Range-Wide Phylogeography of Blanding's Turtle [Emys

Ecology, Evolution and Organismal Biology Ecology, Evolution and Organismal Biology Publications

1-8-2019 Range-wide phylogeography of Blanding’s Turtle [Emys (= Emydoidea) blandingii] Mark A. Jordan Purdue University

Victoria Mumaw Purdue University

Natalie Millspaw Purdue University

Stephen W. Mockford Acadia University

Fredric J. Janzen Iowa State University, [email protected] Follow this and additional works at: https://lib.dr.iastate.edu/eeob_ag_pubs Part of the Genetics Commons, Natural Resources Management and Policy Commons, and the Population Biology Commons The ompc lete bibliographic information for this item can be found at https://lib.dr.iastate.edu/ eeob_ag_pubs/323. For information on how to cite this item, please visit http://lib.dr.iastate.edu/ howtocite.html.

This Article is brought to you for free and open access by the Ecology, Evolution and Organismal Biology at Iowa State University Digital Repository. It has been accepted for inclusion in Ecology, Evolution and Organismal Biology Publications by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Range-wide phylogeography of Blanding’s Turtle [Emys (= Emydoidea) blandingii]

Abstract Documentation of intraspecific eg netic lineages and their evolutionary history can provide insight for current and future conservation and management actions. The lB anding’s Turtle, Emys (=Emydoidea) blandingii, is a long-lived species with a relatively narrow latitudinal distribution centered around the Great Lakes, but extending from Nebraska to Nova Scotia. It is listed as endangered or threatened throughout most of its range mainly due to habitat loss. Microsatellite loci have been predominantly used to test and generate hypotheses concerning the number of evolutionarily significant units and the history of lineage diversification in this species. Here we describe haplotypes from two mitochondrial and three nuclear loci generated from 32 localities across the species’ range to provide an additional perspective on these hypotheses. Haplotype and nucleotide diversity were low in both sets of loci, with mitochondrial polymorphism comparable to the lowest found in any North American freshwater turtle. Spatial analyses of population differentiation supported the presence of two lineages with a boundary in eastern Ontario that is roughly associated with the Appalachian Mountains as proposed by Mockford et al (2007). We suggest that the low diversity in these loci is likely related to periodic range contractions and expansions associated with glacial cycles and that the two lineages recovered result from a deeper history of diversification. Our results further support previously identified evolutionary significant units and help to reconcile population structure found at smaller spatial scales, outcomes that will better inform conservation decision making for the species.

Keywords mitochondrial loci, nuclear loci, Pleistocene glaciation, lineage diversification, low genetic diversity

Disciplines Ecology and Evolutionary Biology | Genetics | Natural Resources Management and Policy | Population Biology

Comments This is a manuscript of an article published as Jordan, Mark A., Victoria Mumaw, Natalie Millspaw, Stephen W. Mockford, and Fredric J. Janzen. "Range-wide phylogeography of Blanding’s Turtle [Emys (= Emydoidea) blandingii]." Conservation Genetics (2019). doi: 10.1007/s10592-018-01140-6. Posted with permission.

This article is available at Iowa State University Digital Repository: https://lib.dr.iastate.edu/eeob_ag_pubs/323 Manuscript Click here to access/download;Manuscript;Jordan_BT_congen_ms_5.0_rev.d Click here to view linked References

2 Title: Range-wide Phylogeography of Blanding’s Turtle [Emys (=Emydoidea) blandingii]

3 Authors: Mark A. Jordan1*, Victoria Mumaw1, Natalie Millspaw1, Stephen W. Mockford2,

4 Fredric J. Janzen3

5 Addresses:

6 1Department of Biology, Purdue University Fort Wayne, Fort Wayne, IN 46805 USA

7 2Biology Department, Acadia University, Wolfville, Nova Scotia B4P 2R6 Canada

8 3Department of Ecology, Evolution and Organismal Biology, Iowa State University, Ames, IA

9 50011 USA

10 *Corresponding author: [email protected], 260-481-6315 (phone), 260-481-6087 (FAX)

11 Abstract: Documentation of intraspecific genetic lineages and their evolutionary history can

12 provide insight for current and future conservation and management actions. The Blanding’s

13 Turtle, Emys (=Emydoidea) blandingii, is a long-lived species with a relatively narrow latitudinal

14 distribution centered around the Great Lakes, but extending from Nebraska to Nova Scotia. It is

15 listed as endangered or threatened throughout most of its range mainly due to habitat loss.

16 Microsatellite loci have been predominantly used to test and generate hypotheses concerning the

17 number of evolutionarily significant units and the history of lineage diversification in this

18 species. Here we describe haplotypes from two mitochondrial and three nuclear loci generated

19 from 32 localities across the species’ range to provide an additional perspective on these

20 hypotheses. Haplotype and nucleotide diversity were low in both sets of loci, with mitochondrial

21 polymorphism comparable to the lowest found in any North American freshwater turtle. Spatial

22 analyses of population differentiation supported the presence of two lineages with a boundary in

23 eastern Ontario that is roughly associated with the Appalachian Mountains as proposed by

24 Mockford et al (2007). We suggest that the low diversity in these loci is likely related to

25 periodic range contractions and expansions associated with glacial cycles and that the two

26 lineages recovered result from a deeper history of diversification. Our results further support

27 previously identified evolutionary significant units and help to reconcile population structure

28 found at smaller spatial scales, outcomes that will better inform conservation decision making for

29 the species.

30 Key Words: mitochondrial loci, nuclear loci, Pleistocene glaciation, lineage diversification, low

31 genetic diversity

33 Introduction

34 Phylogeographic analysis is a cornerstone of conservation genetics. With the goal of conserving

35 genetic diversity within and among populations, the characterization of genetic lineages assists

36 with guiding the release of captively bred individuals or the translocation of individuals from

37 other populations. One challenge in realizing this goal is reconciling genetic information derived

38 from different molecular markers and sample localities. Adding new datasets to previously

39 studied species often yields differing, but ultimately deeper, perspectives on the level,

40 distribution, and history of intraspecific genetic variation.

41 Turtles have proven to be an interesting group for molecular analyses. Rates of DNA sequence

42 evolution are slow (Avise et al. 1992; Shaffer et al. 2013) and this should limit the rate of

43 diversification. Limited diversification is expected to be further prevalent in species that occur in

44 formerly glaciated regions (Lenk et al. 1999; Weisrock & Janzen 2000; Starkey et al. 2003;

45 Spinks & Shaffer 2005; Rosenbaum et al. 2007; Amato et al. 2008; Rödder et al. 2013).

46 Following range expansion, it is expected that there will be a broad distribution of few genetic

47 variants due to a limited number of founders and/or recolonization from a bottlenecked

48 population isolated in a refugium.

49 Blanding’s Turtle, Emys (=Emydoidea) blandingii, has been the subject of several conservation

50 genetic studies over the past decade. It has a northern distribution, ranging from Nebraska east

51 through the Great Lakes region, with disjunct populations in the northeastern United States and

52 Nova Scotia (Figure 1). Blanding’s Turtle is listed as threatened or endangered throughout most

53 of its geographic range and this is likely due to habitat loss, road mortality, and poaching

54 (Congdon et al. 2008). Its geographic range has been impacted by Pleistocene glaciations and the

55 current distribution may have been established from one or more glacial refugia (Mockford et al.

56 2007; Rödder et al. 2013). An early microsatellite analysis suggested the presence of three

57 evolutionarily significant units: 1) populations of the Great Lakes region west of the Appalachian

58 Mountains, 2) disjunct populations in New York and New England east of the Appalachians and

59 3) an isolated population in Nova Scotia (Mockford et al. 2007). More recently, microsatellite

60 analyses conducted with additional loci have revealed finer population structure on regional

61 scales including: the western portion of the geographic range from Nebraska to Illinois

62 (Sethuraman et al. 2014), within Ontario (Davy et al. 2014), within Illinois (Anthonysamy et al.

63 2017), and within Wisconsin (Reid et al. 2017). Further analyses are necessary to better

64 understand these patterns and reconsider the number of evolutionarily significant units.

65 To date, microsatellites have provided the bulk of population genetic information in Blanding’s

66 Turtle. This is likely due in part to low levels of sequence polymorphism found in mitochondrial

67 (mtDNA) and nuclear (nuDNA) loci in the species (Spinks & Shaffer 2009) which can constrain

68 resolution of population structure. However, analyses of mtDNA and nuDNA sequences are

69 better suited to understanding deeper intraspecific history due to their relatively slower overall

70 mutation rates, and this information can be synthesized with more rapidly evolving markers to

71 guide conservation decisions (Wang 2010; Weeks et al. 2011; Epps & Keyghobadi 2015). In

72 addition, differences in structure among microsatellites, mtDNA, and nuDNA, can occur due to a

73 range of processes, including variation in coalescence time, selective sweeps, and sex-biased

74 dispersal (Zink & Barrowclough 2008; Karl et al. 2012; Toews & Brelsford 2012). Comparisons

75 among markers have been well exemplified in the closely related Western Pond Turtle (Spinks &

76 Shaffer 2005; Spinks et al. 2010; Spinks et al. 2014) and European Pond Turtle (Lenk et al.

77 1999; Velo-Antón et al. 2008; Vamberger et al. 2015), but there is no geographically broad

78 sample of sequence polymorphism in E. blandingii. Here we carry out a range-wide

79 phylogeographic analysis of Blanding’s Turtle to: 1) characterize sequence variation in

80 mitochondrial and nuclear loci and 2) use these data to evaluate phylogeographic hypotheses to

81 inform conservation work.

82 Materials and Methods

83 Samples and Genetic Loci

84 Samples were primarily acquired from tissues collected as part of previous studies (Mockford et

85 al. 2007; Davy et al. 2014; Sethuraman et al. 2014; McCluskey et al. 2016; Reid et al. 2017), but

86 also included novel collections and museum specimens. A total 81 turtles representing 32

87 localities were included in the analysis (Fig. 1, Appendix A). The largest sample originated from

88 a locality in Nebraska (n = 10) while no samples were available from a disjunct region in the

89 states of Maine, New Hampshire, and Massachusetts. Although many samples consisted of

90 previously extracted DNA, some were available as tissues or blood. DNA was extracted using

91 the Qiagen DNAeasy Blood and Tissue Kit for both tissue types.

92 Taq-mediated polymerase chain reactions (PCR) were used to amplify loci in a volume of 25 µl.

93 Partial sequences of two mtDNA loci [control region (cr) and cytochrome b (cytb)] that were

94 expected to show intraspecific polymorphism based on prior studies on related species (Lenk et

95 al. 1999; Spinks & Shaffer 2005; Fritz et al. 2007) were used (Appendix B). Thermocycling for

96 the control region followed Spinks and Shaffer (2005). For both cytb segments a touchdown

97 protocol was employed (initial denaturation at 94°C for 2 min; 40 cycles denaturing at 94°C for

98 30 s, annealing at 60°C decreasing 0.5°C each cycle to 50°C for 30 s, extension at 72°C for 60 s;

99 followed by a final extension at 72°C for 10 min). A sample of ten nuDNA loci characterized in

100 E. marmorata by Spinks et al. (2014) were evaluated for amplification and polymorphism

101 (AMY1A, FSHR, NB10179, NB00504, NB12303, NB17367, NB10015, TRAF6, PTGER4, and

102 NB10005) using the abovementioned touchdown protocol but with hot start (initial denaturation

103 at 94°C for 10 min). Due to inconsistent amplification or a lack of polymorphism, only three of

104 the original ten nuDNA markers were retained for analysis: AMY1A, NB10005 and NB17367.

105 The former two loci are anonymous while the third locus (NB17367) is a partial proteasome

106 gene. All sequencing was carried out by Beckman Coulter Genomics

107 (www.beckmangenomics.com) using both forward and reverse primers for each individual.

108 Statistical Analysis

109 Sequences were verified, edited, trimmed, and aligned in Geneious v. 7.1.2 (Kearse et al. 2012).

110 We inspected aligned sequences by eye to identify single nucleotide polymorphisms (SNP). For

111 nuDNA markers, we reconstructed haplotypes using PHASE v. 2.1 (Stephens et al. 2001;

112 Stephens & Donnelly 2003) implemented in DnaSP v. 6.12.01 (Rozas et al. 2017). For each

113 locus we used 500 steps for burn-in, 1 thinning interval, and 1000 iterations in Markov chain-

114 Monte Carlo to estimate posterior probabilities of haplotypes under the default model of

115 recombination. A threshold probability of ≥ 0.7 was used to include a haplotype call (c.f.

116 Garrick et al. 2010) and three runs of the model were compared to verify haplotype assignments.

117 Descriptive statistics, nucleotide diversity, number of haplotypes, and haplotype diversity were

118 calculated in DnaSP. Samples missing data for one or more loci were excluded from multi-locus

119 analyses. In all analyses involving more than one locus, mitochondrial loci were concatenated

120 while nuclear loci were treated as independent samples of the genome.

121 Phylogeographic structure

122 We visualized the evolutionary relationship among samples using the Neighbor-Net algorithm

123 (Bryant & Moulton 2004) employed in SplitsTree v. 4.14.6 (Huson & Bryant 2006). Such

124 phylogenetic networks allow description of the ambiguity in phylogenetic datasets due to

125 sampling error or conflicting signal among loci. Networks for mtDNA and nuDNA were

126 constructed from uncorrected p-distances on concatenated and multiple loci, respectively. We

127 used POFAD v. 1.07 (Joly & Bruneau 2006) to calculate distance matrices for both sets of loci,

128 an approach that was chosen because it is able to account for allelic variation in nuDNA. Due to

129 limited polymorphism and a lack of resolution, networks for individual loci are not presented.

130 We assessed the presence and nature of phylogeographical structure using analyses included in

131 SPADS 1.0 (Dellicour & Mardulyn 2014). Global variation among localities was tested using

132 ɸST based on Euclidean distance of haplotypes (AMOVA , Excoffier et al. 1992), and the level of

133 phylogeographic signal was tested using the difference between NST and GST (Hardy &

134 Vekemans 2002). For statistical inference, randomizing individuals across localities is used for

135 ɸST while the difference between NST and GST relies on randomizing haplotypes across localities

136 (Dellicour & Mardulyn 2014). We also used SPADS to evaluate the phylogeographic structure

137 hypothesized by Mockford et al. (2007). Sample localities were classified into three groups:

138 Midwestern United States (localities 1 – 26), Ontario and New York (localities 27 – 31), and

139 Nova Scotia (locality 32). This classification assumes that the Nebraska sample locality,

140 unsampled in the earlier study (Mockford et al. 2007), would cluster with other Midwest

141 localities. It should also be noted that our dataset lacked samples from one of the northeastern

142 regions (Massachusetts, New Hampshire, and Maine) and had limited coverage in New York. In

143 These analyses tested the hierarchical levels of variation among hypothesized groups (ɸCT ),

144 variation across localities range-wide (ɸST), and variation across localities within groups (ɸSC )

145 permuted following Excoffier et al. (1992).

146 We further used SPADS to cluster sample localities without designating a priori groups using

147 spatial analysis of molecular variance (SAMOVA) (Dupanloup et al. 2002) and Monmonier’s

148 algorithm implemented in BARRIER v. 2.2 (Manni et al. 2004). SAMOVA seeks to maximize

149 differentiation among population clusters while accounting for relative geographic position,

150 whereas BARRIER directly identifies genetic breaks and therefore provides indirect evidence of

151 population structure. Both analyses rely on a network of sample localities generated by Delaunay

152 triangulation and Voronoi tessellation (described in Manni et al. 2004). SAMOVA criteria were

153 set as the number of hypothesized populations, K, ranging from two to ten with 10,000 iterations

154 (the number of repetitions of steps 5-9 of the SAMOVA algorithm) applied in ten replicate runs.

155 When interpreting the SAMOVA results, the best value for K was chosen using the highest ɸCT

156 value provided there was more than one individual in the cluster. For nuDNA loci, the multi-

157 locus ɸ-statistics are a weighted average of the individual loci but for mtDNA loci we analyzed a

158 concatenated sequence to account for the non-independent inheritance of cytb and cr. The

159 BARRIER analysis was conducted using interindividual genetic distance matrices generated in

160 SPADS (IID2, an uncorrected p-distance derived from Miller 2005; Miller et al. 2006).

161 Following the recommendations of Manni et al. (2004), we tested for the presence of isolation by

162 distance since algorithm may construct barriers due to large geographic distances among a few

163 localities. Alleles in Space (Miller 2005) was used to test for the relationship between IID2 and

164 geographic distance, and if a correlation was found a matrix of IID2 residuals was used to

165 identify barriers. Finally, the GDisPAL function described in SPADS was employed to visualize

166 the interpolation of interindividual genetic distances across the geographic range. We ran the

167 function in R version 3.4.1 (R Core Team 2018) using a raster of the Blanding’s Turtle range, the

168 geographic coordinates of the sample localities, and residual IID2. The resulting interpolation

169 surface was saved as a raster file and projected as a map in QGIS version 2.18.10 (Team 2009).

170 Demographic analysis

171 To investigate the hypothesis of population growth following glacial retreat, we calculated

172 neutrality statistics derived from expectations under coalescent theory (reviewed in Ramos-

173 Onsins & Rozas 2002) in DnaSP v. 6.12.01 (Rozas et al. 2017). Statistics included were

174 Tajima’s D (Tajima 1989), Fu’s Fs (Fu 1997), and Ramos-Onsins and Rozas’ R2 (Ramos-Onsins

175 & Rozas 2002), each having different effectiveness under various demographic scenarios, levels

176 of recombination, and sample sizes (Ramírez-Soriano et al. 2008). We employed coalescent

177 simulations, assuming a null hypothesis of constant population size over time and an infinite

178 allele model of mutation (Standard Neutral Model in DnaSP), to calculate 95% confidence

179 intervals and assess whether the observed metrics deviated from neutrality. The simulations

180 were run with 1000 replicates and employed values of θ estimated from the data.

181

182 Results

183 Genetic Diversity

184 Between 44 and 67 individuals were genotyped for each locus, with a combined total of 42 for

185 mtDNA loci and 55 for nuDNA loci (Table 1). Mitochondrial cr and cytb had similar levels of

186 polymorphism with four segregating sites each, all of which were substitutions except for one

187 deletion in cr. The number of haplotypes (cr = 6, cytb = 7) and haplotype diversity were similar,

188 but nucleotide diversity was slightly lower in cytb. The number of segregating sites ranged from

189 two to six in each of the three nuDNA loci (Table 1) and all were substitutions. We found a total

190 of 14 haplotypes across nuDNA loci, with haplotype number varying from three (NB17367) to

191 six (AMY1A). A relatively low haplotype diversity was observed for NB10005 and this is due

192 to most haplotypes being rare in the sample. Levels of nucleotide diversity were comparable to

193 those found in mtDNA loci but with a relatively high value found in AMY1A.

194 Phylogeographic structure

195 Descriptive relationships among samples in split networks suggested some geographic separation

196 but with conflicting partitions as indicated by parallel edges (Fig. 2). The mtDNA network (Fig.

197 2A) was simpler and relatively less ambiguous. Except for one individual from western Ontario,

198 samples from Nova Scotia, New York, and eastern Ontario were divided from other samples.

199 Nebraska individuals and one individual from eastern Iowa also appeared to be separated from

200 other samples. The nuDNA network (Fig. 2B) was much less resolved with multiple parallel

201 edges and samples from across geographic range occupying similar splits. However, there is

202 some suggestion that samples from Nova Scotia sort together.

203 We found more support for differentiation in statistical inference of haplotype variation. The

204 null hypothesis of panmixia across the geographic range (global ɸST) was rejected in all loci, but

205 only concatenated mtDNA loci suggested a strong signal of phylogeographic structure (NST-GST,

206 Table 2). Although the data tended to support the presence of three phylogeographic groups

207 (ɸCT) hypothesized by Mockford et al. (2007), there was a lack of structure in mitochondrial CR

208 and nuclear AMY1A in AMOVA. The SAMOVA analysis supported the presence of three

209 phylogeographic clusters in concatenated mtDNA data and two phylogeographic clusters in

210 multi-locus nuDNA data. Sample localities in the three mtDNA clusters included: Grant County,

211 Nebraska (locality 1), the midwestern US east to Parry Sound, Ontario (localities 1 – 2, 4 – 7, 9,

212 11 – 17, 21 - 27), and eastern Ontario to Nova Scotia (localities 28 - 29, 31 - 32). The two

213 clusters in nuDNA loci included only Nova Scotia samples in one (locality 32) and all other

214 sample localities in the other (localities 1, 3, 5 – 10, 12, 14 – 17, 19 – 26, 27 - 29). When the

215 clusters were used to define groups in AMOVA, there was an overall increase in ɸCT values

216 relative to the a priori grouping hypothesized by Mockford (Table 2). This was particularly the

217 case with mtDNA data where each locus and the concatenated loci increased by at least 0.25.

218 Individual nuDNA data appeared to be less resolved, but the multi-locus ɸCT of increased from

219 0.29 in a priori groups to 0.46 in SAMOVA clusters.

220 We found a positive correlation between interindividual geographic distance and genetic distance

221 in both sets of markers (mtDNA r = 0.62, P <0.0001; nuDNA r = 0.41, P < 0.0001). Residuals

222 from these matrices (residual IID2) resulted in similar clustering in BARRIER as those found in

223 SAMOVA (not shown). When barriers were sequentially added in the mtDNA dataset, the first

224 barrier split off localities in eastern Ontario, New York and Nova Scotia (localities 28 - 29, 31 –

225 32) from the rest of the range before the second barrier split off Nebraska samples. In the

226 nuDNA dataset, the one barrier led to the clustering of Jefferson, New York (locality 29) with

227 Nova Scotia. These results were represented in the interpolation of the residuals of both datasets

228 using the GDisPAL function (Fig. 3).

229 Demographic analysis

230 Overall, we did not find support for a historical bottleneck followed by range expansion in the

231 neutrality tests (Table 3). Only the nuclear locus NB10005 had statistical support for expected

232 negative values in Tajima’s D under this scenario.

233 Discussion

234 Low Genetic Variation

235 The difference in levels of genetic variation among species provides useful context for

236 considering diversification history within Blanding’s Turtle (Table 4). We found that sequence

237 polymorphism in E. blandingii is low in the loci of both genomes, confirming data collected at a

238 more limited scale in an interspecific analysis (Spinks & Shaffer 2009). The low variability is

239 particularly evident in mtDNA, loci which have a longer history of use in phylogeography and

240 are better characterized. Taxa closely related to E. blandingii have more segregating sites

241 regardless of sample size for mtDNA loci. For example, the number of segregating sites is

242 seven- to ten-fold higher in E. marmorata for control region and cytochrome b, respectively.

243 Meanwhile, only G. muhlenbergii (Rosenbaum et al. 2007) has levels of genetic variation

244 comparable to E. blandingii. Therefore, although mtDNA sequence variation is low in turtles

245 overall (Avise et al. 1992), species such as E. blandingii and G. muhlenbergii represent the lower

246 bound of this variability. Meanwhile, levels of nuDNA locus polymorphism are similar to E.

247 marmorata but the sample size used for E. mamorata was an order of magnitude lower.

248 Geographic range expansion and contraction associated with glacial cycles during the

249 Pleistocene could be an important contributor to this pattern (Hewitt 2000). North America

250 experienced three such cycles in the past 320,000 years, with the most recent glaciation

251 (Wisconsin) ending ~20,000 years ago. In species with northern distributions, genetic

252 bottlenecks can occur as distributions contract during cold periods leaving minimal genetic

253 variation among future colonizers. If colonization is rapid following deglaciation, the limited

254 polymorphism is expected to be distributed over a large geographic area. The Blanding’s Turtle

255 distribution appears to be particularly sensitive to glaciation (Rödder et al. 2013). In an analysis

256 of 59 North American species, climate–based models of the paleogeographic range of E.

257 blandingii show it to have among the smallest refugia and that these refugia occur for extended

258 time periods during glacial events. Other turtle species with northern distributions such as G.

259 insculpta and G. muhlenbergii suggest similar correspondence between variability in geographic

260 range size (Rödder et al. 2013) and reduced genetic variation (Table 4). Meanwhile, the same

261 models applied to the more genetically diverse E. marmorata indicate a more stable distribution

262 and the potential for larger refugial populations.

263 Given the patterns of range dynamics and overall levels of genetic variation among species, we

264 expected neutrality tests within E. blandingii to deviate from a null model of constant population

265 size and neutral evolution. Demographic expansion following glaciation should increase the

266 number of recent mutations leading to higher haplotype diversity relative nucleotide diversity.

267 For the majority of the loci and tests, this expectation was not met (Table 3), tempering our

268 ability to make a strong conclusion concerning post-glacial expansion of effective population

269 size.

270 Other explanations for low genetic variation in turtle species include: slow rates of molecular

271 evolution related to long generation time (Avise et al. 1992; Shaffer et al. 2013), recent history of

272 diversification (Weisrock & Janzen 2000), and selective sweeps (Rosenbaum et al. 2007; Amato

273 et al. 2008). Although each mechanism may have an important role in particular species, or

274 turtles in general, none appear to account for the observed data. Blanding’s Turtle does have a

275 relatively long generation time (estimated at 37 years, Congdon et al. 1993), but it is comparable

276 to Wood Turtle (36-47 years, van Dijk & Harding 2011), a species with an order of magnitude

277 higher level of cytb polymorphism (Table 3). The divergence time hypothesis suggests

278 evolutionary rates may not be slow but rather that recently evolved species have not had

279 sufficient time to develop sequence polymorphism. Although estimates of divergence dates

280 within Emys encompass a wide range (9 to 35 mybp), the most extensive phylogenetic analysis

281 of the group to date provides no evidence that E. blandingii has a much more recent origin than

282 any of its congeners (Spinks et al. 2016). Finally, the selective sweep hypothesis is often invoked

283 to explain low levels of mitochondrial polymorphism since directional selection on a particular

284 locus will lead to correlated evolution across the mitochondrial genome due to linkage. Such

285 selection is not expected to impact variation in unlinked nuDNA or microsatellite loci. The

286 broadly concordant population structure we observe in range-wide analyses among the three sets

287 of loci (discussed below) is more consistent with neutral evolutionary processes rather than

288 selection acting in parallel in two or more lineages. Clearly, characterization of many more loci

289 is needed to resolve questions surrounding the low genetic variation in the species and develop a

290 more detailed understanding of population history (c.f. Spinks et al. 2014).

291 Phylogeographic Structure

292 Despite the limited polymorphism we observed, the presence of two genetic groups with a likely

293 boundary located in eastern Ontario was supported by both sets of loci (Table 2, Fig. 3). This

294 pattern generally corresponds with the structure found by Mockford et al. (2007) but with some

295 notable differences that likely contribute to the incomplete sorting of haplotypes (Fig. 2) and the

296 better performance of the SAMOVA analysis. Mockford et al. (2007) identified a primary

297 boundary through eastern Ontario and northern New York, and a secondary boundary within the

298 eastern region that divided Massachusetts and Nova Scotia populations from Ontario and New

299 York populations. Our mitochondrial dataset supported the presence of the first boundary but did

300 not recover separation within the eastern region. Additionally, mtDNA separated the Nebraska

301 locality and grouped Parry Sound, Ontario near Lake Huron (not sampled by Mockford) with

302 midwestern US populations. Nuclear data led to the identification of a boundary in the eastern

303 part of the range but phylogeographic resolution was weaker (Table 2) with the SAMOVA

304 analysis and the interpolation of interindividual distances differing in the position of the putative

305 boundary (Nova Scotia singly isolated, or Nova Scotia clustered with Jefferson, New York).

306 Overall, our results point to the Appalachian Mountains as a fundamental barrier within the

307 geographic range, with possible colonization to eastern Ontario of eastern lineages through the

308 Hudson River valley (Mockford et al. 2007).

309 Two explanations have been proposed for the history of E. blandingii populations that could

310 explain the observed phylogeographic structure. In one scenario (Schmidt 1938; Smith 1957),

311 climate warming following glacial retreat during the Hypsithermal (9000 - 4000 ybp) led to the

312 expansion of prairie habitat eastward into the Atlantic region. Blanding’s Turtle populations are

313 thought to have followed this expansion only to retreat later as the climate cooled, leaving

314 relictual populations in the eastern part of the geographic range. An alternative explanation is

315 that refugia may have occurred both along the Atlantic coastal plain and in the southern Great

316 Plains regions during glaciations (Bleakney 1958a; Bleakney 1958b), therefore suggesting a

317 periodic history of isolation. Blanding’s Turtle fossils occur predominantly in the Great Plains

318 and Midwest, dating as early as the late Hemiphillian (5 mybp), (Parmley 1992; Holman 1995),

319 but there is one specimen from eastern Pennsylvania from the late Irvingtonian (0.75 mybp)

320 (Parris & Daeschler 1995). The possibility of an Atlantic refuge is supported by this fossil

321 observation and is concordant with molecular lineages found within Chrysemys picta (Starkey et

322 al. 2003) and G. insculpta (Amato et al. 2008). The slow rate of molecular evolution in turtles in

323 general (Avise et al. 1992; Shaffer et al. 2013), combined with these observations, suggests to us

324 that the population structure represents ancient polymorphism from an earlier history of

325 divergence.

326 The broad pattern of phylogeographic structure that we observed provides an interesting contrast

327 with the consistent population genetic clustering found in microsatellite analyses conducted at

328 smaller geographic scales (Mockford et al. 2005; Davy et al. 2014; Sethuraman et al. 2014;

329 McCluskey et al. 2016; Anthonysamy et al. 2017; Reid et al. 2017). It is clear from these studies

330 that the addition of more microsatellite loci increases the power to resolve finer genetic

331 differences among populations (pairwise FST < 0.2). This differentiation has been linked most

332 often to the history of post-glacial dispersal and the boundaries of large watersheds (Davy et al.

333 2014; Sethuraman et al. 2014; McCluskey et al. 2016), but recent work has begun to find subtle

334 structure related to fragmentation in contemporary landscapes (Anthonysamy et al. 2017; Reid et

335 al. 2017). The levels of genetic variation in microsatellite loci found in these studies is also

336 moderately high, an observation that is likely due to the maintenance of alleles during

337 bottlenecks in long-lived species with overlapping generations (Kuo & Janzen 2004; Willoughby

338 et al. 2013; Davy et al. 2014) and the relatively high mutation rate of these markers. We

339 therefore suggest that the contrast between sequence and microsatellite datasets is one of

340 temporal scale, where the latter is more appropriate for detecting processes operating more

341 recently within and among populations (Wang 2010; Epps & Keyghobadi 2015).

342 Management for genetic populations

343 The conditions that have made Blanding’s Turtle populations at risk of continued decline persist.

344 Genetic analyses are fundamental to determining which populations are at the greatest risk, how

345 habitat connectivity can lead to gene flow, and where individuals might be drawn from when

346 population augmentation becomes necessary. The observation that contemporary gene flow is

347 nearly non-existent among Blanding’s Turtle populations in most regions (Davy et al. 2014;

348 Sethuraman et al. 2014; Anthonysamy et al. 2017; Reid et al. 2017) and apparent inbreeding

349 depression (Sethuraman et al. 2014) makes the possibility of such interventions more acute.

350 Combined genetic information available for Blanding’s Turtle points to hierarchical levels of

351 variation that should guide direct management. At the level of the geographic range, our

352 generally concordant results between mitochondrial and nuclear loci reinforce the evolutionarily

353 significant units (ESU) identified by Mockford et al. (2007) that were based on a handful of

354 microsatellite loci. These ESUs represent deeper, historical divergence within the species with

355 the Appalachian Mountains as a barrier. Therefore, management at the scale of the geographic

356 range should seek to maintain populations representing this diversity under the additional

357 assumption that adaptive differences exist between regions (Crandall et al. 2000). Meanwhile,

358 the lack of natural gene flow identified in contemporary populations indicates a level of

359 demographic independence that is aligned with the concept of management units, distinct

360 populations nested within the larger ESUs (Palsbøll et al. 2007). The more recent work using

361 expanded arrays of microsatellite loci has uncovered important population structure that should

362 prominently factor into management decisions, as exemplified by the genetic clusters discovered

363 west of the Appalachian Mountains.

364 The synthesis of information derived from phylogeographic and landscape genetic analyses

365 provides an opportunity to conduct management more effectively (Epps & Keyghobadi 2015).

366 For example, conservation efforts where the two lineages meet in Ontario and New York (Fig. 3)

367 should be considered carefully given that neighboring populations near the contact zone are

368 expected to have a longer history of divergence than that found within lineages. Although

369 adaptive variation across the range has not been investigated in the species, translocations across

370 this boundary should elevate the risk of outbreeding depression resulting from disruption of co-

371 adapted gene complexes developed over longer time. Additionally, landscape genetic analyses

372 that identify habitat connectivity associated with gene flow can benefit from controlling for such

373 phylogeographic breaks. This overall approach can be adopted within the eastern and western

374 portions of the range to identify more recent, but significant, histories of divergence to more

375 effectively promote population connectivity and resilience through habitat restoration and

376 translocation.

377 Acknowledgements

378 This research was supported by the Indiana Academy of Science and the IPFW Honors Program.

379 We thank Christina Davy, Mike Finkler, Rusty Gonser, Dave Mifsud, Brendan Reid, and Sasha

380 Tetzlaff for sharing samples with us. The Field Museum and the Royal Ontario Museum were

381 helpful in providing additional samples. We also appreciate the insights shared by Phil Spinks on

382 choosing polymorphic nuclear loci.

383 Literature Cited 384 385 Amato ML, Brooks RJ, Fu J (2008) A phylogeographic analysis of populations of the wood 386 turtle (Glyptemys insculpta) throughout its range. Mol Ecol, 17, 570-581. 387 Anthonysamy WJB, Dreslik MJ, Douglas MR, Thompson D, Klut GM, Kuhns AR, Mauger D, 388 Kirk D, Glowacki GA, Douglas ME, Phillips CA (2017) Population genetic evaluations 389 within a co-distributed taxonomic group: A multi-species approach to conservation planning. 390 Anim Conserv, 21, 137-147. 391 Avise JC, Bowen BW, Lamb T, Meylan AB, Bermingham E (1992) Mitochondrial DNA 392 evolution at a turtle's pace: evidence for low genetic variability and reduced 393 microevolutionary rate in the Testudines. Mol Biol Evol, 9, 457-473. 394 Bleakney JS (1958a) A zoogeographical study of the amphibians and reptiles of eastern Canada. 395 National Museum of Canada Bulletin, 155, 1-119. 396 Bleakney S (1958b) Postglacial dispersal of the turtle Chrysemys picta. Herpetologica, 14, 101- 397 104. 398 Bryant D, Moulton V (2004) Neighbor-net: an agglomerative method for the construction of 399 phylogenetic networks. Mol Biol Evol, 21, 255-265. 400 Congdon JD, Dunham AE, van Loben Sels R (1993) Delayed sexual maturity and demographics 401 of Blanding's Turtles (Emydoidea blandingii): implications for conservation and management 402 of long‐ lived organisms. Conserv Biol, 7, 826-833. 403 Congdon JD, Graham TE, Herman TB, Lang JW, Pappas MJ, Brecke BJ (2008) Emydoidea 404 blandingii (Holbrook 1838) - Blanding's Turtle. In:Rhodin AGJ, Pritchard PCH, van Dijk PP, 405 Saumure RA, Buhlmann KA, Iverson JB (eds) Conservation Biology of Freshwater Turtles 406 and Tortoises: A Compilation Project of the IUCN/SSC Tortoise and Freshwater Turtle 407 Specialist Group Chelonian Research Monographs No 5. p.^pp. 015.011-015.012. 408 Crandall KA, Bininda-Emonds OR, Mace GM, Wayne RK (2000) Considering evolutionary 409 processes in conservation biology. Trends Ecol Evol, 15, 290-295. 410 Davy CM, Bernardo PH, Murphy RW (2014) A Bayesian approach to conservation genetics of 411 Blanding’s turtle (Emys blandingii) in Ontario, Canada. Conserv Gen, 15, 319-330. 412 Dellicour S, Mardulyn P (2014) SPADS 1.0: a toolbox to perform spatial analyses on DNA 413 sequence data sets. Mol Ecol Res, 14, 647-651. 414 Dupanloup I, Schneider S, Excoffier L (2002) A simulated annealing approach to define the 415 genetic structure of populations. Mol Ecol, 11, 2571-2581. 416 Epps CW, Keyghobadi N (2015) Landscape genetics in a changing world: disentangling 417 historical and contemporary influences and inferring change. Mol Ecol, 24, 6021-6040. 418 Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecular variance inferred from metric 419 distances among DNA haplotypes: applications to human mitochonridal DNA. Genetics, 420 131, 479-491. 421 Fritz U, Guicking D, Kami H, Arakelyan M, Auer M, Ayaz D, Fernández CA, Bakiev AG, 422 Celani A, Džukić G (2007) Mitochondrial phylogeography of European pond turtles (Emys 423 orbicularis, Emys trinacris)–an update. Amphib-Reptilia, 28, 418-426. 424 Fu Y-X (1997) Statistical tests of neutrality of mutations against population growth, hitchhiking 425 and background selection. Genetics, 147, 915-925. 426 Garrick RC, Sunnucks P, Dyer RJ (2010) Nuclear gene phylogeography using PHASE: dealing 427 with unresolved genotypes, lost alleles, and systematic bias in parameter estimation. BMC 428 Evol Biol, 10, 118.

429 Hardy OJ, Vekemans X (2002) SPAGeDI: a versatile computer program to analyse spatial 430 genetic structure at the individual or population levels. Mol Ecol Notes, 2, 618-620. 431 Hewitt G (2000) The genetic legacy of the Quaternary ice ages. Nature, 405, 907-913. 432 Holman JA (1995) Pleistocene Amphibians and Reptiles in North America. Oxford University 433 Press, New York. 434 Huson DH, Bryant D (2006) Application of phylogenetic networks in evolutionary studies. Mol 435 Biol Evol, 23, 254-267. 436 Joly S, Bruneau A (2006) Incorporating allelic variation for reconstructing the evolutionary 437 history of organisms from multiple genes: an example from Rosa in North America. Syst 438 Biol, 55, 623-636. 439 Karl SA, Toonen R, Grant W, Bowen B (2012) Common misconceptions in molecular ecology: 440 echoes of the modern synthesis. Mol Ecol, 21, 4171-4189. 441 Kearse M, Moir R, Wilson A, Stones-Havas S, Cheung M, Sturrock S, Buxton S, Cooper A, 442 Markowitz S, Duran C (2012) Geneious Basic: an integrated and extendable desktop 443 software platform for the organization and analysis of sequence data. Bioinformatics, 28, 444 1647-1649. 445 Kuo C-H, Janzen FJ (2004) Genetic effects of a persistent bottleneck on a natural population of 446 ornate box turtles (Terrapene ornata). Conserv Gen, 5, 425-437. 447 Lenk P, Fritz U, Joger U, Wink M (1999) Mitochondrial phylogeography of the European pond 448 turtle, Emys orbicularis (Linnaeus 1758). Mol Ecol, 8, 1911-1922. 449 Manni F, Guerard E, Heyer E (2004) Geographic patterns of (genetic, morphologic, linguistic) 450 variation: how barriers can be detected by using Monmonier's algorithm. Hum Biol, 76, 173- 451 190. 452 McCluskey EM, Mockford SW, Sands K, Herman TB, Johnson G, Gonser RA (2016) Population 453 genetic structure of Blanding's Turtles (Emydoidea blandingii) in New York. J Herpetol, 50, 454 70-76. 455 Miller MP (2005) Alleles In Space (AIS): computer software for the joint analysis of 456 interindividual spatial and genetic information. J Hered, 96, 722-724. 457 Miller MP, Bellinger MR, Forsman ED, Haig SM (2006) Effects of historical climate change, 458 habitat connectivity, and vicariance on genetic structure and diversity across the range of the 459 red tree vole (Phenacomys longicaudus) in the Pacific Northwestern United States. Mol Ecol, 460 15, 145-159. 461 Mockford S, McEachern L, Herman T, Snyder M, Wright JM (2005) Population genetic 462 structure of a disjunct population of Blanding’s turtle (Emydoidea blandingii) in Nova 463 Scotia, Canada. Biol Conserv, 123, 373-380. 464 Mockford SW, Herman TB, Snyder M, Wright JM (2007) Conservation genetics of Blanding's 465 turtle and its appliction in the identification of evolutionary significant units. Conserv Gen, 8, 466 209-219. 467 Palsbøll PJ, Bérubé M, Allendorf FW (2007) Identification of management units using 468 population genetic data. Trends in Ecology and Evolution, 22, 11-16. 469 Parmley D (1992) Turtles from the late Hemphillian (latest Miocene) of Knox County, Nebraska. 470 Tex J Sci, 44, 339-348. 471 Parris DC, Daeschler E (1995) Pleistocene turtles of Port Kennedy cave (late Irvingtonian), 472 Montgomery County, Pennsylvania. J Paleontol, 69, 563-568. 473 R Core Team (2018) R: A Language and Environment for Statistical Computing. R Foundation 474 for Statistical Computing, Vienna, Austria.

475 Ramírez-Soriano A, Ramos-Onsins SE, Rozas J, Calafell F, Navarro A (2008) Statistical power 476 analysis of neutrality tests under demographic expansions, contractions and bottlenecks with 477 recombination. Genetics, 179, 555-567. 478 Ramos-Onsins SE, Rozas J (2002) Statistical properties of new neutrality tests against population 479 growth. Mol Biol Evol, 19, 2092-2100. 480 Reid BN, Mladenoff DJ, Peery MZ (2017) Genetic effects of landscape, habitat preference and 481 demography on three co-occurring turtle species. Mol Ecol, 26, 781-798. 482 Rödder D, Lawing AM, Flecks M, Ahmadzadeh F, Dambach J, Engler JO, Habel JC, Hartmann 483 T, Hörnes D, Ihlow F (2013) Evaluating the significance of paleophylogeographic species 484 distribution models in reconstructing Quaternary range-shifts of Nearctic chelonians. PloS 485 ONE, 8, e72855. 486 Rosenbaum PA, Robertson JM, Zamudio KR (2007) Unexpectedly low genetic divergences 487 among populations of the threatened bog turtle (Glyptemys muhlenbergii). Conserv Gen, 8, 488 331-342. 489 Rozas J, Ferrer-Mata A, Sánchez-DelBarrio JC, Guirao-Rico S, Librado P, Ramos-Onsins SE, 490 Sánchez-Gracia A (2017) DnaSP 6: DNA sequence polymorphism analysis of large data sets. 491 Mol Biol Evol, 34, 3299-3302. 492 Schmidt KP (1938) Herpetological evidence for the postglacial eastward expansion of the steppe 493 in North America. Ecology, 19, 396-407. 494 Sethuraman A, McGaugh SE, Becker ML, Chandler CH, Christiansen JL, Hayden S, LeClere A, 495 Monson-Miller J, Myers EM, Paitz RT, Refsnider JM, VanDeWalle TJ, Janzen FJ (2014) 496 Population genetics of Blanding's turtle (Emys blandingii) in the midwestern United States. 497 Conserv Gen, 15, 61-73. 498 Shaffer H, Minx P, Warren D, Shedlock A, Thomson R, Valenzuela N, Abramyan J, Amemiya 499 C, Badenhorst D, Biggar K, Borchert G, Botka C, Bowden R, Braun E, Bronikowski A, 500 Bruneau B, Buck L, Capel B, Castoe T, Czerwinski M, Delehaunty K, Edwards S, Fronick C, 501 Fujita M, Fulton L, Graves T, Green R, Haerty W, Hariharan R, Hernandez O, Hillier L, 502 Holloway A, Janes D, Janzen F, Kandoth C, Kong L, de Koning A, Li Y, Literman R, 503 McGaugh S, Mork L, O'Laughlin M, Paitz R, Pollock D, Ponting C, Radhakrishnan S, Raney 504 B, Richman J, St John J, Schwartz T, Sethuraman A, Spinks P, Storey K, Thane N, Vinar T, 505 Zimmerman L, Warren W, Mardis E, Wilson R (2013) The western painted turtle genome, a 506 model for the evolution of extreme physiological adaptations in a slowly evolving lineage. 507 Genome Biology, 14, R28. 508 Smith PW (1957) An analysis of post-Wisconin biogeography of the prairie peninsula region 509 based on distributional phenomena among terrestrial vertebrate populations. Ecology, 38, 510 205-218. 511 Spinks PQ, Shaffer HB (2005) Range-wide molecular analysis of the western pond turtle (Emys 512 marmorata): cryptic variation, isolation by distance, and their conservation implications. Mol 513 Ecol, 14, 2047-2064. 514 Spinks PQ, Shaffer HB (2009) Conflicting mitochondrial and nuclear phylogenies for the widely 515 disjunct Emys (Testudines: Emydidae) species complex, and what they tell us about 516 biogeography and hybridization. Syst Biol, 58, 1-20. 517 Spinks PQ, Thomson RC, McCartney-Melstad E, Shaffer HB (2016) Phylogeny and temporal 518 diversification of the New World pond turtles (Emydidae). Mol Phylogenet Evol, 103, 85-97.

519 Spinks PQ, Thomson RC, Shaffer HB (2010) Nuclear gene phylogeography reveals the historical 520 legacy of an ancient inland sea on lineages of the western pond turtle, Emys marmorata in 521 California. Mol Ecol, 19, 542-556. 522 Spinks PQ, Thomson RC, Shaffer HB (2014) The advantages of going large: genome‐ wide 523 SNPs clarify the complex population history and systematics of the threatened western pond 524 turtle. Mol Ecol, 23, 2228-2241. 525 Starkey DE, Shaffer HB, Burke RL, Forstner MRJ, Iverson JB, Janzen FJ, Rhodin AGJ, Ultsch 526 GR (2003) Molecular systematics, phylogeography, and the effects of Pleistocene glaciation 527 in the painted turtle (Chrysemys picta) complex. Evolution, 57, 119-128. 528 Stephens M, Donnelly P (2003) A comparison of bayesian methods for haplotype reconstruction 529 from population genotype data. The American Journal of Human Genetics, 73, 1162-1169. 530 Stephens M, Smith NJ, Donnelly P (2001) A new statistical method for haplotype reconstruction 531 from population data. Am J Hum Genet, 68, 978-989. 532 Tajima F (1989) Statistical method for testing the neutral mutation hypothesis by DNA 533 polymorphism. Genetics, 123, 585-595. 534 Team QD (2009) QGIS Geographic Information System. Open Source Geospatial Foundation. 535 Toews DP, Brelsford A (2012) The biogeography of mitochondrial and nuclear discordance in 536 animals. Mol Ecol, 21, 3907-3930. 537 Vamberger M, Stuckas H, Sacco F, D'Angelo S, Arculeo M, Cheylan M, Corti C, Lo Valvo M, 538 Marrone F, Wink M (2015) Differences in gene flow in a twofold secondary contact zone of 539 pond turtles in southern Italy (Testudines: Emydidae: Emys orbicularis galloitalica, E. o. 540 hellenica, E. trinacris). Zool Scr, 44, 233-249. 541 van Dijk PP, Harding J (2011) Glyptemys insculpta (errata version published in 2016), The 542 IUCN Red List of Threatened Species. http://dx.doi.org/10.2305/IUCN.UK.2011- 543 1.RLTS.T4965A11102820.en. Accessed 18 April 2018 544 Velo-Antón G, García-París M, Rivera AC (2008) Patterns of nuclear and mitochondrial DNA 545 variation in Iberian populations of Emys orbicularis (Emydidae): conservation implications. 546 Conserv Gen, 9, 1263-1274. 547 Wang IJ (2010) Recognizing the temporal distinctions between landscape genetics and 548 phylogeography. Mol Ecol, 19, 2605-2608. 549 Weeks AR, Sgro CM, Young AG, Frankham R, Mitchell NJ, Miller KA, Byrne M, Coates DJ, 550 Eldridge MD, Sunnucks P (2011) Assessing the benefits and risks of translocations in 551 changing environments: a genetic perspective. Evolutionary Applications, 4, 709-725. 552 Weisrock DW, Janzen FJ (2000) Comparative molecular phylogeography of North American 553 softshell turtles (Apalone): implications for regional and wide-scale historical evolutionary 554 forces. Mol Phylogenet Evol, 14, 152-164. 555 Willoughby JR, Sundaram M, Lewis TL, Swanson BJ (2013) Population decline in a long-lived 556 species: the wood turtle in Michigan. Herpetologica, 69, 186-198. 557 Zink RM, Barrowclough GF (2008) Mitochondrial DNA under siege in avian phylogeography. 558 Mol Ecol, 17, 2107-2121. 559

Table 1. Summary statistics for mitochondrial and nuclear sequences, respectively, in Emys blandingii. Data presented are: the number of base pairs (bp), the number of polymorphic sites (S), the number of individuals genotyped (N), the number of haplotypes (haps), haplotype diversity (H +/- standard deviation) and nucleotide diversity (π +/- standard deviation).

Genome Locus bp S N haps H π

Mitochondrial Control region 579 4 67 6 0.63 ± 0.05 0.0023 ± 0.0002

Cytochrome b 1096 4 44 7 0.65 ± 0.07 0.0013 ± 0.0001

Concatenated 1675 8 42 9 0.68 ± 0.07 0.0015 ± 0.0002

Nuclear AMY1A 400 4 58 6 0.52 ± 0.04 0.0037 ± 0.0002

NB10005 535 6 57 5 0.28 ± 0.05 0.0017 ± 0.0004

NB17367 703 2 55 3 0.56 ± 0.03 0.0009 ± 0.0001

Table 2. Tests of genetic structure within and among loci. Overall statistics indicate the presence of structure (Global ɸST) range-wide and the degree of phylogeographical structure (NST-GST). One set of AMOVA results use the three phylogeographical regions hypothesized by Mockford et. al (2007) with sample localities classified as described in the text. The second set of AMOVA results uses the phylogeographic groups identified in SAMOVA. Between the two AMOVA analyses, the higher of the two ɸCT is shown in bold type. Statistical significance is based on 1000 permutations following procedures specific to each analysis as described in the SPADS 1.0 manual (Dellicour & Mardulyn 2014).

Genome Locus Overall AMOVA AMOVA (K = 3 c.f Mockford et al. 2007) (SAMOVA KmtDNA = 3, KnuDNA = 2)

Global ɸST NST-GST ɸCT ɸST ɸSC ɸCT ɸST ɸSC

Mitochondrial Control region 0.765*** 0.030 0.402 0.823*** 0.704*** 0.871** 0.854*** -0.123

Cytochrome B 0.830*** 0.253* 0.629*** 0.892*** 0.709*** 0.877*** 0.901*** 0.186

Concatenated 0.772*** 0.336*** 0.631*** 0.848*** 0.588*** 0.895*** 0.871*** -0.229

Nuclear AMY1A 0.180** 0.012 -0.103 0.087 0.172 -0.110 0.055 0.148

NB10005 0.412*** 0.012 0.653* 0.604*** -0.144 0.770* 0.768*** -0.009

NB17367 0.215*** 0.005 0.221** 0.147 -0.094 0.164 0.169 0.007

Multi-locus 0.260*** 0.040 0.298*** 0.337*** 0.056 0.453*** 0.499*** 0.084 * P < 0.05, ** P < 0.01, *** P < 0.001

Table 3. Results of neutrality tests conducted in DnaSP v. 6.12. For each measure, the observed value of the statistic is given alongside 95% confidence intervals and p-values estimated from coalescent simulations that assume constant population size and an infinite-sites model of mutation. Samples include localities in the western portion of the range (localities 1 – 27) to avoid the possible influence of population structure on the tests.

Genome Locus n Tajima’s D Fu’s Fs Ramos-Onsins and Roza’s R2

Observed Simulated P Observed Simulated P Observed Simulated P 95% CI 95% CI 95% CI

Mitochondrial Control region 57 1.46 -1.58 – 2.02 0.91 0.33 -3.94 – 4.97 0.63 0.19 0.05 – 0.25 0.94

Cytochrome B 37 0.20 -1.56 – 2.09 0.61 -1.27 -2.84 – 4.77 0.22 0.13 0.05 – 0.23 0.57

Concatenated 33 0.13 -1.63 – 2.03 0.58 -0.23 -3.43 – 5.18 0.48 0.13 0.06 – 0.21 0.57

Nuclear AMY1A 94 1.78 -1.62 – 2.14 0.96 1.73 -4.89 – 5.71 0.81 0.18 0.04 – 0.19 0.97

NB10005 94 -1.55 -1.39 – 2.06 0.02 -2.33 -3.94 – 3.60 0.09 0.03 0.02 – 0.25 0.08

NB17367 94 0.93 -1.43 – 2.16 0.81 1.27 -3.94 – 4.40 0.77 0.16 0.02 – 0.24 0.84

Multi-locus 94 0.39 -1.05 – 1.22 0.77 0.22 -2.55 – 2.53 0.61 0.12 0.05 – 0.17 0.81

Table 4. Comparison of the proportion of segregating sites among related taxa (Emys orbicularis, Emys marmorata, Glyptemys muhlenbergii,and Glyptemys insculpta) and Emys blandingii. The geographic extent and size of the samples used for comparison is shown.

Genome Locus Species S bp p-distance N Geographic Source Range

Mitochondrial Control region E. blandingii 4 579 0.007 68 Full This study

E. orbicularis 10 660 0.015 51 Partial Velo-Anton et al. (2008)

E. marmorata 31 630 0.049 135 Full Spinks and Shaffer (2005)

G. muhlenbergii 2 750 0.003 74 Full Rosenbaum et al. (2007)

Cytochrome B E. blandingii 4 1096 0.004 44 Full This study

E. orbicularis 21 1031 0.020 406 Full Vamberger et al. (2015)

E. marmorata 47 1070 0.044 21 Full Spinks and Shaffer (2009)

G. muhlenbergii 4 1179 0.003 74 Full Rosenbaum et al. (2007)

G. insculpta 20 750 0.027 117 Full Amato et al. (2008)

Nuclear AMY1A E. blandingii 4 400 0.010 58 Full This study

E. marmorata 4 518 0.008 8 Full Spinks et al. (2014)

NB10005 E. blandingii 6 535 0.011 57 Full This study

E. marmorata 6 639 0.009 8 Full Spinks et al. (2014)

NB17367 E. blandingii 2 703 0.003 55 Full This study

E. marmorata 2 644 0.003 8 Full Spinks et al. (2014)

Figure 1. Map of the geographic range of Emys blandingii and sample localities of this study. Locality numbers are described in the text and are identified in Appendix A. The maximum of extent of the Wisconsin glaciation (18,000 ybp) is shown. Figure 2. Split networks of uncorrected p-distances from concatenated mitochondrial (A) and multi-locus nuclear (B) loci using the NeighborNet algorithm. Nodes are labelled with the state or province of the sample with numbers in parentheses representing more than one individual. * indicates samples from western Ontario (locality 27). Figure 3. Interpolation of interindividual genetic distances for mitochondrial (A) and nuclear (B) DNA sequences across the geographic range of Emys blandingii. Genetic distances are based on residuals of the relationship between genetic and geographic distance calculated in SPADS 1.0. Lightly colored regions indicate areas of relatively greater genetic distance.