Rangewide studies of genetic parameters can elucidate patterns and processes that operate only over large geographic scales. Herein, we present a rangewide population genetic assessment of the eastern box turtle Terrapene c. carolina, a species that is in steep decline across its range. To inform conservation planning for this species, we address the hypothesis that disruptions to demographic and movement parameters associated with the decline of the eastern box turtle has resulted in distinctive genetic signatures in the form of low genetic diversity, high population structuring, and decreased gene flow. We used microsatellite genotype data from (n = 799) individuals from across the species range to perform two Bayesian population assignment approaches, two methods for comparing historical and contemporary migration among populations, an evaluation of isolation by distance, and a method for detecting barriers to gene flow. Both Bayesian methods of population assignment indicated that there are two populations rangewide, both of which have maintained high levels of genetic diversity (HO = 0.756). Evidence of isolation by distance was detected in this species at a spatial scale of 300 – 500 km, and the Appalachian Mountains were identified as the primary barrier to gene flow across the species range. We also found evidence for historical but not contemporary migration between populations. Our prediction of many, highly structured populations across the range was not supported. This may point to cryptic contemporary gene flow, which might in turn be explained by the presence of rare transients in populations. However these data may be influenced by historical signatures of genetic connectivity because individuals of this species can be long-lived.
Citation: Kimble SJA, Rhodes Jr. OE, Williams RN (2014) Unexpectedly Low Rangewide Population Genetic Structure of the Imperiled Eastern Box Turtle Terrapene c. carolina. PLoS ONE 9(3): e92274. https://doi.org/10.1371/journal.pone.0092274
Editor: Adam Stow, Macquarie University, Australia
Received: January 18, 2013; Accepted: February 20, 2014; Published: March 19, 2014
Copyright: © 2014 Kimble et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Support for this project was provided by the Indiana Department of Natural Resources, Division of Fish and Wildlife, Wildlife Diversity Section, State Wildlife Improvement Grant E2-08-WDS15; Indiana Division of Forestry Grant E-9-6-A558; and the Department of Forestry and Natural Resources, Purdue University. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The quantification of genetic parameters of species is basic to understanding their natural history. This is especially important in declining species in need of appropriate conservation approaches . Species in decline, however, have by definition undergone demographic reductions , which may confound our ability to differentiate between natural and anthropogenically induced changes in genetic parameters. For example, when studying long-lived species investigators must distinguish between pre-disturbance (and therefore presumably stable) and anthropogenically induced genetic patterns to inform management strategies , . Furthermore, population genetic studies involving declining species are often confined to drawing broad management conclusions from limited data on a few individuals or populations, resulting in deviations from analytical assumptions which can negatively affect the reliability of results .
Wide-ranging species of conservation concern, whose interpopulation migration patterns are influenced by habitat fragmentation, also could be expected to exhibit genetic evidence of disruption to large- and small-scale movement behaviors . Such species are often characterized by high dispersal (intrapopulation movement) and migration (interpopulation movement) habits, and thus are expected to be vulnerable to habitat fragmentation , . Unfortunately, the genetic attributes of many declining, wide-ranging species are poorly studied over large spatial scales, despite the fact that many are marked for conservation management planning.
A genetic population is commonly defined in the field as a group of conspecifics that are genetically similar and are to varying degrees separated from other populations of conspecifics and are likely to be more locally adapted . Populations of long-lived, wide-ranging species should be managed at geographic scales that are appropriate for conservation planning at multiple levels of biological resolution , . Management plans conducted at geographic scales significantly smaller than that of the population may fail to incorporate mechanisms that maintain genetic diversity , while those conducted at scales larger than that of the population may lead to the loss of locally adapted genes . For example, anthropogenic interpopulation movement of genotypes may introduce alleles that are locally maladaptive in the receiving population , . In addition, rangewide approaches are tremendously useful for identification of significant management units, i.e., genetic populations for which management plans should be made for species or populations of conservation concern ,  even those with low population differentiation . Determination of the appropriate scale for management of turtles, in particular, is of paramount importance  because approximately 40% of Chelonians worldwide are considered endangered or vulnerable , . Simply removing local sources of extrinsic stressors, without regard to range-wide reservoirs of genetic resources, may not be sufficient mitigation against loss of genetic diversity because future threats such as climate change and novel disease may overwhelm genetically depauperate species , .
The eastern box turtle Terrapene c. carolina is a declining species whose conservation plans are difficult to develop because of the complexities of studying a species that is cryptic, long-lived (>100 years in the wild; N. Karraker, unpublished data; ), and whose generations overlap. It is a terrestrial turtle species that historically ranged across much of the eastern United States , but which has suffered substantial demographic declines –, likely due to some combination of habitat destruction and fragmentation, road mortality, collection, and disease . Understanding the effects of demographic declines on patterns of genetic diversity and structure in box turtles is a necessary precursor to developing management strategies for their protection, but previous studies suggest that such studies should be conducted at scales larger than single states –.
Our intent here was to explore range-wide genetic patterns of the eastern box turtle as a means to inform conservation planning for this species and provide a model for other species with similar traits and demographic histories. To accomplish this we first tested the hypothesis that genetic isolation by distance is significant across the species range due to the limited migration ability of this species. Second, we tested the hypothesis that habitat loss and fragmentation has formed multiple, geographically discrete and genetically differentiated populations across the species range. Because habitat reduction increases distance among patches of suitable habitat, we also tested a third hypothesis that increasing isolation due to habitat fragmentation has caused a reduction in the number of migrants among populations.
We conducted searches for eastern box turtles via visual encounter by car and on foot across much of their range (Figure 1). We sampled at multiple geographic scales to determine at what spatial extent populations occur. To this end, we sampled at eight sites in Indiana (a state whose forests are heavily fragmented by agriculture; ), in the four states surrounding Indiana (Illinois, Kentucky, Ohio, and Michigan), and from another nine states across the species range. We intentionally avoided sampling in the south and southwestern parts of the range where the eastern box turtle is sympatric with two other subspecies of T. carolina  to avoid confounding results with alleles from different subspecies. For each individual we recorded UTM coordinates, morphometric data, sex, activity, and any unusual markings or signs of injury or disease. We took tissue samples for genetic analysis, usually blood (∼10μL), following the protocol of Kimble and Williams . We assigned each turtle a unique number and filed a corresponding pattern of notches into the marginal scutes  to identify recaptures. These notches were subsequently sealed with surgical adhesive. Individuals were processed as quickly as possible and released immediately at the point of capture.
Due to the resolution, some marks represent more than one sample. Range data after Dodd (2001).
All animals we handled were done so in accordance with the Purdue Animal Care and Use Protocol 07-037 and amendments thereto. For all animals we handled we obtained all relevant permissions and permits from the appropriate government agencies, land trusts, and property managers before sampling began. We sampled in the Chattahoochee National Forest area of Georgia under permits from the USDA Forest Service Chattahoochee-Oconee National Forest office and Georgia Department of Natural Resources, Wildlife Resources Division; in Illinois under a permit from the Illinois Department of Natural Resources; in Indiana under permits from the Indiana Department of Natural Resources, Fish and Wildlife Division and Division of Nature Preserves and under permissions from NICHES Land Trust and Wabash College; in New York under a permit from the New York Department of Environmental Conservation, Division of Fish, Wildlife and Marine Resources; in North Carolina under a permit from the North Carolina Wildlife Resources Commission, Division of Wildlife Management; and in Ohio under the Ohio Department of Natural Resources, Division of Wildlife. Samples from other states were collected by collaborators working under their own permits.
We digested tissue samples using a modified proteinase K protocol and extracted DNA with a phenol-chloroform-isoamyl alcohol . We resuspended purified DNA in 50 μL of TLE (10 mM tris, 0.1 mM EDTA, pH 8.0) and quantified DNA concentration on a spectrophotometer (NanoDrop 8000, Thermo Scientific, Wilmington, DE). We then diluted all DNA samples to 20 ng/μL in pure water prior to PCR.
We carried out PCR using 11 microsatellite loci developed specifically for the eastern box turtle . We combined the 11 loci into three multiplexes and two singletons (Table 1). All reactions contained 60 ng DNA template, 10 mM tris-HCl, 0.05 mg/mL BSA, 50 mM KCl, 0.9 mM MgCl2, 0.2 mM of each dNTP, 0.3 U of Taq polymerase, and multiplex-specific concentration of end-labeled fluorescent primers in a total reaction volume of 10 μL. We analyzed all PCR products on an automatic sequencer (ABI 3730XL, Applied Biosystems, Foster City, CA). We automatically scored genotypes with Genemapper (version 3.7, Applied Biosystems, Foster City, CA) and checked each call manually at least twice. We reamplified ∼10% of all genotypes to ensure repeatability and reamplified any that disagreed a third time. We used Cervus (version 3.0.3; ) to check for accidentally duplicated samples and in the case of duplication removed all but the first sample taken from an individual.
To test for the presence of null alleles and large allelic dropout, we used Microchecker (version 2.2.3; ). To assess how well our dataset and each population met Hardy-Weinberg equilibrium (HWE) assumptions, we used the web version of Genepop (version 4.0.10; ) to test for a significant deficit of heterozygotes using an exact method  with default parameters. We used a Markov chain Monte Carlo (MCMC) method  to assess significant deviations from HWE. We quantified deviations from Hardy-Weinberg equilibrium (HWE) by estimating FIS in Genepop with 10,000 iterations.
Isolation by distance
We evaluated rangewide isolation by distance by using a Mantel test  performed in Alleles in Space (Ais, version 1.0; ) to determine whether a significant correlation existed between pairwise matrices of Nei's  genetic distances and geographic distances matrices. We log10-transformed the geographic distance matrix to meet the assumption of normality  and used 1,000 randomized replicates to assess significance. To estimate the geographic scale at which IBD may begin to operate, we also performed a complementary spatial autocorrelation analysis in Ais, using 10 classes of equal distance and 1,000 replicates to assess significance.
Determination of population number, individual assignments, and barriers
To improve analytical robustness against violations of software-specific assumptions, a common approach in population genetics , we used two Bayesian analyses to infer the size, shape, and individual membership of populations. The first approach was in Geneland (version 4.0.0; ) in program R  without spatial priors. We set Geneland to search for the most likely number of populations (k) from 1–25 over 1,000,000 MCMC iterations with thinning every 100th iteration and a burn-in of 50,000. We assumed uncorrelated allele frequencies to avoid artifacts of uneven geographic sampling  and the potential for overestimation of k associated with the use of correlated frequency allele models . We set the Poisson process maximum to 800 and the maximum number of nuclei allowed for the Poisson-Voronoi tessellation to 2,400 , . We ran the MCMC 10 times for each value of k and used the highest mean probability density value as the inferred k value. We also estimated the probability of population assignment for each individual.
To corroborate results, we used the Bayesian algorithm Structure (version 2.3; ). We performed 10 independent runs for each value of k from 1 to 25 for 1,000,000 iterations, including a 50,000 iteration burn-in. We ran the model with default settings except we used an admixture model, and with allele frequencies uncorrelated (as with Geneland). To visualize and infer the most likely value of k we used Structure Harvester (version 0.6.92; ), which employs the Δk method of Evanno et al . Structure also assigns a probability value to each the population assignment of each individual. For both Geneland and Structure we analyzed each population individually for substructuring. We compared the results of both algorithms and used the population assignment with the highest confidence per individual for further analyses. Finally, we compared the individuals assigned to each population by both Geneland and Structure for agreement. We tested resulting genetic populations for violations of HWE and dropped loci with relatively high null allele estimates as necessary.
To estimate the location of specific natural or anthropogenic barriers to gene flow which might have contributed to population structure identified across the species range, we used the maximum difference algorithm of Monmonier , which has recently been applied to landscape genetics , –. This process builds a connectivity network of the sample locations using Delaunay triangulation, and then estimates the barriers among them by following contiguous connectivity links between samples that represent the highest genetic distances. We used this method in Ais using genetic distances corrected for geography (“pseudoslopes”). We set the number of barriers to k - 1, with k being the number of populations identified by Geneland and Structure.
Previous studies in the population genetics of Terrapene species suggest that populations may operate at geographic scales greater than most management jurisdictions, such as state or national parks, and perhaps even larger than a single state , , . If this is the case, managers of local sites cannot manage for the entire population but instead must work with the genetic reality of the individuals under their control. To this end, we also report results for each management unit in which we sampled at least nine individuals (Table S1).
Quantification of migration among populations
We used the genetic population clusters identified by the Bayesian analyses to detect recent migration among them using Bayesass (version 3; ), a genotype-based Bayesian platform appropriate even for populations that do not meet assumptions of HWE. We ran 1,000,000 iterations with a burn-in of 100,000 and sampling intervals of 100. We set the mixing parameters for the MCMC chain to 0.02 for migration, 0.1 for inbreeding, and 0.05 for allele frequencies so that the resulting migration parameter swapping acceptance rates were between 20% and 40% as suggested by the authors. We ran the analysis five times to check for convergence and visualized chain mixing, convergence, and burn-in values in Tracer (version 1.5; ).
We used Migrate (version 3.1.1; ) to estimate historical migration rates (among populations identified by the Bayesian clustering analyses). Under the maximum likelihood (ML) framework, we used a Brownian motion model of mutation, suitable for microsatellite data that likely do not adhere to a strictly stepwise model of mutation. We used five independent replicates of 10 short chains 10,000 iterations in length, three long chains 100,000 iterations in length, and four heated static chains at temperatures 1.0, 1.5, 3.0 and 10,000.
Sample collection and laboratory
We collected tissue samples from 1,603 wild eastern box turtles from across much of the species range (Figure 1). We successfully resolved all quality control disagreements among PCR amplifications and between scorers and excluded all individuals with more than three loci for which genotype data could not be resolved (n = 45). Thus, the final data set included 1,558 individuals. For all genetic analyses, we randomly selected 24 individuals from the two locations where we sampled deeply for other purposes (the Hardwood Ecosystem Experiment in south-central Indiana: n = 627; and Oak Ridge, Tennessee: n = 182), to even the sampling distribution across sites. This resulted in a total of 799 individuals from which data were used for all analyses of genetic structure, gene flow and IBD.
Mean allelic richness from the sample sites ranged from 7.6 to 33.6 (Table S1). When all 799 samples were pooled, there was a significant deviation from Hardy-Weinberg equilibrium across all 11 loci (χ2 = ∞, df = 22, p<0.001). Microchecker estimated the potential presence of numerous null alleles at low frequencies (Table S1). All but two of the management unit populations (e.g., Patuxent Wildlife Research Center in Maryland) were also significantly out of equilibrium, but these two have low sample sizes (9 and 12). A few loci suffered from heterozygote deficiencies in some populations (Table S2).
Isolation by distance
Isolation by distance was significant across the species range (Figure 2; r = 0.13, p<0.001). Additionally, the spatial autocorrelation analysis estimated that the pairwise genetic distance in the 300 – 500 km distance class (Ay = 0.795) exceeded the mean pairwise genetic distance for all pairwise distances (Av = 0.793). This indicates that significant isolation begins to operate at this geographic distance and that panmixia operates at geographic scales less than this threshold , .
Determination of population number, individual assignments, and barriers
The nonspatial model Geneland reported k = 2 in 7 of 10 runs. One of the runs reporting k = 2 also had the highest mean of probability density value and so we used this as the point estimate of the number of populations. Secondary runs on each of the two clusters revealed no substantial substructuring, i.e., population assignment probabilities were low. The final map (Figure 3) shows the population boundary roughly following the spine of the Appalachian Mountains. The nonspatial model Structure also indicated that k = 2 according to the point value estimation method of Evanno et al . More than 85% of individuals were assigned a population assignment probability of >0.9. Secondary analyses on these two clusters also showed no substantial substructuring, i.e., all Δk values were low. Individuals with admixed ancestry did not tend to cluster along the boundary between the Western and Eastern populations, a signature of a zone of interbreeding. Structure and Geneland assigned 95.6% of individuals to the same two population groups.
The probability map of an individual turtle belonging to the Western population was generated by Geneland and increases with darker shades of gray. The probability of an individual belonging to the Eastern population would be proportionally opposite. Note that the border between the populations follows the Appalachian Mountains.
There was a nearly universal lack of Hardy-Weinberg equilibrium, despite the fact that the Bayesian methods for population delineation construct populations by minimizing deviations from HWE and linkage. This might be caused by technical reasons (e.g., null alleles; ), or by the violation of HWE assumptions that are not appropriate for box turtles ). These deviations may be explained by the high polymorphisms at our loci, which range from 16 to 83 alleles per locus (mean: 36.1) and which resulted in the detection of many rare alleles. As allelic richness increases at a locus, the likelihood that they will be found as homozygotes declines in a finite sample size of the population, increasing the likelihood of deviations from HWE . Furthermore, while HE and HO estimates vary by locus in each population, mean HE and HO estimates are similar across populations which suggests that departures from HWE are artifacts of the algorithms and do not compromise the interpretations of our data (Table S2). It has also been demonstrated that distinguishing between deviations from HWE and the presence of null alleles can be difficult, and the common tactic of excluding “problematic” loci may result in a loss of the most informative loci .
Quantification of migration among populations
Because both Structure and Geneland were strongly concordant, we coded Ais to find one Monmonier barrier between them. The resulting line was largely congruent with the map that Geneland generated, drawing the barrier along a line running north to south from Pennsylvania through Maryland, Virginia, North Carolina and South Carolina. We refer to the two resulting populations hereafter as “Western” and “Eastern”. Though both Geneland and Structure operate by minimizing linkage disequilibrium and departures from HWE, both populations were also significantly out of HWE (Table S1). Both the Western and Eastern populations had three loci with a null allele rate 0.10 – 0.12 but exclusion of these from both the Bayesian approaches and the test for HWE had not substantial effect on the results.
Bayesass gave no evidence for recent immigration between populations as estimates included zero in their credible intervals. By contrast, Migrate estimated a historical migration rate of 15.2 migrants/generation from the Western population into the Eastern, and 17.4 migrants/generation from East to West.
Life history traits of Terrapene species make the formulation of clear hypotheses about the geographic patterns of genetic populations difficult. Much of the natural history evidence for Terrapene suggests that members of this genus should maintain population structure at relatively small geographic scales. For example, one T. ornata individual was recaptured 27 times over ten years within 7.6 m of its initial capture location  and home range size for T. carolina has been reported between 0.02 and 187.67 ha , . Observational data suggest that juvenile dispersal may be very short distances, approximately 100 m or fewer , . Ultimately, evidence of low dispersal and highly conserved adult home ranges suggests that members of this genus should display a high degree of population structure at small spatial scales.
Alternatively, there is some evidence to the contrary. First, the existence of transient adult box turtles has been hypothesized due to the lack of recapture, despite intense effort spanning decades , , . Furthermore, individuals with transient behavior have been observed in two radiotelemetry studies (T. c. triunguis, ; S. Kimble, T. c. carolina, unpublished data.) where adult males traveled roughly linear paths that were many times longer than the width of the standard home range (∼10 km over two active seasons, ; ∼7 km over 1 active season, S. Kimble, unpublished data), and have been observed mating along the way . Second, T. carolina box turtles are not closely tied to bodies of water , and thus may suffer from fewer geographic constraints on gene flow than do aquatic turtles. Third, while little is known about juvenile Terrapene dispersal , parent-offspring pairs have been found up to 27.1 km apart [S. Kimble, unpublished data] suggesting some mechanism for higher gene flow than is currently appreciated.
Our data support the latter theory by indicating that there are only two populations across most of the range of the eastern box turtle. Furthermore the Appalachian Mountains may act or have acted as a barrier to gene flow at the continental scale, although eastern box turtles are currently known to inhabit all but the highest altitudes in North Carolina . The finding that populations operate at large geographic scales is supported by previous work in the Terrapene species complex. In T. ornata, individuals from two sites 120 km apart were found to be genetically panmictic . In Terrapene c. carolina, Marsack and Swanson  found that individuals separated by 30 to 70 km in southwestern Michigan also constituted a single population. Hagood  and Butler and colleagues  documented low genetic structure in T. c. carolina across larger distances of 160 km and 250 km, respectively. These studies suggest that the approach taken in this study, evaluating population genetic patterns at the scale of the species range, is the appropriate approach for the eastern box turtle.
Our data also indicate that isolation by distance is operating over relatively large spatial scales across the range of this species. The spatial autocorrelation analysis demonstrated that at approximately 300 – 500 km, mean pairwise genetic distances begin to exceed the average for the entire data set, suggesting a geographic extent at which populations operate. The same analysis by Hagood  in T. c. carolina returned a similar result of 450 to 650 km, which is approximately the distance from the Appalachian Mountains to the edge of the species range. These results, combined with the Geneland and Structure results, describe a species that though apparently highly philopatric, has (or had) high gene flow across vast areas. Though box turtles are reported throughout much of Appalachia , cryptic barriers to gene flow such as terrain and elevation may cause subtle population barriers in turtles  that may result in the Appalachians serving as a modest barrier.
Overall, we detected little evidence that habitat fragmentation is so far affecting population genetic structure in eastern box turtles. The shape and scope of the populations appear to be more consistent with a historical landscape than with current patterns of landscape fragmentation (Fig 1). Individuals from as far distant as eastern Tennessee and southwestern Michigan were assigned to the same population and few private alleles were detected from samples across these geographically distant populations (Table S1). The exception was possibly a signal of incipient decline: we did find evidence that migration between the two rangewide populations has recently been reduced or eliminated, a result expected in the presence of increasing habitat fragmentation. However, generation times can be long in Terrapene, with longevity in the wild at least 100 years (N. Karraker, unpublished data; ). As few as three generations could have passed since European settlers started clearing large swaths of forests in the range of the eastern box turtle , suggesting the idea that the genetic signatures we see in Terrapene box turtles may be largely historical. The loss of migration events between the two populations may be the first signal of reduced gene flow.
Furthermore, though we do not know what historical levels were, genetic diversity appears to remain high. Allelic richness and observed heterozygosity are high (Table S1). Long-lived species that have experienced demographic declines yet retain high genetic diversity include the Nile crocodile  and the harpy eagle . However, while the former has experienced population growth under the protection of a CITES listing, the latter has not exhibited any such rebound. To date, all long-term demographic studies of Terrapene have documented declines –, suggesting that box turtles require more active management to further prevent declines and extirpations.
This work represents the first rangewide study of population genetics patterns in the Terrapene carolina complex and provides insights into the ecology of a subspecies that appears to have only begun to exhibit the effects of habitat fragmentation within the last few generations. Indeed, our data indicate that genetic diversity remains high across the range of this Terrapene subspecies, yet the detection of historical gene flow and the lack of recent immigration events are signatures of recently reduced population connectivity. The loss of genetic diversity is a major threat to chelonian species worldwide , . However, in long-lived species with overlapping generations, signatures of genetic loss may be masked for decades or even centuries . Turtles are some of the longest-living vertebrates on the planet and yet many have suffered severe demographic declines, necessitating immediate management plans.
Rangewide population genetic parameter values for the eastern box turtle Terrapene. c. carolina.
We wish to thank the many individuals who generously helped collect eastern box turtle tissue samples from across the species range, including M. Allender, M. Baragona, J. Beane, K. Buhlmann, N. Burgmeier, J. Butler, D. Carlson, V. Clarkston, K. Creely, M. Cook, M. Cross, T. Despot, A. Durso, N. Engbrecht, E. Estabrook, J. Faller, S. Foertmeyer, A. Garcia, B. Geboy, S. Hagood, J. Hall, K. Hanauer, C. Hennessy, P. Henry, A. Hoffman, J. Greathouse, T. Green, J. Groves, L. & M. Jackson, T. Jedele, B. Johnson, S. Johnson, L. Keener-Eck, J. Kissel, A. Krainyk, S. Klueh, V. Kinney, N. Levitte, K. Lilly, J. MacNeil, J. Mitchell, T. Mitchell, K. Norris, J. Petranka, H. Powell, K. Powers, S. Price, J. Richards, J. Riegel, S. Ritchie, J. Shuey, B. Shaffer, P. Spinks, G. Stephens, C. Szwed, C. Tabaka, B. Thompson, B. Tomson, C-C. Tsai, T. Tuberville, M. Turnquist, B. Weigel, K. Westerman, M. Wildnauer, L. Woody, and the staff of the Woodlands Nature Station at Land Between the Lakes National Recreation Area. We wish to thank all members of the Williams lab for valuable help and willingness in improving the manuscript. R. Burke, G. Dharmarajan, K. Dodd, J. Fike, R. Howard, G. Nyberg, N. Karraker, S. Klueh, M. Kremer, M. Lannoo, N. Lichti, Z. Olson, K. Smith, B. Pauli, M. Pochon, J. Robb, and E. Latch also gave important support for the completion of this project.
Conceived and designed the experiments: SJAK OER RNW. Performed the experiments: SJAK. Contributed reagents/materials/analysis tools: OER RNW. Wrote the paper: SJAK. Conceived of the ideas: SJAK OER RNW. Collected and analyzed the data: SJAK.
- 1. Frankham R, Ballou JD, Briscoe DA (2002) Introduction to Conservation Genetics. Cambridge University Press. Cambridge. 642 p.
- 2. Whitlock MC, McCauley DE (1999) Indirect measures of gene flow and migration: FST≠1/(4Nm+1). Heredity 82: 117–125.
- 3. Avise JC, Ball RM, Arnold J (1988) Current versus historical population sizes in vertebrate species with high gene flow: a comparison based on mitochondrial DNA lineages and inbreeding theory for neutral mutations. Mol Biol Evol 5: 331–344.
- 4. Kuo C-H, Janzen FJ (2004) Genetic effects of a persistent bottleneck on a natural population of Ornate Box Turtles (Terrapene ornata). Conserv Genet 5: 425–437.
- 5. Lowe WH, Allendorf FW (2010) What can genetics tell us about population connectivity? Mol Ecol 19: 3038–3051.
- 6. Ewers RM, Didham RK (2006) Confounding factor in the detection of species responses to habitat fragmentation. Biol Rev 81: 117–142.
- 7. Steen DA, Gibbs JP (2004) Effects of roads on the structure of freshwater turtle populations. Conserv Biol 18: 1143–1148.
- 8. Aresco MJ (2005) Mitigation measures to reduce highway mortality of turtles and other herpetofauna at a north Florida lake. J Wildl Manage 69: 549–560.
- 9. Hartl DL, Clark AG (2007) Principles of Population Genetics, 4th edn. Sinauer Associates, Sunderland, Massachusetts.
- 10. Rousset F (1997) Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance. Genetics 145: 1219–1228.
- 11. Pearse DE, Arndt AD, Valenzuela N, Miller BA, Cantarelli V, et al. (2006) Estimating population structure under nonequilibrium conditions in a conservation context: continent-wide population genetics of the giant Amazon river turtle, Podocnemis expansa (Chelonia; Podocnemididae). Mol Ecol 15: 985–1006.
- 12. Nei M, Maruyama T, Chakraborty R (1975) The bottleneck effect and genetic variability in populations. Evol 29: 1–10.
- 13. Lesica O, Allendorf FW (1999) Ecological genetics and the restoration of plant communities: mix or match? Restor Ecol 7: 42–50.
- 14. McKay JK, Christian CE, Harrison S, Rice KJ (2005) “How local is local? – A review of practical and conceptual issues in the genetics of restoration. Restor Ecol. 13: 432–440.
- 15. Waples RS (1995) Evolutionary significant units and the conservation of biological diversity under the Endangered Species Act. Amer Fish Soc Symp 17: 8–27.
- 16. Alacs EA, Janzen FJ, Scribner KT (2007) Genetic issues in freshwater turtle and tortoise conservation. Chel Res Mon 4: 107–123.
- 17. Gibbons JW, Scott DE, Ryan TJ, Buhlmann KA, Tuberville TD, et al. (2000) The global decline of reptiles, déjà vu amphibians. Biosci 50: 653–666.
- 18. Cook RP, Brothernon DK, Behler JL (2010) Inventory of amphibians and reptiles at the Williams Floyd Estate, Fire Island National Seashore. Natural Resouce Report NPS/NCBN/NRTR-2010/380. U.S. Department of the Interior, National Park Service, Fort Collins, Colorado.
- 19. Dodd CK (2001) North American Box Turtles: A Natural History. University of Oklahoma Press, Norman, Oklahoma. 231 p.
- 20. Stickel LF (1978) Changes in a box turtle population during three decades. Copeia 1978: 221–225.
- 21. Williams EC, Parker WS (1987) A long-term study of a box turtle (Terrapene carolina) population at Allee Memorial Woods, Indiana, with emphasis on survivorship. Herpetol 43: 328–335.
- 22. Hall RJ, Henry PFP, Bunck CM (1999) Fifty-year trends in a box turtle population in Maryland. Biol Cons 88: 165–172.
- 23. Hagood S (2009) Genetic differentiation of selected Eastern Box Turtle (Terrapene carolina carolina) populations in fragmented habitats, and a comparison of road-based mortality rates to population size. PhD Dissertation, University of Maryland.
- 24. Marsack K, Swanson BJ (2009) A genetic analysis of the impact of generation time and road-based habitat fragmentation on Eastern box turtles (Terrapene c. carolina). Copeia 4: 647–652.
- 25. Butler JM, Dodd Jr CK, Aresco M, Austin JD (2011) Morphological and molecular evidence indicates that the Gulf Coast box turtle (Terrapene carolina major) is not a distinct evolutionary lineage in the Florida Panhandle. Biol J Linn Soc Lond 102: 889–901.
- 26. Miller BK (1993) Wildlife trends in Indiana. Hoosier Farmland Wildlife Notes 1: 2.
- 27. Kimble SJA, Williams RN (2012) Temporal variance in hematologic and plasma biochemical reference intervals for free-ranging eastern box turtle (Terrapene carolina carolina). J Wildl Dis 48: 799–802.
- 28. Ernst CH, Barbour RW, Hershey MF (1974) A new coding system for hardshelled turtles. J Ky Acad Sci 35: 27–28.
- 29. Sambrook J, Russell DW (2001) Molecular Cloning: A Laboratory Manual, 3rd edn. Cold Springs Harbor Press, Cold Springs Harbor, New York. 2344 p.
- 30. Kimble SJA, Fike JA, Rhodes Jr OE, Williams RN (2011) Identification of 12 polymorphic microsatellite loci for the eastern box turtle (Terrapene carolina carolina). Conserv Genet Resour 3: 65–67.
- 31. Kalinowski ST, Taper ML, Marshall TC (2007) Revising how the computer program Cervus accommodates genotyping error increases success in paternity assignment. Mol Ecol 16: 1099–1006.
- 32. Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) Microchecker: software for identifying and correcting genotyping errors in microsatellites. Mol Ecol Notes 4: 535–538.
- 33. Raymond M, Rousset F (1995a) Genepop (version 1.2): Population genetics software for exact tests and ecuminicism. J Hered 86: 248–249.
- 34. Raymond M, Rousset F (1995b) An exact test for population differentiation. Evol 49: 1280–1283.
- 35. Guo SW, Thompson EA (1992) Performing the exact test of Hardy-Weinberg proportion for multiple alleles. Biometrics 48: 361–372.
- 36. Mantel N (1967) Detection of disease clustering and a generalized regression approach. Cancer Res 27: 209–220.
- 37. Miller MP (2005) Alleles in Space (Ais): computer software for the joint analysis of interindividual spatial and genetic information. J Hered 96: 722–724.
- 38. Nei M (1972) Genetic distance between populations. Am Nat 106: 283–292.
- 39. Dharmarajan G, Beasley JC, Fike JA, Rhodes Jr OE (2009) Population genetic structure of raccoons (Procyon lotor) inhabiting a highly fragmented landscape. Can J Zool 87: 814–827.
- 40. Latch EK, Scognamillo DG, Fike JA, Chamberlain MJ, Rhodes Jr OE (2008) Deciphering ecological barriers to North American River Otter (Lontra canadensis) gene flow in the Louisiana landscape. J Hered 99: 265–274.
- 41. Guillot G, Mortier F, Estoup A (2005) Geneland: a computer package for landscape genetics. Mol Ecol Notes 5: 712–715.
- 42. R Development Core Team (2009) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.r-project.org.
- 43. Serre D, Pääbo S (2004) Evidence for gradients of human genetic diversity within and among continents. Genome Res 14: 1679–1685.
- 44. Falush D, Stephens M, Pritchard JK (2003) Inference of population structure using multilocus genotype data: linked loci and correlated allele frequencies. Genet 164: 1567–1587.
- 45. Dupanloup I, Schneider S, Excoffier L (2002) A simulated annealing approach to define the genetic structure of populations. Mol Ecol 11: 2571–2581.
- 46. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genet 155: 945–959.
- 47. Earl DA, vonHoldt BM (2012) Structure Harvester: a website and program for visualizing Structure output and implementing the Evanno method. Conserv Genet Resour 4: 359–361.
- 48. Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software Structure: a simulation study. Mol Ecol 14: 2611–2620.
- 49. Monmonier MS (1973) Maximum-difference barriers: an alternative numerical regionalization method. Geo Anal 3: 245–261.
- 50. Manel S, Schwartz L, Luikart L, Taberlet P (2003) Landscape genetics: combining landscape ecology and population genetics. Trends Ecol Evol 18: 189–197.
- 51. Palmé AE, Su Q, Rautenberg A, Manni F, Lascoux M (2003) Postglacial recolonization and cpDNA variation of silver birch, Betula pendula. Mol Ecol 12: 201–212.
- 52. Tsai Y-H E, Manos PS (2010) Host density drives the postglacial migration of the tree parasite, Epifagus virginiana. Proc Natl Acad Sci USA 107: 17035–17040.
- 53. Wilson GA, Rannala B (2003) Bayesian inference of recent migration rates using multilocus genotypes. Genet 163: 1177–1191.
- 54. Rambaut A, Drummond AJ (2007) Tracer v.5, available at beast.bio.ed.ac.uk/Tracer.
- 55. Beerli P, Felsenstein J (2001) Maximum likelihood estimation of a migration matrix and effective population sizes in n subpopulations by using a coalescent approach. Proc Natl Acad Sci USA 98: 4563–4568.
- 56. Wright S (1943) Isolation by distance. Genet 25: 114–138.
- 57. Sokal RR, Wartenberg DE (1983) A test of spatial autocorrelation analysis using an isolation-by-distance model. Genet 105: 219–237.
- 58. Lemer S, Rochel E, Planes S (2011) Correction method for null alleles in species with variable microsatellite flanking regions, a case study of the black-lipped pearl oyster Pinctada margaritifera. J Hered 102: 243–246.
- 59. Dharamarajan GD, Beatty WS, Rhodes OE Jr (2013) Heterozygote deficiencies caused by a Wahlund effect: dispelling unfounded expectations. J Wild Manag 77: 226–234.
- 60. Metcalf E, Metcalf AL (1970) Observations on ornate box turtles (Terrapene ornata ornata Agassiz). Trans Kans Acad Sci 73: 96–117.
- 61. Currylow AF, MacGowan BJ, Williams RN (2012) Short-term forest management effects on a long-lived ectotherm. Plos One 7(7): e40473.
- 62. Madden R (1975) Home range, movements, and orientation in the eastern box turtle Terrapene carolina carolina. PhD dissertation, City University of New York.
- 63. Burke RL, Capitano W (2012) Eastern box turtle, Terrapene carolina, neonate overwintering ecology on Long Island.New York. Chel Conserv Biol 10: 256–259.
- 64. Schwartz CW, Schwartz ER (1974) The three-toed box turtle in central Missouri: its population, home range, and movements. Part I. Missouri Department of Conservation, Jefferson City, Missouri.
- 65. Kiester AR, Schwartz CW, Schwartz ER (1982) Promotion of gene flow by transient individuals in an otherwise sedentary population of box turtles (Terrapene carolina triunguis). Evol 36: 617–619.
- 66. Richtsmeier RJ, Bernstein NP, Demastes JW, Black RW (2008) Migration, gene flow, and genetic diversity within and among Iowa populations of Ornate Box Turtles (Terrapene ornata ornata). Chel Conserv Biol 7: 3–11.
- 67. Latch EK, Boarman WI, Walde A, Fleischer RC (2011) Fine-scale analysis reveals cryptic landscape genetic structure in desert tortoises. Plos One 6: e27794.
- 68. Steyaert LT, Knox RG (2008) Reconstructed historical land cover and biophysical parameters for studies of land-atmosphere interactions within the eastern United States. J Geophys Res 113: 1–27.
- 69. Bishop JM, Leslie AJ, Bourquin SL, O’Ryan C (2009) Reduced effective population size in an overexploited population of the Nile crocodile (Crocodylus niloticus) Biol Cons. 142: 2335–2341.
- 70. Lerner HR, Johnson JA, Lindsay AR, Kiff LF, Mindell DP (2009) It's not too late for the harpy eagle (Harpia harpyja): high levels of genetic diversity and differentiation can fuel conservation programs. PLOS One 4: e7336.
- 71. Victory ER, Glaubitz JC, Rhodes Jr OE, Woeste KE (2006) Genetic homogeneity in Jugland nigra (Juglandaceae) at nuclear microsatellites. Am J Bot 93: 118–126.