Migration, Gene Flow, and Genetic Diversity Within and Among Iowa Populations of Ornate Box Turtles (Terrapene ornata ornata)
ABSTRACT
Like many fragmented reptile populations, the ornate box turtle (Terrapene ornata ornata) is located in isolated, often small, populations in eastern Iowa. If populations are to remain viable, genetic diversity within these populations must be maintained, which is done most efficiently by migration among populations. Population viability was accessed using 3 microsatellite loci to determine population genetic structure in 2 locally dispersed subpopulations of ornate box turtles. Although these subpopulations were determined to be 1 large population with the normal range of heterozygosity, further examination revealed evidence of genetic divergence from a once larger population that is now geographically separated into fragments. We concluded that the central population studied was genetically healthy, and with proper management that continues to promote gene flow, the population should remain viable in the near future.
The worldwide decline of amphibians is well documented (e.g., Lannoo 2005). Although declines among amphibians have received considerable media attention, reptilian declines also appear significant due to pollutants, invasive species, habitat loss, and increased predation along edges of fragmented habitats (Gibbons et al. 2000). Fragmented habitats, without regular gene flow between populations, also reduce genetic diversity within small, isolated populations, partially explained by metapopulation theory (Levins 1969).
Among reptiles, terrestrial turtles are limited in their dispersal abilities, which limit gene flow. Therefore, success of recolonization by terrestrial turtles after a population crash may depend on patch and/or corridor habitat quality, the distance separating subpopulations from the main population, or migration barriers such as rivers or roads (Lidicker and Koenig 1996). Turtle home ranges are often fragmented by highways and housing developments. In such fragmented conditions, marginal populations generally show reduced numbers and are expected to have lower average fitness (Weins 1997; Rubin et al. 2001). All of these conditions potentially apply to the ornate box turtle (Terrapene ornata ornata), which is protected in Indiana, threatened in Iowa, and endangered in Wisconsin (Swarth and Hagood 2005). All of these states contain isolated populations on the margins of their US range (Conant and Collins 1991).
In eastern Iowa, most ornate box turtle populations are scattered and small. However, the Hawkeye Wildlife Area (HWA) of Johnson County, Iowa has a relatively large population (Fig. 1). Two intensely studied subpopulations within the HWA (Mallard Pools and Greencastle) are separated by less than 1 km with no apparent barriers to immigration (Fig. 2). However, studies by Bernstein et al. (2007) indicated that ornate box turtles have a high degree of philopatry and that turtles do not travel between the 2 sites.
Citation: Chelonian Conservation and Biology 7, 1; 10.2744/CCB-0653.1
Citation: Chelonian Conservation and Biology 7, 1; 10.2744/CCB-0653.1
The HWA ornate box turtle population is one of the 2 largest in Iowa (over 600 turtles have been marked in the 75-ha area that contains the 2 subpopulations); the only other population that exceeds 100 individuals in eastern Iowa exists near Muscatine, approximately 120 km away (Fig. 1).
Herein, we analyze data from microsatellite loci to compare levels of genetic diversity among these populations to determine if there is genetic evidence to support the hypothesis of no migration between the HWA subpopulations.
METHODS
Turtles were marked with a modified Cagle (1939) system and located at HWA by walking through the sites at various times of day during April and early May. As described in Bernstein et al. (2007), HWA is largely composed of sandy habitats, which are required by ornate box turtles (Dodd 2001; Bernstein and Black 2005). Blood samples were collected from the Mallard Pool subpopulation (n = 47) and from the Greencastle subpopulation (n = 41). In addition, 14 samples were collected from Muscatine, Iowa, for comparison (Fig. 1).
DNA Extraction
Whole blood was extracted from the caudal vein in the tail using disposable 1-cc syringes. The blood was placed in lysis buffer (Seutin et al. 1991; Pearse et al. 2001) and stored at 20°C. As a backup, blood was absorbed onto filter-paper tabs, placed into 1.0-mL cryogenic storage tubes, and stored at −70°C. DNA was extracted using standard phenol-chloroform techniques (Hillis et al. 1990). When necessary, DNA was extracted from frozen filter tabs using Chelex 100 (Sigma, St. Louis, MO) following Walsh et al. (1991).
Microsatellite Analysis
If microsatellites are flanked by highly conservative sequences, primers may work on multiple, closely related species (Primmer et al. 1996). Therefore, polymerase chain reaction (PCR) primers reported by Pearse et al. (2001), who successfully amplified microsatellites from the confamilial painted turtle (Chrysemys picta), were chosen for this study. Likewise, PCR conditions also followed Pearse et al. (2001). The 3 loci examined were designated CP2, CP3, and CP10. PCR amplifications were carried out using a Primus thermal cycler (MWG-Biotech, High Point, NC). Both positive and negative controls were used in each PCR batch.
PCR products were size-fractionated at the Iowa State University Sequencing and Synthesis Facility. The program Genographer 1.6 (Benham 2001) was utilized for electromorph size determination and genotype scoring of individuals. All resulting chromatograms were also examined by eye to ensure accuracy of scoring (e.g., determine presence of stutter peaks, etc.) and multiple reference individuals were included in different runs to standardize size determinations between batches and to confirm reproducibility of results.
Because the nature of the tandem repeats was not reported by Pearse et al. (2001) one individual was directly sequenced (using the same PCR primers) for each locus using standard cycle-sequencing procedures at the DNA Sequencing and Synthesis Facility at Iowa State University. This sequencing confirmed that the microsatellites were dinucleotide repeats with a motif of CA. We did regenotype individuals with genotypes that fell outside the normal range.
Statistical Analysis
Three different population combinations were analyzed for each statistical test. In the first group, Mallard Pools, Greencastle, and Muscatine were analyzed as separate populations. The second group combined Mallard Pools with Greencastle to form the Hawkeye population, and this was compared with Muscatine. The third group compared Mallard Pools with Greencastle as 2 separate populations, leaving out Muscatine.
With the exception of calculations for RST, an FST analog based on allele size, all analyses were performed using Genepop version 3.3 (Raymond and Rousset 1995). Expected and observed heterozygosities and homozygosities were compared per locus and per population; an excess of homozygosity indicated either an inbred population or the existence of a null allele.
The Markov chain method in Genepop version 3.3 (Raymond and Rousset 1995) was used to test for deviation from Hardy-Weinberg equilibrium (Guo and Thompson 1992). The dememorization number was set to 1000, the number of batches was set to 500, and the number of iterations per batch was set to 1000 to achieve a standard error (SE) < 0.01. The result of this test was the fixation index (Fis), a measure of heterozygote deficiency or excess for codominant diploid alleles where Fis = 0 indicates the meeting of Hardy-Weinberg expectations and Fis = 1 signals no heterozygotes.
Genotypic linkage disequilibrium, whether one locus was independent from another, was also tested. Genepop accomplishes this using contingency tables for all loci and estimating the p value by the Markov chain method (Raymond and Rousset 1995). The dememorization number was set to 1000, the number of batches was set to 1000, and the number of iterations per batch was set to 1000.
The genotypic distribution across populations was also calculated using Genepop version 3.3 (Raymond and Rousset 1995). For each locus, a contingency table was created, and the p value was estimated for each population or pair of populations by using the Markov chain method (Raymond and Rousset 1995). The dememorization number was 1000, the number of batches was 500, and the number of iterations per batch was 1000. FST, a measure of population differentiation, was also generated as defined by Weir and Cockerham (1984).
R ST was generated by FSTAT (Goudet 1995) and SPAGeDI version 1.0 (Hardy and Vekemans 2002). SPAGeDI uses spatial coordinates to analyze pairwise comparisons. Latitude and longitude were taken directly from locations of turtles plotted in ArcView version 3.2 (ESRI, Redlands, CA) and were converted to decimal by the Universal Transverse Mercator conversion program from the US Geological Survey.
RESULTS
All 3 loci were polymorphic, with 7 alleles at locus CP3, 3 at CP2, and 10 at CP10 (Table 1). There were 2 unique alleles at Mallard Pools, allele 145 at CP3 and allele 225 at CP10, and Muscatine did not show alleles 208, 215, 225, and 230 from CP10 (Table 1).
Expected and Observed Heterozygosities
No significant differences in heterozygosity were indicated for CP2 and CP3 (Table 2). However, Fis values indicated heterozygotic deficiency for locus CP10 for the populations at Mallard Pools (Fis = 0.34) and Greencastle (Fis = 0.42). This signaled a heterozygotic deficiency, likely the result of 1 or more null alleles at CP10. Therefore, all subsequent tests were performed on the full data set and also on the 2 loci, CP2 and CP3, alone, to alleviate potential effects of a null allele.
Hardy-Weinberg Exact Tests
CP2 and CP3 did not deviate significantly from Hardy-Weinberg equilibrium (Table 3). When 3 loci were examined, the highest and lowest p values were found in Muscatine ( pCP2 = 0.1114; pCP3 = 0.9403). However, the genotype frequencies for CP10 deviated significantly from Hardy-Weinberg with the Mallard Pools and Greencastle populations (Table 3). Subsequently, fixation indices were low for all loci except at CP10 for Mallard Pools and Greencastle and CP2 for Muscatine.
Genotypic Linkage Disequilibrium
Two cases of significant linkage were indicated by the analysis. In both cases, the linkage involved the loci CP10 and CP2 and occurred within the Greencastle and Mallard Pool populations ( p = 0.02 in both cases) (Table 4). Between CP2 and CP3, the genotype frequencies did not deviate significantly, indicating that the loci were not linked. Because of the likelihood of the presence of a null allele and possible linkage, CP10 was dropped from all further analyses.
Genotype Differentiation
Muscatine and the pooled HWA populations showed significant differentiation for CP2 (Table 5). Populations differed significantly at CP2 between Muscatine and Mallard Pools and also for both loci analyzed for comparisons of Muscatine and Greencastle (Table 6).
FST and RST
Analysis of the 2 loci indicated no significant differences between the populations using either FST or RST (Table 7). FST values ranged from −0.0108 to 0.0686, and traditional RST values were also low, ranging from −0.016 to 0.055 (Table 7). The addition of spatial data using SPAGeDI resulted in similarly low RST values (−0.157 to −0.0081) (Table 7).
Identical Alleles
After comparing all sites for identical alleles, 1 pair of turtles was found within Muscatine, 1 pair within Mallard Pools, and 2 pairs within Greencastle.
DISCUSSION
Both FST and RST values indicate that the 2 putative populations at HWA share a high degree of gene flow (where 1.0 = fixation and 0.0 = panmixia) and were essentially 1 panmictic population. In their study of Blanding's turtles (Emydoidea blandingii), Mockford et al. (2007) reported FST values ranging from 0.0 to 0.465 and RST values ranging from −0.016 to 0.507. Thus, our genetic information does not support the lack of evidence for movement between the Mallard Pools and Greenscastle subpopulations from a decade of mark–recapture and radiotracking data (Bernstein et al. 2007). Similarly, the FST and RST analyses indicate that the HWA and Muscatine populations were not differentiated. However, the results of the genotypic distribution test, the more sensitive test, indicated that the Muscatine population was relatively isolated, given the significant subdivision at CP2. This differentiation is not surprising given the much greater geographic separation between these 2 localities (Fig. 1).
We suggest 3 factors that may explain the lack of subdivision within the Hawkeye populations (Greencastle and Mallard Pools) despite mark–recapture data to the contrary as well as the genetic differences between the HWA and Muscatine populations: 1) recent colonization, 2) recent population subdivision, and 3) gene flow within HWA that has been undetected by field studies.
Recent Colonization
What could explain why FST and RST analyses indicated that Muscatine and HWA were not genetically distinct populations? One potential explanation is the relatively recent colonization of ornate box turtles in Iowa following the Wisconsinan glaciation, which likely would result in low levels of genetic variability (Hewitt 1996).
Fossil evidence places Terrapene south and west of Iowa prior to the Pleistocene (Milstead 1967; Moodie and Van Devender 1978; Dodd 2001); during the Wisconsinan glaciation, only a small portion of central Iowa was glaciated (Prior 1991). As the ice retreated north (approx. 14,000 years ago), wetlands, sandy plains, and prairie vegetation replaced tundra and forests. These new niches provided opportunities for fauna (Pielou 1991), and this was probably the time that ornate box turtles migrated into Iowa, initially into southern Iowa.
By the early 1800s, surveys reported that nearly 70%–80% of Iowa was prairie, suitable habitat for T. ornata ornata (Dinsmore 1994). Brumfiel (1919) found numerous turtles in Johnson County which he described as the “dry land tortoise,” and soil characteristics suggest that ornate box turtles potentially could have been widely distributed throughout the county, especially the northern third where HWA is located. In addition, small, fragmented, remnant, extant populations in Iowa indicate a probable wider past distribution, especially across the southern half of the state. However, the relatively recent time frame that box turtles have inhabited Iowa, Illinois, and Wisconsin may not have allowed for allopatric differentiation of the turtle populations. Detection of allopatric differentiation may be hindered by the relative longevity, high nest predation, and juvenile mortality of box turtles (Dodd 2001; Bowen et al. 2004).
However, the statistical results of the more sensitive genotypic distribution analysis between Muscatine and HWA indicated that the Muscatine population was significantly different in genotypic distribution and was isolated from the HWA. This alternative could have resulted from a recent bottleneck of either the Muscatine or HWA populations or, alternatively, from either a greater or an earlier isolation of the 2 populations than outlined above.
Population Differentiation
In contrast to recent colonization, the populations at HWA and Muscatine were just beginning to differentiate, possibly resulting from fragmentation of panmictic populations in which gene flow was recently hindered by rivers, roads, and/or farms (e.g., Curtin 1997).
Hunting, trapping, and the destruction of habitat led to declines, extirpation, and extinction of many species by the end of the 19th century (Dinsmore 1994), and habitat destruction continues to threaten ornate box turtles today (Iowa Department of Natural Resources 2002).
Approximately 0.12% of original Iowa prairie remains in isolated fragments (Dinsmore 1994). Although HWA has both natural and reconstructed prairies, the area around HWA has been farmed for most of the last 100 years, and turtles regularly utilize nearby agricultural fields in daily movements and during reproduction (Bernstein and Black, unpubl. data, 2000). However, we also noted turtle mortality from farm machinery, and long-term farming in an area can most likely extirpate a population.
In addition to farming, mortality on roads can create barriers to gene flow by fragmenting populations (e.g., Doroff and Keith 1990). Muscatine and HWA have increasingly been separated by roads and major highways (Iowa Department of Transportation 1999). According to Keller and Largiader (2003), fragmentation caused by roads reduced gene flow and genetic variability in ground beetles (Carabus violaceus), and Gibbs (1998) observed restricted amphibian movements because of roads in southern Connecticut. Anderson (1956) reported a 10-year decline in ornate box turtles in Missouri as traffic increased on a highway, and similar to observations on desert tortoises (Gopherus agassizii) in California (Boarman et al. 1997), we have observed ornate box turtles at HWA killed by cars.
However, given the long life span of ornate box turtles and the short period of time that farming and roads have been in Iowa, it seems unlikely that these modern activities alone would be the current cause of Muscatine's isolation. Rivers can also be boundaries or barriers to dispersal and gene flow, and although ornate box turtles are capable of swimming, they are poor swimmers and less likely to cross the several large rivers between HWA and Muscatine (Fig. 1). In addition, distance may also be a limiting factor: 3.3 km is the maximum homing distance for T. ornata noted by Metcalf and Metcalf (1970). Although Schwartz and Schwartz (1982) noted one Terrapene carolina triunguis (transient) that traveled 10.0 km in a straight line, crossing highways and, after much hesitation, the Moreau River in Missouri, Bernstein et al. (2007) documented much smaller home ranges and movements at HWA. Therefore, rivers and distance, in conjunction with habitat fragmentation and roads, are the likely causes of any genetic isolation between Muscatine and HWA. However, neither of the first 2 hypotheses explains gene flow within HWA.
Gene Flow Within HWA
If HWA is a large central population, then no genetic differences would be found between individuals in different locations. Radiotracking and mark–recapture studies may not detect migrants and transients between the subpopulations, and gene flow could be maintained by matings between individuals within the subpopulations and transients (Kiester et al. 1982) and/or consecutive matings across overlapping home ranges between Mallard Pools and Greencastle. The latter conclusion is supported by the knowledge of unmarked turtles found between the 2 subpopulations (Bernstein and Black, unpubl. data, 2000) along with evidence of home ranges that both overlap between individuals and span multiple habitats (Bernstein et al. 2007). These data suggest a larger, panmictic population of ornate box turtles outside of our study areas at HWA, a conclusion supported by observations and reports of ornate box turtles outside of Mallard Pools and Greencastle.
Also, similarities in allelic distribution may be due to the age of the turtles as blood was collected from 15–30-year-old adults. Furthermore, because each area has abundant resources as well as nesting and overwintering sites, long-range movements between sites may not occur.
However, the distinct alleles found in the 3 turtles at the Mallard Pools (1 turtle with allele 145 at CP3 and 2 turtles with allele 225 at CP10), may indicate only slight contact with the Greencastle site. Newman and Squire (2001) suggested that subtle differentiation between wood frogs (Rana sylvatica), separated by small distances of a few kilometers, indicated that the populations acted as a metapopulation with only slight interactions. Thus, Mallard Pools and Greencastle could drift to fixation for 1 of the 2 new alleles if there is only slight contact between the sites.
Extreme weather events can also affect ornate box turtle home ranges and may bring ornate box turtles together at HWA. In 1993, much of the Greencastle subpopulation was isolated onto high points as floodwaters inundated the area, a situation partially repeated in spring 2005. In contrast, we also noted extensive movements between wetlands during a drought period, and these atypical dispersions are similar to the hypothesis of Rowe et al. (2000) that Natterjack toads (Bufo calamita) may cross areas at low tide and maintain gene flow. However, it is unclear if any of these weather-induced dispersal events affected mating patterns of ornate box turtles at HWA.
Therefore, these data can possibly be explained by using all 3 hypotheses. The recent glaciations may not have allowed HWA and Muscatine populations to significantly differentiate; although, the process of separation may have begun due to natural barriers and fragmentation of landscapes by human activity. Finally, the high degree of gene flow within HWA was indicative of one central panmictic population.
Future Research and Management Implications
Analysis of more microsatellite loci might reveal a clearer picture of gene flow because the null allele plus the nearly monomorphic CP2 allele could have affected results, and CP3 showed polymorphism. In addition, specific primers for ornate box turtles might eliminate the problems of the null allele. An alternative for future study would be to examine mitochondrial DNA (mtDNA) markers to determine mating patterns within the HWA (Balloux et al. 2000; Johnson et al. 2003; Bouzat and Johnson 2004).
Of course, more study using microsatellites or mtDNA may only confirm that turtles from the 2 sites at HWA are currently part of a single panmictic population with gene flow within a large, dispersed population. The population at HWA appears to extend beyond the boundaries of the studied sites, and gene flow within this large population may promote genetic diversity and long-term stability under current conditions (Kuo and Janzen 2004, 2007). However, as Dodd et al. (1994) and Bowen et al. (2004) recommended for box turtles in other locations, we recommend that all efforts be made to provide continuous habitat throughout the range of the ornate box turtles at HWA to prevent edge effects and the mortality that results. Additionally, the turtles should remain protected from collection and other unnatural causes of mortality.
Map of Iowa indicating locations of the Hawkeye Wildlife Area (HWA) and Muscatine.
Aerial photo of Hawkeye Wildlife Area showing locations of Mallard Pools and Greencastle subpopulations. Note farm fields and houses south of Mallard Pools. (Photo from Iowa Geological Survey, Iowa Department of Natural Resources, Iowa City, IA.)
