Genetic Characteristics of Restored Elk Populations in Kentucky

Translocations are a common management practice to restore or augment populations. Understanding the genetic consequences of translocation efforts is important for the long‐term health of restored populations. The restoration of elk (Cervus canadensis) to Kentucky, USA, included source stocks from 6 western states, which were released at 8 sites in southeastern Kentucky during 1997–2002. We assessed genetic diversity in restored herds and compared genetic similarity to source stocks based on 15 microsatellite DNA loci. Genetic variation in the restored populations was comparable to source stocks (x̄ allelic richness= 3.52 and 3.50; x̄ expected heterozygosity= 0.665 and 0.661 for restored and source, respectively). Genetic differentiation among all source and restored populations ranged from 0.000 to 0.065 for pairwise FST and 0.034 to 0.161 for pairwise Nei's DA. Pairwise genetic differentiation and Bayesian clustering revealed that stocks from Utah and North Dakota, USA, contributed most to restored populations. Other western stocks appeared less successful and were not detected with our data, though our sampling was not exhaustive. We also inferred natural movements of elk among release sites by the presence of multiple genetic stocks. The success of the elk restoration effort in Kentucky may be due, in part, to the large number of elk (n= 1,548), repeated releases, and use of diverse source stocks. Future restoration efforts for elk in the eastern United States should consider the use of multiple stock sources and a large number of individuals. In addition, preservation of genetic samples of founder stock will enable detailed monitoring in the future. © 2020 The Authors. The Journal of Wildlife Management published by Wiley Periodicals, Inc. on behalf of The Wildlife Society.

Humans have transplanted animals and plants outside their native ranges for millennia (Hofman and Rick 2018). In modern times, the translocation of individuals is one of the most widely employed and successful management techniques for wildlife populations. Translocations have been used for the restoration or augmentation of extirpated populations, to relocate nuisance individuals, to establish new populations of economically desirable animals, to establish additional populations of threatened or endangered species, and to relieve overabundance (Nielsen 1988). Early translocations often experienced mixed success due to primitive capture, handling, and transport methods. As transplantation techniques improved, issues involving animal behavior, adaptive genetic variation, and long-term effects became apparent. For instance, captive stocks are unsuitable for restoration of many species of upland birds because they lack appropriate predator avoidance and foraging behaviors (Leopold 1944). The restoration of white-tailed deer (Odocoileus virginianus) to the southeastern United States was one of the most successful wildlife management actions in history, yet the use of diverse source stocks has left a legacy of altered breeding dates throughout the region (Sumners et al. 2015). Furthermore, not all source stocks were successful, possibly because of differences in adaptive genetic variation between the source populations and the new environment (DeYoung et al. 2003). Similarly, restoration of wild turkey (Meleagris gallopavo) using wild stocks has been successful, yet restored stocks have expanded and introgressed into unique gene pools (Latch et al. 2006, Seidel et al. 2013. Although the long-term success of translocations is ultimately dictated by availability of habitat (Bouzat et al. 2009), understanding the success or failure of active management remains a primary concern.
Translocations are commonly used to augment lowdensity populations or, in wildlife restoration efforts, to return populations to their historical ranges (Griffith et al. 1989). Unless carefully planned, translocations can reduce population viability through loss of genetic variation. Small populations are prone to losing genetic diversity via a process termed genetic drift because of stochastic differences in reproduction and survival among individuals (Wright 1931). Founder events occur when few individuals initiate a population, thus causing a reduction in overall genetic diversity (Tarr et al. 1998, Broders et al. 1999, Hedrick et al. 2001, Cardoso et al. 2009). Serial bottlenecking occurs through the use of restored populations as source stocks for later restoration efforts, where genetic diversity may be lost in each successive restoration event step via the founder effect or genetic drift (Clegg et al. 2002, Gautschi et al. 2002, Lambert et al. 2005. Elk (Cervus canadensis) were once widespread in the eastern United States but were extirpated by the late 1700s to mid-1800s because of habitat loss and overhunting (Bryant and Maser 1982). Highly valued as a game animal, elk have been the subject of numerous restoration efforts in North America since the twentieth century. Early reintroduction efforts began in the western United States using remnant populations of elk from the Greater Yellowstone Ecosystem. These recovery efforts were largely successful in that restored populations exhibited high genetic diversity and low differentiation, indicating rapid growth and gene flow after initial establishment (Hicks et al. 2007).
Efforts to restock elk to the eastern United States, including the Midwest and within the Appalachians, have resulted in mixed success, often due to post-introduction mortality through hunting, poaching, disease, and vehicle collisions (Popp et al. 2014, Keller et al. 2015. Limited population growth over time has resulted in cases where reintroduced elk exhibit low genetic diversity many years after restoration (Conard et al. 2010). For example, current populations of restored elk in Pennsylvania, USA, exhibit comparatively low levels of genetic variation, attributed to serial bottlenecking and genetic drift (Williams et al. 2002). Pennsylvania's elk herd is also highly differentiated from western populations, including its stock source, probably because of founder effect and genetic drift.
Genetic researchers have reported a correlation between genetic variation and the health and survival of wildlife populations (Honeycutt 2000). Unlike historical management actions, modern restoration programs have the ability to assess the genetic effects of translocations as part of post-release monitoring of restoration efforts. During the late 1990s, the Kentucky Department of Fish and Wildlife Resources (KDFWR) embarked on an elk restoration project using stock sources from western states. Overall, 1,548 elk were translocated from source populations in Arizona, Kansas, New Mexico, North Dakota, Oregon, and Utah, USA (G. S. W. Jenkins, KDFWR, personal communication). Reintroduced elk expanded rapidly and the elk population in Kentucky presently numbers about 11,000 animals, by far the largest herd east of the Mississippi River (KDFWR 2016). To date, no analysis of genetic structure has been conducted, and the relative influence of the different stock sources on the current population is unknown. We analyzed genetic variation and structure in the recently restored elk herd in Kentucky and compared genetic variation in the restoration area to 5 of the stock sources. Specific objectives were to determine genetic diversity and differentiation among stock sources, to examine the genetic structure within the Kentucky population, to determine the relative influence of different stock sources, and to examine gene flow between release sites.

STUDY AREA
Elk were released in southeastern Kentucky within a 16-county, 1.2 million-ha restoration zone. The restoration zone is located next to the Cumberland Plateau and is comprised of 79% mixed-deciduous forest (Braun 1950), 10% surface mines, 9% agriculture, and 2% urban (Olsson et al. 2007 (Fig. 1). The climate in the restoration zone was temperate humid continental (Overstreet 1984), with an average winter temperature of 4°C and an average summer temperature of 24°C (McDonald and Blevins 1965).
Six different western states served as sources for translocated elk to Kentucky (Table 1). Utah was the largest source and contributed 1,027 elk from the Ensign Ranch (near Castle Rock, UT), the Deseret Ranch (northwest of Evanston, WY), Mount Nebo, Ogden Valley, and near the towns of Echo, Henefer, Morgan, and Nephi. Elsewhere, stocks were obtained from Theodore Roosevelt National Park, North Dakota, the Maxwell Refuge near Canton, Kansas, Flagstaff, Arizona, Ute Mountain and the Uracca

Sample Collection and Amplification
We obtained muscle tissue samples from hunter-harvested elk during the 2008 and 2009 male and female hunts in the 8 southeastern Kentucky restoration zones. Samples of released stocks were not available, so we also obtained hunterharvested tissues from the source states during 2008-2009 from the same herd units as the founding individuals. We froze and stored all samples at −20°C until DNA extraction. We isolated DNA from tissue samples using the Qiagen® DNeasy™ Tissue Kit (QIAGEN Genomics, Germantown, MD, USA). We amplified 15 microsatellite DNA loci for all samples. All loci were tetranucleotide repeats, and included C01, T107, C143, C180, and C273 (Meredith et al. 2005), and T123, T156, T172, T193, T268, T108, C217, T26, T501, and T507 ( Jones et al. 2002). We amplified loci in 4 multiplex reactions, as described by Meredith et al. (2007), and loaded the resulting products onto an automated genetic analyzer for separation and detection (3130xl, Applied Biosystems, Foster City, CA, USA). We determined allele size calls for each locus using GeneMapper 4.0 (Applied Biosystems). We implemented several different quality control procedures during the genotyping of the samples. First, we modified all reverse primers with a 5′-GTTTCTT sequence to reduce split peaks and encourage +A addition (Meredith et al. 2007). This resulted in almost no detectable stutter peaks. Because the markers amplified only tetranucleotide repeats, all allele bins were separated by 4 base pairs; combined with the lack of stutter, the potential for allele calling errors was greatly reduced. We included a standard reference DNA sample in every polymerase chain reaction run to confirm consistency of electrophoretic mobility of fragments among runs. We called only alleles that were ≥100 on the fluorescence intensity scale; if allele-like peaks were <100 units, we scored the allele as missing. Finally, we re-amplified and re-scored ≥10% of samples to assess consistency of allele calls.

Data Analysis
We indexed genetic diversity based on allelic richness (A; El Mousadik and Petit 1996), computed in HP-RARE (Kalinowski et al. 2007), to facilitate comparisons among samples. We calculated observed heterozygosity (H O ) and expected heterozygosity (H E ) using ARLEQUIN 3.1 (Excoffier et al. 2005). We computed the inbreeding coefficient (F IS ) using ARLEQUIN 3.1 to evaluate departures from equilibrium within samples and computed 95% confidence intervals to determine significant differences from zero. Linkage disequilibrium within restored populations may indicate admixture between stock sources; we tested for linkage disequilibrium within and among samples using ARLEQUIN 3.1. We assessed genetic structure using genetic distances and Bayesian clustering. First, we computed pairwise F ST using the Reynolds et al. (1983) coancestry coefficient among all sites using ARLEQUIN 3.1 (Excoffier et al. 2005); we assessed statistical significance based on 100 permutations of genotypes between populations. Genetic distances such as F ST rely on an underlying evolutionary model (Nei and Kumar 2000) and may not perform as expected in the presence of admixture. Therefore, we also computed Nei's D A (Nei et al. 1983), which does not assume an evolutionary model and is more efficient at recovering simulated population relationships than many other genetic distances (Takezaki and Nei 1996); we computed pairwise D A values using ADEGENET ( Jombart 2008). Next, we evaluated contribution of source stocks to restored populations using the Bayesian clustering algorithm STRUCTURE 2.3  (Pritchard et al. 2000(Pritchard et al. , 2010. We used the admixture model and assumed correlated allele frequencies. We used the LOCPRIOR designation for source stocks, where sampling location acts as a weak prior to inform clustering (Hubisz et al. 2009), and modeled the Kentucky samples as unknowns. We modeled the assumed number of genetic clusters (K) from 1-10, with an initial burn-in of 50,000 Markov chain Monte Carlo (MCMC) repetitions, followed by 150,000 MCMC repetitions for data collection. We repeated the runs for each K 5 times. We determined the most likely number of clusters based on the change in the likelihood function between each successive cluster (ΔK; Evanno et al. 2005) using the online software program STRUCTURE HARVESTER (Earl and vonHoldt 2012).
We performed an additional series of runs at the most likely K, K−1 and K+1, to ensure model convergence. The additional runs consisted of a 100,000 MCMC burn-in followed by 200,000 MCMC repetitions of data collection, with 10 iterations at K, K−1, and K+1.

RESULTS
We collected 421 samples from Kentucky and 5 of the 6 states that provided stock sources: Arizona, New Mexico, North Dakota, Oregon, and Utah. We were unable to procure samples from Kansas. Mean proportion of loci typed was 91% and no monomorphic loci were present in any population. Although there were private alleles at 5 loci (C01, C180, T107, T193, and T128), all were at low frequency, present in only 1 or 2 copies, and uninformative for assignment purposes. Allelic richness and observed heterozygosity were similar among all populations, and ranged from 3.3-3.7 and 0.643-0.682, respectively (Table 2; Table S1, available online in Supporting Information). Oregon had the highest number of locus pairs that displayed linkage disequilibrium (27), followed by the Raven restoration zone with 23; the Letcher County restoration zone had the least with only 1 pair of loci in linkage disequilibrium. Estimates of F IS ranged from −0.038 to 0.048, with a slight excess of heterozygotes in 3 of 5 source stocks and 4 of 8 release populations, although all 95% confidence intervals overlapped 0. Statistically significant pairwise F ST ranged from 0.012 to 0.065, with the lowest differentiation between New Mexico and the Pike County restoration site in Kentucky and the highest between the Raven and Redbird WMA restoration sites in Kentucky (Table 3) Based on the initial Bayesian clustering runs, we determined that 3 clusters showed the greatest change in likelihood values (Fig. 2). The additional runs at K = 3 revealed clear assignment of the North Dakota population and Utah population to distinct, separate clusters (Fig. 3). The Arizona and Oregon populations grouped together into a third cluster, whereas the New Mexico population had nearly equal membership proportions between all 3 clusters. Overall, elk in Kentucky displayed high levels of admixture at the individual and site level, with most membership proportion assigned to clusters representing Utah and North Dakota. The Blue Diamond, the Orr Property, Redbird Wildlife Management Area, and Czar-Martin County restoration sites clustered with Utah, whereas the Raven site clustered with North Dakota. Letcher and Pike counties, and the Starfire site were admixed between the Utah and North Dakota clusters. Although the ΔK method suggested best fit at K = 3 clusters, we also interpreted K = 4 clusters    because Arizona and Oregon populations formed 2 distinct groups at K = 4. We found no further differentiation within the Kentucky populations.

DISCUSSION
Overall, the elk reintroduction in Kentucky appears to be the most successful restoration effort for elk in the eastern United States, as evidenced by rate of population growth, total population size, and our genetic assessment of the herd. Growth of the elk herd in Kentucky from a release of 1,548 animals to the current estimate of 11,000 individuals suggests that the population rapidly expanded postintroduction (KDFWR 2016). Rapid growth of introduced populations and use of diverse source stocks can alleviate loss of genetic diversity via founder effects and genetic drift (DeYoung et al. 2003, Conard et al. 2010. Furthermore, the size and diversity of founding stock is associated with success of restoration efforts (Smith et al. 1976;Leberg 1990Leberg , 1993. We were unable to directly address the factors that contributed to the success, but the Kentucky restoration involved a larger founder stock and number of stock sources compared to other elk restoration efforts in the eastern United States (Popp et al. 2014). Although previous researchers of elk restoration reported that both of these factors had little effect on the success of translocation, Kentucky represents a substantial increase in effort compared to other cases (Popp et al. 2014). Popp et al. (2014) reported that across all eastern elk restorations, stock size and diversity were not associated with translocation success; however, they indicated that Kentucky was an exception to this rule. Additionally, Conard et al. (2010) reported that number of restocked individuals and diversity of stock sources only marginally  Evanno et al. (2005). We ran STRUCTURE using the LOCPRIOR model with elk stock sources as known populations and the elk in Kentucky, USA, modeled as unknowns using genetic samples collected in 2008-2009. The highest change in ΔK is the likeliest number of clusters; therefore, our analysis broke elk populations into 3 major groups.  (Nei et al. 1983;upper diagonal) and F ST (Reynolds et al. 1983;lower diagonal)  explained genetic diversity in restocked elk populations, instead stable or increasing post-introduction populations played an important role. Kentucky restoration efforts were conducted over half a decade with each successive cohort of translocated elk providing large numbers of new individuals. By using more individuals and stocks than other states and substantially augmenting stock each year, Kentucky established a viable elk population that was able to quickly grow. We observed levels of genetic diversity in the restored Kentucky populations comparable to that of the stock sources used for restoration. Although levels of heterozygosity based on different genetic loci are not directly comparable across studies, our findings were similar to western elk populations reported by Hicks et al. (2007), where expected heterozygosity ranged from 0.51-0.60. Heterozygosity found in the Kentucky population was substantially higher than other restored populations of elk east of the Mississippi River; for example, expected heterozygosity in the elk herd in Pennsylvania was 0.254 (Williams et al. 2002). This result is consistent with the demographic history and number and source of stocks used (1,548 in Kentucky compared to 177 in Pennsylvania). Additionally, the elk population in Kentucky has rapidly grown since stocking in the late 1990s, whereas the population in Pennsylvania only grew to an estimated 833 by 2013 from the original population in the early 1900s (Popp et al. 2014). We documented clear genetic structure within the Kentucky herd as a result of stock sources, specifically the North Dakota and Utah stocks. North Dakota was used to stock only the Raven restoration site in Kentucky, and our STRUCTURE analysis showed clear evidence of this. Utah, on the other hand, provided the largest percentage of elk to the restoration effort and provided elk to all but one (Raven) restoration site. Other than those individuals that grouped with North Dakota, all but 2 elk in Kentucky were predominately assigned to the Utah cluster.
The clear distinction between North Dakota and all other populations except for Kentucky was not reported in previous comparisons of restocked western populations where North Dakota, specifically Theodore Roosevelt National Park, was included (Hicks et al. 2007). One explanation for the genetic distance between North Dakota and other populations is that North Dakota's elk population was stocked from the intermediate location of Wind Cave National Park in South Dakota (Anderson 1958). This additional step may have contributed to the differentiation between North Dakota and the other stock sites, all of which were founded with stocks from Yellowstone National Park. We found no clear evidence of stock contributions from the Arizona, New Mexico, or Oregon sources in our analysis of the Kentucky herd. This result is not conclusive and may be due to sampling error because our sampling was not exhaustive. Nonetheless, the Bayesian clustering results indicated that the Arizona and Oregon stocks were welldifferentiated and should be apparent if they were present in elk in Kentucky. Mortality after translocations can be high in some cases, and stocks from these sources may not have survived long enough after restoration to provide a genetic contribution. Although we were unable to sample from the Kansas elk herd, that stock source represented only 2.5% of the translocated individuals, and in both release sites they were the smallest portion of stock sources. It seems improbable that inclusion of Kansas samples in our analysis would have changed our main conclusions. It is more likely that Utah stocks predominate in the Kentucky herd because of the high number of founders; North Dakota stocks are largely pure in the Raven site because those were the only elk released in that area.
There appears to be some dispersal between restoration sites, as evidenced by assignment of some individuals from the Letcher County, Pike County, and Starfire restoration sites to the North Dakota cluster. These sites are all geographically proximate to the Raven restoration site where all the elk from North Dakota were translocated. The Kentucky elk herd has not been observed to conduct seasonal migratory movements (Wichrowski et al. 2005). Larkin et al. (2004), however, documented translocated adult elk in Kentucky moving up to 23 km within 12 months following release. Letcher County, Pike County, and Starfire restoration sites are all within 40 km from the Raven restoration site and, therefore, are well within range of post-translocation dispersal. Future sampling may show an increase in homogenization across the region as elk continue to disperse through population growth and movement.
The success of the Kentucky restoration effort runs counter to a host of examples of low success rates of elk reintroductions across the eastern United States and Canada. Popp et al. (2014) documents the failure of 40% of eastern elk reintroductions as a result of genetic bottlenecking, overhunting, and mortality due to disease. We found that elk in Kentucky show little evidence of genetic bottlenecking with high levels of genetic diversity. Additionally, through a well-regulated hunting season since 2001, the Kentucky herd has continued to grow and expand, showing little indication of any detrimental effects of harvest rates (KDFWR 2016). Although meningeal worm (Parelaphostrongylus tenuis) was a concern in the past among elk in Kentucky (Larkin et al. 2003), the population has not been significantly affected by the disease in the long term. Finally, although the restoration zone for elk in Kentucky was chosen primarily for its low human population and to minimize conflicts with agriculture (Larkin et al. 2001), the presence of reclaimed surface mines (10% of the restoration zone; Olsson et al. 2007) provided early successional vegetation beneficial to elk upon their release. Adequate habitat, post-reintroduction cannot be discounted as a key factor in the success of elk restoration efforts in Kentucky.
Understanding how species respond to restoration efforts through translocations helps biologists learn from past mistakes and prepare for future conservation efforts. Leberg (1990) calls for careful consideration to be taken with regard to genetic variation in species targeted for restoration efforts. To this end, we have seen translocation success stories across North America with rapid population growth and high genetic variation in white-tailed deer (DeYoung et al. 2003), wild turkeys (Seidel et al. 2013), and western populations of elk (Hicks et al. 2007). The continued growth and success of the elk herd in Kentucky after restocking in the late 1990s further demonstrates the importance of using multiple stock sources and translocating large numbers of individuals repeatedly over several years. Genetic variation within the Kentucky herd can be attributed to ≥2 stock sources, and they exhibit high levels of genetic diversity. Wildlife managers should look to the elk model in Kentucky as a blueprint for future translocation efforts.

MANAGEMENT IMPLICATIONS
The effort, time, and expense of restoring populations of large game animals make it imperative to use every tool to ensure success. In the case of the elk restoration project in Kentucky, the Kentucky Department of Fish and Wildlife Resources used a large number of translocated stock from 6 sources to combat chances of inbreeding, founder's effect, and serial bottlenecking. These efforts resulted in a robust elk herd that rapidly grew despite hunting and presence of meningeal worm. Kentucky elk display high genetic diversity, with genetic contributions of ≥2 stock sources still apparent. When resources allow, state agencies should attempt to diversify their stock sources and work to ensure rapid initial growth. Furthermore, preservation of tissues from the founding individuals will enable more precise tracking of the contribution of founding individuals and stocks to restored populations.