OUP user menu

Reintroduction and Genetic Structure: Rocky Mountain Elk in Yellowstone and the Western States

Jason F. Hicks, Janet L. Rachlow, Olin E. Rhodes Jr., Christen L. Williams, Lisette P. Waits
DOI: http://dx.doi.org/10.1644/06-MAMM-A-051R1.1 129-138 First published online: 28 February 2007


Translocation is a common tool for restoring wildlife populations; however, potential genetic consequences include reduced levels of diversity within and increased divergence among populations. Elk (Cervus elaphus) were extirpated across much of North America by the early 20th century, but subsequent translocation programs restored the species to much of its historic range. The effects of these reintroductions on current patterns of genetic diversity in the western United States are largely unknown. We predicted that populations initiated with few founders and those experiencing slow postreintroduction growth would exhibit lower levels of diversity than other reintroduced populations. We used 12 microsatellite markers to examine patterns of genetic variability across 5 reintroduced populations of elk and 2 source herds from the Greater Yellowstone Ecosystem. The northern and southern Yellowstone source herds, which migrate to wintering areas separated by more than 260 km, exhibited similar levels of genetic diversity and high levels of gene flow, identified through both direct (i.e., assignment tests) and indirect measures. Levels of genetic diversity also were relatively high in all populations (unbiased heterozygosity, HE = 0.51–0.60; allelic richness based on a sample size of 21, AR21 = 3.3–4.0) and did not differ significantly between source and reintroduced populations or among reintroduced populations. We observed low to moderate levels of differentiation (Weir and Cockerham's FST statistic, θ = 0.01–0.08) and small genetic distances (Nei's standard genetic distance, DS = 0.02–0.11) between populations. The relatively high levels of genetic diversity and low differentiation observed among our sampled populations are in stark contrast to observations of low diversity and high differentiation among isolated reintroduced populations of elk in the eastern United States. These results suggest that gene flow that includes other elk populations in the western United States may aid in preserving genetic diversity and limiting genetic divergence.

Key words
  • Cervus elaphus nelsoni
  • elk
  • gene flow
  • genetic diversity
  • Greater Yellowstone Ecosystem
  • microsatellites
  • reintroduction
  • Rocky Mountain elk
  • Yellowstone National Park

Translocation of wildlife is a common tactic for restoring extirpated populations and curbing loss of biodiversity (Griffith et al. 1989). Recently, researchers have begun to assess the genetic consequences of founder events and postreintroduction demography in reestablished populations (Fitzsimmons et al. 1997; Mock et al. 2004; Rhodes et al. 1995; Williams et al. 2000). Small numbers of founders and low growth rates following reintroduction can result in prolonged small effective population sizes, increased genetic drift, and higher rates of inbreeding, all of which can reduce genetic diversity within and increase divergence among reintroduced populations (Leberg 1993; Sjöberg 1996; Stockwell et al. 1996). Reductions in genetic diversity can impact individual fitness and consequently reduce population viability and evolutionary potential (Reed and Frankham 2003; Saccheri et al. 1998). Therefore, conservation strategies designed to reestablish wildlife populations should evaluate the consequences of translocations on genetic structure.

Even in the absence of population manipulation, many life history characteristics including mating systems, dispersal, and migration can influence gene flow and result in substantial genetic structure within and among populations (Bohonak 1999; Hedrick 2000; Ross 2001). The effects of these characteristics can be quite cryptic, and genetic investigations often have yielded results that were not predicted based on observations of animal behavior (e.g., Gibbs et al. 2000; Hasselquist and Sherman 2001). Migration commonly results in population overlap on shared wintering areas for many birds, and high levels of genetic population structure result from fidelity of individuals to specific breeding sites (Kimura et al. 2002; Rhodes et al. 1995; Webster et al. 2002). Some terrestrial mammals, such as Rocky Mountain elk (Cervus elaphus nelsoni) and barren ground caribou (Rangifer tarandus granti), exhibit opposite migratory patterns, occupying spatially segregated winter ranges and shared summer ranges. Although largely unknown, the potential effects of migration patterns on genetic structure are relevant for species in which population management includes translocation efforts.

The widespread reduction and subsequent restoration of elk populations in North America provides an opportunity to examine genetic structure in both reintroduced and natural migratory populations. Before European settlement, an estimated 10 million elk inhabited North America (Seton 1927). However, by 1922 numbers were reduced to approximately 90,000 and the species had been extirpated across most of its former range (Bryant and Maser 1982). One of the remaining strongholds for elk was Yellowstone National Park (YNP). Efforts to restore extirpated populations began in the early 1900s when Rocky Mountain elk were captured in the Greater Yellowstone Ecosystem (GYE) and reintroduced into numerous areas in the western States (O'Gara and Dundas 2002). From 1912 to 1967, >13,500 elk were translocated to locations throughout North America, including Mexico and Canada (Robbins et al. 1982), and most reintroduced populations were established with individuals from the GYE (Wolfe et al. 2002).

Multiple elk herds exist in the GYE including the Gallatin, Crandall-Sunlight, North Fork Shoshone, South Fork Sho-shone, Greybull, Madison-Firehole, Paradise Valley, northern Yellowstone, and Jackson herds. Almost 50 years of research on movements of tagged elk suggests apparently low dispersal among herds in the GYE (Anderson 1958; Berger 2004; Boyce 1989; Craighead et al. 1972; Houston 1982; Smith and Anderson 2001; Smith and Robbins 1994). Although some individuals from each of the GYE elk herds occupy summer ranges within YNP, nearly all of the herds migrate, at least in part, to discrete winter ranges outside of the YNP. Given the migratory nature of these populations and documented low rates of interchange among herds, the degree of genetic structure within the GYE elk is unknown. Analysis of 3 polymorphic allozymes failed to document genetic differentiation among 3 GYE herds (Glenn and Smith 1993); however, gene flow within the GYE has not been investigated using highly polymorphic molecular markers, such as micosatellites, which allow for fine-scale resolution of population structure.

In this study, we evaluated how reintroduction histories have impacted current genetic structure in Rocky Mountain elk in the western United States. Because the elk populations we studied were established with founders from 2 migratory GYE herds, we also examined gene flow and genetic structure within the GYE. Our specific objectives were to quantify rates of gene flow between the northern and southern GYE elk herds using both direct and indirect measures of gene flow; determine if levels of genetic variation were lower in reintroduced populations of elk than in the GYE source herds; compare levels of genetic diversity among reintroduced populations of elk with differing founding characteristics; and evaluate the degree of divergence among the reintroduced elk populations. We predicted that reintroduced populations with relatively few founders and slow postreintroduction growth would exhibit lower levels of diversity than their source populations as well as other populations established with greater numbers of founders. An understanding of how founding characteristics shape current genetic structure is important for evaluating the success of translocation programs in population restoration (Hedrick 2005).

Materials and Methods

Study populations.—We sampled 2 elk herds in YNP that served as the sources for elk reintroductions across the western States. The northern Yellowstone elk herd (nYNP) occupies a winter range on the northern border of YNP (Fig. 1). Since 1986, yearly population estimates have ranged from 8,000 to 19,000 with 8,335 elk counted on the northern range in 2004 (White and Garrott 2005). The winter range of the Jackson elk herd includes much of the southern GYE (sGYE). During the winter months, approximately 10,000 elk from this herd reside on the National Elk Refuge with an additional 5,000 elk in the neighboring river drainages (Smith and Anderson 1998; Smith and Robbins 1994).

Fig. 1

Winter and summer ranges for the northern Yellowstone National Park (nYNP) and southern Greater Yellowstone Ecosystem (sGYE) elk herds within the Greater Yellowstone Ecosystem (shaded area) in the western United States. These 2 elk herds occupy widely separated ranges during winter, but many individuals from both herds occupy shared summer range in the southern portion of Yellowstone National Park. Bold arrows indicate general directions of migrations between seasonal ranges.

Although the nYNP and sGYE elk herds occupy independent winter ranges, significant overlap exists in their summer ranges. Most elk in the nYNP herd migrate south seasonally, and occupy 4,700 km2 within YNP (Houston 1982). nYNP elk have been documented east and south of Yellowstone Lake during the summer months. The sGYE herd moves north and summers throughout much of the southern GYE including in areas south of Yellowstone Lake within YNP. This region is the most common area of overlap for elk from the nYNP and sGYE herds (Fig. 1).

Reintroduced populations from 5 western states spanning much of the historic range of Rocky Mountain elk were included in our study (Table 1). Each of those populations was established with individuals from GYE with a few exceptions (Fig. 2). The elk herd in Theodore Roosevelt National Park, North Dakota (TR), was founded with elk translocated from Wind Cave National Park (WCNP—C. C. Anderson, in litt.); however, the WCNP population was previously established with GYE elk. The Vermejo Park Ranch in northern New Mexico (VR) was originally stocked with 12 elk from Colorado and 50 elk from YNP and 43 years later, the population was supplemented with ≥150 animals from YNP (Bailey 1931; Robbins et al. 1982; Wolfe 1977). Last, we could not confirm the origin of the 1st single male released into the Wichita Mountain Wildlife Refuge (WM) in southeastern Oklahoma in 1908. All subsequent releases into WM consisted of GYE elk. The nYNP elk herd served as the direct source of founders for the VR and northwestern Arizona (AR) reintroductions, and the indirect source, through WCNP, for the TR herd. The sGYE herd was the direct source of elk for establishment of the Chesnimnus, Oregon (OR), and WM herds and the indirect source for the TR population (Fig. 2). Three of the reintroduced populations (TR, WM, and VR) are enclosed within fenced areas, and 2 (AR and OR) exist in areas where movement into or out of the population is possible.

Fig. 2

Locations of reintroduced populations of elk included in this study: Chesnimnus in northeastern Oregon (OR), northwestern Arizona (AR), Vermejo Park Ranch (VR), Wichita Mountain Wildlife Refuge (WM), and Theodore Roosevelt National Park (TR). Circle size represents the 2004 population estimate (small, <500; medium, 500–5,000; large, >5,000). Gray shading represents initial growth rates (none = unknown, light = slow, medium = stable, dark = rapid). Dashed arrows signify translocations from the northern Yellowstone National Park herd and solid arrows signify translocations from the southern Greater Yellowstone Ecosystem herd. The dotted arrow from Wind Cave National Park indicates animals that were translocated to TR after establishment of the Wind Cave population from the northern Yellowstone National Park herd. Founder sizes are adjacent to the translocation arrow for each reintroduced population. Note that the VR population was supplemented with >150 elk from the Greater Yellowstone Ecosystem in 1957.

View this table:
Table 1

Number of founders and postreintroduction histories for 5 elk populations established with individuals from 2 source herds in Yellowstone National Park (sGYE = southern Greater Yellowstone Ecosystem herd and nYNP = northern Yellowstone National Park herd).

PopulationSource herdFounding historyaInitial postreintroduction growthCurrent size estimateReferences
Chesnimnus, Oregon (OR)sGYE1912: 13 (5 f, 1 m, 7 c)Stableb2,400Bailey 1936
1913: 15V. Coggins, in litt.
Northwestern Arizona (AR)nYNP1960s: 35Unknown1,900R. Riley, in litt.
Vermejo Park Ranch (VR)1911: 12 (9 f, 3 m)cRapid: 3,400 elk by 1934Bailey 1931
nYNP1914: 508,000–10,000Bryant and Masser 1982
1957: ≥150 (supplemented)Robbins et al. 1982; Wolfe 1977
Wichita Mountain Wildlife Refuge (WM)sGYE1908: 1 (m)d 1911: 5 (4f, 1 m) 1912: 15 (12 f, 3 m)Slow: first 15–20 years500Halloran 1963 Caire et al. 1989 Walter and Leslie 2002
Theodore Roosevelt National Park (TR)Wind Cave National Park (original source nYNP)1985: 47 (38 f, 9 m)Stable: 300–400 by 1993543C. Anderson, in litt.; M. Oehler, pers. comm.; Westfall 1993
  • a f = female, m = male, c = calf.

  • b Immediate postreintroduction growth was not documented, but the population was large enough in 1917 and 1918 to support transfer of 15 elk to Crater Lake National Park, Oregon, and a group of unknown size to Wallowa Lake, Washington.

  • c The 1911 Vermejo Park translocation was from Routt County, Colorado (unknown origin).

  • d Origin of this animal is unknown.

We predicted that genetic variability would differ significantly among reintroduced populations of elk given a priori knowledge of founder numbers and postreintroduction growth (Table 1). The WM population had both the fewest number of founders (n = 21) and the slowest postreintroduction growth (Bryant and Maser 1982). Additionally, this population may have experienced a bottleneck in the late 1960s and early 1970s when roughly 70% of the population was used for translocation into eastern Oklahoma (Bryant and Masser 1982; Caire et al. 1989). The founding characteristics of the WM herd (i.e., number of founders and postreintroduction growth rates) were similar to those of an isolated elk herd in Pennsylvania founded from GYE elk (n = 34 and population size < 50 for 50 years) that exhibited very low levels of genetic diversity (expected heterozygosity = 0.254) and mean number of alleles per locus (A = 1.9—Williams et al. 2002). Therefore, we anticipated that genetic variability in the WM population would be similar to levels documented in the Pennsylvania herd, lower than the other reintroduced populations we sampled, and lower than the GYE source population. In contrast, the VR population had both the highest number of founders and likely the most rapid postreintroduction growth of all the reintroduced populations we examined (Bailey 1931; Robbins et al. 1982). Therefore, we expected the VR population to exhibit the highest levels of genetic variability among reestablished populations and those closest to the GYE source population.

Sample collection and DNA extraction.—Samples of skin or muscle tissue were collected from individual elk within each of 5 reintroduced populations and from the nYNP source herd. Muscle tissue samples were collected from hunter-harvested elk in OR (n = 27), AR (n = 17), VR (n = 34), WM (n = 43), and TR (n = 22). Additionally, skin samples were collected from skulls (n = 23) of elk with malformed antlers harvested on the Hualapai Indian Reservation in northern Arizona (Rachlow et al. 2003). No differences in allele frequencies were detected between deformed and nondeformed elk in northwestern Arizona (Hicks and Rachlow 2006), so all Arizona samples were combined into 1 group (n = 40) for analysis. Personnel in YNP provided tissue samples from nYNP elk. No live animals were used in this research. Tissue samples were frozen and stored at −20°C before DNA extraction with the exception of the OR samples, which were preserved in 100% pure ethanol and stored at room temperature. We isolated DNA from all tissue samples using standard protocols for the Qiagen tissue kit (Qiagen Co., Valencia, California). Skin samples from the Hualapai Indian Reservation were extracted using a modified Qiagen tissue protocol with an overnight soak in lysis buffer (Applied Biosystems, Foster City, California). We followed guidelines of the American Society of Mammalogists for care and use of mammals in research (Animal Care and Use Committee 1998).

Micro satellite analysis.—We selected microsatellite loci for inclusion in our study based on potential variability, amplification efficacy, and application in previous studies of elk populations. Polymerase chain reaction was used to amplify extracted DNA at the following 12 dinucleotide (CA/GT) microsatellite loci: BL42, BM203, BM415, BM4107, BM4208, BM5004, BM6506, BM848, BM888, FCB193, MAF109, and RM006 (Bishop et al. 1994; Buchanan and Crawford 1992; Kossarek et al. 1993; Swarbrick and Crawford 1992). Protocols for microsatellite amplification followed Williams et al. (2002). We ran negative controls with each set of reactions to detect polymerase chain reaction contamination and lane-to-lane leakage during allele scoring. Polymerase chain reaction products were electrophoresed using 6% acrylamide gels and an ABI 377 DNA automated sequencer (Applied Biosystems). We used GENESCAN software to extract and track gel lanes, and GENOTYPER 2.5 software (Applied Biosystems, Foster City) to size alleles.

Genotypes at 9 loci (BL42, BM415, BM4107, BM4208, BM5004, BM888, FCB193, MAF109, and RM006) were previously determined for the sGYE (Jackson) herd (n = 20—Williams et al. 2002). To accomplish genotype agreement, we selected 5 individuals from the study of Williams et al. (2002) that represented the diversity of alleles, and DNA was amplified and scored at each loci as described above. We determined correction factors to incorporate all genotypes from the previous study. Additionally, all sGYE elk samples were geno-typed at locus BM6506. Therefore, a total of 10 loci was used for analyses contrasting nYNP and sGYE elk herds.

Statistical tests.—We tested microsatellite loci for linkage and Hardy–Weinberg equilibrium using GENEPOP 3.4 (Raymond and Rousset 1995a). Pairwise locus tests for linkage equilibrium and per locus departures from Hardy–Weinberg in all populations were adjusted for multiple tests using sequential Bonferroni corrections (Rice 1989). We calculated mean observed hetero-zygosity (HO) and unbiased heterozygosity (HE) for all populations using Cervus version 2.0 (Marshall et al. 1998). Because of unequal samples sizes among populations we calculated allelic richness (AR) per population based on the minimum sample size in each analysis using FSTAT version (http://www2.unil.ch/popgen/softwares/fstat.htm; Petit et al. 1998); subscripts following AR values indicate the sample size. Allele frequency tables created using GENEPOP 3.4 were used to tally the number of private alleles and determine the number of rare alleles (i.e., frequency <10%) present in each population (Raymond and Rousset 1995a).

We 1st examined differences between the nYNP and sGYE herds, which represent the source of founders for the rein-troduced populations included in our study. Statistical comparisons of mean HE (arc-sine transformed) and mean AR between the 2 GYE herds were made using 2-tailed, paired t-tests. Differences in per locus allele frequency distributions between nYNP and sGYE herds were investigated with Fisher's exact test calculated using GENEPOP 3.4 (Raymond and Rousset 1995, 1995b). We quantified indirect gene flow and genetic distance between nYNP and sGYE herds using Weir and Cockerham's FST statistic (θ), and Nei's standard genetic distance (DSDieringer and Schlotterer 2003; Nei 1978; Weir and Cockerham 1984). Additionally, we estimated the number of migrants per generation (Nm = [(1 − FST)/4FST]) necessary to account for the observed FST value for the nYNP and sGYE herds.

Two types of assignment tests were used to document direct gene flow between the nYNP and sGYE herds. First, we used a frequency-based assignment test to calculate the likelihood of each individual's multilocus genotype occurring in the population in which it was sampled versus the individual's likelihood of origin in the other sampled population (Arrendal et al. 2004; Kyle and Strobeck 2001; Paetkau et al. 1995). Second, a Bayesian assignment approach based on clustering individuals into respective subpopulations was implemented with the program STRUCTURE version 2.1 (Pritchard et al. 2000). We initially used STRUCTURE version 2.1 to estimate the number of sub-populations (K) using 4 independent values (K = 1–4). Burn-in length and number of Monte Carlo Markov chains were increased to 200,000 burn-in and 2,000,000 post–burn-in after initial runs that displayed a lack of convergence of summary statistics and inconsistencies in posterior probabilities. Our 2nd step was to assign individual elk to the postulated subpopulations defined in the previous step using each individual's percentage ancestry (q), the proportion of an individual's multilocus genotype relegated to each subpopulation. Assignment was conducted using an a priori confidence threshold of q ≥ 0.90 (Cegelski et al. 2003; Lucchini et al. 2004).

On a regional scale, we explored genetic differentiation between population pairs using several approaches. We assessed differences in HE (arc-sine transformed) and AR between source and reintroduced population pairs and between each reintroduced population pair using 1-tailed (null hypothesis: source ≥ reintroduced) and 2-tailed paired t-tests, respectively (StatView 5.0.1, SAS Institute Inc., Cary, North Carolina). Fisher's exact test was used to assess global differences in allele frequencies when incorporating all populations and reintroduced populations without the Yellowstone source populations. Additionally, population differentiation and genetic distance between source and reintroduced population pairs and between each reintroduced population pair were measured using θ and DS (Dieringer and Schlötterer 2003). We calculated all θ and DS values using Microsatellite Analyzer (MSA—Dieringer and Schlötterer 2003), and the statistical significance of each pairwise θ and the global θ value was determined by permutation. P-values for each pairwise θ were adjusted for multiple tests using a strict Bonferroni correction (Dieringer and Schlotterer 2003).


Genetic structure within the GYE.—Patterns of genetic diversity did not differ significantly between the nYNP and sGYE elk herds. Across both herds, no significant deviations from Hardy–Weinberg or linkage equilibrium were detected after adjusting for multiple comparisons (P > 0.05). Both HE and AR were similar in both herds (nYNP: HE = 0.59, AR17 = 3.87; sGYE: HE = 0.56, AR17 = 3.28), and consequently no significant pairwise differences in HE (t = 0.72, d.f. = 9, P = 0.49) or AR (t = 1.42, d.f. = 9, P = 0.19) were revealed. Indirect measures of gene flow also indicated little to no differentiation between herds. Fisher's exact test revealed no significant differences in allele frequency distributions between herds (P = 0.271). Similarly, FST identified no significant departure from panmixia (θ = 0.004; P = 0.281), and we observed little genetic distance between herds (DS = 0.005). The number of migrants per generation necessary to accommodate this level of gene flow between herds was estimated at 62. These results indicated a lack of genetic structure within and across these 2 GYE herds.

Assignment tests also indicated a lack of differentiation and high gene flow between the 2 GYE elk herds. The frequency-based assignment test (Paetkau et al. 1995) correctly assigned only 50% and 70% of nYNP and sGYE individuals, respectively, indicating high exchange rates between the 2 herds. Using the program STRUCTURE (Pritchard et al. 2000), we were unable to resolve the number of subpopulations (K) with a high degree of certainty. We observed more variability in posterior probabilities across runs for a given K than were estimated for different Ks, despite simulation with very long Monte Carlo Markov chains (200,000 burn-in and 2,000,000 post–burn-in). Additionally, q (percent ancestry) per individual across independent runs indicated very weak membership to any particular group within each K. For example, at K = 2, K = 3, and K = 4, no individual q value was greater than 0.557, 0.376, and 0.263, respectively, which was markedly lower than the established threshold of 0.90. Thus, ubiquitously weak membership across individuals for K = 2–4 also best supported a scenario of a single genetic population for the nYNP and sGYE elk herds.

Because of similar levels of genetic diversity and little to no genetic differentiation between nYNP and sGYE herds, we used genetic data from the nYNP herd as representative of the source population for all comparative analyses between source and reintroduced populations. This allowed us to use the full complement of loci for all comparisons between source and reintroduced populations.

Regional genetic structure.—The total number of alleles per locus across the data set (n = 189 individuals from nYNP and 5 reintroduced populations) ranged from 2 to 11 with a global mean of 5.25. Amplification success was >97% per locus and population. We detected no violations of linkage equilibrium (table-wide P > 0.05), but 2 loci (BM4208 and BM888) in the AR population deviated significantly from Hardy–Weinberg proportions after adjusting for multiple comparisons (P = 0.0005 and 0.0003, respectively). No other populations exhibited deviations from Hardy–Weinberg expectations at those loci, and globally neither locus was out of Hardy–Weinberg equilibrium. Therefore, we retained the complete AR data set for analyses.

We observed only small differences in measures of genetic diversity across populations. Although the YNP source population exhibited relatively high levels of diversity, no reintroduced populations had significantly lower HE (d.f. = 11, P > 0.05) or AR21 (d.f. = 11, P > 0.05) values after correcting for multiple comparisons (Table 2). Furthermore, pairwise comparisons between all reintroduced populations did not uncover a significant disparity in HE (d.f. = 11, P > 0.05) or AR21 (d.f = 11, P > 0.05) after a sequential Bonferroni correction. As expected, diversity measures (HE, Ho, and AR21) were highest in the source population, closely followed by the AR and OR populations (Table 2). Contrary to our predictions, however, the VR population displayed the lowest HE and 2nd lowest HO and AR21.

View this table:
Table 2

Genetic diversity in the northern Yellowstone National Park source herd (nYNP) and reintroduced elk populations in western North America: Chesnimnus, Oregon (OR), northwestern Arizona (AR), Theodore Roosevelt National Park (TR), Wichita Mountain Wildlife Refuge (WM), and Vermejo Park Ranch (VR). Sample size (n), observed heterozygosity (HO), expected heterozygosity (HE), allelic richness (AR21), number of unique alleles (U), and number of rare alleles (R) are reported for each population.

PopulationnHO(SD)HE (SE)AR21 (SE)URa
nYNP230.61(0.11)0.60 (0.08)4.04 (1.23)319
AR400.55(0.11)0.59 (0.06)3.75 (0.72)415
OR270.57(0.06)0.60 (0.08)3.76 (0.85)213
WM430.49(0.13)0.52(0.11)3.32 (0.95)010
VR340.50(0.18)0.51 (0.14)3.40 (1.16)214
TR220.52(0.13)0.55 (0.10)3.57 (1.00)313
  • a Rare alleles <10% frequency.

The number of rare and unique alleles per population revealed patterns similar to measures of heterozygosity and allelic richness. Among reintroduced populations, the number of rare and unique alleles (Table 2) was highest in AR (n = 15 and 4, respectively), and lowest in WM, which had 10 rare alleles and no unique alleles. The nYNP source herd, in comparison, had 1 fewer unique allele (n = 3) than the AR elk but 4 additional rare alleles (n = 19). It is important to note that although WM elk had no unique alleles and roughly half the number of rare alleles as nYNP elk, it also had the largest sample size (n = 43) of all populations.

In contrast to the homogeneity of genetic diversity, we observed low to moderate population differentiation among the herds we sampled. Fisher's exact test revealed that allele frequency distributions differed significantly among populations (P < 0.001), and the global θ (0.05) also was significant (P < 0.001), supporting differentiation. Pairwise θ values for source-reintroduced population combinations indicated relatively low (θ = 0.032–0.061), but significant pairwise differentiation between nYNP and WM, VR, and TR elk (P = 0.002 for each 0; Fig. 3). Pairwise DS values followed the same trend as the 0 estimates (Fig. 3).

Fig. 3

Genetic differentiation (0) and genetic distance (Ds) between the source population (Yellowstone National Park) and 5 reintroduced elk populations in western North America. Elk in Theodore Roosevelt National Park (TR) were translocated from Wind Cave National Park, which was established with elk from the northern Yellowstone National Park (nYNP) herd. Significant differences (P < 0.05) from panmixia (θ = 0) according to permutation tests with Bonferroni corrections are indicated by an asterisk (*).

Allele frequency distributions differed among the 5 reintroduced populations when the nYNP source population was excluded from analysis, indicating some differentiation among reestablished populations. All θ values for reintroduced population pairs were low to moderate and significant (P < 0.05), ranging from 0.02 for AR–TR elk to 0.08 for AR–VR elk (Table 3). Genetic distances mirrored the θ values. Although allele frequencies differed statistically among populations, the relatively low level of differentiation and the short genetic distances between reintroduced population pairs suggested that only modest genetic divergence has occurred since the elk populations were reestablished.

View this table:
Table 3

Genetic differentiation (θ) and genetic distance (DS) between pairs of reintroduced elk populations in western North America: Chesnimnus, Oregon (OR), northwestern Arizona (AR), Theodore Roosevelt National Park (TR), Wichita Mountain Wildlife Refuge (WM), and Vermejo Park Ranch (VR). Pairwise estimates of θ are above the diagonal and pairwise DS values are below the diagonal.

  • * Indicates significance (P < 0.05) using permutation tests and Bonferroni corrections.


Wildlife translocations can be used to augment demographically or genetically depauperate populations through supplementation, or to reintroduce species into locales from which they have been extirpated. However, one potential risk of rein-troductions is a loss of genetic variability associated with small founding group size, a situation that was particularly common with historic reintroduction events (Berry 1986; Maudet et al. 2002; Nei et al. 1975). Studies examining isolated, reintroduced populations have detected significant reductions in genetic variability for some species such as fishers (Martes pennantiWilliams et al. 2000), Merriam's turkeys (Meleagris gallopavo merriamiMock et al. 2004), bighorn sheep (Ovis canadensisHedrick et al. 2001), and moose (Alces alcesBroders et al. 1999). Populations of other mammals, including, sea otters (Enhydra lutrisLarson et al. 2002) and white-tailed deer (Odocoileus virginianusDeYoung et al. 2003) seem not to have suffered any major losses in genetic diversity after reintroduction. Our study provided the opportunity to evaluate the genetic consequences of past reintroduction efforts on current patterns of genetic variability in a species that experienced widespread extirpation and subsequent population reestablishment across it's historical range.

Levels of genetic diversity within the elk populations we studied (HE, = 0.55, range = 0.51–0.60) were comparable to levels documented in other closely related species. Values were slightly lower than those reported for other cervids, such as Norwegian reindeer and Canadian caribou (Rangifer tarandus; HE, ≈ 0.69 and 0.75, respectively—Cote et al. 2002; Wilson et al. 1997) and white-tailed deer (HE, ≈ 0.71—DeYoung et al. 2003), but similar to diversity levels described for nonbottlenecked populations of Rocky Mountain elk in Canada (HE ≈ 0.48—Polziehn et al. 2000) and the United States (HE ≈ 0.56—Williams et al. 2002). Data for other subspecies of elk in North American, C. e. nannodes (tule elk; HE = 0.18) and C. e. elaphus (Roosevelt elk; HE = 0.34), revealed lower levels of genetic diversity than those observed for C. e. elaphus in this study (Polziehn et al. 2000; Williams et al. 2004).

The 2 Yellowstone source herds (nYNP and sGYE) from which most translocations of Rocky Mountain elk in North America have originated displayed comparable genetic variability and no significant population differentiation. High levels of gene flow identified through both direct and indirect measures reinforce a lack of genetic substructure within the GYE. Previous research on movements of tagged elk identified seemingly low dispersal among herds in the GYE with exchange between herds being inversely proportional to geographic distance between their winter ranges (Boyce 1989; Houston 1982). Consequently, dispersal was lowest between the nYNP and sGYE herds (≈ 260 km between winter ranges) with less than 1% of 358 tagged elk from northern Yellowstone emigrating to the sGYE herd during 1924–1958 (W. H. Kittams, in litt.). Additionally, recent studies of radiocollared elk conducted during 1978–1984 (n = 97) and 1990–1994 (n = 164 juvenile males) indicated no dispersal of sGYE elk to the winter range occupied by the nYNP herd (Smith and Anderson 2001; Smith and Robbins 1994). Although the number of migrants estimated using indirect measures of gene flow (FST) should be interpreted with caution because underlying assumptions often are violated (Whitlock and McCauley 1999), the method provides a crude estimate of relative levels of gene flow. Given the relatively large number of estimated migrants per generation (Nm = [(1 – FST)] 4FST] ≈ 62) necessary to account for the nYNP–sGYE FST of 0.004, it seems unlikely that enough individuals moved between the herds to account for the large amount of gene flow observed.

Alternatively, or perhaps in concert with limited dispersal between herds, significant gene movement may result from interherd mating on shared summer ranges in southern YNP. It was estimated that 4,100 elk from the sGYE herd summered in southern YNP during the 1960s and more than 6% (44) of 786 sightings of nYNP elk occurred south and east of Yellowstone Lake during this same time period (Cole 1969; Craighead et al. 1972). Moreover, Smith and Robbins (1994) recently estimated that 28% of the elk wintering on the National Elk Refuge summer in YNP. Intermingling of the nYNP and sGYE herds on common summer habitat occurs through the rut (i.e., mating season) in mid-September and into October and November when the herds begin to migrate back to their respective winter ranges (Boyce 1989; Craighead et al. 1972; Houston 1982). Therefore, it is likely that mating between individuals from both herds is at least partially responsible for the similarities in genetic diversity and low divergence associated with the nYNP and sGYE herds.

We predicted that the GYE elk, which served as the source for translocations, would exhibit higher levels of genetic diversity than the reintroduced populations. Although not statistically significant after corrections for multiple comparisons, this trend was apparent in all measures of genetic diversity (Table 2).

Additionally, the nYNP population exhibited the greatest number of rare alleles among the populations we sampled.

Based on number of founders and estimated growth of the reintroduced populations following establishment, we predicted that the WM herd would exhibit relatively low diversity. The WM population indeed exhibited the lowest levels of variation (AR21 and HO), and 2nd lowest HE among populations that we sampled (Table 2). However, levels of genetic diversity did not differ significantly from the other populations we sampled and were not as low as those observed in reintroduced elk populations with similar founding histories. A Pennsylvania elk herd reestablished with 34 founders from the GYE that remained small (<50 individuals) for 50 years postreintroduction had considerably lower levels of diversity (HE = 0.254, A = 1.9—Williams et al. 2002) than the WM herd (HE = 0.52, A = 3.42).

The relatively robust genetic diversity present in the WM population may be the result of genetic contributions of unknown individuals at the reintroduction site before trans-location or undocumented supplementations after reestablishment, but both of these possibilities seem unlikely. Initial records of the WM founding group were well documented, and biologists surveyed the Wichita Mountains area in the early 1900s before the reintroduction effort and found no elk in the area (Halloran 1963). Therefore, it is unlikely that enough elk were present in the Wichita Mountains to alter the genetic landscape of the founding group. Indeed, despite relatively higher diversity, the lack of private alleles and the low number of rare alleles in the WM population (Table 2) suggest that the population suffered a founder effect. Perhaps the most likely explanation for higher levels of diversity in the WM herd is more rapid population growth than originally postulated. Detailed population data are not available to retrospectively test this hypothesis, but some historic information suggests that an estimated 300 elk may have been present in the population by 1925 (Halloran 1963).

Contrary to our expectations, the VR population exhibited relatively low genetic diversity and relatively high divergence from the source population. In addition, VR elk were most differentiated from the GYE elk, which may have been influenced by genetic contributions from 12 founding individuals from Colorado (Table 1). Recent research with reintroduced populations of wild turkeys has demonstrated that mixtures of founding individuals from genetically differentiated source populations can leave measurable genetic legacies for many decades (Latch and Rhodes 2006). However, statistically significant differences in θ should frequently occur because in most scenarios there is some deviation from absolute panmixia (Waples 1998). In fact, given the time since separation (in some cases 93 years, ≈ 18 elk generations) and assumed isolation of the reintroduced populations sampled, it is surprising that such little differentiation was documented. Although we did not detect much divergence in neutral genetic variation, it is possible that differences in quantitative traits may be more pronounced, especially if selection for those traits varied among the reintroduced populations. Variation in neutral markers only poorly reflects adaptive differences among the populations (Reed and Frankham 2001).

Regional gene flow in the western United States may contribute to the homogenous levels of diversity and relatively low differentiation among our sampled populations. Results from genetic analyses of reintroduced elk populations (YNP source) in the northern Rocky Mountains of Canada revealed similar genetic distances (DS) to those documented in our study; gene flow is thought to occur between those populations, albeit limited between some locales (Polziehn et al. 2000). For example, frequent movements of tagged elk between Kootenay and Yoho national parks, Canada, have been documented (Gibbons 1978), and the genetic distance between those herds was 0.043 (Polziehn et al. 2000), which is equal to or greater than genetic distances observed between AR (northwestern Arizona), OR (northeastern Oregon), and TR (western North Dakota) populations (Table 3). Although, few empirical data exist concerning long-distance movements or dispersal of elk, individuals have been documented moving hundreds of kilometers within a few days; 1 radiocollared male from Montana was ultimately located in the Kansas City, Missiouri, area after traveling >1,500 km (Wisdom and Cook 2000). Although such records likely represent extreme movements, it is possible that enough regional gene flow may be occurring to account for the low population differentiation and short genetic distances observed among some of the sampled reintroduced populations in this study.

Although regional gene flow is a plausible explanation for the high genetic continuity among the reintroduced populations in our study, it also is possible that not enough time (21–93 years; ≈ 5–18 elk generations) has elapsed since the reintroductions for drift to have significantly altered population allele frequencies. For example, the 2 most recently established populations (AR and TR) exhibited the lowest levels of divergence among the reintroduced populations we studied, and those populations were among the most similar to the source population. Additionally, population growth after the reintroduction events also may have been rapid enough to limit the effects of genetic drift.

Translocations to restore elk within their historic range are on-going, and several reintroductions have been conducted in the past 10 years (e.g., Kentucky, Wisconsin, and Ontario, Canada—Larkin et al. 2001). Our results imply that postreintroduction growth and subsequent gene flow may play important roles in maintaining genetic variability in reintroduced populations. Isolation and limited growth have particularly impacted eastern populations of elk in Ontario and Pennsylvania, which displayed very low levels of genetic diversity. On the contrary, a gene flow matrix in the western United States may help to maintain genetic variation and limit divergence among elk populations.


This research was supported by the Rocky Mountain Elk Foundation, University of Idaho, and Purdue University. We are grateful to the many people who provided DNA, tissue, and blood samples including J. Baker, V. Coggins, K. Gunther, D. Hawk, M. Oehler, D. Smith, and D. Walter. We appreciate the technical assistance of C. Anderson, L. Murfitt, and C. Watson. D. Onorato and C. Miller provided valuable assistance with the analyses. Our paper was improved by comments and insights shared by J. Aycrigg, J. Horne, P. Joyce, A. Metge, C. Miller, D. Onorato, J. Peek, and D. Roon.


  • Associate Editor was Carey Krajewski.

Literature Cited

View Abstract