impetus to the Sustainable Development of European Aquaculture”, emphasizes the need for compatibility between aquaculture and the environment. In close cooperation with salient regulatory bodies such as ICES, and associated interactions at the level of the Commission, such as DG-Mare and DG-RTD, and the Technical and Economic Committee for Fisheries (STECF), strategies based partly on the input from AquaTrace should be developed to ensure that data collected during the AquaTrace project will be optimally implemented in an ecosystem-based approach to aquaculture. Such activity is secured by the fact that AquaTrace participants interact actively with ICES, FAO and STECF.

Implications for the development of aquaculture in the EU
Compared to many other regions worldwide, the EU faces major disadvantages hampering the development of its aquaculture sector. One major impediment is limited access to space for marine aquaculture. However the European Commission has stated to the European Parliament and the European Council that aquaculture activity provides great opportunities to the EU particularly with respect to technology development and innovation, the setting of standards and certification processes at EU and international level. Also through properly implemented knowledge and technology transfer the EU aquaculture industry can contribute to the supply of healthy and safe aquatic food products and help to decrease the current significant EU dependence on imports. To increase competitiveness, EU aquaculture products should enter the market as high quality products based on their environmental performance, elevated health standards and effective traceability schemes. Through the research activities provided by AquaTrace over the last four years the project has provided a valuable contribution to the development of competitive European aquaculture based on advanced technologies. AquaTrace has actively engaged with relevant stakeholders and also the European Commission to maximise the potential of uptake of knowledge and methods which have emerged from the project.


Main dissemination activities

The AquaTrace project has encompassed a wide range of outlets for dissemination of project information and results. This includes a very active website (https://aquatrace.eu/). The site has continuously reported AquaTrace progress to scientists, stakeholders, decision and policy makers as well as the general public. At the same time it has embraced and announced relevant information from the scientific community, such as meetings, workshops and conferences. AquaTrace widened the dissemination channels by also developing Twitter and Facebook accounts. These channels have been extensively used by consortium members and stakeholders interested in the project. We previously produced and AquaTrace video (available at the AquaTrace website and at our Facebook site). Five newsletters have been produced and disseminated, thus disclosing significant output from AquaTrace through various channels. As a consequence of the many channels of dissemination we have reduced the frequency of e-Newsletters, as much of the news was published through other media (see above). As deliverables within the project we have included an AquaTrace postcard and an introduction video outlining the content and context of the project which can be found at the project homepage. Project participants have produced more than 45 oral conference and workshop presentations and also provided a number of conference posters. In addition, they have delivered a high number of lectures and seminars related to AquaTrace at academic institutions across Europe. In relation to the general public a number of newspaper magazine and web-based articles have been published along with TV appearances. Many scientific manuscripts are in preparation or have been submitted, however; already four scientific papers have been published in well-recognized international scientific journals. Stakeholder sessions have taken place at all AT consortium meetings. This includes Turkish, Italian and Spanish stakeholders from the industry, NGO’s as well as local and central authorities involved in aquaculture, fisheries and nature management. Furthermore, the essence of AquaTrace was presented at a dedicated and well attended AquaTrace session at AE2016 in Edinburgh (September 22nd, 2016). The AquaTrace session included 9 oral presentations focused on the development of new molecular genomic tools for subsequent assessment and evaluation of the genetic impact of escaped farmed fish on the wild conspecifics. The majority of the presenters were affiliated with the project, thus representing public research institutions across Europe. In addition one presentation was provided by a US researcher in order to give an outside and global perspective of the focal issues. The session was initiated by a general overview of selective breeding programs and seed market in Europe. The presentation revealed that in general fish culture in Europe is moving towards dedicated selective breeding programs, where molecular genetic tools are being increasingly used. This includes parentage testing but also a number of genomic selection programs are in action. The increased use of selective breeding is also reflected in the seed production, where fewer larger breeding programs are supplying seed to many fish growers across Europe. Subsequently, the session focused on presentations of detailed results emerging directly from the AquaTrace project. First the coordinator provided a general overview of the project, including the participating partners, the rationale and ambition of the project, as well as a flavor of some of the results in the different work-packages. The introduction was followed by specific presentations on the development of molecular genomic tools for the three target species, sea bream, sea bass and turbot. SNP (Single Nucleotide Polymorphism) tools were presented including inferences of population structure in wild Atlantic and Mediterranean populations as well as in farmed populations. These results formed the basis of a presentation on the power for detecting escaped fish in the wild including their genetic interactions (introgression) with wild fish. Specifically assignment based and relatedness based methods were tested and evaluated for different traceability panels. The project´s common garden experiments on Atlantic salmon and brown trout were presented with the aim to assess possible fitness consequences of introgression between wild and farmed fish. A presentation on the fitness consequences in brown trout following a river stocking program was provided. One of the conclusions was that at even at a relatively low proportion of ‘farmed’ genetic material in an individual, trait differences could still be observed compared to wild fish. I.e. growth and maturation rates were elevated in fish with farmed input. The total outcome of the project formed the basis of the final presentation, which was focused on providing a risk assessment in relation to escaped fish from aquaculture on wild fish populations. Thus, the integrated information on the studied species in relation to population structure of wild/farmed fish and the rates and consequences of introgression was evaluated in the context of the general development of aquaculture in Europe. I.e. that aquaculture broodstocks are more intensively selected and seed production restricted to fewer main players. The session was generally well attended with between 40 and 70 people, who provided many excellent questions to the presenters.

In conclusion, the dissemination of the AquaTrace concept, results and implications have been very extensive, with diverse channels of output across different media. Likewise, the audience has included all stakeholders from the scientific community, the aquaculture industry, managers, policy makers and the general public. The dissemination of outputs will continue in the coming years, securing the legacy of AquaTrace.


Exploitation of results

Exploitable scientific results
The AquaTrace project includes a series of exploitable scientific results. First of all the genomic sequences and genome assembly for seabass and seabream provide the basis for ‘genome mining’ in relation to searching for genes and genomic regions, which can be used by scientists and the industry for further applications in both basic and applied research in wild and farmed populations. Likewise, the SNPs identified from ddRAD sequencing of turbot sea-bass and seabream can be exploited further in relation to delineating population structure and traceability in wild populations as well as for parentage and ‘Marker Assisted Selection’ in breeding programs by the industry. The trial validation of the molecular markers for traceability scenarios showed that the developed markers and related assays were generally robust and of high quality. Likewise laboratory SOPs could be developed as well as protocols for result analysis and data interpretation. This has resulted in the uptake of the AquaTrace SNPs by the industrial partner Labogena. This will involve all three of the marine species providing a list of ~200 SNPs for each species in order to develop panels for their Sequenome genotyping platform. These panels will include both outlier loci, and some neutral loci. In addition, the population genetic analyses provided insights on population structure of wild populations, as well as the origin and genetic variability of farmed populations. This data can be used by managers and policy makers to develop new management paradigms for wild populations, including impact of farmed fish on wild populations as outlined in the risk assessment and detailed in the section above. Likewise, the data can be explored in order to guide the foundation of new farmed broodstocks from wild and/or farmed populations. Finally, the documented trait (growth and maturation) differences among wild, farmed and introgressed salmon and trout and the associated QTL’s can be used by the salmonid farming industry in relation to faster and more efficient breeding, but also by scientists and managers to investigate and use in relation to evaluating the genetic impact of farmed salmonids on their wild con-specifics in a changing environment.


Exploitable technical results
The project also achieved a number of technical results which can be exploited by other researchers and the industry. The extensive collection of thousands wild and farmed seabass, seabream and turbot covering both Atlantic and Mediterranean populations and farms represents a unique resource, which can be used in future research projects. All samples have been properly stored for future reference, while sampling data are available through the AquaTrace database. Likewise, the salmon and trout families, including extracted DNA and RNA can be used for further analyses of the inheritance of trait differences among farmed, introgressed and wild fish. In addition the bioinformatics pipeline for ddRAD analysis is available for genomic data processing for similar types of data. Briefly, the pipeline is a series of analytical procedures used to handle very large volumes of genomic sequence data. The pipeline produced here aimed to combine all of the DNA sequences produced for all samples from the different libraries in the bioinformatics analysis. Customization involved evaluating multiple parameters within the sequence alignment and SNP calling stages to optimize the accuracy and the number of SNPs discovered while minimizing data loss. Thus the complete pipeline can be used as the basis for ddRAD sequence data analysis in other projects. Finally, AquaTrace has produced an extensive review and evaluation of assignment methods and software in relation to traceability methods. The review is not only relevant in a farmed/wild fish context but has relevance across a wide range of scientific disciplines.

Exploitable results for policy
The survey on selective breeding programs implemented in the European aquaculture industry, provides an extensive review of the status of selective breeding in the six most important species within the aquaculture industry in Europe today. It does not only provide a snapshot, but also provides inference on the development as it is comparable to a previous survey from 2008. Besides reviewing breeding practices it also evaluates the implementation of genomic tools and points to the research priorities of the industry. Thus, this already published paper is likely to have a significant impact on a wide range of policy initiatives regarding the industry and in particular for setting up national and European research programs to address the emerging needs of the sector in terms technological development including genomics. The web based presentations of species leaflets for seabass, seabream, turbot, Atlantic salmon and brown trout, includes all presently available biological information such as life-history traits, genetic and genomic characteristics, ecological, conservation, fisheries and aquaculture specific issues. As such they provide an extensive resource of information for all stakeholders. Specifically regarding policy they offer the opportunity to evaluate the status of the species in Europe, their current management and the most eminent knowledge gaps which need to be filled to assure sustainable management. The information from the species leaflets also feed directly into the most important policy deliveries of the AquaTrace project, the risks assessment white paper and policy document (highlights and summary of the white paper). The risk assessment collates and reviews all available previous information and the knowledge gained by AquaTrace on the target species. This includes a comprehensive overview of the distribution of genetic variation in farmed and wild populations including evidence of escapees and introgression, a thorough evaluation of tools for identifying farmed escapees and farm/wild hybrids and insights on the consequences of hybridization from the model species. Based on all available information AquaTrace provides a detailed risk assessment including eight concrete recommendations and management support tables which can be readily used for policy in order to mitigate genetic impacts of marine aquaculture escapees on wild populations. The implementation of these recommendations would provide a large step towards controlling and alleviating the effects of farmed escapes on wild populations and thus lead to a more environmentally sustainable and economically prosperous European aquaculture.


List of Websites:
https://aquatrace.eu/
final1-def-postcard-front.tif