Dispersal in the Coahuilan Box Turtle

1 Running title: Dispersal in the Coahuilan box turtle In press, Molecular Ecology 2 3 4 CONTRASTING DEMOGRAPHIC AND GENETIC ESTIMATES OF DISPERSAL IN THE 5 6 ENDANGERED COAHUILAN BOX TURTLE: A CONTEMPORARY APPROACH TO 7 8 CONSERVATION 9 10 Jennifer G. Howeth1,4, Suzanne E. McGaugh2, Dean A. Hendrickson1,3 11 12 MEC-08-0399 Final 13 1 Section of Integrative Biology 14 University of Texas at Austin 15 1 University Station C0930 16 Austin, Texas 78712 17 Phone: 512 475 8669 18 Fax: 512 471 3878 19 20 2 Department of Ecology, Evolution, and Organismal Biology 21 Iowa State University 22 251 Bessey Hall 23 Ames, Iowa 50011 24 25 3 Texas Natural Science Center, Texas Natural History Collection 26 University of Texas at Austin 27 PRC 176 / R4000 28 10100 Burnet Road 29 Austin, Texas 78758 30 31 4 Corresponding author: [email protected] 32 Abstract: 249 / 250 words 33 Main text: 7,641 / 8,000 words 34 Figures: 5; Supplementary Figures: 2 35 Tables: 3; Supplementary Tables: 2 36 37 38 Keywords: connectivity, isolation by distance, metapopulation, microsatellite, mark-recapture, 39 habitat fragmentation

40 Abstract 41 42 The evolutionary viability of an endangered species depends upon gene flow among subpopulations 43 and the degree of habitat patch connectivity. Contrasting population connectivity over ecological and 44 evolutionary timescales may provide novel insight into what maintains genetic diversity within 45 threatened species. We employed this integrative approach to evaluating dispersal in the critically 46 endangered Coahuilan box turtle (Terrapene coahuila) that inhabits isolated wetlands in the desert- 47 spring ecosystem of Cuatro Ciénegas, Mexico. Recent wetland habitat loss has altered the spatial 48 distribution and connectivity of habitat patches; and we therefore predicted that T. coahuila would 49 exhibit limited movement relative to estimates of historic gene flow. To evaluate contemporary 50 dispersal patterns, we employed mark-recapture techniques at both local (wetland complex) and 51 regional (inter-complex) spatial scales. Gene flow estimates were obtained by surveying genetic 52 variation at nine microsatellite loci in seven subpopulations located across the species’ geographic 53 range. The mark-recapture results at the local spatial scale reveal frequent movement among wetlands 54 that was unaffected by inter-wetland distance. At the regional spatial scale, dispersal events were 55 relatively less frequent between wetland complexes. The complementary analysis of population 56 genetic substructure indicates strong historic gene flow (global FST = 0.01). However, a relationship of 57 genetic isolation by distance across the geographic range suggests that dispersal limitation exists at the 58 regional scale. Our approach of contrasting direct and indirect estimates of dispersal at multiple spatial 59 scales in T. coahuila conveys a sustainable evolutionary trajectory of the species pending preservation 60 of threatened wetland habitats and a range-wide network of corridors. 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83

84 Introduction 85 86 The degree of dispersal between habitats and subpopulations can profoundly impact the demographic 87 and evolutionary trajectory of a species (Bohonak 1999; Clobert et al. 2004). Dispersal affects these 88 trajectories by mediating the abundance and exchange of individuals between subpopulations (Tilman 89 et al. 1997; Hanski 1999) and the distribution of alleles across the landscape through gene flow 90 (Hastings & Harrison 1994; Manel et al. 2003; Storfer et al. 2007). As a consequence of the critical 91 role of connectivity in the maintenance of genetic diversity and adaptive potential, understanding 92 patterns of dispersal in endangered species remains a central focus of conservation assessments 93 (Allendorf & Luikard 2007). In empirical studies, dispersal rates are typically evaluated from either an 94 ecological or ‘direct’ approach where movement is tracked across the landscape or an evolutionary or 95 ‘indirect’ approach where dispersal is inferred from genetic data (reviewed in Bohonak 1999). A strict 96 evolutionary approach to evaluating dispersal may detect high gene flow in a threatened species but 97 could fail to acknowledge fragmentation-induced restriction in contemporary movement. Thus, by 98 adopting one approach to evaluating dispersal, only partial information about population genetic 99 structure and the determining mechanism may be elucidated and the results can thereby misinform 100 conservation efforts. 101 102 Evaluating dispersal over ecological and evolutionary timescales can assess both patterns of 103 gene flow and the demographic processes that may maintain them (e.g. Watts et al. 2004; Wilson et al. 104 2004; Boulet et al. 2007). Patterns of gene flow between subpopulations may serve as a fingerprint of 105 connectivity prior to habitat alteration and the consequential disruption of movement (Palsboll et al. 106 2007; Schwartz et al. 2007). In such cases, the combined effects of fragmentation-induced changes in 107 modern movement and effective population sizes may not maintain historic allele frequencies in 108 subsequent generations (Palsboll et al. 2007; Schwartz et al. 2007). In this paper, we suggest that the 109 degree to which direct observations of dispersal correspond to the indirect estimate gives an indication 110 of how well current demographic processes relate to historic connectivity in recently fragmented 111 populations. This complementary assessment of dispersal in endangered species experiencing habitat 112 loss can consequently expose the potential for a human-induced shift in evolutionary trajectory. 113 114 Wetland turtles are among the organisms most threatened by habitat loss and degradation 115 (Parker & Whiteman 1993; Joyal et al. 2001), and thus these taxa may be especially vulnerable to 116 fragmentation-induced decoupling of contemporary demography and historic gene flow. Species 117 inhabiting the natural patch-matrix mosaic of wetland habitats often exhibit metapopulation dynamics 118 (Joyal et al. 2001; Marsh & Trenham 2001), and as wetlands are lost or fragmented, the stepping 119 stones facilitating dispersal are removed (Parker & Whiteman 1993; Bohonak & Jenkins 2003). For 120 such species that are patchily distributed, landscape characteristics including interpatch distance 121 (Marsh et al. 2000) and patch size (Hanski 1994; Hanski 1999) are important in regulating patch 122 dynamics and ultimately spatial genetic structure (Manel et al. 2003; Storfer et al. 2007). In 123 freshwater turtles, population genetic structure is shaped by both local dispersal behavior within a 124 subpopulation (Freedberg et al. 2005) and regional (long-distance) dispersal between subpopulations 125 which determines the extent and frequency of stepping stone movement (Spinks & Shaffer 2005). At 126 the local scale, a single subpopulation may utilize multiple wetland habitat patches that are part of a 127 larger complex of wetlands (Joyal et al. 2001). At the regional scale, stepping stone behavior and 128 barriers to dispersal can regulate the exchange of alleles between subpopulations and may generate 129 patterns of genetic isolation by geographic distance (Spinks & Shaffer 2005). Fragmentation-induced

130 regional dispersal limitation in wetland turtles can lead to closed subpopulations that eventually suffer 131 from reduced genetic variation, are vulnerable to inbreeding depression and drift, and consequently 132 may leave individuals with a limited capacity to adapt to changing environments (Templeton et al. 133 1990; Parker & Whiteman 1993; Kuo & Janzen 2004). 134 135 In this study, we contrast patterns of movement and historic gene flow in the critically 136 endangered Coahuilan box turtle, Terrapene coahuila (Emydidae), endemic to the desert-spring 137 ecosystem of Cuatro Ciénegas, Mexico to test for effects of habitat fragmentation on contemporary 138 dispersal. The isolated 84,000 ha Cuatro Ciénegas valley supports over 70 described endemic species 139 (Secretaría del Medio Ambiente y Recursos Naturales 1999) and serves as a focal landscape for 140 conservation in human-altered environments (Abell et al. 2000; Hendrickson et al. in press). The rich 141 biodiversity and fragmented aquatic habitats of Cuatro Ciénegas recently prompted declaration of the 142 region as a UNESCO Biosphere Reserve (UNESCO-MAB 2006). Terrapene coahuila, the only 143 aquatic species of the genus, inhabits both permanent and seasonal wetlands that are widely distributed 144 across the Cuatro Ciénegas valley (Webb et al. 1963; Brown 1974; Van Dijk et al. 2007). These 145 wetland habitats have become substantially fragmented over the last half century due to water 146 extraction and diversion via a complex canal system (Minckley 1992; Abell et al. 2000; Souza et al. 147 2006; Van Dijk et al. 2007; Hendrickson et al. in press). As a consequence, T. coahuila is listed by the 148 World Conservation Union (IUCN) and the United States Fish and Wildlife Service (USFWS) as 149 ‘endangered’ because of the habitat loss and associated restricted geographic range (USFWS 1973; 150 Van Dijk et al. 2007). The geographic distribution of the box turtle has been reduced from an 151 estimated 600 km2 in the 1960’s to approximately 360 km2 in 2002 (Van Dijk et al. 2007). 152 153 Wetland habitat loss has resulted in relatively isolated subpopulations of T. coahuila (Van Dijk 154 et al. 2007), thereby potentially constraining dispersal across the inhospitable desert matrix. 155 Importantly, fifty years of diffuse habitat fragmentation should not result in significant consequences 156 for allele frequencies or a shift in evolutionary trajectory given the species’ estimated 15-year 157 generation time and life-span of over 65 years (both estimates derived by analogy with congeners; 158 Miller 2001; Van Dijk et al. 2007). This buffer of a long-generation time, relative to the period of 159 habitat fragmentation, highlights the need to evaluate whether contemporary dispersal patterns 160 correspond to estimates of pre-fragmentation connectivity. We therefore compare measures of local 161 and regional dispersal acquired from mark-recapture methods with range-wide migration estimates 162 obtained from microsatellite markers to examine if contemporary demography and historic gene flow 163 are decoupled in T. coahuila. 164 165 The following predictions were made about local and regional dispersal, population genetic 166 substructure, and their relationship in T. coahuila. (i) Current regional movement between wetland 167 complexes will be limited as compared to genetic estimates of regional dispersal. (ii) Current local 168 movement within a wetland complex will be more frequent than regional movement, and will depend 169 upon interpatch distance and patch size. (iii) Population genetic differentiation will exist across the 170 geographic range due to regional dispersal limitation imposed by the desert matrix and the mountain 171 Sierra San Marcos which divides the geographic range into two lobes (Fig. 1). (iv) Likewise, genetic 172 isolation by distance is expected across the species’ geographic range. 173 174 175

176 Materials and Methods

177 Research was conducted in the Área de Protección de Flora y Fauna Cuatro Ciénegas, located in the 178 Sierra Madre Oriental of central Coahuila, Mexico between 26o 45’ 00”: 27o 00’ 00” N and 101o 48’ 179 49”: 102o 17’ 53” W (Fig. 1). The Cuatro Ciénegas valley is characterized by Chihuahuan Desert biota 180 and approximately 200 saline ponds and wetlands formed from geothermal springs. Over the last half 181 century, diffuse wetland habitat loss across the Cuatro Ciénegas valley has fragmented once 182 contiguous wetlands into spatially discrete wetland complexes separated by large extents (several 183 kilometers) of desert matrix. Each wetland complex contains multiple wetland habitat patches that are 184 separated from one another by relatively small extents (tens of meters) of matrix. In this study, we 185 consider a wetland complex or ‘site’ to be the local scale of spatial sampling, and two or more wetland 186 complexes to be the regional scale of spatial sampling. 187 188 Population sampling and molecular methods 189 190 To evaluate population genetic substructure, seven wetland complexes located range-wide within the 191 Cuatro Ciénegas basin were sampled opportunistically for Terrapene coahuila (Fig. 1). Tissue 192 samples (0.2 cc blood) were collected from 15 to 27 individuals at each site for a total of 156 193 individuals (71 male, 80 female, 9 indistinguishable adult, 16 subadult; Table 1). Samples were 194 preserved in buffer solution (0.01M Tris, 10mM EDTA, 0.01M NaCl, and 1% SDS) upon collection 195 and stored at -20° C. Whole genomic DNA was extracted from blood with a Qiagen DNeasy® tissue 196 kit (Qiagen Inc., Valencia, CA, USA). Genetic diversity within T. coahuila was evaluated with nine 197 nuclear microsatellite markers originally developed for bog turtles (Glyptemys muhlenbergii; King & 198 Julian 2004) and which were composed of either trinucleotide (GmuB8) or tetranucleotide (GmuD16, 199 GmuD21, GmuD40, GmuD62, GmuD70, GmuD90, GmuD107 GmuD121) repeat motifs. Allelic 200 variation at each locus was assessed by polymerase chain reaction (PCR) in 12.5µL reactions 201 containing 0.4µM HEX or FAM-labeled forward primer, 0.4µM reverse primer, 1X PCR buffer, 2mM 202 MgCl2, 0.08µL Taq polymerase, 0.01µM dNTP, and 2.5µL 10-100 ng/µL DNA. Reactions were 203 performed in a thermal cycler (Techne TC-412 or Eppendorf Mastercycler Gradient) under the 204 following conditions: 94° C for 2 min, 34 cycles of 94° C for 30s, 56° C for 30s, 72° C for 30s, and a 205 final step of 72° C for 7 min. After the addition of 0.25µL GeneScan-500 ROX internal size standard, 206 samples were genotyped individually on an ABI Prism 3100 Genetic Analyzer (PE Applied 207 Biosystems, Foster City, CA). GeneScan files were analyzed with Genotyper (v. 3.6NT; PE Applied 208 Biosystems), and allele sizes were visually confirmed relative to the size standard and controls. Micro- 209 checker (v. 2.2.3; Van Oosterhout et al. 2004) was used to test for null alleles and allelic dropout using 210 1000 Monte Carlo simulations and a Bonferroni corrected 95% confidence interval. 211 212 Voucher specimens for the tissue samples were deposited at the National Collection of 213 Amphibians and Reptiles, Institute of Biology, Universidad de Autónoma de México (UNAM, México 214 D.F, Mexico) under appropriate Mexican collection and Convention on International Trade of 215 Endangered Species (CITES) Appendix 1 permits. 216 217 Gene flow and spatial genetic structure 218 219 Linkage disequilibrium between pairs of microsatellite loci and departures from Hardy-Weinberg 220 within each sample site and locus were evaluated with a Markov Chain approximation of an exact test

221 performed in GENEPOP (web version 3.4; Raymond & Rousset 1995). Associated probability values 222 were corrected for multiple comparisons using a Bonferroni adjustment for a significance level of 0.05. 223 To evaluate variation in allelic diversity within sample sites, the number of total, effective, and private 224 alleles (A, Ae, Ap), and observed and expected heterozygosity (HO, HE), were calculated in GenAlEx (v. 225 6; Peakall & Smouse 2006). Additionally, allelic richness estimates derived from rarefaction and 226 corrected for variable sample sizes, and an inbreeding coefficient (FIS), were calculated in FSTAT (v. 227 2.9.3.2; Goudet 1995). To compare differentiation among sites, pairwise site FST coefficients (Weir & 228 Cockerham 1984) were calculated in ARLEQUIN (v. 2.0.1.1; Schneider et al. 2000). Significant 229 pairwise site differentiation was determined by comparing the actual FST value to a null FST 230 distribution generated from 10 000 genotypic permutations between sites. Probability values were 231 Bonferroni corrected at a significance level of 0.05. 232 233 To further evaluate genetic structure at the local spatial scale and to test whether more closely 234 related individuals are located within a wetland complex, inter-individual relatedness estimates were 235 generated for all pairs of individuals within each site. Relatedness coefficients following Queller and 236 Goodnight (1989; rqg ) were calculated in GenAlEx and 95% confidence intervals were determined for 237 each site mean using 10 000 bootstrap simulations. Departure from the null hypothesis of panmixia at 238 each site was assessed by comparing the site mean relatedness coefficient to a distribution of simulated 239 relatedness coefficients generated from 10 000 genotypic permutations across sites. 240 241 In accord with our hypothesis of differences in genotypic composition between sites west and 242 east of Sierra San Marcos, a hierarchical analysis of molecular variance (AMOVA; Excoffier et al. 243 1992) using 100 000 permutations in ARLEQUIN was used to examine partitioning of genetic 244 variation within and between the western lobe of the basin (Laguna Grande) and the eastern lobe (six 245 sites east of Sierra San Marcos). Evidence for range-wide isolation by geographic distance was 246 evaluated as the correlation between Rousset’s (1997) genetic distance [FST/ (1-FST)] and geographic 247 distance for all pairwise site combinations with a Mantel test using 30 000 Monte Carlo matrix 248 permutations in IBDWS (v. 3.14; Jensen et al. 2005). Geographic distances used in the analysis were 249 calculated from latitude / longitude coordinates determined in the field using a global positioning 250 system (Trimble Geoexplorer® GPS). All GPS data were differentially corrected with Pathfinder 251 Office software (v. 2.80, Trimble Navigation Ltd., Sunnyvale, CA, USA). Pairwise site straight line 252 (Euclidian) distances were calculated on the World Wide Web 253 (http://www.wcrl.ars.usda.gov/cec/java/lat-long.htm). A dummy site, 26° 56’ 22.362” N, 102° 8’ 254 57.577” W (NAD27-Mexico), located northwest of the tip of Sierra San Marcos, was used as an 255 intermediate passing point between Laguna Grande and the six eastern sites. 256 257 Gene flow across the geographic range of T. coahuila was evaluated with a maximum 258 likelihood coalescent-based approach. Population differentiation was too low to detect recent 259 migration events using assignment tests. Because we had no a priori knowledge of which sites or 260 groups of sites supported distinct populations, each of the seven sample sites was represented 261 independently in the analysis of gene flow. Evidence for migration between sites was evaluated as 262 4Nm, an estimate derived by multiplying θ (4Neµ, where Ne is the effective population size and µ is the 263 mutation rate/ generation) and M (m/µ, where m is the immigration rate) in MIGRATE (v. 2.3; Beerli 264 & Felsenstein 2001). The Brownian motion approximation of the stepwise mutation model, assuming 265 a constant microsatellite mutation rate, was employed. Starting θ and M values were generated from 266 FST estimates. A final matrix of 4Nm was produced by averaging over five replicates of the following

267 settings: 20 000 tree burn-in, 10 short chains with 100 000 trees sampled and 5 000 trees recorded 268 (sampling increment of 20), and 3 long chains with 1 000 000 trees sampled and 50 000 trees recorded 269 (sampling increment of 20). Several shorter preliminary runs were initiated previously to determine 270 the parameter values required to reach relatively stable distributions for this final long run. Significant 271 differences in pairwise site unidirectional migration rates were evaluated by 95% confidence intervals 272 around the migration estimates, where no overlap in confidence intervals indicates asymmetric gene 273 flow between site pairs. 274 275 Movement at local and regional spatial scales 276 277 To detect the spatial extent and frequency of movement in T. coahuila, we employed mark-recapture 278 methods that have been used previously to detect dispersal in Terrapene (Kiester et al. 1982; Dodd et 279 al. 2006) and in this species in particular (Brown 1974). To estimate T. coahuila dispersal capacity 280 and rate at the local spatial scale, interpatch movement was evaluated at 37 wetlands within the 281 complex of Los Gatos (Fig. 1) over seven mark-recapture efforts from 19 May – 24 July 2002 (67 d), a 282 time period in which T. coahuila interpatch movements have been documented previously (Brown 283 1974). On average, there were 9.4 days (2.6 SD) between capture efforts. Turtles were located 284 visually by walking linear transects within, and a circular transect around, each wetland once per 285 capture effort. Other aquatic turtle capture techniques (e.g. hoop nets, basking traps) are not 286 appropriate for this semiaquatic species (Brown 1974). Each capture effort was conducted during peak 287 turtle activity, from 0700 to 1100 hr, over a period of two days with two to four biologists 288 participating. All turtles encountered were marked with a numbered gray vinyl tag (Floy Tag, Seattle, 289 WA, USA) affixed to a right pleural scute and were released at the site of capture. This method of 290 tagging is relatively robust, as tag loss was minimal with an average of 2.1 tags lost (3.2 SD) per 291 month, for a total of 30 tags lost in this study. At the local spatial scale, dispersal rate was determined 292 as the proportion of turtles recaptured in a different wetland from where they were originally marked 293 over the 67 day period, standardized to seven days, the minimum time between local capture events. 294 295 To determine the role of patch area in local dispersal and area-abundance relationships, latitude 296 / longitude and area of each of the 37 wetlands was determined from differentially corrected GPS 297 coordinates. Inter-wetland dispersal distances were defined as the minimum Euclidian distance 298 between source wetland edge and target wetland edge, and were calculated in Pathfinder Office. The 299 relationship of turtle abundance (individuals / effort) as a function of wetland area (ha) was determined 300 with linear regression. Abundance and area values were log (x+1) transformed to reduce variance 301 prior to regression analysis in STATISTICA (v. 6.1, Statsoft Inc., Tulsa, Oklahoma, USA). 302 303 To evaluate movement at the regional spatial scale, turtles at the wetland complexes of 304 Antiguos Mineros and Los Gatos (Fig. 1) were surveyed over approximately one year, a time period in 305 which large scale movements have been previously detected in the species (Webb et al. 1963; Brown 306 1974). Antiguos Mineros and Los Gatos are separated by 9.71 km and represent the greatest nearest- 307 neighbor inter-site distance in this study and thus the maximum possible distance for nearest-neighbor 308 migrant exchange. Turtles at Antiguos Mineros were marked over seven capture efforts from 22 May - 309 3 July 2002. Turtles at the four largest wetlands of Antiguos Mineros (sum area = 2.23 ha) and the two 310 largest wetlands of Los Gatos (Fig. 1) were captured (not marked due to time constraints) 311 approximately once a month over the following dates: Antiguos Mineros, 27 May 2002 – 16 May 2003 312 (355 d) and Los Gatos, 30 June 2002 – 28 July 2003 (394 d). Specifically, at Los Gatos the average

313 time between recapture events was 35.7 days (14.8 SD). At Antiguos Mineros, there were 32.2 days 314 (11.5 SD) between recapture events. Daily surveys occurred during temperature-dependent peak turtle 315 activity, May-October from 0700 to 1100 hr and November-April from 1100 to 1400 hr. Regional 316 dispersal rates for turtles at Antiguos Mineros and Los Gatos were determined as the proportion of 317 turtles recaptured in a different wetland complex from where they were originally marked over the 318 sampling period (355 d, 394 d respectively) standardized to seven days to facilitate comparison with 319 the local dispersal rate. 320 321 Results 322 323 Gene flow and spatial genetic structure

324 Measures of allelic richness and heterozygosity revealed high levels of genetic diversity across the 325 geographic range of Terrapene coahuila (Table 1, Table S1, Supplementary material). The 326 microsatellite loci were polymorphic, ranging from 6 alleles (GmuD21) to 17 alleles (GmuD40), with a 327 rarefied richness estimate by locus and sample site ranging from 2.67 alleles (GmuD21, Las Salinas) to 328 10.84 alleles (GmuD40, Charcos Prietos; Table S1, Supplementary material). There was no indication 329 of large allele dropout at any locus. There was significant evidence (P < 0.05) for null alleles at one 330 locus, GmuD70, although estimated null allele frequencies were low (< 0.076 by all estimation 331 methods). The test for linkage disequilibrium in pairwise comparisons of loci within sites 332 demonstrated no linkage between loci. The sample site Pozas Azules supported the greatest number of 333 alleles (A), while the eastern peripheral site Antiguos Mineros supported the fewest (Table 1). The 334 number of effective alleles (Ae), a richness measure corrected for site-specific expected heterozygosity, 335 varied by less than one allele across the geographic range, thus suggesting relatively uniform 336 demographic processes operating range-wide. Private alleles (Ap), a measure of population 337 differentiation which includes both alleles unique to a sample site or an individual (singleton), were 338 detected at Laguna Grande, Las Salinas, Tio Cándido, and Pozas Azules, with the greatest frequency 339 found at Tio Cándido. Observed levels of heterozygosity (HO) were high across sites, ranging from 340 0.69 to 0.81, and comparable to expected heterozygosity with each site and locus in Hardy-Weinberg 341 equilibrium (Fisher’s exact test; all sites, all loci, P > 0.05). There was a slight trend of decline in 342 genetic diversity from west to east across the geographic range of T. coahuila with the two 343 westernmost sites, Laguna Grande and Las Salinas, supporting the highest levels of observed 344 heterozygosity and allelic diversity (Table 1; Table S2, Supplementary material). 345 346 Site-specific FIS values gave no indication of inbreeding and implied strong connectivity 347 between sites and large effective population sizes (FIS range: -0.08 to 0.03; Table 1). In agreement 348 with this result, low levels of relatedness (rqg) between individuals within sites suggest that the 349 majority of sites were at least moderately connected and did not differ from a null model of panmixia 350 (rqg range: -0.03, 0.05; Fig. S1, Supplementary material). Individuals within Antiguos Mineros and 351 Charcos Prietos, however, were significantly more related than what would be expected under range- 352 wide panmixia (Antiguos Mineros P = 0.034, Charcos Prietos, P = 0.006). Thus, these localities may 353 support relatively cohesive subpopulations. 354 355 Low levels of population differentiation among all sites indicate strong historic gene flow, with 356 a global FST of 0.011 (95% CI: 0.006, 0.016). Pairwise site FST values were also low and ranged from - 357 0.006 to 0.041 (Table 2). The only significant differences in pairwise site genetic differentiation were

358 the comparisons with Laguna Grande and four eastern sites: Charcos Prietos, Pozas Azules, Los Gatos, 359 and Antiguos Mineros. Additionally, there was a significant positive correlation between pairwise site 360 genetic (Rousset’s) distance and geographic distance (Mantel test; P = 0.016, y = 0.002x-0.010, r = 361 0.733, Z = 5.383, Fig. 2), indicating a range-wide relationship of isolation by distance. This 362 relationship of isolation by distance dissolved, however, when Laguna Grande was excluded from the 363 analysis (P = 0.490, y = 0.001x-0.006, r = 0.031, Z = 1.012). The hierarchical AMOVA revealed no 364 significant variation in genetic structure between the western lobe of the basin (Laguna Grande) and 365 the eastern lobe (six eastern sites; FCT = 0.022, P = 0.144). There was significant, albeit small (<1%), 366 variation explained by among site differences in genetic differentiation in the eastern sites (FSC = 367 0.006, P < 0.001). The majority of variation (97.24%) was found among individuals within sites (FST 368 = 0.028, P < 0.001). 369 370 Relatively high range-wide migration rates (4Nm) suggest that T. coahuila was once distributed 371 continuously across the basin (Table 3). Levels of gene flow varied substantially between sites, with a 372 minimum estimate of 0.709 4Nm and a maximum of 10.754 4Nm. The significant asymmetric gene 373 flow estimates indicate migration from Laguna Grande northeast towards the eastern lobe (Fig. 3). 374 Additionally, the eastern core site of Tio Cándido experienced relatively strong immigration from all 375 sites sampled except Charcos Prietos. These results reinforce the findings of the analysis of population 376 differentiation (FST) by showing a high degree of mixing in the eastern lobe of the geographic range. 377 Although 67% of the pairwise site migration estimates reported in Table 3 demonstrate significant 378 asymmetric dispersal, the bidirectional migration estimates for a given site pair often differed by only a 379 small amount (Fig. 3). 380 381 Movement at local and regional spatial scales 382 383 The mark-recapture analysis at the local scale (wetland complex) revealed that Terrapene coahuila 384 movement within the Los Gatos wetland complex was relatively frequent at a rate of 1.03% wk -1. 385 Specifically, over the course of 67 days, 481 turtles were marked at Los Gatos with 151 individuals 386 (31.39%) recaptured at least once for a total of 207 recapture events. Of the turtles recaptured, 15 387 individuals (9.93%) were found in a different wetland within the wetland complex than where they 388 were captured initially, and two individuals moved twice over the sampling period for a total of 17 389 dispersal events. Average dispersal distance for these individuals was 279.96 ± 198.85 m SD, with 390 29.76 ± 16.61 SD maximum days between events. The dispersal events within the wetland complex 391 suggest that inter-patch dispersal was not limited by distance, and thus local dispersal capacity was 392 relatively high (Fig. 4a). The most common dispersal route was between the two largest wetlands in 393 the complex (Fig. S2, Supplementary material), a distance of 513 m (Fig. 4a). Of the 17 dispersal 394 events, 11 (64.71%) involved migration to or from the largest wetland in the complex; whereas, the 395 remaining 6 events (35.29%) involved dispersal among 8 smaller wetlands (mean size: 0.15 ± 0.12 ha 396 SD). This suggests, relative to the reduced proportion (8.15%) of turtles originally marked in small 397 wetlands (16 wetlands: mean size: 0.10 ± 0.14 ha SD) that did not move or were not recaptured, that 398 these minor wetlands serve as stepping stone habitat. An analysis of area-abundance further indicates 399 that large wetlands support more individuals than small wetlands (P < 0.001, y = 1.465x+0.030, r = 400 0.893; Fig 4b). 401 402 Regional dispersal dynamics between Los Gatos and Antiguos Mineros revealed long-distance 403 dispersal in T. coahuila (Fig. 5). Temporal variation in T. coahuila proportional abundance at Los

404 Gatos demonstrated that the subpopulation was moderately stable over a period of 394 days and varied 405 by less than 15% (Fig. 5a). In contrast, abundance at Antiguos Mineros fluctuated markedly and the 406 local population collapsed due to high emigration rates. At Antiguos Mineros, 147 turtles were 407 marked and 25 individuals (17.01%) were recaptured over 355 days. Of those individuals recaptured, 408 5 (20%) were recaptured at Los Gatos during June and July 2002, coinciding with the period of 409 emigration at Antiguos Mineros (Fig. 5b). This result corresponds to a dispersal rate of 0.39 % wk-1 410 between Antiguos Mineros and Los Gatos, with a straight-line dispersal distance of 9.71 km. One 411 Antiguos Mineros migrant captured at Los Gatos in July 2002 was recaptured again at Antiguos 412 Mineros the following April, providing insight into seasonal migration encompassing 330 days 413 (marked Antiguos Mineros May 2002; recaptured Los Gatos July 2002; recaptured Antiguos Mineros 414 April 2003; Fig. 5b). In contrast to the relatively high dispersal rate from Antiguos Mineros to Los 415 Gatos, only one Los Gatos individual out of 228 recaptured over a period of 394 days was found at 416 Antiguos Mineros, yielding a dispersal rate of 0.01 wk -1 between Los Gatos and Antiguos Mineros. 417 Collectively, the bidirectional dispersal events between Los Gatos and Antiguos Mineros were 418 separated by 67.43 ± 92.46 SD maximum days. 419 420 Discussion 421 422 The approach used in this study to evaluate dispersal in the endangered Terrapene coahuila 423 acknowledges subpopulation connectivity occurring over both ecological and evolutionary timescales 424 and consequently provides novel insight into the maintenance of genetic diversity. Demographic and 425 genetic estimates of dispersal in T. coahuila demonstrate that the pattern and frequency of current 426 movement correspond well with levels of historic gene flow. This general congruence between 427 dispersal estimates suggests no fragmentation effects in T. coahuila movement behavior. The direct 428 estimates of movement from the mark-recapture study revealed a high local and regional dispersal 429 capacity and thus potential for gene flow between subpopulations, while the indirect estimates of 430 dispersal elucidated substantial genetic diversity, little differentiation among subpopulations, no 431 inbreeding, and relatively large migration rates. Gene flow was not entirely unconstrained across the 432 geographic range, however. In accord with our hypothesis, genetic differentiation between the 433 subpopulation west of Sierra San Marcos (Laguna Grande) and subpopulations located to the east 434 generated a range-wide relationship of isolation by geographic distance. From a conservation 435 perspective, the results highlight the relatively unique genetic diversity harbored west of Sierra San 436 Marcos and convey a sustainable evolutionary trajectory of T. coahuila pending preservation of a 437 range-wide network of wetland habitats and corridors. 438 439 Patterns of movement behavior by Terrapene coahuila suggest a strong potential for migrant 440 exchange and thus provide a possible mechanism for the observed weak population genetic 441 substructure. More specifically, the microsatellite analysis revealed relatively high unidirectional 442 migration estimates between Antiguos Mineros and Los Gatos (9.1 4Nm, AM – LG; 3.1 4Nm, LG – 443 AM; Table 3); and the regional recapture data confirmed the possibility of asymmetric spatial coupling 444 with unidirectional dispersal rates of 0.39% wk -1 from Antiguos Mineros to Los Gatos and 0.01% wk - 445 1 from Los Gatos to Antiguos Mineros. While there is no widely accepted level of dispersal at which 446 subpopulations become demographically correlated (Waples & Gaggiotti 2006; Palsboll et al. 2007), 447 the regional movement behavior by T. coahuila suggests that the two wetland complexes serve as 448 subpopulations within a metapopulation. Migration of turtles at Antiguos Mineros is likely an annual 449 response to the fluctuating environments of the seasonal wetlands which predictably dry each summer

450 (Howeth, personal observation). As a consequence, regional dispersal in T. coahuila appears to be 451 largely regulated by the hydroperiod of seasonal wetland habitats and is relatively less limited by 452 dispersal distance. Interestingly, the significant relatedness of individuals at Antiguos Mineros 453 indicates distinct subpopulation structure and the possibility of natal homing (Freedberg et al. 2005), 454 thereby providing additional support for metapopulation processes operating at the regional spatial 455 scale. 456 457 A high rate of interpatch dispersal at the local spatial scale in Los Gatos (1.03% wk -1), and the 458 correspondingly low relatedness estimate, suggests a well-mixed subpopulation with random mating 459 and a large effective population size. Movement behavior within the wetland complex was more 460 strongly controlled by patch size rather than patch proximity, as larger wetland patches were the 461 preferred targets of local dispersers, regardless of distance. Large local dispersal distances of T. 462 coahuila illustrate that inter-wetland distance does not constrain connectivity, at least at the spatial 463 scale encompassed in this study, as there is no evidence for a distance decay relationship in the 464 dispersal kernel. Patterns of patch occupancy in T. coahuila reflect the behavior of a variety of taxa 465 that exhibit patch dynamics, where incidence and abundance positively correlate with patch area 466 (Hanski 1994). Although studies of dispersal and patch dynamics in aquatic turtles are limited, we find 467 that T. coahuila inter-wetland dispersal distances, frequency of movement, and elevated abundance in 468 large wetland patches approximates the behavior of another North American wetland turtle species, 469 Clemmys guttata (Joyal et al. 2001). In their study, Joyal et al. (2001) come to similar conclusions 470 about the role of dispersal in wetland turtle population dynamics at the local scale of a wetland 471 complex. Our findings of unconstrained and frequent movement between wetlands at the local scale 472 suggest that a wetland complex supports a single subpopulation rather than a metapopulation. 473 474 Despite genetic and demographic evidence for relatively unconstrained dispersal, there was a 475 significant relationship of isolation by distance across the geographic range of T. coahuila (Fig. 2). 476 The relationship is further reflected in significant pairwise site FST comparisons which demonstrate 477 differences in genotypic composition between Laguna Grande and four eastern localities (Table 2). 478 The range-wide pattern of isolation by distance was driven by unique genotypes at Laguna Grande, and 479 was likely in part influenced by distance and / or barrier-based isolation created by Sierra San Marcos, 480 as there was no evidence for isolation by distance and strong support for mixing in the eastern lobe of 481 the species’ range. The high degree of mixing in the eastern range lobe is especially well-illustrated by 482 migration patterns at the core site of Tio Cándido, where immigration was significantly higher than 483 emigration for all sites sampled except Charcos Prietos (Fig. 3). It is possible that over the larger 484 distances necessitated by Sierra San Marcos, the inhospitable terrestrial matrix could constrain 485 stepping stone movement among wetlands and result in such patterns of isolation by distance (Thomas 486 et al. 1998; Hurt & Hedrick 2004; Spinks & Shaffer 2005). Interestingly, however, historic migration 487 around Sierra San Marcos was stronger from west to east, indicating that the western side of the basin 488 may have served as a source of migrants (Fig. 3). 489 490 Comparing population structure in T. coahuila to patterns of genetic structure in sympatric 491 aquatic species endemic to Cuatro Ciénegas provides additional insight into evolutionary history. 492 Phylogeographic analyses of relatively dispersal-limited fishes, Cyprinodon spp. (Carson & Dowling 493 2006), and the spring snail Mexipyrgus churinceanus (Johnson 2005), indicate a unique evolutionary 494 history for subpopulations west of Sierra San Marcos. Mitochondrial DNA haplotype structure in M. 495 churinceanus revealed a bottleneck in subpopulations west of Sierra San Marcos dating back to the

496 Pleistocene, approximately 50 000 years before present (Johnson 2005), which may be associated with 497 the common phylogeographic break in turtles, fishes, and snails. Microevolutionary processes 498 structuring fine-scaled genetic variation in fishes and snails, however, likely acted at different spatial 499 scales from T. coahuila. For example, in M. churinceanus, haplotype frequencies reflected patterns of 500 isolation by distance within drainages (Johnson 2005). Cyprinodon spp. further exhibited strong 501 mtDNA structuring within multiple drainages located basin-wide (Carson & Dowling 2006). The 502 mechanisms shaping genetic structure in these taxa will likely not operate at the scale of a drainage in 503 T. coahuila given the species’ semiaquatic ecology and long-distance dispersal. Collectively, the 504 findings from Johnson (2005), Carson & Dowling (2006), and this study indicate that the western lobe 505 of the basin should be considered a high priority conservation site based upon the unique genetic 506 diversity and substantial localized aquatic habitat loss and degradation at Laguna Grande and 507 associated wetlands (documented in Johnson 2005; Van Dijk et al. 2007; Hendrickson et al. in press). 508 Immediate protection of the Laguna Grande wetland complex and other remaining wetland habitats 509 west of Sierra San Marcos is critical to conserve genetic variation in a suite of endangered sympatric 510 species (sensu Moritz & Faith 1998). 511 512 The evolutionary trajectory of T. coahuila appears sustainable pending preservation of a range- 513 wide network of wetland habitats. Continuous population structure across the geographic range 514 generated near-panmictic allele frequencies and yielded high genetic diversity and heterozygosity. As 515 a consequence, inbreeding depression and genetic drift are not of immediate concern in this species. 516 The threat of continued wetland habitat fragmentation and the associated ecological and genetic 517 consequences proves imminent, however (Souza et al. 2006; Van Dijk et al. 2007; Hendrickson et al. 518 in press). Minimizing extinction risk and maximizing adaptive potential in T. coahuila depends 519 critically upon ceasing habitat loss and preserving corridors that maintain metapopulation processes. 520 Ironically, much of modern habitat connectivity, including connections between Los Gatos and 521 Antiguos Mineros, is facilitated by abandoned canals that retain water and wetland vegetation year- 522 round and effectively serve as corridors (Howeth, personal observation). The next step in evaluating 523 conservation of T. coahuila could include simulating post-fragmentation patterns of gene flow between 524 sites and parameterizing the models with the site-specific allele frequencies determined from this 525 study. Connectivity between sites can be modified (decreased) according to hypotheses of future 526 localized habitat loss based upon predictive hydrologic models. The consequences of increasing 527 isolation for allele frequencies would identify critical corridors to maintain or restore. 528 529 Studies of neutral genetic variation in freshwater turtles reveal surprisingly high levels of 530 diversity (FitzSimmons & Hart 2007). Microsatellite variation and weak population differentiation in 531 T. coahuila is generally within the range represented by other aquatic turtles in the family Emydidae. 532 For example, pairwise population FST values range from 0.000 to 0.182 in Glyptemys insculpta 533 (Tessier et al. 2005), from -0.001 to 0.179 in Malaclemys terrapin (Hauswaldt & Glenn 2005) and 534 from 0.000 to 0.465 in Emydoidea blandingii (Mockford et al. 2007). All of these emydids are 535 officially recognized as threatened or endangered and have experienced substantial habitat 536 fragmentation and associated reductions in population size over the last century. While E. blandingii 537 shows distinct population substructure, populations of the other two species are relatively 538 undifferentiated. Taken together, our findings suggest that the long generation time of turtles may 539 provide a buffer to the loss of genetic diversity from habitat fragmentation. 540

541 In conclusion, this study employs an integrative and relatively novel approach to evaluating 542 dispersal and fragmentation effects at multiple spatial scales in a long-lived endangered vertebrate. 543 Direct observations of movement yielded insight into the spatial extent and frequency of dispersal and 544 thus the potential for evolutionarily effective gene flow. Complementary indirect estimates of 545 dispersal detected effective gene flow and elucidated patterns of population genetic substructure. It is 546 important to emphasize that direct approaches to evaluating dispersal usually assay movement and 547 demographic processes over one or two generations while indirect approaches yield an average over 548 tens to hundreds of generations and can detect effective gene flow and infrequent migration events 549 (Allendorf & Luikard 2007). To date, studies that take a similar comparative demographic and genetic 550 approach to evaluating dispersal typically address taxa with relatively short generation times, such as 551 damselflies (e.g. Coenagrion mercuriale; Watts et al. 2007) and fish (e.g. brook charr, Salvelinus 552 fontinalis; Wilson et al. 2004). In such species, allele frequencies may closely track fragmentation 553 effects and thus an integrative approach to evaluating dispersal can be relatively less informative. For 554 taxa with moderate to long generation times and high levels of historic gene flow, however, the 555 comparative dispersal assessment may provide an important framework to test for fragmentation- 556 induced limitation of contemporary movement. 557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572 References 573 574 Abell RA, Olson DM, Dinerstein E, et al. (2000) Freshwater Ecoregions of North America: A

575 Conservation Assessment. Island Press, Washington, D.C.

576 Allendorf FW, Luikard G (2007) Conservation and the Genetics of Populations. Blackwell Publishing,

577 Singapore.

578 Beerli P, Felsenstein J (2001) Maximum likelihood estimation of a migration matrix and effective

579 population sizes in n subpopulations by using a coalescent approach. Proceedings of the

580 National Academy of Sciences of the United States of America, 98, 4563-4568.

581 Bohonak AJ (1999) Dispersal, gene flow, and population structure. Quarterly Review of Biology, 74,

582 21-45.

583 Bohonak AJ, Jenkins DG (2003) Ecological and evolutionary significance of dispersal by freshwater

584 invertebrates. Ecology Letters, 6, 783-796.

585 Boulet M, Couturier S, Cote SD, Otto RD, Bernatchez L (2007) Integrative use of spatial, genetic, and

586 demographic analyses for investigating genetic connectivity between migratory, montane, and

587 sedentary caribou herds. Molecular Ecology, 16, 4223-4240.

588 Brown WS (1974) Ecology of the aquatic box turtle, Terrapene coahuila (Chelonia, Emydidae) in

589 Northern Mexico. Bulletin of the Florida State Museum, Biological Sciences, 1-67.

590 Carson EW, Dowling TE (2006) Influence of hydrogeographic history and hybridization on the

591 distribution of genetic variation in the pupfishes Cyprinodon atrorus and C. bifasciatus.

592 Molecular Ecology, 15, 667-679.

593 Clobert J, Ims RA, Rousset F (2004) Causes, consequences, and mechanisms of dispersal. In: Ecology,

594 Genetics, and Evolution of Metapopulations (eds. Hanski I, Gaggiotti OE), pp. 307-335.

595 Elsevier Academic Press, Burlington, Massachusetts.

596 Van Dijk PP, Flores-Villela O, Howeth J (2007) Terrapene coahuila. In: 2007 IUCN Red List of

597 Threatened Species. IUCN, www.iucnredlist.org downloaded on 15 March 2008.

598 Dodd CK, Ozgul A, Oli MK (2006) The influence of disturbance events on survival and dispersal rates

599 of Florida box turtles. Ecological Applications, 16, 1936-1944.

600 Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecular variance inferred from metric

601 distances among DNA haplotypes: application to human mitochondrial DNA restriction data.

602 Genetics, 131, 479-491.

603 FitzSimmons, NN, Hart KM (2007) Genetic studies of freshwater turtles and tortoises: A review of the

604 past 70 years. In: Defining Turtle Diversity: Proceedings of a Workshop on Genetics, Ethics,

605 and Taxonomy of Freshwater Turtles and Tortoises (eds. Shaffer HB, FitzSimmons NN,

606 Georges A, Rhodin, AGJ). Chelonian Research Monographs, 4, 15-46.

607 Freedberg S, Ewert MA, Ridenhour BJ, Neiman M, Nelson CE (2005) Nesting fidelity and molecular

608 evidence for natal homing in the freshwater turtle, Graptemys kohnii. Proceedings of the Royal

609 Society B-Biological Sciences, 272, 1345-1350.

610 Goudet J (1995) FSTAT (Version 1.2): a computer program to calculate F-statistics. Journal of

611 Heredity, 86, 485-486.

612 Hanski I (1994) Patch occupancy dynamics in fragmented landscapes. Trends in Ecology & Evolution,

613 9, 131-135.

614 Hanski I (1999) Metapopulation Ecology. Oxford University Press, New York.

615 Hastings A, Harrison S (1994) Metapopulation dynamics and genetics. Annual Review of Ecology and

616 Systematics, 25, 167-188.

617 Hauswaldt JS, Glenn TC (2005) Population genetics of the diamondback terrapin (Malaclemys

618 terrapin). Molecular Ecology, 14, 723-732.

619 Hendrickson, DA, Marks JC, Moline AB, Dinger EC, Cohen AE (in press 2008) Combining ecological

620 research and conservation: a case study in Cuatro Ciénegas, Coahuila, México. In: Aridland

621 Springs in North America: Ecology and Conservation (eds. Stevens, L, Meretsky, VJ).

622 University of Arizona Press, Tucson.

623 Hurt C, Hedrick P (2004) Conservation genetics in aquatic species: general approaches and case

624 studies in fishes and springsnails of arid lands. Aquatic Sciences, 66, 402-413.

625 Jensen JL, Bohonak AJ, Kelley ST (2005) Isolation by distance, web service. BMC Genetics, 6, 6.

626 Johnson SG (2005) Age, phylogeography and population structure of the microendemic banded spring

627 snail, Mexipyrgus churinceanus. Molecular Ecology, 14, 2299-2311.

628 Joyal LA, McCollough M, Hunter ML (2001) Landscape ecology approaches to wetland species

629 conservation: a case study of two turtle species in southern Maine. Conservation Biology, 15,

630 1755-1762.

631 Kiester AR, Schwartz CW, Schwartz ER (1982) Promotion of gene flow by transient individuals in an

632 otherwise sedentary population of box turtles (Terrapene carolina triunguis). Evolution, 36,

633 617-619.

634 King TL, Julian SE (2004) Conservation of microsatellite DNA flanking sequence across 13 Emydid

635 genera assayed with novel bog turtle (Glyptemys muhlenbergii) loci. Conservation Genetics, 5,

636 719-725.

637 Kuo CH, Janzen FJ (2004) Genetic effects of a persistent bottleneck on a natural population of ornate

638 box turtles (Terrapene ornata). Conservation Genetics, 5, 425-437.

639 Manel S, Schwartz MK, Luikart G, Taberlet P (2003) Landscape genetics: combining landscape

640 ecology and population genetics. Trends in Ecology & Evolution, 18, 189-197.

641 Marsh DM, Rand AS, Ryan MJ (2000) Effects of inter-pond distance on the breeding ecology of

642 tungara frogs. Oecologia, 122, 505-513.

643 Marsh DM, Trenham PC (2001) Metapopulation dynamics and amphibian conservation. Conservation

644 Biology, 15, 40-49.

645 Miller JK (2001) Escaping senescence: demographic data from the three-toed box turtle (Terrapene

646 carolina triunguis). Experimental Gerontology, 36, 829-832.

647 Minckley WL (1992) Three decades near Cuatro Cienegas, Mexico: photographic documentation and a

648 plea for area conservation. Journal of the Arizona-Nevada Academy of Science, 26, 89-118.

649 Mockford SW, Herman TB, Snyder M, Wright JM (2007) Conservation genetics of Blanding's turtle

650 and its application in the identification of evolutionarily significant units. Conservation

651 Genetics, 8, 209-219.

652 Moritz C, Faith DP (1998) Comparative phylogeography and the identification of genetically divergent

653 areas for conservation. Molecular Ecology, 7, 419-429.

654 Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) MICRO-CHECKER: software for

655 identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes, 4,

656 535-538.

657 Palsboll PJ, Berube M, Allendorf FW (2007) Identification of management units using population

658 genetic data. Trends in Ecology & Evolution, 22, 11-16.

659 Parker PG, Whiteman HH (1993) Genetic diversity in fragmented populations of Clemmys guttata and

660 Chrysemys picta marginata as shown by DNA fingerprinting. Copeia, 841-846.

661 Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetics software

662 for teaching and research. Molecular Ecology Notes, 6, 288-295.

663 Queller DC, Goodnight KF (1989) Estimating relatedness using genetic markers. Evolution, 43, 258-

664 275.

665 Raymond M, Rousset F (1995) GENEPOP (version 1.2): Population genetics software for exact tests

666 and ecumenicism. Journal of Heredity, 86, 248-249.

667 Rousset F (1997) Genetic differentiation and estimation of gene flow from F-statistics under isolation

668 by distance. Genetics, 145, 1219-1228.

669 Schneider S, Roessli D, Excoffier L (2000) Arlequin ver. 2.000: A software for population genetics

670 data analysis. Genetics and Biometry Laboratory, University of Geneva, Switzerland.

671 Schwartz MK, Luikart G, Waples RS (2007) Genetic monitoring as a promising tool for conservation

672 and management. Trends in Ecology & Evolution, 22, 25-33.

673 Secretaría del Medio Ambiente y Recursos Naturales (1999) Programa del Manejo del Área de

674 Protección de Flora y Fauna Cuatro Ciénegas México. Instituto Nacional de Ecología,

675 Tlacopac, Mexico.

676 Souza V, Espinosa-Asuar L, Escalante AE, et al. (2006) An endangered oasis of aquatic microbial

677 biodiversity in the Chihuahuan Desert. Proceedings of the National Academy of Sciences of the

678 United States of America, 103, 6565-6570.

679 Spinks PQ, Shaffer HB (2005) Range-wide molecular analysis of the western pond turtle (Emys

680 marmorata): cryptic variation, isolation by distance, and their conservation implications.

681 Molecular Ecology, 14, 2047-2064.

682 Storfer A, Murphy MA, Evans JS, et al. (2007) Putting the 'landscape' in landscape genetics. Heredity,

683 98, 128-142.

684 Templeton AR, Shaw K, Routman E, Davis SK (1990) The genetic consequences of habitat

685 fragmentation. Annals of the Missouri Botanical Garden, 77, 13-27.

686 Tessier N, Paquette S, Lapointe FJ (2005) Conservation genetics of the wood turtle (Glyptemys

687 insculpta) in Quebec, Canada. Canadian Journal of Zoology, 83, 765-772.

688 Thomas EP, Blinn DW, Keim P (1998) Do xeric landscapes increase genetic divergence in aquatic

689 ecosystems? Freshwater Biology, 40, 587-593.

690 Tilman D, Lehman CL, Kareiva P (1997) Population dynamics in spatial habitats. In: Spatial Ecology:

691 The Role of Space in Population Dynamics and Interspecific Interactions (eds. Tilman D,

692 Kareiva P). Princeton University Press, Princeton.

693 UNESCO-MAB (2006) World network of biosphere reserves. UNESCO, Paris, France.

694 USFWS (1973) Title 50 – Wildlife and Fisheries; Chapter 1 – Bureau of Sport Fisheries and Wildlife,

695 Fish and Wildlife Service, Department of the Interior; Part 17 – Conservation of Endangered

696 Species and Other Fish or Wildlife, Amendments to Lists of Endangered Fish and Wildlife.

697 Federal Register, 38, 14678.

698 Waples RS, Gaggiotti O (2006) What is a population? An empirical evaluation of some genetic

699 methods for identifying the number of gene pools and their degree of connectivity. Molecular

700 Ecology, 15, 1419-1439.

701 Watts PC, Rouquette JR, Saccheri J, Kemp SJ, Thompson DJ (2004) Molecular and ecological

702 evidence for small-scale isolation by distance in an endangered damselfly, Coenagrion

703 mercuriale. Molecular Ecology, 13, 2931-2945.

704 Watts PC, Rousset F, Saccheri IJ, et al. (2007) Compatible genetic and ecological estimates of

705 dispersal rates in insect (Coenagrion mercuriale : Odonata : Zygoptera) populations: analysis

706 of 'neighbourhood size' using a more precise estimator. Molecular Ecology, 16, 737-751.

707 Webb JK, Minckley WL, Craddock JE (1963) Remarks on the Coahuilan Box Turtle, Terrapene

708 coahuila (Testudines, Emydidae). Southwestern Naturalist, 8, 89-99.

709 Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure.

710 Evolution, 38, 1358-1370.

711 Wilson AJ, Hutchings JA, Ferguson MM (2004) Dispersal in a stream dwelling salmonid: inferences

712 from tagging and microsatellite studies. Conservation Genetics, 5, 25-37.

713

714

715 716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731 Acknowledgements

732 We thank Sari Albornoz, Margie Crisp, Luke McEachron, and especially Jack Siegrist for assistance in 733 the field. The SEMARNAT-Cuatro Ciénegas staff and Arturo Contreras graciously provided logistical 734 support. We are particularly indebted to Mathew Leibold and Fred Janzen for their financial support, 735 use of laboratory equipment, and enduring patience. Darrin Hulsey, Lisette de Senerpont Domis, the 736 Janzen Lab, and two anonymous referees provided comments that greatly improved earlier versions of 737 the manuscript. This project was funded by Chelonian Research Foundation and Nature Conservancy 738 grants to JGH and a Doctoral Dissertation Improvement Grant, NSF DEB 0508068, to Mathew 739 Leibold and JGH. JGH was supported by NSF DEB 0235579 and NSF DEB 0717370 to Mathew 740 Leibold. SEM was supported by an NSF Graduate Research Fellowship. Research was conducted 741 under SEMARNAT (3054, 5573) and CITES (072019, 140459) permits to JGH. 742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

758

759 Figure Legends

760 Figure 1. Maps from left to right: (1) The general location of the study area, Cuatro Ciénegas, in the 761 state of Coahuila, Mexico. (2) The Cuatro Ciénegas valley (bold outline) and the current geographic 762 range of the endemic Coahuilan box turtle, Terrapene coahuila (gray polygon). The geographic range 763 of T. coahuila is divided into a western and eastern lobe by the mountain, Sierra San Marcos. Seven 764 localities (circles) were represented in the population genetic analysis. The town of Cuatrociénegas de 765 Carranza (diamond) is the human population center of the valley and supports agriculture for the 766 region. (3) Los Gatos wetland habitats (gray polygons) served as the focal complex for the local 767 spatial scale mark-recapture study. 768 769 Figure 2. Genetic distance [FST/ (1-FST)] as a function of geographic distance (km) for all pairwise 770 comparisons of the seven Terrapene coahuila sample sites. Black circles represent site comparisons 771 within the eastern range lobe while gray circles indicate pairwise comparisons which include Laguna 772 Grande (Fig. 1). The relationship of isolation by distance in T. coahuila is driven by pairwise 773 comparisons with Laguna Grande. There is no evidence for regional dispersal limitation in 774 subpopulations to the east of Sierra San Marcos. Regression line incorporates all pairwise site 775 comparisons. 776 777 Figure 3. Patterns of asymmetric gene flow across the geographic range of Terrapene coahuila (gray 778 polygon). Arrows indicate the difference in pairwise site unidirectional migration (4Nm) rates where 779 95% confidence intervals do not overlap for the two estimates. Line thickness corresponds to the 780 difference in pairwise site migration rates: thin line, 4Nm = 0-2, medium line, 4Nm = 2.1-4, and thick 781 line, 4Nm = >4. See Table 3 for migration estimates and associated confidence intervals. Site name 782 abbreviations correspond to collection localities labeled in Figure 1. 783 784 Figure 4. Local dispersal by Terrapene coahuila within the wetland complex Los Gatos, Cuatro 785 Ciénegas, Mexico. (a) Proportion of T. coahuila local dispersal events by dispersal distance (Euclidian; 786 n = 17 dispersal events). There is no evidence for dispersal limitation at the local spatial scale. (b) 787 Terrapene coahuila abundance [mean (individuals / capture effort)] as a function of habitat patch 788 (wetland) area in hectares for each of the 37 wetlands surveyed at Los Gatos. Larger wetlands support 789 more individuals. Regression line incorporates turtle abundance at each wetland size per sampling 790 effort; 95% CI, dashed lines. 791 792 Figure 5. Regional dispersal by Terrapene coahuila between the wetland complexes of Los Gatos and 793 Antiguos Mineros in Cuatro Ciénegas, Mexico over the course of approximately one year. (a) 794 Proportion of T. coahuila captures per monthly capture effort at Los Gatos (n = 828 captures; June 795 2002 – July 2003, no data for December 2002 or June 2003) and Antiguos Mineros (n = 231 captures; 796 May 2002 – May 2003, no data for December 2002). (b) Proportion of Antiguos Mineros migrants in 797 the total capture at Los Gatos on the sampling date (gray bars) and proportion of Los Gatos migrants in 798 the total capture at Antiguos Mineros on the sampling date (black bars). Date on x-axis includes month 799 and year abbreviations. + denotes individual returning to original capture site.

Table 1. Summary of genetic diversity across nine microsatellite loci in Terrapene coahuila at seven sample sites in Cuatro Ciénegas, Mexico: sample size for microsatellites (N), number of alleles (A), number of effective alleles (Ae), number of private alleles (Ap), observed heterozygosity (HO), expected heterozygosity (HE), and an inbreeding coefficient (FIS ). Values are mean (SE). Sites are ordered from west to east.

Site N A Ae Ap HO HE FIS

Laguna Grande 15 6.56 (0.80) 4.17 (0.52) 0.22 (0.15) 0.81 (0.06) 0.72 (0.05) -0.08

Las Salinas 21 7.22 (0.91) 4.66 (0.69) 0.11 (0.11) 0.77 (0.07) 0.73 (0.05) -0.03

Tio Cándido 23 7.56 (0.92) 4.08 (0.60) 0.44 (0.18) 0.72 (0.05) 0.71 (0.04) 0.01

Charcos Prietos 27 7.67 (1.13) 4.26 (0.80) 0.00 0.70 (0.07) 0.68 (0.07) -0.01

Pozas Azules 24 7.89 (1.03) 4.39 (0.70) 0.33 (0.17) 0.71 (0.07) 0.71 (0.06) 0.01

Los Gatos 22 6.78 (0.78) 4.14 (0.60) 0.00 0.75 (0.05) 0.71 (0.05) -0.03

Antiguos Mineros 24 6.44 (0.96) 4.01 (0.63) 0.00 0.69 (0.06) 0.70 (0.05) 0.03

Table 2. Pairwise site FST values from nine microsatellite loci in Terrapene coahuila in Cuatro Ciénegas, Mexico. Site matrix includes FST values (above diagonal) for seven sample sites and the associated probability values (italicized; below diagonal). FST values in bold indicate significant pairwise differences after Bonferroni correction for multiple comparisons. Sites are ordered from west to east.

Site Laguna Las Tio Charcos Pozas Los Antiguos Grande Salinas Cándido Prietos Azules Gatos Mineros

Laguna Grande 0.010 0.020 0.040 0.022 0.027 0.041

Las Salinas 0.035 0.010 0.006 0.005 0.002 0.010

Tio Cándido 0.006 0.037 0.010 0.000 0.012 0.016

Charcos Prietos <0.001 0.117 0.026 -0.006 0.005 0.008

Pozas Azules <0.001 0.140 0.459 0.895 0.002 0.005

Los Gatos 0.001 0.562 0.022 0.118 0.294 0.012

Antiguos Mineros <0.001 0.045 0.007 0.063 0.182 0.032

Table 3. Migration matrix representing the number of migrants per generation (4Nm) for the seven Terrapene coahuila samples sites in Cuatro Ciénegas, Mexico. Migration estimates are unidirectional and convey immigration rates to a target site from a source site. 95% confidence intervals in italics. Sites are ordered from west to east.

Source Laguna Las Tio Charcos Pozas Los Antiguos Grande Salinas Cándido Prietos Azules Gatos Mineros Target

Laguna 3.364 1.322 2.334 3.734 1.349 1.533 Grande 2.763-4.047 0.958-1.767 1.835-2.917 3.097-4.454 0.982-1.796 1.124-2.035

Las 5.689 0.709 6.070 4.250 4.876 2.999 Salinas 4.830-6.642 0.436-1.079 5.181-7.035 3.511-5.085 4.076-10.177 2.386-3.707

Tio 5.068 7.731 4.626 10.118 10.416 5.869 Cándido 4.021-6.282 6.387-9.262 3.626-5.798 8.596-11.807 8.886-12.114 4.729-7.174

Charcos 2.516 2.188 3.056 5.351 3.981 4.786 Prietos 2.034-3.070 1.735-2.714 2.514-3.668 4.631-6.140 3.363-4.670 4.108-5.534

Pozas 1.702 2.774 6.849 6.187 6.383 4.720 Azules 1.283-2.203 2.227-3.401 5.972-7.803 5.360-7.096 5.536-7.310 4.000-5.521

Los 1.653 3.069 7.417 5.218 10.754 9.149 Gatos 1.157-2.252 2.399-3.857 6.389-8.539 4.360-6.182 9.423-12.173 7.998-10.406

Antiguos 1.519 2.325 3.512 3.353 3.222 3.101 Mineros 1.167-1.936 1.880-2.836 2.964-4.123 2.810-3.959 2.697-3.810 2.583-3.682 25

FIGURE 1

Texas Los Gatos Cuatrociénegas

Cuatro Las Salinas Ciénegas Charcos Prietos Los Gatos

Tio Cándido 0 500 1000 meters S Mexico i Laguna e r Grande r a Pozas Azules S a n

M Antiguos a r c Mineros o s

0 10 20 Kilometers

FIGURE 2

0.05

0.04 Laguna Grande Eastern sites

) 0.03 ST F 0.02 / (1- ST

F 0.01

0.00

-0.01 0 5 10 15 20 25 30 35

Geographic distance (km)

FIGURE 3

LS CP LG TC

LAG PA Sierra San Marcos AM

FIGURE 4

0.5 (a) ( 0.4

0.3

0.2

0.1 Proportion of dispersal events of dispersal Proportion

0.0 0 0 0 0 0 0 0 0 0 0 0 -5 10 15 20 25 30 35 40 45 50 55 0 1- 1- 1- 1- 1- 1- 1- 1- 1- 1- 5 10 15 20 25 30 35 40 45 50 Local dispersal distance (m)

(b) 2.6

2.2

1.8

1.4

1.0

0.6 Log (individuals / effort) (individuals Log 0.2

-0.2 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 Log (patch area, ha)

FIGURE 5

0.35 (a) Los Gatos 0.30 Antiguos Mineros

0.25

0.20

0.15

0.10 Proportion of total captures of total Proportion 0.05

0.00

0.05 (b) + 0.03

0.00 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 0 0 l 0 0 0 t 0 0 0 0 0 r 0 r 0 0 0 l 0 ay un u ug ep c ov ec an eb a p ay un u Proportion migrants Proportion M J J A S O N D J F M A M J J Date

Supplementary Material. Howeth et al. Dispersal in the Coahuilan box turtle

Table S1. Rarefied allelic richness at nine microsatellite loci in Terrapene coahuila.

Microsatellite Loci

Site GmuB8 GmuD16 GmuD21 GmuD40 GmuD62 GmuD70 GmuD90 GmuD107 GmuD121

Antiguos Mineros 3.974 7.147 3.000 9.853 5.716 7.728 4.966 6.577 3.798

Charcos Prietos 4.514 9.649 3.414 10.842 6.129 8.586 5.282 7.260 2.869

Laguna Grande 6.000 6.926 3.000 9.660 6.929 10.000 4.000 6.995 4.931

Las Salinas 4.666 9.175 2.667 9.811 7.642 8.775 5.883 6.525 4.550

Los Gatos 5.492 9.560 3.831 8.200 4.982 8.709 4.594 6.368 4.615

Pozas Azules 5.405 8.963 4.995 10.325 5.531 9.230 6.283 7.417 2.931

Tio Cándido 4.798 8.853 3.602 8.852 6.174 10.186 5.407 7.626 4.199

Table S2. Nei’s gene diversity at nine microsatellite loci in Terrapene coahuila.

Microsatellite Loci

Site GmuB8 GmuD16 GmuD21 GmuD40 GmuD62 GmuD70 GmuD90 GmuD107 GmuD121 Site Mean

Antiguos Mineros 0.628 0.850 0.626 0.890 0.723 0.772 0.763 0.774 0.394 0.713

Charcos Prietos 0.714 0.892 0.533 0.885 0.667 0.848 0.704 0.752 0.247 0.694

Laguna Grande 0.755 0.795 0.440 0.776 0.812 0.885 0.762 0.852 0.605 0.742

Las Salinas 0.706 0.886 0.469 0.876 0.836 0.858 0.802 0.789 0.474 0.744

Los Gatos 0.658 0.882 0.591 0.843 0.774 0.845 0.733 0.772 0.427 0.725

Pozas Azules 0.727 0.882 0.609 0.873 0.667 0.864 0.769 0.792 0.329 0.724

Tio Cándido 0.714 0.860 0.543 0.803 0.630 0.885 0.758 0.798 0.510 0.722

FIGURE S1

0.15

0.10 )

qg 0.05

0.00

-0.05 Relatedness (r

-0.10

-0.15 e s o s s s s d na id to le to ro an li d ie zu a e r a án Pr A G in G S C s s s M a as o o Lo s un L Ti rc oza uo g a P ig La h t C An

Site

Figure S1. Inter-individual relatedness (Queller and Goodnight relatedness coefficient mean, 95% CI) at the local scale, within each of the seven Terrapene coahuila samples sites. Open circles denote 95% confidence intervals for the null distribution defining expectations of relatedness under panmixia. Two sites, Antiguos Mineros and Charcos Prietos, supported individuals that are significantly more related than under the null model of panmixia. Sites are ordered from west to east.

FIGURE S2

1.0

0.8

0.6

0.4 Proportion

0.2

0.0 0.01-0.50 0.51-1.00 1.01-1.50 1.51-2.00 2.01-2.50 2.51-3.00 3.01-3.50 3.51-4.00 4.01-4.50 4.51-5.00 18.01-18.50 Patch area (ha)

Figure S2. Distribution of wetland patch area (ha) in the Los Gatos wetland complex, as shown in Figure 1 (n = 37 patches). Note break in x-axis scale.