“A cleaner break”: Genetic divergence between geographic groups and sympatric phenotypes revealed in ballan wrasse (Labrus bergylta)

Abstract Capture and long‐distance translocation of cleaner fish to control lice infestations on marine salmonid farms has the potential to influence wild populations via overexploitation in source regions, and introgression in recipient regions. Knowledge of population genetic structure is therefore required. We studied the genetic structure of ballan wrasse, a phenotypically diverse and extensively used cleaner fish, from 18 locations in Norway and Sweden, and from Galicia, Spain, using 82 SNP markers. We detected two very distinct genetic groups in Scandinavia, northwestern and southeastern. These groups were split by a stretch of sandy beaches in southwest Norway, representing a habitat discontinuity for this rocky shore associated benthic egg‐laying species. Wrasse from Galicia were highly differentiated from all Scandinavian locations, but more similar to northwestern than southeastern locations. Distinct genetic differences were observed between sympatric spotty and plain phenotypes in Galicia, but not in Scandinavia. The mechanisms underlying the geographic patterns between phenotypes are discussed, but not identified. We conclude that extensive aquaculture‐mediated translocation of ballan wrasse from Sweden and southern Norway to western and middle Norway has the potential to mix genetically distinct populations. These results question the sustainability of the current cleaner fish practice.


| INTRODUC TI ON
Unlike capture-based fisheries, global aquaculture production has increased markedly over the past decades and has major prospects of further expansion (FAO, 2018). As such, it is often viewed as one of the most important sources of food for the growing human population. However, depending on the nature, location, volume of production, and farmed species in question, aquaculture may elicit a variety of negative effects on wild populations and the surrounding environment. For salmonid aquaculture that is conducted in open sea-cages in both the Atlantic and Pacific oceans, sea lice, which infest both farmed and wild salmonids, represent one of the most | 6121 SELJESTAD ET AL. significant and persistent challenges (Taranger et al., 2015;Torrissen et al., 2013).
Of the species used as cleaner fish, ballan wrasse is the most valuable due to its hardiness and large size (Blanco Gonzalez & de Boer, 2017;Skiftesvik, Bjelland, Durif, Johansen, & Browman, 2013). It is a hard-bottom associated species, with a natural distribution from the Iberian Peninsula to a northern terminus in Trøndelag, mid-Norway (Quignard & Pras, 1986). Reaching an age of more than 25 years (Dipper, Bridges, & Menz, 1977;Dipper & Pullin, 1979) and a size of up to 60 cm (Quignard & Pras, 1986), it is the largest and longest living of the wrasse in northern Europe. Ballan wrasse is a monandric protogynous hermaphrodite; that is, all fish are born female and change sex to become males at ~6 years of age, usually before reaching 40 cm in length (Darwall et al., 1992;Dipper & Pullin, 1979;Muncaster, Norberg, & Andersson, 2013). Protogynous life histories tend to result in highly female skewed population sex ratios (Allsop & West, 2004), leading to smaller effective population size (N e ), which in turn increases the likelihood of developing population structure (D'Arcy, 2013). In addition, ballan wrasse show strong site affiliation and have small home ranges (Mucientes, Irisarri, & Villegas-Ríos, 2019;Skiftesvik, Durif, Bjelland, & Browman, 2015;Villegas-Ríos, Alós, et al., 2013). Males also keep harems of females with which they spawn on the benthic substrate (Darwall et al., 1992). The combination of these traits is thus likely to generate population structure (Dipper et al., 1977;Dipper & Pullin, 1979).
Ballan wrasse are phenotypically diverse and display two main phenotypes, spotty and plain, which can be found in sympatry.
In Galicia, northwest Spain, these two phenotypes have different common names and display differing life histories, with spotty fish growing larger and faster than plain fish that invest in earlier reproduction (Villegas-Ríos, Alonso-Fernández, Domínguez-Petit, Alonso-Fernández, Domínguez-Petit, & Saborido-Rey, 2013). A genetic difference between these two phenotypes in Galicia has been revealed through microsatellite DNA analysis (Quintela et al., 2016), and assortative mating has been proposed as a mechanism to maintaining the difference (Villegas-Ríos, Alonso-Fernández, Domínguez-Petit, et al., 2013). These two phenotypes can also be found in other parts of the distribution range (Villegas-Ríos, Alonso-Fernández, Fabeiro, Alonso-Fernández, Fabeiro, Bañón, & Saborido-Rey, 2013), but the genetics underlying the phenotypic divergence is yet to be examined in these areas.
The majority of wrasses used as cleaner fish in aquaculture are wild captured. Currently, the Norwegian labrid fishery is divided into three management regions, where the allocation of quotas, fishing-gear regulations, and closed seasons differs. However, knowledge of population genetic structure was not available when these management areas were initially decided upon. Of the total 18 million wrasse quota, 4 million are allocated to the southernmost management region where there are virtually no salmonid farms. From this area, wrasses are transported by trucks or boats to salmonid farms in western and middle Norway. In addition, approximately 1 million wrasses are imported annually from Sweden, and permits have also been granted for imports from Denmark although a commercial fishery is yet to be established in the latter region (Rueness et al., 2019).
Little is still known about the fate of cleaner fish in sea-cages, but escapes and hybridization with wild populations have been documented (Faust, Halvorsen, Andersen, Knutsen, & André, 2018).
Furthermore, deliberate release of wrasses following the end of salmonid production has also been reported (Taranger, Svåsand, Kvamme, Kristiansen, & Boxaspen, 2014). Thus, this practice can lead to gene flow between previously isolated populations, as suggested for goldsinny (Jansson et al., 2017) and reported for corkwing wrasse (Faust et al., 2018). This type of human-mediated movement of organisms between areas, termed translocation, has been identified as an increasing threat to biodiversity and genetic diversity worldwide (Laikre, Schwartz, Waples, & Ryman, 2010;Prenter, MacNeil, Dick, & Dunn, 2004), through co-introduction of (novel) disease vectors and/or parasites and by leading to extrinsic outbreeding depression and/or breakdown of population structure (André et al., 2011;Glover et al., 2017).
The population genetic structure of ballan wrasse has yet to be studied in Scandinavia. Given the widespread harvest and translocation of this species, such studies are required to provide sound management advice. Thus, the present study was designed with two primary objectives: (a) to investigate the population genetic structure of ballan wrasse in Scandinavia, including comparative analysis with Galician populations, and (b) to investigate the relationship between genetic and phenotypic variation in this species in both Scandinavia and Galicia. In order to achieve these objectives, we genotyped 1,058 ballan wrasse from Scandinavia and Galicia in Spain, using a newly developed SNP panel and existing microsatellite markers.

| Sampling
A total of 1,058 ballan wrasse were sampled from 16 locations in Norway, two in Sweden and one in Galicia, Spain (Figure 1). At five locations (see Table 1), fish were photographed for phenotypic analysis of plain and spotty morphs. These individuals were visually sorted according to phenotype (see example in Figure 2) in order to investigate the potential connection between phenotype and genetic background and were kept separate for analyses where appropriate.
Scandinavian fish were caught using fyke nets or pots. Fish were killed and thereafter sampled by trained individuals in accordance F I G U R E 1 Map of the study area of ballan wrasse (Labrus bergylta). (a) Scandinavian sampling locations. Northwestern Scandinavian locations indicated by squares, southeastern by circles. Approximate genetic cline center and range in Scandinavia indicated by line and area shaded in green. (b) Galician sampling location. See Table 1 for detailed information about the sampling locations with EU Directive 2010/63/EU and national legislations. Galician samples were obtained from a local fish market in A Coruña (43.359 N, 8.402 W). All fin clips were stored in 96% ethanol for later analysis.
Putative SNPs, identified from the procedure above, were imported in R v.3.5.0 (R Core Team, 2018) and filtered to only retain SNPs with the highest confidence of polymorphism based on the Phred-scaled genotype likelihoods of the.vcf output file from bcftools. To cover as much as possible of the genome, as well as to avoid strong linkage between markers, one SNP was randomly selected from each of the 120 largest contigs of the assembly. A total of 106 of these identified SNPs were then sorted into four genotyping multiplexes, each consisting of 24-29 SNPs named BaWr-1 to −106 (see Tables S1-S4 for multiplex designs).

| Genotyping
Total genomic DNA was extracted from fin clips using the Qiagen DNeasy 96 Blood & Tissue Kit following the manufacturer's instructions (Qiagen, 2016). All samples were genotyped for the 106 SNPs on a MassARRAY ® platform using an iPLEX ® reaction (Agena Bioscience, n.d.). SNPs were amplified in 5 ml of PCR master mix consisting of 0.5 μM Multiplex primer mix (one of four multiplexes), 25 mM dNTP, 25 mM MgCl 2 , PCR buffer, 5 U/μl PCR enzyme, ~15 ng/μl DNA, and HPLC-grade water. Cycling conditions were TA B L E 1 Summary statistics of sampled ballan wrasse (Labrus bergylta) using 82 single nucleotide polymorphisms (SNPs). Lat and Lon are latitude and longitude of the sampling location, respectively; N is the number of individuals sampled; HWE and LD are number of deviations from Hardy-Weinberg equilibrium and linkage equilibrium, respectively; H e and H o are expected and observed heterozygosity, respectively; and F IS is the inbreeding coefficient. Numbers in parentheses (B) are after sequential Bonferroni corrections. Galician sample listed both as total sample and as sorted by morphotype (Plain/Spotted) at 56°C, and 1 min at 72°C, and finished with 5 min at 72°C. After initial PCR, the remaining primers and nucleotides were removed using Shrimp alkaline phosphatase (SAP) and followed by iPLEX (Agena Bioscience™) Extension Reaction, with extension primer (UEP) mix according to multiplex (see Tables S1-S4 in Appendix S1).
SAP cycling conditions were 40 min at 37°C, followed by 5 min at 85°C, while iPLEX cycling conditions were 30 s at 94°C, followed by 40 cycles consisting of 5 s at 94°C and 5 cycles of 5 s at 52°C and 5 s at 80°C. The iPLEX extension reaction was finished off by 3 min at 72°C. All thermocycler programs were set infinitely to 4°C after the cycling program was run to prevent product degradation.
Extension products were desalted using Clear Resin before transfer onto a SpectroCHIP Array by automated nanodispenser (Agena Bioscience™ RS1000). SpectroCHIP arrays were analyzed using a

MALDI-TOF Mass Spectrometer (Agena Bioscience™ MassARRAY
Dx Analyzer) that assigns genotypes to sample fragments according to mass. The results were called and control checked in TyperAnalyzer v.4.1.83 (Agena Bioscience™), and all genotypes with a mass height below 0.4 were filtered.
In addition to genotyping the entire dataset with SNPs as described above, a subset of the samples (N = 513) were also geno-

| Data analysis
Prior to any statistical analysis, extensive data quality checks were performed on the SNPs to ensure good clustering and low levels of missing data (<30%). Twenty-four of the SNPs did not pass the quality checks, leaving the final dataset to consist of 82 SNPs. Individuals missing ≥50% of genotype calls were excluded from the analysis.
Summary statistics, sample size, observed and expected heterozygosity, as well as inbreeding coefficient (F IS ) were computed using the packages adegenet and hierfstat (Goudet, 2005;Jombart, 2008) as implemented in R. Hardy-Weinberg equilibrium (HWE) tests were conducted for each sample and locus combination using the package pegas (Paradis, 2010) in R, using 1,000 Monte Carlo permutations. Statistical significance was assessed after post hoc sequential Bonferroni correction (Holm, 1979). Potential linkage disequilibrium (LD) between all locus pairs per population as well as across all populations was examined using the program GENEPOP 7 (Rousset, 2008) with 10,000 steps of dememorization, 100 batches, and 5,000 iterations per batch. As with HWE, significance was assessed after Bonferroni correction. All population structure analyses were conducted with the SNPs, while the microsatellite dataset was only used for genetic diversity analyses by calculating the number of alleles and allelic richness per sample using the hierfstat R package.  (2008)), and PCAdapt (Luu, Bazin, & Blum, 2017). Following a consensus approach, only those loci marked as under selection by at least two programs were treated as outliers. BayeScan was run by setting sample size to 10,000 and the thinning interval to 50 as suggested by Foll and Gaggiotti (2008). Loci with a posterior probability over 0.99 were retained as outliers, corresponding to a Bayes factor > 2 (i.e., "decisive selection" (Foll & Gaggiotti, 2006) (Cattell, 1966) was used to select the number of principal components (K) to identify SNPs deviating from neutrality.
Bayesian clustering analysis was carried out in STRUCTURE v.2.3.4 (Pritchard, Stephens, & Donnelly, 2000) using an admixture model and correlated allele frequency without population information. The analysis was conducted using the program ParallelStructure which speeds up analytical time by dispersing runs onto multiple processors in tandem (Besnier & Glover, 2013).
Ten runs with a burn-in of 100,000 and 1,000,000 Markov chain Monte Carlo (MCMC) iterations were performed for K = 1-5. To determine the most likely number of genetic clusters within the data, STRUCTURE output was analyzed using two approaches.
The ad hoc summary statistic ΔK and mean LnP(K) of Evanno, Regnaut, and Goudet (2005) were calculated; in addition, four alternative statistics (MedMed, MedMean, MaxMed, and MaxMean) were estimated using StructureSelector (Li & Liu, 2018). Finally, the ten iterations of the selected Ks were averaged with CLUMPP v.1.1.1 (Jakobsson & Rosenberg, 2007) using the FullSearch algorithm and the G' pairwise matrix similarity statistic and graphically displayed using bar plots. Genetic clustering was also visualized using discriminant analysis of principal components (DAPC; Jombart, Devillard, and Balloux (2010)) in adegenet, with grouping based on sampling locations. The number of principal components kept was determined from the number of PCAs needed to explain more than 80% of the variation. Genetic structure was assessed both using the total suite of SNP markers and using only the putative neutral loci ( Figures S6 and S7). Additionally, STRUCTURE was also run separately for each location with phenotype sorted samples at K = 2. Pairwise F ST matrices (Nei, 1987)  The reference cline was built using STRUCTURE Q-score for the total dataset and, in both cases, the best cline model was decided upon AIC scores. Clines were considered significantly displaced if the two log-likelihood unit support limits of the cline center did not overlap with the STRUCTURE Q-score (Qb = 1 − Qs).
Isolation-by-distance (IBD) analysis was computed for the entire Scandinavian dataset as well as within the clusters detected by pairwise F ST and STRUCTURE conducted using the R packages adegenet and ade4 (Dray & Dufour, 2007). Nei's genetic distance was calculated by adegenet, while geographic distance was approximated by following the coastline in Google Maps.

| Outlier tests
The three approaches implemented to detect outlier SNPs pro- Pairwise F ST analysis, STRUCTURE analysis, and DAPC were conducted both with and without outliers and because only marginal differences in the results were observed between the two sets of loci ( Figure S2), all markers were retained for subsequent analyses.

| Genetic differentiation and allele frequency patterns
Genetic differentiation among the Scandinavian samples was small to moderate with pairwise F ST values ranging from 0.001 to 0.041 for the 82 SNPs (Table 3). Two distinct significantly differentiated genetic clusters were observed representing NW and SE Scandinavia (Table 3). An exception to this was the NW sample Analysis of IBD showed that genetic differentiation (F ST /1−F ST ) increased with increasing distance through the entire Scandinavian study area (Figure 4a). A week IBD relationship was also observed within NW (Figure 4b), while no such relationship could be found within the SE cluster (Figure 4c).

| Cline analysis
The reference cline based on the STRUCTURE Q-score fitted a fixL Gothenburg. In contrast, the remaining 36 SNPs showed clines with centers ranging between 295 and 1,386 km from the starting point (see Figure S3 and Table S5 for full details). Seven of the cline centers were found to overlap with the reference cline ( Figure S3) with one of these markers, BaWr-46, identified as an outlier putatively under selection within Scandinavia.
F I G U R E 5 STRUCTURE bar plots (82 SNPS; K = 3) for ballan wrasse (Labrus bergylta) sampled at 19 locations in Scandinavia and Galicia, Spain. Vertical bars represent individuals, and coloration indicates assignment probability to each of the three genetic clusters. STRUCTURE analysis was also conducted without three outlier loci with no change in overall clustering ( Figure S1 in Appendix S1) Shading of DA eigenvalues indicates DAs kept. DAPCs were also conducted without three outlier loci with little change in clustering ( Figure  S3 in Appendix S1) F I G U R E 7 Geographic cline analysis for ballan wrasse (Labrus bergylta) across a 1,500-km transect along the Scandinavian coast from Sauhestøya in Trøndelag, Norway, to Gothenburg, Sweden. Cline fitting the STRUCTURE Q-score, with the center of the cline depicted by a vertical dashed line and the width highlighted in green

| Genetic differentiation and Bayesian clustering of phenotype sorted samples
Moderately large and highly significant genetic differentiation was observed between the sympatric spotty and plain wrasse sampled in Galicia, as revealed by the 82 SNPs (pairwise F ST = 0.025, p < .001; Table 3). This was also reflected in the DAPC (Figure 6a). In stark contrast however, no genetic differentiation was revealed among  Table S6).
Of the STRUCTURE runs (K = 2) for each of the phenotype sorted samples, only plain and spotty Galician phenotypes showed distinct clustering when genotyped for SNPs (Figure 8). For Scandinavian samples, no separation of phenotype was observed ( Figure S4).

| D ISCUSS I ON
This study presents the most extensive analysis of population genetic structure in ballan wrasse thus far. We identified two distinct genetic clusters splitting populations in NW and SE Scandinavia and discovered very large genetic differences between all Scandinavian and Galician samples. The split in Scandinavia was located in southwestern Norway across a region that also includes sandy beaches, representing a habitat discontinuity for this species. Finally, sympatric plain and spotty phenotypes were genetically distinct in Galicia, but not in Scandinavia.

| Genetic break in Scandinavia
A distinct break in population genetic structure was observed in Scandinavia, splitting ballan wrasse into NW and SE clusters. This break is located on the coast of southern Rogaland, in southwestern Norway, an area of coastline defined by the longest continuous stretch of sandy bottom substrate along the Norwegian coast (the Jaeren beaches; ~26 km; Blanco Gonzalez, Knutsen, and Jorde (2016)). This sandy bottom would function as a habitat discontinuity for this species and is likely to be a major reason for the observed genetic break. Ballan wrasse have a strong preference for rocky habitats, and hard-bottom substrates are necessary for them to lay their sticky benthic eggs (Darwall et al., 1992;Quignard & Pras, 1986;Villegas-Ríos, Alós, et al., 2013). Genetic breaks in fish populations associated with discontinuities in rocky substrates by sandy areas have been described previously (e.g., Bernardi (2000), Riginos and Nachman (2001)). In particular, a similar break in the population genetic structure of the corkwing wrasse has also been observed at this exact location in Norway (Blanco Gonzalez et al., 2016;Faust et al., 2018;Mattingsdal et al., 2020), as well as in the hard-bottom associated kelp Laminaria hyperborea (Evankow et al., 2019). It should be noted that for corkwing wrasse, the genetic break across the beaches of Jaeren aligns with considerable differences in life-history traits and has important implications for management (Halvorsen et al., 2016). Therefore, investigations of life-history traits in ballan wrasse could yield a deeper understanding of the biological consequences of the observed genetic barrier.
A result that should be noted is linked to the sample from Årdal in the NW cluster. This sample was not genetically distinct from any of the other Scandinavian sampling locations. Whether this is caused by Årdal being a very small sample (i.e., by chance), close to the center of the break between the two main genetic clusters, or whether this is due to aquaculture practice is not known and cannot be determined based on the current dataset. Future studies would benefit from including multiple other samples from areas close to the genetic break, and new markers designed to distinguish between the genetic clusters identified here and potential hybridization between them.

| Colonization of Scandinavia
The majority of the North Sea was covered by an ice sheet during the last glacial maximum (LGM;Lambeck, 1995;Shennan et al., 2000).
F I G U R E 8 STRUCTURE bar plot (82 SNPs; K = 2) of Galician ballan wrasse (Labrus bergylta) phenotypes including example photographs of plain and spotted fish  (Almada et al., 2017;Almada, Almada, Francisco, Castilho, & Robalo, 2012). In addition, multiple possible glacial refugia have been proposed for other marine species, such as around the British Isles (Almada et al., 2012) and even in some parts of the North Sea (Gysels, Hellemans, Pampoulie, & Volckaert, 2004;Robalo et al., 2011). The observed differences in genetic variation and degree of genetic differentiation between ballan wrasse in different parts of Europe may therefore in part be caused by recolonization from different glacial refugia as proposed by Almada et al. (2017).
A decrease in mtDNA nucleotide and haplotype diversity was found by D'Arcy, Mirimin, and FitzGerald (2013)  with observations from corkwing wrasse (Knutsen et al., 2013;Robalo et al., 2011) that has a similar ecology and biology to ballan wrasse.
As ballan wrasse is a coastal species that displays a strong affiliation to rocky substrates, colonization of Scandinavia from southern latitudes would most likely have followed the coastline northward.
This suggests that colonization could have followed the coastline of mainland Europe to Scandinavia and northward, or alternatively, via the British Isles and across the North Sea to western Norway and thereafter both northward and southward in Scandinavia. The latter has been proposed as the main route of colonization of Scandinavia for the corkwing wrasse (Knutsen et al., 2013;Mattingsdal et al., 2020). Our data also suggest this to be most likely for ballan wrasse as we observed greater genetic differentiation between SE Scandinavia and Galicia than between NW Scandinavia and Galicia.
This also fits better with climatological records which show that western Norway was probably available prior to Skagerrak and other areas in eastern Scandinavia (Stroeven et al., 2016), and geological records, showing that the English channel and large areas of the North Sea was dry land during and immediately following the LGM (Lambeck, 1995;Shennan et al., 2000). In order to clarify this, future studies of ballan wrasse need to encompass samples from the entire distribution range, especially the British Isles and Denmark to Atlantic France.

| Phenotypic and genetic variation
In Galicia, sympatric spotty and plain phenotypes have been shown to display overlapping yet differing life-history strategies, with plain fish investing more in reproduction while spotty fish invest more in growth (Villegas-Ríos, Alonso-Fernández, Fabeiro, et al., 2013).
These observations indicate that the two phenotypes represent different genetic groups that have even been suggested to be potentially different subspecies (Quintela et al., 2016;Villegas-Ríos, 2013). Quintela et al. (2016) reported a weak but significant genetic differentiation between sympatric plain and spotty phenotypes of ballan wrasse in Galicia using microsatellites. The results of our study have expanded upon this previous work by analyzing phenotyped samples from multiple locations with a panel of SNPs developed here.
For the Galician samples, the 82 SNPs revealed clear genetic differentiation between both phenotypes (F ST = 0.025; Table 3). As such, the present study provides the most compelling evidence yet that in the southern part of its distribution range, spotty and plain ballan wrasse represent distinct groups. However, and in stark contrast, no genetic differentiation was detected between the sympatric spotty and plain phenotypes in any of the four Scandinavian samples where this was investigated (see Table 1 for locations).
Previous authors have suggested that the two sympatric phenotypes in Galicia could be maintained through assortative mating (Villegas-Ríos, Alonso-Fernández, Domínguez-Petit, et al., 2013). It is also possible that other mechanisms are at play; for example, the two phenotypes may spawn at subtly different times or areas or that the two phenotypes have selective advantages in different parts of the environment. Whatever the underlying mechanisms, a significant question is why they are at play in the southern range of this species, but not in Scandinavia? Spot patterns in fish may be both environmentally and genetically determined (Jørgensen et al., 2018;Kause, Ritola, Paananen, Eskelinen, & Mantysaari, 2003;Skaala & Jørstad, 1987), and as such, differ-  (Jørgensen et al., 2018), could be used to disentangle this issue.

| Management implications
The management of wrasse in Scandinavia is currently facing two major challenges, both linked to their extensive use as cleaner fish to delouse farmed salmonids in sea-cages. These challenges include the sustainable harvest of wild populations to provide cleaner fish for the aquaculture industry, and the potential aquaculture-driven inadvertent translocation of wrasse from southern Norway and Sweden to western and middle Norway. Results from the present study are highly relevant in order to address both these issues.
We observed a distinct genetic difference between ballan wrasse in SE and NW Scandinavia. Therefore, our data clearly demonstrate that the inadvertent translocation of wrasse, via current aquaculture practice, from Sweden and southern Norway to western and middle Norway, will result in mixing fish from highly distinct genetic groups (and potentially phylogenetic lineages). Therefore, this aquaculture practice requires revision in light of our data. Studies of other species of wrasse subjected to the same aquaculture-driven harvest and translocation regimes have indicated escapes and potential hybridization between translocated individuals and local populations (Faust et al., 2018;Jansson et al., 2017). However, while the genetic markers implemented here provided a clear picture of population genetic structure between NW and SE Scandinavia, they do not provide enough statistical power to accurately identify potential escaped fish or hybrids sampled in the middle part of Norway. In corkwing wrasse, several thousand SNPs were screened to develop population-diagnostic loci that could accurately identify escapees, hybrids, and backcrosses (Faust et al., 2018). We therefore suggest that a more comprehensive set of genetic markers is required to investigate the degree to which translocated ballan wrasse survive and hybridize with local wild populations.
Harvest of wild wrasses has dramatically risen in Norway and Sweden (Norwegian Directorate of Fisheries, n.d.; Rueness et al., 2019) in order to meet the aquaculture industry's demand for cleaner fish. This sudden increase in exploitation of a previously noncommercial species raises the question of sustainability and potential extinction of local populations (Halvorsen et al., 2017).
Failure to take the spatial and temporal extent of populations into account when harvesting living resources can lead to differential exploitation and potentially overexploitation (Allendorf, England, Luikart, Ritchie, & Ryman, 2008; Kerr et al., 2017). The concern that labrids, and especially ballan wrasse, are disappearing from multiple areas in Norway have already been raised both by the public and by fishermen that report having to move to new fishing locations as catches decline (Grefsrud et al., 2018). This suggests that ballan wrasse are very stationary, which has also been shown in studies of site fidelity in Galicia (Mucientes et al., 2019). The data presented here show some evidence of genetic structuring within the SE and NW Scandinavian clusters. This needs further investigation, combining biological, fishery and genetic data, in order to better understand connectivity within regions, and thus the potential for exhaustion of local populations.
In Galicia, the two phenotypic groups are already sold and marketed separately (Villegas-Ríos, Alonso-Fernández, Fabeiro, et al., 2013) and are therefore likely subject to different fishing pressures.
The findings of this study could thus be used as a scientific basis for upholding this practice, and as a basis for the future management of these two groups as separate stocks.

ACK N OWLED G M ENTS
We would like to acknowledge colleagues Kjell Nedraas and Håkon Otterå for some of the samples used here, as well as the efforts of commercial fishermen. This study was financed by the Norwegian Department for Industry and Fisheries, the Norwegian Research Council project "Ecogenome," and the Swedish Research Council FORMAS.

CO N FLI C T O F I NTE R E S T
None declared.