Conservation Genetics of Freshwater Turtles · amplification of two loci in the snapping turtle (chelydra serpentina) 12 2 abstract 12 2.1 primer note 13 2.2 acknowledgments 15 2.3

i

Conservation Genetics of Freshwater Turtles

by

Christina Maria Davy

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy in Zoology

Department of Ecology and Evolutionary Biology University of Toronto

© Copyright by Christina M. Davy, 2013

ii

Conservation Genetics of Freshwater Turtles

Christina M. Davy

Doctor of Philosophy in Zoology

Department of Ecology and Evolutionary Biology University of Toronto

2013

Abstract

Turtles have long life spans, overlapping generations and promiscuous mating systems. Thus,

they are an ideal model system with which to investigate the application of conservation genetics

methods and assumptions to long-lived organisms. Turtles are also one of the most threatened

groups of vertebrates and conservation genetics studies are essential to effective recovery of

turtle species. This thesis has two main objectives: 1) to evaluate some common population

genetics assumptions with respect to turtles and other long-lived organisms, and 2) to collect

important information on the population genetics of threatened turtles in Ontario, which can be

used to inform species recovery. In Chapters Two and Three, I describe the development of

novel microsatellite markers for the snapping turtle and spiny softshell. In Chapter Four I

demonstrate significant genetic structure in populations of the endangered spotted turtle in

Ontario, and find that “bottleneck tests” may fail to detect recent population declines in small

turtle populations. I also show that spotted turtles do not show the typical correlation between

population size and genetic diversity. In Chapter Five I use microsatellite markers developed in

Chapter Two and document population structure in the widespread snapping turtle for the first

time. I compare these results with results from Chapter Four to test the traditionally accepted

hypothesis that genetic diversity is reduced in small, isolated populations compared to large,

connected populations. As in Chapter Four, my results suggest that the usual patterns of genetic

iii

structure and loss of diversity may not apply to turtles. In Chapter Six I conduct a conservation

genetics study of the endangered Blanding’s turtle. Finally, in Chapter Seven I combine results

from spotted, snapping and Blanding’s turtles to test whether vagility predicts population

structure, genetic diversity and significant barriers to gene flow in three species sampled across a

single landscape. Analyses reveal minimal congruence in barriers to gene flow and the three

species show unexpected and contrasting patterns of diversity across the landscape. Discordant

patterns among species highlight areas for further research and shed light on possible cryptic

behaviour, and I discuss potential further directions for research in the Summary.

iv

Acknowledgments

First of all, thank you to Leif, Katja and the Davy and Einarson families for all your support and

“grenzenloses Vertrauen”.

I wrote this thesis in the first person because I’ve been informed that that is how one writes a

thesis. While I will happily take credit for the work I have done (and of course for anything that

requires correction or improvement), I dislike this convention because this research was

supported in many ways by a great many wonderful people. If annoyingly long lists of names

appear below, it is because I have been fortunate to have help of many kinds from many corners.

Dr. Bob Murphy is an incredible supervisor and his generosity towards his students is practically

limitless. I need to point out his eccentricities simply because I think it will make him happy. He

has many, and they kept life at the ROM entertaining, hilarious and inspiring. Although Bob has

worked with virtually every possible “herp” out there, this project represents (to my knowledge)

his first foray into the world of North American turtles. I am grateful that he was so open-minded

about the direction my thesis took, and I hope he has enjoyed the experience too. I am also

grateful for the numerous “extra” opportunities Bob provides to students in his lab; in my case,

studies of Ctenosaura and the Seri Indians, whipping frogs from Vietnam, multiple paternities in

desert tortoises and some pretty great teaching experiences. I have learned so much, and laughed

quite a bit too.

Thank you to the rest of my advisory committee, Dr. Deborah McLennan and Dr. Chris Wilson.

Chris was my population genetics guru and helped me make some sense out of biogeographic

patterns that seemed arbitrary at first. Deborah’s detailed and thoughtful comments on my

manuscripts greatly improved the clarity of the final version, and she has inspired me to keep

improving my writing. I am grateful to my committee for providing enough guidance to get me

going while also giving me room to take a slightly exploratory approach to this project.

Bob, Deborah and Chris - thank you also for being so understanding and encouraging when I

announced that I would be taking maternity leave.

v

Thank you to my fellow graduate students at the ROM, especially Christopher Blair, Ida

Conflitti, Pedro Bernardo, Pamela Wong and Andre Ngo, for discussions, relevant and otherwise,

and for sharing parts of the journey. Amy Lathrop, Kristen Choffe, Oliver Haddrath and

Christopher Blair introduced me to molecular laboratory analyses and answered endless

questions about reaction recipes and lab methods.

While I was working in Mauritius over Christmas 2007, I was chatting with another field

technician and asked her if she would consider working on a turtle project in Canada. She

(unexpectedly) said yes, and the crazy lady proceeded to join me four summers of field work

catching turtles in Ontario. To Suzanne Coombes – thank you. I might have been able to do it

without you, but if so, it would have been much less fun. Ashley Leifso joined the party a little

later but has been equally wonderful, and kept things running smoothly through buckets of baby

softshells and later on in the lab, where we did manage to finish genotyping (10 days before the

baby arrived).

This project is built on the previous studies of turtles in Ontario, many of which originated in the

labs of Dr. Ron Brooks and, more recently, Dr. Jacqueline Litzgus. I thank both of them for their

hard work, inspiring research and dedication to the study and conservation of turtles. Dr. Jackie

Litzgus, David Seburn, Scott Gillingwater and Joe Cebek provided both encouragement and

helpful suggestions during my thesis work, and Dr. Fred Schueler, Dr. Francis Cook and Dr.

Frank Ross also provided helpful suggestions and useful information. Dr. Brock Fenton provided

much-needed random moments of bat-related distraction and inspiration, and helped me maintain

perspective.

Irene and Till Davy, Leif Einarson, Dr. Jackie Litzgus, Johnston Miller and David Seburn made

helpful comments on previous versions and their suggestions were very much appreciated. Joe

Crowley, John Urquhart and James Patterson provided helpful comments and several memorable

debates (as a result of which I’m still unsure whether Blanding’s turtles really do or do not occur

on the Bruce Peninsula).

Thank you to the Litzgus lab – especially Amanda Bennett, James Baxter-Gilbert, Matt Keevil,

James Patterson, Megan Rasmussen, Julia Riley and Katherine Yagi – for hours of turtle-talk,

laughter, and “adopting” me into their lab at conferences.

vi

Working with species at risk in Ontario is constrained by a vast tangle of red tape. This is

probably necessary, but it can slow things down considerably. When I told Bob that I wanted to

work in Ontario, he informed me that he was happy with this, as long as I took care of the

permits. I have since learned the reason why...So for helping me to navigate the maze of

regulations and obtain the six different pieces of paper I required each year, I thank Amelia

Argue, Corina Brdar, Melody Cairns, Stephanie Chan, Sarah Crosgrey, Tammy Dobbie, Sandy

Dobbyn, Mike Gatt, Ron Gould, Pud Hunter, Alistair Mackenzie, Andrew Promaine, Emily

Slavik, Roxanne St. Martin, Scott Sutton, Scott Taylor and all the other MNR and Canada Parks

staff whom I did not contact directly but who helped get the necessary papers processed. Phew.

Access to sites and logistical support were generously provided by the Ausable Bayfield

Conservation Authority, Jackie Litzgus, Ontario Hydro, Ontario Nature (Mark Carabetta and

John Urquhart), Ontario Parks, Parks Canada, Megan Rasmussen, Rick MacArthur, David

Seburn, the Nature Conservancy of Canada, South Nation Conservation Authority, Anne and

Katherine Yagi. Additional samples were contributed by Brennan Caverhill, Joe Cebek, Scott

Gillingwater, Bob Johnson, Jeremy Rouse, David Seburn and Jim Trottier.

My work on the south shore of Lake Huron would not have been possible without

accommodations generously provided by the Fraser-Green family in 2008, and by Mrs.

Stephanie Donaldson in 2009 – 2011.

A large portion of this thesis was made possible by the support of Wildlife Preservation Canada.

To WPC and to Elaine Williams, my Conservation Fairy Godmother – thank you. I would not

have been able to start or complete my studies without the support of an Ontario Graduate

Scholarship and a CGS from NSERC. The Blanding’s turtle chapter was funded largely by a

SARRFO grant through the Toronto Zoo; thanks to Bob Johnson and Julia Phillips for making

this happen. Pedro Bernardo helped get the Blanding’s turtle genotyping done on time. Finally,

thanks to the Till Eulenspiegel Foundation for filling in the gaps where necessary.

It takes a lot of swanp-walking to find a lot of spotted turtles. Thanks to James Baxter-Gilbert,

Christopher Blair, Hope Brock, Mark Carabetta, Suzie Coombes, Laila Copes, Joe Crowley, Eric

Davy, Leif Einarson, Kari Jean, Chris Law, Ashley Leifso, Jackie Litzgus, Steve Marks, Melissa

Oddie, Karen Paquette, James Patterson, Crystal Roberston, Michelle Scheerder, David Seburn,

Will (Kum C.) Shim, Emily Slavik, Dan Storisteanu, John Urquhart, Silu Wang, Amy Whitear,

vii

and everyone else who came out and helped with surveys. Thanks also to everyone who has

helped or is helping with turtle projects that are not included in this thesis (head-starting projects,

righting-time experiments, etc.) – it continues to be an adventure and I am so lucky to have such

wonderful people on the “turtle team”.

This was not an uncomplicated journey. The details of the various road-blocks I encountered are

not important - but all the support I had while overcoming them is. I am privileged to have had

the opportunity to spend more than four years immersed in a subject I love. I have learned so

much, and I am inspired by how much more there is to learn.

To my friends – thank you for helping me to stay grounded.

Returning to my parents, Veronika, Eric, my grandmother, my in-laws (and sibling-in-laws!),

and especially to Leif – the gratitude I feel for your support is something I would rather express

to you directly than share publicly in a thesis, so I’ll do that.

Finally, thanks to my parents for introducing me to the turtles... and thanks, of course, to the

turtles.

viii

Table of Contents

Table of Contents

ACKNOWLEDGMENTS IV

TABLE OF CONTENTS VIII

LIST OF TABLES XII

CHAPTER 1 CONSERVATION GENETICS AND FRESHWATER TURTLES: A GENERAL

INTRODUCTION 1

1 1

1.1 OUTLINE 1

1.2 CONSERVATION GENETICS: OBJECTIVES AND CHALLENGES 1

1.3 HOW SMALL IS SMALL AND WHAT ARE WE MEASURING? 3

1.4 TURTLES AND CONSERVATION GENETICS OF LONG-LIVED ORGANISMS 4

1.5 CONSERVATION GENETICS OF FRESHWATER TURTLES IN ONTARIO 6

1.6 REFERENCES 8

CHAPTER 2 CHARACTERIZATION OF TEN NOVEL MICROSATELLITE LOCI AND CROSS-

AMPLIFICATION OF TWO LOCI IN THE SNAPPING TURTLE (CHELYDRA SERPENTINA) 12

2 ABSTRACT 12

2.1 PRIMER NOTE 13

2.2 ACKNOWLEDGMENTS 15

2.3 REFERENCES 15

CHAPTER 3 ISOLATION AND CHARACTERIZATION OF ELEVEN NOVEL POLYMORPHIC

MICROSATELLITE LOCI IN THE SPINY SOFTSHELL TURTLE (APALONE SPINIFERA) 18

3 ABSTRACT 18

3.1 PRIMER NOTE 19

3.2 ACKNOWLEDGEMENTS 21

3.3 REFERENCES 21

ix

CHAPTER 4 CONSERVATION GENETICS OF THE ENDANGERED SPOTTED TURTLE DO NOT

SUPPORT A RELATIONSHIP BETWEEN GENETIC VARIATION AND POPULATION SIZE. 25

4 ABSTRACT 25

4.1 INTRODUCTION 26

4.2 METHODS 28

4.2.1 SAMPLE COLLECTION AND GENOTYPING 28

4.2.2 POPULATION GENETICS ANALYSES 30

4.2.3 DEVELOPMENT OF GENETIC ASSIGNMENT TESTS FOR CANADIAN CL. GUTTATA 32

4.2.4 ANALYSES OF GENETIC BOTTLENECKS 32

4.3 RESULTS 32

4.4 DISCUSSION 35

4.4.1 BIOGEOGRAPHY AND CONSERVATION GENETICS OF CLEMMYS GUTTATA 36

4.4.2 MANAGEMENT IMPLICATIONS 37

4.4.3 LONG-LIVED ORGANISMS (TURTLES) AND LOSS OF DIVERSITY IN FRAGMENTED POPULATIONS 38

4.4.4 BOTTLENECK TESTS AND LONG-LIVED ORGANISMS 39

4.5 ACKNOWLEDGEMENTS 40

4.6 REFERENCES 41

CHAPTER 5 UNEXPECTED PATTERNS OF GENETIC DIVERSITY IN TWO SYMPATRIC SPECIES

OF TURTLE. 59

5 ABSTRACT 59

5.1 INTRODUCTION 60

5.2 METHODS 61

5.2.1 STUDY SPECIES 61

5.2.2 DATA COLLECTION AND ANALYSES – CHELYDRA SERPENTINA 62

5.2.3 DATA COLLECTION AND ANALYSES – CLEMMYS GUTTATA 64

5.2.4 INTERSPECIFIC COMPARISONS 64

5.2.5 DATA ACCESSIBILITY 64

5.3 RESULTS 64

5.3.1 BAYESIAN CLUSTERING ANALYSES 65

5.3.2 POPULATION DIFFERENTIATION 65

5.3.3 INTERSPECIFIC COMPARISON 66

5.4 DISCUSSION 66

x

5.4.1 SUMMARY 70


5.6 REFERENCES 71

CHAPTER 6 CONSERVATION GENETICS OF BLANDING’S TURTLE (EMYS BLANDINGII) IN

ONTARIO, CANADA. 85

6 ABSTRACT 85

6.1 INTRODUCTION 86

6.2 METHODS 88

6.3 RESULTS 90

6.4 DISCUSSION 92


6.6 REFERENCES 96

CHAPTER 7 GENOTYPES AND GHOSTS: COMPARATIVE LANDSCAPE GENETICS REVEALS

INCONGRUENT BARRIERS TO GENE FLOW AMONGST THREE SPECIES OF FRESHWATER

TURTLE 108

7 ABSTRACT 108

7.1 INTRODUCTION 109

7.2 METHODS 112

7.2.1 STUDY SPECIES AND RELATIVE DISPERSAL ABILITY 112

7.2.2 BAYESIAN DELINEATION OF POPULATION BOUNDARIES 112

7.2.3 BARRIER ESTIMATION WITH MONMONIER’S ALGORITHM 113

7.2.4 ESTIMATION OF MIGRATION AMONG POPULATIONS 113

7.3 RESULTS 114

7.3.1 BAYESIAN DELINEATION OF POPULATION BOUNDARIES 114

7.3.2 BARRIER ESTIMATION WITH MONMONIER’S ALGORITHM 115

7.3.3 ESTIMATION OF MIGRATION AMONG POPULATIONS 115

7.4 DISCUSSION 116

7.4.1 COMPARATIVE LANDSCAPE GENETICS OF FRESHWATER TURTLES 116

7.4.2 LONG-LIVED ORGANISMS AND LANDSCAPE GENETICS 119

7.4.3 CONSERVATION IMPLICATIONS 121

7.5 REFERENCES 123

xi

CHAPTER 8 SUMMARY AND CONCLUSIONS 138

8 138

8.1 SUMMARY 138

8.2 SHORT-TERM STUDIES OF LONG-LIVED ORGANISMS – DIRECTIONS FOR FUTURE RESEARCH 140

8.3 REFERENCES 141

COPYRIGHT ACKNOWLEDGEMENTS 144

xii

List of Tables

Table 2.1. Characteristics of ten novel and two cross-amplified microsatellite loci for 127

Chelydra serpentina sampled from across southern Ontario. N=number of individuals

genotyped; k = number of alleles; Ho = observed heterozygosity; He = expected heterozygosity;

PI = probability of identity. Primer sequences shown include a 5’ M13 tail (5’-TGT AAA ACG

ACG GCC AGT-3’) on forward primers and a 5’ GTTTCTT pigtail on reverse primers.

MteD111 and MteD9 are from Hackler et al. (2006), with M13 tail (F) and pigtail (R) added.

Loci which are not in Hardy-Weinberg equilibrium (p < 0.01) are indicated with a *.

Table 3.1. Primer sequences and amplification conditions for 11 novel polymorphic loci for

Apalone spinifera. Temp = primer-specific annealing temperature (°C). Primer sequences shown

include a 5’ M13 tail (5’-TGT AAA ACG ACG GCC AGT-3’) on forward primers and a 5’

pigtail (GTTTCTT) on reverse primers.

Table 3.2. Characteristics of 11 novel polymorphic loci for 15 Apalone spinifera from southern

Ontario and 30 individuals of unknown origin. N=number of individuals genotyped; k = number

of alleles; Ho = observed heterozygosity; He = expected heterozygosity; PI = probability of

identity.

Table 4.1. Summary statistics for eleven microsatellite loci originally developed for the

Glyptemys muhlenbergii (King and Julian 2004) and amplified in 256 Clemmys guttata from

southern Ontario. Temp. = annealing temperature (°C) used in PCR amplification. * indicates an

initial touchdown of 1°C/cycle from 10°C above the annealing temperature, followed by a

constant annealing temperature for the remaining cycles; N = number of individuals amplified at

each locus; k = number of alleles; Ne = number of effective alleles; HO = observed

heterozygosity; HE = expected heterozygosity; PI = probability of identity, PIsibs = Probability

of identity for siblings at a locus.

Table 4.2. Number of alleles (private alleles in parentheses), observed and expected

heterozygosities (HO and HE), and estimated frequency of a null allele (for each locus across all

populations), for 256 Clemmys guttata from southern Ontario, by locus and populations (see

Figure 1 for definition of site acronyms).

xiii

Table 4.3. Genetic diversity (heterozygosity, allelic richness and private allelic richness) of

sampled regions, genetic populations and sites for 253 Clemmys guttata genotyped at 11

microsatellite loci. Allelic and private allelic richness are rarefacted to account for variation in

sample sizes (Kalinowski 2004). Pop1–5 = genetic clusters supported by both STRUCTURE and

TESS analyses. Georgian Bay is considered independently. HO = observed heterozygosity

averaged across all loci; HE = expected heterozygosity averaged across all loci. See Figure 1 and

text for definitions of site acronyms.

Table 4.4: Pairwise values FST (below the diagonal) and Dest (Jost 2008, above diagonal) for 13

putative subpopulations of Clemmys guttata sampled across southern ON (N = 253; see Figure 1

for definition of site acronyms). Sites in the Golden Horseshoe, Georgian Bay and the Bruce

Peninsula are analyzed together. FST values in italics are not significant (p > 0.05).

Table 4.5: Hierarchical analysis of molecular variance (AMOVA; Excoffier et al. 1992)

conducted in ARLEQUIN. Each source of variation was significant (p = 0.000). Tested populations

were those identified by both STRUCTURE and TESS analyses with GB treated as a separate, sixth

population. Subpopulations refer to sampling sites except GB, GH and BP which are treated as

single subpopulations.

Table 4.6. Assignment of individuals in GENECLASS analysis based on sampling sites; 66.3% of

individuals were assigned correctly. Shaded areas indicate clustering of sites in genetic

populations supported by both STRUCTURE and TESS. Sampling sites correspond to Figure 1.

Table 4.7. Summary of bottleneck tests in published studies of population genetics of tortoises

and freshwater turtles. Loci = the number of loci used in bottleneck tests (in some cases this was

lower than the total number amplified). Individuals = the maximum-minimum and mean ()

number of individuals genotyped per tested population. When only populations above a certain

size limit were used, only these populations were included in the summary. Where values for loci

and individuals are in bold this indicates that the minimum sampling recommendations for tests

in BOTTLENECK were met.

Table 5.1. Summary statistics for 11 microsatellite loci (Hackler et al. 2007; Davy et al. 2012)

amplified in 167 Chelydra serpentina from southern Ontario. N = number of individuals

successfully amplified at each locus; k = number of alleles; Ne = number of effective alleles; HO

xiv

= observed heterozygosity; HE = expected heterozygosity; PI = probability of identity; PIsibs =

Probability of identity for siblings at a locus. Locus MteD111 showed evidence of potential null

alleles and was excluded from all multi-locus analyses.

Table 5.2. Genetic diversity in 167 Chelydra serpentina sampled across southern Ontario based

on 10 microsatellite loci. Populations (Pop) and subpopulations (SP) were identified with

Bayesian clustering analyses (see text for details). GH = Golden Horseshoe, EO1 = Eastern

Ontario 1. Number of alleles (private alleles in parentheses); HO and HE: observed and expected

heterozygosities; N: sample size per tested unit. Estimated frequency of a null allele was

calculated for each locus across all populations. Summary statistics are presented for locus

MteD111, but this locus was excluded from calculations of mean heterozygosity and allelic

richness.

Table 5.3. Population differentiation (Dest above the diagonal, FST below) for four subpopulations

and two admixed groups of Chelydra serpentina identified by STRUCTURE analysis (Figure 3).

FST values in bold are significant (p < 0.05). Subpopulations (SP) are described in the text. GH =

Golden Horseshoe. EO = EO1 and EO3.

Table 5.4. Hierarchical partitioning of molecular variance with AMOVA (Excoffier et al. 1992).

All sources of variation with AMOVA (Excoffier et al. 1992). All sources of variation were

significant (p < 0.02).

Table 6.1. Genetic diversity at 12 microsatellite loci for 97 Emys blandingii from southern

Ontario. Temp. (optimal annealing temperature (°C) determined from temperature gradients of

initial PCR reactions) sample size (N), allelic richness (k), observed and expected heterozygosity

(HO, HE) and two measures of probability of identify (PI, PISibs) are shown for each locus. Total

values show mean ± standard error for N, k, Ne, HO and HE, and PI/PIsibs values with all loci

included.

Table 6.2. Number of alleles (number of private alleles in parentheses) and observed and

expected heterozygosities (HO and HE) for 97 Emys blandingii sampled across southern Ontario

and genotyped at 12 microsatellite loci. Loci Gmu– from King and Julian (2004). Loci Eb– from

Osentoski et al. (2002). Acronyms for sampling areas are defined in Figure 1. Estimated

frequency of a null allele is based on analysis of the entire data set following Brookfield (1996).

xv

No loci showed consistent evidence for null alleles when sampling areas were analyzed

independently. HO = observed heterozygosity; HE = expected heterozygosity; Ar = allelic

richness; PAr = private allelic richness.

Table 6.3. Genetic differentiation of Emys blandingii among sites in Ontario with N ≥ 15. All

FST values are significant (p < 0.05). All FST values were significant (p < 0.05). Average

historical number of migrants per generation (Nm) was calculated following Barton and Slatkin

(1986).

Table 6.4. GENECLASS results for Bayesian assignment tests. Values represent the proportion

of individuals from each sampled population assigned to each population. Values in bold indicate

the proportion of individuals from each sampled population assigned correctly to their source

population. Grey shaded areas indicate the two larger genetic clusters identified by TESS and

STRUCTURE.

Table 7.1. Life history, distribution and behavioral traits of Clemmys guttata, Chelydra

serpentina and Emys blandingii. Global conservation status is determined by the International

Union for Conservation of Nature (IUCN); Canadian conservation status is determined by the

Committee on the Status of Endangered Wildlife in Canada (COSEWIC).

Table 7.2. Mean, range, and maximum and minimum 95% confidence interval of effective

population size (Ne) estimated in ONeSAMP (Tallmon et al. 2008) for populations of three

freshwater turtles in southern Ontario, Canada. N = number of populations for which estimates

were obtained; mean Ne = mean estimated effective population size; s.d. = standard deviation;

range = minimum – maximum estimate. N is lower than the number of populations sampled

because Ne estimates for sites from which < 20 individuals were sampled were not included

(estimated Ne from sites with low sample sizes were all < 25).

Table 7.3. Average historic number of migrants per generation (Nm, Barton and Slatkin 1986).

xvi

List of Figures

Figure 4.1. Approximate location of sampled sites. LE = Lake Erie; LH = Lake Huron; BP =

Bruce Peninsula; GB = Georgian Bay; HC = Hastings County; DOR = dead on road; EO =

Eastern Ontario.

Figure 4.2. A) Population structure inferred by STRUCTURE and TESS for increasing values of K.

Colours indicating populations in the K = 5 model (marked with an asterisk) match colours used

in Figure 4. B) Estimated ln probability of the data (L(K)) for STRUCTURE analyses at increasing

values of K with 8 independent runs at each. C) ∆K (Evanno et al. 2005) calculated from (B). D)

TESS results: decreasing deviance information criterion (DIC) with increasing Kmax.

Figure 4.3. Principal Coordinates Analysis plot based on Dest (Table 4) for populations (a, b) and

based on genetic distance for individuals (c).

Figure 4.4. Genetic population structure identified by STRUCTURE and TESS with K = 5 (Figure

2). Georgian Bay was assigned to different populations by the two analyses. Hypothesized

dispersal routes for Clemmys guttata colonizing Canada after glacial retreat are indicated by the

large grey arrows.

Figure 5.1. Sampling sites for 167 Chelydra serpentina (blue squares) sampled in this study and

256 Clemmys guttata (yellow squares) sampled in Chapter 3. Bi-coloured squares indicate sites

where both species were sampled. Insert shows pairs of sampling areas used for comparisons of

genetic diversity between species. LE1 = Lake Erie 1; LE2 = Lake Erie 2; LH1 = Lake Huron 1;

LH 2 = Lake Huron 2; BP = Bruce Peninsula; GB = Georgian Bay; GH = Golden Horseshoe; N

= North of Golden Horseshoe; Kaw. = Kawartha Lakes area; Alg. = Algonquin Provincial Park;

HC = Hastings County; LO = north-east shore of Lake Ontario; EO = Eastern Ontario. Base map

modified from http://www.aquarius.geomar.de/omc/make_map.html and used under the GNU

Free Documentation license.

Figure 5.2. Results of Bayesian clustering analyses for increasing values of K, the number of

genetically distinct populations represented in the sample following analyses described in

Methods. Structure results for K = 1 – 8 : A) Log likelihood (L(K)) of the data (mean ± standard

xvii

deviation); B) ∆K following Evanno et al. (2005). Tess results for Kmax = 2–14; C) Deviance

information criterion (mean ± standard deviation) following analyses described in Methods.

Figure 5.3. Results of Bayesian clustering analyses for a range of models with increasing values

of K inferred using STRUCTURE and TESS. Models shown here are those that best fit the data based

on criteria described in Methods. Population structure in Cl. guttata across the same landscape is

shown for comparison (from Chapter 3). Colours used for subpopulations in the K = 4 model are

consistent with colours used in Figure 4.

Figure 5.4. Principle component analysis of genetic distance for 167 Ch. serpentina based on 10

microsatellite loci. A and B: PCoA of populations based on Dest ; C and D: PCoA based on

genetic distance among individuals labelled by sampling site.

Figure 5.5. Heterozygosity and effective population sizes of Ch. serpentina and Cl. guttata

compared across five pairs of sites (Figure 1, inset). HO: observed heterozygosity. HE: expected

heterozygosity. Ne: effective population size.

Figure 6.1. Approximate location of collection areas for Emys blandingii sampled across

southern Ontario. Top right inset indicates species range in North America (shown in red).

Sampling was focused on sites indicated with grey squares: LE = Lake Erie; GH = Golden

Horseshoe; PSD = Parry Sound District; KAW = Kawartha Lakes; EO = Eastern Ontario.

Sample sizes are included in each site marker. Grey triangles indicate extra samples included

opportunistically (each triangle represents an individual turtle): LHsouth = south shore of Lake

Huron; LHnorth = north shore of Lake Huron; ALG = Algonquin Provincial Park. Variation in

sample sizes results from differential sampling effort; differences in sample sizes are not

reflective of variation in actual population sizes. Base map modified from

http://www.aquarius.geomar.de/omc/make_map.html and used under the GNU Free

Documentation license; range map modified from COSEWIC (2005).

Figure 6.2. Principal Coordinates analysis of sampling areas (A, B) and individuals (C) for 91

Emys blandingii sampled from across southern Ontario based on 12 microsatellite loci.

Figure 6.3. Population structure inferred by Bayesian inference for 91 Emys blandingii collected

across southern Ontario. A) TESS results showing decreasing deviance information criterion

(DIC) with increasing values of Kmax. B) STRUCTURE results, mean estimated ln probability of

xviii

the data (L(K)) for increasing values of K, and ∆K, the second order rate of change of L(K)

following Evanno et al. (2005). Site abbreviations are explained in Figure 1.

Figure 7.1. Sampling sites and groupings of sites used for BARRIER analysis. Colors indicate the

sampled species at each site: yellow = Clemmys guttata, blue = Chelydra serpentina, red = Emys

blandingii. Small triangles indicate individual samples from otherwise unsampled sites. Squares

with dashed lines indicate the four areas used for comparison of Nm estimates. BARRIER analysis

considered genetically continuous samples with N > 12 as sampling units (inset, bottom right,

based on STRUCTURE results). ALG = Algonquin Provincial Park; BP = Bruce Peninsula; EO =

Eastern Ontario; GB = Georgian Bay; GH = Golden Horseshoe; HC = Hastings County; KAW =

Kawartha Lakes; LE = Lake Erie; LH = Lake Huron; LO = Lake Ontario; N = area north of GH

and south of GB; PSD = Parry Sound District. SP = subpopulation. Base map modified from


Documentation license.

Figure 7.2. Genetic population structure inferred in A) TESS, and B) STRUCTURE for Clemmys

guttata (yellow/brown), Chelydra serpentina (blue), and Emys blandingii (red). SP =

subpopulation. See Figure 1 for explanation of site abbreviations. Inferred clusters are plotted on

maps to the right of each set of results. Division of Cl. guttata samples under a K = 2 model

(implemented in STRUCTURE; see Chapter 3) is shown by a dashed black line on the bar plot and

the map for comparison.

Figure 7.3. Barriers to gene flow identified with Monmonier’s algorithm. Colored numbers

indicate sampling sites; thin green lines indicate boundaries between populations based on

Delaunay triangulation. A) Clemmys guttata (yellow; N = 253), B) Chelydra serpentina (blue; N

= 167) and C) Emys blandingii (red; N = 91). Estimates are based on 5,000 bootstrap replicates

of genetic distance matrices (Nei’s absolute distance). The thickness of each line and the

numbers in black text indicate the strength of bootstrap support. D) Barriers and sampling sites

for the three species overlaid on top of one another; barriers with bootstrap support > 0.90 are

marked with a dashed line.

Figure 7.4. Significant barriers (bootstrap support > 0.90) inferred using Monmonier’s algorithm

for Clemmys guttata (yellow dashed lines), Chelydra serpentina (blue dashed lines) and Emys

blandingii (red dashed lines).

xix

Figure 7.5. Average number of migrants per generation for Clemmys guttata, Chelydra

serpentina and E. blandingii estimated following Barton and Slatkin (1986) among four sites at

which all three species were sampled.

1

Chapter 1 Conservation genetics and freshwater turtles: a general

introduction

1

1.1 Outline

Conservation genetics is essentially the study of population genetics in small populations. The

field differs from traditional population genetics because its goal is to inform effective recovery

of threatened species, and the importance of genetics informing species recovery is well

documented. Much of the theory behind conservation genetics is based on studies of organisms

with relatively short life-spans and generation times, because these characteristics make them

good models for multi-generational studies. Long-lived organisms with overlapping generation

times may show different responses to population fragmentation than these studies predict. Thus,

effective conservation genetics studies of long-lived organisms must consider these potential

differences. Turtles provide an ideal model group to test the application of conservation genetics

theory to long-lived organisms. Because many species of turtles are threatened, turtles are also a

group for which conservation genetics studies are desperately needed so that effective

conservation actions can be implemented. In Ontario, seven of the eight resident species of turtle

are listed as “at-risk”, but no information about genetic structure in these populations is

available. In this thesis I develop genetic markers for two species of turtle and undertake

conservation genetics studies of three species. I use the results of these studies to test the

application of several widely accepted conservation genetics hypotheses to long-lived organisms.

1.2 Conservation genetics: objectives and challenges

The field of conservation genetics was explicitly established in the late 1900s, but it is strongly

rooted in classical population genetics. In the early 1900s, renewed interest in the work of

Mendel (1866) set the stage for advances in the study of quantitative genetics. Significant early

models in population genetics, such as “Hardy-Weinberg equilibrium” (Hardy, 1908), still

inform contemporary studies of the relationships among populations. The effects of genetic drift

on allele frequencies in a population were considered by Fisher (1922), who discussed the

possibility of fixation of alleles over time. This line of inquiry continued with investigations of

2

the potential impacts of inbreeding and increased genetic load (Wright 1922; Haldane 1926,

1937; Fisher, 1949). Wright (1931) coined the term “genetic drift” and demonstrated that drift

has a much stronger effect on small populations than on large populations. Wright (1931) also

introduced the concept of “effective population size” (Ne), the number of breeding individuals in

an idealized population that produces the observed loss of heterozygosity from one generation to

the next. The importance of natural selection in causing populations to diverge was investigated

(Haldane 1924; Fisher 1930), and the impacts of drift and selection on natural populations are

still being debated in the literature (Keller and Waller 2002, Sutton et al. 2011). Similarly, the

null hypothesis of isolation by distance (Wright 1943, Malécot 1948) still provides a useful

framework in which to test hypotheses about genetic structuring of populations.

The consideration of genetic changes in small populations in the context of conservation biology

began in the late 1900s. Gilpin and Soulé (1986) identified four “extinction vortexes”, biological

phenomena that can lead to extinction in small populations. Two of these involved loss of

genetic variation through genetic drift in small populations, through inbreeding depression and

increased genetic load, or through long-term fixation of alleles by genetic drift leading to reduced

adaptive potential.

Thus, the field of conservation genetics applies the study of population genetics to small and

declining populations (Frankham et al. 2002). The relative importance of evolutionary forces

such as genetic drift, mutation, and natural selection change as population sizes decline. Random

factors such as genetic drift and demographic stochasticity have a greater impact on genetic

diversity in small populations and can cause allele frequencies in such populations to change

dramatically over a few generations (Fisher 1922, Wright 1931). Genetic drift in small

populations may also cause the random loss of some alleles and the eventual fixation of others.

Additionally, random mating in small populations leads to increased inbreeding, which may

cause reduced fitness (inbreeding depression, as reviewed by Charlesworth and Willis 2009).

When genetic diversity has been significantly reduced in a population fragment, long-term

recovery of the population may be difficult even if short-term population growth can be easily

achieved (Ewing et al. 2008). For example, recovered populations of the pink pigeon (Columba

mayeri) have reduced fitness due to inbreeding (Swinnerton et al. 2004), and lethal recessive

alleles have led to a high incidence of chondrodystrophy (a form of dwarfism) in a captive

breeding program for the California condor (Gymnogyps californianus) (Ralls et al. 2000).

3

Thus, understanding the distribution of genetic diversity in small populations is essential to

effective population management and can facilitate genetic management of populations in

especially dire circumstances. Successful genetic management of threatened populations results

in a quantifiable increase in fitness and/or population growth. Examples include genetic rescue in

adders (Vipera berus), bighorn sheep (Ovis canadensis), Florida panthers (Puma concolor) and

lakeside daisies (Hymenoxys herbacea) (Demauro 1994; Madsen et al. 1994; 2004; Land and

Lacy 2000; Tallmon et al. 2004; Hogg et al. 2006; Miller et al. 2012). Conversely, attempts to

recover species without consideration of genetic management has led, for example, to extreme

inbreeding depression in reintroduced populations of koala (Phasolarctos cinereus; Houlden et

al. 1996; Sherwin et al. 2000) and outbreeding depression in reintroduced populations of ibex

(Capra ibex; Turcek 1951; Grieg 1979).

The major distinction between conservation genetics and other population genetics research is

the explicit objective of contributing to the preservation and recovery of threatened populations

and species (Frankham et al. 2002). Therefore, conservation genetics studies aim to generate

recommendations for maintaining genetic diversity in threatened species.

1.3 How small is small and what are we measuring?

Studies on the biology and genetics of “small” populations require definition of “small” and of

which part of a population is included in the measurement. The maintenance of genetic variation

in a population relies not on its census population size (the estimated total number of individuals

in the population based on methods such as capture-mark-recapture) but rather on its genetic

effective population size (Ne). Wright (1938, 1969) defined genetic Ne as the number of

individuals that would cause the observed loss of heterozygosity per generation (1/2N) in an

idealized population. Natural populations deviate from the assumptions of Wright’s idealized

model, for example, due to non-random mating, unequal sex ratios or high variance in

reproductive success among individuals. Thus, Ne in wild populations is usually lower than Nc

(the census population size), and Ne is the “population size” of interest in conservation genetics

studies (Frankham 1996; Jamieson and Allendorf 2012).

Estimates of Ne:Nc ratios in wild populations average 0.10–0.15, but estimates vary from <

0.001 to > 0.30 among species with differing life history strategies (Frankham 1996; Palstra and

Ruzannte 2008). In general, Ne:Nc appears to be higher in low-fecundity species than in high-

4

fecundity species, but the great amount of intraspecific variation in this ratio precludes an

accurate prediction (Palstra and Ruzzante 2008; Luikart et al. 2010). Estimates of Nc are usually

based on methods such as capture-mark-recapture, while Ne can be estimated from genetic data

(e.g. Tallmon et al. 2008; Waples and Do 2008).

So how small is “too small”? Franklin (1980) suggested that Ne > 50 is required to avoid the

deleterious effects of inbreeding in wild populations, and that Ne > 500 is required for

maintaining long-term genetic viability of a population. These estimates were derived based on

population genetics theory and Franklin did not suggest that they should apply to all species in

all circumstances. Nevertheless, a Google Scholar search for the so-called “50/500 rule”

identifies over 2,800 peer-reviewed papers citing, discussing, and criticizing this “rule of

thumb,” and the definition of “small” population size is the subject of ongoing debate. There is

little consensus on either the number of individuals required to maintain sufficient genetic

diversity for long-term population persistence, or even if such a number can be estimated with

current methods (e.g. Brook et al. 2011; Flather et al. 2011a, 2011b; Traill et al. 2007; 2010;

Jamieson and Allendorf 2012). However, there is some consensus that the minimum number of

individuals required for long-term genetic persistence of most species is probably several

thousand as this generally predicts Ne > 500 (Traill et al. 2010; Flathers et al. 2011).

Loss of genetic variation through genetic drift or inbreeding is not the only (nor necessarily the

largest) threat to small populations (Lande 1988). Demographic impacts on genetically healthy

populations can cause extirpation and extinction. Thus, mitigating the factors causing population

decline is as crucial to species recovery as maintenance of genetic diversity. Nevertheless,

genetic diversity is vital to the long-term persistence of a species and must be considered in

recovery plans for threatened species (Frankham et al. 2002; Jamieson and Allendorf 2012).

1.4 Turtles and conservation genetics of long-lived organisms

Turtles are one of the most threatened groups of vertebrates, with nearly 50% of species listed as

threatened by the International Union on the Conservation of Nature (IUCN;

http://www.iucnredlist.org/; Turtle Conservation Coalition 2011). However, data on genetic

population structure are unavailable for many species (Alacs et al. 2007). This thesis investigates

patterns of genetic diversity in three species of turtle in Ontario: the spotted turtle (Clemmys

guttata), the snapping turtle (Chelydra serpentina) and the Blanding’s turtle (Emys blandingii). I

5

also develop genetic markers for future studies of the spiny softshell (Apalone spinifera). All

four species are considered “at risk” by the Committee on the Status of Endangered Wildlife in

Canada (COSEWIC) and the Committee on the Status of Species at Risk in Ontario

(COSSARO). However, genetic population structure in these species has not been studied.

Therefore, the number of genetically distinct populations of each species in Ontario is unknown.

As well as providing species-specific information about threatened turtles in Ontario, my thesis

uses turtles as a model to test the genetic impacts of population fragmentation on long-lived

organisms. Most species of turtles have long generation times, for example > 40 years in E.

blandingii (COSEWIC 2005). Most species of turtle are also extremely long-lived; Cl. guttata

may live for 110 years (Litzgus 2006) and Ch. serpentina may live > 100 years (R. Brooks,

unpublished data, in COSEWIC 2008). Thus, the impacts of current anthropogenic landscape

modifications will have strong demographic effects on populations of turtles long before a

genetic signature of that effect becomes detectable.

Studies of mitochondrial DNA show extremely low intraspecific differentiation in several turtle

species in the north-eastern United States and Canada (Amato et al. 2008; McGaugh et al. 2008;

Phillips et al. 1996). Fortunately, microsatellite markers are more informative for population

genetics research on turtles (Tessier et al. 2005). Recently, several genetic studies of turtles in

North America have compared genetic diversity among “isolated” or “fragmented” populations

and “continuous” or “unfragmented” populations (e.g. Rubin et al. 2001; Kuo and Janzen 2004;

Richtsmeier et al. 2008; Pittman et al. 2011; Banning-Anthonysamy 2012). Other studies have

investigated the effects of predefined barriers to gene flow such as dams or urban development

(e.g. Bennett et al. 2010). The questions posed by these studies are central to conservation

biology. Unfortunately, the results are difficult to interpret without the right context and

information, which are unavailable for most species of turtle. Specifically, what is a “normal”

level of genetic variation in populations of turtles in North America, and how large, small,

fragmented or continuous are most populations of turtles? How does population structure vary

among species? Are most species panmictic across large distances or do they show evidence of

historical population structure pre-dating major anthropogenic landscape modification? How do

mating systems or patterns of paternity affect genetic diversity in turtles and other organisms

with overlapping generations?

6

What is clear from the literature is that populations of freshwater turtles may be connected across

relatively large distances (> 100km; Bennet et al. 2010; Banning-Anthonysamy et al. 2012), and

that even dramatic population declines may not be detectable genetically (Kuo and Janzen 2004;

Pittman et al. 2011). Therefore, I do not attempt to define populations a priori in this thesis but

rather collect samples across a wide geographic range and use a suite of analyses to determine

the number of genetic populations represented in the samples. This approach allows an unbiased

assessment of the number of genetic populations present in Ontario, which can be used to

improve population management plans.

Comparative studies provide an opportunity to test some of the conservation genetics hypotheses

discussed above. For example, can factors such as dispersal ability, rarity, fecundity, or

population size predict variation in genetic structure among species (Frankham 1996; Mitton

1997)? How does the dispersal ability of different turtle species affect the placement or number

of barriers to gene flow across the landscape? Do turtle species share common barriers to gene

flow across the landscape? Here, I use data from Cl. guttata, Ch. serpentina and E. blandingii to

investigate these questions. All three species share long generation times and life spans, but they

differ in their fecundity, vagility and abundance.

1.5 Conservation genetics of freshwater turtles in Ontario

In Chapters Two and Three I apply 454 “shotgun” sequencing to characterize and develop

primers for polymorphic microsatellite markers for the snapping turtle (Ch. serpentina) and the

spiny softshell (Apalone spinifera). These markers can be used for population genetics studies of

both species, and for investigations of relatedness and paternity.

In Chapter Four I investigate patterns of genetic diversity among populations within a species. I

use genetic and field data from populations of the spotted turtle (Cl. guttata) to test the

relationship between census population size and genetic diversity. Both theoretical and empirical

data show that these factors are correlated across a wide range of taxa, and I test whether or not

this correlation is also present in an endangered turtle. I use 11 microsatellite markers to

investigate genetic structure in Cl. guttata across southern Ontario. This is the first study of

genetic population structure in Cl. guttata, and I propose management units based on the results.

The ability of traditionally used “bottleneck tests” to detect recent population declines in Cl.

7

guttata is also tested. The use of these tests in other studies of turtles and tortoises is evaluated in

a literature review.

In Chapter Five I compare patterns of genetic diversity between two species, Cl. guttata and Ch.

serpentina. Variation in fecundity, rarity, vagility and abundance between these two species is

used to test several widely accepted hypotheses about the factors affecting genetic diversity. I

apply the microsatellite markers developed in Chapter Two to characterize population structure

in Ch. serpentina across southern Ontario, and compare the results to those obtained for Cl.

guttata in Chapter Four. I also use approximate Bayesian computation of effective population

size (Ne) to compare Ne between these species, and consider the effect of variation in Ne on

genetic diversity.

In Chapter Six I conduct a conservation genetics study of Blanding’s turtle (Emys blandingii)

across southern Ontario. Ontario is poorly represented in previous studies of the genetic structure

of E. blandingii populations (one south-eastern site at St. Lawrence Islands National Park is

sampled by Mockford et al. 2007). I characterize population structure in E. blandingii based on

data from 12 microsatellite loci and test the null hypothesis that the “Great Lakes-St. Lawrence”

population defined by COSEWIC is a single, panmictic population (COSEWIC 2005). I also

compare genetic diversity of Ontario E. blandingii to values reported by Mockford et al. (2007)

to further test the hypothesis that genetic variation is greater in the continuous main range of the

species than in isolated eastern populations.

In Chapter Seven I combine microsatellite data from the previous three chapters to compare

relative migration rates, population boundaries and barriers to gene flow among Cl. guttata, Ch.

serpentina and E. blandingii. I infer population boundaries using Bayesian clustering analyses

and infer barriers to gene flow using Monmonier’s algorithm. Based on these data, I test the

hypothesis that vagility can predict genetic population structure. I identify common areas of high

gene flow and common barriers to gene flow among species.

Finally, in the Conclusions I summarize the major findings of my dissertation and discuss

possible future directions for research.

8

1.6 References Alacs EA, Janzen FJ, Scribner KT (2007) Genetic issues in freshwater turtle and tortoise

conservation. Chel Res Monogr 4:107–123

Amato ML, Brooks RJ, Fu J (2008) A phylogeographic analysis of populations of the wood turtle (Glyptemys insculpta) throughout its range. Mol Ecol 17:570–581

Banning-Anthonysamy WJ (2012) Spatial ecology, habitat use, genetic diversity, and reproductive success: measures of connectivity of a sympatric freshwater turtle assemblage in a fragmented landscape. PhD dissertation, University of Illinois at Urbana-Champaign.

Bennett AM, Keevil M, Litzgus JD (2010) Spatial ecology and population genetics of northern map turtles (Graptemys geographica) in fragmented and continuous habitats in Canada. Chel Conserv Biol 9:185–195

Brook BW, Bradshaw JA, Trail LW, Frankham R (2011) Minimum viable population size: not magic, but necessary. Trends Ecol Evol 26:619–620

Charlesworth D, Willis JH (2009) The genetics of inbreeding depression. Nat Rev Genet 10:783-96

COSEWIC (2005) COSEWIC assessment and update status report on the Blanding's Turtle Emydoidea blandingii in Canada. Committee on the Status of Endangered Wildlife in Canada. Ottawa. viii + 40 pp. (www.sararegistry.gc.ca/status/status_e.cfm)

COSEWIC (2008) COSEWIC assessment and status report on the snapping turtle Chelydra serpentina in Canada. Committee on the Status of Endangered Wildlife in Canada. Ottawa. vii + 47 pp. (www.sararegistry.gc.ca/status/status_e.cfm)

Demauro MM (1994) Development and implementation of a recovery program for the federally threatened lakeside daisy (Hymenoxys acaulis var. glabra). In: Bowles ML, Whelan CJ (eds.) Restoring of endangered species: Conceptual Issues, Planning and Implementation. Cambridge University Press, Cambridge, UK, pp 298-321

Ewing SR, Nager RG, Nicoll MAC, Aumjaud A, Jones CG, Keller LF (2008) Inbreeding and loss of genetic variation in a reintroduced population of Mauritius kestrel. Conserv Biol 22:395–404

Fisher RA (1922) On the dominance ratio. P Roy Soc Edinb 42:321–341

Fisher RA (1930) The genetical theory of natural selection. Clarendon Press

Fisher RA (1949) The theory of inbreeding. Oliver and Boyd, Edinburgh.

Franklin IR (1980) Evolutionary change in small populations. In: Soulé ME, Wilcox BA (eds.) Conservation Biology: an Evolutionary–Ecological Perspective. Sinauer Associates pp 135–150

Frankham R (1996) Relationship of genetic variation to population size in wildlife. Conserv Biol 10:1500–1508

Frankham R, Ballou JD, Briscoe DA (2002) Introduction to Conservation Genetics. Cambridge University Press, Cambridge, U.K.

9

Frankham R (2010) Challenges and opportunities of genetic approaches to biological conservation Biol Conserv 143:1919–1927

Flather CH, Hayward GD, Beissinger SR, Stephens PA (2011) Minimum viable populations: is there a ‘magic number’ for conservation practitioners? Trends Ecol Evol 26:307–316

Flather CH, Hayward GD, Beissinger SR, Stephens PA (2011) A general target for MVPs: unsupported and unnecessary. Trends Ecol Evol 26:620–622

Gilpin ME, Soulé ME (1986) Minimum viable populations: The processes of species extinctions. In Soulé M (ed.) Conservation biology: The science of scarcity and diversity, pp. 13–34. Sinauer Associates, Sunderland, Massechussets

Grieg JC (1979) Principles of genetic conservation in relation to wildlife management in Southern Africa. S Afr J Wildl Res 9:57-78

Haldane JBS (1924) The mathematical theory of natural and artificial selection. Part I. Trans Cambridge Philos Soc 23:19–41

Haldane JBS (1926) A mathematical theory of natural and artificial selection. Math Proc Cambridge 23: 363–372

Haldane JBS (1937) The effect of variation on fitness. Am Nat 71: 337–349

Hardy GH (1908) Mendelian proportions in a mixed population. Science 28: 49–50

Hogg JT, Forbes SH, Steele BM, Luikart G (2006) Genetic rescue of an insular population of large mammals. Proc R Soc Lond B Biol Sci 273:1491–1499

Houlden BA, England PR, Taylor AC, Greville WD, Sherwin WB (1996) Low genetic variability of the koala Phasolarctos cinereus in southeastern Australia. Mol Ecol 5:269–281

Jamieson IG, Allendorf FW (2012) How does the 50/500 rule apply to MVPs? Trends Ecol Evol 27:578–584

Land DE, Lacy RC (2000) Introgression levels achieved through Florida panther genetic restoration. Endangered Species Updates 17:99–103

Lande R (1988) Genetics and Demography in Biological Conservation. Science 241:1455-1460

Litzgus JD (2006) Sex differences in longevity in the spotted turtle (Clemmys guttata). Copeia 2006:281–288

Luikart G, Ryman N, Tallmon DA, Schwartz MK, Allendorf FW (2010) Estimation of census and effective population sizes: the increasing usefulness of DNA-based approaches. Conserv Genet 11:355–373

Keller LF, Waller DM (2002) Inbreeding effects in wild populations. Trends Ecol Evol 17: 230–241

Kuo CH, Janzen FJ (2004) Genetic effects of a persistent bottleneck on a natural population of ornate box turtles (Terrapene ornata). Conserv Genet 5:425–437

Madsen T, Shine R, Olsson M, Wittzell H (1999) Restoration of an inbred Adder population. Nature 402:34–35

Madsen T, Ujvari B, Olsson M (2004) Novel genes continue to enhance population growth of an inbred population of Adders (Vipera berus). Biol Conserv 120:145–147

10

Malécot G (1948) Les Mathematiques de l’Hérédité. Masson et Cie, Paris.

McGaugh SE, Eckerman CM, Janzen FJ (2008) Molecular phylogeography of Apalone spinifera (Reptilia, Trionychidae). Zool Scr 37:289–304

Mendel G (1866) Versuche über Pflanzen-Hybriden. Reprinted in 1951: J Hered 42: 3–47

Miller JM, Poissant J, Hogg JT, Coltman DW (2012) Genomic consequences of genetic rescue in an insular population of bighorn sheep (Ovis canadensis). Mol Ecol 21:1583–1596

Mitton JB (1997) Selection in natural populations. Oxford University Press: Oxford.

Mockford SW, Herman TB, Snyder M, Wright JM (2007) Conservation genetics of Blanding’s turtle and its application in the identification of evolutionarily significant units. Conserv Genet 8:209–219

Palstra FD, Ruzzante PE (2008) Genetic estimates of contemporary effective population size: what can they tell us about the importance of genetic stochasticity for wild population persistence? Mol Ecol 17:3428–3447

Phillips CA, Dimmick WW, Carr JL (1996) Conservation genetics of the common snapping turtle (Chelydra serpentina). Conserv Biol 10:397–405

Pittman SE, King T, Faurby S, Dorcas ME (2011) Genetic and demographic status of an isolated bog turtle (Glyptemys muhlenbergii) population: implications for the conservation of small populations of long-lived animals. Conserv Genet 12:1589–1601

Ralls K, Ballou JD, Rideout BA, Frankham R (2000) Genetic management of chondrodystrophy in California condors. Anim Conserv 3:145–153

Richtsmeier RJ, Bernstein NP, Demastes JW, Black RW (2008) Migration, gene flow, and genetic diversity within and among Iowa populations of ornate box turtles (Terrapene ornata ornata). Chel Conserv Biol 7:3–11

Rubin CS, Warner RE, Bouzat JL, Paige KN (2001) Population genetic structure of Blanding’s turtles (Emydoidea blandingii) in an urban landscape. Biol Conserv 99:323–330

Sherwin WB, Timms P, Wilcken J, Houldne B (2000) Analysis and conservation implications of koala genetics. Conserv Biol 14:639–649

Sutton JT, Nakagawa S, Robertson BC, Jamieson IG (2011) Disentangling the roles of natural selection and genetic drift in shaping variation at MHC immunity genes. Mol Ecol 20: 4408–4420

Swinnerton K, Groombridge JJ, Jones CG, Burn RW, Mungroo Y (2004) Inbreeding depression and founder diversity among captive and free-living populations of the endangered pink pigeon Columba mayeri. Anim Conserv 7:353–364

Tallmon DA, Luikart G, Waples RS (2004) The alluring simplicity and complex reality of genetic rescue. Trends Ecol Evol 19:489–496

Tallmon DA, Koyuk A, Luikart GH, Beaumont MA (2008) ONeSAMP: a program to estimate effective population size using approximate Bayesian computation. Mol Ecol Resour 8: 299–301

Tessier N, Paquette S, Lapointe FJ (2005) Conservation genetics of the wood turtle (Glyptemys insculpta) in Quebec, Canada. Can J Zool 83:765–772

11

Traill LW, Bradshaw CJA, Brook BW (2007) Minimum viable population size: a meta-analysis of 30 years of published estimates. Biol Conserv 139:159–166

Traill LW, Brook BW, Frankham R, Bradshaw CJA (2010) Pragmatic population viability targets in a rapidly changing world. Biol Conserv 143:28–34

Turcek FJ (1951) Effect of introductions on two game populations in Czechoslovakia. J Wildl Manage 15:113–114

Turtle Conservation Coalition. (2011). Turtles in Trouble: The World’s 25+ Most Endangered Tortoises and Freshwater Turtles—2011. Rhodin AGJ, Walde AD, Horne BD, van Dijk PP, Blanck T, Hudson R (eds.): IUCN/SSC Tortoise and Freshwater Turtle Specialist Group, Turtle Conservation Fund, Turtle Survival Alliance, Turtle Conservancy, Chelonian Research Foundation, Conservation International, Wildlife Conservation Society, and San Diego Zoo Global, 54 pp. Lunenburg, MA

Waples RS, Do C (2008) ldne: a program for estimating effective population size from data on linkage disequilibrium. Mol Ecol Resour 8:753–756

Wright S (1922) Coefficients of inbreeding and relationship. Am Nat 56: 330–338

Wright S (1931) Evolution in Mendelian populations. Genetics 16: 97–159

Wright S (1943) Isolation by distance. Genetics 28: 114–138

Wright S (1969) Evolution and the Genetics of Populations. Vol. 2. The Theory of Gene Frequencies, University of Chicago Press.

12

Chapter 2 Characterization of ten novel microsatellite loci and cross-

amplification of two loci in the snapping turtle (Chelydra serpentina)1

Christina M. Davy1,2*, Ashley E. Leifso3, Ida M. Conflitti1,2 and Robert W. Murphy1,2 1 Department of Ecology and Evolutionary Biology, 25 Willcocks St., University of

Toronto, Toronto, ON, M5S 3B2, Canada 2Department of Natural History, Royal Ontario Museum, 100 Queen’s Park, Toronto,

Ontario, M5S 2C6, Canada. 3 Wildlife Preservation Canada, RR #5, 5420 Highway 6 North, Guelph, Ontario, N1H

6J2, Canada

2 Abstract

We used 454 (‘‘shotgun’’) sequencing to obtain a partial genomic library for the snapping turtle

(Chelydra serpentina). We characterized ten microsatellite loci from these sequences and tested

cross-amplification of loci originally developed for the alligator snapping turtle (Macrochelys

temminckii). We genotyped 127 individuals from Ontario at twelve loci. The number of alleles

per locus ranged from 1 to 14; heterozygosity ranged from 0.157 to 0.850. These loci will be

used to study population genetic structure in this long-lived reptile and may cross-amplify in two

closely related species.

Keywords: microsatellite; Chelydra serpentina; Macrochelys temminckii

1 This chapter is published in Conservation Genetics Resources 4:695–698 (DOI: 10.1007/s12686-012-9624-7). The

original publication is available at www.springerlink.com. The co-authors grant permission to include this chapter and its appendix in the thesis, and authorize the use of the thesis by the National Library.

13

2.1 Primer Note

The snapping turtle (Chelydra serpentina) is a long-lived species found from southern Canada

east of the Rocky Mountains southwards to the Gulf Coast and Florida (Ernst and Lovich 2009).

The International Union for the Conservation of Nature (IUCN) lists the species as Least

Concern. However, its life history makes it vulnerable to over-harvesting (van Dijk 2011).

Reliable microsatellite markers for this species may provide valuable information for potential

conservation initiatives. We use 454 sequencing to develop novel microsatellite markers, and

cross-amplify nine microsatellite markers previously developed for the alligator snapping turtle

(Macrochelys temminckii; Hackler et al. 2006).

For microsatellite development, we isolated genomic DNA from snapping turtle blood using

phenol–chloroform procedure (Sambrook et al. 1989) and cleaned the DNA using EtOH

precipitation. We obtained a partial genomic library by sequencing on a Roche GS Junior

(Roche, Branford, CT) at Trent University’s Natural Resources DNA Profiling and Forensics

Centre. The run produced 127,778 sequences averaging 423 base pairs, which we searched for

tri- and tetra-nucleotide microsatellites using the program MSATCOMMANDER (Faircloth

2008). We examined potential target loci by eye to identify loci with appropriate flanking

regions for primer design and designed 40 primer pairs using Primer 3 (Rozen and Skaletsky

2000). We labelled forward primers with a fluorescent 5’ M13 tail and labelled reverse primers

with a 5’ pigtail (GTTTCTT; Brownstein et al. 1996) to facilitate adenylation.

We collected blood samples from 127 Ch. serpentina from across southern Ontario by caudal

venipuncture and stored the blood on FTA cards (Whatman, Inc.). We extracted genomic DNA

from each card following the method suggested by Smith and Burgoyne (2004) for samples with

nucleated erythrocytes. We ran a temperature gradient with two DNA samples at each locus and

used the optimal temperature for all subsequent PCR reactions. Amplification followed the

methods of Schuelke (2000), using 4.0 lL of M13-labelled forward primer, 0.66 lL each of

pigtailed reverse primer (Eurofins MWG Operon) and a 6-carboxyfluorescein dye (6-FAM;

Eurofins MWG Operon) and 1.0 lL of DNA eluate (6–9 ng). We also tested cross-amplification

of nine microsatellite loci developed for the alligator snapping turtle (Macrochelys temminckii;

Hackler et al. 2006) using a 12.5 lL PCR reaction containing 0.6 lL of forward primer and 1.0 lL

each of reverse primer and 6-FAM. Two of the nine alligator snapping turtle loci amplified

14

successfully in the snapping turtle. PCR cycling parameters followed King and Julian (2004)

with annealing temperatures adjusted for each locus as summarized in Table 1. We visualized

length of the amplified fragments using a 3730 DNA Analyzer (Applied Biosystems) with

GS(500) Liz (Applied Biosystems) as a size standard and scored genotypes using

GENEMARKER (SoftGenetics, State College, PA). One to two homozygous samples were

subsequently sequenced at each locus to confirm identity of the amplified fragments, and five

percent of the sampled individuals were genotyped twice at each locus to assess genotyping

error.

We successfully amplified ten novel loci and cross-amplified two of the nine alligator snapping

turtle loci (MteD9 and MteD111). We genotyped 127 Ch. serpentina from Ontario to

characterize these 12 loci. We found no ambiguities in the genotypes of the individuals amplified

and genotyped multiple times. Sequencing of homozygotes confirmed the identity and motifs of

the amplified fragments.

We used GENALEX v6.0 (Peakall and Smouse 2006) to quantify the number of alleles per locus

(k), calculate observed and expected heterozygosity (HO and HE) and probability of identity (PI)

for each locus. We used GENEPOP 4.0.10 (Raymond and Rousset 1995; Rousset 2008) to test

for linkage disequilibrium and deviations from Hardy–Weinberg equilibrium (HWE).

The number of alleles per locus ranged from 1 to 14. Heterozygosity ranged from 0.157 to 0.850.

Table 2.1 summarizes the primer sequences, amplification conditions and characteristics of the

12 characterized loci. None of the 66 pairwise comparisons between loci showed evidence of

linkage disequilibrium after Bonferroni correction for multiple comparisons. Two loci (Cs18 and

MteD111) showed significant deviations from HWE (p < 0.01). Three alleles at the tri-nucleotide

locus Cs16 differed by only one base pair. Sequencing of homozygous individuals and successful

replication of these genotypes in independent amplifications both demonstrated that these are

unique alleles and are not the result of stutter. One locus (Cs14) was monomorphic in the tested

samples. Because they are all from the northern limits of this species’ range low genetic diversity

is expected, but it may be variable in southern populations.

The family Chelydridae contains two other species (Phillips et al. 1996): the Central American

Snapping Turtle (Chelydra rossignoni), listed by the IUCN as Vulnerable, and the South

15

American snapping turtle (C. acutirostris), which remains to be assessed. Cross-amplification of

these markers could facilitate conservation genetic analyses of these two closely related and

poorly understood species.

2.2 Acknowledgments

This project was supported by a Canada Collection grant from Wildlife Preservation Canada

(WPC) to CD and a National Science and Engineering Research Council (NSERC) Discovery

Grant (A3148) to Robert W. Murphy. Genotyping costs were offset by the generous assistance of

the Schad Foundation. Ida Conflitti assisted with sequencing to confirm homology of M.

temminckii and Ch. serpentina sequences. CD and IC were funded by Canada Graduate

Scholarships from NSERC. Ashley Leifso assisted with microsatellite genotyping and was

funded by WPC through a Science Horizons grant from Environment Canada. We thank Dr. C.

Kyle, E. Kerr and M. Harnden at the NRDPFC (Trent University) for conducting 454

sequencing. Sample collection was conducted with the permission of the Government of Ontario

and Ontario Parks and followed Animal Use Protocol 2010-14 (Royal Ontario Museum).

2.3 References Brownstein MJ, Carpten JD, Smith JR (1996) Modulation of nontemplated nucleotide addition

by taq DNA polymerase: primer modifications that facilitate genotyping. BioTech 20:1004–1010.

Ernst C, Lovich J (2009) Turtles of the United States and Canada 2nd ed. Johns Hopkins University Press, Baltimore.

Faircloth BC (2008) msatcommander: detection of microsatellite repeat arrays and automated, locus-specific primer design. Mol Ecol Resour 8:92–94

Hackler JC, van den Bussche RAV, Leslie DM (2006) Characterization of microsatellite DNA markers for the alligator snapping turtle, Macrochelys temminckii. Mol Ecol Notes 7:474–476

King TL, Julian SE (2004) Conservation of microsatellite DNA flanking sequences across 13 Emydid genera assayed with novel bog turtle (Glyptemys muhlenbergii) loci. Conserv Genet 5:719–725

Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Mol Ecol Notes 6:288–295


16

Raymond M, Rousset F (1995) GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. J Hered 86:248–249

Rousset F (2008) Genepop'007: a complete reimplementation of the Genepop software for Windows and Linux. Mol Ecol Resour 8:103–106

Rozen S, Skaletsky HJ (2000) Primer3 on the WWW for general users and for biologist programmers. In: Krawetz S, Misener S (eds) Bioinformatics Methods and Protocols: Methods in Molecular Biology. Humana Press, Totowa, NJ, pp 365-386. Source code available at http://fokker.wi.mit.edu/primer3/

Sambrook J, Fritsch EF, Maniatis T (1989) Molecular cloning—a laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, New York, NY

Schuelke M (2000) An economic method for the fluorescent labeling of PCR fragments. Nat Biotechnol 18:233–234

Smith LM, Burgoyne LA (2004) Collecting, archiving and processing DNA from wildlife samples using FTA® databasing paper. BMC Ecol 4:4: http://www.biomedcentral.com/1472-6785/4/4

van Dijk PP (2011) Chelydra serpentina. In: IUCN 2011. IUCN Red List of Threatened Species. Version 2011.2. <www.iucnredlist.org>. Downloaded on 12 December 2011.

17

Table 2.1. Characteristics of ten novel and two cross-amplified microsatellite loci for 127 Chelydra serpentina sampled from across southern Ontario. N=number of individuals genotyped; k = number of alleles; Ho = observed heterozygosity; He = expected heterozygosity; PI = probability of identity. Primer sequences shown include a 5’ M13 tail (5’-TGT AAA ACG ACG GCC AGT-3’) on forward primers and a 5’ GTTTCTT pigtail on reverse primers. MteD111 and MteD9 are from Hackler et al. (2006), with M13 tail (F) and pigtail (R) added. The “*” indicates loci that are not in Hardy-Weinberg equilibrium (p < 0.01).

Locus Primer Sequence (5’ – 3’)*

Repeat

motif

Annealing

temperature

(°C) Size (bp) N k Ho He PI

Cs08 F: TGTAAAACGACGGCCAGTGCTGGACATGTACGTGCAAA AGAT 54 157–198 127 11 0.850 0.859 0.0356

R: GTTTCTTTGTATATGTCCTTTAGGCATTTATGTG

Cs12 F: TGTAAAACGACGGCCAGTGGCATTTCTAGCCAACAGGA AGAT 58 197 – 250 126 12 0.841 0.808 0.0536

R: GTTTCTTAGCGGTGTGCTTTTCTCAGT

Cs14 F: TGTAAAACGACGGCCAGTCAGACAGGGGCTGTTGAGTC AGG 60 231 57 1 - - 1.000

R: GTTTCTTGGCAGCTCTGTGTGTCAGTC

Cs16 F: TGTAAAACGACGGCCAGTTCCAAATGCAGCTCTCTTCA AAT 58 193 – 198 127 4 0.701 0.678 0.1645

R: GTTTCTTCCTTGCCATCTCGAACAAAT

Cs17 F: TGTAAAACGACGGCCAGTTGGAAACTCTCCTTGTCTGTCC ACC 60 290 – 305 127 6 0.654 0.691 0.1399

R: GTTTCTTGAGGCACTTTATTAATTCCTACCTCT

Cs18* F: TGTAAAACGACGGCCAGTTGGTGGTTTCTCTGGAAGTTTT AATT 56 268 – 276 127 3 0.157 0.159 0.7169

R: GTTTCTTTTCTGCTTTTAACCTGCACTCA

Cs19 F: TGTAAAACGACGGCCAGTTTGAGTGGTCTACGGGAACC AGG 58 194 – 206 127 5 0.472 0.469 0.3158

R: GTTTCTTTGACGGTTTCCTAGCGGTAT

Cs22 F: TGTAAAACGACGGCCAGTCGGCAGAAGATAAGAGGCATT AAAT 56 321 – 333 127 4 0.378 0.357 0.4339

R: GTTTCTTTGGTAGGGTTGCTCATGAAA

Cs24 F: TGTAAAACGACGGCCAGTTGTTTCCATTCCAAACACCTG ACC 61 417 – 432 127 3 0.551 0.575 0.2540

R: GTTTCTTGCAACACTGCTTCCCTTCAT

Cs25 F: TGTAAAACGACGGCCAGTTGTGTTGTCACAGGGCACTT ATC 60 220 – 229 127 3 0.567 0.560 0.2921

R: GTTTCTTAAATGGACTGCGGACACTTC

MteD9 F: TGTAAAACGACGGCCAGTCCAGATGCTAGTCTCACACC TAGA 60 261 – 285 127 6 0.740 0.743 0.1058

R: GTTTCTTGCTTACTGGAATTAACCTCATG

MteD111* F: TGTAAAACGACGGCCAGTTCCACAAACTCCCATCTTC TAGA 60 175 – 191 120 14 0.558 0.865 0.0324

R: GTTTCTTCCACACGGAAAAATCTATCTAC

18

Chapter 3 Isolation and characterization of eleven novel polymorphic

microsatellite loci in the spiny softshell turtle (Apalone spinifera)2

Christina M. Davy1,2*, Ida M. Conflitti1,2, Daniel M.L. Storisteanu3 and Robert W. Murphy1,2

1 Department of Ecology and Evolutionary Biology, University of Toronto, 25 Willcocks St., Toronto, ON, M5S 3B2, Canada

2 Department of Natural History, Royal Ontario Museum, 100 Queen’s Park, Toronto, Ontario, M5S 2C6, Canada.

3 Department of Medicine, Addenbrooke’s Hospital, University of Cambridge, Hills Road, Cambridge, CB2 2QQ, United Kingdom

3 Abstract

We isolated and characterized 11 microsatellite loci for the spiny softshell turtle (Apalone

spinifera) from a partial genomic library obtained using 454 sequencing technology. Genotypes

of 15 individuals from southern Ontario and 30 individuals of unknown origin contained 6 to 20

alleles per locus and the level of heterozygosity ranged from 0.229 to 0.800. These markers

would be useful for population genetics studies and enforcement activities such as assignment of

illegally traded individuals to their population of origin.

Keywords: microsatellite; next-generation sequencing; turtle

2 This chapter is published in Conservation Genetics Resources 4:759–761 (DOI: 10.1007/s12686-012-9638-1).

The original publication is available at www.springerlink.com. The co-authors grant permission to include this chapter and its appendix in the thesis, and authorize the use of the thesis by the National Library.

19

3.1 Primer Note

The spiny softshell turtle (Apalone spinifera) is a widely distributed species ranging from

southern-eastern Canada to north-eastern Mexico (Ernst and Lovich 2009). The species is listed

as Least Concern by the International Union for the Conservation of Nature (IUCN). However,

particular populations are considered to be at risk (van Dijk 2011) and the Canadian populations

of A. spinifera are listed as Threatened (COSEWIC 2002). Despite lack of data on harvest levels

in many parts of the species’ range, recorded exports of A. spinifera from the North America to

Asian food markets increased dramatically in recent years and doubled between 2006 and 2008

(IUCN/SSC Tortoise and Freshwater Turtle Specialist Group 2010). Illegal harvest poses a

significant threat to at-risk populations of A. spinifera. Here, we present a suite of variable

microsatellite markers for A. spinifera that could be used to answer a range of research questions

as well as for conservation enforcement (for example, development of assignment tests).

We used phenol-chloroform extraction (Sambrook et al. 1989) to isolate genomic DNA from a

whole blood sample of A. spinifera stored in 95% EtOH and cleaned the DNA using standard

EtOH precipitation. A partial genomic library was obtained by sequencing on a Roche GS Junior

(Roche, Branford, CT) using the next generation sequencing facilities at Trent University's

Natural Resources DNA Profiling and Forensics Centre.

The GS Junior run produced 137,054 sequences averaging 415 base pairs in length. We searched

sequences for tri-, tetra- and penta-nucleotide microsatellites with the program

MSATCOMMANDER (Faircloth 2008) and designed 40 primer pairs using the software Primer

3 (Rozen and Skaletsky 2000). We added a 5’ M13 tail to forward primers to facilitate

fluorescent labelling, and a 5’ pigtail (GTTTCTT; Brownstein et al. 1996) to reverse primers to

facilitate adenylation.

PCR amplification followed the method of Schuelke (2000) and cycling parameters followed

King and Julian (2004) with annealing temperatures adjusted for each locus (Table 3.1). We used

a 3730 DNA Analyzer (Applied Biosystems) to visualize length of the amplified fragments by

comparison to GS(500) Liz size standard (Applied Biosystems). We scored genotypes using

GENEMARKER (SoftGenetics, State College, PA).

20

Eleven of the 40 primer pairs amplified unambiguous, replicable alleles. We used GENALEX v6.0

(Peakall and Smouse 2006) to quantify the number of alleles per locus (k), calculate observed

and expected heterozygosity (Ho and He) and probability of identity (PI) for each locus. We used

GENEPOP 4.0.10 (Raymond and Rousset 1995; Rousset 2008) to test for linkage disequilibrium

and deviations from Hardy-Weinberg equilibrium (HWE).

We obtained blood from 15 individual A. spinifera from southern Ontario by caudal

venipuncture following approved Animal Use Protocols and stored it on FTA cards (Whatman

Inc.). We prepared DNA for PCR following the protocols of Smith and Burgoyne (2004) for

processing FTA cards containing blood with nucleated erythrocytes. Because the sampled

Ontario population is near the northern limit of the species’ range we expected genetic variation

to be relatively low. Thus, we also collected 30 tissue (muscle) samples from A. spinifera

carcasses confiscated by Government of Ontario wildlife enforcement staff. We isolated DNA

from the muscle samples using phenol-chloroform extraction (Sambrook et al. 1989). The exact

origin of these individuals was unknown, but we assumed that they represented a wider

geographic distribution than the samples from Ontario and included them to better investigate

polymorphism in these markers. We genotyped a total of 45 individuals at 11 loci. We also

sequenced alleles from homozygous loci to confirm that the amplified fragments were

homologous to those obtained through 454 sequencing.

Table 3.2 summarizes the characteristics of each locus for the samples from Ontario and those of

unknown origin. The overall number of alleles per locus ranged from 6 to 20 and the level of

heterozygosity ranged from 0.229 to 0.800. No linkage disequilibrium was detected when the

population from Ontario is analyzed separately. However, when considering the entire dataset, 5

of the 55 pairwise comparisons between loci showed evidence of linkage disequilibrium after

Bonferroni correction for multiple comparisons (As18 and As07; As18 and As15; As15 and

AsB12; AsB09 and As12; and As15 and As B09).

All loci were in HWE in the samples from Ontario (p < 0.05) with the exception of AsB14 (p =

0.048). Only one locus (AsB08) meets the expectations of HWE in the 30 samples of unknown

origin (p < 0.05). However, these samples probably do not represent a single population and

should not be treated as such.

21

These 11 polymorphic loci will facilitate studies of the population genetics of A. spinifera. Due

to the unconventional use of individuals for whom the exact location of origin is not known,

these results are not intended to provide a robust genetic profile of any particular population.

Rather, the results demonstrate the utility of these variable loci for population genetics studies in

this species, including population assignment tests and potential conservation enforcement.

3.2 Acknowledgements

This project was made possible by a Canada Collection grant from Wildlife Preservation Canada

to C.D. and a National Science and Engineering Research Council (NSERC) Discovery Grant

A3148 to Robert W. Murphy. Generous assistance from the Schad Foundation offset genotyping

costs. Ida Conflitti performed sequencing to confirm sequences motifs; both CD and IC were

funded by NSERC Canada Graduate Scholarships. Daniel Storisteanu assisted with genotyping

and was funded by an NSERC Undergraduate Summer Research Award. We thank Dr. Chris

Kyle, Emily Kerr and Matthew Harnden at the NRDPFC (Trent University) for conducting 454

sequencing. Sample collection was conducted with the permission of the Government of Ontario

following Animal Use Protocol 2010-14 (Royal Ontario Museum, Toronto, Canada). We thank

Rick Andrews and the Lake Ontario Enforcement Unit for access to tissues from confiscated

turtles.

3.3 References Bacher J, Hennes LF, Gu T, Tereba A, Micka KA, Sprecher CJ, Lins AM, Amiott EA, Rabbach

DR, Taylor JA, Helms C, Donis-Keller H, Schumm JW (1999) Pentanucleotide repeats: highly polymorphic genetic markers displaying minimal stutter artifact. In: Proceedings from the ninth international symposium on human identification, Orlando, pp 24–37

Brown DJ, Faralloa VR, Dixon JR, Baccusa JT, Simpson TR, Forstner MRJ (2011) Freshwater Turtle Conservation in Texas: Harvest Effects and Efficacy of the Current Management Regime. J Wildl Manag 75:486–494

Brownstein MJ, Carpten JD, Smith JR (1996) Modulation of nontemplated nucleotide addition by taq DNA polymerase: primer modifications that facilitate genotyping. BioTechnol 20:1004–1010.

COSEWIC 2002. COSEWIC assessment and update status report on the spiny softshell turtle Apalone spinifera in Canada. Committee on the Status of Endangered Wildlife in Canada. Ottawa. vii + 17 pp.

Ernst C, Lovich J (2009) Turtles of the United States and Canada 2nd ed. Johns Hopkins University Press, Baltimore.

22

Faircloth BC (2008) msatcommander: detection of microsatellite repeat arrays and automated, locus-specific primer design. Mol Ecol Resour 8:92–94

IUCN/SSC Tortoise & Freshwater Turtle Specialist Group. 2010. A study of progress on conservation of and trade in CITES-listed tortoises and freshwater turtles in Asia. In: CoP15, Inf. 22. Convention on international trade in endangered species of wild fauna and flora. Fifteenth meeting of the conference of the parties, Doha (Qatar). 13–25 March. <http://www.cites.org/common/cop/15/inf/E15i-22.pdf>.

King TL, Julian SE (2004) Conservation of microsatellite DNA flanking sequences across 13 Emydid genera assayed with novel bog turtle (Glyptemys muhlenbergii) loci. Conserv Genet 5:719–725

Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Mol Ecol Notes 6:288–295

Raymond M, Rousset F (1995) GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. J Hered 86:248–249

Rousset F (2008) Genepop'007: a complete reimplementation of the Genepop software for Windows and Linux. Mol Ecol Resour 8:103–106

Rozen S, Skaletsky HJ (2000) Primer3 on the WWW for general users and for biologist programmers. In: Krawetz S, Misener S (eds) Bioinformatics Methods and Protocols: Methods in Molecular Biology. Humana Press, Totowa, NJ, pp 365-386. Source code available at http://fokker.wi.mit.edu/primer3/.


Schuelke M (2000) An economic method for the fluorescent labeling of PCR fragments. Nat Biotechnol 18:233–234


van Dijk, P.P. 2011. Apalone spinifera. In: IUCN 2011. IUCN Red List of Threatened Species. Version 2011.2. <www.iucnredlist.org>. Downloaded on 19 December 2011

23

Table 3.1. Primer sequences and amplification conditions for 11 novel polymorphic loci for

Apalone spinifera. Motif = repeat motif of microsatellite. Temp = primer-specific annealing

temperature (°C). Primer sequences shown include a 5’ M13 tail (5’-TGT AAA ACG ACG GCC

AGT-3’) on forward primers and a 5’ pigtail (GTTTCTT) on reverse primers.

Locus Primer Sequence (5’ – 3’) Motif Temp

As07 F: TGTAAAACGACGGCCAGTACGACGCCAAAATTTGAGTT AGAT 52

R: GTTTCTTACTTTTGTTCCTCCGGGTTT

As12 F: TGTAAAACGACGGCCAGTTGATCATTGTCTCTTGGCAGTC ATGGT 52

R: GTTTCTTGTGATTGCAGCAGCGAAATA

As13 F: TGTAAAACGACGGCCAGTCCCACTGGGATTGCTAACTT CTTT 60

R: GTTTCTTTGGATGAAGAAATTGCATGG

As14 F: TGTAAAACGACGGCCAGTGTGGCTGAAAAGGCAAGACT GATT 58

R: GTTTCTTTGCAAAATGGACCTTGAACA

As15 F: TGTAAAACGACGGCCAGTTGGCCTTAGGCAAGTCTTTT GTTT 54

R: GTTTCTTGAGCCTACATCTGCAATGGTT

As18 F: TGTAAAACGACGGCCAGTTTTAATTCCTGAGAGGGACACTG GTTT 58

R: GTTTCTTGCAGTAAAGGGCAAAACCAG

AsB07 F: TGTAAAACGACGGCCAGTTTCAGTAAGAAAGTTGTAAATCTTGAA AAC 61

R: GTTTCTTATATGGCCCTTGACCTCACA

AsB08 F: TGTAAAACGACGGCCAGTGCCGCATCAGCTTTGTTAAG AAC 58

R: GTTTCTTTTCCCTGTGCTTACCTGGTC

AsB09 F: TGTAAAACGACGGCCAGTCTGCTTCACCCCTTCTCTGA ATC 58

R: GTTTCTTAGGCATCGGATACAAACAGG

AsB12 F: TGTAAAACGACGGCCAGTTGCCAGAATCTTCAAAAGCA AAT 58

R: GTTTCTTCTCCTGTGAGCCAGGTCAGT

AsB14 F:TGTAAAACGACGGCCAGTTGTTGCAAACACAGTTGGAA AAT 58

R: GTTTCTTTGCCAGAAAGAAATCACCAA

24

Table 3.2. Characteristics of 11 novel polymorphic loci for 15 Apalone spinifera from southern

Ontario and 30 individuals of unknown origin. N=number of individuals genotyped; k = number

of alleles; Ho = observed heterozygosity; He = expected heterozygosity; PI = probability of

identity.

southern Ontario (N=15)

confiscated individuals (unknown origin; N=30)

Locus Size (bp) N k Ho He PI

N k Ho He PI

As07 191–307 14 11 0.714 0.860 0.034 27 14 0.556 0.767 0.069

As12 243–309 15 5 0.600 0.638 0.178 30 7 0.700 0.733 0.113

As13 172–236 15 5 0.600 0.702 0.146 30 17 0.900 0.925 0.011

As14 216–224 15 1 0.000 0.000 1.000 26 6 0.308 0.567 0.229

As15 262–342 15 9 0.867 0.820 0.056 27 16 0.630 0.833 0.045

As18 252–305 13 5 0.385 0.337 0.453 27 12 0.630 0.871 0.029

AsB07 179–205 15 2 0.333 0.358 0.476 30 9 0.367 0.527 0.250

AsB08 209–230 15 3 0.333 0.380 0.423 28 8 0.821 0.844 0.044

AsB09 142–180 15 3 0.333 0.640 0.206 30 10 0.500 0.814 0.055

AsB12 256–283 15 5 0.667 0.709 0.136 27 5 0.259 0.628 0.197

AsB14 237–252 13 6 0.385 0.589 0.195 30 6 0.667 0.821 0.057

25

Chapter 4 Conservation genetics of the endangered spotted turtle do not

support a relationship between genetic variation and population size.

Formatted for Biological Conservation.

4 Abstract

The hypothesis that genetic variation is affected by population size is widely accepted in

conservation genetics. Here, I test this hypothesis in a particularly long-lived vertebrate, and test

the efficacy of “bottleneck” tests in long-lived species. The endangered spotted turtle (Clemmys

guttata) is restricted to small, disjunct and declining populations, and can be used to model the

application of conservation genetics theory to long-lived organisms with overlapping

generations. I genotyped 256 individuals at 11 microsatellite loci and used a suite of

conservation genetics analyses to investigate population structure across the Canadian range of

Cl. guttata. Within-site allelic richness ranged from 3.18 to 4.49; observed heterozygosity ranged

from 0.510 to 0.743. Although allelic richness was correlated with population size,

heterozygosity and private allelic richness were not. Bottleneck tests failed to detect population

declines in 12 of 13 tested sites. A literature review discovered that bottleneck tests in 17 of 18

studies of tortoises and freshwater turtles had insufficient sampling, potentially resulting in Type

I and II errors. Bayesian analyses identified a minimum of five genetic populations and a

maximum of 10 genetically differentiated subpopulations which are demographically

independent. Genetic population structure of Cl. guttata appeared to reflect patterns of post-

glacial colonization rather than current landscape modifications. These results can improve

management and recovery plans for the endangered spotted turtle and demonstrate that long-

26

lived organisms such as turtles may not show the generally accepted relationship between

genetic diversity and population size.

Keywords: Clemmys guttata, landscape genetics, microsatellites, Ontario, STRUCTURE, TESS

4.1 Introduction

Genetic drift can significantly impact small or fragmented populations (Ewing et al., 2008) by

driving the stochastic loss of genetic diversity and increasing inbreeding. These impacts can

reduce fitness and evolutionary potential. Conservation genetics aims to mitigate these impacts

and maintain adaptability to environmental changes in threatened species by preserving genetic

diversity (Frankham, 1996; Frankham et al., 2002). The emerging fields of spatial and landscape

genetics emphasize that genetic population structure is shaped by historic and current landscape

structure (Manel et al., 2003; Guillot et al., 2009). Spatial analyses can also be integrated into

conservation genetics studies to explicitly consider the effects of current and historic landscapes

on the status of a species and its potential for recovery.

Small populations of organisms with long-life spans, overlapping generations and promiscuous

mating systems violate some of the assumptions of conservation genetics theory (Frankham et

al., 2002). Such populations may experience the genetic effects of population fragmentation

more slowly than typical model organisms such as Drosophila melanogaster or Caenorhabditis

elegans. For example, most species of turtle have delayed maturity, long generation times (> 25

years), long life-spans and polygamous or promiscuous mating systems (Congdon et al., 1993;

1994; Litzgus, 2006; Davy et al., 2011). Therefore, turtles are an effective model system to study

the effects of population fragmentation on long-lived organisms. Given that almost 50% of turtle

species are threatened and require protection (www.iucnredlist.org), population genetics studies

of turtles are also a conservation priority (Alacs et al., 2007).

Freshwater turtles often show no detectable genetic population structure on small spatial scales

(< 80 km; Bennet et al., 2010; Banning-Anthonysamy, 2012), but analyses of population

structure on a larger spatial scale can reveal the historical drivers of patterns across the landscape

(e.g. Tessier et al., 2005; Pearse et al., 2006; Stepien et al., 2009; Echelle et al., 2010). Taken in

27

the context of the current, modified landscape, such analyses can identify areas that should be

prioritized for mitigation measures and facilitate recovery plans to maximize the preservation of

genetic diversity. For example, these analyses can identify situations where translocations or the

development of wildlife corridors can re-connect fragmented, historically continuous

populations. They can also identify cases where increased connectivity across the landscape

could be detrimental to species recovery, for example, by leading to outbreeding depression or

genetic homogenization among strongly differentiated populations.

The genetic effects of recent, anthropogenic habitat and population fragmentation are usually not

detectable in turtles (Rubin et al., 2001; Kuo and Janzen, 2004; Marsack and Swanson, 2009;

Pittman et al., 2011), even when dispersal is restricted (Bennett et al., 2010). Similarly, small

populations of turtles that have declined significantly in recent years often show no evidence of

recent genetic bottlenecks (Kuo and Janzen, 2004; Mockford et al., 2005; Marsack and Swanson,

2009; Spradling et al., 2010; Pittman et al., 2011). One possible explanation is that the long life

span of turtles buffers small populations against the genetic effects of habitat fragmentation and

bottlenecks (Kuo and Janzen, 2004; Marsack and Swanson, 2009, Bennet et al., 2010). However,

this result may also be affected by sampling bias.

The commonly used program BOTTLENECK requires a minimum of 10 loci and 30 individuals per

tested population to achieve reasonable statistical power in tests of heterozygote excess and in

the qualitative mode-shift test (Piry et al., 1999). Unfortunately, the collection of large sample

sizes of threatened turtles from multiple populations is logistically challenging, and the

development of appropriate molecular markers is costly. Tests may therefore be performed with

suboptimal sample sizes (see below). Furthermore, analyses of simulated data suggest that these

tests may not accurately detect bottlenecks even when the test requirements are met (Peery et al.,

2012). Thus, bottleneck tests may lead to incorrect conclusions about the loss of genetic diversity

in threatened populations of turtles and other long-lived organisms. Native, severely reduced

populations can serve to explore this problem.

The spotted turtle, Clemmys guttata (Schneider, 1792), is globally endangered due to habitat loss

and illegal collection (van Dijk, 2011). Across its range, this species occurs in small, isolated

populations (van Dijk, 2011; COSEWIC, 2004) precluding a genetic rescue effect (Tallmon et

28

al., 2004). Low vagility (dispersal ability) may result in low levels of gene flow between isolated

populations because this species has high site fidelity and individuals rarely travel > 2 km per

year (Litzgus, 1996; Seburn, 2003; Ernst and Lovich, 2009). Most Canadian populations now

contain < 150 individuals (COSEWIC, 2004). These small, isolated, and declining populations

provide an excellent test case for commonly used genetic bottleneck tests.

In this study I evaluate genetic diversity and population structure among Canadian populations of

Cl. guttata. I use Cl. guttata to test the following conservation genetics hypotheses:

1) That native, declining populations of a long-lived organism with overlapping generations

will show the genetic impacts of fragmentation and population decline predicted by conservation

genetics theory (Frankham, 1996), namely, reduced genetic diversity in smaller populations

relative to larger populations, and significantly reduced genetic diversity in populations relative

to the metapopulation;

2) That traditional bottleneck tests can be used effectively in studies of threatened

populations of long-lived organisms. This is tested using multiple populations of Cl. guttata

along with a literature review of other species of turtles.

I evaluate the data in the framework of recovery of this endangered species, and I test the

efficacy of assignment tests for identifying the origin of Canadian Cl. guttata.

4.2 Methods

4.2.1 Sample collection and genotyping

I conducted capture-mark-recapture surveys for Cl. guttata at 13 sites across southern Ontario

from April 2008 to October 2011. Sampled sites represented most of the known, extant

populations of Cl. guttata in Canada (COSEWIC, 2004). Several sites were surveyed in

collaboration with researchers working on existing long-term projects. Approximate locations of

sampling sites are shown in Figure 4.1. Detailed location information is not provided to avoid an

increase in illegal collection.

Turtles were captured by hand, except for one opportunistic capture in a hoop trap. Each turtle

was sexed, measured, photographed, and marked by shell notching (Cagle, 1939). I collected

29

0.05–0.10 mL of blood from mature individuals by caudal venipuncture and stored samples on

FTA cards (Whatman, Inc., Clifton, NJ, USA). Muscle was sampled from freshly dead turtles

encountered during surveys or dead on the road (DOR). Bone samples were taken from older

carcasses or empty shells. Several additional samples were contributed by other researchers.

I sampled 25–30 individuals/population where possible and > 10% of the estimated population at

all other sampling sites. Published estimates of population size were used to estimate the

proportion of the population sampled. Where published estimates were unavailable, I estimated

population size (N) from mark-recapture data using the program MARK (White and Burnham,

1999), under a closed-capture model (data not shown). Capture probability was allowed to vary

with time to account for differences in survey effort and field conditions among survey years.

Genomic DNA was extracted from FTA cards and muscle following Davy et al. (2012). To

extract DNA from old turtle shells, I removed a small piece of bone, ground it into a fine

powder, and processed it using the QIAamp ® DNA Investigator Kit (Qiagen Inc., Valencia,

CA).

PCR was conducted for 11 microsatellite loci originally developed for the Bog Turtle (Glyptemys

muhlenbergii; King and Julian, 2004). Amplification followed the methods of Schuelke (2000);

using 4.0 µL of M13-labelled forward primer, 0.66 µL each of pigtailed reverse primer (Eurofins

MWG Operon) and a 6-carboxyfluorescein dye (6-FAM; Eurofins MWG Operon), and 1.0 µL of

DNA eluate (6-9 ng). PCR cycling parameters followed King and Julian (2004) with annealing

temperatures optimized for each locus (Table 4.1). Fragment lengths were visualized using a

3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA) with size standard GS(500)

Liz (Applied Biosystems). I scored genotypes with an RFU (relative fluorescence units) peak >

200 using GENEMARKER (SoftGenetics, State College, PA). Amplification was repeated for

genotypes with a weak signal (< 200 RFU).

Genotyping error was assessed by re-extracting and re-genotyping approximately three percent

of the samples (8/256) at each locus. I used duplicate, independent samples from the same

individual where possible (Pompanon et al., 2005). In three cases, duplicate extractions were

taken from a single FTA card.

30

4.2.2 Population genetics analyses

Genotypes were checked for evidence of null alleles and long-allele drop-out using MICRO-

CHECKER v.2.2.3 (van Oosterhout et al., 2004). I used GENALEX v.6.0 (Peakall and Smouse,

2006) to quantify observed and expected heterozygosity (HO and HE) and probability of identity

(PI) for each locus, for each site and globally. Because all sampled populations were small, I also

calculated PISibs, the probability of identity taking into account the possibility that close

relatives were sampled (Taberlet and Luikart, 1999; Waits et al., 2001). Linkage equilibrium and

deviations from Hardy-Weinberg equilibrium (HWE) were tested in GENEPOP v.4.0.10 (Raymond

and Rousset, 1995; Rousset, 2008). A sequential Bonferroni correction was applied to account

for multiple pairwise comparisons (Rice, 1989). Each site was also tested for heterozygote deficit

(indicating inbreeding) or heterozygote excess (which could indicate inbreeding avoidance or a

recent bottleneck). Allelic richness (Ar) and private alleleic richness (PAr) were adjusted for

unequal sample sizes by rarefaction in HP-RARE v.1.0 (Kalinowski, 2004; 2005).

Effective population size (Ne) for each sampled site was estimated by approximate Bayesian

computation in ONeSAMP (Tallmon et al., 2008) using prior lower and upper bounds of 4 and

200 on each estimate of Ne. I used Pearson’s correlation coefficient to test for significant

correlations between N, Ne and genetic diversity (HO, Ar and PAr). Genetic diversity (HO, HE,

Ar and PAr) were compared among sites and genetic populations using Friedman’s two-way

analysis of variance by ranks. Correlations and Friedman`s test were conducted in SPSS v. 20.0

(SPSS Inc. Chicago, Illinois).

I calculated absolute differentiation among sites, Dest (Jost, 2008), using SMOGD v.1.2.5

(Crawford, 2010). I also quantified differentiation using FST and assessed significance with

10,000 randomizations in FSTAT (Goudet, 1995). Both Dest and FST could have been biased by

small sample sizes. Therefore sites BP1 and BP2, GH1 and GH2, and GB1 and GB2, which were

each < 30 km apart, were combined for these calculations. Correlations between genetic distance

(Dest) and Euclidean distance (Wright, 1943) were tested using IBDWS v.3.23 (Jensen et al.,

2005). Significance of matrix correlations was assessed with a Mantel test (Mantel, 1967) with

30,000 randomizations. Geographic genetic structure was also visualized using principal

coordinates analysis (PCoA) in GENALEX to ordinate genetic distance (Dest) among sampled sites.

31

No a priori assumptions were made about the amount of structure present because this is the first

study of genetic structure in Cl. guttata. Instead, I used three programs that employ Bayesian

inference to detect population structure in genetic data. First, I tested the relative probability of a

series of models ranging from 1 to13 populations (K) using STRUCTURE v.2.3.4 (Pritchard et al.,

2000), which used Bayesian inference to assign individuals to distinct clusters based on their

genotypes by minimizing disequilibrium in each cluster. Each run involved 750,000 generations

with a burn-in of 75,000 generations. The model assumed correlated allele frequencies (Falush et

al., 2003) and historical admixture between populations (Pritchard et al., 2000). Eight runs were

conducted at each value of K using the LOCPRIOR function to include sampling information

(collection site of each individual) in the analysis.

We compiled the output of the 104 runs with STRUCTURE HARVESTER v.0.6.92 (Earl and

vonHoldt, 2012) and used two methods to estimate K, the most probable number of genetically

distinct populations represented in the data. The increase in pr(X|K), the probability of the data

given a particular value of K, typically plateaus at the most likely value of K (Pritchard et al.,

2000). The ad hoc ∆K method (Evanno et al., 2005) implemented in STRUCTURE HARVESTER was

also used to calculate the second-order rate of change in log likelihood between successive

values of K, which typically peaks at the appropriate value of K. We used the Greedy and

LargeKGreedy algorithms in CLUMPP v.1.1.2 (Jakobsson and Rosenberg, 2007) to permute and

combine results from independent runs. Genetic clusters identified by STRUCTURE were

visualized with DISTRUCT v.1.1 (Rosenberg, 2004).

The second and third programs used to explore population structure were TESS v.2.3.1 (Chen et

al., 2007) and GENELAND v.4.0.2 (Guillot et al., 2005), which explicitly considered spatial

information. The TESS analysis assumed an admixture model (Durand et al., 2009) and

considered increasing values of Kmax (the maximum number of populations in the dataset) from 2

to 9, with 10 runs at each Kmax. Each run included 50,000 sweeps with a burn-in of 10,000

sweeps and run data were assessed to ensure convergence. The most likely value of K was

determined based on the value at which the decreasing deviance information criterion (DIC)

values reached a point of inflection and the number of distinct clusters stabilized (Chen et al.,

2007). In GENELAND, I assumed correlated allele frequencies between populations. I conducted

32

10 runs of 1,000,000 iterations each, exploring a range of K values from 1 to 15. A burn-in of

10% was applied post-processing.

Analysis of molecular variance (AMOVA) was performed in ARLEQUIN v.3.1 (Excoffier et al.,

2005). Partitioning of genetic variance was quantified within and among sampled sites and

genetic populations identified with clustering analyses.

4.2.3 Development of genetic assignment tests for Canadian Cl. guttata

Assignment tests in GENECLASS v.2.0 (Piry et al., 2004) used the method of Rannala and

Mountain (1997) with 100,000 simulated individuals. Assignment tests first considered each

sampled site as an independent population, and were repeated using the genetic populations

identified by STRUCTURE and TESS. I also used the posterior probability of assignment of each

individual to the clusters identified by STRUCTURE (i.e. the individual q-matrix) to determine

whether individuals could be accurately assigned to their site or populations of origin based on

genetic data. Individuals were assigned to a cluster (site or population) when q > 0.5 (Latch et

al., 2006).

4.2.4 Analyses of genetic bottlenecks

Genetic bottlenecks were inferred using BOTTLENECK, which tested for bottlenecks in the past

2Ne–4Ne generations (Piry et al., 1999). Five sites had sample sizes > 29, meeting the criteria

recommended for this program (Table 4.2), and the remaining sites were also tested. I tested for

heterozygote excess (Cornuet and Luikart, 1997) using 1,000 replicates under the two-phase

model (TPM; Di Rienzo et al., 1994), with a variance of 12 among multiple steps. Parameters

were set to 95% single-step mutations and 5% multiple-step mutations, and the Wilcoxon test

was used to assess statistical significance of heterozygote excess as recommended for < 20 loci

(Piry et al., 1999). The qualitative mode shift test (Luikart et al., 1997) was performed for all 13

sampled sites. Both tests were also applied to the populations identified by STRUCTURE and TESS.

4.3 Results

I sampled 254 Cl. guttata from 13 sites (mean N = 19.5, s.d. = 10.6, range = 4–35; Table 4.2)

representing approximately 10% of the total estimated Canadian population and most of the

known Canadian range. I also sampled three DOR individuals that were not found near known

33

populations. Pairwise distances between sampled sites ranged from 3.2 to 670.0 km. (mean =

277.2; s.d. = 147.5).

Eleven polymorphic loci amplified successfully (Table 4.1). Duplicated analyses of genotypes

were identical in all cases, indicating negligible genotyping error. Genotypes with weak signal

strength (> 200 RFU) received identical scores when the amplification was repeated and the

signal increased. One sample from EO2 produced genotypes with three peaks. The triple peaks

were replicated at all loci with four independent re-extractions and amplifications and the

individual was subsequently excluded from the study.

MICRO-CHECKER identified homozygote excess across the entire dataset but did not detect

homozygote excess in any individual sites. Thus, homozygote excess in the entire dataset was

likely a result of deviation from HWE due to population genetic structure rather than null alleles.

Exact tests detected no deviations from HWE in the overall dataset. Four loci showed deviations

from HWE in at least one population, but these were not significant after Bonferroni correction.

GENEPOP detected heterozygote excess in populations EO1 and EO2 (p < 0.05). Heterozygote

deficit detected in populations LH1, GH1 and GH2 was not significant after Bonferroni

correction. Two pairs of loci (GmuD107 and GmuD21, and GmuD16 and GmuD79) showed

evidence of linkage disequilibrium in the whole dataset, but this was not consistent among

populations. Values of PI and PIsibs reached < 0.001 with inclusion of 4 and 8 loci, respectively.

Number of alleles per locus ranged from 3 to 18, and HO ranged from 0.484 at GmuB08 to 0.891

at GmuD87 (mean HO = 0.689; Table 4.1). Within sites, HO ranged from 0.614 to 0.743 (mean

HO = 0.672; Table 4.2). Sites LE1, GB, HC and EO2 had private alleles, and several alleles were

restricted to only two or three sites. Allelic richness (Ar) ranged from 3.18 to 4.49; private allelic

richness (PAr) ranged from 0.1 to 0.28 (Table 4.3).

In STRUCTURE analyses ∆K showed an initial maximum at K = 2, with two increasingly smaller

peaks at K = 5 and K = 8 indicating possible hierarchical population structure (Figure 4.2). At K

= 2, southeastern and southwestern Ontario split (Figure 4.2). At K = 5, the following clusters

were resolved: LH1 and LH2 (mean q = 0.80, s.d. = 0.14); BP1 and BP2 (mean q = 0.91, s.d. =

0.03); LE1, LE2, GH1, GH2, GB1, and GB2 (mean q = 0.77, s.d. = 0.10); HC (mean q = 0.97,

34

s.d. = 0.01); and EO1 and EO2 (mean q = 0.86, s.d. = 0.18). At K = 8, sites EO1 and GB (GB1

and GB2) became distinct, but no biologically relevant eighth cluster was apparent.

The average change in DIC between Kmax values tested in TESS was 150.01. Runs reached a

point of inflection at five clusters (mean DIC = 16460.830, s.d. = 129.039) and plateaued at Kmax

= 6 (Figure 4.2). Although the DIC continued to decline past K = 5 no new clusters emerged in

the individual q-matrices at K = 6 (Figure 4.2d,e). The TESS clusters corresponded to those

identified by STRUCTURE for K = 5, except that Georgian Bay clustered with the Bruce Peninsula

rather than with Lake Erie and the Golden Horseshoe. Georgian Bay was considered as a

separate, sixth population for further population-level analyses.

GENELAND analysis gave the highest probability to K = 15 (log likelihood = -3611.943). Five

“ghost populations” were inferred and these were disregarded (Guillot et al., 2005). The analysis

identified 10 clusters corresponding to sampled sites, with pairs of sites separated by < 30 km

assigned to single clusters (BP1 and 2, GB1 and 2, GH1 and 2, and HC and the three nearby

DOR samples). Thus, Bayesian analyses identified five genetic populations with Georgian Bay

potentially forming a sixth. Ten subpopulations corresponding to the GENELAND clusters nested

within these populations. Placement of the DOR samples varied with the different methods

(Figure 4.2) and they were not included in further population-level analyses.

Estimated census population size (mean = 110.8, median = 58, s.d. = 119, range = 12 – 423) was

positively correlated with Ar (r = 0.563, p = 0.004), but not with either PAr (r = 0.435, p = 0.138)

or HO (r = 0.358, p = 0.229). Estimated Ne (mean = 26.18, median = 33.48, s.d. = 13.00, range =

6.44 – 45.21) was also positively correlated with Ar (r = 0.705, p = 0.010) but not with PAr (r =

0.191, p = 0.552) or HO (r = 0.325, p = 0.302). Friedman’s test comparing heterozygosity and

allelic richness among the K= 2 and K = 5 models and the subpopulations (Table 4.3) found that

HE, Ar and PAr increased significantly with each level of structure (HE: χ2 = 21.294, d.f. = 2 p =

0.000, Ar: χ2 = 19.633, d.f. = 2, p = 0.000, PAr: χ2 = 22.240, d.f. = 2, p = 0.000). However, HO

values of sites were not significantly different from larger populations (χ2 = 0.824, d.f. = 2, p =

0.662).

Values of FIS were not significant after correction for multiple comparisons. All pairwise FST

values were significant after correction for multiple comparisons, with the exception of LE2 and

35

GH. Pairwise Dest values (Table 4.4) were up to two times larger than FST values and ranged

from 0.227 (LH1 vs. EO2) to near zero (0.014; LE2 vs. GH). Mantel tests detected a significant

overall correlation between genetic and Euclidian distance (Z = 2687.4073, r = 0.3856, p =

0.018).

AMOVA of the K = 5 model with Georgian Bay considered separately showed that variation

within populations accounted for 91.79 % of the variation in the data (Table 4.5, AMOVA: ΦST =

0.082, p < 0.0001). Significant variation also occurred among subpopulations (ΦSC = 0.046, p <

0.0001) and between the populations (ΦCT = 0.038, p < 0.0001). PCoA showed each

subpopulation occupying distinct coordinate space, with the first three axes accounting for 73.4%

of total variation (Figure 4.3).

Assignment tests in GENECLASS had a 66.3% success rate (167 individuals correctly assigned)

when assigning individuals to their sampling site (Table 4.6). When the genetic populations

identified by STRUCTURE and TESS were considered assignment accuracy increased to 77% and

78.6%, respectively. STRUCTURE correctly assigned all individuals to their cluster of geographic

origin (q > 0.5) based on the K = 5 model.

BOTTLENECK did not detect heterozygote excess in any sites or populations. Only one site

showed evidence of a mode shift; this site had both a sample and census population size of four.

In a literature review of 18 studies of tortoises and freshwater turtles that used bottleneck tests

(Table 4.7), 14 studies did not use the recommended number of loci and 15 studies applied the

test to samples of > 30 individuals. Only one study met both requirements (Kuo and Janzen,

2004).

4.4 Discussion

Clemmys guttata shows significant genetic structure across its Canadian range that cannot be

explained by either geographic proximity of sites or isolation by distance alone. Heterozygosity

does not appear to vary with population size or effective population size in this species. Two

commonly used bottleneck tests failed to detect recent population declines; these tests may have

limited use in studies of threatened turtles and other long-lived organisms.

36

4.4.1 Biogeography and conservation genetics of Clemmys guttata

Patterns of genetic structure in Cl. guttata are reflected in other taxa across this landscape.

Barriers to gene flow among the Great Lakes occur for the walleye, Sander vitreus (Stepien et

al., 2009). Limited historic gene flow between central and southeastern Ontario exists for small-

mouth bass, Micropterus dolomieu (Borden and Krebs, 2009), S. vitreus (Stepien et al., 2009),

channel darters, Percina copelandi (Kidd et al. 2011), and snapping turtles, Chelydra serpentina

(Chapter 4). The presence of a distinct population of Cl. guttata (site HC) in the Moira River

watershed also occurs in P. copelandi (Kidd et al., 2011), suggesting that this area has a

colonization history distinct from that of nearby watersheds. These broad-scale patterns

apparently reflect historic processes and barriers to gene flow rather than influences of the

current landscape.

The division between southwestern and southeastern Ontario (Figure 4.2a) suggests two distinct

routes for Cl. guttata colonizing Canada after the retreat of the Laurentide ice sheets. Two likely

colonization routes are shown in Figure 4.4, and I hypothesize two independent Pleistocene

glacial refugia. The southern Appalachian Mountains contain the first potential refugium, where

a Pleistocene refugium has also been inferred for wood turtles (Glyptemys insculpta; Amato et

al., 2008). This may have been the source of the current populations in southeastern Ontario,

which could have entered Ontario either from the east, through present-day Quebec, or from the

south by crossing the St. Lawrence River. To the southwest of our study area, a rich mid-

Holocene fossil record from Indiana, Ohio and Michigan includes Cl. guttata and several other

species of turtles, and suggests rapid post-glacial colonization of the Great Lakes region from

nearby refugia (Holman, 1992). Thus, I infer a second refugium near present-day southern

Indiana, to the south of the last glacial maximum. As Cl. guttata dispersed northward into

Ontario, habitat succession and shifting water levels may have isolated peripheral populations

that gradually diverged via genetic drift. As humans colonized Ontario, the relatively slow

effects of genetic drift would have been compounded by anthropogenic habitat modification and

further population declines due to a combination of habitat destruction and hunting pressures.

Today, anthropogenic impacts maintain isolation of populations and subpopulations.

Genetic differentiation of populations on the Bruce Peninsula is also documented in black bears

(Ursus americanus; Pelletier et al., 2011) and Massasauga rattlesnakes (Sistrurus catenatus;

37

Gibbs et al., 1997). Repeated fires burning through the peninsula in the late 1800s may have

caused bottlenecks that could potentially explain the differentiation of Bruce Peninsula

populations. Long-term isolation may also explain the differentiation. Populations of S. catenatus

on the Bruce Peninsula have apparently been isolated since before the arrival of Europeans to

North America (Gibbs et al., 1997) and this scenario also seems plausible for Cl. guttata.

Genetic structure of Cl. guttata in southwestern Ontario differs strikingly from that of the eastern

foxsnake (Pantherophis gloydi), a marshland-prairie specialist that shows significant genetic

structure along the north shore of Lake Erie (Row et al., 2010). Pantherophis gloydi shares the

marshland habitat preferences of Cl. guttata but can also exploit a variety of other open habitats.

Populations of P. gloydi on the north shore of Lake Erie have a broader distribution than Cl.

guttata. Nevertheless, P. gloydi has distinct genetic units along a shoreline where Cl. guttata

does not. Row et al. (2010) suggested that the pattern observed in P. gloydi is the result of

reduced dispersal due to habitat conversion for agriculture and development. However, the

distribution of populations of Cl. guttata in this area has been reduced more dramatically than the

distribution of populations of P. gloydi. The difference in genetic population structure between

the two species may be driven by differing generation times: approximately 5 years for P. gloydi

versus > 25 years for Cl. guttata, which may reach > 100 years of age (COSEWIC, 2004;

Litzgus, 2006; COSEWIC, 2008). This difference could allow populations of P. gloydi to

express greater effects of genetic isolation than longer-lived Cl. guttata.

Current migration among subpopulations of Cl. guttata is highly unlikely and the sites sampled

here are demographically independent (COSEWIC 2004). Reduced gene flow among sites has

resulted in genetic differentiation of subpopulations (Figure 4.2, Table 4.4). Occasional human-

facilitated translocations occur (S. Gillingwater, pers. comm.; F. Ross, pers. comm.) but

clustering analyses (Figure 4.2) suggest that haphazardly translocated individuals have not

impacted the genetic profile of any sampled subpopulations.

4.4.2 Management implications

Management units (MUs) are populations whose allele frequencies have diverged significantly

(Moritz, 1994), and that lack significant, current genetic exchange with neighbouring populations

(Palsbøll et al., 2006). The 10 subpopulations of Cl. guttata meet these criteria and, therefore,

38

represent 10 potential MUs. Effective recovery planning must account for the specific

circumstances of each MU because population size, habitat type, habitat quality, and specific

threats to persistence differ greatly among subpopulations. The larger populations identified

through Bayesian inference also meet two of the first criteria for Designatable Units (DUs), the

subspecific categorization recognized under Canadian law (Green, 2005). However, the final

criterion for DUs is variation in risk of extinction among potential units. The risk of extirpation

is high for all known subpopulations of Cl. guttata (Enneson and Litzgus, 2009). Thus,

categorization of Canadian subpopulations of Cl. guttata as DUs may not be justifiable at this

time.

The population of Cl. guttata in Hastings County is distinct from all other sampled populations.

It is particularly vulnerable to stochastic events because it is very small (N < 50), and no

genetically similar Canadian populations can be paired with it to facilitate a genetic rescue

(Tallmon et al., 2004). Although there is no evidence for the risk of extinction being higher for

this population than for the others, the Hastings County population should be prioritized for

protection because it represents a distinct genetic unit within Cl. guttata that is probably not

represented elsewhere in Canada.

Genetic assignment tests for Canadian Cl. guttata show a potential for repatriation of poached

individuals to their genetic populations of origin. Unfortunately, fine-scale discriminations (i.e.

between subpopulations) are not accurate enough to justify repatriations (< 70% accuracy for

subpopulations). Profiling of populations in the United States will further strengthen our

understanding of the genetic structure of Cl. guttata and may allow assignment of individuals

back to their source populations with greater confidence. Increased sample sizes and the

incorporation of additional markers might also increase the accuracy of identification. However,

in some cases larger sampling is not feasible. I sampled more than 80% of the census population

size at some sites, and these sample sizes are difficult to increase.

4.4.3 Long-lived organisms (turtles) and loss of diversity in fragmented populations

My results are consistent with the findings that small, isolated populations of long-lived

organisms typically retain high heterozygosity (reviewed by Vargas-Ramirez et al., 2012). This

39

is encouraging for population recovery because it gives managers time to stabilize populations

before loss of diversity becomes a concern (Kuo and Janzen, 2004; Marsack and Swanson,

2009). Furthermore, heterozygosity is often an important predictor of individual fitness

(Frankham et al., 2002). However, comparison of heterozygosity among species sampled from

different landscapes may be confounded by landscape effects.

Allelic richness is correlated with population size in Cl. guttata and may provide a more useful

measure than heterozygosity for comparing absolute genetic diversity among small and declining

populations of long-lived organisms (see also Petit et al., 1995). Heterozygosity is not correlated

with N or Ne in Cl. guttata, and even sites with critically reduced populations (N < 50) may

retain HO comparable to larger populations. Unexpectedly high heterozygosity may indicate an

undetected heterozygote advantage. Alternately, overlapping generations in turtle populations

may slow the loss of heterozygosity as population size declines, in which case high

heterozygosity would be expected in most turtle species. Several authors have made this

suggestion, but a comparative approach that eliminates confounding landscape effects is required

to test this prediction.

4.4.4 Bottleneck tests and long-lived organisms

The evaluation of bottleneck tests for detection of population declines in turtles raises two

independent concerns: 1) the repeated, uncritical use of these tests in the literature with

inadequate sample sizes, and 2) the ability of the tests to detect declines when the test

requirements are met. As the literature review demonstrated, many studies of turtles and tortoises

that use BOTTLENECK used small sample sizes and/or insufficient number of loci (Piry et al.,

1999; Table 4.7). This study also has small sample sizes, which are common in studies of

threatened taxa. It is not my intention to question the overall validity of previous studies but

rather to highlight that a basic sampling problem may be leading to questionable conclusions

regarding the true impacts of bottlenecks in declining populations of long-lived organisms such

as turtles.

The failure of the heterozygosity excess and mode shift tests to detect declines in populations of

Cl. guttata — even when the requirements of the test were met — is consistent with the

limitations of bottleneck tests already demonstrated in both natural and simulated populations

40

(Cristescu et al., 2010; Pittman et al., 2011; Peery et al., 2012). The consequences of bottlenecks

can complicate species recovery (Frankham et al., 2002) even if the genetic signature of the

bottleneck is not statistically detectable. The tests used here may fail to detect bottlenecks

because insufficient time has passed since the bottleneck event (Mockford et al., 2005) or

because long generation times may mask the genetic signature of bottlenecks (Marsack and

Swanson, 2009; Bennett et al., 2010); or they may fail because of sampling issues or limitations

inherent in the method (Peery et al., 2012). In all cases, incorrect conclusions about the genetic

health of threatened populations may slow recovery efforts or remove focus from populations on

the brink of extirpation or extinction. If bottleneck tests are used, they should be combined

wherever possible with direct evidence, such as long-term mark-recapture studies. This

combination will facilitate a more accurate evaluation of the demographic and genetic history of

a population, and its consequences for conservation and recovery measures.

4.5 Acknowledgements

This research was generously supported by the Government of Ontario (Species at Risk

Stewardship Fund grant to CMD and RWM), Wildlife Preservation Canada (Canada Collection

grant to CMD) and the National Science and Engineering Research Council of Ontario (NSERC

Discovery Grant to RWM; Canada Graduate Scholarship to CMD). Thanks to S. Coombes and a

large number of volunteers for assistance with field work. Site access and logistical support were

provided by C. Brdar, M. Cairns, J. Cebek, S. Gillingwater, J. Litzgus, M. Rasmussen, D.

Seburn, K. Yagi, A. Yagi , the Ausable Bayfield Conservation Authority, the Nature

Conservancy of Canada, Ontario Parks, Ontario Nature, Ontario Hydro and Parks Canada.

Genotyping costs were offset by the generous support of the Schad Foundation. Research

methods were approved under animal use protocols ROM2008-11, 2009-02, 2009-21 and 2010-

14) from the Animal Care Committee of the Royal Ontario Museum, under permits1045769,

1049600, 1062210, 1067079, SR-B-001-10 and AY-B-013-11 from the Ontario Ministry of

Natural Resources and under research authorizations from Ontario Parks and Parks Canada.

J. Litzgus, D. McLennan, R. Murphy and D. Seburn provided valuable comments on earlier

versions of the manuscript.

41

4.6 References Alacs, E.A., Janzen, F.J., Scribner, K.T., 2007. Genetic issues in freshwater turtle and tortoise

conservation. Chel. Res. Monogr. 4, 107–123.

Amato, M.L., Brooks, R.J., Fu, J., 2008. A phylogeographic analysis of populations of the wood turtle (Glyptemys insculpta) throughout its range. Mol. Ecol. 17, 570–581.

Banning-Anthonysamy, W.J., 2012. Spatial ecology, habitat use, genetic diversity, and reproductive success: measures of connectivity of a sympatric freshwater turtle assemblage in a fragmented landscape. PhD dissertation, University of Illinois at Urbana-Champaign.

Bennett, A.M., Keevil, M., Litzgus, J.D., 2010. Spatial ecology and population genetics of northern map turtles (Graptemys geographica) in fragmented and continuous habitats in Canada. Chel. Conserv. Biol. 9, 185–195.

Borden, W.C., Krebs, R.A., 2009. Phylogeography and postglacial dispersal of smallmouth bass (Micropterus dolomieu) into the Great Lakes. Can. J. Fish. Aquat. Sci. 66, 2142–2156.

Cagle, F., 1939. A system of marking turtles for future identification. Copeia 1939, 170–173.

Chen, C., Durand, E., Forbes, F., François, O., 2007. Bayesian clustering algorithms ascertaining spatial population structure: a new computer program and a comparison study. Mol. Ecol. Notes 7, 747–756.

Congdon, J.D., Dunham, A.E., van Loben Sels, R.C., 1993. Delayed sexual maturity and demographics of Blanding’s turtles (Emydoidea blandingii): Implications for conservation and management of long-lived organisms. Conserv. Biol. 7, 826–833.

Congdon, J.D., Dunham A.E., van Loben Sels R.C., 1994. Demographics of common snapping turtles (Chelydra serpentina): Implications for conservation and management of long-lived organisms. Am. Zool. 34, 397–408.

Cornuet J.M., Luikart G., 1997. Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics 144, 2001–2014.

COSEWIC, 2004. COSEWIC assessment and update status report on the spotted turtle Clemmys guttata in Canada. Committee on the Status of Endangered Wildlife in Canada. Ottawa. vi + 27 pp. (www.sararegistry.gc.ca/status/status_e.cfm).

COSEWIC. 2008. COSEWIC assessment and update status report on the eastern foxsnake Elaphe gloydi, Carolinian population and Great Lakes/St. Lawrence population, in Canada. Committee on the Status of Endangered Wildlife in Canada. Ottawa. vii + 45 pp. (www.sararegistry.gc.ca/status/status_e.cfm).

Crawford, N.G., 2010. SMOGD: software for the measurement of genetic diversity. Mol. Ecol. Resour. 10, 556–557.

Cristescu, R., Sherwin, W.B., Handasyde, K., Cahill, V., Cooper, D.W., 2010. Detecting bottlenecks using BOTTLENECK 1.2.02 in wild populations: the importance of the microsatellite structure. Conserv. Genet. 11, 1043–1049.

42

Cunningham, J., Baard, E.H.W., Harley, E.H., O'Ryan C., 2002. Investigation of genetic diversity in fragmented geometric tortoise (Psammobates geometricus) populations. Conserv. Genet. 3, 215–223.

Davy, C.M., Edwards, T., Lathrop, A., Bratton, M., Hagan, M., Henen, B., Nagy, K., Stone, J., Hillard, L.S., Murphy, R.W., 2011. Polyandry and multiple paternities in the threatened Agassiz’s desert tortoise, Gopherus agassizii: conservation implications. Conserv. Genet. 12, 1313–1322.

Davy, C.M., Leifso, A.E., Conflitti, I.M., Murphy, R.W., 2012. Characterization of 10 novel microsatellite loci and cross-amplification of two loci in the snapping turtle (Chelydra serpentina). Conserv. Genet. Resour. 4, 695–698.

Di Rienzo, A., Peterson, A.C., Garza, J.C., Valdes, A.M., Slatkin, M., Freimer N.B., 1994. Mutational processes of simple-sequence repeat loci in human populations. P. Natl. Acad. Sci. USA. 91, 3166–3170.

Durand, E., Jay, F., Gaggiotti, O.E., François O., 2009. Spatial inference of admixture proportions and secondary contact zones. Mol. Biol. Evol. 26, 1963–1973.

Earl, D.A., vonHoldt, B.M., 2012. STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conserv. Genet. Resour. 4, 359–361.

Echelle, A.A., Hackler, J.C., Lack, J.B., Ballard, S.R., Roman, J., Fox, S.F., Leslie D.M., Van Den Bussche, R.A., 2010. Conservation genetics of the alligator snapping turtle: cytonuclear evidence of range-wide bottleneck effects and unusually pronounced geographic structure. Conserv. Genet. 11, 1375–1387.

Edwards, T., Schwalbe, C.R., Swann D.E., Goldberg, C.S., 2004. Implications of anthropogenic landscape change on inter-population movements of the desert tortoise (Gopherus agassizii). Conserv. Genet. 5, 485–499.

Enneson, J.J., Litzgus, J.D., 2009. Stochastic and spatially explicit population viability analyses for an endangered freshwater turtle, Clemmys guttata. Can. J. Zool. 87, 1241–1254.

Ernst, C.H., Lovich, J.L., 2009. Turtles of the United States and Canada, 2nd ed, Johns Hopkins University Press, Baltimore.

Escalona, T., Engstrom, T.N., Hernandez, O.E., Bock, B.C., Vogt R.C., Valenzuela, N., 2009. Population genetics of the endangered South American freshwater turtle, Podocnemis unifilis, inferred from microsatellite DNA data. Conserv. Genet. 10, 1683–1696.

Evanno, G., Regnaut, S., Goudet J., 2005. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol. Ecol. 14, 2611–2620.

Ewing, S.R., Nager, R.G., Nicoll, M.A.C., Aumjaud, A., Jones, C.G., Keller, L.F., 2008. Inbreeding and loss of genetic variation in a reintroduced population of Mauritius kestrel. Conserv. Biol. 22, 395–404.

Excoffier, L., Laval, G., Schneider, S., 2005. ARLEQUIN ver. 3.0: An integrated software package for population genetics data analysis. Evol. Bioinform. Online 1, 47–50.

43

Falush, D., Stevens, M., Pritchard, J.K., 2003. Inference of population structure using multilocus genotype data: linked loci and correlated allele frequencies. Genetics 164, 1567–1587.

Frankham, R., 1996. Relationship of genetic variation to population size in wildlife. Conserv. Biol. 10, 1500–1508.

Frankham, R., Ballou, J.D., Briscoe, D.A., 2002. Introduction to Conservation Genetics. Cambridge University Press, Cambridge, U.K.

Fritz, U., Alcalde, L., Vargas-Ramírez, M., Goode, E.V., Fabius-Turoblin, D.U., Praschag P., 2012. Northern genetic richness and southern purity, but just one species in the Chelonoidis chilensis complex. Zool. Scr. 41, 220–232.

Gibbs, H.L, Prior, K.A., Weatherhead, P.J., Johnson, G., 1997. Genetic structure of populations of the threatened eastern Massasauga rattlesnake, Sistrurus c. catenatus: evidence from microsatellite DNA markers. Mol. Ecol. 6, 1123–1132.

Goudet, J., 1995. FSTAT (Version 1.2): A computer program to calculate F-statistics. J. Hered. 86, 485–486.

Green, D.M., 2005. Designatable units for status assessment of endangered species. Conserv. Biol. 19, 1813–1820.

Guillot, G., Mortier, F., Estoup, A., 2005. GENELAND: a program for landscape genetics. Mol. Ecol. Notes 5, 712–715.

Guillot, G., Leblois, R., Coulon A., Frantz, A.C., 2009. Statistical methods in spatial genetics. Mol. Ecol. 18, 4734–4756.

Hauswaldt, J.S., Glenn, T.C., 2005. Population genetics of the diamondback terrapin (Malaclemys terrapin). Mol. Ecol. 14, 723–732.

Holman, J.A., 1992. Late Quaternary herpetofauna of the central Great Lakes region, U.S.A.: zoogeographical and paleoecological implications. Quaternary Sci. Rev. 11, 345–351.

Jakobsson, M., Rosenberg, N.A., 2007. CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics 23, 1801–1806.

Jensen, J.L., Bohonak, A.J., Kelley, S.T., 2005. Isolation by distance, web service. BMC Genetics 6, 13. v.3.23 http://ibdws.sdsu.edu/

Jost, L., 2008. GST and its relatives do not measure differentiation. Mol. Ecol. 17, 4015–4026.

Kalinowski, S.T., 2004. Counting alleles with rarefaction: private alleles and hierarchical sampling designs. Conserv. Genet. 5, 539–543.

Kalinowski, S.T., 2005. HP-Rare: A computer program for performing rarefaction on measures of allelic diversity. Mol. Ecol. Notes 5, 187–189.

Kidd, A., Reid S., Wilson, C., 2011. Local and regional population genetic structure of the threatened channel darter in Ontario and Quebec. Poster presentation at American Fisheries Society 41st Annual Meeting, Seattle, Washington, Sept. 4–8, 2011.

44

King, T.L., Julian, S.E., 2004. Conservation of microsatellite DNA flanking sequences across 13 emydid genera assayed with novel bog turtle (Glyptemys muhlenbergii) loci. Conserv. Genet. 5, 719–725.

Kuo, C.H., Janzen, F.J., 2004. Genetic effects of a persistent bottleneck on a natural population of ornate box turtles (Terrapene ornata). Conserv. Genet. 5, 425–437.

Latch, E.K., Dharmarajan, G., Glaubitz, J.C., Rhodes, O.E., 2006. Relative performance of Bayesian clustering software for inferring population substructure and individual assignment at low levels of population differentiation. Conserv. Genet. 7, 295–302.

Litzgus, J.D., 1996. Life history and demography of a northern population of spotted turtles, Clemmys guttata. MSc Thesis, University of Guelph, Ontario. 145 pp.

Litzgus, J.D., 2006. Sex differences in longevity in the spotted turtle (Clemmys guttata). Copeia 2006, 281–288.

Luikart, G., Allendorf, F.W., Cornuet, J.M., Sherwin, W.B., 1997. Distortion of allele frequency distributions provides a test for recent population bottlenecks. J. Hered. 89, 238–247.

Manel, S., Schwartz, M.K., Luikart, G., Taberlet, P., 2003. Landscape genetics: combining landscape ecology and population genetics. Trends Ecol. Evol. 18, 189–197.

Mantel, N., 1967. The detection of disease clustering and a generalized regression approach. Cancer Res. 27, 209–220.

Marsack, K., Swanson, B.J., 2009. A genetic analysis of the impact of generation time and road-based habitat fragmentation on eastern box turtles (Terrapene c. carolina). Copeia 2009, 647–652.

Mockford, S.W., McEachern, L., Herman, T.B., Snyder, M., Wright, J.M., 2005. Population genetic structure of a disjunct population of Blanding’s turtle (Emydoidea blandingii) in Nova Scotia, Canada. Biol. Conserv. 123, 373–380.

Moritz, C., 1994. Defining ‘evolutionarily significant units’ for conservation. Trends Ecol. Evol. 9, 373–375.

Murphy, R.W., Berry, K.H., Edwards, T., McLuckie, A.M., 2007. A genetic assessment of the recovery units for the Mojave population of the desert tortoise, Gopherus agassizii. Chelonian Conserv. Biol. 6, 229–251.

Paetkau, D., Slade, R., Burden M., Estoup, A., 2004. Direct, real-time estimation of migration rate using assignment methods: a simulation-based exploration of accuracy and power. Mol. Ecol. 13, 55–65.

Palsbøll, P.J., Bérubé M., Allendorf, F.W., 2006. Identification of management units using population genetic data. Trends Ecol. Evol. 22, 11–16.

Peakall, R., Smouse, P.E., 2006. GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Mol. Ecol. Notes 6, 288–295.

Pearse, D.E., Arndt, A.D., Valenzuela, N., Miller, B.A., Cantarelli, V., Sites J.W. Jr., 2006. Estimating population structure under nonequilibrium conditions in a conservation context: continent-wide population genetics of the giant Amazon river turtle, Podocnemis expansa (Chelonia; Podocnemididae). Mol. Ecol. 15, 985–1006.

45

Peery, M.Z., Kirby, R., Reid, B.N., Stoelting, R., Doucet-Bëer, E., Robinson, S., Vásquez-Carrillo, C., Pauli, J.N., Palsbøll, P.J., 2012. Reliability of genetic bottleneck tests for detecting recent population declines. Mol. Ecol. 21, 3403–3418.

Pelletier, A., Obbard, M.E., White, B.N., Doyle, C., Kyle, C.J. 2011. Small-scale genetic structure of American black bears illustrates potential postglacial recolonization routes. J. Mammal. 92,629-644.

Perez, M., Leblois, R., Livoreil, B., Bour, R., Lambourdiere, J., Samadi, S., Boisselier, M., 2011. Effects of landscape features and demographic history on the genetic structure of Testudo marginata populations in the southern Peloponnese and Sardinia. Biol. J. Linn. Soc. 105, 591–606.

Petit RJ, El Mousadi A, Pons O (1995) Identifying populations for conservation on the basis of genetic markers. Conserv. Biol. 12, 844–855.

Piry, S., Luikart, G., Cornuet, J.M., 1999. BOTTLENECK: A program for detecting recent effective population size reductions from allele frequency data. J. Hered. 90, 502–503.

Piry, S., Alapetite, A., Cornuet, J.M., Paetkau, D., Baudouin, L., Estoup, A., 2004. GENECLASS2: a software for genetic assignment and first-generation migrant detection. J. Hered. 95, 536–539.

Pittman, S. E., King, T., Faurby, S., Dorcas, M.E., 2011. Genetic and demographic status of an isolated bog turtle (Glyptemys muhlenbergii) population: implications for the conservation of small populations of long-lived animals. Conserv. Genet. 12, 1589–1601.

Pompanon, F., Bonin, A., Bellemain, E., Taberlet, P., 2005. Genotyping errors: causes, consequences and solutions. 2005. Nat. Rev. Genet. 6, 847–846.

Pritchard, J.K., Stephens, M., Donnelly, P.J., 2000. Inference of population structure using multilocus genotype data. Genetics 155, 945–959.

Rannala, B., Mountain J. L., 1997. Detecting immigration by using multilocus genotypes. P. Natl. Acad. Sci. USA. 94, 9197–9221.

Raymond, M., Rousset, F., 1995. GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. J. Hered. 86, 248–249.

Rice, W.R., 1989. Analyzing tables of statistical tests. Evolution 43, 223–225.

Richter, S.C., Jackson, J.A., Hinderliter, M., Epperson, D., Theodorakis, C.W., Adams, S.M., 2011. Conservation genetics of the largest cluster of federally threatened gopher tortoise (Gopherus polyphemus) colonies with implications for species management. Herpetologica 67, 406–419.

Rosenberg, N.A., 2004. DISTRUCT: a program for the graphical display of population structure. Mol. Ecol. Notes 4, 137–138.

Rousset, F., 2008. GENEPOP̀007: a complete reimplementation of the GENEPOP software for Windows and Linux. Mol. Ecol. Res. 8, 103–106.

Row, J.R., Blouin-Demers, G., Lougheed, S.C., 2010. Habitat distribution influences dispersal and fine-scale genetic population structure of eastern foxsnakes (Mintonius gloydi) across a fragmented landscape. Mol. Ecol. 19, 5157–5171.

46

Rubin, C.S., Warner, R.E., Bouzat, J.L., Paige, K.N., 2001. Population genetic structure of Blanding’s turtles (Emydoidea blandingii) in an urban landscape. Biol. Conserv. 99, 323–330.

Schuelke, M., 2000. An economic method for the fluorescent labeling of PCR fragments. Nat. Biotechnol. 18, 233–234.

Schwartz, T.S., Karl, S.A., 2005. Population and conservation genetics of the gopher tortoise (Gopherus polyphemus). Conserv. Genet. 6, 917 – 928.

Seburn, D.C., 2003. Population structure, growth, and age estimation of spotted turtles, Clemmys guttata, near their northern limit: an 18-year follow-up. Can. Field Nat. 117, 436–439.

Spradling, T.A., Tamplin, J.W., Dow, S.S., Meyer, K.J., 2010. Conservation genetics of a peripherally isolated population of the wood turtle (Glyptemys insculpta) in Iowa. Conserv. Genet. 5, 1667–1677.

Stepien, C.A., Murphy, D.J., Lohner, R.N., Sepulveda, O.J., Haponski, A.E., 2009. Signatures of vicariance, postglacial dispersal and spawning philopatry: population genetics of the walleye Sander vitreus. Mol. Ecol. 18, 3411–3428.

Taberlet, P., Luikart G., 1999. Non-invasive genetic sampling and individual identification. Biol. J. Linn. Soc. 68, 41–55.

Tallmon, D.A., Luikart G., Waples, R.S., 2004. The alluring simplicity and complex reality of genetic rescue. Trends Ecol. Evol. 19, 489–496.

Tallmon, D.A., Koyuk, A., Luikart, G.H., Beaumont, M.A., 2008. ONeSAMP: a program to estimate effective population size using approximate Bayesian computation. Mol. Ecol. Res. 8, 299-301.

Tessier, N., Paquette, S., Lapointe, F.J., 2005. Conservation genetics of the wood turtle (Glyptemys insculpta) in Quebec, Canada. Can. J. Zool. 83, 765–772.

van Dijk, P.P., 2011. Clemmys guttata. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.1. <www.iucnredlist.org>. Downloaded on 21 August 2012

van Oosterhout, C., Hutchinson, W.F., Wills, D.P.M., Shipley, P., 2004. MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data. Mol. Ecol. Notes 4, 535–538.

Vargas-Ramirez M., Stuckas, H., Castaňo-Mora, O.V., Fritz., U., 2012. Extremely low genetic diversity and weak population differentiation in the endangered Colombian river turtle Podocnemis lewyana (Testudines: Podocnemididae) Conserv. Genet. 13, 65–77.

Velo-Antón, G., Becker C.G., Cordero-Rivera, A., 2011. Turtle carapace anomalies: the roles of genetic diversity and environment. PLoS ONE, 6:e18714.

Waits, L.P., Luikart, G., Taberlet, P., 2001. Estimating the probability of identity among genotypes in natural populations: cautions and guidelines. Mol. Ecol. 10, 249–256.

White, G.C., Burnham, K. P., 1999. Program MARK: Survival estimation from populations of marked animals. Bird Study 46 (supplement), 120–138.

Wright, S., 1943. Isolation by distance. Genetics 28, 114–138.

47

Table 4.1. Summary statistics for 11 microsatellite loci originally developed for the Glyptemys muhlenbergii (King and Julian 2004)

and amplified in 256 Clemmys guttata from southern Ontario. Temp. = annealing temperature (°C) used in PCR amplification. “*”

indicates an initial touchdown of 1°C/cycle from 10°C above the annealing temperature, followed by a constant annealing temperature

for the remaining cycles; N = number of individuals amplified at each locus; k = number of alleles; Ne = number of effective alleles;

HO = observed heterozygosity; HE = expected heterozygosity; PI = probability of identity, PIsibs = Probability of identity for siblings

at a locus.

Locus Temp ( °C) N k Ne HO HE PI PIsibs GmuA19 61.5 245 3 2.955 0.624 0.662 0.189 0.466 GmuB08 58* 250 3 1.993 0.484 0.498 0.372 0.594 GmuD16 58 250 11 5.710 0.768 0.825 0.052 0.35 GmuD21 58* 242 9 4.225 0.769 0.763 0.08 0.388 GmuD55 58* 245 10 4.498 0.702 0.778 0.073 0.379 GmuD79 61.5 252 7 4.892 0.758 0.796 0.071 0.37 GmuD87 61.5 238 18 13.309 0.891 0.925 0.01 0.29 GmuD88 61.5 243 12 2.761 0.547 0.638 0.158 0.47 GmuD107 54 240 15 8.641 0.858 0.884 0.024 0.313 GmuD114 60 249 4 2.241 0.518 0.554 0.279 0.543 GmuD121 56 254 10 3.936 0.657 0.746 0.1 0.402

48

Table 4.2. Number of alleles (private alleles in parentheses), observed and expected heterozygosities (HO and HE), and estimated

frequency of a null allele (for each locus across all populations), for 256 Clemmys guttata from southern Ontario, by locus and

populations (see Figure 1 for definition of site acronyms).

LH1 LH2 BP1 BP2 LE1 LE2 GH1 GH2 GB1 GB2 HC EO1 EO2* DOR1 DOR2

GmuB08 Number of alleles 2 2 2 2 2 3 2 2 2 3 2 2 2 – 1 Estimated null allele HO 0.414 0.433 0.500 0.556 0.533 0.455 0.480 0.333 0.333 0.647 0.438 0.571 0.471 – –

frequency = 0.00 HE 0.428 0.495 0.375 0.489 0.491 0.541 0.461 0.500 0.278 0.493 0.342 0.459 0.457 – – N 29 30 4 27 30 11 25 6 6 17 16 14 34 0 1

GmuD16 Number of alleles 7 8 5 6 8 8 8 4 4 9 6 5 5 (1) – 2 Estimated null allele HO 0.828 0.800 1.000 0.593 0.700 0.917 0.792 0.667 0.667 0.941 0.688 0.643 0.824 – –

frequency = 0.00 HE 0.815 0.782 0.688 0.763 0.812 0.813 0.821 0.653 0.625 0.794 0.693 0.617 0.734 – – N 29 30 4 27 30 12 24 6 6 17 16 14 34 0 1

GmuD55 Number of alleles 5 8 3 7 9 6 7 3 4 6 6 7 7 – 2 Estimated null allele HO 0.414 0.600 0.750 0.741 0.833 0.583 0.591* 0.400 0.750 0.882 0.688 0.929 0.853 – –

frequency = 0.134 HE 0.542 0.639 0.531 0.768 0.764 0.476 0.697 0.540 0.688 0.739 0.668 0.778 0.804 – – N 29 30 4 27 30 12 22 5 4 17 16 14 34 0 1

GmuD79 Number of alleles 5 6 4 7 6 5 7 4 3 5 4 4 5 – 1 Estimated null allele HO 0.759 0.800 0.750 0.815 0.871 0.727 0.750 0.429 0.667 0.765 0.750 0.929 0.647 – –

frequency = 0.00 HE 0.748 0.784 0.719 0.720 0.793 0.669 0.700 0.367 0.611 0.619 0.686 0.640 0.554 – – N 29 30 4 27 31 11 24 7 6 17 16 14 34 0 2

GmuD88 Number of alleles 6 6 3 5 7 (1) 4 7 2 4 (1) 3 3 4 6 (1) – 2 Estimated null allele HO 0.370* 0.517 0.500 0.577 0.667 0.545 0.720 0.000 0.500 0.400 0.188* 0.615 0.765 – –

frequency = 0.173 HE 0.517 0.650 0.531 0.649 0.655 0.442 0.694 0.278 0.514 0.464 0.365 0.559 0.707 – – N 27 29 4 26 30 11 25 6 6 15 16 13 34 0 1

GmuD107 Number of alleles 8 11 5 9 11 8 12 5 6 6 5 (1) 5 9 2 2 Estimated null allele HO 0.815* 0.833 1.000 0.913 0.867 0.917 0.818 0.571 1.000 0.882 0.800 0.786 0.935 – –

frequency = 0.00 HE 0.776 0.842 0.781 0.854 0.862 0.795 0.857 0.673 0.806 0.777 0.716 0.615 0.830 – – N 27 30 4 23 30 12 22 7 6 17 15 14 31 1 1

GmuD114 Number of alleles 3 4 3 3 3 3 2 2 2 3 2 2 4 – 1 Estimated null allele HO 0.379 0.533 0.500 0.444 0.500 0.750 0.478 0.333 0.500 0.529 0.375 0.643 0.706 – –

frequency = 0.00 HE 0.372 0.534 0.406 0.529 0.545 0.538 0.496 0.444 0.375 0.562 0.469 0.497 0.643 – – N 29 30 4 27 30 12 23 6 6 17 16 14 34 0 1

GmuD121 Number of alleles 7 6 2 4 6 7 7 3 5 5 3 3 5 2 (1) 2 Estimated null allele HO 0.724 0.933 0.250 0.519 0.613 0.545 0.615 0.667 0.833 0.588 0.750 0.643 0.588 – –

frequency = 0.078 HE 0.781 0.796 0.469 0.578 0.743 0.727 0.753 0.611 0.694 0.683 0.639 0.582 0.500 – –

49

N 29 30 4 27 31 11 26 6 6 17 16 14 34 1 2

GmuD21 Number of alleles 5 7 1 6 7 5 6 4 4 5 (1) 5 5 6 1 2 Estimated null allele HO 0.852 0.852 0.000 0.538 0.800 0.900 0.769 0.714 0.833 0.600 0.688 0.917 0.912 – –

frequency = 0.00 HE 0.752 0.799 0.000 0.498 0.773 0.735 0.791 0.602 0.597 0.484 0.678 0.708 0.787 – – N 27 27 4 26 30 10 26 7 6 15 16 12 34 1 1

GmuD87 Number of alleles 12 15 4 9 15 12 13 7 6 11 4 7 13 – 2 Estimated null allele HO 0.852 0.900 0.750 0.846 0.933 0.900* 0.875 0.833 1.000 0.938 0.643 1.000 0.971 – –

frequency = 0.00 HE 0.861 0.866 0.563 0.770 0.844 0.890 0.882 0.708 0.778 0.873 0.630 0.830 0.891 – – N 27 30 4 26 30 10 24 6 6 16 14 10 34 0 1

GmuA19 Number of alleles 3 3 3 3 3 3 3 3 3 3 3 3 3 – 2 Estimated null allele HO 0.786 0.700 0.750 0.593 0.567 0.364 0.480 0.667 0.500 0.667 0.867 0.500 0.606 – –

frequency = 0.00 HE 0.663 0.626 0.531 0.656 0.638 0.665 0.655 0.625 0.611 0.624 0.660 0.630 0.496 – – N 28 30 4 27 30 11 25 6 6 15 15 14 33 0 1

* One individual from this site was excluded because of repeated triplicate peaks in its pherograms. Data from this individual are

excluded from this table.

50

Table 4.3. Genetic diversity (heterozygosity, allelic richness and private allelic richness) of sampled regions, genetic populations and

sites for 253 Clemmys guttata genotyped at 11 microsatellite loci. Allelic and private allelic richness are rarefacted to account for

variation in sample sizes (Kalinowski 2004). Pop1–5 = genetic clusters supported by both STRUCTURE and TESS analyses. Georgian

Bay is considered independently. HO = observed heterozygosity averaged across all loci; HE = expected heterozygosity averaged

across all loci. See Figure 1 and text for definitions of site acronyms.

Region Genetic

population Sampling Site

HO HE Allelic richness

Private allelic richness

Southwestern Ontario 0.679 0.728 6.75 1.72 Pop1 0.680 0.724 5.52 0.23

LH1 0.654 0.659 4.01 0.18 LH2 0.718 0.71 4.44 0.25

Pop2 0.687 0.703 4.64 0.16 BP1 0.614 0.509 3.18 0.19 BP2 0.649 0.661 3.92 0.13

Pop3 0.644 0.662 6.88 0.57 LE1 0.717 0.72 4.49 0.24 LE2 0.691 0.663 4.37 0.35 GH1 0.67 0.71 4.48 0.28 GH2 0.51 0.546 3.35 0.1

Georgian Bay 0.708 0.651 5.31 0.35 GB1 0.689 0.598 3.81 0.24 GB2 0.713 0.647 3.98 0.35

Southeastern Ontario 0.718 0.707 6.11 1.08 Pop4 0.625 0.595 3.34 0.18

HC 0.625 0.595 3.34 0.18 Pop5 0.749 0.694 6.11 0.61

EO1 0.743 0.629 3.53 0.16 EO2 0.742 0.673 4.1 0.35

51

Table 4.4. Pairwise values FST (below the diagonal) and Dest (Jost 2008, above diagonal) for 13 putative subpopulations of Clemmys

guttata sampled across southern ON (N = 253; see Figure 1 for definition of site acronyms). Sites in the Golden Horseshoe, Georgian

Bay and the Bruce Peninsula are analyzed together. FST values in italics are not significant (p > 0.05).

LH1 LH2 BP LE1 LE2 GB GH HC EO1 EO2

LH1 0.039 0.16 0.104 0.1 0.095 0.072 0.161 0.171 0.229

LH2 0.031 0.137 0.067 0.04 0.068 0.065 0.111 0.142 0.164

BP 0.091 0.069 0.102 0.052 0.104 0.088 0.15 0.084 0.161

LE1 0.061 0.041 0.058 0.045 0.062 0.021 0.118 0.133 0.142

LE2 0.067 0.042 0.047 0.032 0.077 0.013 0.144 0.094 0.172

GB 0.071 0.049 0.065 0.044 0.052 0.06 0.136 0.12 0.098

GH 0.051 0.045 0.057 0.022 0.009 0.043 0.123 0.089 0.116

HC 0.134 0.095 0.11 0.097 0.124 0.12 0.104 0.157 0.226

EO1 0.112 0.078 0.082 0.067 0.079 0.082 0.055 0.138 0.102

EO2 0.107 0.078 0.096 0.078 0.082 0.078 0.073 0.135 0.089

Table 4.5. Hierarchical analysis of molecular variance (AMOVA; Excoffier et al. 1992) conducted in ARLEQUIN. Each source of

variation was significant (p < 0.0001). Tested populations were those identified by both STRUCTURE and TESS analyses with GB treated

as a separate, sixth population. Subpopulations refer to sampling sites except GB, GH and BP which are treated as single

subpopulations.

Source of variation Sum of squares Variance components (σ2)

% variation

Among populations 102.237 0.158 3.829

Among subpopulations within populations

59.966 0.180 4.384

Within subpopulations 1637.154 3.778 91.787

Total 1799.357 4.116

52

Table 4.6. Assignment of individuals in GENECLASS analysis based on sampling sites; 66.3% of individuals were assigned correctly.

Shaded areas indicate clustering of sites in genetic populations supported by both STRUCTURE and TESS. Sampling sites correspond to

Figure 1.

Assigned to: LH1 LH2 BP1 BP2 LE1 LE2 GH1 GH2 GB1 GB2 HC EO1 EO2

LH1 20 6 1 2

LH2 1 26 2

BP1 2 2

BP2 1 18 8

LE1 2 26 2

LE2 1 2 3 6

GH1 1 3 1 5 1 15

GH2 1 4 2

GB1 2 1 1 2

GB2 3 3 2 9

HC 4 1 11

EO1 1 1 1 9 2

Sam

ple

so

urc

e:

EO2 1 1 2 30

53

Table 4.7. Summary of bottleneck tests in published studies of population genetics of tortoises and freshwater turtles. Loci = the

number of loci used in bottleneck tests (in some cases this was lower than the total number amplified). Individuals = the maximum-

minimum and mean () number of individuals genotyped per tested population. When only populations above a certain size limit

were used, only these populations were included in the summary. Where values for loci and individuals are in bold this indicates that

the minimum sampling recommendations for tests in BOTTLENECK were met.

Source Species Loci Individuals Mode shift detected?

Significant heterozygosity excess detected?

Significantly decreased M-ratio detected?

Cunningham et al. (2002)

Psammobates geometricus

8 25–28; = 26.3 No No 3 of 3 tests significant.

Edwards et al. (2004)

Gopherus morafkai

7 9–38; = 18.8 -- No No

Kuo and Janzen (2004)

Terrapene ornata

11 73–74; = 73.5 No No No

Mockford et al. (2005)

Emydoidea blandingii

5 27–43; = 36.7 No No --

Schwartz and Karl (2005)

Gopherus polyphemus

9 11–26; = 19.1 -- 5 of 14 tests significant --

Hauswaldt and Glen (2005)

Malaclemys terrapin

6 12–56; = 24.8 -- No No

Pearse et al. (2006)

Podocnemis expansa

9 16–37; = 26.6 -- 5 of 11 tests significant 11 of 11 tests significant

Murphy et al. (2007)

Gopherus agassizii

11 18–83; = 41.9 No 2 of 15 tests significant No

Escalona et al. (2009)

Podocnemis unifilis

5 14–55; = 28.4 No No Yes, 10 of 11 significant

Marsack and Swanson (2009)

Terrapene carolina carolina

8 40–70; = 54.3 No 2 of 3 tests significant No

Echelle et al. (2010)

Macrochelys temmincki

7 N ≥ 10 per tested population Y, 4 of 12 tests significant

No Yes, 10 of 12 significant

54

Spradling et al. 2010

Glyptemys insculpta

9 51 and 80 No No No

Pittman et al. 2011

Glyptemys muhlenbergii

18 8.9–34.7; = 15.5*

-- 2 of 6 tests significant --

Richter et al. 2011

Gopherus polyphemus

9 11–40, = 18.8 -- No --

Velo-Anton et al. 2011

Emys orbicularis 7 23–36; = 29.7 No 3 of 9 tests significant 5 of 9 tests significant

Vargas-Ramirez et al. 2012

Podocnemis lewyana

10 4–49; = 21 2 of 3 tests significant §

7 of 7 tests significant 7 of 7 tests significant

Perez et al. 2012 Testudo marginata

11 18 No Significant under IAM but not under TPM (Piry et al. 1999)

Yes

Fritz et al. 2012 Chelonoidis chilensis

10 21 -- No --

* Pittman et al. (2011) present mean sample size per locus

§ Tests of seven sites were not significant; when sites were grouped into three populations, two of these demonstrated a significant

mode shift.

55

Figure 4.1. Approximate location of sampled sites. LE = Lake Erie; LH = Lake Huron; BP =

Bruce Peninsula; GB = Georgian Bay; GH = Golden Horseshoe; HC = Hastings County; DOR =

dead on road; EO = Eastern Ontario.

56

Figure 4.2. A) Population structure inferred by STRUCTURE and TESS for increasing values of K.

Colours indicating populations in the K = 5 model (marked with an asterisk) match colours used

in Figure 4.4. B) Estimated ln probability of the data (L(K)) for STRUCTURE analyses at

increasing values of K with 8 independent runs at each. C) ∆K (Evanno et al. 2005) calculated

from (B). D) TESS results: decreasing deviance information criterion (DIC) with increasing Kmax.

57

Figure 4.3. Principal Coordinates Analysis plot based on Dest (Table 4.4) for populations (a, b)

and based on genetic distance for individuals (c).

58

Figure 4.4. Genetic population structure identified by STRUCTURE and TESS with K = 5 (Figure

2). Georgian Bay was assigned to different populations by the two analyses. Hypothesized

dispersal routes for Clemmys guttata colonizing Canada after glacial retreat are indicated by the

large grey arrows.

59

Chapter 5 Unexpected patterns of genetic diversity in two sympatric species

of turtle.

Formatted for Molecular Ecology.

5 Abstract

Studies of conservation genetics in natural populations often assumed that the genetic diversity

of wild populations can be predicted by the population size, behavior and ecology of the study

species. I used approximate Bayesian computation to estimate effective population size (Ne) and

a suite of spatial genetics analyses to compare genetic diversity in two sympatric species of

freshwater turtles, Chelydra serpentina and Clemmys guttata, sampled across southern Ontario,

Canada. The results did not support the hypothesis that higher vagility, fecundity and population

size predicted higher genetic diversity. Bayesian clustering analyses revealed significant

population structure in both species across the study area. Despite substantial differences in

contemporary population sizes, estimates of Ne were unexpectedly comparable between species.

The different in the Ne:Nc ratios in these two species may result from behavioural differences

(specifically mating and nesting behaviour) that serve to increase reproductive variance in the

snapping turtle, depressing Ne relative to Nc, while decreasing reproductive variance in the

spotted turtle with the opposite effect. Unexpectedly high Ne in Cl. guttata may have improved

the outlook for recovery of this endangered species. In contrast, depressed Ne estimates in Ch.

serpentina suggested that this “common” species may be particularly vulnerable to the genetic

impacts of population declines and overharvesting. These results illustrated the dangers of

making assumptions about the genetic health of populations of “common” and “rare” species.

60

Keywords: microsatellite, effective population size, heterozygosity, STRUCTURE, TESS, landscape

genetics

5.1 Introduction

A central tenet of conservation biology involves understanding the drivers that affect genetic

variation within and among populations because genetic variation predicts the evolutionary

potential of a species (Fisher 1930; Frankham 1995a). For example, genetic drift, the stochastic

shift in allelic frequencies over time due to random sampling of alleles between generations,

tends to affect smaller populations more strongly than larger ones. Consequently, smaller

populations are considered especially vulnerable to the negative effects of genetic drift,

specifically loss of heterozygosity, loss of allelic richness and potential inbreeding depression

(Frankham 1995a; Frankham et al. 2002).

The theoretical relationship between population size (N) and the rate of genetic drift is clear in

the equation 1/(2Ne), where Ne is the effective population size, an estimate of the relative number

of breeding individuals per generation. The equation estimates the stochastic per-generation loss

of heterozygosity and the probability that an allele will be lost from one generation to the next

due to random sampling. In wild populations, Ne typically averages 10% of the census

population size (Frankham 1995b). Thus, the effect of genetic drift inversely correlates with

population size. This predicts that small, fragmented populations will lose genetic diversity more

quickly than large, connected ones (Nei et al. 1975; Frankham et al. 2002; Ewing et al. 2008).

Meta-analyses show a positive correlation between population size and genetic diversity in many

taxa, which is consistent with this prediction (Soulé 1976; Frankham 1996; Leimu et al. 2006).

Other correlates of genetic diversity include dispersal ability, fecundity and range size

(Frankham 1996; Mitton 1997).

Recent studies of the population genetics of turtles reveal high heterozygosity across most

species (summarized in Vargas-Ramirez et al. 2012), even in bottlenecked populations in which

reduced diversity is expected (e.g. Kuo & Janzen 2004; Marsack & Swanson 2009; Chapter 3).

The extreme longevity of turtles (over 100 years in some cases) is usually invoked to explain this

phenomenon (Kuo & Janzen 2004; Alacs et al. 2007). The only empirical test of this hypothesis

61

to date compared a long-lived species of turtle to a short-lived species of snake sampled from

different landscapes; generation time did not significantly impact on diversity (Howes et al.

2009).

In this study I compare genetic diversity and population structure in two sympatric species of

freshwater turtle across a single landscape to understand whether drivers of genetic diversity are

consistent among long-lived organisms. I test the hypothesis that snapping turtles (Chelydra

serpentina) will exhibit higher genetic diversity than spotted turtles (Clemmys guttata) based on

the following five, widely accepted predictions (Frankham 1996; Mitton 1997): (1) genetic

variation will be higher in a species with larger population size than a species with smaller

population size, considering both census and effective population sizes; (2) genetic variation will

be higher and populations will show less structure in a species with larger ranges and higher

dispersal ability than in a species with smaller ranges and lower dispersal ability; (3) genetic

variation will be higher in a widespread species than in a restricted species; and (4) genetic

variation will be higher in a species with high fecundity than in a species with low fecundity.

5.2 Methods

5.2.1 Study species

The widespread snapping turtle (Chelydra serpentina) occurs across the eastern United States

and south-eastern Canada, and has been introduced to locations on the west coast of North

America. Fecundity is high with a mean clutch size of 35.2 (Ernst & Lovich 2009). Chelydra

serpentina is considered abundant across its range, but recent data suggest population declines

due to over-harvesting (van Dijk 2011). Dispersal ability is relatively high, and radio-tracking

studies and observational studies have regularly recorded movements of 2 to 5 km (Ernst &

Lovich 2009; J. Paterson, pers. comm.). Estimated generation time in Ontario is 31 years and

longevity may exceed 100 years in the wild (Galbraith & Brooks 1987a; Galbraith et al. 1989; R.

Brooks, unpublished data, in COSEWIC 2008).

The endangered spotted turtle (Clemmys guttata) is distributed along the east coast of the United

States from Florida northwards to Maine and westwards through southern Ontario and into

Illinois. Fecundity is low with a mean clutch size of 3.5 (Ernst & Lovich 2009). Populations of

Cl. guttata are typically small and are geographically isolated across its range (Ernst & Lovich

62

2009). Low dispersal ability compounds the isolation of populations: individuals may not move

more than 500m in a year, and rarely undertake movements > 2 km (Litzgus 1996; Ernst &

Lovich 2009; Banning-Anthonysamy 2012). Many populations of Cl. guttata are in decline due

to habitat loss and illegal collection and several local extirpations have been recorded

(COSEWIC 2004; Ernst & Lovich 2009). Most Ontarian populations have < 200 individuals, and

several populations have a census size < 50 (COSEWIC 2004; C. Davy, unpublished data).

Generation time is > 25 years (COSEWIC 2004) and longevity is high, potentially up to 110

years (Litzgus 2006).

5.2.2 Data collection and analyses – Chelydra serpentina

Figure 5.1 shows approximate locations of collection sites for Ch. serpentina. Specific location

details were withheld to avoid exposing populations to increased harvesting pressure. Mature Ch.

serpentina were captured by hand, in dip-nets and in hoop traps baited with sardines. Blood

sampling and DNA extraction followed Davy et al. (2012).

Further blood samples stored in heparin were also obtained from turtles rehabilitated at the

Kawartha Turtle Trauma Centre (Peterborough, Ontario). I isolated genomic DNA from

heparinized blood with a phenol-chloroform extraction (Sambrook et al. 1989) and cleaned the

DNA using EtOH precipitation.

I genotyped samples at 11 species-specific polymorphic microsatellite loci following Davy et al.

(2012). Genotypes with an RFU (relative fluorescence units) peak > 200 were scored and PCR

was repeated for genotypes with a weaker signal. Seven samples (4%) were extracted twice and

genotyped twice at each locus to assess genotyping error. Duplicate, independently taken blood

samples (Pompanon et al. 2005) were available from only one individual. Other duplicate

extractions were taken from single samples.

5.2.2.1 Analysis

Genotypes were checked for errors and for evidence of stuttering, long allele dropout and null

alleles using MICRO-CHECKER v.2.2.3 (vanOosterhout et al. 2004). I used the method of

Brookfield (1996) to estimate frequencies of null alleles.

63

Chelydra serpentina is still relatively widespread in Ontario

(http://www.ontarionature.org/protect/species/reptiles_and_amphibians/map_snapping_turtleSO.

html). Therefore, I considered panmixia to be likely and did not assign individuals to

“populations” a priori. Instead, genetically continuous populations were defined based on

analyses of the data using STRUCTURE V.2.3.4 (Pritchard et al. 2000) and TESS V.2.3.1 (Chen et

al. 2007), following the parameter settings outlined in Chapter 3. Eight models were tested using

STRUCTURE ranging from panmixia (K = 1) to a highly structured population (K = 8). TESS was

run for a range of Kmax values from two to nine, with 10 independent runs at each Kmax.

STRUCTURE and TESS output were compiled following Chapter Four.

I used GENALEX (Peakall & Smouse 2006) to calculate the number of alleles at each locus, the

mean observed and expected heterozygosity (HO and HE) of each sampled site, and

heterozygosity of genetic clusters identified by STRUCTURE and TESS. Deviations from Hardy-

Weinberg Equilibrium (HWE) were assessed using GENEPOP v.4.0.10 (Raymond & Rousset

1995; Rousset 2008). Deviation from HWE was tested in sampling areas with N > 18, and for

populations identified by STRUCTURE and TESS. Each test was run with 1,000 iterations. A

sequential Bonferroni correction was applied to multiple pairwise comparisons (Rice 1989).

ARLEQUIN (Excoffier et al. 2005) was used to calculate FIS for each population and pairwise FST

values for population pair, and I assessed significance with 10,000 randomizations. I also

estimated absolute pairwise differentiation (Dest, Jost 2008) using SMOGD (Crawford 2010).

Allelic richness and private allelic richness were rarefacted using HP-RARE (Kalinowski 2004;

2005) to account for variation in sample size. Genetic distances (Dest) were ordinated among

populations and among individuals using principal coordinates analysis (PCoA) in GENALEX.

Correlations between genetic distance (Dest) and Euclidean distance between sample sites

(isolation by distance; Wright 1943) were tested for significance using IBDWS V.3.23 (Jensen et

al. 2005).

Effective population size was estimated for each site and each inferred population using

approximate Bayesian computation, implemented in ONeSAMP (Tallmon et al. 2008). Estimates

of Ne for each site were given prior lower and upper bounds of 4 and 400. Estimates of Ne for

each inferred genetic population were constrained between 4 and 1,500.

64

5.2.3 Data collection and analyses – Clemmys guttata

Data collection and analysis for Cl. guttata paralleled those for Ch. serpentina and were

described in detail in Chapter 3.

5.2.4 Interspecific comparisons

Genetic diversity (HO, HE and Ar) and Ne were compared twice between species using SPSS

v.20.0 (SPSS Inc. Chicago, Illinois). Diversity and Ne between species were compared between

the two species at five paired sampling sites using a paired Wilcoxon signed ranks test. Where

possible, samples were collected from both species at one location. Nearby sites were paired for

comparisons when exact overlap was not possible due to differences in distribution of the two

species (Figure 5.1, inset). Secondly, normality of the data was confirmed using a Shapiro-Wilks

test and an independent samples t-test was used to compare mean diversity and Ne between

species across all sampled sites.

5.2.5 Data Accessibility

Microsatellite genotypic data for both species and all raw data used in analyses were archived at

the Royal Ontario Museum. Specific locations of sample collection have been withheld at the

request of the Ontario Ministry of Natural Resources (OMNR); these data were archived with the

OMNR Natural Heritage Information Centre (http://nhic.mnr.gov.on.ca/).

5.3 Results

I genotyped 167 Ch. serpentina for 11 microsatellite loci (Table 5.1). No evidence of genotyping

error due to stuttering or long-allele drop-out was found. Evidence for null alleles was detected

only at locus MteD111 (Hackler et al. 2006; Table 5.2). Consequently, this locus was excluded

from all further analyses. The PI and PIsibs reached values < 0.01 with inclusion of two and six

loci, respectively. Mean pairwise distance between sampling sites was 255 km (s.d. 127.89;

range = 41 – 544). Isolation by distance was significant among sampled sites (Z = 173.435, r =

0.267, p = 0.038).

65

5.3.1 Bayesian clustering analyses

Results from TESS showed minimal decrease in DIC values from K = 2 to K = 13 (mean ∆ DIC =

17.9). The q-matrix stabilized at K = 2 (mean DIC = 7815.94, s.d. = 0.89; Figures 5.2 and 5.3)

dividing south-western Ontario from all other sampling areas. Population A included sites LE1

and LH1 (mean q = 0.97, s.d. =0.05). Population B included all other sites (mean q = 0.92, s.d. =

0.08). STRUCTURE analysis indicated that K = 2 and K = 4 were the models that best explained

the data (Figures 5.2 and 5.3). At K = 2, STRUCTURE also resolved Population A (mean q = 0.92,

s.d. = 0.03) and Population B (0.92, s.d. = 0.04). Estimated Ne of Population A was 119.63; Ne

of Population B was 195.38 individuals.

At K = 4, STRUCTURE divided each population into two subpopulations (Figure 5.3). Site LE1

(subpopulation 1; mean q = 0.88, s.d. = 0.02; Ne = 29.71) separated from LH1 (subpopulation 2;

mean q = 0.83, s.d. = 0.02; Ne = 43.24). Sites LH2, BP, GB and N (subpopulation 3; mean q =

0.72, s.d. = 0.12; Ne = 32.63) separated from Alg., Kaw., LO, EO2 and two samples from EO1

(subpopulation 4; mean q = 0.82, s.d. = 0.09; Ne = 62.82). A cline occurred between

subpopulations 3 and 4, and to the east of subpopulation 4. The cline was represented by

extensive admixture in all GH samples, 17 EO1 samples and the single sample from EO3. All

admixed individuals were genetically intermediate between subpopulations 3 and 4.

An independent STRUCTURE analysis of population B using identical methods indicated no

further sub-structure beyond subpopulations 3 and 4 (data not shown). Thus, the dataset

consisted of two distinct populations, each containing two subpopulations. This configuration

was used for all further tests, with the admixed sites GH and EO1 considered separately.

Estimated Ne of GH and EO1 was 6.46 and 22.33, respectively, but the GH estimate is not

robust due to low sample size (N = 6).

Population structure was higher in Cl. guttata than in Ch. serpentina across approximately the

same spatial scale (Figure 5.3).

5.3.2 Population differentiation

Sampling site EO1 exhibited heterozygote deficit (p = 0.002). No other deviations from HWE or

linkage equilibrium occurred. Allelic richness ranged from 3.000 to 3.440 and HO ranged from

66

0.480 to 0.650 (Table 5.2). Inbreeding was not significant within subpopulations; FIS ranged from

-0.020 to 0.010 (all 95% confidence intervals overlapped zero). Low but significant levels of

differentiation were detected between populations A and B and among all four subpopulations

(Table 5.3). Locus Cs18 was fixed for a single allele in subpopulation 3 (N = 22). Similarly,

locus 22 was fixed for a single allele in the samples from GH (N = 6).

Variation within subpopulations accounted for 95.34 % of the variation in the dataset (Table 5.4,

AMOVA: ΦST = 0.047, p = 0.000). Significant variation was also detected among subpopulations

within the two populations (ΦSC = 0.028, p = 0.000) and between the populations (ΦCT = 0.019,

p = 0.014).

The first two principal coordinates of the PCoA accounted for 95.42% of variation among

subpopulations. The first component (PCo1) divided LE1 and LH1 from all other sites (Figure

5.4a), while PCo2 divided LH1 from the other distinct subpopulations (Figure 5.4b). At the level

of individuals, several samples from different populations overlapped in principal coordinate

space but a low level of structure was apparent (Figure 5.4c).

5.3.3 Interspecific comparison

There was no difference in Ne (Z = 0.944; p = 0.345), Ar (Z = -0.944, p = 0.345) or PAr (Z = -

1.483, p = 0.138) between Cl. guttata and Ch. serpentina at the paired sites, but Clemmys guttata

had significantly higher HO (Z = -2.023, p = 0.043) and HE (Z = -2.023, p = 0.043) than Ch.

serpentina (Figure 5.5).

Values of Ne, HO and HE across all sampled sites met assumptions of normality. Unpaired tests

of values from all sites also showed that Ne was not different between species (t = 1.265, d.f. =

17, p = 0.223) and that HO and HE were significantly higher in Cl. guttata than Ch. serpentina

(HO: t = 3.937, df = 17, p = 0.001; HE: t = 3.791, df = 17, p = 0.001).

5.4 Discussion

A comparative approach to population genetics can identify or rule out potential drivers of

diversity across a landscape. For example, Howes et al. (2009) show that genetic variation is

comparable in populations of the long-lived Blanding’s turtle (Emydoidea blandingii) and the

67

black rat snake (Pantherophis obsoleta), which has a much shorter generation time. However,

these two species were sampled across two distinct landscapes introducing potential confounding

landscape effects. In this study, I compared two equally long-lived species across a single

landscape to test the hypothesis that a common, relatively vagile, abundant species with high

fecundity (the snapping turtle: Chelydra serpentina) should exhibit higher genetic diversity than

an endangered, less vagile, rare species with low fecundity (the spotted turtle: Clemmys guttata).

As predicted, population structure across the study landscape was higher in Cl. guttata than Ch.

serpentina. On the other hand, heterozygosity was higher in Cl. guttata than in Ch. seprentina.

Quite unexpectedly, estimates of Ne were comparable between the two species, which is

surprising given the well-documented disparity in abundance and therefore population size

between them. Heterozygosity is related to Ne. The unexpected patterns of heterozygosity might,

therefore, be best explained by considering factors other than longevity that can affect the Ne:N

ratio (summarized by Charlesworth 2009). In the following paragraphs I discuss two possible

explanations for the study results: 1) differential variance in reproductive success resulting from

different nest success, mating systems and mate choice behaviour, and 2) stochastic effects

during post-glacial colonization events.

High variance in reproductive success among individual males, females, or both sexes causes Ne

to decrease (Hedrick 2000; Karl 2008; Galbraith 2008; Charlesworth 2009). Reproductive

success has not been robustly quantified in any population of freshwater turtle. Nevertheless,

behavioural observations may provide evidence for differential variance between Ch. serpentina

and Cl. guttata. Male Ch. serpentina at some sites may defend territories that provide access to

females and fight other males who enter their territory (Galbraith et al. 1987b). Male Ch.

serpentina are also thought to coerce the smaller females to mate (Berry & Shine 1980) but this

assumption remains to be verified. These strategies may maximize the success of dominant

individual males. However, they may also increase average variance in male reproductive

success by effectively removing less dominant males from the breeding pool. In contrast, Cl.

guttata aggregate to breed after emerging from hibernation. Although males chase and

occasionally bite females they are pursuing, territorial behaviour has not been observed and

aggregations may contain several individuals of both sexes (Ernst & Lovich 2009, Liu et al. in

review). Therefore, breeding aggregations of Cl. guttata may serve to increase the frequency of

68

mate encounters and multiple mating, and to decrease the overall variance in reproductive

success of both males and females.

Mating systems and patterns of paternity also affect Ne (Sugg and Chesser 1994; Karl 2008).

Turtles exhibit promiscuous mating systems (polygynandry) and often exhibit multiple paternity

(Galbraith et al. 1993; Uller & Olsson 2008; Davy et al. 2011). Multiple mating increases Ne

relative to monogamy, but multiple paternities in single clutches reduce Ne unless the paternal

contributions are equal (Zbinden et al. 2007; Karl 2008). In multiply-sired small clutches such as

those of Cl. guttata variance in paternal contribution is likely lower than in large clutches (> 20

eggs) such as those laid by Ch. serpentina.

Some mate choice behaviours can increase heterozygosity, and can similarly impact Ne:N ratios

by affecting average variance in reproductive success. For example, inbreeding avoidance is

well-documented in a number of species (Pusey 1987; Johnson et al. 2010; Dunn et al. 2012;

Varian-Ramos & Webster 2012). Inbreeding avoidance serves to minimize the relatedness of an

offspring’s parents, typically maximizing heterozygosity of offspring and maintaining Ne above

the levels expected with inbreeding. Heterozygosity sometimes correlates with factors such as

survivorship, immunity, or reproductive success (Frankham et al. 2002). Thus, inbreeding

avoidance may also maximize offspring fitness (Foerster et al. 2003; Fossøy et al. 2008; but see

Balloux et al. 2004).

Little is known about mate choice in either Cl. guttata or Ch. serpentina. However, my data

suggest mate choice in Cl. guttata is not random because it is unlikely that multiple populations

containing < 50 individuals can randomly maintain high heterozygosity. Thus, it is possible that

aspects of the mating system in Cl. guttata such as possible inbreeding avoidance might buffer

genetic diversity, at least for awhile. This buffering effect, when present, is a boon for

conservation strategies because it might “buy us more time” in the race to rescue an endangered

species. On the other hand, if increased reproductive variance in Ch. serpentina may be

depressing Ne within populations, then the opposite is true and populations may be affected

more strongly by declines than is currently recognized. Further study incorporating direct

observation, genetic profiling of adults and paternity testing of hatchlings in one or more

populations can serve to test for potential inbreeding avoidance in Cl. guttata. A parallel study of

69

Ch. serpentina or other sympatric species would allow direct comparison of the average

relatedness of mating pairs among species.

Differences in nesting behaviour may also affect reproductive success. Chelydra serpentina

produces on average 10 times as many eggs per clutch as Cl. guttata, but predation of Ch.

serpentina nests exceeds 90% at some sites, as I have observed at sites LE1 and LH1. This

increases variance in reproductive success among individuals and family groups, both of which

reduce Ne relative to N (Karl 2008; Charlesworth 2009). It is possible that the less obvious nests

of Cl. guttata have higher average survivorship despite lower fecundity, thus reducing variance

in reproductive success in Cl. guttata relative to Ch. serpentina. Direct evidence would be

required to test this hypothesis; in particular, data from multiple sites that account for inter-site

variation in nest success.

Stochastic or unknown historic factors can also cause unexpected patterns of diversity among

populations. For example, Cl. guttata likely colonized Ontario from two or more independent

refugia after the retreat of the Laurentide Glacier (Chapter 3). If Ch. serpentina colonizing

Ontario came from a single refugium, but Cl. guttata came from multiple refugia, then the higher

genetic diversity in populations of Cl. guttata may reflect this different history. Fossil evidence,

however, suggests this is not the case because Ch. serpentina appears to be one of the first

species to re-enter Ontario after the end of the last ice age, and fossil Ch. serpentina are known

from a greater number of Holocene locations than Cl. guttata (Holman 1992; Holman &

Andrews 1994). Thus, Ch. serpentina appears to have also spent glacial Pleistocene periods in

multiple refugia. Additionally, the pattern of higher genetic diversity in Cl. guttata is consistent

across multiple sampling sites and populations. Stochastic effects would be more likely to affect

a single population than to cause a consistent trend across > 500 km. Therefore, I consider

factors involved in differential reproductive success to be the more likely explanation for the

unexpectedly high levels of heterozygosity in Cl. guttata.

This is the first study to detect genetic population structure in Ch. serpentina. A range-wide

study of Ch. serpentina mitochondrial DNA (Phillips et al. 1996) failed to detect population

structure and extremely low mtDNA variation was reported across the southwestern portion of

the range (Walker et al. 1998). Similarly, no structure was found based on microsatellite

70

genotypes across a small geographic scale in Illinois (60 km; Banning-Anthonysamy 2012). The

discovery of genetic structure across several hundred kilometres in Ontario was thus unexpected,

but no previous studies of Ch. serpentina have employed microsatellite markers across a broad

geographic scale.

Galbraith (2008) predicted that northern populations of Ch. serpentina should have lower genetic

diversity and structure than southern populations. This prediction was based on geographic

variance in clutch size of Ch. serpentina, which is correlated with latitude and predicts higher

reproductive variance in northern females. Founder effects during post-glacial colonization

would also predict lower diversity in northern populations relative to southern populations

(Galbraith, 2008). My results show that variation in genetic diversity in Ch. serpentina is

sufficient that microsatellite genotyping of samples collected across a broad geographic scale can

be used to test these predictions.

5.4.1 Summary

Biologists often incorporate information about factors such as longevity, vagility, fecundity and

population size into explanations of why levels of genetic diversity vary among species (e.g.

Frankham 1996, Howes et al. 2009, Pittman et al. 2011). This study indicates that behavioural

differences that impact reproductive success may have a greater impact on genetic diversity than

these traditionally considered factors, and should be explicitly considered in our analyses (see

also Gregory et al. 2012). Because of the complicated interactions among all of these factors, it

should not be assumed a priori that small populations of endangered species will be genetically

depauperate compared to abundant species with large populations. Similarly, it cannot be

assumed a priori that genetic diversity in widespread and relatively abundant species will

necessarily be high. In this study, genetic diversity in the common Ch. serpentina at the northern

edge of its range was unexpectedly low, possibly due to a combination of reduced reproductive

variance based on the mating system coupled with high levels of nest predation. Overharvesting

across the range of Ch. serpentina is likely causing significant population declines, although the

data necessary for robust evaluations of population size or trends are not available (van Dijk,

2011). In a world in which conservation decisions are often predicated upon estimates of

population size and species rarity, species such as Ch. serpentina are rarely prioritized for

conservation measures. This study raises the possibility that anthropogenic change might have a

71

more dramatic effect than predicted on some widespread species because they have low genetic

diversity despite being abundant. In such species we are “losing time” and are not even aware of

it. I thus recommend that future population genetic studies be coupled with further studies on

behaviour and ecology of the study species to build a more robust framework on which to base

conservation decisions.

5.5 Acknowledgments

This research was generously supported a Canada Collection grant to CMD from Wildlife

Preservation Canada and by the National Science and Engineering Research Council of Ontario

(NSERC Discovery Grant to RWM; Canada Graduate Scholarship to CMD). Thanks to S.

Coombes and a large number of volunteers for assistance with field work. Site access and

logistical support were provided by M. Cairns, J. Urquhart, the Ausable Bayfield Conservation

Authority, Ontario Parks, Ontario Nature and Parks Canada. Genotyping costs were offset by the

generous support of the Schad Foundation. Research methods were approved under animal use

protocols ROM2008-11, 2009-02, 2009-21 and 2010-14) from the Animal Care Committee of

the Royal Ontario Museum, under permits1045769, 1049600, 1062210, 1067079, SR-B-001-10

and AY-B-013-11 from the Ontario Ministry of Natural Resources and under research

authorizations from Ontario Parks and Parks Canada. D. McLennan, R. Murphy, J. Miller and L.

Einarson provided valuable comments on earlier versions of the manuscript.


conservation. Chelonian Research Monographs, 4, 107–123.

Balloux F, Amos W, Coulson T (2004) Does heterozygosity estimate inbreeding in real populations? Molecular Ecology, 13, 3021–3031.


Berry JF, Shine R (1980) Sexual size dimorphism and sexual selection in turtles (Order: Testudines). Oecologia, 44, 185–191.

Brookfield JFY (1996) A simple new method for estimating null allele frequency from heterozygote deficiency. Molecular Ecology, 5, 453–455.

72

Charlesworth B (2009) Effective population size and patterns of molecular evolution and variation. Nature Reviews Genetics, 10, 195–205.

Chen C, Durand E, Forbes F, François O (2007) Bayesian clustering algorithms ascertaining spatial population structure: a new computer program and a comparison study. Molecular Ecology Notes, 7, 747–756.

COSEWIC (2004) COSEWIC assessment and update status report on the spotted turtle Clemmys guttata in Canada. Committee on the status of endangered wildlife in Canada. Ottawa. vi + 27 pp. (www.sararegistry.gc.ca/status/status_e.cfm).

COSEWIC (2008) COSEWIC assessment and status report on the Snapping Turtle Chelydra serpentina in Canada. Committee on the status of endangered wildlife in Canada. Ottawa. vii + 47 pp. (www.sararegistry.gc.ca/status/status_e.cfm).

Crawford NG (2010) SMOGD: software for the measurement of genetic diversity. Molecular Ecology Resources, 10, 556-557.

Davy CM, Edwards T, Lathrop A, Bratton M, Hagan M, Nagy K, Stone J, Hillard LS, Murphy RW (2011) Polyandry and multiple paternities in the threatened Agassiz’s desert tortoise, Gopherus agassizii: conservation implications. Conservation Genetics, 12, 1313–1322.

Davy CM, Leifso AE, Conflitti IM, Murphy RW (2012) Characterization of 10 novel microsatellite loci and cross-amplification of two loci in the snapping turtle (Chelydra serpentina). Conservation Genetics Resources, 4,695–698.

Dunn SJ, Clancey E, Waits LP, Byers JA (2012) Genetic evidence of inbreeding avoidance in pronghorn. Journal of Zoology, 288, 119–126.

Ernst CH, Lovich JL (2009) Turtles of the United States and Canada, 2nd ed, Johns Hopkins University Press, Baltimore, Maryland.

Evanno G, Regnaut S, Goudet J, (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology 14, 2611–2620.

Ewing SR, Nager RG, Nicoll MAC, Aumjaud A, Jones CG, Keller LF (2008) Inbreeding and loss of genetic variation in a reintroduced population of Mauritius Kestrel. Conservation Biology, 22, 395–404.

Excoffier L, Laval G, Schneider S (2005) Arlequin ver. 3.0: An integrated software package for population genetics data analysis. Evolutionary Bioinformatics Online, 1, 47–50.

Fisher RA (1930) The genetical theory of natural selection. Clarendon Press

Foerster K, Delhey K, Johnsen A, Lifjeld JT, Kempenaers B (2003) Females increase offspring heterozygosity and fitness through extra-pair matings. Nature, 425, 714–717.

Fossøy F, Johnsen A, Lifjeld JT (2008) Multiple genetic benefits of female promiscuity in a socially monogamous passerine. Evolution, 62, 145–156.

Frankham R (1995a) Conservation genetics. Annual Review of Genetics, 29, 305–327.

Frankham R (1995b) Effective population size/adult population size ratios in wildlife: a review Genetical Research, 66, 95–107.

73

Frankham R (1996) Relationship of genetic variation to population size in wildlife. Conservation Biology, 10, 1500–1508.

Frankham R, Ballou JD, Briscoe DA (2002) Introduction to conservation genetics. Cambridge University Press, Cambridge, U.K.

Galbraith DA, Brooks RJ (1987a) Survivorship of adult females in a northern population of common snapping turtles, Chelydra serpentina. Canadian Journal of Zoology, 65, 1581–1586.

Galbraith DA, Chandler MW, Brooks RJ (1987b) The fine structure of home ranges of male Chelydra serpentina: are snapping turtles territorial? Canadian Journal of Zoology, 65: 2623–2629.

Galbraith DA, Brooks RJ, Obbard ME (1989) The influence of growth rate on age and body size at maturity in female snapping turtles Chelydra serpentina. Copeia, 1989, 896–904.

Galbraith DA, White BN, Brooks RJ, Boag PT (1993) Multiple paternity in clutches of snappng turtles (Chelydra serpentina) detected using DNA fingerprints. Canadian Journal of Zoology, 71, 318–324.

Galbraith DA (2008) Population biology and population genetics. In: The Biology of the Snapping Turtle (eds Steyermark AC, Finkler MS and Brooks RJ), pp 168–180. Johns Hopkins University Press, Baltimore, Maryland.

Gregory AJ, Kaler RSA, Prebyl TJ, Sandercock BK, Wisely SM (2012) Influence of translocation strategy and mating system on the genetic structure of a newly established population of island ptarmigan. Conservation Genetics, 13, 465–474.

Hackler JC, van den Bussche RAV, Leslie DM (2006) Characterization of microsatellite DNA markers for the alligator snapping turtle, Macrochelys temminckii. Molecular Ecology Notes, 7, 474–476.

Hedrick PW (2000) Genetics of Populations, 2nd edn, pp. 244–255, Jones & Bartlett Publishers, Sudburg, MA.

Holman JA (1992) Late Quaternary herpetofauna of the central Great Lakes region, U.S.A.: zoogeographical and paleoecological implications. Quaternary Science Review, 11, 345–351.

Holman JA, Andrews KD (1994) North American Quaternary cold-tolerant turtles: distributional adaptations and constraints. Boreas, 23, 44–52.

Howes BJ, Brown JW, Gibbs HL, Herman TB, Mockford SW, Prior KA, Weatherhead PJ (2009) Directional gene flow patterns in disjunct populations of the black ratsnake (Pantheropis obsoletus) and the Blanding’s turtle (Emydoidea blandingii). Conservation Genetics, 10, 407–417.

Jensen JL, Bohonak AJ, Kelley ST (2005) Isolation by distance, web service. BMC Genetics, 6, 13. v.3.23 http://ibdws.sdsu.edu/

Johnson AM, Chappell G, Price AC, Rodd HF, Olendorf R, Hughes KA (2010) Inbreeding Depression and Inbreeding Avoidance in a Natural Population of Guppies (Poecilia reticulata). Ethology, 116, 448–457.

74

Jost L (2008) GST and its relatives do not measure differentiation. Molecular Ecology, 17, 4015–4026.

Kalinowski ST (2004) Counting alleles with rarefaction: private alleles and hierarchical sampling designs. Conservation Genetics, 5, 539–543.

Kalinowski ST (2005) HP-Rare: A computer program for performing rarefaction on measures of allelic diversity. Molecular Ecology Notes, 5, 187–189.

Karl SA (2008) The effect of multiple paternity on the genetically effective size of a population Molecular Ecology, 17, 3973–3977.

Kuo CH, Janzen FJ (2004) Genetic effects of a persistent bottleneck on a natural population of ornate box turtles (Terrapene ornata). Conservation Genetics, 5, 425–437.

Leimu R, Mutikainen R, Koricheva J, Fischer M (2006) How general are positive relationships between plant population size, fitness and genetic variation? Journal of Ecology, 94, 942–952.

Litzgus JD (1996) Life history and demography of a northern population of spotted turtles, Clemmys guttata. MSc Thesis, University of Guelph, Ontario. 145 pp.

Litzgus JD (2006) Sex differences in longevity in the spotted turtle (Clemmys guttata). Copeia, 2006, 281-288.

Liu Y, Davy CM, Shi HT, Murphy RW (In review) Sex in the half-shell: a review of the history, signal-function, and evolution of courtship behavior in freshwater turtles. In review, Chelonian Conservation and Biology.

Marsack K, Swanson BJ (2009) A Genetic Analysis of the Impact of Generation Time and Road-Based Habitat Fragmentation on Eastern Box Turtles (Terrapene c. carolina). Copeia, 4, 647–652.

Mitton JB (1997) Selection in natural populations. 240 pp. Oxford University Press: Oxford.

Nei M, Maruyama T, Chakraborty R (1975) Bottleneck effect and genetic variability in populations. Evolution, 29, 1–10.

Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6, 288–295.

Phillips CA, Dimmick WW, Carr JL (1996) Conservation genetics of the common snapping turtle (Chelydra serpentina). Conservation Biology, 10, 397–405.

Pittman SE, King T, Faurby S, Dorcas ME (2011) Genetic and demographic status of an isolated bog turtle (Glyptemys muhlenbergii) population: implications for the conservation of small populations of long-lived animals. Conservation Genetics, 12, 1589–1601.

Pompanon F, Bonin A, Bellemain E, Taberlet P (2005) Genotyping errors: causes, consequences and solutions. Nature Reviews Genetics, 6, 847–846.

Pritchard JK, Stephens M, Donnelly PJ (2000) Inference of population structure using multilocus genotype data. Genetics, 155, 945–95

Pusey AE (1987) Sex-biased dispersal and inbreeding avoidance in birds and mammals. Trends in Ecology and Evolution, 2, 295–299.

75

Raymond M, Rousset F (1995) GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. Journal of Heredity, 86, 248–249.

Rice WR (1989) Analyzing tables of statistical tests. Evolution, 43, 223–225.

Rousset F (2008) Genepop’007: a complete reimplementation of the Genepop software for Windows and Linux. Molecular Ecology Resources, 8, 103–106.

Sambrook J, Fritsch EF, Maniatis T (1989) Molecular cloning—a laboratory manual. 2nd edn. Cold Spring Harbor Laboratory Press, New York.

Soulé ME (1976) Allozyme variation, its determinants in space and time. In: Molecular evolution (ed Ayala FJ), pp. 60–88. Sinauer Associates, Sunderland, Massacheusetts.

Sugg DW, Chesser RK (1994) Effective population size with multiple paternity. Genetics, 137, 1147–1155.

Tallmon DA, Koyuk A, Luikart GH, Beaumont MA (2008) ONeSAMP: a program to estimate effective population size using approximate Bayesian computation. Molecular Ecology Resources, 8, 299–301.

Uller T, Olsson M (2008) Multiple paternity in reptiles: patterns and processes. Molecular Ecology, 17, 2566–80.

van Dijk PP (2011) Chelydra serpentina. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.1. <www.iucnredlist.org>. Downloaded on 24 September 2012.

van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes, 4, 535–538.

Vargas-Ramirez M, Stuckas H, Castaňo-Mora OV, Fritz U (2012) Extremely low genetic diversity and weak population differentiation in the endangered Colombian river turtle Podocnemis lewyana (Testudines: Podocnemididae) Conservation genetics, 13, 65–77.

Varian-Ramos CW, Webster MS (2012) Extrapair copulations reduce inbreeding for female red-backed fairy-wrens, Malurus melanocephalus. Animal Behaviour, 83, 857–864.

Walker D, Moler PE, Buhlmann KA, Avise JC (1998) Phylogeographic uniformity in mitochondrial DNA of the snapping turtle (Chelydra serpentina). Animal Conservation, 1, 55–60.

Wright S (1943) Isolation by distance. Genetics, 28, 114–138.

Zbinden JA, Largiader CR, Leippert F, Margaritoulis D, Arlettaz R (2007) High frequency of multiple paternity in the largest rookery of Mediterranean loggerhead sea turtles. Molecular Ecology, 16, 3703–3711.

76

Table 5.1. Summary statistics for 11 microsatellite loci (Hackler et al. 2007; Davy et al. 2012)

amplified in 167 Chelydra serpentina from southern Ontario. N = number of individuals

successfully amplified at each locus; k = number of alleles; Ne = number of effective alleles; HO

= observed heterozygosity; HE = expected heterozygosity; PI = probability of identity; PIsibs =

Probability of identity for siblings at a locus. Locus MteD111 showed evidence of potential null

alleles and was excluded from all multi-locus analyses.

Locus N k Ne HO HE PI PIsibs

Cs08 165 11 7.167 0.861 0.860 0.035 0.328

Cs12 159 12 5.226 0.818 0.809 0.054 0.359

Cs16 164 4 3.260 0.622 0.693 0.152 0.441

Cs17 165 5 2.998 0.630 0.666 0.159 0.457

Cs18 163 3 1.173 0.141 0.147 0.737 0.861

Cs19 167 5 2.127 0.503 0.530 0.263 0.551

Cs22 159 4 1.533 0.333 0.348 0.444 0.687

Cs24 163 3 2.298 0.521 0.565 0.262 0.533

Cs25 161 3 2.290 0.578 0.563 0.288 0.540

MteD9 152 6 4.162 0.737 0.760 0.094 0.318

MteD111 151 14 4.324 0.561 0.701 0.027 0.394

77

Table 5.2. Genetic diversity in 167 Chelydra serpentina sampled across southern Ontario based on 10 microsatellite loci. Populations

(Pop) and subpopulations (SP) were identified with Bayesian clustering analyses (see text for details). GH = Golden Horseshoe, EO1

= Eastern Ontario 1. Number of alleles (private alleles in parentheses); HO and HE: observed and expected heterozygosities; N: sample

size per tested unit. Estimated frequency of a null allele was calculated for each locus across all populations. Summary statistics are

presented for locus MteD111, but this locus was excluded from calculations of mean heterozygosity and allelic richness.

POPA SP 1 SP 2 PopB SP 3 SP 4 GH EO1

Cs08 Number of alleles 10 8 9 11 9 11 6 8 Estimated null allele HO 0.855 0.852 0.857 0.864 0.826 0.881 0.875 0.850

frequency: 0.00 HE 0.845 0.821 0.844 0.853 0.813 0.852 0.797 0.806

N 55 27 28 110 23 59 8 20

Cs12 Number of alleles 12 10 11 9 8 9 6 7

Estimated null allele HO 0.852 0.815 0.889 0.800 0.773 0.821 0.714 0.800

frequency: 0.00 HE 0.809 0.765 0.837 0.799 0.769 0.800 0.694 0.798

N 54 27 27 105 22 56 7 20



frequency: 0.00 HE 0.593 0.559 0.610 0.716 0.598 0.745 0.704 0.656

N 55 27 28 109 24 58 7 20



frequency: 0.00 HE 0.743 0.709 0.744 0.605 0.518 0.632 0.367 0.648

N 55 27 28 110 24 59 7 20



frequency: 0.00 HE 0.150 0.252 0.035 0.146 0.000 0.128 0.227 0.289

N 55 27 28 108 22 58 8 20

78



frequency: 0.00 HE 0.218 0.205 0.228 0.624 0.669 0.604 0.625 0.544

N 55 27 28 112 24 60 8 20



frequency: 0.00 HE 0.276 0.237 0.309 0.383 0.598 0.258 0.000 0.480

N 55 27 28 104 22 56 6 20



frequency: 0.00 HE 0.616 0.612 0.605 0.532 0.455 0.542 0.357 0.584

N 55 27 28 108 23 58 7 20



frequency: 0.00 HE 0.521 0.431 0.536 0.581 0.590 0.561 0.500 0.584

N 55 27 28 106 22 57 7 20

MteD9 Number of alleles 6 6 5 6 6 6 5 6


frequency: 0.00 HE 0.740 0.735 0.581 0.767 0.654 0.771 0.750 0.798

N 55 27 28 97 21 52 4 20

MteD111* Number of alleles 11 (1) 8 7 (1) 13 (3) 9 12 (2) 5 9


frequency: 0.3017 HE 0.846 0.787 0.807 0.881 0.822 0.865 0.667 0.783

N 53 27 23 98 23 50 6 19

Mean HO 0.547 0.541 0.553 0.588 0.577 0.588 0.480 0.650

Mean HE 0.551 0.533 0.533 0.601 0.566 0.601 0.502 0.619

Allelic richness 4.63 3.86 3.61 4.92 3.63 3.91 4.16 3.96

Private allelic richness 0.63 0.18 0.27 0.91 0.35 0.22 0.38 0.28

* MteD111 was excluded from all analyses, including mean heterozygosity and allelic richness presented in this table, due to possible presence of null alleles.

79

Table 5.3. Population differentiation (Dest above the diagonal, FST below) for four

subpopulations and two admixed groups of Chelydra serpentina identified by STRUCTURE

analysis (Figure 3). FST values in bold are significant (p < 0.05). Subpopulations (SP) are

described in the text. GH = Golden Horseshoe. EO = EO1 and EO3.

SP 1 SP 2 SP 3 SP 4 GH EO

Subpopulation 1 0 0.013 0.043 0.016 0.025 0.032

Subpopulation 2 0.024 0 0.029 0.023 0.005 0.002

Subpopulation 3 0.097 0.064 0 0.006 0.006 0.006

Subpopulation 4 0.045 0.034 0.036 0 0.001 0.000

GH 0.035 0.042 0.010 -0.004 0 0.000

EO 0.042 0.030 0.014 0.005 0.001 0

Table 5.4. Hierarchical partitioning of molecular variance in Chelydra serpentina from southern

Ontario with AMOVA (Excoffier et al. 1992). All sources of variation were significant (p <

0.02).

Source of variation Sum of squares Variance components (σ2)

% variation

Among populations 16.75 0.058 1.92

Among subpopulations within populations

26.91 0.083 2.73

Within subpopulations 916.55 2.890 95.34

Total 960.22 3.031

80

Figure 5.1. Sampling sites for 167 Chelydra serpentina (blue squares) sampled in this study and

256 Clemmys guttata (yellow squares) sampled in Chapter 3. Bi-coloured squares indicate sites

where both species were sampled. Insert shows pairs of sampling areas used for comparisons of

genetic diversity between species. LE1 = Lake Erie 1; LE2 = Lake Erie 2; LH1 = Lake Huron 1;

LH 2 = Lake Huron 2; BP = Bruce Peninsula; GB = Georgian Bay; GH = Golden Horseshoe; N

= North of Golden Horseshoe; KAW = Kawartha Lakes area; ALG = Algonquin Provincial Park;

HC = Hastings County; LO = north-east shore of Lake Ontario; EO = Eastern Ontario. Base map

modified from http://www.aquarius.geomar.de/omc/make_map.html and used under the GNU

Free Documentation license.

81

Figure 5.2. Results of Bayesian clustering analyses of Ch. serpentina for increasing values of K,

the number of genetically distinct populations represented in the sample following analyses

described in Methods. Structure results for K = 1 – 8 : A) Log likelihood (L(K)) of the data

(mean ± standard deviation); B) ∆K following Evanno et al. (2005). TESS results for Kmax = 2–

14; C) Deviance information criterion (mean ± standard deviation) following analyses described

in Methods.

82

Figure 5.3. Results of Bayesian clustering analyses of Ch. serpentina for a range of models with

increasing values of K inferred using STRUCTURE and TESS. Models shown here are those that

best fit the data based on criteria described in Methods. Population structure in Cl. guttata across

the same landscape is shown for comparison (from Chapter 3). Colours used for subpopulations

in the K = 4 model are consistent with colours used in Figure 4.

83

84

Figure 5.4. Principle component analysis of genetic distance for 167 Ch. serpentina based on 10

microsatellite loci. A and B: PCoA of populations based on Dest ; C: PCoA based on genetic

distance among individuals labelled by sampling site.

Figure 5.5. Heterozygosity and effective population sizes of Ch. serpentina and Cl. guttata

compared across five pairs of sites (Figure 1, inset). HO: observed heterozygosity. HE: expected

heterozygosity. Ne: effective population size. Error bars show standard deviation of HO and HE

and 95% confidence intervals of Ne estimates.

85

Chapter 6 Conservation genetics of Blanding’s turtle (Emys blandingii) in

Ontario, Canada.

Formatted for Conservation Genetics.

6 Abstract

Blanding’s turtle, Emys blandingii, is a globally endangered species with a range centred on the

Great Lakes. Several disjunct populations occur along the East Coast of North America. Previous

studies suggest that gene flow may be uninterrupted in the Great Lakes portion of the range.

However, E. blandingii is restricted to relatively small populations across its range and,

therefore, panmixia across large geographic distances is unlikely. Here, Bayesian analyses of

population structure among samples collected across southern Ontario (N = 97) rejected a null

hypothesis of panmixia. These data were used to identify potential management units. Ontario

contains four distinct genetic clusters of E. blandingii and these should be considered as

independent management units. Preliminary evidence suggests that further structure may be

present in some poorly sampled areas, and these deserve further consideration. Genetic diversity

at sampled sites is comparable to that reported for other freshwater turtles. Comparison between

this study and previous work confirms reduced genetic diversity in disjunct eastern populations

compared to populations centred on the Great Lakes. Genetic diversity in E. blandingii is not

correlated with latitude, which may indicate post-glacial dispersal of this species from multiple

Pleistocene glacial refugia.

Keywords: population structure, STRUCTURE, TESS, GENECLASS, heterozygosity

86

6.1 Introduction

Blanding’s turtle, Emys (=Emydoidea) blandingii is a moderately sized freshwater species found

in the northeastern United States and southern Canada. The main portion of its range is centred

on the Great Lakes region. Disjunct populations occur in New York, Massachusetts, and Nova

Scotia (Figure 1; Ernst and Lovich 2009). Mean age of maturity in a well-studied Michigan

population is 17.5 years, generation time is approximately 37 years, and longevity exceeds 75

years (Congdon and van Loben Sels 1991; Congdon et al. 1993; Brecke and Moriarty 1989). One

consequence of this life history is that populations are sensitive to any increase in the mortality

rate of reproductive adults (Congdon et al. 1993). Even a small increase in adult mortality can

cause significant population declines. A number of factors including road mortality, illegal

collection, and habitat degradation are currently causing such declines. Therefore, E. blandingii

was recently up-listed from Least Concern to Endangered by the International Union for

Conservation of Nature (IUCN; van Dijk and Rhodin 2011).

Using random amplified polymorphic DNA (RAPD) markers and microsatellites, Mockford et

al. (1999; 2005; 2007) and Rubin et al. (2001) quantified genetic variation in E. blandingii across

the species’ range. Band-sharing analyses of RAPD data showed that the disjunct Nova Scotian

population differed genetically from central populations (Mockford et al. 1999; Rubin et al.

2001). Within Nova Scotia, analyses of microsatellite data based on FST values suggested

significant differentiation among three subpopulations despite separation by < 30 km. However,

very little population structure was detected in the main portion of the range based on samples

from Minnesota, Wisconsin, Illinois, Michigan, and Ontario (Mockford et al. 2005). Based on

these data, Mockford et al. (2007) proposed that E. blandingii comprised three Evolutionarily

Significant Units (ESU): 1) the Nova Scotian population; 2) isolated populations in

Massachusetts and New York; and 3) populations extending from the Great Lakes. The ESU

concept does not apply to the legal conservation of turtles in either the USA or Canada, where

protection is based on the concepts of distinctive population segments (Pennock and Dimmick

2002) and designatable units (DUs; Green 2005), respectively. In Canada, the Committee on the

Status of Endangered Wildlife in Canada (COSEWIC) recognizes E. blandingii in Nova Scotia

and the populations around the Great Lakes as two DUs.

87

Recent studies highlight concerns with analyses based on FST. For example, Jost (2008)

demonstrated that FST and related measures of diversity do not necessarily measure actual

population differentiation. He proposed an alternative, more accurate measure (Dest). Jost (2008)

also pointed out that statistical significance of FST was primarily a factor of sample size and may

be biologically meaningless. Howes et al. (2009) summarized further concerns with analyses of

population structure based on FST, including that assumptions of these analyses often were not

met in natural populations (Whitlock and McCaughley 1999) and that FST does not reflect

contemporary gene flow (Paetkau et al. 2004). FST and related measurements can provide

information about historical migration rates, but they are not appropriate measures of population

differentiation or structure (Jost 2009).

Bayesian methods for detection of population structure and connectivity (e.g. Pritchard et al.

2000; Chen et al. 2007) do not rely on the assumptions of FST-based analyses. Howes et al.

(2009) applied Bayesian methods to the data of Mockford et al. (2005) to study population

connectivity in the three subpopulations of E. blandingii in Nova Scotia. The results

demonstrated moderate historic and current gene flow among all three subpopulations, and

clustered the two nearest subpopulations together indicating that they were genetically

continuous. Bayesian analysis of three E. blandingii populations separated by <60km in Illinois

showed no evidence of population structure (Banning-Anthonysamy 2012). These results are

congruent with the biology of Blanding’s turtle. Emys blandingii is extremely vagile, and

although it is dependent on wetland habitats, individuals may travel > 10 km overland (Power

1989). Thus, population structure in this species is more likely to occur on a relatively large

geographic scale (> 100 km).

Ontario has a large portion of the core range of E. blandingii, but previous studies included only

11 samples from one site in southeastern Ontario, St. Lawrence Islands National Park (Mockford

et al. 2007). Presence-absence data show that the distribution of E. blandingii in Ontario is not

continuous (Ontario Nature Herpetofaunal Atlas,

http://www.ontarionature.org/protect/species/reptiles_and_amphibians/map_blandings_turtle.ht

ml). Gaps in occurrence records may reflect historic or current barriers to gene flow and suggest

that populations may not be panmictic across the province. Nevertheless, COSEWIC currently

88

considers the “Great Lakes/St. Lawrence population,” comprised of all E. blandingii in Ontario

and Quebec, as a single unit for management and recovery purposes (COSEWIC 2005).

Here, I use three Bayesian analyses and a principal coordinates analysis to investigate population

structure in E. blandingii across > 500 km in southern Ontario. I investigate the level of

population structure and genetic diversity present at sampled sites to test the hypothesis that

populations of E. blandingii around the Great Lakes show little differentiation and consist of a

single genetic unit. Further, I compare genetic variation (heterozygosity) among populations in

Ontario and the populations studied by Mockford et al. (2007) to test two hypotheses: a) that

variation will be lower in disjunct eastern populations than in populations around the Great

Lakes, as suggested by Mockford et al. (2005); and b) that variation will decrease with proximity

to the northern limit of the species’ range.

6.2 Methods

I collected DNA from E. blandingii across southern Ontario between 2008 and 2011, and

additional samples were contributed by other researchers and government biologists (Figure 6.1).

Turtles were captured by hand or in hoop traps baited with sardines at sites LE, GH, EO, ALG.,

LHsouth, and LHnorth. Blood was taken by caudal venipuncture with a sterile syringe and

blotted onto FTA cards (Whatman Inc., Clifton, NJ) for storage. All individuals were released at

their initial capture site. Blood was extracted from FTA cards following Smith and Burgoyne

(2004). At site PSD, muscle samples were collected from road-killed individuals. At KAW,

blood samples were taken from turtles injured on local highways and rehabilitated at the

Kawartha Turtle Trauma Centre (Peterborough, Ontario); these blood samples were stored in

heparin before analysis. Extraction of DNA from muscle and heparinized blood followed the

phenol-chloroform procedure of Sambrook et al. (1999) and extracted DNA was cleaned with

EtOH precipitation. Four additional blood samples were collected from captive E. blandingii at

Scales Nature Park (Orillia, Ontario) that were from Ontario, but whose exact locations of origin

were unknown.

Samples were amplified at four microsatellite loci developed for E. blandingii (Eb09, Eb11,

Eb17 and Eb19; Osentoski et al. 2002). These loci were used by Mockford et al. (2007) and,

therefore, allowed for some direct comparison of diversity between the two studies. In addition, I

89

amplified 13 loci from Glyptemys muhlenbergii that cross-amplified in E. blandingii (GmuB08,

GmuD16, GmuD21, GmuD28, GmuD55, GmuD70, GmuD87, GmuD88, GmuD89, GmuD90,

GmuD93, GmuD107 and GmuD121; King and Julian 2004). Amplification and allele scoring

followed Chapter 3, using the locus-specific annealing temperatures listed in Table 6.1.

Genotyping error was assessed by including positive controls with each PCR reaction and re-

amplifying approximately 6% of the samples.

Evidence for null alleles and long allele drop-out was assessed with MICRO-CHECKER

(vanOosterhout et al. 2004) using 1,000 iterations. Frequency of null alleles was calculated with

the method of Brookfield (1996). I calculated the number of alleles per locus, observed

heterozygosity (HO) and expected heterozygosity (HE) in GENALEX (Peakall and Smouse 2006).

Allelic richness was rarefacted to correct for unequal sample sizes in HP-RARE (Kalinowski

2004; 2005). Linkage disequilibrium and deviations from Hardy-Weinberg equilibrium (HWE)

were tested in GENEPOP v.4.0.1 (Raymond and Rousset 1995; Rousset 2008). Significance levels

were corrected for multiple comparisons following Rice (1989).

I assessed genetic differentiation among sample sites by calculating absolute differentiation (Dest,

Jost 2008) of all sites with N ≥ 15 in SMOGD (Crawford, 2010). For purposes of comparison with

previous studies and for calculations of historic migration rates (Nm) I also used FSTAT (Goudet,

1995) to calculate pairwise FST and assessed significance with 10,000 randomizations. Isolation

by distance (IBD, a significant correlation between geographic and genetic distance, Wright

1943) was assessed using IBDWS (Jensen et al. 2005) with an input matrix of Dest and pairwise

distances (km) between sites.

Within-population heterozygosity (HO) from Ontario sites was compared to HO values reported

in Mockford et al. (2007) using an independent-samples t-test in SPSS v.20.0 (IBM-SPSS,

Chicago, IL) after testing normality of the data. Comparisons were made for each locus sampled

in both studies and across all sampled loci. Comparisons were made between sampled sites in

Ontario and the “western” sites from Mockford et al (2007; all sampled sites west of the

Appalachian Mountains). I also compared sites east of the Appalachian Mountains to western

populations, combining study sites from Ontario with western sites from Mockford et al. (2007).

90

Pearson’s correlation coefficient was used to test for significant relationships between latitude

and HO among sites surrounding the Great Lakes.

Population structure was assessed by Bayesian inference (BI) in STRUCTURE V.2.3.4 (Pritchard et

al. 2000) and TESS V.2.3.1 (Chen et al. 2007) following the run parameters outlined in Chapter 3.

STRUCTURE considered possible K values (number of genetically distinct populations) from one

to six with 10 independent runs at each value of K. TESS considered possible Kmax values

(maximum possible number of populations represented by the data) from two to eight, with 10

independent runs at each Kmax.

Assignment tests were conducted in GENECLASS V.2.0 (Piry et al. 2004) using the Bayesian

method of Rannala and Mountain (1997), with 100,000 iterations and a Type I error level of

0.05. This duplicates the analyses conducted by Howes et al. (2009), allowing a reasonable level

of comparison between studies. Assignment tests considered only sampling areas with six or

more samples. Individual samples from other sites and samples of unknown origin were then

assessed by the program as “unknown”, and assigned to the most similar sampling area.

Population structure was also visualized with principal coordinates analysis (PCoA) in GENALEX,

based on Dest for sampled sites and on Nei’s unbiased genetic distance for individuals.

6.3 Results

Loci Eb09, Eb11, GmuD70, GmuD89, and GmuD90 either did not amplify, or could not be

scored consistently despite multiple adjustments of PCR conditions. Thus, 12 loci were used for

analyses. In total, 116 samples were collected but several yielded degraded DNA and were

successfully amplified at only five or six loci. These samples were excluded and a total of 97

individuals (91 individuals from known locations) were genotyped at > 10 loci and included in

the final analysis.

All duplicated genotypes were identical. MICRO-CHECKER found evidence for potential null

alleles at three loci (Eb19, GmuD93 and GmuD107). However, when the four largest samples

were tested independently, potential null alleles were not consistent among sites; only EO and

GH showed evidence for nulls, and only at locus Eb19.

91

Deviations from HWE were detected at locus Eb19 in PSD, GH, and EO, but not in KAW or LE.

Evidence for LD was detected across the entire dataset between two pairs of loci: GmuD55–

GmuD107 and GmuD28–GmuD107. However, LD was not detected when testing sampled areas

independently and, therefore, I accepted the null hypothesis of linkage equilibrium. PI and PIsibs

decreased to < 0.01 with the inclusion of three and six loci, respectively. The 12 loci exhibited 3–

16 alleles (mean 8.917, s.d. = 4.187), and HO ranged from 0.253 at locus GmuD21 to 0.845 at

locus GmuD28 (Table 6.1).

Summary statistics for all sampling sites are shown in Table 6.2. Pairwise values of Dest ranged

from 0.010 to 0.156 (mean = 0.083, s.d. = 0.044, Table 6.3). Values of pairwise FST ranged from

0.039 to 0.099 (mean 0.072, s.d. = 0.021). Nm among sites averaged 2.432, and Nm between

each pair of sites ranged from 0.095 (GH–PSD) to 3.380 (PSD–EO; Table 6.3). No evidence

suggested significant isolation by distance among the four sites with N > 12 (Z = 194.900, r =

0.233, p = 0.301).

In the PCoA of sampling sites, the first principal coordinates axis accounted for 60.19% of total

variation. This axis separated sites LE and GH from PSD, Kaw, and EO (Figure 6.2). When the

PCoA was conducted at the individual level individuals clustered by site but with overlap

indicating that differentiation in this dataset may have occurred along a gradient rather than

along sharply defined boundaries.

GENECLASS assigned individuals from LE, GH, PSD, KAW, and EO to their area of origin with

69% accuracy (Table 6.4). Samples from KAW were assigned to PSD (N=5) or GH (N=1).

When KAW was removed from assignment tests, overall accuracy increased to 79%. The two

samples from the north shore of Lake Huron were assigned to PSD. The two samples from

Algonquin Park and the sample from the south shore of Lake Huron were not assigned to any

sampled clusters (p < 0.01).

The deviance information criterion (DIC) in the TESS analysis decreased gradually from Kmax = 2

with no clear point of inflection (mean ∆DIC = 54.2, Figure 6.3A). Individual q-matrices

stabilized at Kmax = 2; no clearly defined new clusters appeared at higher values of Kmax,

although potential admixture from a third population became apparent in site EO at Kmax = 3.

The first resolved population included LE, GH, and LHsouth (mean q = 0.969, s.d. = 0.094). The

92

second population included all other samples (mean q = 0.709, s.d. = 0.377). One sample from

PSD was assigned with approximately equal probability to both populations (0.493 vs. 0.507).

STRUCTURE resolved the same two populations as TESS at K = 2 (Figure 6.3B). At K = 3, LE and

LHsouth separated from population GH with evidence of admixture remaining between the two

clusters. At K = 4, EO separated from a final population consisting of PSD, KAW, LHnorth,

ALG, and Hastings County.

Heterozygosity data were normally distributed and Levene’s test indicated equal variances (F =

0.09, p = 0.927). Data from the two loci used both in this study and in Mockford et al. (2007;

Eb17 and Eb19) were combined for comparison. Observed heterozygosity in the Great Lakes

portion of the species’ range was significantly higher at locus Eb17 (t = -3.621, d.f. = 15, p =

0.003) but not at locus Eb19 (t = -1.823, d.f. = 15, p = 0.088). When mean heterozygosity across

all loci in both studies was compared, HO was significantly higher in western populations (t = -

3.749, d.f. = 15, p = 0.002) than in the disjunct eastern populations. No difference in

heterozygosity occurred between the western populations sampled by Mockford et al. (2007) and

the populations sampled in this study (t = -0.413, df = 10, p = 0.688).

Latitude and HO were not correlated (Pearson’s correlation = 0.056, N = 11, p = 0.869). Site EO

had substantially higher heterozygosity than the site from Ontario sampled by Mockford et al.

(2007) (0.636 compared to 0.48) despite their proximity (< 40 km apart).

6.4 Discussion

Three independent Bayesian analyses and a principle components analysis reveal consistent

population structure in E. blandingii in southern Ontario. Sampling localities in this study

include two genetic populations and four subpopulations, refuting the hypothesis of panmixia in

E. blandingii in Ontario. Assignment tests identify individuals to their subpopulation of origin

with relatively high accuracy considering the small sample sizes available for this study.

Intensive urban development and expanding road networks make current migration between

these four subpopulations unlikely.

Bayesian assignment of individual samples to the larger dataset suggests that the population on

the north shore of Lake Huron at the northern extreme of the species’ range may be continuous

93

with the population in PSD. Assignment of individuals from KAW to PSD reflects genetic

continuity between these two areas (Table 6.4; Figure 6.3). However, assignment tests cannot

determine a likely origin for the sample from the south shore of Lake Huron or the two samples

from Algonquin Park. Thus, these areas may be genetically distinct, especially because E.

blandingii in both locations are apparently isolated from other nearby populations (Ontario

Nature Reptile and Amphibian Atlas).

Based on my results, I propose four tentative management units (MUs) for E. blandingii in

Ontario: Lake Erie, Golden Horseshoe, Georgian Bay-Parry Sound District and Eastern Ontario.

These four areas are unlikely to qualify as DUs under Canadian law (Green 2005) because I am

not aware of any evidence to suggest that they are subject to significantly different risks of

extinction. However, they appear to be both demographically and genetically independent of one

another and this should be considered when planning for population management and recovery.

Future studies of geographically disjunct areas of occurrence such as Algonquin Park and the

south shore of Lake Huron may identify further MUs or clarify relationships between these sites

and the proposed MUs. Currently under-sampled areas should not be considered part of the four

tentative management units until genetic data are available to confirm this categorization.

Genetic diversity (HO) is significantly lower in the disjunct eastern populations than in

populations around the Great Lakes. This pattern was first shown by Mockford et al. (2007) and

is not altered by the inclusion of additional populations from Ontario. Diversity in E. blandingii

does not vary with latitude. A negative correlation between genetic diversity and latitude is

expected in turtles in North America (Galbraith 2008) because colonization following the last ice

age proceeded from south to north making founder effects more likely in northern populations,

although this has not been tested in other species of turtle in Ontario. However, E. blandingii has

a compressed latitudinal range and likely underwent east-west migrations as well as north-south

migrations after the last ice age. Fossil evidence places E. blandingii in southern Indiana 15–14

ka BP, and fossils are also known from Indiana and Michigan 6–4 ka BP (Holman 1992).

Although some populations might have used Pleistocene refugia in the southern Atlantic plain

(Bleakney 1958), it is probable that other populations persisted near the Great Lakes throughout

the Wisconsonian ice age, rapidly recolonizing the Great Lakes area as the ice sheets retreated

(Holman 1992). This hypothesis places a major refugium for E. blandingii south of the centre of

94

the Great Lakes portion of the current range. Gradual expansion from this refugium to the sites

considered in this study is consistent with the similar levels of genetic diversity reported from a

range of central populations.

Lower values of Nm for sites in Ontario compared to those in Nova Scotia (Mockford et al.

2005) are probably due in part to geographic distance. The Nova Scotian sites compared by

Mockford et al. (2005) were 15 – 25 km apart, with Nm = 1.76 – 5.8, and they estimated Nm =

0.54 – 0.74 between Nova Scotia and a Michigan population approximately 1510 km in distance.

Ontario populations sampled here were 151 – 516 km apart and estimated Nm values were

intermediate between the two extremes reported by Mockford et al. (2005).

Several sites have private alleles at one or more loci, indicating possible effects of genetic drift.

However, no alleles are fixed and heterozygosity is comparable to that reported for other

populations of turtles (Vargas-Ramirez et al. 2012). Heterozygosity in continental chelonian

species ranges from 0.33 (Podocnemis lewyana) to 0.76 (Astrochelys radiata and Malaclemys

terrapin), and the mean heterozygosity of sampled populations of E. blandingii in Ontario (0.64)

is within this range. If population sizes can be stabilized (or kept stable), there is no reason to

believe loss of genetic diversity is cause for immediate concern at these sites. Thus, recovery

plans need not consider genetic management measures at this time. Instead, effort should be

made to mitigate high adult mortality and low recruitment (Congdon et al. 2008). Increasing or at

the very least maintaining population size is the most effective way to prevent loss of genetic

diversity in threatened populations (Frankham et al. 2002).

Although active genetic management appears to be unnecessary, the genetic structure

demonstrated here should be considered when planning measures that will increase or modify

habitat connectivity. For example, anthropogenic features that fragment habitat (e.g. highways,

urban development) also reduce gene flow among population fragments. Where possible, the

effect of this fragmentation can be mitigated using tools such as wildlife underpasses, or

corridors of suitable habitat. Alternatively, actions such as translocations that involve moving

individuals across the landscape should include explicit consideration of genetic structure and

social interactions of turtles (Chapters 4 and 6). Mixing of genetic populations can have serious

consequences for fitness if locally adapted genes or co-adapted gene complexes are disrupted

95

(outbreeding depression, Templeton 1986). For example, Sletvold et al. (2012) demonstrated a

47% fitness reduction when individuals from two populations of a nectariferous orchid

(Gymnadenia conopsea) located 1.6 km apart were crossed. The situation was reversed in a study

concerning the translocation of Bighorn sheep (Miller et al. 2012); more outbred individuals (i.e.

individuals with more introduced alleles) lived longer and had higher reproductive success than

individuals who were not affected by the genetic rescue. Increased fitness in outbreeding

Bighorn sheep supports the efficacy of facilitated rescue effects on declining populations (Hogg

et al. 2006; Miller et al. 2012). There is no evidence to suggest that mixing of populations of E.

blandingii is likely to cause outbreeding depression, but minimal genetic data exist for this

species and the possibility has not been investigated.

Interestingly, results from both STRUCTURE and TESS suggest possible past translocations of

individuals between populations (Fig.2; individuals with an approximately 50% probability of

membership to two populations may be first-generation offspring of migrants who mated with

residents). Collection of individual turtles by members of the public occurs regularly and these

individuals are often released elsewhere than their collection site (F. Ross, pers. comm.; S.

Gillingwater, pers. comm.; C. Davy, unpublished data). There are no data on the frequency of

these casual translocations, but attempts to maintain existing genetic structure of populations are

unlikely to succeed without public education. Such efforts should explain not only the laws that

prohibit collection of turtles in Ontario, but also the impact that collection and translocation can

have on wild turtle populations, as well as clarifying the low chance of survival for their former

pets after release.

This study addresses the first two areas for research in genetics of turtles recommended by Alacs

et al (2007) : 1) “… identification of genetic discontinuities at landscape and species levels to

delineate management units, and 2) Predicting effects of landscape-level changes and

concomitant changes in population demography and movement patterns on apportionment of

genetic diversity within and among populations.” I achieve the delineation of management units

based on existing genetic discontinuities. Application of Bayesian methods to identify and

profile populations across the central range of E. blandingii will likely reveal further population

structure at appropriate spatial scales. Future studies should more clearly delineate boundaries

among populations and significant barriers to gene flow including those hypothesized by

96

Mockford et al. (2007). For example, the Appalachian Mountains may have played a role in the

isolation of the disjunct eastern populations. However, FST values suggest that individuals from

New York were most similar to individuals in St. Lawrence Islands National Park (Ontario).

Perhaps these populations are historically connected via the Delaware water gap or a similar

landscape feature. Alternatively, perhaps Bayesian analyses will reveal a completely different

pattern of structure than was previously suggested.

6.5 Acknowledgments

Sample collection was accomplished with the assistance of Sue Carstairs, Brennan Caverhill,

Suzanne Coombes, Joe Crowley, Jacqueline Litzgus, James Paterson, James Baxter-Gilbert, Jim

Trottier, Julia Riley, Jeremy Rouse, David Seburn, John Urquhart and Amelia Whitear. Jeff

Hathaway and Jenny Pierce allowed me to sample E. blandingii at Scales Nature Park. Pedro

Bernardo assisted with the laboratory analyses. Field collection was funded in part by a Canada

Collection grant from Wildlife Preservation Canada to CD. Laboratory analyses were funded by

a Species at Risk Research Fund for Ontario grant from the Government of Ontario; I thank Bob

Johnson, Julia Philips and Robert Murphy for collaborating on this grant. Comments from

Robert Murphy and Deborah McLennan improved an earlier version of this manuscript.


conservation. Chelon Res Monogr 4:107-123

Banning-Anthonysamy WJ (2012) Spatial ecology, habitat use, genetic diversity, and reproductive success: measures of connectivity of a sympatric freshwater turtle assemblage in a fragmented landscape. Dissertation, University of Illinois at Urbana-Champaign

Barton NH, Slatkin M (1986) A quasi-equilibrium theory of the distribution of rare alleles in a subdivided population. Heredity 56:409-416

Bleakney JS (1958) A zoogeographic study of the amphibians and reptiles of eastern Canada. Natl Mus Can Bull 155:1-119

Brecke BJ, Moriarty JJ (1989) Emydoidea blandingii (Blanding’s turtle). Longevity. Herpetol Rev 20:53

Brookfield JFY (1996) A simple new method for estimating null allele frequency from heterozygote deficiency. Mol Ecol 5:453-455

97

Chen C, Durand E, Forbes F, François O (2007) Bayesian clustering algorithms ascertaining spatial population structure: a new computer program and a comparison study. Mol Ecol Notes 7:747-756

Congdon JD, van Loben Sels RC (1991) Growth and body size in the Blanding’s turtles (Emydoidea blandingii): relationships to reproduction. Can J Zool 69:239-245

Congdon JD, Dunham AE, van Loben Sels RC (1993) Delayed sexual maturity and demographics of Blanding’s Turtles (Emydoidea blandingii): implications for conservation and management of long-lived organisms. Conserv Biol 7:826-833

Congdon JD, Graham TE, Herman TB, Lang JW, Pappas MJ, Brecke BJ (2008) Emydoidea blandingii (Holbrook 1838) – Blanding’s turtle. In: Rhodin AGJ, Pritchard PCH, van Dijk PP, Saumure RA, Buhlmann KA, Iverson JB (eds.). Conservation biology of freshwater turtles and tortoises: a compilation project of the IUCN/SSC Tortoise and Freshwater Turtle Specialist Group. Chelon Res Monogr 5:015.1-015.12. doi:10.3854/crm.5.015.blandingii.v1.2008, http://www.iucn-tftsg.org/cbftt/


Crawford NG (2010) SMOGD: software for the measurement of genetic diversity. Mol Ecol Res 10:556-557

Ernst CH, Lovich JL (2009) Turtles of the United States and Canada, 2nd ed, Johns Hopkins University Press, Baltimore, Maryland

Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol 14:2611-2620

Frankham R, Ballou JD, Briscoe DA (2002) Introduction to conservation genetics. Cambridge University Press, Cambridge

Galbraith DA (2008) Population biology and population genetics. In: Steyermark AC, Finkler MS, Brooks RJ (eds) The biology of the snapping turtle. Johns Hopkins University Press, Baltimore, Maryland pp 168-180

Goudet J (1995) FSTAT (Version 1.2): A computer program to calculate F-statistics. J Hered 86:485-486

Green DM (2005) Designatable units for status assessment of endangered species. Conserv Biol 19:1813-1820

Hogg JT, Forbes SH, Steele BM, Luikart G (2006) Genetic rescue of an insular population of large mammals. Proc R Soc Lond B Biol Sci 273:1491-1499.

Holman JA (1992) Late Quaternary herpetofauna of the central Great Lakes region, U.S.A.: zoogeographical and paleoecological implications. Quaternary Sci Rev 11:345-351

Howes BJ, Brown JW, Gibbs HL, Herman TB, Mockford SW, Prior KA, Weatherhead PJ (2009) Directional gene flow patterns in disjunct populations of the black ratsnake (Pantherophis obsoletus) and the Blanding’s turtle (Emydoidea blandingii). Conserv Genet 10:407-417

98

Jensen JL, Bohonak AJ, Kelley ST (2005) Isolation by distance, web service. BMC Genet 6:13. v.3.23 http://ibdws.sdsu.edu/

Jost L (2008) GST and its relatives do not measure differentiation. Mol Ecol 17:4015-4026

Jost L (2009) D vs. GST: Response to Heller and Siegismund (2009) and Ryman and Leimar. Mol Ecol 18:2088-2091

Kalinowski ST (2004) Counting alleles with rarefaction: private alleles and hierarchical sampling designs. Conserv Genet 5:539-543

Kalinowski ST (2005) HP-Rare: A computer program for performing rarefaction on measures of allelic diversity. Mol Ecol Notes 5:187-189

King TL, Julian SE (2004) Conservation of microsatellite DNA flanking sequences across 13 Emydid genera assayed with novel bog turtle (Glyptemys muhlenbergii) loci. Conserv Genet 5:719-725

Miller JM, Poissant J, Hogg JT, Coltman DW (2012) Genomic consequences of genetic rescue in an insular population of bighorn sheep (Ovis canadensis). Mol Ecol 21:1583-1596

Mockford SW, Snyder M, Herman TB (1999) A preliminary examination of genetic variation in a peripheral population of Blanding’s turtle, Emydoidea blandingii. Mol Ecol 8:323-327

Mockford SW, McEachern L, Herman TB, Snyder M, Wright JM (2005) Population genetic structure of a disjunct population of Blanding’s turtle (Emydoidea blandingii) in Nova Scotia, Canada. Biol Conserv 123:373-380

Mockford SW, Herman TB, Snyder M, Wright JM (2007) Conservation genetics of Blanding’s turtle and its application in the identification of evolutionarily significant units. Conserv Genet 8:209-219

Osentoski MF, Mockford S, Wright JM Snyder M, Herman TB, Hughes CR (2002) Isolation and characterization of microsatellite loci from the Blanding’s turtle, Emydoidea blandingii. Mol Ecol Notes 2:147-149

Paetkau D, Slade R, Burden M, Estoup A (2004) Direct, real-time estimation of migration rate using assignment methods: a simulation-based exploration of accuracy and power. Mol Ecol 13:55-65

Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Mol Ecol 6:288-295

Pennock DS and Dimmick WW (1997) Critique of the Evolutionarily Significant Unit as a Definition for “Distinct Population Segments” under the U.S. Endangered Species Act. Conservation Biology 11:611-619

Piry S, Alapetite A, Cornuet JM, Paetkau D, Baudouin L, Estoup A (2004) GeneClass2: a software for genetic assignment and first-generation migrant detection. J Hered 95:536-539

Power TD (1989) Seasonal movements and nesting ecology of a relict population of Blanding’s turtles (Emydoidea blandingii) in Nova Scotia. M.Sc. Thesis, Acadia University, Wolfville, Nova Scotia.

99

Pritchard JK, Stephens M, Donnelly PJ (2000) Inference of population structure using multilocus genotype data. Genetics 155:945-959

Rannala B, Mountain JL (1997) Detecting immigration by using multilocus genotypes. P Natl Acad Sci USA 94:9197-9221

Raymond M, Rousset F (1995) GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. J Hered 86:248-249

Rice WR (1989) Analyzing tables of statistical tests. Evolution 43:223-225

Rousset F (2008) Genepop’007: a complete reimplementation of the Genepop software for Windows and Linux. Mol Ecol Res 8:103-106

Rubin CS, Warner RE, Bouzat JL, Paige KN (2001) Population genetic structure of Blanding’s turtles (Emydoidea blandingii) in an urban landscape. Biol Conserv 99:323-330


Sletvold N, Grindeland JM, Zu P, Ågren J (2012) Strong inbreeding depression and local outbreeding depression in the rewarding orchid Gymnadenia conopsea. Conserv Genet 13:1305-1315.


Templeton AR (1986) Coadaptation and outbreeding depression. In: Soulé M (ed) Conservation biology: the science of scarcity and diversity. Sinauer, Sunderland, Massachusetts, pp 105-116

van Dijk PP, Rhodin AGJ (2011) Emydoidea blandingii. In: IUCN 2011. IUCN Red List of Threatened Species. Version 2011.2. <http://www.iucnredlist.org>. Downloaded on 15 December 2011

van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data. Mol Ecol Notes 4:535-538

Vargas-Ramirez M, Stuckas H, Castaňo-Mora OV, Fritz U (2012) Extremely low genetic diversity and weak population differentiation in the endangered Colombian river turtle Podocnemis lewyana (Testudines: Podocnemididae) Conserv Genet 13:65-77

Whitlock MC, McCaughley DE (1999) Indirect measures of gene flow and migration: FST ≠ 1/(4Nm + 1). Heredity 82:117–125

Wright S (1943) Isolation by distance. Genetics 28:114-138

100

100

Table 6.1. Genetic diversity at 12 microsatellite loci for 97 Emys blandingii from southern Ontario. Temp. (optimal annealing

temperature (°C) determined from temperature gradients of initial PCR reactions) sample size (N), allelic richness (k), observed and

expected heterozygosity (HO, HE) and two measures of probability of identify (PI, PISibs) are shown for each locus. Total values show

mean ± standard error for N, k, Ne, HO and HE, and PI/PIsibs values with all loci included.

Temp. N k HO HE PI PISibs

GmuB08 58 96 7 0.406 0.403 0.378 0.643

GmuD16 56 95 13 0.811 0.818 0.055 0.357

GmuD21 58 95 3 0.253 0.230 0.617 0.789

GmuD28 61 97 16 0.845 0.862 0.034 0.328

GmuD55 56 96 13 0.792 0.816 0.056 0.356

GmuD87 54 88 11 0.659 0.723 0.123 0.419

GmuD88 58 96 11 0.792 0.848 0.040 0.336

GmuD93 58 95 4 0.421 0.552 0.294 0.548

GmuD107 58 96 11 0.771 0.854 0.038 0.332

GmuD121 58 94 8 0.766 0.725 0.103 0.413

Eb17 58 95 6 0.705 0.742 0.109 0.406

Eb19 58 92 4 0.478 0.704 0.140 0.433

Total 94.583 ± 0.701 8.917 ± 1.209 0.642 ± 0.057 0.690 ± 0.057 0.000 0.000

101

101

Table 6.2. Number of alleles (number of private alleles in parentheses) and observed and expected heterozygosities (HO and HE) for

97 Emys blandingii sampled across southern Ontario and genotyped at 12 microsatellite loci. Loci Gmu– from King and Julian (2004).

Loci Eb– from Osentoski et al. (2002). Acronyms for sampling areas are defined in Figure 1. Estimated frequency of a null allele is

based on analysis of the entire data set following Brookfield (1996). No loci showed consistent evidence for null alleles when

sampling areas were analyzed independently. HO = observed heterozygosity; HE = expected heterozygosity; Ar = allelic richness; PAr

= private allelic richness.

ALG EO LHnorth PSD GH KAW LE LHsouth

GmuB08 Number of alleles 2 4 (1) 3 (1) 5 (1) 4 3 2 1 Estimated null allele HO 0.500 0.591 0.500 0.522 0.214 0.333 0.150 –

frequency = 0.00 HE 0.375 0.583 0.625 0.436 0.199 0.292 0.139 – N 2 22 2 23 14 6 20 1

GmuD16 Number of alleles 3 8 2 8 9 (1) 7 7 1 Estimated null allele HO 1.000 0.905 0.500 0.826 0.800 0.667 0.800 –

frequency = 0.00 HE 0.625 0.796 0.375 0.751 0.767 0.819 0.743 – N 2 21 2 23 15 6 20 1

GmuD21 Number of alleles 1 2 2 2 2 3 (1) 2 2 Estimated null allele HO 0.000 0.091 1.000 0.174 0.286 0.800 0.350 –

frequency = 0.00 HE 0.000 0.087 0.500 0.159 0.245 0.580 0.289 – N 2 22 2 23 14 5 20 1

GmuD28 Number of alleles 2 9 (1) 3 11 8 5 10 2 Estimated null allele HO 0.000 0.818 1.000 0.870 0.733 1.000 0.900 –

frequency = 0.00 HE 0.500 0.789 0.625 0.843 0.791 0.722 0.851 – N 2 22 2 23 15 6 20 1

GmuD55 Number of alleles 4 9 2 8 5 8 (2) 6 1 Estimated null allele HO 1.000 0.773 0.500 0.773 0.800 1.000 0.850 –

frequency = 0.000 HE 0.750 0.784 0.375 0.826 0.664 0.819 0.711 – N 2 22 2 22 15 6 20 1

102

102

GmuD87 Number of alleles 3 5 (1) 2 8 (3) 6 3 5 (1) 2 Estimated null allele HO 0.500 0.636 1.000 0.696 0.818 0.250 0.579 –

frequency = 0.000 HE 0.625 0.727 0.500 0.641 0.736 0.656 0.677 – N 2 22 2 23 11 4 19 1


frequency = 0.00 HE 0.625 0.790 0.750 0.855 0.640 0.778 0.800 – N 2 22 2 23 14 6 20 1


frequency = 0.148 HE 0.375 0.567 0.375 0.491 0.408 0.500 0.531 – N 2 22 2 23 14 6 20 1

GmuD107 Number of alleles 2 8 4 9 6 6 7 2 Estimated null allele HO 1.000 0.773 1.000 0.783 0.643 0.833 0.750 –

frequency = 0.073 HE 0.500 0.721 0.750 0.823 0.694 0.694 0.659 – N 2 22 2 23 14 6 20 1

GmuD121 Number of alleles 3 7 3 6 5 6 5 1 Estimated null allele HO 0.500 0.818 1.000 0.762 0.867 1.000 0.600 –

frequency = 0.000 HE 0.625 0.751 0.625 0.718 0.598 0.800 0.484 – N 2 22 2 21 15 5 20 1

Eb17 Number of alleles 2 4 2 5 5 (1) 3 5 1 Estimated null allele HO 0.500 0.591 0.500 0.636 0.733 0.600 0.950 –

frequency = 0.000 HE 0.375 0.699 0.375 0.636 0.709 0.460 0.696 – N 2 22 2 22 15 5 20 1

Eb19 Number of alleles 2 4 2 4 4 3 3 2 Estimated null allele HO 0.500 0.364 0.500 0.591 0.250 0.600 0.550 –

frequency = 0.234 HE 0.375 0.673 0.375 0.577 0.642 0.660 0.594 – N 2 22 2 22 12 5 20 1 Mean HO 0.583 0.636 0.75 0.672 0.578 0.688 0.64 – Mean HE 0.479 0.664 0.521 0.646 0.591 0.648 0.598 – Ar – 5.09 – 5.25 4.8 – 4.64 – PAr – 0.62 – 0.39 0.53 – 0.33 –

103

103

Table 6.3. Genetic differentiation of Emys blandingii among sites in Ontario with N ≥ 15. All

FST values are significant (p < 0.05). All FST values were significant (p < 0.05). Average

historical number of migrants per generation (Nm) was calculated following Barton and Slatkin

(1986).

Approximate distance (km) Dest FST Nm

Lake Erie–Golden Horseshoe 151 0.057 0.066 1.350

Lake Erie–Parry Sound District 310 0.062 0.062 1.442

Lake Erie–Eastern Ontario 516 0.100 0.089 1.029

Golden Horseshoe–Parry Sound District 266 0.156 0.100 0.952

Golden Horseshoe–Eastern Ontario 367 0.143 0.099 1.013

Parry Sound District–Eastern Ontario 337 0.064 0.040 3.380

Table 6.4. GENECLASS results for Bayesian assignment tests. Values represent the proportion

of individuals from each sampled population assigned to each population. Values in bold indicate

the proportion of individuals from each sampled population assigned correctly to their source

population. Grey shaded areas indicate the two larger genetic clusters identified by TESS and

STRUCTURE.

Sampled population Assigned population

LE GH PSD KAW EO

LE 0.8 0 0.2 0 0

GH 0.07 0.67 0.26 0 0

PSD 0 0 0.74 0.04 0.22

KAW 0 0.17 0.83 0 0

EO 0 0 0.27 0 0.73

104

104

Figure 6.1. Approximate location of collection areas for Emys blandingii sampled across

southern Ontario. Top right inset indicates species range in North America (shown in red).

Sampling was focused on sites indicated with grey squares: LE = Lake Erie; GH = Golden

Horseshoe; PSD = Parry Sound District; KAW = Kawartha Lakes; EO = Eastern Ontario.

Sample sizes are included in each site marker. Grey triangles indicate extra samples included

opportunistically (each triangle represents an individual turtle): LHsouth = south shore of Lake

Huron; LHnorth = north shore of Lake Huron; ALG = Algonquin Provincial Park. Variation in

sample sizes results from differential sampling effort; differences in sample sizes are not

reflective of variation in actual population sizes. Base map modified from


Documentation license; range map modified from COSEWIC (2005).

105

105

Figure 6.2. Principal coordinates analysis of sampling areas (A, B) and individuals (C) for 91

Emys blandingii sampled from across southern Ontario based on 12 microsatellite loci.

106

106

107

107

Figure 6.3 (previous page) Population structure inferred by Bayesian inference for 91 Emys

blandingii collected across southern Ontario. A) TESS results showing decreasing deviance

information criterion (DIC) with increasing values of Kmax. B) STRUCTURE results, mean

estimated ln probability of the data (L(K)) for increasing values of K, and ∆K, the second order

rate of change of L(K) following Evanno et al. (2005). Site abbreviations are explained in Figure

6.1.

108

108

Chapter 7 Genotypes and ghosts: comparative landscape genetics reveals

incongruent barriers to gene flow amongst three species of freshwater turtle

Formatted for Conservation Genetics.

7 Abstract

The genetic connectivity of populations is determined by levels of gene flow across the

landscape, which is affected strongly by landscape features. Understanding population

connectivity is a priority for conservation because maintenance of additive genetic diversity

within populations affects their probability of persistence. Comparative approaches to landscape

genetics can help to prioritize areas for applied conservation approaches that increase

connectivity of multiple species across the landscape, such as wildlife corridors. Here, I

compared population structure in three sympatric species of turtle with varying dispersal ability.

I used Bayesian clustering analyses, Monmonier’s algorithm, and estimates of gene flow based

on data from microsatellite markers to identify areas of genetic connectivity and barriers to gene

flow that were shared among species. Monmonier’s algorithm revealed significant but discordant

barriers to gene flow in all three species, and boundaries between populations inferred with

Bayesian clustering analyses were also incongruent among species. Dispersal ability based on

previously published radio-telemetry studies did not predict either estimated gene flow or the

number of significant barriers to gene flow. Apart from a possible common barrier to gene flow

near the base of the Bruce Peninsula, genetic structure in the three species differed strongly,

precluding generalization of biogeographic patterns among species. The discrepancy between the

genetic results and previous ecological studies suggested that we may need to re-evaluate our

109

109

understanding of these relatively well-studied species, and highlighted potential areas for future

research.

Keywords: Clemmys guttata, Chelydra serpentina, Emys blandingii, Ontario, BARRIER,

dispersal

7.1 Introduction

The genetic connectivity of populations affects their long-term probability of extinction and is,

therefore, a priority for conservation (Frankham and Ralls 1998; Frankham et al. 2002). Recently

developed methods allow the inference of genetic population structure, rates of gene flow among

populations and spatial patterns of gene flow based on genetic data (Wilson and Rannala 2003;

Chen et al. 2007; Guillot et al. 2009). It can take many generations for the effects of a changing

landscape to be genetically detectable (Landguth et al. 2010, Blair et al. 2012). Therefore,

genetic structure in long-lived organisms may indicate the effects of past, but not current,

landscapes. A comparative approach can be used both to test hypotheses about the historic

distribution and structure of populations and maximize the effectiveness of applied conservation

measures by identifying common patterns of genetic population structure among species.

Genetic connectivity is measured in terms of gene flow among populations, and differs from

demographic connectivity, which determines the impact of immigrants on a population’s growth

rate and size but does not necessarily affect its genetic profile (Lowe and Allendorf 2010). In

large populations, allopatric speciation may result from the loss of connectivity followed by

genetic divergence over time. However, in small, threatened populations, genetic connectivity

may be vital to persistence. Genetic drift gradually erodes genetic diversity in small, isolated

populations and reduces their long-term adaptive potential (Frankham et al. 2002). Without

connectivity to neighboring populations there is no possibility of a rescue effect (augmentation

of the gene pool by reproductively successful immigrants; Thrall et al. 1998; Tallmon et al.

2004). Thus, understanding genetic structure of threatened populations is essential for their

effective conservation and recovery.

110

110

Anthropogenic barriers to gene flow often have significant demographic and genetic effects on

wildlife, but the genetic impacts may be difficult to detect in long-lived organisms (Bennett et al.

2009; Bennett et al. 2010). Thus, studies testing the genetic impact of anthropogenic barriers

(dams, highways, urban, and agricultural development) on long-lived species often fail to detect

an effect (Kuo and Janzen 2004; Marsack and Swanson 2009; Pittman et al. 2011). This does not

necessarily indicate that the tested barriers are permeable because the genetic signatures of

barriers develop over generations. It may take 10 - 200 generations for a new barrier to modify

the genetic profile of affected populations sufficiently for detection, and up to 15 generations for

the removal of a barrier to be detectable (Landguth et al. 2010, Blair et al. 2012). As a result,

tests of genetic connectivity in long-lived organisms are especially unlikely to detect effects of

relatively recent anthropogenic landscape modifications. This applies even if the demographic

impact of the modification is devastating. For example, the endangered spotted turtle (Clemmys

guttata) has a generation time > 25 years (COSEWIC 2004). Extant populations of Cl. guttata

are extremely isolated from one another and the isolation is maintained by current habitat

modifications that make gene flow among them impossible (COSEWIC 2004). However, genetic

structure among populations of Cl. guttata in Ontario most likely reflects the signature of a

landscape inhabited > 500 - 5,000 years ago. Population genetic structure may therefore indicate

historical landscape effects, while population persistence is affected by the current landscape

structure.

The field of landscape genetics involves measurements of genetic connectivity of populations

across landscapes and investigations into how landscape features affect gene flow (Manel et al.

2003, Epps et al. 2007). Genetic and spatial data can be integrated in Bayesian inference of

population structure (Chen et al. 2007; Guillot et al. 2005) to define the geographic limits of

genetic populations. More complex analyses integrate resistance layers to explicitly test the

effects of different landscape features and habitat types on gene flow among populations.

Resistance layers describe the relative ease of dispersal of a study organism or the relative rate of

gene flow through different habitat types (O’Brien et al. 2006; Wang et al. 2008). They allow

explicit tests of hypotheses related to landscape structure when integrated into least cost path

models (Adriaensen et al. 2003) or when considered using circuit theory (McRae 2006).

Unfortunately, assigning costs to resistance layers requires data such as dispersal distances,

habitat selection and relative survivorship of individuals in different habitats that are not

111

111

available for most wild species. As a result, parameterization of resistance layers often relies on

“expert opinion” or direct evidence from telemetry studies with small sample sizes (Spear et al.

2010).

When data are insufficient to assign accurate values to resistance layers, population-level

analyses provide a simpler but more robust alternative. For example, Manni et al. (2004) use

Delaunay triangulation to connect populations in a single geometric network, and apply

Monmonier’s maximum difference algorithm (Monmonier 1973) to identify boundaries between

neighboring populations where the change in genetic distance is significant. This method

provides less fine-scale information about gene flow across the landscape than least cost path

models or circuit theory but relies on fewer assumptions. It is also well-suited to clustered

sampling designs.

Genetic and demographic connectivity of populations can be increased using conservation tools

ranging in scale from small wildlife underpasses beneath large highways to translocations or

large wildlife corridors. The financial cost of these mitigation measures is significant. Therefore,

the most economical mitigation measures will target multiple species. Comparative approaches

to landscape genetics (DiLeo et al. 2010; Goldberg & Waits 2010; Cyr and Angers 2011) can

identify areas of historic connectivity for multiple species. Such areas could be prioritized for

mitigation measures. Comparative studies can also identify pairs of populations that have been

isolated for many generations, and assign a lower priority for mitigation to the area separating

them compared to areas of historic connectivity.

Interpretation of genetic population structure in the context of direct evidence from field research

provides a more holistic view of a species’ behavior and may highlight knowledge gaps in both

types of research. The objective of this study is to test congruence of detectable barriers to gene

flow in three species of sympatric freshwater turtles that have differing dispersal abilities. I test

the hypothesis that species with higher vagility experience fewer barriers to gene flow, and I use

Bayesian analyses from Chapters 3, 4 and 5 and analyses based on Monmonier’s algorithm to

identify common genetic boundaries and barriers to gene flow among species.

112

112

7.2 Methods

7.2.1 Study species and relative dispersal ability

I compared genetic population structure in the spotted turtle (Cl. guttata), the Blanding’s turtle

(Emys (=Emydoidea) blandingii) and the snapping turtle (Chelydra serpentina). These species

have similar generation times, but direct evidence from radio-telemetry studies demonstrates that

their vagility differs substantially. Clemmys guttata typically move less than 500 m in a year and

rarely move farther than 2 km (Litzgus 1996; Rasmussen and Litzgus 2010; Ernst and Lovich

2009; Banning-Anthonysamy 2012). Chelydra serpentina may undertake movements of >10 km

between wetlands or to find a suitable nesting site, although overland movements are typically

shorter (summarized in Ernst and Lovich 2009; Obbard and Brooks 1980; J. Paterson pers.

comm.). Emys blandingii may also migrate several kilometres to nest and have been recorded

migrating > 10km overland (COSEWIC 2005; Power 1989). This direct evidence was used as a

proxy for vagility and I categorized the dispersal ability of Cl. guttata, Ch. serpentina and E.

blandingii as being low, moderate or high, respectively (Table 7.1).

7.2.2 Bayesian delineation of population boundaries

Microsatellite data were compiled from three previous studies, using 11 loci for Cl. guttata

(Chapter 3, N = 253), 10 for Ch. serpentina (Chapter 4, N = 167) and 12 for E. blandingii

(Chapter 5, N = 91). Sampling sites are shown in Figure 7.1.

Population differentiation was calculated using Dest (Jost 2008) and Nei’s absolute differentiation

(DST, Nei, 1973) in SMOGD (Crawford 2010) and MSANALYZER (Dieringer and Schlötterer

2003). Populations were defined based on Bayesian inference in the programs STRUCTURE

(Pritchard et al. 2000) and TESS (Chen et al. 2007), as described in Chapters 3–5. Genetically

distinct clusters from each species were used as independent units (“genetic populations”) for

barrier estimation (Figure 7.1, inset).

Effective population sizes of E. blandingii populations were estimated in ONeSAMP

(Tallmon et al. 2008) and compared to estimates for Cl. guttata and Ch. serpentina (Chapter 4)

using a one-way ANOVA in SPSS v.20.0 (SPSS Inc., Chicago, Illinois).

113

113

7.2.3 Barrier estimation with Monmonier’s algorithm

Barriers to gene flow were also estimated for each data set using Monmonier’s maximum

distance algorithm and Delaunay triangulation implemented in BARRIER V.2.2 (Manni et al.

2004). BARRIER used pairwise matrices of genetic and geographic distance among sampling sites

to infer barriers to gene flow relative to site locations. These analyses were based on measures of

pairwise genetic distance (Nei’s D, Dest and FST) among populations. Because measures of

population differentiation are sensitive to small sample sizes (Kalinowski 2005), only

populations with N > 12 were included in these analyses.

The analysis was run first with Dest matrices (Jost 2008) because Dest provided the most accurate

available measure of population differentiation. Bootstrap replicates were required to test

significance of barriers, but these could not be calculated for Dest (N. Crawford, pers. comm.).

Therefore, the analysis was re-run with Nei’s absolute difference (DST, Nei 1973) to verify

congruence between barriers based on the two measures. Finally, 5,000 bootstrap replicates of

DST were used to determine the significance of each inferred barrier. Bootstrap support > 0.90

was considered significant.

7.2.4 Estimation of migration among populations

The average historical number of migrants per generation (Nm) was calculated for each pair of

genetically differentiated clusters within each species following the private alleles method of

Barton and Slatkin (1986). This value is a historical average of the number of individuals

exchanged among populations per generation and it does not represent contemporary gene flow.

Rather, it provides a basis for comparison of historic, genetic population connectivity among

species. Estimates of Nm and pairwise distances between sites were log-transformed to achieve a

normal distribution. Pearson`s correlation coefficient was used to test the relationship between

geographic distance and Nm.

Estimates of Nm were also compared directly among the four areas where sufficient samples

were available from all three species: LE1, GH, GB/PSD and EO1. For Ch. serpentina,

subpopulation 3 was used for GB/PSD comparisons (Figure 7.1, inset; Fig. 2). These data

remained non-normal after transformation. Therefore, I tested for differences in Nm among

species with Friedman’s test for related samples in SPSS v.20.0 (SPSS Inc., Chicago, Illinois).

114

114

The number of first-generation migrants in each population was estimated in GENECLASS

V. 2.0 (Piry et al. 2004) using the Bayesian method of Rannala and Mountain (1997), and re-

sampling 100,000 times with the Markov Chain Monte Carlo method of Paetkau (2004).

GENECLASS estimated the likelihood of each individual’s genotype originating in the population

where it was sampled (L = L_home, the likelihood of sampling an individual`s genotype in a

population based on the genetic profile of that population). The estimate L= L_home was

appropriate in this case because it does not assume that all existing populations were sampled

(Piry et al. 2004).

I considered also applying the Bayesian method of Wilson and Rannala (2003) to estimate

contemporary gene flow. However, Faubert et al. (2007) showed that this method performed

poorly when FST < 0.05 and several tested population pairs met this criterion (Chapter 3; 4; 5).

7.3 Results

7.3.1 Bayesian delineation of population boundaries

Comparison of population structure inferred previously with STRUCTURE and TESS (Chapters 3,

4, 5) revealed a substantial lack of geographic congruence in inferred boundaries among species

(Figure 7.2). For example, three sampled sites along the shore of Lake Huron (LH1, LH2 and

BP) were clustered differently in Cl. guttata (LH1 and LH2 vs. BP) than in Ch. serpentina (LH1

vs. LH2 and BP). Samples from GH formed a potentially distinct subpopulation in E. blandingii,

while GH grouped with samples from the northwest shore of Lake Erie in Cl. guttata, and

grouped with Georgian Bay and the Bruce Peninsula in Ch. serpentina.

A general east-west split occurred in all three species but its location was inconsistent. In Cl.

guttata, STRUCTURE resolved HC, EO1 and EO2 into a single eastern cluster at K = 2 and

grouped all other samples together. In Ch. serpentina, all samples from LH2 eastwards,

including GH, cluster together at K = 2. In E. blandingii, LE and GH separate from all other

samples at K = 2.

Effective population sizes estimated in ONeSAMP did not differ significantly among species

(ANOVA: F = 0.165, d.f = 2, p = 0.850). Average estimated Ne and ranges of the estimates for

each species are listed in Table 7.2.

115

115

7.3.2 Barrier estimation with Monmonier’s algorithm

Monmonier’s algorithm resolved significant barriers for each species (Figures 7.3, 7.4) and

indicated further barriers that approached significance. Barriers showed some potential overlap

near the base of the Bruce Peninsula but were otherwise spatially dissimilar. Monmonier’s

algorithm detected fewer significant boundaries in each dataset than Bayesian clustering.

7.3.3 Estimation of migration among populations

GENECLASS identified four potential first-generation migrants (p < 0.01) in samples of Cl.

guttata. Two of these (one each from LH2 and EO2) were assigned most strongly to their source

population, indicating either that they were migrants from unknown, unsampled populations or

that they were not in fact migrants. One Cl. guttata from LE1 was implicated as a potential

migrant from GB1, and an individual from EO1 was implicated as a potential migrant from EO2.

Estimates of Nm between sites ranged from 0.33 to 3.03, with an overall Nm of 1.74 among all

sampled sites (Table 7.3a).

No first generation migrants were detected among sampled Ch. serpentina populations (α =

0.01). The value of Nm between populations A and B was 3.78; average Nm among all sampled

sites was 2.62. Pairwise Nm among subpopulations ranged from 4.585 to 1.016 (Table 7.3b).

GENECLASS detected 27 E. blandingii as possible first-generation migrants (p < 0.01); of these,

13 were assigned most strongly to their population of origin. The 14 others included three

potential migrants in PSD (two from LE, two from KAW); three in GH (one from LE, two from

PSD), four in KAW (three from PSD, one from GH), one in LE (from PSD), and two in EO (one

from PSD, one from KAW). Estimates of Nm ranged from 0.952 to 3.380 (Table 7.3c).

Geographic distance was not correlated with Nm within any species or over all species

(Pearson`s correlation coefficient, r = -0.114, N = 66, p = 0.363). Estimates of Nm did not differ

among species at the four sites tested with Friedman’s test (N = 6, d.f. = 2, χ2 = 4.000, p = 0.135,

Figure 7.5) or when comparing the means of all pairwise Nm among species (ANOVA: F =

1.874, d.f. = 2, p = 0.162).

116

116

7.4 Discussion

Genetic structure and barriers to gene flow in three sympatric species of turtle sampled across >

500 km show remarkably little congruence. This variation represents the most important finding

of this study and demonstrates that genetic population structure from one species cannot predict

structure in similar species in the same landscape. Furthermore, variance in patterns of gene flow

and diversity cannot be explained simply by variation in the dispersal ability of these three

species. These results are inconsistent with predictions based on previous studies of these

species’ spatial ecology and behavior, and overall have important implications for the

conservation of genetic diversity in communities of threatened species.

7.4.1 Comparative landscape genetics of freshwater turtles

A general east-west break occurs in Cl. guttata, Ch. serpentina and E. blandingii across southern

Ontario, with further sub-structuring of populations within each eastern and western cluster

(Figure 7.2a; Chapters 7.3, 7.4, 7.5). However, the location of this break is discordant among the

three species.

In Cl. guttata, significant barriers isolate the populations from Hastings County and the Bruce

Peninsula from their nearest neighbors. The Hastings County samples are differentiated from all

other sampled Cl. guttata (Chapter 3). Given similar patterns of differentiation recorded in

channel darters from the same watershed (Kidd et al. 2011), it would be informative to sample

Ch. serpentina and E. blandingii from this area as well. Unfortunately, samples of Ch. serpentina

and E. blandingii from Hastings County were not available for this study. The barrier isolating

Cl. guttata in the Bruce Peninsula is not reflected in Ch. serpentina because samples from LH2

occur in the same genetic population as BP, GB and N. However, individual-based analyses in

STRUCTURE and TESS show significant differentiation between Ch. serpentina from LH1 and

LH2, sites between which Cl. guttata are genetically continuous (Figure 7.2). Thus, dispersal

along the Lake Huron shoreline is disrupted in both Cl. guttata and Ch. serpentina, but in

different places. These patterns likely reflect differing colonization routes following the end of

the last ice age, 6–4 ka BP, because the lag time needed to detect effects of genetic barriers may

be as long as 200 generations (Landguth et al. 2010; > 5,000 years for these species). The current

landscape may be maintaining this genetic structure or the removal of a previous barrier may

117

117

have occurred but may not yet be detectable. Given that Cl. guttata currently occurs in only a

few disjunct populations in Ontario (the majority of them sampled here), it is more likely that the

remaining extant populations of Cl. guttata will remain isolated.

The western barrier to gene flow inferred for E. blandingii extends towards the base of the Bruce

Peninsula and coincides roughly with the shift in allele frequencies in Ch. serpentina between

sites LH1 and LH2. Resolution of this barrier’s location is especially limited because only one

sample of E. blandingii was obtained from site LH1 in four years of extensive mark-recapture

surveys. Emys blandingii was common at LH1 in the mid-1990s (J. Skevington, pers. comm.)

but this population has apparently declined severely over the last 10 years (C. Davy, unpublished

data). North of LH1, E. blandingii becomes rare. Only four records of the species exist from the

Bruce Peninsula (Ontario Nature Reptile and Amphibian Atlas; J. Paterson and J. Urquhart, pers.

comm.). No robust populations are known from the Bruce Peninsula despite the presence of

suitable habitat (J. Crowley, pers. comm.) and these four reports may represent released animals

or rare long-distance migrants. Therefore, there may be a small E. blandingii population on the

Bruce Peninsula but the lack of large populations north of LH1 indicates a real gap in

distribution rather than a sampling bias. This gap is consistent with the placement of the inferred

barrier to gene flow in E. blandingii south of LH2.

Overall, the landscape of south-western Ontario was apparently more permeable to Cl. guttata

than to E. blandingii or Ch. serpentina, while the opposite pattern occurs in eastern Ontario

(from Parry Sound district eastwards). South-western Ontario is characterized by sand and clay-

soil substrates, while a large portion of central and eastern Ontario (including PSD, Alg, KAW,

EO2 and EO3) is located on the Canadian Shield. The observed pattern suggests potential

variation in landscape permeability among these species, and this pattern deserves further

consideration.

Estimates of Nm < 10 among all populations indicate that none of the sampled populations are in

drift connectivity, the genetic connectivity required to maintain approximately equal allele

frequencies among populations (Lowe and Allendorf 2010). Lack of drift connectivity across the

study area is also indicated by Bayesian clustering analyses that indicate K > 1 for all three

species. Maintenance of inbreeding connectivity, the genetic connectivity required to prevent

118

118

inbreeding depression, requires a minimum of one migrant per generation (Mills and Allendorf

1996). Each species in this study has Nm < 1.0 between one or more pairs of sites. However,

some of the tested sites are hundreds of kilometers apart, and they do not represent all extant

populations of these species in southern Ontario. The occurrence of Nm > 1.0 in all species

between two or more sites demonstrates that freshwater turtles were historically able to maintain

low levels of gene flow across large landscapes despite evidence for significant barriers to gene

flow.

Estimates of Nm based on private alleles (Barton and Slatkin 1986) represent an average number

of immigrants exchanged between populations per generation. This estimate is a historic average

and cannot reflect the effects of severe habitat destruction in southern Ontario in the past 200

years. As a result, it is surprising that apparent dispersal ability does not appear to predict Nm or

the number of barriers to gene flow estimated by Monmonier’s algorithm. The average Nm

among populations did not differ among species. Monmonier’s algorithm estimated two barriers

for Emys blandingii, a single barrier for Ch. serpentina surrounding the population at the Golden

Horseshoe, and only three barriers for Cl. guttata., which was less than expected based on this

species’ apparently low dispersal tendency. Clustering in TESS and STRUCTURE identified a

greater number of genetic clusters than were inferred based on boundary estimation in BARRIER

(consistent with the findings of Blair et al. 2012). However, the clusters estimated by TESS and

STRUCTURE show similar incongruence among species to the barriers estimated using

Monmonier’s algorithm. All three analyses support a hypothesis of greater historic landscape

permeability in south-western Ontario for Cl. guttata, and in central and eastern Ontario for E.

blandingii and Ch. serpentina, as noted above.

The long generation times of turtles may result in sufficient movement per generation to

maintain migration rates between distant populations, even in species with low vagility.

However, perhaps our understanding of vagility, which influences demographic connectivity of

populations, is not a good predictor of actual gene flow across the landscape, which influences

genetic connectivity (Lowe and Allendorf 2010). The genetic results are somewhat

counterintuitive when considered in the context of the relative vagility of the species. For

example, although populations of Cl. guttata are not in drift connectivity, sufficient gene flow

exists (or existed before significant landscape modification occurred) to prevent significant loss

119

119

of alleles from populations. The relatively low vagility of Cl. guttata has been documented in

numerous studies (e.g. Litzgus 1996; 2004; Seburn 2003; Kaye et al. 2006; Rasmussen and

Litzgus 2010; Yagi and Litzgus 2012) and predicts relatively low gene flow among populations.

Yet although significant structure exists within Cl. guttata in Ontario, no populations are fixed

for alleles at any loci and estimates of Nm are comparable to those for other species. One

possible explanation is that multiple unknown populations exist or existed recently across the

landscape. While there are probably some unknown populations of Cl. guttata in Ontario, I

consider this explanation highly unlikely given the amount of survey effort expended by

professional biologists and amateur naturalists across the province. As discussed in Chapter 4,

non-random mating could also maintain genetic diversity in small populations of Cl. guttata, and

this possibility should be explored further.

On the other hand, Ch. serpentina is relatively widespread across southern Ontario, and

telemetry studies regularly record movements of many kilometers within a single year (e.g.

Obbard and Brooks 1980; Paterson et al. 2012). High gene flow among populations seems

especially likely because females will migrate long distances to nest, which should serve to

disperse their genetic material across large distances. In spite of this apparently high vagility, Ch.

serpentina individuals in subpopulation 3 (SP3; sites LH2, BP, GB and N) are fixed for an allele

at locus Cs18, while individuals from GH are fixed for a single allele at locus Cs22. The

Euclidean distance between sites N and GH is less than 50 kilometers, but the genetic evidence

demonstrates that these sites have been isolated for several generations. A combination of direct

evidence (radio telemetry) and further genetic sampling targeted along boundaries between

identified populations could shed light on the mechanisms that maintain genetic differentiation

on small spatial scales in a species with apparently high dispersal ability.

7.4.2 Long-lived organisms and landscape genetics

Landscape genetics strives to understand the effect of landscape features on the genetic structure

of populations (Manel et al. 2003; Holderegger and Wagner 2008). This is an important and

appealing objective, especially in the context of current, rapid anthropogenic landscape

modification (e.g. Amos et al. 2012). However, evidence for the effects of specific, recent

landscape modifications on long-lived freshwater turtles is either equivocal or lacking (e.g. Kuo

and Janzen 2004; Marsack and Swanson 2010; Bennett et al. 2010; Pittman et al. 2011; Banning-

120

120

Anthonysamy 2012). Based on their long generation times, simulation studies predict this result

(Landguth et al. 2010), a point that many of these studies independently acknowledge.

Simulation studies (Balkenhol et al. 2009; Landguth et al. 2010; Blair et al. 2012) show that it is

inadvisable to investigate the impact of a specific, recent landscape modification (or

modifications) on long-lived organisms by testing for differences in allele frequencies around the

potential “barrier” because a genetic signature can take many generations to develop. This

renders hypotheses about the effects of new barriers to gene flow impossible to test based solely

on genetic data. At the very least, such studies should simultaneously investigate genetic

structure elsewhere in the landscape to provide a context in which the data can be more

accurately interpreted. Ideally, direct evidence of changes in demographic connectivity should

also be obtained (for example, from radio-tracking or capture-mark-recapture studies; Lowe and

Allendorf 2010; Segelbacher et al. 2010). Whichever approach is taken, researchers must always

interpret their data with the understanding that genetic signatures in populations of long-lived

organisms generally reflect the ghost of historic landscapes.

When genetic connectivity of long-lived organisms is a question of interest I recommend an

approach similar to the one taken here. Geographically representative and intensive sampling of

populations across a wide geographic range (relative to the dispersal ability of the species) will

avoid sampling within a panmictic area and obtaining uninformative results. Identification of

broad-scale population structure and barriers to gene flow (if possible, using more sophisticated

methods to detect barriers than those used here) will provide the context necessary to study the

long-term effects of potential anthropogenic barriers to gene flow. However, genetic methods

will not detect effects of recent landscape modifications in long-lived organisms or species with

low levels of dispersal (Landguth et al. 2010; Cyr and Angers 2011) and should probably not be

used to do so.

The Introduction of this Chapter provides a brief discussion of resistance layers and the

challenges inherent in their parameterization (Spear et al. 2010; Braunisch et al. 2010).

Landscape resistance describes the relative ease with which a species can move through different

parts of the landscape. Resistance is ideally quantified using direct, empirical measures of the

relative cost of movement through different habitat types across the landscape but may also be

121

121

based on “expert opinion” (Segelbacher et al. 2010). This study illustrates the potential risks of

assigning costs to resistance layers based on expert opinion (Segelbacher et al. 2010) because an

“expert opinion” of landscape resistance based on vagility of Cl. guttata, Ch. serpentina and E.

blandingii would be inconsistent with the genetic data.

To my knowledge, resistance layers have never been quantified for any species of turtle.

Parameterization of resistance layers for turtles will require specific data that could be collected

alongside ongoing field studies. Such analyses could be extremely informative, but the challenge

is the accurate parameterization of resistance layers. In less robust organisms such as

amphibians, possible correlates of landscape resistance include the relative risk of dehydration in

different habitat types (see Mazerolle and Desrochers 2005; Stevens et al. 2006). No obvious

corollary exists for turtles, but possible measurements could include the relative probability that

different species will cross roads and highways of various sizes or move between patches of

suitable habitat separated by agricultural, urbanized, forested, and other less suitable habitat

types. Variation among populations and habitat types is a further challenge because spatial

ecology and habitat preferences may vary across the range of a species, between sexes and

among individuals (Litzgus et al. 2004; Edge et al. 2010; Rasmussen and Litzgus 2010; Paterson

et al. 2012 ). Temporal variation in spatial ecology and dispersal behavior may also occur as a

landscape changes over time (Yagi and Litzgus 2012). Nevertheless, finding a way to

parameterize resistance layers for turtles and combining these with in-depth sampling will

provide a more detailed understanding of demographic and genetic connectivity in natural and

modified landscapes.

7.4.3 Conservation implications

Maintenance of genetic population structure includes increasing gene flow among historically

connected, recently isolated sites, and avoiding increased gene flow among historically isolated

sites (Frankham et al. 2002). Comparative population genetics of multiple species allows

identification of shared areas of high gene flow among populations and facilitates the

prioritization of areas for mitigation measures. These could include wildlife corridors, highway

underpasses and restoration of riparian zones that might decrease landscape resistance for the

three species. Unfortunately, I detected no substantial overlap in areas of gene flow between

122

122

sampling sites, which makes it more difficult to suggest mitigation measures that will impact all

three species equally.

On the other hand, some overlap occurred in the approximate locations of barriers: all three

species appear to have experienced historic barriers to gene flow in the area south of the Bruce

Peninsula. Artificially increasing connectivity between historically isolated populations can alter

existing genetic structure and decrease overall genetic diversity by genetically homogenizing the

metapopulation (Frankham et al. 2002). Facilitated breeding among historically isolated

populations may also lead to outbreeding depression that can cause reduced fitness of offspring

from differentiated populations (Templeton 1986). Outbreeding depression is unlikely among

recently diverged populations (Frankham et al. 2011), but it only takes a few generations to alter

existing genetic structure of a population. Thus, an area where several species experience a

barrier to dispersal and gene flow would be a poor choice for large-scale measures to increase

connectivity.

Similar disparity in population structure among species occurs in two sympatric snakes in

southwestern Ontario (DiLeo et al. 2010). Discordant patterns of gene flow and population

structure in the eastern garter snake (Thamnophis sirtalis) and the eastern foxsnake (Mintonius

gloydi) may result from differing effects of habitat fragmentation causing drastically different

landscape permeability in the two species. Discordant patterns may result from differing

effective population size (Ne), because populations with smaller Ne are affected more strongly

by genetic drift and diverge more quickly as a result (DiLeo et al. 2010). However, my results

show overall discordance among species without comparable differences in Ne. It is unlikely that

any single factor can explain the lack of correspondence observed here.

Thus, the most important finding of this study is the overall disparity of genetic population

structure among species. Analysis of genetic population structure is time-consuming and costly,

and the ability to generalize genetic population structure from a studied species to other, similar

species would be very useful. However, my results demonstrate that population structure of one

species cannot predict structure in another. The three species of turtle sampled here have

different microhabitat preferences but have similar current distributions and are sympatric in

many locations. They share similar post-glacial colonization histories and life-history strategies

123

123

(Holman 1992; Chapter 4). Yet populations of these three species are structured very differently

across the landscape, with patterns of genetic structure that are inconsistent with our

understanding of their spatial ecology based on direct evidence from field studies.

The identification of historic and persistent barriers to gene flow can highlight gaps in our

knowledge of well-studied species; in this case, genetic methods indicate that current

assumptions about the relative dispersal abilities of three species of turtle may be inaccurate.

Directions for further research include the combination of targeted genetic sampling along

boundaries between populations with field studies of spatial ecology. Integrating genetic

methods with field studies will allow testing of further hypotheses about fine-scale patterns of

gene flow across the landscape in these three species and result in a more holistic understanding

of their biology.

Finally, consideration must be given to the low effective population sizes of turtles in Ontario.

All but two sampled populations have Ne < 50, the often quoted theoretical lower limit required

to avoid the short-term deleterious effects of inbreeding (Franklin 1980). The traditional estimate

for effective population size required to maintain genetic diversity in the long-term is 500

(Franklin 1980). In wild populations the minimum effective population size actually required for

long-term persistence varies substantially among species and is likely to be significantly larger

than the “50:500 rule” suggests (Traill et al. 2007; Traill et al. 2010). Thus, effective population

sizes for turtles in Ontario – including Ch. serpentina, which until recently was called the

“common” snapping turtle – are probably too low to avoid the genetic impacts of population

decline over the coming generations. Demographic impacts may prove more harmful to

populations than a gradual increase in inbreeding or loss of allelic richness, but these data

provide further evidence that rapid action is required to conserve these long-lived but highly

threatened species.

7.5 References Adriaensen F, Chardron JP, DeBlust G, Swinnend E, Villalbad S, Gulinckd H, Matthysen E

(2003) The application of ‘least-cost’ modelling as a functional landscape model. Landscape Urban Plan 64:233–247

Amos JN, Bennett AF, MacNally R, Newell G, Pavlova A, Radford JQ, Thomson JR, White M, Sunnucks P (2012) Predicting landscape-genetic consequences of habitat loss, fragmentation and mobility for multiple species of woodland birds. PLoS ONE 7: e30888

124

124

Balkenhol N, Waits LP, Dezzani RJ (2009) Statistical approaches in landscape genetics: an evaluation of methods for linking landscape and genetic data. Ecography 32:818–830


Barton NH, Slatkin M (1986) A quasi-equilibrium theory of the distribution of rare alleles in a subdivided population. Heredity 56:409–416

Bennett AM, M Keevil, JD Litzgus (2009) Demographic differences among populations of Northern Map Turtles (Graptemys geographica) in intact and fragmented sites. Can J Zool 87:1147–1157

Bennett AM, Keevil M, Litzgus JD (2010) Spatial ecology and population genetics of northern map turtles (Graptemys geographica) in fragmented and continuous habitats in Canada. Chel Conserv Biol 9:185–195

Blair C, Weigel DE, Balazik M, Keeley ATH, Walker FM, Landguth E, Cushman S, Murphy M, Waits L, Balkenhol N (2012) A simulation-based evaluation of methods for inferring linear barriers to gene flow. Mol Ecol Res 12:822–833

Braunisch V, Segelbacher G, Hirzel AH (2010) Modelling functional landscape connectivity from genetic population structure: a new spatially explicit approach. Mol Ecol 19:3664-3678

Chen C, Durand E, Forbes F, François O (2007) Bayesian clustering algorithms ascertaining spatial population structure: a new computer program and a comparison study. Mol Ecol Notes 7:747–756

Congdon JD, Graham TE, Herman TB, Lang JW, Pappas MJ, Brecke BJ (2008) Emydoidea blandingii (Holbrook 1838) – Blanding’s turtle. In: Rhodin AGJ, Pritchard PCH, van Dijk PP, Saumure RA, Buhlmann KA, Iverson JB (eds.). Conservation biology of freshwater turtles and tortoises: a compilation project of the IUCN/SSC Tortoise and Freshwater Turtle Specialist Group. Chelon Res Monogr 5:015.1-015.12. doi:10.3854/crm.5.015.blandingii.v1.2008, http://www.iucn-tftsg.org/cbftt/

COSEWIC (2004) COSEWIC assessment and update status report on the spotted turtle Clemmys guttata in Canada. Committee on the Status of Endangered Wildlife in Canada. Ottawa. vi + 27 pp. (www.sararegistry.gc.ca/status/status_e.cfm).



Crawford NG (2010) SMOGD: software for the measurement of genetic diversity. Mol Ecol Resour 10:556–557

125

125

Cyr F, Angers B (2011) Historical process lead to false genetic signal of current connectivity among populations. Genetica 139:1417–28

DiLeo MF, Row JR, Lougheed SC (2010) Discordant patterns of population structure for two co-distributed snake species across a fragmented Ontario landscape. Divers Distrib 16:571–581

Dieringer D, Schlötterer C (2003) Microsatellite analyser (MSA): a platform independent analysis tool for large microsatellite data sets. Mol Ecol Notes 3:167–169

Edge CB, Steinberg BD, Brooks RJ, Litzgus JD (2010) Habitat selection by Blanding’s Turtles (Emydoidea blandingii) in a relatively pristine landscape. Écoscience 17:90–99

Epps CW, Wehausen JD, Bleich VC, Torres SG, Brashares JS (2007) Optimizing dispersal and corridor models using landscape genetics J Appl Ecol 44:714–724

Ernst CH, Lovich JL (2009) Turtles of the United States and Canada, 2nd ed, Johns Hopkins University Press, Baltimore, Maryland

Faubert P, Waples R, Gaggiotti O (2007) Evaluating the performance of a multilocus Bayesian method for the estimation of migration rates. Mol Ecol 16:1149–1166

Frankham R, Ralls K (1998) Inbreeding leads to extinction. Nature 392:441–442

Frankham R, Ballou JD, Briscoe DA (2002) Introduction to conservation genetics. Cambridge University Press, Cambridge

Frankham R, Ballou JD, Eldridge MDB, Lacy RC, Ralls K, Dudash MR, Fenster CB (2011) Predicting the probability of outbreeding depression. Conserv Biol 25:465–475

Franklin R (1980) Evolutionary change in small populations. In: Soulé ME, Wilcox BA (eds.), Conservation Biology: An Evolutionary Ecological Perspective. Sinauer Associates, Sunderland, Massachusetts, pp135–140

Galbraith DA, Chandler MW, Brooks RJ (1987) The fine structure of home ranges of male Chelydra serpentina: are snapping turtles territorial? Can J Zool 65:2623–2629

Galbraith DA, Brooks RJ, Obbard ME (1989) The influence of growth rate on age and body size at maturity in female snapping turtles Chelydra serpentina. Copeia 1989:896–904

Goldberg CS, Waits LP (2010) Comparative landscape genetics of two pond-breeding amphibian species in a highly modified agricultural landscape. Mol Ecol 19:3650–3663

Guillot G, Mortier F, Estoup A (2005) GENELAND: a program for landscape genetics. Mol Ecol Notes 5:712–715

Guillot G, Leblois R, Coulon A, Frantz AC (2009) Statistical methods in spatial genetics. Mol Ecol 18:4734–4756

Holderegger R, Wagner HH (2008) Landscape genetics. BioScience 58:199–207

Holman JA (1992) Late Quaternary herpetofauna of the central Great Lakes region, U.S.A.: zoogeographical and paleoecological implications. Quaternary Sci Rev 11:345–351

Jost L (2008) GST and its relatives do not measure differentiation. Mol Ecol 17:4015–4026

126

126

Kalinowski ST (2005) Do polymorphic loci require large sample sizes to estimate genetic distances? Heredity 94:33–36

Kaye DR, Walsh KM, Rulison EL and Ross CC (2006) Spotted turtle use of a culvert under relocated Route 44 in Carver, Massachusetts. In: Irwin CL, Garrett P, McDermott KP (eds) Proceedings of the 2005 International Conference on Ecology and Transportation. Center for Transportation and the Environment, North Carolina State University, Raleigh, NC, pp 426-432

Kidd A, Reid S, Wilson C (2011) Local and regional population genetic structure of the threatened channel darter in Ontario and Quebec. Poster presentation at American Fisheries Society 41st Annual Meeting, Seattle, Washington, Sept. 4–8, 2011.


Landguth EL, Cushman SA, Schwartz MK, McKelvey KS, Murphy M, Luikart G (2010) Quantifying the lag time to detect barriers in landscape genetics. Mol Ecol 19:4179–4191

Litzgus JD (1996) Life history and demography of a northern population of spotted turtles, Clemmys guttata. MSc Thesis, University of Guelph, Ontario. 145 pp.

Litzgus JD, Mousseau TA, Lannoo MJ (2004) Home range and seasonal activity of southern spotted turtles (Clemmys guttata): implications for management. Copeia 2004:804–817

Litzgus JD (2006) Sex differences in longevity in the spotted turtle (Clemmys guttata). Copeia 2006:281–288

Lowe WH, Allendorf FW (2010) What can genetics tell us about population connectivity? Mol Ecol 19:3038–3051

Manel S, Schwartz MK, Luikart G, Taberlet P (2003) Landscape genetics: combining landscape ecology and population genetics. Trends Ecol Evol 18:189–197

Manni F, Guérard E, Heyer E (2004) Geographic patterns of (genetic, morphologic, linguistic) variation: how barriers can be detected by “Monmonier’s algorithm”. Hum Biol 76:173–190

Marsack K, Swanson BJ (2009) A genetic analysis of the impact of generation time and road-based habitat fragmentation on eastern box turtles (Terrapene c. carolina). Copeia 2009:647–652

Mazerolle MJ, Desrochers A (2005) Landscape resistance to frog movements. Can J Zool 83:455–464

McRae BH (2006) Isolation by resistance. Evolution 60:1551–1561

Mills LS, Allendorf FW (1996) The one-migrant-per-generation rule in conservation and management. Conserv Biol 10:1509–1518

Monmonier M (1973) Maximum-difference barriers: An alternative numerical regionalization method. Geogr Anal 3:245–261

Nei M (1973) Analysis of gene diversity in subdivided populations. Proc Nat Acad Sci USA 70:3321–3323

127

127

O’Brien D, Manseau M, Fall A, Fortin MJ (2006) Testing the importance of spatial configuration of winter habitat for woodland caribou: an application of graph theory. Biol Conserv 130:70–83

Obbard ME, Brooks RJ (1980) Nesting migrations of the Snapping Turtle (Chelydra serpentina) Herpetologica 36:158–162

Paetkau D, Slade R, Burden M, Estoup A (2004) Direct, real-time estimation of migration rate using assignment methods: a simulation-based exploration of accuracy and power. Mol Ecol 13:55–65

Paterson JE, Steinberg B, Litzgus JD (2012) Generally specialized or especially general? Habitat selection by snapping turtles (Chelydra serpentina) in central Ontario. Can J Zool 90:139–149

Piry S, Alapetite A, Cornuet JM, Paetkau D, Baudouin L, Estoup A (2004) GeneClass2: A software for genetic assignment and first-generation migrant detection. J Hered 95:536–539


Power TD (1989) Seasonal movements and nesting ecology of a relict population of Blanding’s turtles (Emydoidea blandingii) in Nova Scotia. M.Sc. Thesis, Acadia University, Wolfville, Nova Scotia

Pritchard JK, Stephens M, Donnelly PJ (2000) Inference of population structure using multilocus genotype data. Genetics 155:945–959

Rannala B, Mountain JL (1997) Detecting immigration by using multilocus genotypes. Proc Natl Acad Sci USA USA 94:9197–9221

Rasmussen ML, Litzgus JD (2010) Habitat selection and movement patterns of spotted turtles (Clemmys guttata): effects of spatial and temporal scales of analyses. Copeia 2010:86–96

Seburn, D.C., 2003. Population structure, growth, and age estimation of spotted turtles, Clemmys guttata, near their northern limit: an 18-year follow-up. Can Field Nat 117:436–439

Segelbacher G, Cushman SA, Epperson BK, Fortin MJ, Francois O, Hardy OJ, Holderegger R, Taberlet P, Waits LP, Manel S (2010) Applications of landscape genetics in conservation biology: concepts and challenges. Conserv Genet 11:375–385.

Spear SF, Balkenhol N, Fortin MJ, McRae B, Scribner K (2010) Use of resistance surfaces for landscape genetic studies: considerations for parameterization and analysis. Mol Ecol 19:3576–3591

Stevens VM, Verkenne C, Vandewoestijne S, Wesselingh RA, Baguette M (2006) Gene flow and functional connectivity in the natterjack toad. Mol Ecol 15:2333–2344

Tallmon DA, Luikart G, Waples RS (2004) The alluring simplicity and complex reality of genetic rescue. Trends Ecol Evol 19:489–496

128

128

Tallmon DA, Koyuk A, Luikart GH, Beaumont MA (2008) ONeSAMP: a program to estimate effective population size using approximate Bayesian computation. Mol Ecol Resour 8:299–301

Templeton AR (1986) Coadaptation and outbreeding depression. In: Soulé ME (ed) Conservation biology: the science of scarcity and diversity. Sinauer, Sunderland, Massachusetts pp 105–116

Thrall PH, Richards CM, McCauley DE, Antonovics J (1998) Metapopulation collapse: the consequences of limited gene-flow in spatially structured populations. In: Bascompte J, Sole RV (eds.) Modeling Spatiotemporal Dynamics in Ecology, Springer Verlag, Berlin, pp 83–104

Traill LW, Bradshaw CJA, Brook BW (2007) Minimum viable population size: A meta-analysis of 30 years of published estimates. Biol Conserv 139:159–166

Traill, LW, Brook BW, Frankham R, Bradshaw CJA (2010) Pragmatic population viability targets in a rapidly changing world. Biol Conserv 143:28–34

van Dijk PP (2011) Clemmys guttata. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.2. <www.iucnredlist.org>. Downloaded on 23 October 2012.

van Dijk PP (2012) Chelydra serpentina. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.2. <www.iucnredlist.org>. Downloaded on 23 October 2012.

van Dijk PP, Rhodin AGJ (2011) Emydoidea blandingii. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.2. <www.iucnredlist.org>. Downloaded on 23 October 2012.

Wang YH, Yang KC, Bridgman CL, Lin LK (2008) Habitat suitability modeling to correlate gene flow with landscape connectivity. Landscape Ecol 23:989–1000

Wilson GA, Rannala B (2003) Bayesian inference of recent migration rates using multilocus genotypes. Genetics 163:1177–1191

Yagi KT, Litzgus JD (2012) The effects of flooding on the spatial ecology of spotted turtles (Clemmys guttata) in a partially mined peatland. Copeia 2012:179–190

129

Table 7.1. Life history, distribution and behavioral traits of Clemmys guttata, Chelydra serpentina and Emys blandingii. Global

conservation status is determined by the International Union for Conservation of Nature (IUCN); Canadian conservation status is

determined by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC).

Clemmys guttata

Chelydra serpentina Emys blandingii Source

Estimated longevity > 110 years > 100 years > 75 years Litzgus 2006; R. Brooks, unpublished data, in COSEWIC 2008, Congdon 2008.

Estimated generation time > 25 years ~ 31 years > 40 years COSEWIC 2004, 2005, 2008; Galbraith and Brooks 1987; Galbraith et al. 1989

Average clutch size 3.5 35.2 10.7 Ernst and Lovich 2009

Vagility (dispersal ability) based on telemetry data

low moderate high See Methods

Estimated area of occupancy in Canada

<< 2,000 km2 ~ 858,000 km² < 935 km2 COSEWIC 2004, 2005, 2008

Estimated extent of occurrence in Canada

~ 57,500 km² ~ 1,455,000 km² ~ 74,700 km2 COSEWIC 2004, 2005, 2008

Global conservation status (IUCN)

Endangered Least Concern Endangered van Dijk and Rhodin 2011; van Dijk 2011; 2012

Conservation status in Canada (COSEWIC)

Endangered Special Concern Threatened*

* This status refers specifically to the Great Lakes/St. Lawrence population, which is distributed across the area sampled in this study.

COSEWIC considers the Nova Scotia population as a Designatable Unit (DU, Green 2005) and it is listed as Endangered.

130

Table 7.2. Mean and range of effective population size (Ne) estimated in ONeSAMP (Tallmon

et al. 2008) for populations of three freshwater turtles in southern Ontario, Canada. N = number

of populations for which estimates were obtained; mean Ne = mean estimated effective

population size; s.d. = standard deviation; range = minimum – maximum estimate. N is lower

than the number of populations sampled because Ne estimates for sites from which < 20

individuals were sampled were not included (estimated Ne from sites with low sample sizes were

all < 25).

Species N Mean Ne s.d. Range

Lowest 95% CI lower estimate

Highest 95% CI upper estimate

Clemmys guttata 8 36.8 9.0 26.3 - 55.8 22.8 90.5

Chelydra serpentina 5 38.1 15.7 22.3 - 62.8 24.1 104.1

Emys blandingii 3 29.1 2.7 23.7 - 32.5 20.6 78.3

131

Table 7.3. Average historic number of effective migrants per generation (Nm, Barton and

Slatkin 1986).

A) Clemmys guttata LH1 LH2 BP LE1 LE2 GB GH HC EO1 EO2

LH1 -- 1.646 0.947 1.081 0.631 1.171 1.146 0.479 0.517 0.644

LH2 -- 0.942 1.258 0.942 1.303 1.193 0.812 0.91 1.468

BP -- 0.835 0.967 1.263 1.959 0.519 0.622 0.685

LE1 -- 1.568 1.082 2.129 0.736 0.988 0.64

LE2 -- 2.178 3.033 0.52 1.079 0.856

GB -- 2.248 1.048 1.356 1.184

GH -- 1.19 1.273 1.361

HC -- 0.333 0.496

EO1 -- 0.905

Nm (overall) = 1.737

C) Emys blandingii LE GH PSD EO1

LE1 -- 1.35 1.442 1.029 GH -- 0.952 1.013

PSD -- 3.380 Nm overall = 2.79

B) Chelydra serpentina Population 1 . Population 2 . SP 1 SP 2 SP 3 SP 4 GH EO1

Subpopulation 1 -- 2.850 1.016 4.585 0.989 2.287 Subpopulation 2 -- 1.289 2.027 0.488 1.231 Subpopulation 3 -- 1.344 0.847 0.857 Subpopulation 4 -- 0.496 1.805

GH -- 0.831 EO1 --

Nm (Pop A – Pop B) = 3.777 Nm (overall) = 2.624

132

Figure 7.1. Sampling sites and groupings of sites used for BARRIER analysis. Colors indicate the

sampled species at each site: yellow = Clemmys guttata, blue = Chelydra serpentina, red = Emys

blandingii. Small triangles indicate individual samples from otherwise unsampled sites. Squares

with dashed lines indicate the four areas used for comparison of Nm estimates. BARRIER analysis

considered genetically continuous samples with N > 12 as sampling units (inset, bottom right,

based on STRUCTURE results). ALG = Algonquin Provincial Park; BP = Bruce Peninsula; EO =

Eastern Ontario; GB = Georgian Bay; GH = Golden Horseshoe; HC = Hastings County; KAW =

Kawartha Lakes; LE = Lake Erie; LH = Lake Huron; LO = Lake Ontario; N = area north of GH

and south of GB; PSD = Parry Sound District. SP = subpopulation. Base map modified from


Documentation license.

133

134

Figure 7.2 (previous page). Genetic population structure inferred in A) TESS, and B) STRUCTURE for Clemmys guttata (yellow/brown),

Chelydra serpentina (blue), and Emys blandingii (red). SP = subpopulation. See Figure 7.1 for explanation of site abbreviations. Inferred

clusters are plotted on maps to the right of each set of results. Division of Cl. guttata samples under a K = 2 model (implemented in

STRUCTURE; see Chapter 3) is shown by a dashed black line on the bar plot and the map for comparison.

135

Figure 7.3. Barriers to gene flow identified with Monmonier’s algorithm. Colored numbers indicate sampling sites; thin green lines

indicate boundaries between populations based on Delaunay triangulation. A) Clemmys guttata (yellow; N = 253); B) Chelydra serpentina

(blue; N = 167); and C) Emys blandingii (red; N = 91). Estimates are based on 5,000 bootstrap replicates of genetic distance matrices

(Nei’s absolute distance). The thickness of each line and the numbers in black text indicate the strength of bootstrap support. D) Barriers

and sampling sites for the three species overlaid on top of one another; barriers with bootstrap support > 0.90 are marked with a dashed

line.

136

Figure 7.4. Significant barriers (bootstrap support > 0.90) inferred using Monmonier’s algorithm

for Clemmys guttata (yellow dashed lines), Chelydra serpentina (blue dashed lines) and Emys

blandingii (red dashed lines).

137

Figure 7.5. Average number of migrants per generation (Nm) for Clemmys guttata, Chelydra

serpentina and E. blandingii estimated following Barton and Slatkin (1986) among four sites at

which all three species were sampled.

138

Chapter 8 Summary and Conclusions

8

8.1 Summary

In this thesis I developed molecular markers for conservation genetics studies of turtles and used

a suite of analyses to investigate genetic population structure in three freshwater turtle species.

My study species included the spotted turtle, Clemmys guttata, and the Blanding’s turtle, Emys

blandingii. Both species are listed as endangered by the International Union for Conservation of

Nature (IUCN, van Dijk 2011; van Dijk and Rhodin 2011). I also studied the snapping turtle,

Chelydra serpentina, and the spiny softshell, Apalone spinifera. These species are listed as Least

Concern globally, but some populations show evidence of decline due to overharvesting (van

Dijk 2012). Sufficient genetic variation exists in the novel microsatellite loci developed for Ch.

serpentina and A. spinifera to study patterns of genetic variation across landscapes and delimit

management units (Chapters 2, 3, 5; Alacs et al. 2007). Bayesian clustering analyses, principal

coordinates analysis and Monmonier’s algorithm reveal significant population structure in Cl.

guttata, Ch. serpentina and E. blandingii (Chapters 4 – 7).

In the introduction, I stated that conservation genetics studies should result in conclusions or

recommendations that maintain a species’ genetic diversity. Maintenance of genetic diversity in

Cl. guttata, Ch. serpentina and E. blandingii in Ontario will require explicit consideration of

population structure when developing management strategies. Significant shifts in allele

frequencies among populations of all three species demonstrate that they have been isolated for

many generations. Preliminary data from A. spinifera (Chapter 3) also suggests that significant

genetic variation likely occurs among populations (Chapter 3) and this should be investigated

further. These results suggest that any attempts to artificially mix populations (for example,

through translocations) should increase connectivity only between closely related sites to avoid

genetic homogenization of the Ontario meta-populations. Conversely, efforts to maintain genetic

connectivity among related sites through habitat corridors, wildlife underpasses, translocations

and other applied conservation tools could mitigate the gradual genetic impacts of population

declines. Unfortunately, dissimilarity in patterns of gene flow among species requires each

139

species to be assessed independently for such measures – no “one size fits all” solutions appear

to be possible (Chapter 7).

It is encouraging that there is no evidence of inbreeding in turtles in Ontario (Chapters 3, 4 and

5). Fragmentation of turtle populations in Ontario and increased mortality of adults due to

vehicle-related mortality, hunting (legal and illegal), boat propeller strikes and persecution are

well-documented (e.g. COSEWIC 2004; 2005; 2008; OMSTARRT 2012). The genetic impacts

of these effects are not yet detectable, but these impacts are probably inevitable over the next

generations as populations continue to decline. Mitigation measures such as those suggested

above could help to prevent decline in genetic variation.

Although most populations retain relatively high genetic diversity, effective population sizes

(Ne) are low (<100 in all tested populations, and most estimates are < 50; Chapter 4, 5 and 7).

Numerous demographic studies suggest that even relatively large populations of long-lived

turtles can collapse quickly when adult mortality is increased, and that populations of turtles are

extremely slow to recover after dramatic declines (e.g. Congdon et al. 1993; 1994; Cunnington

and Brooks 1996; Heppell 1998; Enneson and Litzgus 2008). These conclusions are supported

by the low effective population sizes of all three species studied here. Demographic threats to

turtles in Ontario are a more immediate concern than gradual loss of genetic variation, but if

variation is not maintained then the long-term persistence of populations may be compromised.

Overall, the species-specific results fill some of the research gaps identified by Alacs et al.

(2007) and noted in the Draft Recovery Strategy for Ontario turtles (Seburn 2007), including the

genetic delineation of management units. My findings also identify areas where further

resolution of genetic structure would be useful. For example, are populations of Ch. serpentina

and E. blandingii in Hastings County also distinct? How are E. blandingii from Algonquin Park

genetically related to individuals from Parry Sound District or the Kawartha Lakes? These

questions are specifically related to management and recovery of these species and to

identification of biogeographic patterns in turtles in Ontario.

These population and location-specific results will be essential to the effective management and

recovery of individual species, but this thesis also tested the application of general conservation

genetics hypotheses to long-lived organisms. Comparisons of multiple populations of Cl. guttata

show that the general relationship between population size and heterozygosity should not be

140

assumed (Chapter 4), even in extremely small populations of endangered species. This

relationship has been demonstrated in many taxa (Frankham 1996), but is not ubiquitous.

Similarly small, isolated populations of endangered species such as Cl. guttata are not

necessarily genetically depauperate compared to large, wide-spread populations of less

threatened species such as Ch. serpentina (Chapter 5). Additionally, the results of Chapters 4, 5

and 6 demonstrate that longevity itself is an insufficient explanation for the apparently high

genetic diversity exhibited by many populations of turtles (Kuo and Janzen 2004; Pittman et al.

2011; Vargas-Ramirez et al. 2012) because high variation exists among similarly long-lived

species. More attention must be given to the role of behaviour in determining reproductive

variance and genetic diversity among species.

Unexpected similarity in Ne:Nc ratios between species may result from behavioural differences,

including variation in nesting behavior, mating systems and paternal contributions to multiply-

sired clutches (Chapter 5). The Ne:Nc ratio can be affected by several factors (see Introduction;

Charlesworth 2009; Jamieson and Allendorf 2012) but in the case of Cl. guttata and Ch.

serpentina, variance in reproductive success seems to be the most likely cause of variation.

Further research will help to confirm or refute this hypothesis.

8.2 Short-term studies of long-lived organisms – directions for future research

Coming to the end of this thesis, I find myself with a long list of potential directions for future

research. The challenges of unravelling evolutionary processes in long-lived organisms include

an inability to study more than one or two generations within a human life-time, so experimental

manipulations such as those used with other species may not be possible. However, the

combination of genetic methods such as those used here with long-term field studies can

gradually clarify the relative role of behaviour in maintaining genetic diversity in fragmented

populations. Analyses of relatedness, mate choice and paternity in Cl. guttata would be

particularly interesting, especially because the small size of most Cl. guttata populations could

allow sampling of >90% of a population.

The unexpectedly low connectivity of some snapping turtle populations (Chapters 4 and 6) raises

interesting questions about the relationship between dispersal and gene flow. Different, fixed

alleles in Ch. serpentina in the Golden Horseshoe and a nearby subpopulation indicate

141

reproductive isolation of nearby sites although the distance between them is small relative to

movements recorded in telemetry studies. Because female Ch. serpentina may travel long

distances to nest, high gene flow is expected across the landscape (Galbraith 2008). However,

Chapters 4 and 6 indicate that genetic structure may occur on relatively small geographic scales.

Targeted sampling along population boundaries combined with radio telemetry may help to

clarify the relative roles of dispersal and gene flow in shaping populations.

The microsatellite markers developed for Ch. serpentina may also facilitate studies of its tropical

congeners Ch. rossignoni and Ch. acutirostris. These species were only recently separated from

Ch. serpentina and, therefore, their biology is poorly understood (Phillips et al. 1996).

Comparative studies of these three closely related species provide an interesting opportunity to

investigate reptile adaptations to temperate and tropical climates.

Comparative studies of several long-lived turtles allowed me to remove the confounding effects

of longevity from inter-species comparisons. A similar comparative approach of turtles with

different life spans would also be informative. Most turtle species are long-lived, but the chicken

turtle (Deirochelys reticularia) may mature in only two to five years (Buhlmann et al. 2009). It

would be interesting to see how genetic structure and the Ne:Nc ratio in this relatively short-

lived species compare to that of other turtles.

The Order Testudines has survived on Earth since the Triassic (Spotila 2004;

http://www.bbc.co.uk/news/science-environment-19872821). Today, many species of turtle are

threatened and the pressures on populations of turtles continue to increase (Turtle Conservation

Coalition 2011). Given these pressures, I would like to finish by expressing my sincere hope that

turtles will persist well into the future. Turtles provide excellent model organisms for studies of

the implications of long-lived life history strategies. Some species provide important ecosystem

services, and turtles are central to many human cultures (e.g. Moll and Jansen 1995; Campbell

2003; COSEWIC 2008; Griffiths et al. 2011). I hope that we will have the opportunity to gain a

better understanding of their endlessly fascinating biology, and that we can continue to share

Ontario and the rest of the planet with these beautiful and complex creatures.


conservation. Chel Res Monogr 4:107–123

142

Buhlmann KA, Congdon JD, Gibbons JW, Greene JL, Mathis A (2009) Ecology of chicken turtles (Deirochelys reticularia) in a seasonal wetland ecosystem: exploiting resource and refuge environments. Herpetologica 65:39–53

Campbell L (2003) Contemporary culture, use, and conservation of sea turtles. In: Lutz PA, Musick JA, Wyneken J (eds.) The biology of sea turtles Vol. 2. Boca Raton, Florida: CRC Press, pp 307–338

Charlesworth B (2009) Effective population size and patterns of molecular evolution and variation. Nat Rev Genet 10:195–205

Congdon, JD, AE Dunham, van Loben Sels RC (1993) Delayed sexual maturity and demographics of Blanding’s turtles (Emydoidea blandingii): implications for conservation and management for long-lived organisms. Conserv Biol 7:826–833

Congdon, JD, Dunham AE, van Loben Sels RC (1994) Demographics of common snapping turtles (Chelydra serpentina): implications for conservation and management of long-lived organisms. Amer Zool 34:397-408

COSEWIC (2004) COSEWIC assessment and update status report on the spotted turtle Clemmys guttata in Canada. Committee on the Status of Endangered Wildlife in Canada. Ottawa. vi + 27 pp. (www.sararegistry.gc.ca/status/status_e.cfm).



Cunnington DC, Brooks RJ (1996) Bet-hedging theory and eigen elasticity: a comparison of the life histories of loggerhead sea turtles (Caretta caretta) and snapping turtles (Chelydra serpentina). Can J Zool 74:291–296

Enneson JJ, Litzgus JD (2008) Using long-term data and a stage-classified matrix to assess conservation strategies for an endangered turtle (Clemmys guttata). Biol Conserv 141:1560–1568

Frankham R (1996) Relationship of genetic variation to population size in wildlife. Conserv Biol 10:1500–1508

Galbraith DA (2008) Population biology and population genetics. In: Steyermark AC, Finkler MS, Brooks RJ (eds) The biology of the snapping turtle. Johns Hopkins University Press, Baltimore, Maryland pp 168–180

Griffiths CJ, Hansen DM, Jones CG, Zuë N, Harris S (2011) Resurrecting extinct interactions with extant substitutes. Curr Biol 21:1–4

Heppell SS (1998) Application of life-history theory and population model analysis to turtle conservation. Copeia 1998:367–375

Jamieson IG, Allendorf FW (2012) How does the 50/500 rule apply to MVPs? Trends Ecol Evol 27:578–584

143


Moll D, Jansen KP (1995) Evidence for a role in seed dispersal by two tropical herbivorous turtles. Biotropica 27:121–127

OMSTARRT (Ontario multi-species turtles at risk recovery team) (2012) Rationale for Ending the Harvest of Snapping Turtles (Chelydra serpentina) in Ontario. 42 pp.



Seburn DC (2007) Recovery Strategy for Species at Risk Turtles in Ontario (draft). Ontario Multi-Species Turtles at Risk Recovery Team.

Spotila J (2004) Sea Turtles: A complete guide to their biology, behavior, and conservation. Baltimore, Maryland: The Johns Hopkins University Press and Oakwood Arts

Turtle Conservation Coalition. (2011) Turtles in Trouble: The World’s 25+ Most Endangered Tortoises and Freshwater Turtles—2011. Rhodin AGJ, Walde AD, Horne BD, van Dijk PP, Blanck T, Hudson R (eds.): IUCN/SSC Tortoise and Freshwater Turtle Specialist Group, Turtle Conservation Fund, Turtle Survival Alliance, Turtle Conservancy, Chelonian Research Foundation, Conservation International, Wildlife Conservation Society, and San Diego Zoo Global, 54 pp. Lunenburg, MA

van Dijk PP (2011) Clemmys guttata. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.2. <www.iucnredlist.org>. Downloaded on 23 October 2012.

van Dijk PP (2012) Chelydra serpentina. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.2. <www.iucnredlist.org>. Downloaded on 23 October 2012.

van Dijk PP, Rhodin AGJ (2011) Emydoidea blandingii. In: IUCN 2012. IUCN Red List of Threatened Species. Version 2012.2. <www.iucnredlist.org>. Downloaded on 23 October 2012.

Vargas-Ramirez M, Stuckas H, Castaňo-Mora OV, Fritz U (2012) Extremely low genetic diversity and weak population differentiation in the endangered Colombian river turtle Podocnemis lewyana (Testudines: Podocnemididae). Conserv Genet 13:65–77

144

Copyright Acknowledgements

Chapters 2 and 3 are published in Conservation Genetics Resources and are reproduced here with

permission of the co-authors.

Documents

Conservation Genetics of Freshwater Turtles · amplification of two loci in the snapping turtle (chelydra serpentina) 12 2 abstract 12 2.1 primer note 13 2.2 acknowledgments 15 2.3