T Cell Receptor Alpha Chain Genes in the Teleost Ballan Wrasse (Labrus bergylta) Are Subjected to Somatic Hypermutation

Previously, somatic hypermutation (SHM) was considered to be exclusively associated with affinity maturation of antibodies, although it also occurred in T cells under certain conditions. More recently, it has been shown that SHM generates diversity in the variable domain of T cell receptor (TCR) in camel and shark. Here, we report somatic mutations in TCR alpha chain genes of the teleost fish, Ballan wrasse (Labrus bergylta), and show that this mechanism adds extra diversity to the polymorphic constant (C) region as well. The organization of the TCR alpha/delta locus in Ballan wrasse was obtained from a scaffold covering a single copy C alpha gene, 65 putative J alpha segments, a single copy C delta gene, 1 J delta segment, and 2 D delta segments. Analysis of 37 fish revealed 6 allotypes of the C alpha gene, each with 1–3 replacement substitutions. Somatic mutations were analyzed by molecular cloning of TCR alpha chain cDNA. Initially, 79 unique clones comprising four families of variable (V) alpha genes were characterized. Subsequently, a more restricted PCR was performed to focus on a specific V gene. Comparison of 48 clones indicated that the frequency of somatic mutations in the VJ region was 4.5/1,000 base pairs (bps), and most prevalent in complementary determining region 2 (CDR2). In total, 45 different J segments were identified among the 127 cDNA clones, counting for most of the CDR3 diversity. The number of mutations in the C alpha chain gene was 1.76 mutations/1,000 bps and A nucleotides were most frequently targeted, in contrast to the VJ region, where G nucleotides appeared to be mutational hotspots. The replacement/synonymous ratios in the VJ and C regions were 2.5 and 1.85, respectively. Only 7% of the mutations were found to be linked to the activation-induced cytidine deaminase hotspot motif (RGYW/WRCY).

Previously, somatic hypermutation (SHM) was considered to be exclusively associated with affinity maturation of antibodies, although it also occurred in T cells under certain conditions. More recently, it has been shown that SHM generates diversity in the variable domain of T cell receptor (TCR) in camel and shark. Here, we report somatic mutations in TCR alpha chain genes of the teleost fish, Ballan wrasse (Labrus bergylta), and show that this mechanism adds extra diversity to the polymorphic constant (C) region as well. The organization of the TCR alpha/delta locus in Ballan wrasse was obtained from a scaffold covering a single copy C alpha gene, 65 putative J alpha segments, a single copy C delta gene, 1 J delta segment, and 2 D delta segments. Analysis of 37 fish revealed 6 allotypes of the C alpha gene, each with 1-3 replacement substitutions. Somatic mutations were analyzed by molecular cloning of TCR alpha chain cDNA. Initially, 79 unique clones comprising four families of variable (V) alpha genes were characterized. Subsequently, a more restricted PCR was performed to focus on a specific V gene. Comparison of 48 clones indicated that the frequency of somatic mutations in the VJ region was 4.5/1,000 base pairs (bps), and most prevalent in complementary determining region 2 (CDR2). In total, 45 different J segments were identified among the 127 cDNA clones, counting for most of the CDR3 diversity. The number of mutations in the C alpha chain gene was 1.76 mutations/1,000 bps and A nucleotides were most frequently targeted, in contrast to the VJ region, where G nucleotides appeared to be mutational hotspots. The replacement/ synonymous ratios in the VJ and C regions were 2.5 and 1.85, respectively. Only 7% of the mutations were found to be linked to the activation-induced cytidine deaminase hotspot motif (RGYW/WRCY).
Keywords: T cell receptor, ballan wrasse, Tcrα, polymorphism, somatic hypermutation, activation-induced cytidine deaminase motif, teleost inTrODUcTiOn T cells in jawed vertebrates are generally divided into two subtypes αβ and γδ on the basis of the heterodimeric T cell receptor (TCR). The αβ T cells are most abundant in circulation and lymphoid organs, while γδ T cells are found in mucosal and epithelial tissues. In humans and mice, γδ cells represent less than 5% of the total T cell population while in birds and ruminants they constitute more than 40% of the total peripheral lymphocytes (1,2). TCR αβ recognize peptides that are bound to major histocompatibility complex (MHC) molecules. TCR γδ recognize antigens directly, independent of MHC molecules in a manner similar to immunoglobulins (Ig). They are considered as a bridge between the innate and adaptive immune system as they use their receptor as a pattern recognition receptor (3). However, TCR γδ can also recognize phospholipids presented by CD1d molecules, suggesting that presentation by other non-classical MHC or MHC-like molecules might be possible (4). A unique TCRμ subtype first discovered in marsupials has subsequently been found in duckbill platypus (Ornithorhynchus anatinus), indicating that this locus was present in the last common ancestor of all extant mammals (5,6).
T cell receptor molecules have structural and organizational resemblance to the Ig heavy and light chains. TCRα is encoded by variable (Vα) and joining (Jα) gene segments combined with the constant region (Cα) gene, like the Ig light chain. TCRβ is encoded by Vβ, diversity (Dβ), and Jβ gene segments combined with the Cβ gene, like the Ig heavy chain. Rearrangements of V, D, and J gene segments and variability generated in these junctions create an enormous repertoire of receptors with different specificities, providing versatility and diversity to the immune system (7)(8)(9). The mechanism of gene rearrangement is similar in B and T cells. V(D)J recombination is mediated by enzymes encoded by recombination activating genes 1 and 2 which recognize highly conserved recombination signal sequences (RSS). RSS are heptamer and nonamer motifs that flank the V, D, and J gene segments.
Somatic hypermutation (SHM) is a key mechanism generating antibody diversity. In mammals, introduction of mutations in the recombined V(D)J gene in mature B cells is followed by selection of clones with higher affinities, typically in IgG-producing B-cells in germinal centers (24). Activation-induced cytidine deaminase (AID) deaminates cytosine (C) to uracil (U) in single stranded DNA creating U:G mismatch lesions, resulting in point mutations during SHM and double stranded breaks during class switch recombination. SHM occurs during transcription and primarily at RGYW/WRCY hotspot motifs (where G/C is a mutable position and R = A/G, Y = C/T, and W = A/T). SHM creates point mutations at a rate of 10 −3 mutations/bp/generation, a million fold higher than the background genome mutation rate (25). SHM at A/T base pairs (bps) (typically WA/TW motifs) is generated by a mismatch repair mechanism employing polymerase η or other low fidelity polymerases (26). SHM has been detected in Ig genes of both cartilaginous and teleost fish (27)(28)(29). The impact of SHM on affinity maturation in fish needs further studies to be fully understood (29). Somatic mutations within Ig light chain genes of zebrafish were found to be overrepresented at AID hotspot motifs like in mammals. Mutations were most prevalent in the V region, but a significant number of substitutions were introduced in the C region; the mutation frequency decreased slightly with distance from the V region (30).
It was believed that TCR diversity was generated by V(D)J rearrangement and that SHM did not occur in the TCR genes, although SHM was detected in TCRα of mice and in TCRβ of HIV-infected patients (31,32), and in TCR Cβ of two children after in utero stem cell transplantation (33). In the latter study, the frequency of Cβ mutations was significantly higher than in other groups of patients and healthy individuals (33). It was suggested that the lymphocytes of the babies presumably were under chronic activation. The SHM mechanism has also been found to target other genes, including oncogenes (34)(35)(36). A single cell PCR approach on lymph node germinal centers from a healthy person did not reveal SHM in T cells, in contrast to the situation in B cells where IgG clones were mutated (24). On the other hand, AID expression was unexpectedly detected in a subset of T cells in mice (37). More recently, studies of sandbar shark (TCRγ, TCRα) and camel (TCRγ, TCRδ) have shown that SHM occurs in TCRs of phylogenetically distant species (38)(39)(40), indicating a role in the diversification of the pre-immune repertoire. Accordingly, in mice it has been shown that early immature B cells are subjected to SHM, suggesting a role in B cell diversification as well as in the affinity maturation of antibodies (41,42).
The aim of this study was to characterize the TCRα genes in Ballan wrasse and analyze Vα and potential Cα diversity within this species. Ballan wrasse has attracted increasing interest as a "cleaner fish" recently for the biological control of salmon lice in fish farms. Ballan wrasse belongs to Perciformes. Approximately 40% of all fish belong to this group (more than 10,000 species).

MaTerials anD MeThODs samples
Three adult wild fish of Ballan wrasse Labrus bergylta (700-800 g) were caught from fjords near Bergen, Norway. Ethical approval was not required, as the study did not involve transport or experiments on live fish. Fish were killed with a sharp blow to the head immediately after they were caught and tissue samples were stored in RNA-later solution (Ambion). Previously deposited transcriptome data (intestine) from 34 juvenile fish were provided from individuals sampled in a commercial fish farm in Øygarden, Hordaland, Norway (NCBI accession numbers: PRJNA382082 and PRJNA360275).

Pcr-amplification of cDna Fragments and Dna sequencing
Primer construction for TCRα amplification was based on intestinal transcriptome data, genomic sequences and additional Vα sequence information obtained in the course of the present study ( Table 1). Amplification using standard Taq polymerase (Invitrogen) was performed as follows: denaturation at 94°C for 2 min, followed by 35 cycles of denaturation at 94°C (30 s), annealing at 55°C (30 s), and extension at 72°C (1 min/1,000 bps), and final extension for 10 min. Amplification using Accuprime TM Taq DNA polymerase and Accuprime TM High Fidelity Taq DNA polymerase (Invitrogen) was performed as follows: denaturation at 94°C for 2 min, 30 cycles of denaturation at 94°C (30 s), annealing at 55°C (30 s), and extension at 68°C (1 min/1,000 bps). DNA fragments were excised from the gel and further amplified for 5 cycles before cloning into pCR™ 4-TOPO ® vector (Invitrogen). Sequencing was performed at an in-house sequencing facility using Big Dye termination chemistry (Applied Biosystems).

sequence analysis and Phylogeny
DNA/protein sequences were compared to the GenBank/EMBL databases using BLAST. 3 DNA was translated into amino acid sequence using the translate tool available at ExPasy. 4 Multiple alignments were performed using ClustalW. 5 Phylogenetic trees were constructed using MEGA6 software and neighbor joining (NJ) and maximum likelihood (ML) matrixes with 1,000 bootstrap replicates (45).

calculation of Mutability index (Mi) and statistical analysis
Mutability index is a measure of observed/expected number of mutations for a specific nucleotide without target bias. A mutability score of 1 represent unbiased mutation, while higher scores indicate that a specific nucleotide is selected for mutation. Relative frequency of each nucleotide was multiplied by total number of observed mutations within all sequenced clones to calculate the expected number of mutations. Observed numbers of mutations were divided by expected numbers for each nucleotide to calculate MI. Chi-squared analyses of MIs were carried out by comparing observed mutational frequencies to their expected (unbiased) mutational frequencies. P values <0.01 were considered statistically significant.

resUlTs genomic Organization of the Tcr α/δ locus in Ballan Wrasse
Ballan wrasse TCRα sequences were identified by BLAST searches in an intestine transcriptome database using salmon TCRα as query (18). In the course of the present work, genomic sequence data of a heterozygous individual became available, and two scaffolds containing two allelic variants of the Cα gene in Ballan wrasse were identified by BLAST searches of whole genome shotgun data (GenBank): LaB_20160104_scaffold_928 (99,234 nt) and LaB_20160104_scaffold_4467 (16,780 nt). The Cα gene consisted of three exons corresponding to the Ig domain, the connecting peptide (CP) and the transmembrane (TM)/cytoplasmic (CYT) part. In total, 65 putative Jα segments Cα is encoded by three exons shown in green, which are followed by 65 Jα segments. Light blue arrows present potential Jα segments, while dark blue represent the sequences found in cDNA clones. Cδ is encoded by three exons indicated here as Cδ1, Cδ2, and Cδ3. One Jδ segment is shown in yellow and two Dδ segments are represented in gray (Dδ1 and Dδ2). The SMG-7 gene was identified upstream of Cδ and its location is shown by the pink arrow. were found by manual inspection of the region upstream of Cα in scaffold 928. All putative Jα segments contained the highly conserved core motif FGXG or slightly modified versions of this, and splice sites and RSS flanking the Jα exons. The presence of J segments was further confirmed by alignment of transcriptome data with scaffold 928, using the IGV program (46). The TCRα cDNA clones characterized in this study contained 45 of the 65 identified Jα segments. A single Jδ segment and two putative Dδ segments were identified upstream of the Cδ gene. As in other teleosts, the SMG-7 gene was identified further upstream of Cδ (Figure 1; Table S1 in Supplementary Material). Scaffold 4467 was shorter and represented the other allele of the TCRα locus, comprising the Cα gene and 13 Jα segments (Table S2 in Supplementary Material). Several scaffolds containing Vα genes were found in the genomic sequence database, but a complete assembly of all Vα genes was not possible based on the present whole genome shotgun data.

sequence analysis of Tcrα cDna clones
The assembled Ballan wrasse TCRα sequence was used as a basis for primer construction, and cDNAs encoding part of the leader sequence, V/J, and Cα were amplified by PCR. In total, 79 distinct TCRα clones were analyzed from three individuals. The V gene regions of the cDNAs were sorted into four groups based on 75% nucleotide identity ( Figure S1 in Supplementary Material). The translated sequences in group Vα1 showed all conserved characteristics of a V domain, while Vα2, Vα3, and Vα4 lacked the conserved cysteine (Cys) at position 26 (replaced with tyrosine); a pattern which was also seen in bicolor damselfish and olive flounder (16,17). Figure 2 shows representatives of  Wrasse TCR Cα encodes a polypeptide of 112 amino acids. The Cα region can be divided into an Ig domain, CP, TM, and CYT part. The structurally important Cys residues in the Ig domain and CP are conserved. The TM region is the most conserved region containing the positively charged residues lysine and arginine involved in the assembly of the TCR-CD3 complex. Multiple sequence alignment demonstrated that Ballan wrasse TCR Cα has 47.3% sequence identity to pufferfish, 46.6% with salmon, 40.2% with zebrafish, and 38.6% with cod, 31% with mouse, and 26.6% with human ( Figure 3A). The phylogenetic relationship between Ballan wrasse Cα and the orthologous molecules in other species is shown in Figure 3B.

cα somatic Mutations
Molecular cloning of TCRα from thymus cDNA of individual B1 showed that there was a significant number of point mutations in Cα (1.75/1,000 bps). Amplification of the first sample was done with standard Taq polymerase and 5,690 bps were sequenced (18 cDNA clones). In a second experiment, TCRα from spleen cDNA was amplified from individual B4 with high fidelity Taq polymerase. As a control, CD3ε cDNA was amplified under the exact same conditions. The frequency of mutations in Cα was 2.30/1,000 bps versus 0.62/1,000 bps in CD3ε. In total, 6,952 bps of Cα (22 cDNA clones) and 12,992 bps of CD3ε were analyzed. In a third experiment, TCRα was amplified from spleen cDNA with high fidelity Taq polymerase from individual B6, revealing 1.46 mutations/1,000 bps in a total of 12,327 bps (39 cDNA clones) ( Figure 4B; Table 2). All cDNA clones examined were confirmed to be unique, possessing distinct V regions. Two clones were regarded to be artifacts of PCR jumping (i.e., partial extension on cDNA from one gene and final extension on another, resulting in hybrids of two alleles). The replacement to synonymous mutation ratio (R/S) was 1.85 and most replacements were conservative (i.e., the biochemical properties were not changed; hydrophobic, charged, and neutral, etc.). In total, 93% of the nucleotide substitutions were transitions (primarily A to G and T to C). Of all mutations, 37.5% were targeted at WA motifs, whereas 7.5% were targeted at AID motifs. Three WA motifs were present in WRCY (AID hotspot), but the targeted nucleotide was not at the C position (Table S4 in Supplementary Material).

Mutability indices for Mono, Di, and Trinucleotides indicate Targets for cα Mutation
To determine which nucleotides or combinations of adjacent nucleotides were preferentially targeted for mutation in the Cα gene, MIs for mono, di, and trinucleotides were calculated. Chisquare analysis showed that only A nucleotides were significantly targeted for mutation, 93% of which were transitions. C nucleotides were significant cold spots for mutation. MI scores showed that nucleotides were targeted in the order: A > T > G > C for somatic mutations ( Table 3). The dinucleotide MIs revealed that AA, AT, GA, and AG were significant targets for mutation and when the analysis was expanded to trinucleotides, the dinucleotides were most frequently targeted in AAT, AAA, and AGA combinations (Tables S5 and S6 in Supplementary Material).
This pattern of substitution indicates that mutations targeted particular nucleotides and combinations.

Vα somatic Mutations
To study the mutation pattern in one single Vα gene, the B4-34 sequence was selected, and primers (TCR-VF/TCR-CR) were designed to amplify the VJ region (and a relatively short part of Cα). High fidelity Taq polymerase was used for amplification of spleen cDNA from individual B4, and the resulting PCRfragment was cloned. In total, 48 clones were analyzed. The 48 cDNA clones contained 25 different Jα gene segments (Table S1 in Supplementary Material). Out of the 48 clones, 45 were in frame. One clone had a stop codon (caused by point mutation) in the V sequence while two clones had frameshifts at the VJ junction. Alignment of the 48 cDNA clones indicated the presence of two   subgroups, highly similar to B4-34 (differing at two nucleotide positions). The subgroups were treated separately before the total number of substitutions at each position was calculated. Substitutions at positions 82 and 95 were considered allotypic differences as they had two alternative nucleotides represented by approximately 50% each ( Figure S2 in Supplementary Material). Of the total 102 substitutions, transitions (53%) were more common than transversions (47%) with a T/V ratio of 1.12. Transversions were relatively abundant in FR2 and complementary determining region 2 (CDR2) with a T/V ratio of 0.3. When there is no bias toward transition or transversion, the theoretical ratio of transition to transversion is 0.5 for random substitutions.
The overall ratio of replacement to synonymous substitutions was 2.5. Replacement substitutions were more frequent than silent mutations in both CDRs and FRs, except for FR4 where silent mutations were dominant (Table S7 in Supplementary Material). SHM studies in mammals and other teleost Ig variable genes have shown relatively high T/V and R/S mutation ratios (30).

Mono, Di, and Trinucleotide Targets in the Vα1 gene
The MIs for mononucleotides in the B4-34 group showed that nucleotides were preferentially targeted in the order: G > C > A~T. Of total mutations, G nucleotides were mutated 56.8%, followed by   Table 4). Analysis of dinucleotide MIs showed that CG and GC were the preferred targets for mutation, while AT were mutational cold spots (Table S8 in Supplementary Material). GC was found to be a significant target for mutation in human and catfish Ig heavy chain genes and zebrafish Ig light chain genes (28,30,47). When the analysis of the wrasse B4-34 group was expanded to trinucleotides, CG and GC dinucleotides were targeted most in GCG and CGA. Other combinations with significant MIs were AGG, GGT, and GAT (Table S9 in Supplementary Material).

DiscUssiOn
The present study has shown that the TCRα genes in the teleost Ballan wrasse are subjected to SHM, and that this process also introduces some diversity in Cα. Similar to the situation in other teleosts (17, 23) the Cα gene in Ballan wrasse is polymorphic. Analysis of 37 fish identified 6 allotypes of Cα, each with 1-3 amino acid substitutions. Although TCR polymorphism is widespread among teleost fish it is tempting to suggest that some of the TCR Cα diversity observed in teleost cDNA pools might be a result of SHM.
The first attempt to amplify TCRα cDNA was based on a primer pair from a leader exon to the end of Cα, revealing four families of Vα genes. To analyze the SHM in a single Vα gene we selected the B4-34 as template and made the forward primer from the boundary of the leader and Vα exons. The resulting sequences were divided into two groups and treated separately to avoid overestimation of mutation frequencies ( Figure S2 in Supplementary Material). The two groups might represent two highly similar V genes which have been amplified by the B4-34 primer pair.
The distribution of substitutions in TCR VJ and Cα is shown in Figure 5. The substitution rates were plotted against 20 bps nucleotide intervals. The first nucleotide corresponds to the third amino acid of FR1. Primers were designed from the start of the Vα gene and the first codons were, therefore, not included in the calculations. In the VJ region, CDR2 showed the highest frequency of substitutions. In Ballan wrasse TCRα the frequency of AID motifs was found to be much higher in the V/J genes (~8/100 bps) than in the Cα gene (4/100 bps), but no direct relationship was found between mutation frequencies and AID motifs in this study. In the VJ region, 6.8% of the mutations were present in AID motifs, while 9.2% were linked to WA/TW motifs, as compared to 40% in Cα. In a study of Ig variable genes it was found that SHM do occur in the absence of AID motifs and were predominantly G to C substitutions although the mutation frequency was lower than found in the presence of AID motifs (48). Mutability index analysis confirmed that nucleotides were targeted differently in the VJ and Cα region. In the VJ region, G nucleotides were targeted in CG and GC dinucleotides, and at GCG and GGG trinucleotide combinations. The mutations in Cα were primarily on A nucleotides, targeted mostly at AA (WA motif) and AT dinucleotides, indicating a mismatch repair mechanism employing polymerase η or other low fidelity polymerases (26). Replacement substitutions were dominant in both CDR1 and CDR2 and typically conservative. In a study of mouse TCR, extensive diversification by mutagenesis of CDR1 and CDR2 did not affect MHC binding (49), demonstrating that SHM of these regions is acceptable. In both Vα and Cα, replacement mutations were twice as frequent as silent mutations. Mutation frequencies in VJ versus Cα were found to be 4.5/1,000 bps and 1.76/1,000 bps, respectively. Key residues in Cα were conserved, showing that replacement substitutions had no impact on structural stability or interactions with CD3. Targeting of Cα is likely a side effect of SHM in the VJ region, like in the Ig light chain genes of zebrafish (30).
In the initial amplification and cloning of Ballan wrasse TCRα cDNAs about 24% of the clones had stop codons or were out of frame in the VJ region. The abundance of non-functional TCRα transcripts was similar in thymus and spleen. When narrowing the PCR-amplification to the B4-34 gene(s) the frequency of transcripts with stop codons or frame shifts decreased to 6.25%. Thus, it appears that there is a significant amount of non-functional transcripts in circulation, while some subpopulations of functional clones expand. In salmon, it was found that approximately 32% of TCRβ transcripts and 10% of TCRα transcripts in blood lymphocytes had stop codons or were out of frame (18,50). In another study of salmon, about 10% of the TCRα transcripts from thymus were non-functional (20). Corresponding frequencies in rainbow trout thymus were 32% for TCRβ and 12.5% for TCRα (11,51). In the amphibian Mexican axolotl more than 30% of the TCRβ transcripts from thymus and spleen, and 13.6% of TCRα  transcripts from thymus were sterile (52). The high fraction of nonfunctional TCRβ transcripts in ectothermic animals contrasts the situation in mammals where these mRNAs are eliminated (53, 54), although not absolutely (55). A more "leaky" system in coldblooded animals might be a result of less efficient cell proliferation and control mechanisms compared to higher vertebrates. From the first amplification, TCRα clones which were in frame counted for 72% of the Cα mutations. As almost all replacement mutations were conservative there is no reason to believe that these are not incorporated into the functional T cell repertoire. Considering that we find somatic TCRα mutations in thymus as well as in spleen, SHM is probably involved in the diversification of the pre-immune TCR repertoire in Ballan wrasse, at the same time introducing some Cα diversity. A survey of translated TCRα ESTs in public databases revealed many amino acid substitutions in Atlantic salmon TCR Cα as well ( Figure S3 in Supplementary Material). Thus, it is plausible to assume that SHM of TCR is a common phenomenon in teleost fish. However, SHM in mature T cells of wrasse cannot be ruled out either, and is an interesting topic for further research. SHM of TCR is generally believed to be restricted. Typically, an increase in TCR affinity after secondary challenge is minor compared to the affinity maturation of antibodies in mammals (56). It has been suggested that high-affinity TCR clones might be unfavorable due to prolonged binding, impairing serial interactions. On the other hand, somatic mutation of TCR genes in mature T cells has been documented (31,32), suggesting that SHM of TCRs can occur under certain conditions, e.g., chronic activation (33). Vaccination of fish might facilitate conditions that trigger inappropriate immune responses (57,58), but wild caught fish were used in the present analysis, implying that the results presented here are a normal situation in adult fish.
In conclusion, this study has shown that the TCRα in the teleost Ballan wrasse is subjected to SHM. The mutation frequency was highest in CDR2, although mutations were also evident in the constant part and FRs of TCRα. A high-throughput sequencing approach will be an interesting future study that can provide a Teleost TCR Alpha Somatic Hypermutation Frontiers in Immunology | www.frontiersin.org May 2018 | Volume 9 | Article 1101 more complete overview of the effects of SHM during the development of the TCR repertoire in the teleost fish.

eThics sTaTeMenT
All work complied with relevant ethical guidelines and regulations. Ethical approval was not required as the study did not involve transport or experiments on live fish.

aUThOr cOnTriBUTiOns
SB and IH designed the study. SB did experimental work and data analysis. KL and ØS worked on high-throughput sequence data. IH took part in data analysis and supervised the whole study.
acKnOWleDgMenTs We thank Dr. Lindsey J. Moore for comments on the manuscript.

FUnDing
The present study was funded by the Norwegian Research Council (Project number: 244396).