Functional genetic diversity of domestic and wild American mink (Neovison vison)

Abstract The release of domestic organisms to the wild threatens biodiversity because the introduction of domestic genes through interbreeding can negatively impact wild conspecifics via outbreeding depression. In North America, farmed American mink (Neovison vison) frequently escape captivity, yet the impact of these events on functional genetic diversity of wild mink populations is unclear. We characterized domestic and wild mink in Ontario at 17 trinucleotide microsatellites located in functional genes thought to be associated with traits affected by domestication. We found low functional genetic diversity in both mink types, as only four of 17 genes were variable, yet allele frequencies varied widely between captive and wild populations. To determine whether allele frequencies of wild populations were affected by geographic location, we performed redundancy analysis and spatial analysis of principal components on three polymorphic loci (AR, ATN1 and IGF‐1). We found evidence to suggest domestic release events are affecting the functional genetic diversity of wild mink, as sPCA showed clear distinctions between wild individuals near mink farms and those located in areas without mink farms. This is further substantiated through RDA, where spatial location was associated with genetic variation of AR, ATN1 and IGF1.

repeatedly escape and interbreed with native populations, as reported in Ontario with domestic mink, this could result in the introgression of maladaptive alleles into the wild which may lower the fitness of the native population (Bowman, Kidd, Gorman, & Schulte-Hostedde, 2007;Hansen, 2002;Hindar, Ryman, & Utter, 1991).
To assess whether functional gene complexes of domestic mink are compromising those of the introgressant mink populations, we first sought to characterize functional genetic diversity in wild and domestic American mink using microsatellites found in coding regions of the genome and in promoter regions that are noncoding but may be linked to functionally important genes. Sequence polymorphisms in coding regions are responsible for the variation in traits exhibited among individuals-variation that can be targeted by artificial selection and other genetic mechanisms (Borštnik & Pumpernik, 2002). Areas under artificial selection are especially interesting as we expect to see the greatest genetic difference between wild and domestic populations at these loci. With samples from multiple farm and wild sites across Ontario, our first objective was to determine whether trinucleotide microsatellite markers of identical repeat size differed in sequence. Slipstrand mispairing (SSM), a mutation process that results in the addition or deletion of a tandem repeat unit, occurs both in the farm and in the wild (Fan & Chu, 2007;Levinson & Gutman, 1987). SSM is known to cause frameshift mutations, which can change the gene function and thus impact survival in the wild (Fan & Chu, 2007). However, because selection against new mutations causing a change in amino acid repeat sequence or length may be less constrained in captivity, signatures of SSM might be more prevalent in domestic mink. Therefore, we expect to see sequence differences in alleles of the same size between domestic and wild mink.
Secondly, we sought to assess how domestication has affected functional genetic variability in American mink by genotyping individuals at loci thought to be targeted by artificial selection. Studies have shown that domestication can lower genetic diversity in neutral loci within breeding lines; however, little is known about how it affects the genetic diversity of functional genes (Belliveau et al., 1999). Domestication is driven by intense artificial selection which can increase the frequency of an allele until it is fixed in the population, and given American mink undergo intense artificial selection (e.g. for large size), we are likely to see a small range of alleles associated with these traits. In addition to artificial selection, the main genetic mechanisms influencing the domestication of mink are as follows: natural selection, relaxation of natural selection, inbreeding and genetic drift. The effects of domestication on genetic diversity vary depending on the selection pressure placed on different traits.
For example, natural and artificial selection, inbreeding and genetic drift are thought to reduce genetic variability, whereas relaxed natural selection is thought to increase genetic variability through the production of new alleles (Price, 1984). Therefore, the amount of functional genetic variability in farmed mink relative to wild mink should differ by trait and the conditions under which that trait is maintained on each individual farm.
Our third objective was to determine how functional genetic diversity of wild mink populations is affected when domestic mink are introduced. Captive American mink are continuously being released and escaping to the wild (Joergensen, 1985). Bowman et al. (2007) estimated that 38% of free-ranging mink harvested per province per year in Canada were of domestic origin. Furthermore, domestic escapees are persisting within wild mink populations, with one sampled wild population containing only 22% wild individuals (Kidd et al., 2009). Backcrossing of domestic-wild hybrid mink occurs in both directions, confirming that domestic alleles at neutral loci are being introgressed into wild mink populations (Kidd et al., 2009). We hypothesized that the phenotypes unique to captive mink are maladaptive to life in the wild; therefore, alleles common in captive mink are selected against in the wild and consequently occur in wild mink populations at relatively low frequencies. We consider that alleles common in domestic mink confer low fitness in wild mink and therefore should only occur in close proximity to mink farms as a result of recent introgression. We quantified this by testing genetic variation against mink farm density per township using redundancy analysis and spatial analysis of principal components.  Figure 1). Captive mink (n = 143) were donated by three farms in Ontario and a pelting service in Nova Scotia. Samples were separated into five groups: Nova Scotia domestic black (NSB), Ontario domestic black (ONB), Ontario domestic brown (OND), wild (Wild) and free-ranging Ontario mink of mixed origin that cannot be classified as either fully domestic or fully wild (Hybrid) based on Bayesian assignment tests carried out in a previous study (Kidd et al., 2009). Individuals with a mean membership coefficient q ≥ 0.8 were assigned as wild, those with q ≤ 0.2 were domestic, and the remaining (0.2 > q < 0.8) were considered hybrid though they may have been backcrossed to domestic or wild groups (Kidd et al., 2009). DNA was extracted from spleen and liver using a DNeasy Blood & Tissue Kit (Qiagen) and quantified using a Quant-iT Picogreen ® dsDNA Assay Kit (Invitrogen™) on the BMG FLUOstar microplate-reader system (BMG-LabTech).

| Microsatellite genotyping
Genes were selected for their putative adaptive potential. Genes related to functions such as behaviour, development and stress were targeted because they can be artificially selected for by mink ranchers and consequently may show variation between captive and wild mink. We downloaded DNA sequences from the National Center for Biotechnology Information (NCBI) and designed primers (n = 38) around regions containing 6 + repeat units using Geneious 9.0.5 (Kearse et al., 2012). Primers were manufactured by Integrated DNA Technologies. Primers were tested in a PCR cocktail consisting 1 mM PCR buffer (10×), 2 mM MgCl 2 , 0.2 mM dNTPs, 0.1 mg/ml BSA, 0.2 µM each forward and reverse primer, 0.04 U Taq polymerase and 5 ng DNA for a total reaction volume of 8 µl/sample. Thermal-cycling conditions consisted of the following: 95°C for 10 min, 94°C for 30 s, 55°C for 1 min and 72°C for 1 min; 30 cycles of 94°C for 30 s, 50-65°C for 1 min and 72°C for 1 min, followed by 65°C for 15 min.
Products were visualized on a QIAxcel capillary electrophoresis system (Qiagen) using a DNA Screening Kit. Primers that successfully amplified their target (n = 17, Table A1 in Appendix) were then tested for polymorphisms using a subset of samples that was comprised of wild and multiple lines of captive American mink, both males and females, in order to represent the complete sample set and encompass multiple levels of variation (between sex, colour and origin [domestic vs. wild]). The same PCR cocktail was prepared, this time containing a fluorescently labelled forward primer, and amplified with an annealing temperature of 55-61°C depending on the primer (Table 2).
New primers were fluorescently labelled with either HEX or 6-FAM (Integrated DNA Technologies, Inc.) or NED (Applied Biosystems).
A total of 291 American mink were profiled at AR, ATN1, IGF1 and TOB1. Individuals (n = 8) that could be scored at fewer than 70% of the target loci were eliminated from further analysis. We tested for Hardy-Weinberg equilibrium (HWE) and linkage disequilibrium (LE) with GENEPOP 4.5.1 using Markov chain parameters of 5,000 iterations with 10 repetitions (Rousset, 2008). Effective number of alleles was determined using an allelic richness test in Fstat 2.9.3.2 (Goudet, 1995). Population-specific variations in allele distributions were tested with a principal components analysis (PCA) in RStudio 3.3.0 using the adegenet package (Jombart, 2008). Also analysed were 15 neutral microsatellite loci previously used to assess population genetic structure and diversity in the same individuals (Tables A2 and A3

| Apportioning genetic variation to environmental variables
Redundancy analysis (RDA) was used to identify genetic variation significantly associated with environmental gradients. Using the rda function of the "vegan" package in RStudio 3.5.2, the wild and hybrid genetic data sets were used as the response matrix Y, and a set of anthropogenic and spatial variables (mink farm density, road density, latitude and longitude) was used as the explanatory matrix X (Oksanen  interested in potential north-south or east-west gradients in alleles of domestic origin, and therefore, latitude and longitude were included as a measure of location. Latitude has previously shown to be a good predictor of a free-ranging mink's probability of being classified as domestic based on neutral markers (Beauclerc et al., 2013). To obtain mink farm density values, we used ArcMap 10.1 to determine the township each sample was associated with and then divided the number of mink farms in that township by the township area (km 2 ; Statistics Canada, 2006). Similarly, road density values were obtained by dividing the sum of the lengths of all the roads (km) in the township by the township area (km 2 ). RDA was completed for each locus on 159 individuals, and an additional RDA was performed on the same individuals using the entire neutral data set. Outlier alleles were identified using the test statistic Mahalanobis Distance D, which was computed using the covRob function of the "robust" package in R (Capblancq, Luu, Blum, & Bazin, 2018;Wang et al., 2017). Then, we performed a second RDA with only outlier alleles (q-values < 0.1) to identify the environmental variables that are most correlated with the genetic variation of those alleles. Next, to determine whether allele frequencies varied by location, we evaluated the principal components of each functional locus and neutral data set over geographic space using spatial analysis of principal components (sPCA) using the "adegenet" package (Jombart, 2008).

| Microsatellite genotyping
Of the 38 primer pairs tested, 17 amplified a single amplicon at the size expected with no evidence of nonspecific amplification or smearing (Tables A1 and A4 in Appendix). Of these 17 loci, only four displayed length polymorphisms: androgen receptor (AR), atrophin 1 (ATN1), insulin-like growth factor 1 (IGF1) and transducer of ERBB2 (TOB1). A total of 287 American mink were profiled at AR, ATN1, IGF1 and TOB1 resulting in 19 alleles and a mean of 4 alleles per locus (range 2-8; Tables 2 and 3).
We profiled the AR trinucleotide repeat in 135 females and 145 males and found six alleles (range 289-304  black mink ellipse overlapped with wild mink (Figure 2; Table A5 in Appendix). The wild mink population deviated from Hardy-Weinberg equilibrium following Bonferroni correction (Table 4). We profiled the TOB1 trinucleotide repeat in 220 individuals and found two alleles (242 and 245). The frequency of allele 245 was significantly different between populations at Bonferroni-corrected significance thresholds. Allele 245 was more commonly found than 242 in all mink types except for domesticated black mink in Ontario, where both alleles were found to occur in equal frequency (Table 3). Principal components of TOB1 suggest moderate allele frequency similarity between all populations except for between OND and wild mink (Figure 2; Table A5 in Appendix). We found no evidence for deviation from LE nor from HWE at Bonferroni-corrected significance thresholds for each population.
American mink (n = 280) were profiled at 10-15 polymorphic neutral markers, which produced a total of 163 alleles and a mean of 11

| Microsatellite sequencing analysis
Androgen receptor consisted of three repeat sections: a perfect repeat ("GCA"), followed by an imperfect repeat ("GGA GAC CAG TTC TCG") and a second perfect repeat ("GCA"). Allele size variation was due to the number of repeats present in the first perfect ("GCA") repeat section. For atrophin 1, a compound microsatellite containing ("CAG") as well as ("CAA") repeats, the size difference amongst alleles was due to the number of ("CAG") repeats present. No nonsynonymous sequence differences between alleles of the same size were discovered.

| Apportioning genetic variation to environmental variables
Only a small proportion of genetic variation in each functional locus was affected by the environmental variables we tested (AR R 2 adj = 0.0235, ATN1 R 2 adj = 0.0114 and IGF1 R 2 adj = 0.0271; Table 5). Longitude was a significant influence on genetic variation in AR (R 2 = .0413) and ATN1 (R 2 = .0211). With IGF1, it appeared that latitude was correlated with RDA1 ( Figure 3). The follow-up RDA performed on the three outlier alleles detected in IGF1 (103, 105 and 111) revealed that latitude did have an impact on genetic variation in IGF1 (R 2 = .0157), which was especially evident in RDA1, the axis that held most of the genetic variance (70%; Figure 4). The environmental variables had a small but significant effect on neutral loci (R 2 adj = .0411); here, latitude, longitude and road density all explained significant genetic variation (R 2 = .0205; Table 5). We performed RDA on 35 outlier alleles detected in the neutral data set and found that most of the variance was explained by the first two axes (51% and 30%, respectively). Both axes were highly correlated with spatial location: RDA1 was correlated with latitude and RDA2 with longitude ( Figure 4). Spatial analysis of principal components on neutral loci demonstrated a latitudinal gradient in genetic variation between individuals in northern and southern Ontario. The same pattern was observed in functional loci, though the latitude at which the northsouth transition occurred varied by locus ( Figure 5).

| D ISCUSS I ON
Contrary to our prediction that slip-strand mispairing causes sequence changes in domestic mink alleles, sequencing results from AR and ATN1 show no nonsynonymous change in repeat structure between alleles of the same size nor between alleles of varying size. However, it appears that domestication has affected the functional genetic diversity of captive mink populations. Captive mink displayed fewer alleles per locus than the wild population at AR, ATN1 and IGF1. In fact, allele frequencies were so different between captive and wild populations, that domestic release events affected functional genetic diversity of wild mink. As predicted, sPCA showed clear distinctions between wild individuals near mink farms and those located in areas without mink farms. This is further substantiated through RDA, where spatial location was associated with genetic variation of AR, ATN1 and IGF1.

| Genetic diversity of wild and domestic American mink
Only four of 17 amplified loci displayed length polymorphisms: AR, ATN1, IGF1 and TOB; the remaining loci did not show variation within or between mink types. This low proportion of polymorphic loci is not unlike in some other mustelids. For example, a study on functional trinucleotide motifs in genes in fisher (Pekania pennanti) reported only 27% polymorphic loci (Greenhorn, 2016). American   (Bolton et al., 2007). AR is involved in male sexual development and has been linked to aggressive behaviour (Butovskaya et al., 2015;Hurd, Vaillancourt, & Dinsdale, 2010).
We speculate the significant frequency difference in AR allele 298 between wild and domestic mink is due to selective breeding for lower aggression level. Multiple studies report that when breeding mink for domesticated behaviour, aggressive reaction to humans varied from generation to generation, but showed a general decline throughout the selection period (Kizhina et al., 2017;Kulikov et al., 2016). It is likely that in choosing less aggressive animals as the following years breeding stock, mink farmers in Ontario curated herds that are more docile than their wild counterparts. Further work genotyping AR in mink of known aggression level could help to determine whether allele 298 is associated with lower aggression levels.
Insulin-like growth factor 1 (IGF-1) encodes a hormone similar in structure to insulin and is involved stimulating cell growth and The frequency of allele 103 was significantly different between wild and domestic mink; this shift towards allele 103 and other mid-range TA B L E 4 Number of alleles (Na), allelic richness (Ar), expected heterozygosity (He) and observed heterozygosity (Ho) for functional microsatellite loci in wild and domestic mink from three lines: Nova Scotia black (NSB), Ontario black (ONB) and Ontario brown (OND)

TA B L E 5
Proportion of genetic variance of four functional microsatellite loci (AR, ATN1, IGF-1 and TOB1) and a neutral data set explained by environmental variables (Farm Density, Road Density, Latitude and Longitude) using redundancy analysis on free-ranging mink in Ontario IGF-1 alleles may have been a result of artificial selection for large size, though this is purely speculative.
Atrophin 1 (ATN1) encodes a protein thought to be involved in kinase binding, toxin receptor binding and transcription corepressor activity (Wood et al., 2000). The exact function of ATN1 is unknown, but an expansion of the trinucleotide repeat within this gene is responsible for dentatorubral-pallidoluysian atrophy (DRPLA), a neurodegenerative disorder similar to Huntington's disease (Wood et al., 1998). Transducer of ERBB2 (TOB1) encodes a protein thought to function as a tumour suppressor (Zhang et al., 2016). We cannot attribute allele frequency differences between populations at either ATN1 or TOB1 to a particular function.
Overall, captive American mink displayed a lower number of alleles, at both neutral and functional loci, than wild mink. In addition to artificial selection, this loss of variation in captive mink and the difference in allele frequencies between wild and captive mink may be a result of other genetic mechanisms influencing domestication, namely inbreeding, relaxed natural selection or genetic drift. Natural selection accompanies artificial selection in captive populations, usually in the form of reproductive failure or increased infant mortality rate due to inbreeding depression (American Fur Breeder, 1959;Belliveau et al., 1999;Demontis et al., 2011;Price, 1984). While inbreeding is usually a chance phenomenon, intentional breeding of related individuals is practiced in an attempt to obtain or maintain a particular characteristic (Belliveau et al., 1999;Price, 1984). For instance, many pelt colours, known as colour phases, are recessive to the standard brown and must be line bred; however, this breeding strategy can result in inbreeding at loci controlling fur quality traits as well as at traits linked with such loci (Belliveau et al., 1999;Gregorius, 1980).
American mink possess many traits that are thought to impede the domestication process: solitary and territorial social structure, altricial young, extreme wariness to man and extreme agility (Hale, 1969). These behavioural characteristics are not advantageous to F I G U R E 3 Redundancy analysis (RDA) of the relationship between American mink (Neovison vison) alleles and environmental variables at three functional genes (AR (females only), ATN1, IGF1) and a data set of neutral microsatellites. Alleles are represented by points, with outlier alleles coloured orange. Arrows indicate the direction and magnitude of the environmental variables life in captivity, so it is very likely that the first mink herds experienced intense natural selection in the first few generations following capture. However, many characteristics that were once heavily selected for or against in nature are done so at a much lower intensity in captivity. For example, captive mink do not compete for resources, avoid predation or compete for mates; these behaviours essential for survival in the wild lose their significance in captivity, and as a result, genetic and phenotypic variability for these traits are likely to increase (Price, 1984). Relaxed natural selection may result in an increased frequency of alleles that are normally selected against in nature, and while these deleterious alleles may have no effect in a farm environment, they could have negative impacts on mink survival in the wild (Lynch & O'Hely, 2001;Price, 1984;Rhymer & Simberloff, 1996;Snyder et al., 1996). Androgen receptor allele 289 was found only in captive mink populations, and while it is possible the trait associated with this allele is the target of artificial selection, the fact that it was found in each captive population regardless of breeding line suggests captivity has relaxed the natural selection normally acting on this trait in the wild. However, it is not known whether AR allele 289 originated in captivity or existed in the wild before mink were sampled for this study.
Genetic drift is the change in frequency of an allele in a population due to random sampling. The phenomenon is unpredictable and tends to reduce genetic variability within populations while increasing genetic variability between populations (Dobzhansky & Pavlovsky, 1957). Considering the colour-phase strategy in which mink are bred, genetic drift may affect mink differently according to their line and farm. Drift is a common occurrence in captive groups of animals founded by small isolated populations; therefore, captive American mink are highly susceptible to this genetic mechanism (American Fur Breeder, 1959;Price, 1984).

| Apportioning genetic variation to environmental variables
Traditionally, to evaluate genetic differentiation, conservation genetic studies use F ST , but this method requires defining populations, which is unrealistic to do for wild populations that show continuous genetic gradients (Lewontin & Krakauer, 1973;Martins, Caye, Luu, Blum, & François, 2016). To avoid this impracticality, we used RDA, an individual-based multivariate method that can detect significant genetic differentiation through the identification of outlier alleles using the test statistic Mahalanobis distance. We then performed a secondary RDA using the outliers defined by Mahalanobis distance, effectively "creating a space in which we can identify the environmental variables that are most correlated with putatively adaptive variation" (Capblancq et al., 2018;Steane et al., 2014). We discovered putative adaptive variation in wild mink through the identification of outliers in ATN1, IGF1 and neutral loci. In the context of our study, outlier alleles are likely domestic alleles that have been introduced to the wild population as a result of recent mink releases.
The environmental variables used in the RDA support our hypothesis-we found latitude, longitude and road density are all significant influences on outlier alleles. Road density, which we used as a proxy for human population density and therefore also a high density of anthropogenic features, was only significant in neutral loci, and so, we think it unlikely that anthropogenic change is a factor in functional genetic diversity. However, as most of Ontario's mink farms are in the southwestern section of the province (Figure 1), it is reasonable that the variables indicating spatial location are correlated to functional genetic diversity of wild mink populations. Since farms become less frequent the farther north or east one travels, the level of impact from a domestic release event should change as latitude and longitude changes, which is likely why we saw an association Figure 3) and environmental variables at IGF-1 and a data set of neutral microsatellites. Alleles are represented by points, with outlier alleles coloured orange. Arrows indicate the direction and magnitude of the environmental variables of latitude with IGF1 and longitude with AR and ATN1 (Table 5).

F I G U R E 4 Redundancy analysis (RDA) of the relationship between American mink (Neovison vison) outlier alleles (identified in
Additionally, the sPCA of each functional marker shows a clear genetic distinction between individuals located in areas containing mink farms and those located far away from farms, namely those located farther north ( Figure 5).
Interestingly, we found no evidence that mink farm density significantly affected functional allele frequency. In a past study, mink farm density was not the most informative variable in predicting a free-ranging mink's probability of being of domestic origin (Beauclerc et al., 2013). Due to a disparity in mink farm biosecurity measures, not all farms release individuals at the same rate, so some areas of high mink farm density have few escapees, while other areas of high mink farm density have a large number of escapees (Beauclerc et al., 2013).
With this inconsistency, it is unlikely for RDA to identify mink farm F I G U R E 5 Spatial analysis of principal components of (1) AR (females only), (2) ATN1, (3) IGF-1 and (4) neutral loci, performed on freeranging American mink (Neovison vison) in Ontario. Each square represents an individual mink. Genetic differentiation is steepest between individuals with large white squares and those with large black squares 1.

2.
density as a significant variable affecting functional allele frequency, even if the frequency differences are a result of domestic release events. Instead, in concordance with our study, Beauclerc et al. (2013) found latitude to be a better predictor variable.
A study of the native red fox populations suggested that rare haplotypes were the result of the release of captive fox from local fur-farms which may have been replenishing their stock from native populations farther south (Mercon et al., 2017). Farmed mink in Canada have been domesticated for many generations, with relatively few wild mink used to supplement the domestic population due to disease concerns. There is also a commercial trade of domestic mink among farms. Overall, we consider the combined effects of the genetic founders in farms, inbreeding, drift and artificial selection to be the most likely processes leading to differentiation between farmed and wild mink in Ontario, as opposed to the introduction of wild mink alleles from distant locations. While functional genetic diversity of free-ranging mink in areas near mink farms was not exceedingly different from those far away from farms, we advise genetic monitoring of American mink, because as domestic release events continue to occur, the disruption of genetic structure of wild populations surrounding farms will continue to accumulate. Ideally, protecting the genetic integrity of wild mink populations would require preventing domestic release events or establishing a protocol to mitigate the impact of these events. To decrease the number of these events in the future, we advise for increased fencing and security on farms, as recommended by the Mink Biosecurity

Standard-Producers Guide (Canadian Food Inspection Agency, 2016).
If a release event does occur, a protocol for rapid response should be followed to reduce the probability of mink establishing feral populations. Countries such as Spain, Italy, and Estonia have recommended (a) installing surveillance systems to record presence of mink around farms, (b) preparing trapping and baiting equipment for immediate use and (c) training individuals to capture mink (Bonesi & Palazón, 2007).

ACK N OWLED G EM ENTS
We thank the mink farmers of Ontario and Brian Tapscott of OMAFRA for their sample donation. Thank you to Carrie Sadowski for designing our map. We also thank K. Trudeau, B. Redquest and the Wilson Lab technicians for assistance in the laboratory.

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

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly avail-