Interbreeding between local and translocated populations of a cleaner fish in an experimental mesocosm predicts risk of disrupted local adaptation

Abstract Translocation of organisms within or outside its native range carries the risk of modifying the community of the recipient ecosystems and induces gene flow between locally adapted populations or closely related species. In this study, we evaluated the genetic consequences of large‐scale translocation of cleaner wrasses that has become a common practice within the salmon aquaculture industry in northern Europe to combat sea lice infestation. A major concern with this practice is the potential for hybridization of escaped organisms with the local, recipient wrasse population, and thus potentially introduce exogenous alleles and breaking down coadapted gene complexes in local populations. We investigated the potential threat for such genetic introgressions in a large seminatural mesocosm basin. The experimental setting represented a simulated translocation of corkwing wrasse (Symphodus melops) that occurs on a large scale in the Norwegian salmon industry. Parentage assignment analysis of mesocosm's offspring revealed 30% (195 out of 651 offspring) interbreeding between the two populations, despite their being genetically (F ST = 0.094, p < 0.05) and phenotypically differentiated. Moreover, our results suggest that reproductive fitness of the translocated western population doubled that of the local southern population. Our results confirm that human translocations may overcome the impediments imposed by natural habitat discontinuities and urge for immediate action to manage the genetic resources of these small benthic wrasses.

ecological and genetic interactions between native and exogenous populations, remains a common topic of debate for managers and conservationists (Araki & Schmidt, 2010;Laikre et al., 2010;Thomas, 2011). Putative deleterious genetic risks, associated with large-scale releases and translocations, include the loss of genetic variation and adaptations as well as alterations in genetic profiles and population structure (Geller et al., 2010;Hӓnfling, 2007;Laikre et al., 2010).
Experimental studies assessing reproductive fitness between exogenous and locally adapted populations have drawn contradictory results. While many studies suggested lower fitness performance in nonlocal (hatchery-released, unintentionally escaped and translocated) organisms (Araki, Cooper, & Blouin, 2009;Christie, Ford, & Blouin, 2014;Eldridge & Naish, 2007;Glover et al., 2017), others have reported no fitness disadvantage (Berejikian, Van Doornik, Scheurer, & Bush, 2009;Blanco Gonzalez, Nagasawa, & Umino, 2008;Delgado & Glazer, 2007;Hess et al., 2012). Pedigree reconstruction in offspring for stocking has often revealed significant variation in the number of parental contributors and their family sizes, and consequently, in their effective population sizes and rate of inbreeding (Blanco Gonzalez, Taniguchi, & Umino, 2010;Jeong, Blanco Gonzalez, Morishima, Arai, & Umino, 2007). Currently, our knowledge on reproductive fitness of nonlocal populations is highly biased toward studies conducted on salmonids, with very little information on strictly marine fish species (Araki & Schmidt, 2010). The large population sizes and extensive geographic areas commonly occupied by most marine fish make these studies very challenging (see references in Araki & Schmidt, 2010;Blanco Gonzalez, Aritaki, Knutsen, & Taniguchi, 2015). Large-scale mesocosm facilities may offer an alternative to overcome these limitations (Blanco Gonzalez et al., 2010;Jeong et al., 2007). Although such seminatural experimental setup may not fully resemble processes taking place in the wild, it will at least provide a good indication of the spawning performance as compared to laboratory culture conditions (Leggatt et al., 2014).
Sea lice infestation is a major burden for the aquaculture industry causing high salmon mortalities and large economic losses (Blanco Gonzalez & de Boer, 2017;Iversen, 2016). The use of wrasses as cleaner fish has been proposed to be the most economical and environmental friendly solution to combat sea lice infestation (Liu & Bjelland, 2014), despite high mortalities at low temperatures (Bjelland, Simensen, & Kvenseth, 1996;Costello, 1991;Sayer, Reader, & Davenport, 1996). As result, translocation practices have been undertaken in several European countries, where wild wrasses are sex-and size-selectively fished and transported alive from areas of high abundance to salmon farms located in areas where local supply cannot cope with their high demand (Blanco Gonzalez & de Boer, 2017;Riley, Jeffery, Cochrane-dyet, White, & Ellis, 2017). In the UK, for example, up to one million wrasses, mainly ballan wrasse Labrus bergylta, are translocated from the southwest coast to Scottish farms annually (Riley et al., 2017). In Norway, the largest farmed salmon producer in the world, several millions of wild-caught adult wrasses, mainly goldsinny Ctenolabrus rupestris and corkwing wrasse Symphodus melops, from southern regions of Norway and Sweden are translocated to salmon farms located further north in the west coast (Blanco Gonzalez & de Boer, 2017). Several cleaner wrasse species display a strong regional and latitudinal variation in sexual size dimorphism (Halvorsen et al., 2016;Sayer, Gibson, & Atkinson, 1996). Human-mediated translocation may facilitate crossing the geographical boundaries delimited by natural barriers and evolutionary processes (Geller et al., 2010) and could therefore pose a major threat to local wrasse populations if a significant number of wrasses are released from the net pens unintentionally or intentionally (Espeland et al., 2010).
Recent genetic studies of cleaner fishes in northern Europe have revealed marked large-scale population genetic structure, with reduced genetic variability in northern populations (Almada et al., 2017;Blanco Gonzalez, Knutsen, & Jorde, 2016;Jansson et al., 2017;Knutsen et al., 2013;Robalo et al., 2012). The northern Scandinavian populations have shown a further pattern of isolation by distance in the goldsinny (Jansson et al., 2017) and a genetic discontinuity ("break") between southern and western Norwegian populations in the case of corkwing wrasse (Blanco Gonzalez et al., 2016). Moreover, apparent translocated individuals of goldsinny (Jansson et al., 2017) and corkwing wrasse (Faust, Halvorsen, Andersen, Knutsen, & André, 2018) have been found in the proximities of salmon farms in Norway, with indication of hybridization with locals for the latter species. Little is known about how translocated individuals fare in their new environment, their mating preferences versus local conspecifics, and potential isolation mechanisms between translocated and local wrasse populations. Hence, the risk of spreading of exogenous genes in these cleaner fishes is presently unclear, despite their large-scale deployment in the industry.
In the present study, we addressed several critical questions concerning the reproductive behavior of translocated fish: (a) Do local and translocated wild-caught marine fish interbreed? (b) Do locally adapted populations present reproductive fitness advantage over translocated individuals? (c) Do specimens display any mating preference regarding origin or phenotypic traits?

| Study species and experimental setting
Rising sea water temperatures registered in the last decades have favored the increase in abundance of corkwing wrasse in Norway (Barceló, Ciannelli, Olsen, Johannesen, & Knutsen, 2016;Knutsen et al., 2013). The species is nest-building and displays alternative reproductive tactics (Halvorsen et al., 2016;Sayer, Gibson, et al., 1996). Large nesting males build and defend nests from sneaker males (Potts, 1984). These different life-history tactics are particularly pronounced in the west coast of the Scandinavian Peninsula where nesting males show a delay in maturation time and faster growth rates compared to females and sneaker males (Halvorsen et al., 2016).
The present experimental study was conducted at the research facilities of the Institute of Marine Research at Flødevigen in Arendal, on the south coast of Norway. The mesocosm basin has a capacity of approximately 2,000 m 3 , a surface area of 660 m 2 , a maximum depth of 5 m, and the seawater is pumped up from 75 m depth (Moksness, 1982). The geographical location and suitability of these facilities for the study justify moving specimens from the west coast to the south, instead of in the opposite direction which is the one most commonly used by the salmon industry (Blanco Gonzalez & de Boer, 2017). On June 24, 2014, 167 adults from Norheimsund, on the west coast of Norway (for location in a map see Blanco Gonzalez et al., 2016), were collected using baited wrasse pots by a local fisherman and transported alive to Flødevigen. After arrival, all individuals were measured in total length (cm), weighted (g; see Table 1 for details on wrasse samples). The sex of every adult was determined by examination of the urogenital papilla (only present in females and sneaker males but not in nesting males) and inspecting an ejaculate sample of egg/sperm obtained by applying gentle pressure on the abdomen. Following these procedure, we confirmed that all individuals were sexually mature, and they were classified as nesting male, sneaker male, and female. In addition, a small piece of the tail was clipped and stored in 95% ethanol for DNA analysis before they were released into the mesocosm basin. These individuals represent the group of translocated specimens of nonlocal west origin.
On July 1, 2014, a second group of 151 adult fish were collected from Arendal, close to Flødevigen, using baited wrasse pots by a local fisherman and transported to Flødevigen. They were subjected to the same measurements and handling as the earlier sample, except that the ventral side of their bodies was tagged with pink visible implanted elastomer. These specimens comprise the group of breeders of local south origin of the study and were released into the mesocosm basin together with the western group.
The two group of wrasses were maintained at the mesocosm for a full year and allowed to spawn naturally during the upcoming spawning season in spring-summer. At the end of the spawning season, between July 30 and August 19, 2015, a total of 651 offspring were collected with small aquarium nets along the edges of the spawning basin. Offspring were weighted (g) and measured in total length (cm) after the image analysis of the digital photography taken with a Tucsen CMOS IS1000 camera (Tucsen) attached to a Leica MZ16a stereomicroscope (Leica). A small piece of the juvenile's tail was clipped and stored in 95% ethanol for DNA analysis.

| Microsatellite genotyping
Total genomic DNA was extracted from ethanol-preserved tail samples using E.Z.N.A® Tissue DNA kit (Omega Bio-Tek), resuspending the DNA in TE buffer. The ability of microsatellite loci to resolve parentage assignment depends on their number as well as their degree of polymorphism (Villanueva, Verspoor, & Visscher, 2002).
The analysis was conducted on eleven polymorphic microsatellites characterized in previous studies on corkwing wrasse: SMD121, SMA11, SMA103, SMD131, SMD110, SMD112, SMB11, SMC8, SMB101, SMC5, SMD118, SMB101, SMC5, and SMD118, following the same multiplex PCR protocols and dye labeling as previously described (Blanco Gonzalez et al., 2016;Knutsen et al., 2013;Knutsen & Sannaes, 2009). PCR amplifications were carried out in a multiplex reaction of 10 μl volume including 10 pmol of each primer and 1 μl of template DNA, corresponding to 30-50 ng. PCR conditions for the multiplex reaction of three new primers consisted of an initial denaturation step at 94°C for 5 min, followed by 35 cycles of 95°C for 30 s, annealing at 56°C for 60 s and 72°C for 60 s, with a final extension at 72°C for 15 min. One microliter of PCR product was mixed with 10 µl of Hi-Di formamide and 0.8 µl of GeneScan Tm -600 Liz (Applied Biosystems) and run on an ABI 3130XL automated sequencer. Individual genotypes were assessed with GENEMAPPER v. 4.0 (Applied Biosystems). As a guard against potential genotyping errors, all samples were run with the same size standard and on the same machine. In addition, approximately 5% of the samples were randomly subjected to repeated genotyping.  Nei and Chesser (1983), using FSTAT v.2.9.3.2. (Goudet, 1995)  Note: Sample ID, origin, collection date, tag color, sex, sample size, total length range (mean ± standard deviation, SD) in cm, and body weight range (mean ± standard deviation, SD) in g.
The discrimination power of the set of microsatellite loci for parentage analysis was determined by the polymorphic information content (PIC) and the exclusion probability (Q) with CERVUS v.3.0 (Kalinowski et al., 2007). We also used GenAlEx (Peakall & Smouse, 2012) to estimate the probability of identity index (I), an index representing the probability of finding two individuals sharing a multilocus genotype. In addition, we determined the cumulative success rate of parentage allocation ranking the markers according to the exclusion probability of both parents, Excl. P2 option, based on 1,000 offspring simulations using the PFX_Mchoice macro implemented in PARFEX v1.0 (Sekino and Kakehi 2011).

Genetic differentiation among samples was conducted by
Wright's F ST , using Weir and Cockerham's (1984) estimator θ applied to all samples and between pairs of samples. The statistical significance of p values was examined by G tests in GENEPOP package v.4.7.0. (Rousset, 2008) with 100,000 demorization steps, 1,000 batches, and 50,000 iterations per batch. The FDR approach proposed by Benjamini and Yukutieli (2001) was adopted to correct for multiple tests in pairwise tables.

| Pedigree reconstruction and effective number of breeders
Parentage assignment was performed in PAPA v.2.0 (Duchesne, Godbout, & Bernatchez, 2002) and corroborated in CERVUS v.3.0 (Kalinowski et al., 2007). Pedigree reconstruction in PAPA v.2.0 requires a closed system where all putative breeders have been sampled (Duchesne et al., 2002) as was the case in the present study. The analysis was performed with a uniform error of 0.02 on all loci. The assignment analysis in CERVUS is based on the loglikelihood (LOD) score, inferring parental pairs to those breeders with the highest likelihood. The analysis was conducted considering known broodstock sexes. Allocation was considered correct only when trio (offspring and a parental pair) showed no mismatch at any locus, and PAPA and CERVUS showed consistent parental pair assignments. In both programs, prior to offspring assignment, parentage estimates were simulated for 10,000 offspring.
In order to minimize any genotyping errors and ensure a reliable parental-offspring assignment, we examined those offspring not allocated to any parental pair and those allocated to more than one pair (called as "ambiguous" in PAPA). A new assignment test was performed allowing a maximum mismatch at two loci. Allele scoring for trios was revised at all loci and corrected accordingly.
Parental assignment of the few remaining offspring was successfully resolved by repeated genotyping at all loci and re-running the analysis. Based on these results, offspring were classified as "south" when both parents were of south origin, "west" when both parents were from the west coast, or "hybrid" when parents were of different origins.
Once all parental pairs were identified, we estimated the inbreeding effective number of breeders, Nb (Araki, Waples, Ardren, Cooper, & Blouin, 2007;Crow & Kimura, 1970;Waples, 2002), with and without taking parental origin into consideration. This was calculated from the Nb (N), the average number of offspring (k), and the variance in the number of offspring (V k ) from contributing breeders, as assessed from the parental assignments (above): This was calculated separately for male (m) and female (f) parents and combined to obtain the inbreeding effective Nb: We further investigated whether the presence of sneaker males would contribute to increase Nb or not. The analysis was performed by comparing the estimates when offspring from all breeding pairs were included (regardless nesting or sneaking male behavior) to those obtained excluding the offspring produced by sneaker males.

| Assortative mating and reproductive fitness
We used parentage assignment results to test the hypothesis of nonrandom mating between south and west breeders using a 2 × 2 contingency table with Yates correction. The analysis was performed taking into account parental origin (south and west) and sex (female and male). In this analysis, nesting and sneaker males were pooled together due to the small number of sneakers in the samples (see Table 1 for details).
Results of the parentage assignment test were also used to analyze causes of variance in reproductive fitness. For each possible breeding pair, we considered whether offspring were produced or not, parental origin (south and west), sex (nesting males, females and sneaker males), and whether both parents were from same origin or not. Furthermore, we also considered the effects of parental weight/ length ratio (both for males and females) to analyze reproduction success. Since most possible male-female pair combinations did not produce offspring (243 identified

| Genetic diversity and differentiation
We scored the complete genotypes of 318 adults and 651 offspring at eleven microsatellite markers with no missing genotypes. A total of 193 alleles were scored, and overall total gene diversity, H T, was 0.750 ( Deviation from HW expectations (F IS ) was observed in 10 of the 3*11 cases generated from the two parental groups and the offspring at 11 loci, with eight of them remaining statistically significant at the 5% level after the FDR correction ( Over 30% (41 of 135) of pairwise tests for LD were statistically significant after the FDR correction (at the 5% level, data not shown).
Except for the locus pair SMD121-SMB101 in the breeders from the west coast, all significant pairwise tests were detected in the offspring samples. As deviations from HW and from LD are expected in the heterogenous offspring group, representing a mix of parental stocks and hybrids, no action was taken in response to these findings, except as noted above for locus SMB118.

| Pedigree reconstruction and effective number of breeders
High PIC and Q estimates, and low I index, were observed for the set of microsatellite loci ( TA B L E 2 Summary statistics of genetic variability for the parental groups of south and west origin, offspring, and the total among samples at 11 microsatellite markers Note: Bold values indicate significant p values at 5% level after the false discovery rate approach (Benjamini & Hochberg, 1995).  (Table 4). The largest family identified in this study comprised 31 offspring, and it was assigned to the west origin pair M300-F403.
The parental pair M075-F276 contributed to the largest family of south origin with just six offspring ( Figure 1). Meanwhile, the largest hybrid family comprised 12 offspring from the pair M435-F219.
These large differences in number of offspring resulted in high variance in family sizes (V k ), up to more than 2,000 for western nesting males (Table 5) and, together with skewed contribution of the two sexes, resulted in low inbreeding effective Nb of only 17.6 for the total (west + south) spawning population when sneaker males were not considered. Although none of the sneaker males of south origin contributed to the offspring, the participation of 17 sneaker males from the west coast reduced the variance in family size and resulted in an increase in Nb to 22.9 (Table 5).

| Assortative mating and reproductive fitness
Mating between corkwing wrasse breeders of south and west origin appears to occur randomly with no evidence for assortative mating by population origin (χ 2 = 0.27, p = 0.60 after Yates correction df = 1).

| D ISCUSS I ON
This study represents one of the few examples evidencing interbreeding between native and translocated wild populations of non-salmonid marine fish species (Hӓnfling, 2007;Swan et al., 2016). The parental analysis conducted here, in addition to resolving the pedigree of two corkwing wrasse population from south and west Norway, evidenced successful mating between two genetically distinct cleaner fish populations with no evidence for nonrandom mating between them. Thus, there appears to be no intrinsic mechanism against interbreeding and  (Araki & Schmidt, 2010;Geller et al., 2010;Laikre et al., 2010).
Paternal care marine fish species, such as corkwing wrasse, often display territorial behavior during the spawning season when large nesting males show aggressive behavior to guard their nests against small sneakers (Myhre, Forsgren, & Amundsen, 2012;Potts, 1984;Sinopoli et al., 2018;Stiver et al., 2018). At high population densities, frequent aggressive encounters may interrupt courtship interactions and reduce the occurrence of mating episodes (Myhre et al., 2012).
Despite the considerable size of the mesocosm basin (2,000 m 3 ), such aggressive encounters between males have likely occurred in the present experimental setting (Halvorsen et al., 2017). Moreover, high population densities may induce higher variance in mating success (Aronsen, Berglund, Mobley, Ratikainen, & Rosenqvist, 2013;Stiver et al., 2018) and alter relative fitness performance (Kokko & Rankin, 2006). The significantly larger contribution of breeders of west origin compared to those from the south (Table 4) suggests putative fitness advantage in western populations, who are typically found at higher densities in nature (Halvorsen et al., 2016). The fact that all breeders employed in this study were wild fish without prior experience in captivity may have offset any disadvantage related to domestication selection (Araki et al., 2009;Christie et al., 2014).
The offspring evidenced significant differences in their genetic profile compared to either of the two putative parental populations (Table 4). How these differences may affect their fitness parameters and impact local adaptation is yet to be determined. Although the offspring genotyped here were randomly collected from the mesocosm basin over a 3-week period, they only represent a portion of the total offspring born in the basin and we cannot dismiss the possibility of some unintentional sampling bias in the offspring. TA B L E 5 Census of breeders contributing to the offspring (N), average number of offspring (k), variance in reproductive success (V k ), inbreeding effective number of breeders after accounting for variance in family sizes (Nb f,m ) and also including sex ratio (Nb) for each sample origin separately and for the whole dataset Intraspecific latitudinal variance in life-history traits generally is strongly correlated to environmental gradients (Munch & Salinas, 2009). Corkwing wrasse populations display strong regional and latitudinal variation in sexual size dimorphism (Halvorsen et al., 2016;Sayer, Gibson, et al., 1996). Populations inhabiting colder areas along the west coast of Norway appeared to present higher proportion of sneaker males, slower growth rates, and delayed maturation (Halvorsen et al., 2016). Despite such differences in lifehistory traits, we found no evidence for assortative mating in the mesocosm. The presence of sneaker males had positive effects on offspring production (Table 5 and 6) and the effective Nb (Table 7).
These results support previous studies suggesting an important role of sneaker males in sexual size dimorphism and for reproductive suc- in such situations. Similarly to corkwing wrasse, intraspecific geographic variance in genetic and phenotypic traits has been also reported on goldsinny and ballan wrasse, the other two main cleaner wrasses used by the salmon industry (Halvorsen et al., 2016;Jansson et al., 2017;Leclercq, Grant, Davie, & Migaud, 2014;Sayer, Gibson, et al., 1996). Hence, considering that millions of cleaner wrasses are annually translocated in Norway (Iversen, 2016) and the UK (Riley et al., 2017) and released inadvertently and intentionally when the net pens are emptied (Blanco Gonzalez & de Boer, 2017), our results should bring awareness of the putative threat pose by wrasse translocations on fitness performance and the long-term evolutionary potential of recipient populations (Araki, Cooper, & Blouin, 2007;Araki et al., 2009;Eldridge & Naish, 2007;Glover et al., 2017;Laikre et al., 2010).

ACK N OWLED G M ENTS
This work was funded under the Havkyst program of the Norwegian Research Council (Project #234328/MO "Adaptation or plasticity as response to large scale translocations and harvesting over a climatic gradient in the marine ecosystem?"). The publication charges for this article have been funded by a grant from the publication fund of UiT The Arctic University of Norway. We would like thank Mana Naito and Laia Bono for their help during sample collection and DNA extraction. We would also like to thank Stian Stiansen for his assistance with the maintenance of the mesocosm basin. We are also grateful to three anonymous reviewers for their thoughtful comments on an earlier version of the manuscript.

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

AUTH O R CO NTR I B UTI O N S
EBG designed the study, conducted the sampling, and performed the microsatellite analyses. EBG, SHE, and PEJ analyzed the data. EBG wrote the original draft with contribution from all the co-authors.

DATA AVA I L A B I L I T Y
Microsatellite data of every sample genotyped in this study are