NATURAL AND ANTHROPOGENIC INFLUENCES ON POPULATION DYNAMICS IN BUTTERNUT (JUGLANS CINEREA L.)

A Dissertation

Submitted to the Graduate School of the University of Notre Dame in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

by Sean M Hoban

Jeanne Romero-Severson, Director

Graduate Program in Biological Sciences Notre Dame, Indiana April 2010

© Copyright 2010 Sean Michael Hoban

NATURAL AND ANTHROPOGENIC INFLUENCES ON POPULATION DYNAMICS IN BUTTERNUT (JUGLANS CINEREA L.)

Abstract by Sean M. Hoban

The success of a species and the co-existence of communities over time ultimately depends on the stability of populations. While populations are inherently dynamic, contemporary environmental change may create population instability by altering processes such as migration, recruitment, and population growth and decline. A current goal in conservation biology and population ecology is to quantify the short- and longterm impact of anthropogenic activities, including habitat degradation and global climate change, in order to predict changes in populations, species, and communities. This knowledge is elusive, as these activities take place at multiple and often vast spatial and temporal scales, which to some degree limits the use of manipulative ecological studies to disentangle the web of human and anthropogenic influences. However, changes in population dynamics often leave genetic signatures. Biologists can use molecular ecology tools and carefully chosen study sites to investigate large-scale “experiments in action”- natural populations already subject to specific ecological processes. I use this approach and the forest tree butternut (Juglans cinerea

Sean M. Hoban L.) to investigate the influence of: contemporary population decline on the distribution of genetic diversity, contrasting habitats on patterns of recruitment and colonization, and proximity to anthropogenic landscapes on rates of interspecific hybridization. Juglans cinerea represents a suite of early colonizing tree species, as well as a model conservation dilemma. My work shows the utility of a comparative, observational molecular ecology approach to investigating these processes. First, I show that the major influence on population genetic diversity is not dramatic 20th century decline due to an epidemic disease, but rather ancient range shifts following Holocene-era climatic warming. Second, I show that patterns of disturbance in upland and riparian habitats create markedly different opportunities for recruitment and spatial distribution of diversity. Third, I show that while anthropogenic landscapes facilitate extensive hybridization between J. cinerea and an introduced tree, Japanese walnut, forested landscapes greatly limit hybridization. The results of my first study suggest that the reproductive and life history characteristics of forest trees may allow them to endure extreme and rapid environmental change, even if populations are reduced in size and connectivity. However, the next two studies are a reminder that realized dispersal is quite limited in trees, particularly in the absence of suitable colonization sites. Predictions of forests‟ ability to track future climate change are therefore difficult to make. Together, these results emphasize that an effective conservation approach is the preservation of populations, even if small, in situ.

This dedication is for my parents who, as my first teachers, instilled in me a love of learning, as well as a desire to conserve and restore natural places, to help persons less fortunate than I, and to carefully consider how my actions affect the entire planet.

ii

CONTENTS

Figures................................................................................................................................ vi Tables .............................................................................................................................. viii Acknowledgments............................................................................................................... x Chapter 1: Introduction ....................................................................................................... 1 1.1 Scope of dissertation ......................................................................................... 1 1.2 Study system and approach ............................................................................... 3 1.3 Summary of chapters ........................................................................................ 5 1.4 Overall goals ................................................................................................... 13 Chapter 2: Microsatellite marker development................................................................. 15 2.1 Introduction ..................................................................................................... 15 2.2 Methods........................................................................................................... 17 2.3 Discussion ....................................................................................................... 22 Chapter 3: Geographically extensive hybridization between a native and an introduced forest tree .............................................................................................................. 23 3.1 Introduction ..................................................................................................... 23 3.2 Materials and methods: ................................................................................... 25 3.2.1 Species ................................................................................................... 25 3.2.2 Study populations................................................................................... 25 3.2.3 Genotyping ............................................................................................. 26 3.2.4 Genetic analysis ..................................................................................... 26 3.2.5 Hybrid assignment ................................................................................. 27 3.3 Results ............................................................................................................. 27 3.4 Discussion ....................................................................................................... 28 Chapter 4: Range-wide distribution of genetic diversity in the tree Juglans cinerea: a product of historical range shifts, not ecological marginality or recent population decline ................................................................................................................... 33 4.1 Introduction ..................................................................................................... 33 4.2 Materials/ methods .......................................................................................... 37 4.2.1 Overview of methods ............................................................................. 37 4.2.2 Species ................................................................................................... 38 iii

4.2.3 Collections ............................................................................................. 38 4.2.4 Genotyping ............................................................................................. 38 4.2.5 Population diversity statistics ................................................................ 40 4.2.6 Population differentiation statistics........................................................ 40 4.2.7 Bottlenecks ............................................................................................. 41 4.2.8 Qualitative and quantitative predictive variables ................................... 42 4.2.9 Qualitative and quantitative statistical analyses .................................... 43 4.2.10 Bayesian analyses of population structure ........................................... 45 4.2.11 Isolation-by-distance: Mantel tests ...................................................... 46 4.3 Results ............................................................................................................. 46 4.3.1 Descriptive statistics .............................................................................. 46 4.3.2 Population differentiation ...................................................................... 47 4.3.3 Bottlenecks ............................................................................................. 48 4.3.4 t-tests ...................................................................................................... 51 4.3.5 Linear models......................................................................................... 52 4.3.6 GESTE ..................................................................................................... 54 4.3.7 Bayesian analyses of population structure ............................................. 54 4.3.8 Isolation-by-distance: Mantel tests ........................................................ 54 4.4 Discussion ....................................................................................................... 56 4.4.1 Study summary ...................................................................................... 56 4.4.2 Contemporary population decline .......................................................... 56 4.4.3 Ecological marginality of contemporary range edge ............................. 57 4.4.4 Range shifts ............................................................................................ 59 4.4.5 Isolation-by-distance .............................................................................. 60 4.4.6 Conservation context ............................................................................. 61 4.4.7 Conclusions ............................................................................................ 62 Chapter 5: Contrasting spatial genetic structure in two habitat types for butternut: riparian and upland population dynamics .......................................................................... 64 5.1 Introduction ..................................................................................................... 64 5.2 Materials and methods .................................................................................... 71 5.2.1 Species: .................................................................................................. 71 5.2.2 Collections: ............................................................................................ 73 5.2.3 Genotyping:............................................................................................ 75 5.2.4 Population summary statistics: .............................................................. 75 5.2.5 Spatial genetic structure ......................................................................... 76 5.2.6 Comparison between habitat types ........................................................ 80 5.2.7 Changing the reference population for SGS .......................................... 81 5.3 Results ............................................................................................................. 83 5.3.1 Summary statistics ................................................................................. 83 5.3.2 Relatedness and spatial genetic structure ............................................... 84 5.3.3 Comparison between habitats ................................................................ 86 5.3.4 Use of overall reference population ....................................................... 89 5.4 Discussion ....................................................................................................... 90 5.4.1 Comparison between riparian and upland habitats ................................ 90 iv

5.4.2 Comparison of butternut to other tree species ....................................... 95 5.4.3 Implications for conservation of butternut ............................................. 96 5.4.4 Relevance to experimental design ......................................................... 97 5.4.5 Conclusion ............................................................................................. 99 Chapter 6: Human impacted landscapes facilitate hybridization between a native and an introduced tree species ........................................................................................ 100 6.1 Introduction ................................................................................................... 100 6.2 Materials and methods .................................................................................. 104 6.2.1 Species ................................................................................................. 104 6.2.2 Collections ........................................................................................... 106 6.2.3 Genotyping ........................................................................................... 109 6.2.4 Hybrid analysis .................................................................................... 109 6.2.5 Habitat assignment ............................................................................... 110 6.2.6 Tests between habitat type ................................................................... 111 6.2.7 Maternal ancestry ................................................................................. 111 6.2.8 Spatial analysis..................................................................................... 112 6.3 Results ........................................................................................................... 112 6.3.1 Genetic diversity and number of hybrids ............................................. 112 6.3.2 Comparison between habitats .............................................................. 113 6.4 Discussion ..................................................................................................... 116 6.4.1 Comparison between habitats .............................................................. 117 6.4.2 Comparison to other work ................................................................... 120 6.4.3 Butternut canker and hybridization ...................................................... 121 6.4.4 Predictions and future directions ......................................................... 122 Chapter 7: Conclusion..................................................................................................... 126 Appendix A: Supplemental tables .................................................................................. 144 Bibliography ................................................................................................................... 156

v

FIGURES

Figure 3.1 Posterior probability distribution and the most likely genotypic class assignments of hybrid individuals. „Mixed‟ assignments had < 0.50 probability in any of the given hybrid classes. ............................................................................ 29 Figure 3.2 Native range of butternut, and admixture in each collection site. Histograms represent the number of individuals assigned to each genetic class by NEWHYBRIDS. The four individuals with ≤ 0.50 probability in any class (the „mixed‟ hybrids) are not shown. ........................................................................... 29 Figure 4.1 Three models for prediction of range-wide genetic diversity and differentiation. Filled area: contemporary native range. Red to blue gradient: high diversity and low differentiation to low diversity and high differentiation. Yellow circles: populations. (a) contemporary population decline with a southern epicenter, (b) ecological marginality of the periphery, (c) range shift following the LGM and (d) population labels. ............................................................................ 36 Figure 4.2. Principal component analysis based on FST. The first and second principal components account for 0.28 and 0.27 of the variance, respectively. Population labels as in Figure 1 and Table 1. Black circle: north of ice margin at LGM; open circle: south of the ice margin at LGM. ................................................................ 49 Figure 4.3. Bayesian cluster analyses. (a) STRUCTURE (K=11), colors represent the cluster having the highest Q value for that population (one cluster did not have the highest Q in any population, so only ten colors are displayed); (b) BAPS (K=7), colors represent distinct BAPS clusters; (c) Actual Q values per population, showing admixture of 11 clusters including gray cluster not appearing in (a); populations to the left of the arrow are south and populations to the right are north of the ice margin at LGM (double dashed line). ......................................... 55 Figure 5.1: Sampling sites for this study. Blue circles are „riparian,‟ brown circles are „upland,‟ and yellow circles are „other‟. ............................................................... 71 Figure 5.2: Contour maps and photographs from typical riparian (a and b), and upland (c and d) sites. Note the riparian site is clearly within the floodplain (a) and the upland site is far from waterways (b), and the dense forest cover in (d). ............. 74 vi

Figure 5.3. Spatial autocorrelograms, where the black line shows Fij,WITHIN and the green line shows Fij,OVERALL. The correlogram show (a) signal of positive SGS and (b) lack of signal of positive SGS, but a signal of colonization ................................. 78 Figure 5.4. Correlograms for a subset of populations in the study: five riparian (a) and five upland (b) sites. Black line shows Fij using the within population reference. Green line shows Fij using the overall reference. No SGS means that the black line never crosses the 95% confidence interval (grey dashed line). Colonizer means that the green line is always elevated above the 95% confidence interval (high relatedness at all distances). ........................................................................ 85 Figure 5.5. Cartoon of our model of population dynamics in a riparian butternut population ............................................................................................................. 91 Figure 5.6. Cartoon of our model of population dynamics in an upland butternut population ............................................................................................................. 92 Figure 6.1. Sampling locations of trees used in this chapter. Each dot represents at least one tree. ............................................................................................................... 107

vii

TABLES

Table 2.1: Descriptive statistics for microsatellites ......................................................... 20 Table 2.2: Results of experimental verification of null alleles ........................................ 21 Table 4.1 Populations used in the study............................................................................ 39 Table 4.2 Genetic diversity statistics for all populations .................................................. 50 Table 4.3 Results from qualitative tests ............................................................................ 51 Table 4.4 Results from quantitative tests .......................................................................... 53 Table 5.1 Some recent studies of habitat influence on SGS ............................................. 67 Table 5.2. Populations used in this study......................................................................... 70 Table 5.3 Comparison of Sp and F1 statistics observed in butternut to other primarily outcrossing, wind pollinated trees......................................................................... 82 Table 5.4 SGS statistics and population age structure observed ...................................... 87 Table 5.5 Comparison between riparian and upland habitats for SGS and population age structure................................................................................................................. 88 Table 5.6 Results of regression with Sp and Prop(young)................................................ 89 Table 6.1 Populations used in this study......................................................................... 108 Table 6.2 Counts of J. cinerea and non-J. cinerea in each habitat type ......................... 114 Table 6.3 Count in each hybrid class, in each habitat type ............................................. 114

viii

Table 6.4 Occurrences of J. cinerea and J. ailantifolia chloroplast in each hybrid class115 Table 6.5 Occurrences of J. cinerea and J. ailantifolia chloroplast in hybrids in each habitat .................................................................................................................. 116 Table 6.6 Mean pair-wise distance between J. cinerea individuals and between non J. cinerea individuals in six populations ................................................................ 116

ix

ACKNOWLEDGMENTS

First, I would like to acknowledge my advisor, Dr. Jeanne Romero-Severson. For five years, Jeanne has provided immeasurable guidance and knowledge, as well as life lessons. Just as importantly, Jeanne has always expressed confidence in my abilities, and has given me the freedom and encouragement to explore new interests and challenges. I could not have asked for a better mentor. I am also grateful for the unique role each of my other committee members has provided. Nora, Jason, and Scott have consistently reminded me of the value of a carefully constructed, falsifiable hypothesis. Each has also provided invaluable advice and criticism, and their frankness has kept my ideas grounded, taught me to appreciate natural history, and challenged me to be a more precise in my thinking and speaking. Studies of this kind cannot be done within the short period of one graduate career without the assistance of many people. In particular, Tim McCleary, the RomeroSeverson lab manager, has made a huge contribution to my work. Tim has assisted with every aspect of my lab work, shouldered most of the work of planning and executing our collection trips, provided much feedback on writing, performed genotyping, grinding, and other lab work, and has been a good friend. For the past two years, Dan Borkowski has been similarly helpful, and has always been cheerful and supportive. Both Dan and Tim have taught me a lot about forest ecology, and have always been ready to debate my latest research idea. x

Without the assistance of many undergraduate assistants, I would still be grinding twigs, entering spreadsheet data, and labeling sample bags (not to mention racking tips), so I thank: Britney Dennis, Bridgette Sanchez, Betsy Madison, Ellen Luecke, Lindsay Huebner, Paul Goedde, Casey Bouskill, and Marie Pereira. Although our lab went on many collection trips (which have left me with many great memories), many of the butternut collections were made by others. The following people directly collected tissue samples for my project and/ or helped locate trees: Robert White, Jack Frank, Brice Leech, Ashton Berdine, Eagle Brosi, Jeanine Riggleman, Jennifer Murrow, Sunshine Brosi, Ami Sharp, Dale Bergdahl, Diane Burbanke, Barbara Boysen, Judy Loo, Donnie McFee, Robert Anderson, David Gallagher, Mary Ann Buenzow, Paul Berrang, Chris Casey, Henrietta Bennett, Carrie Sweeney, Dwight Slocum, Mark Coggeshall, Pamela Slatten, Sandra Anagnostokis, Rebecca Wright, Dena Garvue, Bryan Connolly, Wayne Kruger, as well as other professionals in the US Forest Service, US National Parks Service, several state Natural Resource/ Fish and Wildlife Departments, and Canadian Ministry of Natural Resources. I would also like to acknowledge assistance from graduate students and staff in the Schlarbaum, McLachlan, and Hellman programs, and staff members from the Connecticut Agricultural Station, Chicago Botanic Garden, Cornell Plantations, Minnesota Arboretum, Morton Arboretum, Rhora‟s Nut Nursery, USDA-GRIN, and the Center for Agroforesty at the University of Missouri at Columbia. The Northern Nut Growers Association has also been an important part of my research. I thank them for providing me with invaluable firsthand knowledge of butternuts and hybrids, as well as a balanced perspective on science and nature. I xi

particularly thank a teacher whom I will always remember, Tucker Hill; a knowledgeable and engaging guide to Indiana‟s hybrid butternuts, Bill Dieter; a patient and thoughtful nurseryman, Ernie Grimo; and an always friendly and helpful NNGA librarian, Jerry Henkin. I also thank many other NNGA members, including Ken Hunt, Ken Bauman, Bud Luers, Clifford England, Malcolm Olsen, Maurice Matthews, Parker Coble, Ralph Johnson, Michael Dolan, Al Beck, and John Brittain. I also owe thanks to my colleagues in the department, especially the members of the EEE cluster who have provided friendship, criticism, and encouragement. In particular I will long remember tea and lunch time conversations with Nathalie Griffiths, Sara Epstein, and Ashlee Baldridge. The following have been great role models as well as critics: Travis Marsico, Kelly Lane, Peter Levi, Chris Patrick, Mia Stephen, Candace Lumbao, Kelly Lane, Peter Levi, Shannon Pelini, Tom Powell, and Andy Forbes. It is doubtful that I would be where I am today, in many ways, without one of my best friends and most interesting scientific colleagues, Rory Carmichael. To Rory I owe much, but in particular I thank him for many late hours discussing genetics, bioinformatics, literature, and philosophy. I also thank Rory for some of the best food I‟ve ever had, the best writing I‟ve produced, and of course the blue label. I would like to thank many other members of the Notre Dame community: Lisa Anderson in the Career Center, who taught me the value of networking, the Interlibrary Loan office, for finding rare books and manuscripts that would otherwise have been inaccessible, Brent Harker, for running my genotyping plates, Loretta, Kathy, and especially Faith for always answering my questions about the department, and Jane Gayou for editing many of the things I‟ve written. xii

The following organizations have funded my research and my travels: the University of Notre Dame Biology Department and Graduate School, the Arthur J Schmitt Foundation (who sponsored my graduate work), the University of Washington, the American Society of Plant Biologists, the University of Tennessee, the Margaret Finley Shackelford Trust, the Healing Stones Foundation, and the Northern Nut Growers Association Last, I thank those persons close to me who have tolerated my long hours of work and my odd passion for butternuts, including my family, friends, pen pals, and of course SCN.

xiii

CHAPTER 1: INTRODUCTION

1.1 Scope of dissertation Ecology is often described as the study of the distribution and abundance of individuals, populations, species, and communities, and how this distribution changes over time. Therefore, ecology is inherently a study of change, at local (e.g., community structure, competition, succession) and global scales (e.g., range shifts, resource cycling, species‟ extinctions). Humans play an increasing role in ecological change, through habitat degradation, overharvesting, introduction of pathogens, and release of greenhouse gases. A challenge in ecology is to describe the integrated spatial and temporal dynamics of both natural and anthropogenic influences, at a range of scales. The long-term goals of this work are to better predict population response to future changes and to facilitate informed conservation policy. To achieve these goals, it will be particularly important to identify the natural and anthropogenic influences that leave the greatest impact on biodiversity, from populations to ecosystems. My dissertation research addresses natural and anthropogenic influences on three aspects of population dynamics: genetic diversity and differentiation among populations, spatial genetic structure within populations, and interspecific gene flow, at spatial scales that range from hundreds of square meters to tens of thousands of square kilometers. By examining these three somewhat disparate topics, I approach some of the most pressing 1

challenges in conservation biology: global climate change, habitat degradation, introduction of non-native species, population collapse, and local extinction. My results provide fresh perspective on the short and long-term response of populations to these influences. I am interested in these challenges in forest trees, for some of the same reasons that make them difficult to work with: long life span, large size, vast native ranges, massive potential reproductive output, and overlapping generations. Forest ecosystems are also fascinating for their remarkable resilience and ability to regenerate. However, a current and crucial knowledge gap in forest ecosystems is a better understanding of the conditions that ensure such resilience. While these attributes make long-term manipulative experiments of forests difficult (although such experiments are necessary), they also make some areas of investigation especially tractable, including those that I am interested in: seed dispersal, regeneration, changes in population size, migration, and interspecific interactions. Carefully selected observational studies of long-term „experiments in action‟ (forests already subject to various natural and anthropogenic influences) can provide a picture of a population‟s response to different degrees of perturbation, enabling comparative studies. Forest trees are particularly important conservation foci, as they provide mast and habitat for wildlife, erosion control, building materials, fuel, and carbon storage. To ensure the stability of these services, it is likely that forest ecosystems will be increasingly managed, a task complicated by multiple factors, including climate change, loss of biodiversity, and habitat fragmentation.

2

1.2 Study system and approach I approach these research directions using butternut (Juglans cinerea L.), a windpollinated, outcrossing North American forest tree that has experienced severe demographic decline in the 20th century. Butternut (also known as white walnut, oilnut, and lemonnut) is native to eastern North American deciduous forests, primarily in riparian areas. Butternut is a useful system for studying the interaction between ecological processes and population dynamics for several reasons. First, a principal requirement of ecological studies is a comparative framework, e.g., evaluating populations subject to contrasting ecological processes. Throughout its range (> 1 million km2), butternut is subject to different ecological processes in different geographical regions, including the effects of a contemporary disease epidemic, butternut canker, which has been most severe in southern states (Chapter 4). Other processes, including habitat type (Chapter 5) and anthropogenic influence (Chapters 3, 6) also have a spatial component. Second, butternut serves as a good model for many fast growing, early colonizing tree species (e.g., Salix, Platanus, Populus, Liriodendron, Betula), which are underrepresented in the literature on population dynamics, but which may be among the first trees to respond to local and global environmental change (Halpern, 1989; Rossetto et al., 2004; Twedt, Wilson, 2002). Butternut also serves as a good model for species that tend to occur at low density. As Charles Darwin wrote, "rarity is the attribute of vast numbers of species in all classes (Darwin, 1859)." While this is now a basic tenet of ecology (Preston, 1962), uncommon trees (with the exception of critically endangered species) are underrepresented in the population dynamic and population genetic literature. Third, while butternut is a rare species, it has ecological value as winter mast

3

for wildlife (the nut is nutritious and rot resistant) as well as economic value for humans, for fine hardwood and a potential source of fungicides and medical uses (Omar et al., 2000; Ostry, Pijut, 2000). Additionally, butternut is currently protected on national forest and crown land, is a species of concern on many state lists, and has recently been named an endangered species in Ontario (Nielsen et al., 2003; Schultz, 2003). Forest biologists have located and surveyed many suitable study sites, and have begun to implement conservation plans (Boysen, 2009; Schlarbaum et al., 1997a; Schultz, 2003). Butternut is also of interest to many private citizens who have a cultural attachment to the tree, and who are eager to assist with these research efforts. Butternut decline is predicted to become more severe under future climate change, increasing loss to disease, and habitat degradation, including loss of riparian systems due to irrigation for agriculture, dams, and changing weather patterns, as well as other changing land-use patterns. My research, by quantifying the effects of past events, will help predict the outcome of current and future ecological change, and prioritize conservation concerns. Specifically, this dissertation work contributes to current and near future conservation management for butternut, including seed collection, habitat improvement and restoration, relocation, and disease studies, as well as contribute to basic knowledge of natural history and population biology. I am primarily interested in changes to population dynamics such as recruitment, migration, and population size. I use genetic tools to detect the signature of these processes. Long-term ecological studies of these processes will complement the genetic studies, but for rare and threatened species, observation of population decline over time is not an option (nor is a time machine). Examining patterns in the distribution and

4

abundance of genetic diversity may reveal the impact of ongoing processes, as well as those that occurred dozens or thousands of years ago (Allendorf, Ryman, 2002; Frankham, 1995; Lowe et al., 2004). For example, reductions in population size (bottlenecks) leave distinct patterns of genetic diversity, and the shape of the genetic signature may reveal information about the length or severity of the bottleneck (Chapter 4). At smaller spatial scales, the spatial distribution of relatives within populations can reveal information about dispersal and recruitment, two processes essential for successful restoration and community stability (Chapter 5). These two processes are linked. Seed and pollen dispersal occur over small scales but have a cumulative large scale-effect, influencing a species‟ ability to, for example, track a changing climate with rapid range shifts while also retaining enough genetic diversity to successfully persist in newly colonized areas (Austerlitz, Garnier-Gere, 2003; Clark, 1998; Hardy, 2009; Petit et al., 2005). Quantifying dispersal, migration, and population size fluctuations using genetic tools will therefore contribute to an understanding of population ecology at multiple scales.

1.3 Summary of chapters To examine these research questions, I first developed genetic tools specific to Juglans cinerea (Chapter 2). Microsatellites, which are highly variable genetic markers, can be used in a wide range of studies, from parentage analysis to biogeography. While microsatellite markers have been developed for a related tree species, J. nigra (Robichaud et al., 2006a; Victory et al., 2006), I developed the first microsatellite markers for J. cinerea. Due to phylogenetic distance between these two species 5

(Manning, 1978), primers for J. nigra may amplify some alleles in J. cinerea, but not others, because of mutations in the primer-binding regions. These unobserved or null alleles can result in misleading departures from expected genetic distributions (Lowe et al., 2004; Van Oosterhout et al., 2004), and would be particularly problematic in studies of local gene flow (Chapter 5) and hybridization (Chapters 3, 6). In Chapter 2, I describe the protocol for developing these markers from a butternut sequence library enriched for microsatellite repeats, and I demonstrate their utility in generating basic population statistics. This work not only enabled the rest of my projects but also provides a permanent genetic resource for future investigations. The next chapter combines a further tool development and optimization project with an exploratory study to validate the practicality of my main questions. This study was sparked by concerns over hybridization between butternut and Japanese walnut (Juglans ailantifolia Carrière). In orchard settings, the two species can cross-pollinate, and the offspring of this cross is such a vigorous, fruitful tree that forest tree biologists have expressed concern over the possibility of an undetected, range-wide genetic invasion (Ostry, Woeste, 2004; Reed, Davidson, 1958). In addition to contributing towards the decline of the native (Hails, Morley, 2005; Wolf et al., 2001), hybridization can also alter or mask the genetic patterns produced by the ecological processes I am interested in (Valbuena-Carabaña et al., 2005), if hybrids are not detected and removed before analysis. Therefore, in order to examine processes such as the impact of habitat on genetic structure, I needed to separate the butternuts from the “non-butternuts,” the J. ailantifiolia and various types of hybrid individuals that have naturalized in butternut‟s native range.

6

Therefore in Chapter 3, I used the markers developed in Chapter 2, along with a set of chloroplast markers (McCleary et al., 2009), to reliably identify the hybrid status of the individuals collected during my first several years of work. As no definitive silvic characters distinguish hybrids from J. cinerea, identification using DNA markers is essential. In this investigation, I quantified rates of hybridization at seven locations. In doing so, I demonstrated that many populations have very low or zero levels of admixture (the amount of genetic material from the introduced species), and in these populations I could proceed to answer my questions focused on the response of butternut populations to range-wide (Chapter 4) and local (Chapter 5) ecological processes. However, the presence of J. ailantifolia chloroplast and many highly probable hybrids in two locations in this preliminary study suggested that hybrids persist and interbreed with the native species in some natural settings (in contrast to expectations for disjunct species), which enabled my questions in Chapter 6. These results are among the first reports of naturally occurring hybridization between native and introduced forest trees, and add to a growing body of evidence that many congeneric but disjunct species retain the ability to hybridize if brought back together in nature. In this (Chapter 3) I describe initial observations, and I return to approach the dynamics of this hybrid invasion, and conservation consequences, in Chapter 6. My next study (Chapter 4) concerned the genetic and demographic impact of three range-wide ecological processes, each occurring at different temporal and spatial scales. I was primarily interested in contrasting the effects of contemporary ecological conditions and both ancient and contemporary population bottlenecks, which work together to shape the distribution and abundance of genetic diversity within a species

7

(Foll & Gaggiotti, 2006, Loveless & Hamrick, 1984). This study is especially relevant in a time of rapid environmental change, shifts in species‟ ranges, and changes in community structure. These questions are also relevant to management activities that involve relocation and ex- situ conservation of subsets of a species‟ total diversity. Identifying those processes that have the greatest impact is a logical first step in developing predictive models for population responses to future environmental change, identifying important thresholds, and guiding management strategies (Hamrick, 2004, Jaramillo-Correa et al., 2009, Petit et al., 2008). My goal in this investigation was to determine whether contemporary population decline, contemporary ecological marginality, or historical range shifts has had the most influence on current patterns of range wide diversity in butternut. The contemporary population decline investigated was that due to butternut canker, but the approach used and the conclusions in this study should apply to other species that have experienced recent declines due to exotic pests and pathogens (Brockerhoff et al., 2006; Hain, 2006; Tkacz et al., 2007), or to abiotic stress (Van Mantgem et al., 2009). The current spatial distribution and abundance of genetic diversity should offer some insight into which processes have resulted in the greatest genetic impact, but previous investigations have had difficulty in distinguishing the effects of these processes (Gaggiotti et al., 2009; Knowles et al., 2007). To overcome this difficulty, I first built a hypothesis-testing framework by constructing explicit predictions under each process. I then implemented a sampling scheme designed to test the range-wide predictions of each model, incorporating 27 populations and nearly 1000 individual trees, in the first rangewide study of genetic diversity in butternut. Using both classical and Bayesian model

8

comparison, as well as other lines of evidence, I established statistical support of range shifts as the clear determinant of range-wide structure and diversity, adding to the already strong body of evidence in support of the range shift hypothesis in both plant and animal taxa (Hewitt, 2000, Jaramillo-Correa et al., 2009). In spite of dramatic (70- 90%) 20th century decline due to the disease butternut canker, contemporary populations show little evidence of genetic bottlenecks. Location within the range edge vs. the range core (a proxy of ecologically marginal conditions) also failed to account for the observed patterns of diversity and differentiation. My results are consistent with those of other investigators, who suggest that brief and even extreme population declines may not have lasting genetic signatures (Brown et al., 2007, Okello et al., 2008, Yao et al., 2007). This should hold true for forest trees as well as other species with long distance dispersal potential and overlapping generations. I anticipate that this work, paired with that of others (Eckert et al., 2008, Lawton, 1993), will strengthen the experimental design and analysis of future investigations of these and other range-wide processes. More specifically, carefully designed sampling schemes and comparative statistical analysis, including recently developed Bayesian methods (Foll, Gaggiotti, 2006), will contribute to this research direction. Moving my research from the largest to the smallest scale of population genetics, I next studied the influence of habitat on seed dispersal, population recruitment, age structure, and metapopulation dynamics (Chapter 5). Ecological conditions, including resource availability and competition, are important determinants of these population processes (Dick et al., 2008; Loveless, Hamrick, 1984; Petit, Hampe, 2006; Vekemans, Hardy, 2004), and quantifying their contribution is an emerging area of population and

9

ecological genetics (Guillot et al., 2009; Hu et al., 2010). These processes can be studied by describing and analyzing patterns of spatial location of related individuals, particularly the correlation between spatial and genetic distance (also called spatial genetic structure, SGS) (Legendre, Fortin, 1989; Levin, 1981; Manel et al., 2003; Smouse, Peakall, 1999). Such knowledge can guide reintroduction, relocation, and habitat restoration efforts , as well as improve the predictive value of models of population response to future local perturbations (e.g., selective harvest, habitat degradation) at a range of spatial and temporal scales (Marquardt, Epperson, 2004; Oddou-Muratorio et al., 2004; Rajora, Pluhar, 2003). However, identification and quantification of the contribution of local conditions to SGS is a relatively recent trend in spatial genetics (Born et al., 2008a), and has been somewhat inconclusive. Previous studies have typically been exploratory or descriptive, and have therefore often led to multiple and largely post hoc explanations for observed SGS (Guillot et al., 2009; Hampe et al., 2010; Pardini, Hamrick, 2008). A hypothesis-testing framework, as well as a comparative statistical approach, could resolve much of this conflict (Hampe et al., 2010; Vekemans, Hardy, 2004; Born et al. 2008b). To explore this emerging question, I contrasted the SGS observed in eight riparian populations with that observed in ten upland populations, to test the hypothesis that riparian sites will exhibit stronger SGS than upland sites. While the transition between these habitats is complex and dynamic (Naiman and, Decamps, 1997; Nilsson, Svedmark, 2002; Tabacchi et al., 1998), I propose a simple but biologically grounded model of population dynamics based on patterns of disturbance and recruitment, which are central forest processes that determine long-term community structure and stability (Franklin et al., 2002; Sousa, 1984). By sampling from a moderate number of populations in each

10

category, I established statistical support of the role of habitat in determining SGS, supporting the model of population dynamics that I proposed. My results are concordant with an emerging consensus that frequency and location of seedling establishment, as well as density of surrounding trees, are influential determinants of SGS (Born et al., 2008b; Pardini, Hamrick, 2008; Sezen et al., 2009). A deeper investigation of the interaction between SGS and colonization dynamics is a promising new direction integrating ecology, population dynamics, and genetics (Born et al., 2008b), that will generate far-reaching implications for habitat management, as well as a greater understanding of the mechanisms by which small populations persist in dynamic landscapes (Oddou-Muratorio et al., 2004; Slavov et al., 2009). My final research direction (Chapter 6) was to revisit the hybrid dynamics from Chapter 3. As noted above, forest biologists have long suspected that many naturally occurring, vigorous, and fertile trees that appear to be J. cinerea are actually hybrid offspring of J. cinerea and Japanese walnut (J ailantifolia), which was planted widely in eastern North America in the late 1800s (Anagnostakis, 2009; Fleguel, 1996b; Hoover, 1919; Reed, Davidson, 1958). However, little is known about where, to what extent, and by what mechanisms hybrid offspring may persist and interbreed (Ostry, Woeste, 2004). In Chapter 3, I demonstrated that hybridization does occur, that large proportions of hybrids (> 0.90) occur in some populations while low or zero hybridization occurs in others. I also suggested that extensive hybridization in two populations is associated with agriculture and other human impacted landscapes, which may have facilitated colonization by both parental species or may have provided a new habitat or resource that hybrids can exploit. My goal in Chapter 6 was to examine a large number of locations

11

that could be classified as continuous forest, fragmented forest, or anthropogenic landscapes. I was interested in the influence of habitat on the extent of hybridization (Anderson, 1948; Petit et al., 2004b; Rieseberg et al., 2003; Wiegand, 1935), but also in its influence on the direction of gene flow and the spatial dynamics of hybrid populations (Brede et al., 2009; Mullen et al., 2008; Petit et al., 2004b; Ruegg, 2008). A greater understanding of how habitat contributes to the spatial and temporal dynamics of hybridization is needed to better evaluate this complex process, enable predictions, and facilitate management decisions (Currat et al., 2009; Gunnell et al., 2008; Wolf et al., 2001). Based on suspected introduction points and dispersal limitations, I propose three specific hypotheses in this chapter: (1) extensive hybridization is limited to sites with high and medium anthropogenic impact, (2) hybrids primarily have ancestry from a J. ailantifolia mother tree, and (3) hybrids exhibit spatial aggregation rather than random distribution in the populations where they occur. Previous studies in salamanders, irises, and sunflowers suggest that habitat can influence the fitness of hybrids (Fitzpatrick, Shaffer, 2004; Fitzpatrick, Shaffer, 2007a; Fitzpatrick, Shaffer, 2007b; Martin et al., 2006; Rieseberg, 1997), but few studies have investigated the influence of habitat on dispersal of introduced genes. This is one of the first investigations of hybrid dynamics between a native and introduced forest tree in North America (for another, see Burgess et al, 2005, 2008), and to my knowledge the largest in spatial scale. By investigating across the entire range, and by specifically focusing on habitat types, I show that hybridization is not advancing in a regional hybrid „front,‟ but rather as „bubbles‟ within suitable habitat types, and that

12

habitat significantly influences both extent and direction of hybridization. Hybridization may exacerbate other conservation concerns (such as those in Chapters 3, 4) by introducing new genetic, demographic, and ecological changes that accelerate the loss of native biodiversity (Levin, 2002; Wolf et al., 2001). On the other hand, hybridization may introduce useful adaptive traits (Schweitzer et al., 2002), such as pest and disease resistance (Adams et al., 2002; Fritz, 1999). Whether hybrids are detrimental or beneficial, an understanding of where introduced genes will move over space and time is currently desirable and timely, and my results present a basis for prediction in butternut as well as other forest trees.

1.4 Overall goals Pattern description for individual species has been a fundamental aspect of conservation genetics, but alone it is insufficient (Fitzpatrick, Shaffer, 2004; Hampe et al., 2010). By structuring my questions around universal processes, I have been able to identify and quantify the contribution of specific ecological mechanisms to population dynamics and population genetic response for butternut, but my conclusions should be applicable to other systems. Additionally, throughout this dissertation, I try to not only answer my research questions but also provide a demonstration of the utility and also the limitations of some of the genetic and statistical tools I use. At the same time, by grounding my work in the biological reality and processes of a species under conservation concern, I have also described useful facets of butternut‟s natural history, and connected my work to conservation management work. I hope to integrate my results with state and federal conservation plans for J. cinerea (including relocation, seed 13

collection, and habitat improvement) to promote maintenance of population level-features like spatial structure and diversity. My research is an engagement with multiple facets of population, ecological, and conservation genetics, with far-reaching implications for how forests response to ecological change.

14

CHAPTER 2: MICROSATELLITE MARKER DEVELOPMENT

2.1 Introduction Butternut (Juglans cinerea L.) is a medium-size hardwood tree native to eastern North American forest ecosystems, most often occurring in riparian zones (Fleguel, 1996c). A lethal and aggressive fungus, Sirococcus clavigignenti-juglandacearum V.M.G. Nair, Kostichka, & Kuntz, which causes butternut canker disease (Nair et al., 1979), has eliminated native butternut populations in several southern states and greatly reduced numbers throughout the natural range (Anderson, LeMadeleine, 1978; Ostry, Woeste, 2004; Schultz, 2003). The pathogen infects the cambium and inner bark of the tree, causing black lesions that disrupt the flow of nutrients through the vasculature. Over several to many years, these cankers coalesce and ultimately kill the tree. The fungus also infects other Juglans species, which show varying levels of susceptibility. Persian walnut (Juglans regia L.), a domesticated species highly valued for nut production, is easily infected (Fleguel, 1996a). Japanese walnut (Juglans ailantifolia Carrière), an Asian species that easily hybridizes with butternut, shows high tolerance. Also, there is evidence of varying tolerance to the disease in wild butternut populations (Schlarbaum et al., 2004), but the genetic basis of tolerance remains unknown. Although previous reports suggest that butternut may have low genetic diversity, thus limiting the species‟ ability to resist exotic pests and pathogens (Fjellstrom, Parfitt, 15

1994; Morin et al., 2000), range-wide investigation of the genetic diversity and population structure of butternut is lacking (Van Sambeek et al., 2002). Additionally, the range of tolerance to the disease that has been observed in natural butternut populations (Schlarbaum et al., 2004) may be due to hybridization with an introduced species, Japanese walnut (Juglans ailantifolia), or native resistance, but the rates of natural hybridization are currently unknown (Ostry, Woeste, 2004; Schlarbaum et al., 2004; Schultz, 2003). Resistance to butternut canker disease, whether in J. cinerea or in related species, could form the basis of a breeding program to reintroduce disease-resistant butternut to eastern forests (Schlarbaum et al., 1997b). Genetic markers designed for genetic diversity investigations, as well as determination of hybrid ancestry, will enable investigations to approach these knowledge gaps in range wide diversity and hybridization, as well as facilitate conservation management of butternut. We present here the first group of microsatellites tailored to J. cinerea investigations. Previous genetic investigations of the Juglans genus have used several DNA marker systems. These include: nuclear genome RFLPs for genetic diversity and phylogeny of the Juglans genus (Fjellstrom, Parfitt, 1995), chloroplast restriction sites for phylogeny of the Juglandaceae family (Smith, Doyle, 1995), RAPD markers to identify Persian walnut cultivars (Nicese et al., 1998), cpDNA and nrDNA sequences for biogeographic history of Juglans (Stanford et al., 2000), isozyme loci for butternut genetic diversity (Morin et al., 2000), chloroplast IGS sequence data for phylogeny and biogeographic history (Aradhya et al., 2004), and chloroplast RFLP data for interspecies identification and taxonomy (Abuín et al., 2004). However, these markers do not provide the versatility and reliability needed in conservation genetics studies of butternut.

16

Additionally, several investigators have developed suites of microsatellite markers for black walnut (Juglans nigra), a valuable hardwood species whose native range overlaps with the central range of butternut. Investigations based on these published primers include parentage analysis and clonal identification (Robichaud et al., 2006b) and interspecies discrimination (Pollegioni et al., 2004). A recent study of black walnut used 12 microsatellites to reveal a remarkable lack of population substructure in the central United States (Victory et al., 2006). However, J. nigra and J. cinerea are not within the same section of Juglans (Manning, 1978) and cannot hybridize, suggesting substantial significant genetic divergence between the two species. Therefore, primers based on the J. nigra clones may amplify some alleles in J. cinerea, but some alleles may not amplify because of mutations in the primer binding regions. These unobserved or null alleles can result in misleading departures from expected allele frequency distributions (Jones et al., 1998). We reduced this risk by using a library of cloned J. cinerea microsatellite sequences for marker development.

2.2 Methods We isolated DNA from freshly cut twigs of a single wild J. cinerea individual from Shannon County, Missouri, using the DNeasy Plant Minikit. Genetic Identification Systems used this DNA for construction of genomic libraries enriched for CA, GA, and TAG repeats as described previously (Jones et al., 2002). They identified 146 microsatellite-containing sequences and designed primer pairs for 95 of them using DesignerPCR version 1.03 (Research Genetics, Inc., Chatsworth, CA). We designed

17

additional primer pairs using Primer3 (Rozen, Skaletsky, 2000). We chose 34 sequences for amplification tests to determine annealing temperatures and quality of amplification. We tested these primers on four J. cinerea individuals. The PCR reactions (10μL) included the following reagents: 1.5 mM MgCl2, 1X PCR buffer [50 mM KCl, 10 mM Tris-HCl ( pH 9.0), 0.1% Triton-X-100], 0.2 mM dNTPs, 4 pM each forward and reverse primer, 4% Bovine Serum Albumin, 0.25 U Takara Ex Taq Polymerase, and 20 ng DNA template. The PCR temperature profile was as follows: 2 minutes at 94 °C; 30 cycles of 94 °C for 30 s, Ta for 30 s, and 72 °C for 30 s; 45 minutes at 60 °C; and 10 minutes at 72 °C on a PTC-225 Peltier Thermal Cycler. Each primer was tested at an annealing temperature (Ta ) of 50 °C, 55 °C, and 60 °C. Amplicons were visualized on 1.5% agarose gels stained with ethidium bromide. Of the 34 primers tested, 13 produced clear, consistently amplifiable, polymorphic bands (Table 1). WellRED fluorescent dye-labeled forward primers and unlabelled reverse primers for the 13 polymorphic microsatellites were used to genotype 63 individuals from a population in central Kentucky. We included three replicates of a positive control (the individual used for creation of the microsatellite library) and one negative control (water) on all plates. Amplicon sizes were detected with an 8-capillary CEQ 8000 Genetic Analysis System, which has a ±1 base pair accuracy using GenomeLab DNA Size Standard 400. After initial genotyping, we used Micro-Checker (Van Oosterhout et al., 2004) to check for the presence of null alleles. Possible null alleles were reported for three loci: jcin_B112, jcin_B212, and jcin_B264. To experimentally verify null alleles, we designed new primers for these loci (Table 2), with the new forward primer offset by at least 30 base pairs from the original forward primer. We used a heterogeneity G-test

18

(Sokal, Rohlf, 1969) to compare observed heterozygosity with the original and redesigned primers. Amplification using the new primers resulted in significantly increased observed heterozygosity for the jcin_B212 locus and non- significant deviations in observed heterozygosity for the other two loci (Table 2). We retained the new primer pair for locus jcin_B212. A significant deficit in observed heterozygosity with two different primer sets at loci jcin_B112 and jcin_B264 may indicate a true deficiency in heterozygotes at these loci. Of the 819 possible genotypes (13 markers on 63 individuals), 797 were recovered (97.3% success). CERVUS v 3.0.3 (Kalinowski et al., 2007) was used to calculate descriptive statistics (Table 1). Amplicon sizes ranged from 103 to 358 base pairs, allowing multiplexing of some loci. The number of alleles observed at a locus ranged from 8 to 23 (mean=13.7), suggesting a high information content potential for these loci in population studies. CERVUS was also used to calculate observed and expected heterozygosities for this population. Four loci (Table 1) showed significant deviation from zygotic equilibrium values (Hardy-Weinberg expected heterozygosity), suggesting a moderate deficiency of heterozygotes. Genepop on the Web (Raymond, Rousset, 1995b) was used to test for gametic disequilibrium (LD). After correction for multiple comparisons, only one of the 78 pair-wise comparisons (jcin_B12 with jcin_B264) showed significant (p < 0.05) LD. This low signature of LD suggests that the markers are not physically linked.

19

TABLE 2.1: DESCRIPTIVE STATISTICS FOR MICROSATELLITES

Repeat

Ta (C)

Na

jcin_A5

AC17

60

13

jcin_B110

CT18

60

11

jcin_B112

CT11

55

8

jcin_B114

CT23

55

19

jcin_B12

CT13

55

9

jcin_B121

CT19

50

14

jcin_B147

CT18

50

10

jcin_B157

CT16

60

9

jcin_B159

CT19

55

16

jcin_B212

CT16

60

23

jcin_B249

CT15

50

13

jcin_B262

CT15

60

20

jcin_B264

AG17

50

13

Locus

Size Range, bp -Exp size

20

195 - 221 -210 212 - 238 -232 271 - 291 -277 249 - 289 -267 156 - 178 -162 171 - 201 -183 327 - 345 -343 158 - 172 -165 103 - 162 -125 196 - 262 -215 170 - 198 -181 316 - 358 -318 222 - 246

-235

HObs

HExp

K

0.806

0.852

62

0.500*

0.644*

62

0.333

0.388

63

0.841

0.895

63

0.762

0.819

63

0.81

0.871

63

0.729

0.786

59

0.794

0.784

63

0.905

0.886

63

0.833*

0.918*

60

0.778*

0.84*

63

0.903

0.887

62

0.581*

0.785*

62

Primer Sequences (5‟ to 3‟) F: AGGATGAAAACTTGGATGTAGA R: TTACGGATCAGGTATGATGAC F: CCTTGTCAGATGCTTATGAAGA R: GAGCAATAACATTTGAAGGTTG F: CCAAGCGAAAGCCAAGTC R: CCACCACCCATTCACTCC F: CTGCCCCTTTCACGTACATA R: CCTTCTCCCTGATGATGACA F: CGTGCCTATGTTCTATCCAC R: CAACGATTTTGTAAGCACAAG F: CAAGCACCAACAATCTAGTAGG R: CACTTATCAAAATGGGGTATCA F: CTAAAGACTTGGCCCGGTTC R: TGGCAAGATGTCATGCAAAG F: CAGCGATGGATGTTTGAGG R: ACTCCGCATAGGTTGTCAGG F: ACTCCAAGTGATGGTGTGG R: ACGGTTTAGCAAGCAGTAATG F: ATTGATGGAGACTCGGTTGC R: CTTAGCCTTGACCCATTAGCA F: TCCGCCTTCAAGTTCTTATGTA R: AATCATCCGCAGAAGTATCATG F: ACCACCATCCTTGTCTCACC R: AATGCCGAGATCGAAGAGC F: GACAGGTGGAAGAGATAGAGTAAGC

R: TGAAGCCAAGGAGACAAAGC

GenBank Accession Number EF408809 EF408812 EF408813 EF408814 EF408811 EF408815 EU076547 EU076548 EF408816 EU076549 EF408818 EU076550 EU076551

Descriptive characters of 13 polymorphic nuclear microsatellite loci in Juglans cinerea. Repeat is the microsatellite motif observed in the cloned allele. Ta is the optimized annealing temperature for each locus. Number of alleles (Na), size range of amplicons (size of the cloned allele in parentheses), observed (Ho) and expected (He) heterozygosities and number of individuals genotyped (K) for each locus are based on individuals from a wild population (N = 63). * indicates significant deviations (p< 0.05) from Hardy-Weinberg equilibrium.

TABLE 2.2: RESULTS OF EXPERIMENTAL VERIFICATION OF NULL ALLELES

Locus name

N1

N2

Ho 1

Ho 2

GH value

Decision

B110

62

33

0.5

0.485

0.009

Retain original primer pair

B212

51

60

0.569

0.833

4.77*

Use new primer pair

B264

62

49

0.581

0.653

0.303

Retain original primer pair

21

Results of experimental verification of null alleles: comparison of observed heterozygosity from two sets of primers for each of three loci that suggested the presence of null alleles using G-test for heterogeneity. Observed heterozygosities (Ho) and number of individuals assayed (N) are given for both the original primer pair (1) and the redesigned primer pair (2). Significant changes in observed heterozygosity with the new primer set are marked with an *. The critical χ2 value used in G-test for heterogeneity for p = 0.05, ν = 1 is 3.84.

Microsatellite-containing sequences were run through a BLAST search (Altschul et al., 1997), and two highly similar (E-value < 1E-20 or Max score > 100) sequences were found using the blastn algorithm. Vitis vinifera whole genome shotgun (WGS) contigs VV78X193609.6 and VV78X277634.23 both had an E-value = 1E-30 and a max score = 141 when compared to locus jcin_B110. These contigs are not recognized as known genes.

2.3 Discussion We plan to use this marker set for describing genetic diversity in J. cinerea populations and elucidation of population dynamics in the presence of butternut canker. This information will be useful for conservation and restoration efforts for this threatened species. We showed that these markers produce a large number of alleles, allowing power for studies of both gene flow and relatedness. The markers show only small departures from equilibrium in a natural population, and do not appear to be linked. These markers are also currently being tested for amplification in other Juglans species, including J. ailantifolia, which will enable identification of non-J. cinerea individuals in natural butternut populations, germplasm collections, and disease screening programs.

22

CHAPTER 3: GEOGRAPHICALLY EXTENSIVE HYBRIDIZATION BETWEEN A NATIVE AND AN INTRODUCED FOREST TREE

3.1 Introduction Natural hybridization between sympatric forest tree species is common (Craft et al., 2002; Lexer et al., 2006; Lexer et al., 2005; Valbuena-Carabaña et al., 2005), but hybridization between introduced species and rare or economically valuable native species has raised concern among conservation geneticists (Hails, Morley, 2005; Mallet, 2005). Successful hybridization can result in adaptive evolutionary change if circumstances favor hybrid progeny (Lewontin, Birch, 1966; Rieseberg, 1997). However, successful genetic invasion involving disjunct species seems unlikely, as hybrids are usually assumed to be intrinsically or extrinsically unfit, e.g., the biological species concept, the Dobzhansky-Muller model of hybrid dysfunction, and the assumption of local adaptation (Bomblies et al., 2007; Coyne, Orr, 2004; Mayr, 1992). Therefore, the outcome of hybridization is debatable, and of theoretical and practical interest (Allendorf et al., 2001; Hails, Morley, 2005; Haygood et al., 2003). Many east Asian woody plant taxa can produce fertile hybrid progeny with their North American sister taxa, including species in Platanus, Ulmus, Castanea, Pinus, and Juglans (Wen, 1999). Cultivated hybrids between varietal forms of Japanese walnut (Juglans ailantifolia Carrière) and the native North American tree butternut (Juglans 23

cinerea L.) are such vigorous, fruitful trees that forest tree biologists have expressed concern over the possibility of an undetected, rangewide genetic invasion (Ostry, Woeste, 2004). Additionally, some varieties of J. ailantifolia and some hybrids have shown remarkable tolerance to Sirococcus clavigignenti-juglandacearum, a pathogenic fungus causing the disease butternut canker, a major cause of decline in butternut populations (Anderson, 1996; Anderson, LeMadeleine, 1978; Nair et al., 1979; Orchard, 1984). The adaptive potential of Japanese walnut genes could have an evolutionary and ecological impact on butternut (specifically for disease dynamics) if the hybrids produce successful descendents, but it is not known whether hybrids persist in natural settings (Ostry, Woeste, 2004; Schlarbaum et al., 1997a). To investigate the possibility of genetic invasion, We developed a set of nuclear microsatellite markers (Hoban et al., 2008) and a set of chloroplast markers (McCleary et al., 2009) to reliably identify an individual as one of the parent species or as a hybrid. Using recently developed Bayesian methods (Anderson, 2003; Anderson, Thompson, 2002), we can identify the type of hybrid with a degree of probability, and using the chloroplast we can ascertain the maternal ancestry of an individual. As no definitive silvic characters distinguish hybrids from J. cinerea (Ross-Davis et al., 2008a), identification using DNA markers is essential. We qualified both types of DNA markers on reference sets (J. ailantifolia N = 69 and J. cinerea N = 71) to assess accuracy of species identification, then genotyped putative butternuts from seven populations in Eastern North America (N = 187). We address the question: “Do interspecific hybrids occur and reproduce in natural populations, and if so, to what degree?”

24

3.2 Materials and methods:

3.2.1 Species J. ailantifolia, native to the mountain forests of Japan, was introduced to North America around 1870, was widely planted, and soon naturalized in woodlots, abandoned fields, and roadsides (Neilson, 1930; Sargent, 1894; Theiss, 1933). J. cinerea, native in eastern North American riparian and bottomland forest ecosystems, has suffered severe decline throughout the 20th century and is currently a species of concern on both state and federal threatened species lists (Fleguel, 1996b; Reed, Davidson, 1958; Rink, 1990; Schultz, 2003).

3.2.2 Study populations Assignment of hybrid status is greatly facilitated by use of a „reference‟ population, a group of individuals of suspected „pure‟ species origin, so we first collected a large number of putative J. ailantifolia and J. cinerea. The hybrid assignment method is not dependent on the putative assignment being correct, as only the genetic data is used, but having a number of „pure‟ genotypes helps to establish the allele frequencies (Anderson, Thompson, 2002; Latch et al., 2006; Maudet et al., 2002; Rosenberg et al., 2003). The J. cinerea reference set consisted of 71 individuals: 11 butternut cultivars from the USDA National Germplasm Repository (Corvallis, Oregon) and 60 individuals from a naturally regenerated, protected population in Kentucky. The J. ailantifolia reference set consisted of 69 individuals from germplasm repositories, arboreta, and botanic gardens. Between 2005 and 2008, we sampled 187 individuals from seven additional 25

populations of J. cinerea where hybridization is suspected. Samples consisted of ~ 500 g of twigs or leaves. DNA was extracted using a modified CTAB protocol.

3.2.3 Genotyping We genotyped with three different chloroplast Cleaved Amplified Polymorphic Sequence (CAPS) markers to detect species-specific chloroplast haplotypes (McCleary et al., 2009). CAPS products were visualized using electrophoresis in 1.5% agarose gels. We also genotyped with eight nuclear microsatellite markers: Jcin_B114, Jcin_B121, Jcin_B147, Jcin_B159, Jcin_B249, Jcin_B262, Jcin_B264 (Hoban et al., 2008), and WGA82 (Robichaud et al., 2006a). Fragment length polymorphisms were visualized on an ABI 3730 XL DNA Analyzer and scored using GENEMAPPER (Applied Biosystems, Inc.). Matching genotypes were identified using identity analysis in CERVUS (Kalinowski et al., 2007). Exact and near (mismatch = 1) matches were removed prior to genetic analysis, as duplicate genotypes bias the allele frequencies used for clustering algorithms (Anderson, 2003; Latch et al., 2006; Rosenberg et al., 2003). We removed 18 individuals from our reference sets and five from our natural populations. After removing duplicate genotypes, we calculated allele frequencies, FST and RST for each species‟ reference set using GENALEX (Peakall, Smouse, 2006) and In statistics using INFOCALC (Rosenberg et al., 2003).

3.2.4 Genetic analysis We used the Bayesian approach implemented by STRUCTURE (Pritchard et al., 2000) to identify admixed individuals. We tested values of K from two to ten and used the method of Evanno et al. (2005) to determine the number of genetic clusters (K) for 26

which the data were most likely. For the most likely K (K=2), we report estimated membership in each cluster for every individual, and 95% posterior probability intervals. We complement this analysis with the Bayesian approach in NEWHYBRIDS (Anderson, Thompson, 2002) to assign individuals to one of six genotypic classes: J. cinerea, J. ailantifolia, F1, F2, BC to J. cinerea, and BC to J. ailantifolia. These categories are based on expected patterns of Mendelian inheritance of alleles. We ran the analysis using the z and s options to incorporate knowledge about our reference sets. We also examined the posterior assignment of individuals in our reference sets.

3.2.5 Hybrid assignment The presence of the J. ailantifolia chloroplast in a native J. cinerea population is preliminary evidence of hybridization in that population. We classify individuals with <0.90 probability of membership in a STRUCTURE species cluster as likely admixed. We also assign each individual to the most likely (probability > 0.50) NEWHYBRIDS genotypic class. Individuals for which no single class had ≥ 0.50 probability are not assigned to a class.

3.3 Results All of the individuals in our reference sets had the chloroplast haplotype of the expected species. STRUCTURE analysis for the reference sets showed that all individuals in our reference groups had > 0.90 probability of belonging to the species expected (either J. cinerea or J. ailantifolia), with one exception suggesting admixture, the “Johnson” butternut cultivar (accession CJUG 10.001 from the USDA National

27

Germplasm Repository, Corvalis, OR). In the seven study sites, using STRUCTURE and our cutoff value (0.90), we report 60 admixed individuals, and 12 J. ailantifolia. Using NEWHYBRIDS, we report 55 probable hybrids, and 10 J. ailantifolia, which suggests general agreement, but occasional disagreement, between the two approaches. The most probable hybrid classes with our cutoff value (0.50) were: 29 F1s, 14 F2s, eight backcrosses to J. cinerea, and four not assigned (Figure 1). No individual had > 0.34 probability of backcross to J. ailantifolia. Of those assigned, twenty-five individuals were assigned to a hybrid class with ≥ 0.90 probability, 12 were assigned with ≥ 0.80 probability, and 15 were assigned with probabilities between 0.80 and 0.50 (Figure 1). Fifty of the 55 putative hybrids (90.9%) contained the J. ailantifolia chloroplast, and the remaining five contained the J. cinerea chloroplast. All seven populations showed evidence of at least one probable hybrid (Figure 2) and non-admixed native J. cinerea occurred at proportions from 0.05 to 0.92.

3.4 Discussion Our data indicate that introgression occurs across a large portion of the native species‟ range. Although hybridization was highest in fragmented, semi-rural landscapes (sites CT and NC), we also found small numbers of probable hybrids in five other locations, including three National Forests. Extensive introgression could alter the gene pool and evolutionary potential of the native species, with potentially cascading ecosystem consequences. Our data suggest caution should be taken in choosing individuals for restoration efforts.

28

Figure 3.1 Posterior probability distribution and the most likely genotypic class assignments of hybrid individuals. „Mixed‟ assignments had < 0.50 probability in any of the given hybrid classes.

Figure 3.2 Native range of butternut, and admixture in each collection site. Histograms represent the number of individuals assigned to each genetic class by NEWHYBRIDS. The four individuals with ≤ 0.50 probability in any class (the „mixed‟ hybrids) are not shown.

29

Most (90.9%) of the probable hybrids contained the J. ailantifolia chloroplast. As the chloroplast is maternally inherited in Juglans (Zhang et al., 2003), this result indicates that most maternal parents of hybrids are J. ailantifolia. One possible explanation is the higher number of J. cinerea and thus higher amount of J. cinerea pollen in the pollen pool, so most paternal parents will be J. cinerea, similar to observations in Populus (Lexer et al., 2005). Alternatively, the explanation may be partial one way intrinsic incompatibility, which is common in angiosperms (Tiffin et al., 2001). Interestingly, all eight probable backcross individuals are to J. cinerea. This also may be due to high representation of J. cinerea pollen in the pollen pool. While eight is an exceedingly small sample size, we can speculate that a long term consequence of repeated backcrossing to J. cinerea could be chloroplast capture, in which the J. ailantifolia chloroplast is retained in a mostly J. cinerea nuclear background. However, interaction between hybrids and butternut canker may preserve large blocks of the J. ailantifolia genome, if J. ailantifolia tolerance has a genetic basis. The extent of J. ailantifolia introgression over time may depend on the strength of selection against J. cinerea and asymmetry in survival of backcross individuals. We note that our Bayesian probabilities are based on a small number of moderately to highly differentiated markers, so assignments are rough estimates. The four unclassifiable individuals and the fifteen individuals with ≤ 0.80 probability in a single hybrid class reflect the limits of our ability to distinguish genotypic classes with the current marker sets. We also note that the two methods, NEWHYBRIDS and STRUCTURE, occasionally were not in agreement about hybrid assignment. Such

30

assignments should be treated with caution. Genotyping with additional well differentiated markers will improve the resolution of this analysis and the number of F1 versus advanced generation hybrids may change. Nevertheless, the presence of J. ailantifolia chloroplast and many highly probable hybrids in multiple populations allows us to strongly suggest that hybrids persist and interbreed with the native species in natural settings, in contrast to expectations for disjunct species. As human migration and commerce among regions and continents will continue, we suggest that natural hybridization is likely to be an increasingly pressing issue for many taxa, including rare native species, cultivated crops and orchard trees. Burgess et al (2008) recently demonstrated that interspecific mating between the introduced Asian white mulberry (Morus alba L.) and the rare North American native red mulberry (Morus rubra L.) occurs more frequently than intrapecific mating in the native mulberry, increasing the risk of local extinction for the native taxon. Others have demonstrated the possibility of gene flow between crop-wild relative complexes (Ellstrand, 2003; Haygood et al., 2003; Snow et al., 1998), as well as many native-introduced taxa (Allendorf et al., 2001; Levin et al., 1996; Wolf et al., 2001). Hybridization may lead to the development of aggressive or weedy introduced species, complicating biological control (Campbell et al., 2006; Gaskin et al., 2009; Mercer et al., 2006). Genotyping using nuclear and chloroplast markers provides an efficient and cost-effective way to detect hybridization in natural populations, and track the progress of hybrid populations over space and time. Future work in this project will incorporate more natural populations, and will investigate the dynamics of hybrid populations over time, the source of initial introduction of exotic

31

genotypes, and the selective factors that determine whether introgression is limited or extensive, such as fitness differences under disease pressure.

32

CHAPTER 4: RANGE-WIDE DISTRIBUTION OF GENETIC DIVERSITY IN THE TREE JUGLANS CINEREA: A PRODUCT OF HISTORICAL RANGE SHIFTS, NOT ECOLOGICAL MARGINALITY OR RECENT POPULATION DECLINE

4.1 Introduction A central and continuing aim of evolutionary biology is to identify and describe the processes that determine the spatial distribution and abundance of genetic diversity. Many distinct processes, working at different time and spatial scales, shape this distribution, including contemporary ecological conditions and both historic and contemporary population bottlenecks (Foll, Gaggiotti, 2006; Loveless, Hamrick, 1984). A pressing challenge in population genetics is to disentangle the effects of each process (Gaggiotti et al., 2009; Grivet et al., 2008; Zellmer, Knowles, 2009). Identifying those having the greatest impact is a logical first step in developing predictive models for population responses to future environmental change (Hamrick, 2004; Jaramillo-Correa et al., 2009; Petit et al., 2008). Our goal in this investigation is to determine whether contemporary population decline, contemporary ecological marginality, or historical range shifts has had the most influence on current patterns of range wide diversity in Juglans cinerea L. (butternut), an outcrossing, wind-pollinated canopy tree native to eastern North America.

33

Contemporary decline may reduce population size and abundance, lower individual reproductive output, decrease migration rates, and increase rates of local extinction. These changes in population demography may result in substantial loss of genetic diversity and increased differentiation due to genetic drift (Allendorf, Ryman, 2002; Fernández-M., Sork, 2007; Frankham, 1995). Both natural and anthropogenic causes of decline may vary in time and in spatial scale (Vellend, 2004), affecting parts of the range intensely and other parts not all (Hutchison, 2003; Jump, Penuelas, 2006). Butternut populations have experienced severe decline in the 20th century, largely due to the introduced fungus Sirococcus clavigignenti-juglandacearum, the agent of butternut canker. Butternut canker has had greatest demographic impact in the southeastern United States, where the disease has caused mortality of 70-90% (Anderson, LeMadeleine, 1978; Ostry, Woeste, 2004; Schlarbaum et al., 1997a). If this disease has had a significant genetic impact, it would be most pronounced in this region, the southern edge of the contemporary range. Contemporary ecological marginality, in particular abiotic conditions towards the periphery of the native range, could also alter population dynamics, reducing local population sizes and weakening connectivity (Gapare et al., 2005; Gaston, 1996; Lawton, 1993; Lönn, Prentice, 2002). An increase in genetic drift and decrease in gene flow should cause reduced diversity and increased differentiation at the range edge (Lesica, Allendorf, 1995; Maruyama, Kimura, 1980). Butternut is typically found in riparian habitats in temperate deciduous forests, but populations persist on the range edges in the transition from the central hardwood forest to the unsuitable habitat of northern boreal

34

forest, southern and eastern pine forest, and western prairie (Fralish, 2002; McDaniel, 1979; Rink, 1990). The influences of contemporary decline and marginal environments take place on a background of historical processes, perhaps the largest of which was the repeated advance and retreat of northern ice sheets during the Pleistocene (Gamache et al., 2003; Hewitt, 2000; Petit et al., 2004a). During each glacial retreat, temperate species recolonized northward in a series of successive founder events which may have resulted in substantial loss of diversity and increased differentiation along the route (Excoffier et al., 2009; Petit et al., 2004a). Butternut has colonized at least half of the contemporary range since the retreat of the Wisconsinian ice sheet ~18,000 years ago (Davis, 1983; Pielou, 1992), and genetic impact of this geographically extensive range shift may still persist in northern populations. A primary problem in distinguishing the relative impacts of these three processes is that predicted genetic patterns for different processes may be concordant across some regions. For example, low genetic diversity in northern edge populations may result from range shifts or from marginal environmental conditions (Figure 1), but most investigations are not explicitly constructed to distinguish between them. Identifying the underlying process will often require sampling across much of the native range (Lawton, 1993; Sagarin, Gaines, 2002; Sagarin et al., 2006). A secondary challenge is to construct a statistical framework that supports the influence of one process and fails to support the others (Eckert et al., 2008; Gaggiotti et al., 2009). Methods to evaluate the relative importance of each in natural systems have only recently emerged (Gaggiotti et al., 2009; Zellmer, Knowles, 2009). 35

Figure 4.1 Three models for prediction of range-wide genetic diversity and differentiation. Filled area: contemporary native range. Red to blue gradient: high diversity and low differentiation to low diversity and high differentiation. Yellow circles: populations. (a) contemporary population decline with a southern epicenter, (b) ecological marginality of the periphery, (c) range shift following the LGM and (d) population labels. Our aim in this study was to explicitly test the relative strengths of contemporary decline, contemporary ecological marginality, and range shifts as predictor variables for the spatial distribution of genetic diversity in butternut. We combined an extensive sampling effort with a comparative statistical framework to determine which process has most influenced genetic diversity and population differentiation in butternut. In our

36

conclusion, we interpret our results in light of past support for each process and discuss the application of these results for conservation of forest trees in a time of climate change.

4.2 Materials/ methods

4.2.1 Overview of methods We employed quantitative metrics to measure the genetic response of each population to ecological processes, for 29 populations covering most of the native range: four measures of diversity (number of alleles, allelic richness, number of rare alleles, heterozygosity), two measures of differentiation (FST, RST) and the bottleneck statistics M-ratio, allele frequency mode shift and heterozygote excess. Then we constructed a statistical framework in which each process of interest is represented by a qualitative and a quantitative predictive variable, similar to the approach of Gapare et al. (2005) and Schwartz et al. (2003). Our goal was to identify the variable(s), and hence the process, with the greatest predictive power for diversity, differentiation, and bottlenecks. We used t-tests and simple linear regression to test the predictive power of each variable alone, and two model comparison procedures to test the predictive power jointly. To complement these model based tests, we used a Bayesian clustering method to test for population genetic structure without the constraints of an a priori model. Lastly, we tested the null hypothesis that none of these processes are influencing population structure, in which case differentiation would be a function of isolation by distance alone. We utilized this suite of analyses (qualitative and quantitative, Bayesian and classical

37

model simplification, and simple and multiple regression) to ensure that our results are not dependent on one metric.

4.2.2 Species Butternut is a monoecious, outcrossing, diploid tree in the Juglans genus (walnuts), typically found at low densities. Trees are generally reproductively mature at 10-15 years and individual trees rarely live longer than 70 years. The fruit, also called a butternut, contributes to mast for wildlife (Fleguel, 1996b; Ostry, Pijut, 2000).

4.2.3 Collections We sampled twigs from butternuts in 29 naturally regenerating forests (Table 1, Figure 1), and recorded their GPS location. Tissue samples were stored at -20C until DNA extraction, as previously described (Hoban et al., 2008).

4.2.4 Genotyping We genotyped individuals for 11 microsatellite markers, ten from butternut (Hoban et al., 2008) and one from black walnut, J. nigra L. (Robichaud et al., 2006a). PCR and genotyping protocols are previously reported (Hoban et al., 2008). As hybridization between butternut and an introduced exotic species, Japanese walnut (J. ailantifolia Carrière), is common at some locations (Hoban, 2009; Ostry, Woeste, 2004), we used both nuclear and chloroplast DNA markers to identify and remove hybrids (Hoban, 2009; McCleary et al., 2009).

38

TABLE 4.1 POPULATIONS USED IN THE STUDY

Pop

N

Lat.

Long.

Disease Impact

Tdisease

CorePeriphery

Dedge (km)

Ice Margin

Dice (km)

Connectivity

1

35

34.75

90.56

High

1940

Periphery

-29

South

210

12.55

2

26

37.5

91.46

High

1940

Core

86

South

330

11.94

3

24

37.28

91.42

High

1940

Core

72

South

330

11.76

4

18

37.22

91.4

High

1940

Core

56

South

350

11.94

5

13

36.96

91

High

1940

Periphery

48

South

350

11.74

6

31

36.85

91.13

High

1940

Periphery

31

South

370

11.88

7

162

36.01

85.81

High

1940

Periphery

-33

South

350

9.94

8

64

37.29

86.01

High

1940

Core

87

South

210

9.48

9

21

38.68

79.36

High

1940

Core

108

South

260

8

10

14

38.7

79.08

High

1940

Core

93

South

270

8.02

11

13

38.85

78.15

High

1940

Periphery

46

South

300

8.17

12

19

38.31

78.63

High

1940

Periphery

32

South

330

8.27

13

18

41.94

78.86

High

1940

Core

297

South

10

7.55

14

26

41.93

79.11

High

1940

Core

301

South

20

7.52

15

18

41.46

78.92

High

1940

Core

352

South

60

7.52

16

45

41.64

79.43

High

1940

Core

334

South

20

7.49

17

21

44.78

72.51

Low

1983

Periphery

30

North

-390

10.42

18

21

44.55

72.9

Low

1983

Periphery

35

North

-410

10.27

19

20

44

72.72

Low

1983

Core

67

North

-330

10.23

20

20

43.09

72.45

Low

1983

Core

144

North

-280

10.31

21

28

45.57

88.53

High

1940

Periphery

6

North

-160

11.27

22

20

44.66

89.01

High

1940

Core

81

North

-60

11.15

23

40

42.93

88.72

High

1940

Core

284

North

-50

10.48

24

26

45.5

76.99

Low

1991

Periphery

-19

North

-370

9.03

25

29

44.71

78.42

Low

1991

Periphery

-12

North

-260

8.39

26

24

44.51

80.15

Low

1991

Periphery

30

North

-270

8.26

27

38

45.98

66.21

Low

1997

Periphery

-281

North

-860

15.48

28

33

46.08

66.73

Low

1997

Periphery

-251

North

-830

15.08

29

37

46.68

65.92

Low

1997

Periphery

-341

North

-920

15.94

39

4.2.5 Population diversity statistics A total of 904 individuals were retained for analysis. Markers were tested for linkage (gametic) disequilibrium (LD) at each population and locus, using GENEPOP (Raymond, Rousset, 1995a). We used Holms‟ correction (Holm, 1979) for multiple tests. We tested for evidence of null alleles using MICRO-CHECKER (Van Oosterhout et al., 2004). We tested per locus and per population deviations from Hardy-Weinberg equilibrium (HWE) using GENEPOP. Using GDA (Lewis, Zaykin, 2001), we calculated per population number of alleles (A), allelic richness (AR), and expected (HE) and observed heterozygosity (HO). We also manually counted the number of „rare alleles‟ which we define as alleles present in five or fewer populations.

4.2.6 Population differentiation statistics To test whether population allele frequencies differ, we used GENEPOP Option 3, which tests the null hypothesis that population allele frequencies are drawn from the same distribution. We calculated pair-wise FST (a general measure of differentiation) between each population using FSTAT (Goudet, 1995) (significance tested using 1,000 permutations). Using SPAGEDI (Hardy, Vekemans, 2002) we calculated RST (a differentiation statistic incorporating allele size information) and performed similar permutation tests. We visually displayed genetic distance between populations using a principal component analysis (PCA) of pair-wise FST generated with GENALEX (Peakall, Smouse, 2006).

40

4.2.7 Bottlenecks We used three tests to detect signatures of population bottlenecks: M-ratio, heterozygote excess, and allele frequency mode shift. The first, M-ratio (Garza, Williamson, 2001), is the ratio of the number of occupied allelic states over the number of possible states (the range from smallest to largest allele). The test is based on the assumption that bottlenecked populations are likely to have more empty allelic states than non-bottlenecked populations due to loss of rare alleles. Garza and Williamson calculated the M-ratio for simulated and actual datasets and found that it accurately detects populations that have undergone large demographic reductions (e.g., effective population size, Ne, from 5000 to 50) at a critical value of 0.68. We calculated the Mratio for each population using ARLEQUIN (Excoffier et al., 2005). The second test identifies a significant heterozygosity excess. We used a onetailed Wilcoxon signed-rank test, implemented in BOTTLENECK (Luikart, Cornuet, 1998) to compare actual heterozygosity to a distribution obtained through a coalescent simulation based on a two-phase microsatellite mutation model. Thirdly, we used a qualitative method, also implemented in BOTTLENECK, to examine a mode shift in allele frequency distribution. In a population at equilibrium, more alleles are expected at low frequency (<0.1) than in any other equal-sized frequency bin (e.g. 0.1-0.2), making the frequency distribution skewed (i.e. „L-shaped‟). During a bottleneck, low frequency alleles will be lost faster than higher frequency alleles, reducing the „L‟ and causing a mode-shift (Luikart et al., 1998; Luikart, Cornuet, 1998). The M-ratio method better detects older bottlenecks, while the heterozygote excess and mode-shift tests detect more recent bottlenecks, because heterozygote excess

41

is reduced rapidly as a population expands and reaches a new migration-drift equilibrium (< 4 Ne generations), while gaps in the allele frequency spectrum can only be filled in by new mutations (or migration) (Williamson-Natesan, 2005).

4.2.8 Qualitative and quantitative predictive variables Qualitative. We assigned each population to one of two categories: high or low disease impact, peripheral or core location, and north or south of the ice margin at last glacial maximum, LGM (Table 1). Populations exposed to butternut canker for more than 60 years were classified as „high impact‟ (Anderson, LeMadeleine, 1978; Schlarbaum et al., 1997a), and populations exposed to butternut canker for one generation or less (~ 25 years) (Hopkin et al., 2001; Schmalz, Bergdahl, 2006) were considered „low impact‟ of the disease. We then defined a peripheral strip, an area around the range boundary wide enough to account for 5% of the range area, ~50 km for butternut‟s range (Little, 1971). Populations within or beyond this strip were classified as peripheral. We also tested a looser definition of periphery, ~10% of the range edge, or 100 km. Lastly, we classified populations as south or north of the southernmost advance of the Wisconsinian ice margin at LGM. Quantitative. Each population was assigned a predictive continuous variable representing the strength of the given process (Table 1). First, we assigned each population the predictor variable Tdisease, the year in which butternut canker was first detected or estimated to be in that state or province. Next, we measured the distance of each population to the nearest range edge (Dedge) with ARCINFO (Environmental Systems Research Institute, 2004), assigning negative values to disjunct populations. Lastly, we calculated the distance from the population to the southern edge of the Wisconsinian ice 42

margin at LGM (Dice), giving negative values to those populations north of the margin. The ice margin provides a relative measure of the minimum colonization distance after the ice receded.

4.2.9 Qualitative and quantitative statistical analyses We used a one tailed t-test to test for significantly different means of the population genetic response variables for each category, similar to the method used by Jump & Peñuelas (2006) to test whether fragmented forests have greater allelic diversity and heterozygosity than non-fragmented populations. Additionally, we use a chi-square test to compare the proportion of bottlenecked populations in each category, for each bottleneck test. Next, we used multiple linear regression for each of the seven response variables, using the three predictor variables, Tdisease, Dedge, and Dice. This is an approach similar to the method used by Vellend (2004) to test the effects of land use history, habitat heterogeneity, and isolation on both species and genetic diversity. We used a linear model with pair-wise interactions: ŷi = β0 + β1X1 + β2X2 + β3X3 + β4X1X2 + β5X1X3 + β6X2X3 + ε where β0 is the intercept, X1 is Tdisease, X2 is Dedge, X3 is Dice, and ŷi is the response variable. We performed model simplification using backwards removal of nonsignificant terms, using R (R-CDT, 2005). The effect of each removal was checked using the Akaike Information Criterion. We also performed simple linear regression for each of the seven response variables using each of three the predictor variables alone, to test their predictive power apart from the others. For all the tests in this section, we report significance both with 43

and without Holm‟s correction for multiple comparisons. Additionally, as populations 27-29 are disjunct, their inclusion may upwardly bias significance values for both the qualitative and quantitative tests. We therefore performed tests with and without these populations. Lastly, we implemented a Bayesian model comparison method using GESTE (Foll, Gaggiotti, 2006), which first estimates population-specific F-statistics and then evaluates the predictive relationship of the variables (e.g. latitude and connectivity). This method generates a posterior probability for each model considered, and can consider any number of predictor variables. A recent study considered 10, which generated 210 (1024) model probabilities. Note that this is a posterior probability (Pr), not a p-value, as in classical model comparison. Models take the form of a general linear model: μ = α0 + α1G1 + α2G2 + … αnGn where μ is the population specific parameter ln(θ) (θ being directly related to FST), and G1-Gn are the environmental variables that may influence θ. In contrast to model simplification, this method estimates the probability of every possible model in a computationally efficient manner, making evaluation of multiple predictor variables feasible. This method allowed us to test for the effects of four additional predictor variables to the original three (Tdisease, Dedge, Dice): sample size, connectivity, latitude, and longitude (Table 1). We calculated connectivity as a measure of the average distance between a population and all other populations (Foll, Gaggiotti, 2006). As in Gaggiotti et al. (2009), we first performed analysis with all factors, and then using only those having the top five cumulative posterior probabilities (for these data, factors with Pr>0.10).

44

Because this is a new approach, we compare values of FST from GESTE to those obtained using GENEPOP, FSTAT, and SPAGEDI (Weir, Cockerham, 1984).

4.2.10 Bayesian analyses of population structure We used two Bayesian clustering programs, BAPS (Corander et al., 2003) and STRUCTURE (Pritchard et al., 2000), to identify population structure without a priori assumptions. Both programs infer the parameters of a model consistent with the multilocus genotypic data provided. The advantage of using a Bayesian approach is that these programs explore a range of possible parameters to produce a posterior distribution of parameters deemed most likely according to the data, without the constraints of a specific model. STRUCTURE simultaneously estimates population allele frequencies and assigns individuals to populations, using the criterion of minimum deviation from HWE within populations. BAPS partitions the data among groups of individuals (i.e. detects which sampling locations are different) based on similarity of population allele frequencies, comparable to a Fisher‟s exact test for population differentiation. The biological interpretation of population structure found using BAPS and STRUCTURE is similar to FST, in that migration and shared ancestry result in weaker structure, and reduced gene flow and increased drift result in stronger structure. Strong structure concordant with geographical contrasts (e.g. the range edge vs. the core) provides support for the process proposed. Both are conservative methods that tend to not report spurious structure. Settings for STRUCTURE: admixture model, infer α, uniform prior on α, allele frequencies correlated, locprior information on, burn-in=25,000, sweeps=50,000, K from one to 15, with five replicates. Settings for BAPS: clustering of groups, K from one to 15, with five replicates. 45

4.2.11 Isolation-by-distance: Mantel tests Wind-pollinated forest trees typically have characteristics that may attenuate the signal of historical and recent bottlenecks, such as long distance dispersal potential, extended juvenile periods, and relatively long life spans (Liepelt et al., 2002; Petit, Hampe, 2006; Robledo-Arnuncio et al., 2005). We therefore test whether differentiation may be due to isolation-by-distance of populations at equilibrium (Hutchison, Templeton, 1999; Kimura, Weiss, 1964; Slatkin, 1993), in which case geographic and genetic distances should be strongly correlated. Using SPAGEDI we performed a Mantel test of pair-wise FST/(1- FST) on the natural logarithm of distance (Rousset, 1997), with 20,000 permutations. We conducted this test with all populations (n=29), and with the following subsets: with disjunct populations removed (n=25), within the core (n=14), south of the ice margin at LGM (n=16), low disease impact (n=10), and the largest BAPS cluster (n=22). The inclusion of connectivity in GESTE is another test of isolation-by-distance (Gaggiotti et al., 2009).

4.3 Results

4.3.1 Descriptive statistics No locus had fewer than five alleles in a given population. The total number of alleles (A) observed across all eleven markers was 265 (mean A per locus=24.1, mean A per population=109). The number of alleles within populations ranged from 174 (65.7% of 265) in population 8 to 78 (29.4% of 265) in population 29).

46

Overall, departures from HWE expectation, LD and null alleles were minimal. Of 1595 possible population-locus combinations, only 32 (2%) showed significant LD. These were distributed across 11 populations, with only one population having > 4 instances. No pair of loci showed significant linkage disequilibrium in more than two populations. Null alleles were reported as likely for marker jcinB264 in 13 of the 29 populations, with the estimated percentage of null alleles in a population ranging from 9% to 21% (mean=15.8%) in these populations. We recalculated all descriptive statistics with and without this locus and used t tests to compare values. We observed no significant differences (data not shown). We observed slight but significant heterozygote deficiencies in 14 populations which were attributable to small departures at particular loci. No population had more than four loci departing from HWE. Small per locus departures from HWE are expected in small populations (Marsico et al., 2009; Schwartz et al., 2003).

4.3.2 Population differentiation Global and pair-wise population differentiation was moderate and highly significant in most cases. Of 406 population pair-wise comparisons, 396 were significant (p < 0.05). The overall test for population differentiation was also significant (p < 0.0001). Mean pair-wise FST was 0.0452 (σ = 0.025, max = 0.126) and mean pair-wise RST was 0.0443 (σ = 0.049, max = 0.271). The PCA based on genetic distances indicates that populations north of the ice margin at LGM tend to be distant both from each other and from the other populations, while populations south of the ice margin exhibit less differentiation (Figure 2).

47

All loci showed some deviation from the stepwise mutation model (i.e. one base pair difference between alleles). Loci displayed between 1% and 36% non-stepwise mutations (mean=7.4%). However, RST values showed strong correlation with FST values, suggesting that RST is robust to moderate violations of the assumptions underlying the mutation model.

4.3.3 Bottlenecks Twelve populations were below the critical value for the M-ratio test, seven populations showed a significant heterozygote excess, and six showed a mode shift in allele frequencies. The only significant biogeographical pattern to these bottlenecks was the M-ratio, which showed a higher proportion of bottlenecked populations north of the ice at LGM (Table 3, Supplemental Table A1). Recent investigations have demonstrated evidence of historical, but not contemporary, bottlenecks by comparing results from these three methods (Henry et al., 2009; Jordan et al., 2008).

48

49 Figure 4.2. Principal component analysis based on FST. The first and second principal components account for 0.28 and 0.27 of the variance, respectively. Population labels as in Figure 1 and Table 1. Black circle: north of ice margin at LGM; open circle: south of the ice margin at LGM.

TABLE 4.2 GENETIC DIVERSITY STATISTICS FOR ALL POPULATIONS

Pop 1

A 10.818

Ar 7.428

Rare alleles 12

HE1 0.788

HO 0.687

FST2 0.051

RST 0.043

M ratio 0.702

Het excess3 N

Mode shift N

2

11.273

8.058

5

0.837

0.815

0.027

0.019

0.704

Y

Y

3

11.091

8.2

6

0.837

0.803

0.029

0.051

0.726

N

N

4

9.727

7.593

2

0.817

0.778

0.033

0.018

0.712

N

N

5

7.091

6.646

4

0.778

0.735

0.061

0.047

0.602

N

N

6

10.273

7.214

2

0.819

0.733

0.034

0.044

0.765

Y

Y

7

15.818

7.786

28

0.839

0.801

0.042

0.041

0.862

N

N

8

14.364

8.37

15

0.85

0.836

0.025

0.02

0.853

Y

N

9

9.909

7.875

7

0.842

0.818

0.036

0.033

0.647

N

N

10

8.545

7.8

4

0.811

0.875

0.052

0.014

0.681

N

N

11

9

8.345

3

0.848

0.806

0.028

0.015

0.664

N

N

12

10.545

8.556

6

0.839

0.794

0.037

0.033

0.685

N

Y

13

9.909

8.216

5

0.859

0.79

0.03

0.009

0.684

N

N

14

10.636

8.037

8

0.84

0.795

0.03

0.019

0.716

N

N

15

9.727

8.065

5

0.842

0.822

0.039

0.097

0.697

N

Y

16

11.091

7.105

6

0.801

0.764

0.047

0.025

0.703

N

N

17

8.545

6.781

1

0.814

0.885

0.047

0.04

0.635

Y

N

18

9.182

7.349

1

0.818

0.805

0.044

0.022

0.665

N

N

19

10.545

8.264

4

0.842

0.826

0.026

0.019

0.687

N

N

20

9.545

7.645

9

0.812

0.789

0.046

0.076

0.679

N

Y

21

7.818

6.193

3

0.776

0.79

0.072

0.1

0.583

N

N

22

8.545

6.62

6

0.766

0.794

0.064

0.044

0.621

Y

Y

23

8.636

6.007

4

0.753

0.723

0.069

0.071

0.603

Y

N

24

8.636

6.747

7

0.778

0.756

0.07

0.076

0.631

N

N

25

10.182

7.634

1

0.836

0.837

0.038

0.079

0.685

N

N

26

10.364

8.085

7

0.858

0.821

0.027

0.016

0.724

N

N

27

8.364

6.197

3

0.783

0.761

0.063

0.078

0.572

Y

N

28

8.909

6.77

5

0.807

0.798

0.059

0.065

0.641

N

N

29

8.273

5.947

2

0.751

0.752

0.088

0.07

0.588

N

N

Population genetic diversity, differentiation and bottleneck statistics. Values are averaged across all loci, except for rare alleles, which are summed across loci. 1 Heterozygosity expected under HWE. 2 Mean pairwise FST and RST. 3Significant heterozygosity excess: bottleneck test.

50

TABLE 4.3 RESULTS FROM QUALITATIVE TESTS

Test Proportion of populations showing bottlenecks (X2 tests)

Seven population genetic variables (t-tests)

Response variable

Contemporary decline (Low/High)

Ecological marginality (C/P)1

Range shift (N/S)1

M ratio

ns

ns

p = 0.003

Het. excess

ns

ns

ns

Mode shift

ns

ns

ns

Allele number

p = 0.04

ns

p = 0.0082,3

Allele richness

ns

p = 0.034

p = 0.0012,3

Rare alleles

p = 0.046

ns

p = 0.0382

Heterozygosity

ns

ns

p = 0.0092,3

FST

ns

p = 0.035

p = 0.0042,3

RST

ns

ns

p = 0.0062,3

M-ratio

p = 0.029

ns

p = 0.0012,3

1

C/P core/periphery, N/S north/south. Also significant (p < 0.05) after Holm's correction. 3 Also significant (p < 0.05) after removing disjunct populations. 2

4.3.4 t-tests We used t-tests to compare the means of the seven response variables (four measures of diversity, two of differentiation, and one of historic bottlenecks) between the following categories: low vs. high disease impact, core vs. periphery, and north vs. south of the ice margin. These categories represent the three processes of interest: contemporary decline, contemporary ecological marginality, and postglacial range shifts.

51

For disease impact contrast, three tests were significant and for the core-periphery contrast, two tests were significant. For the north vs. south of the ice margin contrast, seven tests were significant (Table 3, Supplemental Table A2). Under the alternative definition of the periphery, core-peripheral populations had no significant differences (Supplemental Table A3). After multiple comparison correction and removal of disjunct populations 27-29, the north vs. south of the ice margin comparisons were still significant for all seven and six response variables, respectively, while the other categories were never significant (Table 3, Supplemental Tables A4, 5).

4.3.5 Linear models In these analyses, the predictor variables are quantitative rather than categorical (Tdisease, Dedge, and Dice), and represent the strength of the three processes of interest. Both multiple and simple linear regressions reveal a stronger influence of the range shift process. Using multiple regression, we obtained a minimal adequate model explaining each response variable (Supplemental Table A6). Tdisease was retained in four, Dedge was retained in three, and Dice was a retained as a significant predictor for all seven. In all cases, the single most significant predictor was Dice (Table 4). This result was unaffected by multiple comparison correction (Supplemental Table A7). When populations 27-29 were removed, Dice was still the most significant predictor, and was retained in all models (Table 4, Supplemental Table A8). Using simple regression, Dice was a significant predictor for six response variables (all but rare alleles), Dedge for two, and Tdisease for none (Table 4, Supplemental Table A9). After multiple comparison correction, Dice had significant predictive value for allelic richness, M-ratio, and FST (and near significance

52

for the other four), while Dedge and Tdisease had no significant predictive value for any response variables (Table 4, Supplemental Table A10).

TABLE 4.4 RESULTS FROM QUANTITATIVE TESTS

Response variables

Method

Contemporary decline (Tdisease)

Ecological marginality (Dedge)

Range shift (Dice)

A. p value for predictors evaluated separately

Simple linear regression

Allele number

ns1

ns

p = 0.030

Allelic richness

ns

p = 0.027

p = 0.002

Rare alleles

ns

ns

ns

Heterozygosity

ns

ns

p = 0.034

FST

ns

ns

p = 0.003

RST

ns

p = 0.018

p = 0.019

M-ratio

ns

ns

p = 0.003

B. Retention in the most predictive model for predictor variables evaluated jointly

Multiple regression with model simplification

Bayesian model comparison

Allele number

-

-

X*

Allele richness

X

X

X*

Rare alleles

-

-

X*

Heterozygosity

X

X

X*

FST

X

X

X*

RST

-

-

X*

M-ratio

X

-

X*

-

-

X*

FST

A: p value for the regression using three predictors evaluated separately. B: Retention in the most predictive model for predictor variables evaluated jointly. 1 ns, not significant (p > 0.05). * Most significant predictor

53

4.3.6 GESTE The Bayesian model testing procedure also suggests a stronger influence of the range shift process. When considering all seven predictive variables, the two with the highest cumulative probability were Dice (Pr=0.433) and latitude (Pr=0.306), and the single model with the highest probability (Pr=0.258) included only Dice (Table 4, Supplemental Table A11). Using the five best factors from this initial run, latitude and Dice again resulted in the highest cumulative probability (Supplemental Table A11). The error for the most probable model and the 95% credible interval are within the bounds suggested by Foll and Gaggiotti (2006). Values of FST from GESTE were highly correlated with values from traditional methods of estimating FST, but were slightly inflated.

4.3.7 Bayesian analyses of population structure The clustering methods also suggest population differentiation due to bottlenecks from historical range shift. For both methods, at the most likely K values, population structure is most apparent north of the ice margin (Figure 3).

4.3.8 Isolation-by-distance: Mantel tests A weak but significant pattern of isolation-by-distance was observed across all populations and within each subset examined (Supplemental Table A12). Similar values were obtained with FST instead of FST/(1- FST), and with raw distance values instead of log values (data not shown). Additionally, connectivity was a weak predictor in the GESTE model (Supplemental Table A11), suggesting that distance between populations is a poor predictor of differentiation. 54

55 Figure 4.3. Bayesian cluster analyses. (a) STRUCTURE (K=11), colors represent the cluster having the highest Q value for that population (one cluster did not have the highest Q in any population, so only ten colors are displayed); (b) BAPS (K=7), colors represent distinct BAPS clusters; (c) Actual Q values per population, showing admixture of 11 clusters including gray cluster not appearing in (a); populations to the left of the arrow are south and populations to the right are north of the ice margin at LGM (double dashed line).

4.4 Discussion

4.4.1 Study summary Our study compared the relative influence of three processes (contemporary population decline, contemporary ecological marginality, and historical range shifts) on the distribution of genetic diversity in butternut. We first built a hypothesis-testing framework by describing explicit predictions under each process. Second, we implemented a sampling scheme designed to include populations under high and low influence of each process. Thirdly, we compared the three proposed processes by adapting both traditional and Bayesian techniques used in previous investigations of gene flow, but to our knowledge never used to compare all three processes. With this hypothesis-driven, multistep approach, we have shown that range shifts are the major determinant of range-wide structure and diversity in butternut, and not contemporary population decline or contemporary ecological marginality. Our results provide a basis for further investigations on the postglacial population dynamics of other eastern North American species.

4.4.2 Contemporary population decline We observed a latitudinal cline in diversity opposite to that predicted under contemporary decline. In locations where population decline due to disease has been most severe, diversity is actually greatest. Other investigations have shown that brief and even extreme population declines may not have lasting genetic signatures (Brown et al.,

56

2007; Okello et al., 2008; Yao et al., 2007). The magnitude and the duration of a bottleneck will determine its genetic signature in a population. While drastically reduced, current butternut populations remain at appreciable sizes (> 50 individuals in some populations). Overlap among generations, persistence of a few trees that survive infection, and gene flow among local populations may delay the genetic impact of population reductions. Retention of substantial effective population sizes despite a sharp decline in population census can occur in species in which gene flow among local populations and across generations is high (England et al., 2003; Johansson et al., 2006). Previous studies of genetic diversity in the Juglans genus have produced inconsistent results. Very low genetic diversity for isozymes, among the lowest reported for trees (Vendramin et al., 2008), was reported in butternut populations in Quebec, New Brunswick, and Vermont (Morin et al., 2000). A study in 43 populations of black walnut (J. nigra) across the range core (Victory et al., 2006) detected high diversity and low differentiation, but the effect of range edges was not explored. A previous investigation of genetic diversity in butternut (Ross-Davis et al., 2008b) showed high diversity and low structure in the center of the range, but the authors did not test for hybrid individuals or sample along range edges, limiting their conclusions. While we report high diversity in the center of the range, similar to Ross-Davis et al., we demonstrate a reduction in diversity in northern populations, consistent with Morin et al.‟s first suggestion.

4.4.3 Ecological marginality of contemporary range edge Previous studies have shown lower diversity in marginal populations in both plants and animals (Cassel, Tammaru, 2003; Hutchison, 2003; Schwartz et al., 2003), with the strongest support from investigations pairing demographics with genetics (Lönn 57

and Prentice 2002). However, many studies are not explicitly designed to test alternative hypotheses and authors acknowledge that effects of range shift or anthropogenic disturbance are viable alternative explanations. Although a pattern of low diversity in northern edge populations and high diversity in core populations was evident in butternut, we also observed high diversity at the southern and eastern peripheries. Furthermore, when the three processes are compared, the historical range shift hypothesis is consistently a better predictor. Additionally, peripheral populations were no more likely to show recent or historical bottlenecks than core populations. These observations may have two explanations. Population dynamics (size, abundance, persistence) on the range edge may not be substantially different from those in the core. Alternatively, the dynamics are different but the genetic signature is too weak relative to the effect of range shifts for detection with the methods used. In either case, genetic diversity was clearly not predicted by a population‟s location with respect to the geographic periphery, when the entire periphery was included. We suggest that this may also be true in trees with life history and reproductive biology similar to that of butternut. We anticipate that our results, paired with other critiques (Eckert et al., 2008; Lawton, 1993), will strengthen the experimental design and analysis of future investigations of this process. We emphasize that a decline in diversity from the range center to northern edge does not provide support of either the core-periphery or the range shift process, but rather both, and a specifically designed sampling scheme coupled with a rigorous model-comparison approach is necessary for a better understanding of the genetic consequences of ecological marginality of the range edge.

58

4.4.4 Range shifts We observed a latitudinal gradient in diversity from the range core northward. Using model comparison, we show that this pattern is distinct from core-periphery influences or contemporary population decline, providing convincing support for the effect of range shifts and adding to the already strong body of evidence in support of the range shift hypothesis in both plant and animal taxa (Hewitt, 2000; Jaramillo-Correa et al., 2009). We also detected more population differentiation in northern populations. Long distance dispersal or recolonization from disjunct and isolated refugia, which seem necessary to explain rates of post-glacial colonization in forest trees (Austerlitz, GarnierGere, 2003; Clark, 1998), should lead to severe founder effects. In some investigations, however, leading populations show less differentiation, perhaps due to complex colonization paths or a mixing of divergent lineages from isolated refugia (Gamache et al., 2003; Heuertz et al., 2004; Petit et al., 2003). Our results reinforce the idea that a universal pattern for differentiation is unlikely. Rather, differentiation will depend on the frequency of long distance dispersal events, location of refugia, and stochastic biotic and abiotic events, including local landscape heterogeneity (Excoffier et al., 2009; Magni et al., 2005; Soltis et al., 2006). Our results for pairwise FST, number of alleles, and heterozygosity are similar to values observed in other forest trees, including Fraxinus, Quercus, and Populus (Heuertz et al., 2004; Muir et al., 2004; Smulders et al., 2008). Therefore, we suggest that our finding regarding the influence of range shifts is not limited to butternut. As nearly all eastern North American forest species migrated north during the last glacial retreat, range

59

shifts may have played a key and enduring role in the distribution of diversity within these forests. Several butternut populations deviate from the overall patterns we observe. The Arkansas population (population 1) was a moderately divergent southern population, a finding that is consistent with its disjunction from the main range. The question of how long it has been disjunct, and the possible role of this area as a glacial refuge, deserves future investigation. The Tennessee population (population 7) was the only relatively distinct southern population in clustering analyses (Figure 3). This observation may be due to the unusually high census at this location, which we will examine in another manuscript. Lastly, some populations north of the ice margin showed moderate diversity and differentiation (Table 2, Figures 2-3). We reemphasize the probable role of stochastic processes during range colonization, as well as possible later gene flow that erased signals of founder effects among some populations during or after range shift.

4.4.5 Isolation-by-distance We found weak isolation-by-distance. Several other studies in wind pollinated forest trees have also shown weak isolation-by-distance, even when populations are significantly differentiated (Craft, Ashley, 2007; Heuertz et al., 2004; Smulders et al., 2008). This may be attributed to the fact that isolation-by-distance is predicted to be weak at large spatial scales (Bradbury, Bentzen, 2007), or when populations have not reached migration-drift equilibrium (Slatkin, 1993), such as after range shifts.

60

4.4.6 Conservation context Over 90% of butternut trees are infected with canker in some regions (Nielsen et al., 2003; Ostry, Woeste, 2004; Schmalz, Bergdahl, 2006). Natural regeneration is rare, as seedlings are particularly susceptible to canker (Orchard, 1984). Our work indicates that southern populations, despite recent large-scale population declines, have retained high heterozygosity and a sizeable fraction of the total diversity in the species, and are therefore the most suitable targets for ex-situ conservation collections (Frankham, 1995; Ledig, 1988; Stockwell et al., 2003). However, if local extinctions continue, whole cohorts of trees will perish without producing descendents, an irrecoverable loss of genetic diversity will ultimately occur (Segelbacher et al., 2003). Management of habitat to facilitate regeneration and transplanting of disease tolerant trees back to forested locations should be both feasible and effective at maintaining moderate population sizes in-situ. On the other hand, northern populations have the most severely reduced gene pool. These populations, although in a position to colonize northward in response to climate warming, may have the least capacity to adapt to new environments and future climatic conditions. Unfortunately, due to moderate population differentiation, transplanting across large distances to augment northern diversity may be of concern due to potential genetic and phenological incompatibilities with individuals from other populations, or poor adaptation to the local environment (Savolainen et al., 2007). However, debate continues on the importance of diversity versus differentiation in prioritizing populations (Petit et al., 1998), the adaptive significance of differentiation for neutral genetic markers (Fraser, Bernatchez, 2001), and the potential conflict between

61

maintaining diversity and maximizing fitness (Stockwell et al., 2003). Given the importance of adaptationn and diversity to conservation genetics programs, common garden experiments and functional genetics studies are essential for the long-term success of restoration programs. We do suggest that, particularly for long-lived organisms with overlapping generations and long distance dispersal potential, the minimum viable conservation size (500-1000 individuals) recommended by some may be upwardly biased, and that even small populations (<50) retain conservation value (Allendorf, Ryman, 2002)..

4.4.7 Conclusions Our investigation provides fresh perspective on the genetic consequences of three major biogeographical processes typically investigated in isolation. Based on our finding that contemporary decline and ecological marginality of the periphery have less genetic impact in butternut than large-scale range shifts, we suggest that the life histories and dispersal mechanisms of some wind-pollinated forest trees enables them to withstand all but the most severe and long-lasting bottlenecks. Future investigations will need to focus on the duration of the bottleneck and the multiple mechanisms by which both duration and severity cause loss of genetic diversity. Biotic and abiotic stresses (e.g., hemlock woolly adelgid, emerald ash borer, pine bark beetle) have caused large scale population declines in North American forests (Van Mantgem et al., 2009). Our results, combined with other recent investigations, suggest that the genetic impact of contemporary population losses should be evaluated in the context of Holocene epoch range shifts. Carefully designed sampling schemes and comparative statistical analysis, including recently developed Bayesian methods, will contribute to this research direction. Our 62

results also apply to predictions of the impact of contemporary climate change on butternuts, other forest trees, and organisms with similar life history and gene flow characteristics. Due to reduced diversity, northern populations may be the least capable of adapting to future changes and new environments. Even if tree populations are able to track climate change by shifting ranges, a debatable outcome (Pearson, 2006), we show that the genetic consequences of range shifts in forest trees may endure hundreds of generations.

63

CHAPTER 5: CONTRASTING SPATIAL GENETIC STRUCTURE IN TWO HABITAT TYPES FOR BUTTERNUT: RIPARIAN AND UPLAND POPULATION DYNAMICS

5.1 Introduction Spatial genetic structure (SGS), the correlation between spatial location and genetic relatedness of individuals, has important ecological and evolutionary impacts on local adaptation, inbreeding, and genetic drift, as well as population persistence and community stability (Endler, 1977; Frankham, 2005; Rogers et al., 1999; Saccheri et al., 1998). Ecological mechanisms at the species (e.g., mating system) and population (e.g., density of individuals) levels are important determinants of the strength and pattern of SGS (Dick et al., 2008; Loveless, Hamrick, 1984; Petit, Hampe, 2006; Vekemans, Hardy, 2004), and quantifying their contribution is an emerging area of population genetics (Guillot et al., 2009; Hu et al., 2010). Such knowledge can guide reintroduction, relocation, and habitat restoration efforts (Fischer, Matthies, 1997; Ledig, 1988; Vellend, Orrock., 2009), as well as improve the predictive value of models of population response to future perturbations (e.g., fragmentation, selective harvest, habitat degradation) at a range of spatial and temporal scales (Marquardt, Epperson, 2004; Oddou-Muratorio et al., 2004; Rajora, Pluhar, 2003). Our goal in this paper is to contrast both the strength and pattern of SGS in two contrasting habitat types, riparian and upland, in a forest tree, Juglans cinerea. 64

Terrestrial plants typically exhibit some level of SGS, as both pollen and seed dispersal tend to have leptokurtotic distributions skewed towards small distances (Levin, 1981; Levin, 1988; Loveless, Hamrick, 1984; Petit, Hampe, 2006). A species‟ seed dispersal mechanism, seed size, mating system, and pollination vector contribute towards the pattern and strength of SGS (Cottrell et al., 2003; Dick et al., 2008; Heuertz et al., 2003). For example, species with wind dispersed seeds tend to show weaker SGS than those with gravity dispersed seeds (Hardy et al., 2006; Vekemans, Hardy, 2004). However, while reproductive biology explains much variation in SGS between species, local ecological factors may also significantly influence SGS by changing the density and distribution of individuals, availability of mates, magnitude of reproductive output, and location of seedling establishment (Hardy et al., 2006; Loveless, Hamrick, 1984; Trapnell et al., 2008). Due to habitat specific contributions to SGS, the variation in SGS within a species may exceed the variation between some species (Born et al., 2008a). However, identification and quantification of the role of local conditions in shaping SGS is a relatively recent trend in spatial genetics (Born et al., 2008a), and has been somewhat inconclusive. For example, colonization after disturbance regimes may increase or decrease the spatial genetic structure, depending on the colonizer pool (Oddou-Muratorio et al., 2004; Pardini, Hamrick, 2008), and the maturation of a population (Jones et al., 2006; Jones, Hubbell, 2006). Further, low population density has been implicated in high as well as low SGS (Bizoux et al., 2009; Kyndt et al., 2009). Conflicting results may be due in part to insufficient clarity with regards to the proximal mechanism underlying the observed SGS, as well as lack of specific, falsifiable alternative hypotheses (Born et al., 2008b). While the exploratory and descriptive nature

65

of previous studies is useful for describing both point estimates and a rough measure of variation, it has often led to multiple and largely post-hoc explanations for the level of SGS that is observed (Table 1). A hypothesis-testing framework could resolve much of this conflict (Hampe et al., 2010), especially if this framework is mechanistic, connecting local ecological conditions (e.g., habitat), population biology response (e.g., density of individuals), and population genetics (e.g.,subsequent changes in SGS). A second source of disagreement may be lack of statistical power, as SGS studies typically involve a small number of populations (typically one to four). Statistical support for the contribution of particular ecological mechanisms will require an experimental approach with enough populations to provide statistical power, similar to the approach used in interspecific comparative studies (Vekemans, Hardy, 2004). Our aim is to rigorously explore the connection between habitat, recruitment, and SGS in Juglans cinerea (butternut), an Eastern North American outcrossing, windpollinated canopy tree, whose spatial genetic structure is yet unknown. Butternut is a threatened species, primarily due to the introduced fungal pathogen Sirococcus clavigignenti-juglandacearum, the causal agent of butternut canker, which has caused 7090% mortality in many locations (Nair et al., 1979; Orchard, 1984; Schlarbaum et al., 1997a). Other causes of decline include deer browse, poor quality of remaining trees, logging, and habitat loss (Nielsen et al., 2003; Ostry et al., 1994). An insufficient understanding of the genetic and population dynamics of this species has hindered conservation management (Ostry et al., 2002).

66

TABLE 5.1 SOME RECENT STUDIES OF HABITAT INFLUENCE ON SGS

Study/ Species

67

Gonzales et al 2010 Enterolobium cyclocarpum Hu et al 2010 Fraxinus mandshurica Bizoux et al 2009 Milicia excelsa Valbuena-Carabaña et al 2007 Quercus petraea

Habitat influence on SGS Human disturbance (cattle pastures) Mountain vs. riparian location Forest vs. open field/ savannah Spatial location of recruitment sites

Effect observed SGS is highest when recruits cluster around a parent individual SGS is only observed in mountain sites, where seed dispersal remains local SGS is only observed in dense forests SGS was lowest when suitable recruitment sites were few

Pardini & Hamrick 2008 Albizia julibrissin

Colonization history

SGS was not observed when populations had few founders

Born et al 2008 Aucoumea klaineana

Fragmentation due to agriculture and mining

Similar SGS was observed in both continuous and fragmented forests

Kyndt et al 2009 Adansonia digitata

Density in an agroforestry system

Higher SGS was observed in lower density populations

Vaughan et al 2007 Prunus avium

Selective harvest management

Managed population had reduced SGS

Trapnell et al 2008 Ceratiola ercoides

Frequency of dispersal by birds vs. gravity

The site with more bird dispersal has higher SGS

A sample of recent studies examining the effect of habitat on SGS in trees and woody shrubs.

Mechanism Pasture size and age determines the spatial location of suitable sites for recruitment Extensive seed dispersal may occur via water Distance of seed and pollen movement is greater in open locations Distance of seed movement is high due to few suitable sites for seedlings establishment Number of founders influences the relatedness of individuals within a site; few founders results in high relatedness at all distances Fragmentation reduces density, which should increase SGS, but this is compensated by increased pollen movement Reduced density lowers seed shadow overlap, increasing the likelihood that nearest neighbors are related Selective removal of adults and clearing of areas of dense regeneration removes sibling and parent-offspring clusters Bird dispersal contributes to a greater degree of co-dispersal of siblings into a suitable site

In this study, we sample 21 populations and 853 individuals, and focus on the effect of two habitat types, riparian and upland (Table 2, Figure 1). Riparian (river) habitats are located at the boundary of terrestrial and aquatic ecosystems, and typically support a unique assemblage of species, due to resource supply and disturbance regime. While the transition between these habitats is complex and dynamic (Naiman and, Decamps, 1997; Nilsson, Svedmark, 2002; Tabacchi et al., 1998), and a precise definition of each is difficult, we focus on butternut populations within the floodplain of rivers or streams, compared to butternut populations that occur outside the flood fringe. We propose a simple but biologically grounded model of population dynamics that is based on patterns of disturbance and recruitment, central forest processes that determine longterm community structure and stability (Franklin et al., 2002; Sousa, 1984). Specifically, we propose that the spatial distribution and frequency of recruitment events will differ markedly between habitats, leading to substantially different SGS. First, riparian sites experience frequent and local disturbance from flooding (Auble, 1995; Naiman et al., 1998), while upland sites do not. Both sites will experience disturbance from tree fall and storm damage, but in upland sites this may be more episodic and spatially stochastic (Canham et al., 1990; Greenberg, McNab, 1998; Whitmore, 1989). As an early successional, light demanding species, butternut requires canopy openings for seedling establishment. Therefore we expect that seedling establishment in riparian habitats will occur more frequently, and within the local area, while establishment in upland sites will be infrequent, and suitable sites for establishment (e.g., gaps) will likely be distant from the parent tree. The greater abundance of resources (light and water) in riparian areas may also increase the likelihood of local seedling survival (Klapproth, Johnson, 2009),

68

strengthening SGS further in this habitat. Evidence for this model includes a younger skewed age structure and stronger SGS in riparian areas than in upland areas. In summary, we hypothesize that riparian sites will exhibit stronger SGS than upland sites, due to frequent, local recruitment. We compare this scenario to four alternative hypotheses: (i) no SGS in any population, (ii) SGS exists but is similar at all sites, (iii) SGS varies between sites but is not correlated with habitat, and (iv) SGS is correlated with habitat but is lower in riparian habitats, as water may facilitate long distance seed dispersal (Hu et al., 2010). We use spatial and genetic data, as well size (trunk diameter at breast height) as a proxy for relative age, to test these predictions. Our objectives are (a) to quantify SGS in 21 populations, (b) test the hypothesis that populations in riparian areas will show greater SGS than upland sites and a population age structure skewed towards younger trees, and (c) discuss conservation implications for this rare tree.

69

TABLE 5.2. POPULATIONS USED IN THIS STUDY

Population

70

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Saint Francis National Forest AR Ozark MO Mark Twain MO Butternut Valley TN Mammoth Cave National Park KY Private property, WVA Allegheny Dew Drop PA Allegheny Riverway PA Allegheny Golden Rod PA Barre Town Forest VT Jericho Research Forest VT Putney School VT ChequamegonNicolet National Forest WI Private property, Waupaca WI Private property, Whitewater WI Private property, Renfrew ON Private property, Peterborough ON NottawasgaLake ON GilbertIsland NB KeswickRidge NB Blackville NB

Alleles 10.5 13.4 9.8 14.8 13.8 10.0 9.9 9.8 10.7 8.2 8.8 9.8 7.6 8.3 8.3 8.2 9.9 9.7 8.3 8.5 8.0

Allelic richness 7.4 8.1 7.3 7.7 8.4 8.1 8.3 8.0 7.1 6.8 7.3 7.8 6.2 6.7 6.0 6.7 7.7 7.9 6.2 6.7 6.0

Fis 0.129* 0.037 0.103* 0.047* 0.025 0.018 0.098* 0.049 0.034 -0.085 0.019 0.035 -0.012 -0.028 0.026 0.015 -0.003 0.048 0.026 0.026 0.011

He 0.767 0.818 0.816 0.812 0.830 0.834 0.853 0.813 0.775 0.777 0.800 0.797 0.773 0.763 0.736 0.767 0.819 0.822 0.766 0.796 0.748

Max D for regression 1536 1990 1950 2000 1783 940 1973 938 869 191 596 1458 901 428 424 1718 906 1855 1147 612 421

Minimum and maximum distance to water 50 to 500 5 to 50 5 to 50 5 to 50 5 to 100 5 to 200 50 to 200 5 to 100 5 to 100 100 to 300 200 to 1000 50 to 300 900 to 1500 800 to 1600 1500 to 2500 500 to 1500 200 to 800 50 to 1000 100 to 300 5 to 50 1 to 50

Sample Size 35 68 31 161 64 22 19 23 45 21 21 23 28 20 40 26 29 24 39 33 37

Habitat Upland Riparian Riparian Riparian Riparian Other Upland Riparian Riparian Upland Upland Upland Upland Upland Upland Upland Upland Other Other Riparian Riparian

Populations used in the study, numbered as in Figure 1. *, Fis is significant. Max D for regression is the maximum pair-wise distance between individuals in that population used in the regression. Minimum and maximum distance to water is the linear distance to the nearest river or stream. “Other” refers to populations that we were unable to classify as riparian or upland.

Figure 5.1: Sampling sites for this study. Blue circles are „riparian,‟ brown circles are „upland,‟ and yellow circles are „other‟.

5.2 Materials and methods

5.2.1 Species: Butternut (also called white walnut, oilnut, or lemonnut) is a monecious, diploid tree of the walnut genus native to eastern North American hardwood forests. Male and female flowers are temporally separated (heterodichogamous) on individual trees, and trees are assumed to be highly outcrossing. Trees usually begin producing fruit at approximately 15 years old and rarely live longer than 70 years (Davis, 1966; McDaniel, 1979; Rink, 1990). Populations are typically found along stream banks, floodplains,

71

slopes and other sites with well-drained, fertile soil, but are also found in upland and even mountainous sites, especially those with limestone-based soils. Populations are small, (tens to hundreds of trees scattered over several hundred to thousand meters) and relatively isolated (separated by ten or more km). While butternut has likely never been a common species, its extremely sparse distribution is partly a result of recent decline due to an epidemic disease known as butternut canker. Due to its short life span, fast growth, and intolerance of light competition, butternut is considered an early successional species (Nesom, 2000; Ostry, Pijut, 2000; Reed, Davidson, 1958). The large ovoid fruit (5-10 cm in length and 3-6 cm in width) consists of a seed, also called a butternut, encased in a thick husk. Depending on site conditions and tree health, a tree may produce between zero and several hundred seeds per year, and a mast year (a large „crop‟ of seed) typically occurs every two to three years (Davis, 1966; Ostry, Pijut, 2000; Reed, Davidson, 1958). Flowering and pollination occurs concurrently with bud break in April, and nuts develop throughout summer, although developing nuts will abort under drought or stress (McDaniel, 1956). After seedfall (October or November), butternuts are dispersed by squirrels, large birds, and gravity. Water is another possible, although undocumented, dispersal mechanism, as seeds float and butternut trees often occur along the banks of streams and rivers. The extent of seed dispersal by any of these mechanisms in butternut is not known. Observations in black walnut (J. nigra) and Japanese walnut (J. ailantifolia) suggest that transport by squirrels is primarily short distance, approximately 15-20 m (maximum ~ 100m), and transport by crows is longer distance, up to 1-2 km. Movement via birds is unlikely to play a significant role in gene dispersal, as crows immediately consume the nut (Cristol 2005,

72

Tamura et al 2008, Stapanian and Smith 1984). Seeds require a period of stratification (chilling) before germination in spring. Seed banking is unlikely, as seeds that do not germinate typically die. Stump sprouting is assumed to be rare (Fleguel, 1996b; Schultz, 2003).

5.2.2 Collections: We sampled butternuts from 21 forested locations throughout the Eastern United States and Southern Canada (Table 2), and recorded the GPS location, diameter at breast height (DBH), and height of each tree. As butternut populations are typically small (10 to 100 trees), all or most of the known butternuts were sampled in most sites. Samples of leaves or twigs were transported to the University of Notre Dame, where they were kept at -20 C until DNA extraction, as previously described (Hoban et al., 2008; Hoban, 2009). Sites were classified as riparian or upland (Table 2), based on proximity to a stream or river. In sites classified as „riparian‟, trees were typically within 1 to 50 meters of a stream or creek (within the floodway), while trees in sites classified as „upland‟ were typically at least 100 to 1000 meters from a stream or creek (in or outside of the flood fringe). Riparian sites were typically at a similar elevation as the river or stream, while upland sites were at higher elevation (Figure 2). We were unable to classify three populations (referred to as “Other” in Table 2), so we do not include them in our habitat comparison, but we do calculate all spatial genetic statistics for them. One “other” population was distributed along a gradient from upland to riparian, and the other two were populations in agricultural environments (pasture).

73

74 Figure 5.2: Contour maps and photographs from typical riparian (a and b), and upland (c and d) sites. Note the riparian site is clearly within the floodplain (a) and the upland site is far from waterways (b), and the dense forest cover in (d).

5.2.3 Genotyping: We genotyped individuals for 12 microsatellite markers, 11 from (Hoban et al., 2008), one from (Robichaud et al., 2006a). PCR and genotyping protocols are previously reported (Hoban et al., 2008). Before proceeding, we first checked our samples for hybrids, as hybridization between butternut and an introduced exotic species, Japanese walnut (J ailantifolia) is common in some locations (Anagnostakis, 2009; Hoban, 2009; Ostry, Woeste, 2004), and hybridization can alter SGS (Valbuena-Carabana et al., 2007). Using NEWHYBRIDS, and previous methods (Hoban, 2009; McCleary et al., 2009), any individual with < 0.95 probability of being a pure butternut and/or any individual with a J. ailantifolia chloroplast was removed before further analysis. We also checked our samples for clonal propagates, as vegetative reproduction by stump sprouting will strongly increase SGS as adjacent individuals will have identical genotypes (Vaughan et al., 2007). Using CERVUS v 3.0 (Kalinowski et al., 2007), we screened for identical genotypes, allowing up to two mismatching loci for a match. After removal of clones and hybrids, 809 individuals were retained for genetic and demographic analysis. Markers were tested for linkage (gametic) disequilibrium at each population and locus, using GENEPOP ON THE WEB (Raymond, Rousset, 1995a) Option 2, which tests the null hypothesis that alleles at one locus occur independently from alleles at another locus. We used the log likelihood ratio test, and default values for the Markov chain.

5.2.4 Population summary statistics: We used Arlequin v 3.11 (Excoffier et al., 2005) to calculate observed and expected heterozygosities (HO and HE), and FSTAT v2.9.3.2 (Goudet, 1995) to calculate number of alleles (A) and allelic richness (AR, or number of alleles relative to the number 75

of complete genotypes in the smallest population). We tested for departures from Hardy Weinberg expectations using Wright‟s inbreeding coefficient (FIS) with FSTAT (significance tested using 5,000 permutations).

5.2.5 Spatial genetic structure Within each site we calculated a measure of pairwise genetic relatedness between all individuals, using Nason‟s kinship coefficient (Fij) (Loiselle et al., 1995). A variety of relatedness estimators have been developed (Casteele et al., 2001; Lynch, Ritland, 1999; Wang, 2002). We chose Fij because it demonstrates low variance, efficiently combines information across loci, and exhibits good performance with microsatellite data (Hardy, Vekemans, 2009; Vekemans, Hardy, 2004). The expected Fij value for full siblings is 0.25 and for half siblings is 0.125 (Hardy, Vekemans, 2009). Calculations of the kinship coefficient, and all following analyses were performed with SPAGEDI v 1.3 (Hardy, Vekemans, 2002). We then calculated pairwise distance between individuals, and performed linear regression of Fij on this geographic distance, truncated at 2,000 meters. SGS is predicted to be most apparent at biologically limited distances, i.e. < 20σ, where σ is the average distance between a parent and offspring (Vekemans, Hardy, 2004). Recent investigations of temperate trees, particularly Quercus, suggest that 20σ is between several hundred and several thousand meters (Dutech et al., 2005; Heuertz et al., 2003; Soto et al., 2007; Vaughan et al., 2007). In the absence of data for Juglans, we chose 2,000 m as an upper boundary. Using the same maximum distance across sites also ensures comparability of spatial genetic structure across sites (Vekemans, Hardy, 2004). Twenty-one trees did not have GPS points recorded, so they were not included in spatial analysis. 76

We then constructed spatial autocorrelograms, which is a graphical display of the average Fij observed within each of a set of distance categories, or bins (e.g., the average Fij observed within 0-50 meters) (Smouse, Peakall, 1999; Vekemans, Hardy, 2004). This technique is used to give a visual picture of the change in SGS as distance increases, as well as to identify the distance class at which average relatedness first becomes not significantly different from zero (Doligez et al., 1998; Smouse, Peakall, 1999). Distance intervals were determined through several trials, to achieve the most even distribution of individuals across classes, across all 21 sites, and to maintain an ideal number of 30-100 individuals per class. This balance was best achieved with distance classes whose upper bounds were 50, 100, 175, 250, 400, 600, 1000, and >1000 m. We generated 95% confidence intervals around the null hypothesis of no relatedness within each bin, with 20,000 permutations of genetic data. For one geographically small population, population 10, we used shorter classes. In this population the maximum inter-individual distance was 191 m, so we used distance classes of 25, 50, 75, 100, 150, and 200 m. For each correlogram (Figure 3 shows two examples), we noted whether the average kinship coefficient (Fij) value deviates from the 95% confidence interval for the first distance class (called F1), which represents significant “positive SGS” (Smouse, Peakal, 1999; Vekeman, Hardy, 2004). We also noted the distance class at which average kinship (Fij) first drops into the 95% confidence interval of no relatedness (not significantly different from zero), which we call “maximum distance of positive relatedness.” This distance does not represent neighborhood size (Nb) or any other biologically relevant parameter (Vekemans, Hardy, 2004), but simply represents the maximum distance at which a larger than average relatedness occurs. We note that

77

occasionally after dropping into the 95% confidence interval, the correlogram would become significant again, and then drop down. This common observation is usually interpreted as stochastic noise, rather than an extension of SGS to this distance (Hampe et al., 2010).

Figure 5.3. Spatial autocorrelograms, where the black line shows Fij,WITHIN and the green line shows Fij,OVERALL. The correlogram show (a) signal of positive SGS and (b) lack of signal of positive SGS, but a signal of colonization We then calculated the Sp statistic, a measure of the overall „strength‟ of SGS, which is equal to –b/(1-F1), where F1 is the Fij in the first distance class (this use of F1 is not to be confused with first filial generation, F1, for the offspring of a genetic cross), and b is the slope of the linear regression of Fij on the natural logarithm of distance. To examine the effect of our choice of eight distance classes, we reanalyzed four populations (populations 6, 9, 20, and 21) with sixteen distance classes, and one population with sixty

78

distance classes (population 4), and use paired t-tests to compare Sp values with the different sets. We then examined the age structure of each population, using diameter of the trunk at breast height (DBH, measured approximately 1.5 meters above the ground) as a proxy for age, except at Mammoth Cave, where we did not collect DBH data. In forest trees with sexual reproduction and continuous growth, DBH is often used as a proxy for age, particularly to distinguish between saplings, juveniles, and adults (Jones, Hubbell, 2006; Oddou-Muratorio et al., 2004; Vaughan et al., 2007). While butternut‟s lateral growth is subject to the influence of weather and site conditions, tree ring width does show moderate autocorrelation, and a standard deviation less than the mean, suggesting relatively constant growth (Clark et al., 2008). First, we calculated the mean DBH of trees in each site. We then calculated the proportion of trees in the following size classes: 0-12.7 cm, 12.7 cm-25.4 cm, and 0-25.4 cm (which we call “saplings”, “young”, and “saplings + young”). In black walnut (Juglans nigra), which grows to similar sizes, these categories are called “sapling” and “pole timber” (Bruckerhoff, 2005), and are considered relatively immature. These categories do not represent specific age groups (i.e. 0-10 years). Lastly, we specifically examined the difference in size between pairs of trees with kinship higher than 0.125, the amount expected between half siblings. For these highly related pairs of individuals, we calculated the difference between their DBH (dDBH) to determine whether these highly related individuals are in different size classes (possible parent-offspring) or within the same size class (possible sibling). Estimates of kinship should be useful for describing aggregated information such as this (Queller, Goodnight,

79

1989). Any pairs with dDBH less than 5.1 cm, we consider likely to be of the same size cohort, and thus probable siblings. We repeat this with a slightly higher cutoff, 10.2 cm. We use this as an approximation of the degree to which siblings contribute to SGS (compared to parent-offspring pairs).

5.2.6 Comparison between habitat types We used t-tests to compare the two habitat types, riparian and upland, for the following variables: Sp, F1, mean DBH, maximum distance of positive kinship, proportion saplings (props), proportion young (propy), and the combined proportion young + saplings (props+y). In addition, we separately consider the two values used to derive Sp, the kinship in the first distance class (F1) and the overall slope of the regression (b). This is of interest, as Sp is a relatively recent statistical tool in spatial genetics, and its properties have not been fully explored. Because Sp = -b/(1-F1), and F1 is usually relatively small in outcrossing, low inbreeding populations (Table 3), we propose that Sp is highly biased towards measuring b. To see if Sp accurately captures information embodied in F1 and b, we perform linear regression of Sp on F1 and b. To test for contribution of other population specific factors to SGS, such as the amount of genetic diversity in the population, we performed linear regressions of the Sp statistic and propy on each of the following predictive variables: sample size (N), observed and expected heterozygosity (HE, HO), number of alleles (A), and maximum distance used in the regression analysis (not to be confused with maximum distance of positive kinship). We also perform linear regression of the Sp statistic on the natural log

80

of the distance from the nearest body of water (stream or river) to the closest tree, and to the farthest tree, ln(dnear) and ln(dfar).

5.2.7 Changing the reference population for SGS Relatedness estimators are calculated in the context of a reference population (Lynch, Walsh, 1998). The kinship estimator, Fij, is typically calculated with reference to other individuals in the same population, with the population average assumed to be near zero. However, in all finite populations, average „true‟ relatedness (i.e. if a pedigree was known) will be greater than zero, and so the estimator will be downwardly biased. In small or newly established populations we expect to find possibly related founders and their descendents, e.g., a substantial portion of close relatives. The true relatedness‟ in such a population will be quite high, but standard techniques for calculating kinship will show that the average is zero, an extreme downward bias.

81

TABLE 5.3 COMPARISON OF SP AND F1 STATISTICS OBSERVED IN BUTTERNUT TO OTHER PRIMARILY OUTCROSSING, WIND POLLINATED TREES

Species Fraxinus excelsior

Q. robur (seedling)

Reference Vekemans & Hardy 2004 Vekemans & Hardy 2004 Bizoux et al 2009 Vekemans & Hardy 2004 Dutech et al 2005 Hampe et al 2010 Valbuena-Carabaña et al 2007 Vekemans & Hardy 2004 Vekemans & Hardy 2004 Vekemans & Hardy 2004 Valbuena-Carabaña et al 2007 Jump & Peñuelas 2007 Hampe et al 2010

Juglans cinerea

This study

0.064 (mean) 0.007-0.126 (range)

0.0199 (mean) -0.0002-0.0421 (range)

Q. suber

Soto et al 2007

0.057a

0.0244d

Quercus robur Milicea excels (4 populations) Larix laricina Q. lobata Q. robur (adult) Q. petraea (2 populations) Q. petraea Acer saccharum Pinus strobus Q. pyrenaica (3 populations) Fagus sylvatica

F1

Sp

0.029

0.0020

0.011

0.0030

0.021 (mean) 0.013-0.035 (range)

0.0031 (mean) 0.0002-0.0063 (range)

0.020

0.0045

0.042

0.0046

0.025a

0.005

0.040 (mean) 0.03-0.05a (range)

0.007 (mean) 0.003-0.015 (range)b

0.031

0.0083

0.018

0.0102

0.032

0.0108

0.063 (mean) 0.05-0.075a (range)

0.013 (mean) 0.013-0.013 (range)b

0.040a

0.0141c

0.075a

0.015

Low SGS

High SGS

Note the high Sp and F1 values observed in this study relative to other taxa. a- exact values were not given in the text; values are estimates extracted by visually inspecting correlograms b- Sp values from visually inspecting Fig 5 c- the authors estimated Sp using two distance cutoffs for the regression; this value is from the shorter distance, 200 m d- Sp calculated from the slope (given by authors) and F1 (visual inspection of correlogram)

82

Pardini and Hamrick (2008) recently proposed using a truly unrelated reference population in order to remove this bias. We therefore recalculate pairwise Fij using a reference sample composed of all 21 populations, in which any two randomly chosen individuals are likely to be truly unrelated. We then examine the difference between Fij values calculated using the within population reference sample and Fij values calculated using the overall reference sample (Fij,WITHIN and Fij,OVERALL). The correlogram based on Fij,OVERALL may be different than the correlogram based on Fij,WITHIN. In particular, Pardini and Hamrick note that a correlogram based on Fij,OVERALL may be elevated across all distance categories. This suggests that most individuals are closely related, without regard to spatial location, which in turn suggests a recent colonization. We therefore record whether Fij,OVERALL is elevated across all distances, suggesting a new colonization, or returns to zero relatedness at higher distance classes, a signature of positive SGS.

5.3 Results

5.3.1 Summary statistics Thirty-four clones were removed, so all analyses are based on the remaining 809 trees. LD was observed at 48 of the possible 1362 population-locus-locus combinations (3.5%). Null alleles were observed at 18 of the possible 348 loci-population combinations (5.1%). Only four populations (two riparian and two upland) showed significant FIS values (Table 2), suggesting minimal departures from HWE. These minimal departures from equilibrium are expected in populations in which nonrandom mating or closely related individuals occur.

83

A total of 270 alleles were observed. Average number of alleles per locus, per population ranged from 7.6 to 14.8 (mean=9.9), and no population had fewer than 3 alleles at a given locus. Heterozygosity ranged from 0.67 to 0.84 (mean=0.77), typical of wind pollinated forest trees.

5.3.2 Relatedness and spatial genetic structure Mean pairwise kinship (Fij) within a population ranged from -0.007 to 0.026 (mean = 0.002). The slope of the regression of pairwise relatedness on the natural log of pairwise distance, b, ranged from -0.040 to 0.0003. All but two of the regressions were significant (one tailed test, α=0.05). Both nonsignificant slopes (populations 15 and 16) were in upland sites (Table 4). Mean pairwise relatedness within the first distance class (F1), 50 m, ranged from 0.126 to 0.007 (mean = 0.064). The Sp statistic ranged from 0.042 to -0.0002 (mean=0.020) (Table 4). The Sp statistic did not significantly change for the populations in which we experimented with a greater number of distance classes (paired t-test, two-tailed, p = 0.623). Sp shows an extremely strong correlation with b (R2 = 0.983, p<<0.001), but only a weak correlation with F1 (R2 = 0.121, p=0.046). Visual inspection of the correlograms revealed that five populations showed no significant pattern of SGS, i.e. that average pairwise relatedness was never significantly positive or negative for any distance class (Table 4). All five of these populations were in upland sites. Riparian sites were more likely to show positive SGS (8 out of 8) than upland sites were (5 out of 10) (Chi-square test, p=0.034). For those populations showing SGS,

84

Figure 5.4. Correlograms for a subset of populations in the study: five riparian (a) and five upland (b) sites. Black line shows Fij using the within population reference. Green line shows Fij using the overall reference. No SGS means that the black line never crosses the 95% confidence interval (grey dashed line). Colonizer means that the green line is always elevated above the 95% confidence interval (high relatedness at all distances). 85

the distance at which relatedness stayed positive ranged from 50 m (first distance class) to 250 m (fourth distance class), with a mean of 106 m. Mean DBH within populations ranged from 22.1 cm to 62.5 cm. MinimumDBH – maximumDBH ranged from 32.3 to 100.6 cm (mean = 59.4 cm), showing that all populations were a mix of size classes, i.e. no population was completely even aged. Props ranged from 0 to 0.277 (mean = 0.085), propy ranged from 0 to 0.529 (mean = 0.251), and props+y ranged from 0 to 0.696 (mean = 0.336) (Table 4). The proportion of probable siblings among the high relatedness comparisons ranged between 0.095 and 0.799 (mean = 0.263) for the 5.1 cm cutoff, and between 0.233 and 0.865 (mean = 0.462) for the 10.2 cm cutoff.

5.3.3 Comparison between habitats Riparian habitats were significantly different from upland habitats for both SGS and age (size) structure. Riparian habitats had significantly higher Sp statistics (Spriparian = 0.024, Spupland = 0.015, p = 0.027) and increased distance at which relatedness was positive (driparian = 131.3 m, dupland = 80.0 m, p = 0.037) (Table 5). This difference was observed even though the five populations without SGS were removed for this test (because they were never positive). If the test was repeated with these samples, giving them zero values for distance, the test statistic becomes even stronger (driparian = 131.3 m, dupland = 40.0 m, p = 0.003). Further, riparian habitats had a significantly steeper slope of the regression of genetic distance on geographic distance, b (briparian = -0.022, bupland = 0.015, p = 0.037) and nearly significantly higher relatedness in the first distance class (F1,riparian = 0.080, F1,upland = 0.050, p = 0.053) (Table 5).

86

TABLE 5.4 SGS STATISTICS AND POPULATION AGE STRUCTURE OBSERVED

Habitat

87

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

U R R R R O U R R U U U U U U U U O O R R

Pos Colonizer SGS Y N Y N Y N Y N Y N Y N N N Y N Y N N N Y N Y N Y Y N Y N Y N Y Y N Y N Y N Y N Y N

U/R/O- Upland, River, Other

F1

b

Sp

0.1109 0.1257 0.1071 0.0950 0.1089 0.0642 0.0353 0.0745 0.0285 0.0229 0.0690 0.0660 0.0580 0.0076 0.0065 0.0267 0.1004 0.0936 0.0455 0.0543 0.0422

-0.0198 -0.0197 -0.0125 -0.0271 -0.0215 -0.0319 -0.0095 -0.0227 -0.0132 -0.0149 -0.0230 -0.0184 -0.0251 -0.0148 -0.0086 0.0003 -0.0147 -0.0137 -0.0195 -0.0216 -0.0403

0.022 0.023 0.014 0.030 0.024 0.034 0.010 0.025 0.014 0.015 0.025 0.020 0.027 0.015 0.004 0.000 0.016 0.015 0.020 0.023 0.042

mean

mean

Fij,OVERALL 0.059 0.012 0.035 0.032 0.032 0.031 0.027 0.025 0.052 0.039 0.041 0.043 0.091 0.087 0.083 0.103 0.027 0.016 0.078 0.075 0.129

Fij,WITHIN 0.005 0.001 0.004 0.001 0.026 0.001 0.007 0.003 0.001 -0.006 0.001 0.002 -0.001 -0.002 0.001 0.002 -0.001 0.003 0.000 0.001 0.000

Distance Positive (m) 250 50 100 600 250 175 0 175 600 150 175 175 never never never never 50 50 never 600 400

Mean DBH (cm) 62.41 22.07 28.22 24.43 32.74 38.51 27.20 26.77 48.90 24.54 23.75 38.33 40.51 34.16 32.49 28.35 44.58 50.77 34.32 27.48

Propy+s 0.029 0.676 0.419 0.596 0.095 0.143 0.565 0.545 0.000 0.632 0.696 0.143 0.050 0.200 0.308 0.414 0.125 0.103 0.424 0.568

Riparian habitats also showed a lower mean DBH (DBHriparian = 27.2 cm, DBHupland = 37.2 cm, p = 0.014). Riparian habitats have approximately double the proportion of saplings (sapriparian = 0.150, sapupland = 0.059, p = 0.026), double the proportion of „young‟ trees (youngriparian = 0.392, youngupland = 0.203, p = 0.005), and double the proportion of combined sapling + „young‟ trees (sap+youngriparian = 0.542, youngupland = 0.261, p = 0.003) (Table 5).

TABLE 5.5 COMPARISON BETWEEN RIPARIAN AND UPLAND HABITATS FOR SGS AND POPULATION AGE STRUCTURE.

Response Variable Sp Regression slope (b) Mean relatedness in first distance class (F1) Mean DBH (cm) Sapling proportion Young proportion Sapling/ young combined proportion Positive SGS distance (m)

Riparian Upland p 0.024 0.015 0.027* -0.022 -0.015 0.037* 0.080 0.050 0.053+ 27.2 37.2 0.014* 0.150 0.059 0.026* 0.392 0.203 0.005* 0.542 0.261 0.003* 131.3 80.0 0.037*

* significant + nearly significant

Linear regression further supported these results. Both ln(dnear) and ln(dfar) had significant predictive power for the Sp statistic (near: R2=0.357,p=0.004; far: R2=0.248,p=0.022) and proportion of young (near: R2=0.226,p=0.034; far:

88

R2=0.227,p=0.034), while sample size, number of alleles, gene diversity, and maximum distance for regression had no significant predictive power for either (Table 6).

5.3.4 Use of overall reference population In all populations, we observed an increase in Fij when a reference population of unrelated individuals was used, as expected. Mean per population Fij,OVERALL ranged from 0.012 to 0.129 (mean = 0.053) (Table 4), an average increase of 0.051. We were particularly interested in cases where relatedness is high across all distance classes, never crossing into the 95% confidence interval, indicating recent colonization by a small number of founders (Pardini, Hamrick, 2008). This signature of colonization was observed at four locations (Figure 4). All four were upland sites (Table 4). Three of these locations had not shown an overall pattern of SGS based on Fij,WITHIN (i.e. significantly positive F1 values) and one had. Riparian sites were less likely to be „colonizers‟ (0 out of 8) than upland sites were (4 out of 10) (Chi-square test, p=0.072).

TABLE 5.6 RESULTS OF REGRESSION WITH SP AND PROP(YOUNG)

Predictor variable Number of alleles Expected heterozygosity Observed heterozygosity Sampled individuals (N) Ln(max distance to river) Ln(min distance to river)

Sp Prop(young) 2 p R p R2 0.719 0.007 0.643 0.012 0.582 0.016 0.797 0.003 0.538 0.02 0.805 0.003 0.228 0.009 0.241 0.076 0.022 0.248 0.034 0.227 0.004 0.357 0.034 0.226

89

5.4 Discussion An emerging goal in population genetics is to quantify the ecological factors that shape SGS. Local habitat conditions- as a primary determinant of age structure, recruitment, metapopulation dynamics, and the spatial arrangement of individuals- may strongly influence SGS, and has recently received attention in the literature. However, the statistical power, and utility for future predictive models, of most previous investigations is limited by small numbers of sites and by lack of an a priori hypothesis of the underlying mechanism. We chose a forest tree capable of colonizing both riparian and upland habitats and sampled 21 populations located across the range. This approach has allowed us to statistically support the role of habitat as a determinant of spatial genetic structure. Further, we have demonstrated support for the mechanism by which habitat influences SGS, an altered population age structure, likely due to different opportunities (both distance and frequency) for seedling establishment.

5.4.1 Comparison between riparian and upland habitats We observed stronger SGS in riparian populations using a number of metrics: a significantly greater number of sites showing positive SGS, „stronger‟ SGS (Sp statistic, b, and F1), increased distance of positive relatedness, and a greater likelihood of recent colonization. This supports our proposed hypothesis and rejects the four alternative hypotheses. We summarize the mechanism by which habitat influences SGS as follows (Figures 5, 6). Riparian sites have more frequent disturbance (e.g., seasonal floods), more light, and less competition from flood-intolerant taxa, facilitating more frequent local recruitment of butternut. Other factors may also contribute, including lower herbivory in the open, riparian area, and an earlier age of death due to unstable soil, both 90

91 Figure 5.5. Cartoon of our model of population dynamics in a riparian butternut population

92 Figure 5.6. Cartoon of our model of population dynamics in an upland butternut population

of which would also contribute to more frequent local recruitment. Upland forested sites are characterized by light limitation, as well as fluctuations in water availability. The age distributions (using size as a proxy for age) of the populations further support this mechanism. While DBH measurements do not allow a determination of exact ages of trees, so we cannot directly determine the frequency with which establishment occurs, we can compare two characteristics, the mean DBH and the proportion of young trees. The lower mean DBH and higher proportion of young trees (which are correlated) suggests that riparian sites are more conducive to frequent, local recruitment. We recognize that better conditions in riparian sites may lead to faster growth, such that the DBH in riparian sites may be larger than that in upland sites, for trees of the same age. However, we saw that the DBH distribution in riparian sites is smaller, in spite of this expectation of better growth, so we can suggest that our observations likely reflect a younger age distribution. There were some exceptions to these observations, as several upland sites showed moderate levels of SGS and recruitment. This may have at least three explanations. First, disturbance regimes are variable in magnitude and frequency, and riparian zones vary from several meters of the river to hundreds of meters. Therefore, populations will often fall between the two „model‟ habitat types we have described. Second, the size of the canopy gap in upland sites, which can be quite large after extreme weather events (Peterson, 2000), will determine the number of colonists and the longevity of this population, in turn influencing SGS. Lastly, some sites may forests regenerated after timber harvests or on abandoned farmland. The affect of shifting patterns of human land use on SGS in forest trees are not be simple to predict, as the colonization and

93

recruitment dynamics (frequency and location of possible recruitment sites, resources, and competition) in such a landscape (Duncan, Chapman, 1999) will be unlike either „model‟ habitat we have described. A discussion of these possibilities are beyond our scope but has been approached elsewhere (Cottrell et al., 2003; González-Martínez et al., 2006; Kyndt et al., 2009; Soto et al., 2007). We do find it remarkable that we observe an influence of habitat on SGS in spite of the many unknown variables that characterize a site. Several recent investigations have focused on the effect of habitat on population dynamics and subsequent patterns of SGS, including changes in density and distribution of adults, distribution of suitable recruitment sites, seed dispersal mechanisms, and colonization dynamics (Table 1). Our results are concordant with an emerging consensus that frequency and location of suitable sites for seedling establishment, as well as density of surrounding trees, are influential determinants of SGS. This is an important departure from previous directions in fine scale spatial genetic studies, in which an isolation-bydistance process was assumed, based on distance between parent and offspring (e.g., (Fenster et al., 2003; Heuertz et al., 2003; Rousset, 1997)). While isolation-by-distance may apply in species with high density, continuous stands, a more complex picture is apparent for rarer species. For example, Born et al (2008) and Sezen et al (2009) both emphasize that mother-offspring distances and the isolation-by-distance process may be a poor explanation of SGS in their systems, both tropical trees with scattered distributions. They observed that co-dispersal of siblings into suitable establishment sites is the mechanism creating higher relatedness among neighbors, rather than distance between parent-offspring. Our results, by emphasizing the role of colonization of suitable sites

94

and by showing the high contribution of siblings to SGS, suggest a similar process in temperate species such as butternut, a suggestion also made much earlier (Levin, 1988). This is the opposite of the finding of (Cottrell et al., 2003), in oaks, in which high kinship was primarily between young and old trees. We suggest that rare species, temperate or tropical, have very distinct mechanisms influencing local distribution of diversity, through different recruitment and colonization dynamics. Consequently, both the design and interpretation of SGS studies should carefully distinguish between the two processes.

5.4.2 Comparison of butternut to other tree species The range of Sp and F1 values observed in butternut span nearly the entire range observed in wind-pollinated, outcrossing tree species (Table 3). The mean of these statistics are among the highest values observed in similar species, and are actually in the range for animal pollinated or herbaceous species (Vekemans, Hardy, 2004). This is likely due to three aspects of butternut reproductive biology. First, butternuts are heavy, and are likely primarily dispersed by gravity or rodents. Gravity dispersal, particularly with heavy fruits, is generally limited (Hardy et al., 2006; Kyndt et al., 2009). However, rodent caching may have a more complex influence. Squirrels may transport larger seeds further for caching because of their high nutritional value (Jansen, Forget, 2001; Tamura, Hayashi, 2008), but squirrel caches may often be composed of nuts from the same tree, leading to co-dispersal of sibling groups (Loveless, Hamrick, 1984). This dynamic is beyond the scope of our investigation, but deserves future examination. Second, butternut populations, even in riparian habitats, may be more ephemeral than those of most temperate trees, due to short life span and early successional nature. These metapopulation dynamics may be conducive to close relationships between nearest 95

neighbors (Pardini, Hamrick, 2008). Lastly, the density of butternuts may be a contributing factor to the strong spatial structure. Butternuts are usually distributed sparsely enough that seed shadows do not overlap, so that seedlings near a mother tree are likely from that mother tree, but close enough that pollen flow between nearest neighbors (and likely relatives) is more likely than mating between random individuals. This is the situation in which strong SGS may be most likely to develop (Vaughan et al., 2007), and we suggest that similar SGS may be observed in other early successional species with low density.

5.4.3 Implications for conservation of butternut Our findings suggest that habitat has a strong influence on population dynamics in butternut. As our sampling sites cover much of the range of the species, we can further suggest that the influence of habitat is a range-wide rather than local or regional phenomena. Therefore, habitat may be one of the central factors shaping population dynamics and spatial genetics in this species. The recognition of different population dynamics, and genetic consequences, in different habitat types is an important consideration for conservation of butternut and other species. First, our results suggest that habitat management may be critical for promoting regeneration in butternut, such as creating gaps in upland sites to maintain this metapopulation dynamic and prevent local extinction. It has been recognized previously that even aged stands often lack the gaps needed for butternut, which has contributed to declining populations (Ostry et al., 2002; Ostry et al., 1994). Our results provide genetic evidence of the process underlying this need. A slightly different management focus may be needed in riparian areas. In particular, alterations of the natural riparian disturbance regime (via dams or agricultural 96

diversion) may alter the frequency and magnitude of disturbance as well as water resources, possibly causing local loss of butternut populations (and other riparian species), potentially altering community composition and function (Auble, 1995; Nilsson, Svedmark, 2002; Romano, 2010). For both habitats, our results suggest disturbance is a crucial mechanism facilitating seedling establishment. A greater knowledge of how populations recruit will contribute towards management of recovering populations in butternut and other declining tree species (Ostry et al., 1994; Schultz, 2003). Based on our results, we also suggest guidelines for seed sampling. Over all sites, average relatedness of trees within 30 meters (a normal guideline for sampling) was 0.144, and within 100 meters of each other was 0.117, approximately that of half siblings. Relatedness of individuals between 100 and 200 meters was half this amount, 0.068. Therefore, sampling from trees closer than 100 meters, especially many trees, is somewhat redundant in terms of preservation of genes, and if resources permit, sampling sparsely over wider areas would be more beneficial.

5.4.4 Relevance to experimental design Our study also demonstrates that small sample sizes (minimum = 19, mean = 39) are sufficient to detect moderate to strong SGS, and distinguish this from populations that do not show SGS, even though this is smaller than the ideal sizes suggested by previous work using simulations (Cavers et al., 2005; Epperson, Li, 1997). Other investigations have also demonstrated the presence of SGS with populations sizes as low as 20 to 50 individuals, with only small reduction in precision of various SGS statistics (e.g., Sp) (Born et al., 2008a; Hardy et al., 2006; Hu et al., 2010; Kyndt et al., 2009; Pardini, Hamrick, 2008). This recent trend suggests that, when highly polymorphic markers are 97

used, sampling hundreds of individuals is unnecessary if the objective is to detect the presence and strength of SGS (as opposed to quantifying pollen and seed dispersal distances or pursuing parentage analysis). For our investigation, confidence intervals around zero relatedness with ~30 individuals were 0.020 to -0.020, while with >150 individuals they were only slightly less than half this amount. Recognition of the diminishing returns of sampling high numbers of individuals per site should allow investigations to focus on more populations and fewer individuals. However, the incorporation of more populations underscores the need for a mechanistic hypothesis on which to base choice of study sites. Additionally, there is clearly a lower limit to population sizes in which an investigation of SGS is biologically relevant (i.e. that there will be related and unrelated individuals in the collection), and ~20-30 trees is likely near that limit. A future exploration of the statistical properties of low sample sizes would be insightful. Our study contributes to the growing body of literature employing the Sp statistic, which allows for comparison across populations and species, and is not affected by size of distance classes. However, our study also highlights a drawback for this statistic. We note that, for our data, Sp is overwhelmingly determined by the overall regression slope (b), and is little determined by F1. In many cases, the Sp statistic is nearly identical to the slope, b (Table 2), as noted by Hardy and Vekemans (2009). As F1 provides important information regarding relatedness of nearest neighbors, it may be a key, but often ignored, parameter in spatial genetics and population dynamics. Therefore, while Sp is a descriptor of the „strength‟ of SGS (Vekemans, Hardy, 2004), F1 (which may not be concordant with Sp) and b may capture the „shape‟ of SGS.

98

5.4.5 Conclusion Using a mechanistic hypothesis and a moderate number of populations, we have found statistical support for the influence of habitat type on SGS in butternut. We suggest that habitat may influence SGS in other species, particularly when populations exhibit metapopulation dynamics, occur in suboptimal habitat, or have small population size. In such cases, models of isolation-by-distance based on parent-offspring distances may be insufficient to explain SGS, and new models would be useful. Important parameters for these models may include sibling co-dispersal, frequency and location of recruitment (a function of disturbance), and population stability. Combined with other studies, our results also suggest that variation in SGS within a species may frequently exceed the difference observed between species. The findings that suitable locations for seedling establishment is a crucial determinant of fine scale SGS, and that SGS may rapidly evolve as the habitat and population dynamics change, are both important considerations in increasingly fragmented and degraded habitats. A deeper investigation of the interaction between SGS and colonization dynamics is a promising new direction integrating ecology, population dynamics, and genetics (Born et al., 2008b), that will generate far-reaching implications for habitat management, as well as a greater understanding of the mechanisms by which small populations persist in dynamic landscapes.

99

CHAPTER 6: HUMAN IMPACTED LANDSCAPES FACILITATE HYBRIDIZATION BETWEEN A NATIVE AND AN INTRODUCED TREE SPECIES

6.1 Introduction In the past two decades, conservation biologists have expressed increasing concern over possible consequences of hybridization between introduced and native taxa (Hails, Morley, 2005; Levin et al., 1996; Mallet, 2005), including competitive exclusion of the native taxa and loss of native diversity (Burke, Arnold, 2001; Ellstrand, 2003). Hybridization may also play a role in origin of novel traits that permit colonization of new habitats, which may exacerbate biological control of the introduced species or alter ecosystem function (Campbell et al., 2006; Fritz, 1999). The extent and spatial distribution of hybridization, which determines its evolutionary and ecological impact, may be largely determined by habitat (Metcalf et al., 2008; Milne, Abbott, 2008; Rubidge, Taylor, 2004). A greater understanding of how habitat contributes to the spatial and temporal dynamics of hybridization is needed to better evaluate this complex process, enable predictions, and facilitate management decisions (Wolf et al., 2001). Most investigations of hybridization between native and introduced taxa have focused on relatively short-lived animals (e.g. insects, passerine birds, and salmoniid fish) or annual plants. Despite the ecological importance of forests, the well-known propensity of temperate forest trees to hybridize, and unprecedented rates of 100

anthropogenic modification of habitat, little is known about the environmental factors that determine the outcome of interspecifc hybridization in trees (although see Burgess et al. 2008). The ecological and genetic dynamics that occur in short-lived taxa may not apply to trees, whose life history characteristics include leptokurtotic but fat-tailed dispersal curves, extended juvenile period, prolonged reproductive lifespan, and overlapping generations (Petit, Hampe, 2006). Many non-native plant species have been introduced to North America, purposefully or accidentally (Driesche et al., 2002; Zheng et al., 2004). Despite long (~2-10 million years) periods of separation, many introduced trees produce fertile hybrid progeny with native trees, including species in Morus (Burgess et al., 2005), Platanus (Santamour, 1970), Ulmus (Zalapa et al., 2007), Castanea (Jaynes, 1979), and Juglans (McDaniel, 1979). These and other trees comprise a central ecological role for habitat and mast in eastern North American forests. Many forests are experiencing decline due to abiotic and biotic stress (Van Mantgem et al., 2009). Hybridization may exacerbate this problem by introducing genetic, demographic, and ecological change, including accelerating the loss of native biodiversity (Levin, 2002; Rhymer, Simberloff, 1996) and disrupting coadapted gene complexes that have evolved within a lineage over long time scales (Rubidge, Taylor, 2004). On the other hand, hybridization may introduce useful adaptive traits (Schweitzer et al., 2002), such as pest and disease resistance (Fritz, 1999). While hybridization between sympatric tree species is relatively well characterized (Hamzeh et al., 2007; Valbuena-Carabaña et al., 2005), hybridization between native and introduced trees may be subject to different dynamics, including the historical pattern of introductions, the suitability of habitat, and the speed and direction of geographic spread.

101

In this investigation, we focus on Japanese walnut (Juglans ailantifolia Carrière), which has been planted widely in orchards, along roadsides, and near farmhouses for more than a century (Hoover, 1919; Neilson, 1930; Theiss, 1933). Hybrids between cultivated varieties of Japanese walnut and the native North American tree butternut (Juglans cinerea L.) are such vigorous, fruitful trees (Ashworth, 1969; Gellatly, 1966) that biologists have expressed concern over the possibility of an undetected, range-wide genetic invasion. We recently described natural hybridization in seven locations along the eastern edge of the native range of butternut. Hybridization was extensive in two landscapes with severe anthropogenic disturbance, and limited in less disturbed landscapes with more forest cover (Hoban, 2009). Researchers have recognized that disturbed or otherwise human influenced landscapes can serve as the source of introduction, facilitate colonization by both parental species, or provide a new habitat or resource that hybrids can exploit (Levin et al, 1996; Rieseberg et al 2003; Fitzpatrick and Shaffer 2007). For this investigation, we collected a total of 1611 individuals in 52 locations across the native range, to explore how habitat influences the extent of hybridization (Mahelka 2007, Hails and Morley 2005, Fitzpatrick and Shaffer 2004), the spatial aggregation of hybrid offspring, and direction of introgression (Milne 2008, Hamzeh 2008). Introduction of Japanese walnut has primarily been in small orchards, farms, roadsides, and back yards in small towns and villages. Forest fragments and woodlots regenerated from abandoned farmland will be near these introduction points. Continuous forest, while rarely free of human impact, is least likely to be close to introduction points. Therefore our first hypothesis is that hybridization rates are highest within anthropogenic

102

landscapes, moderate in fragmented forest, and lowest in continuous forest. We sampled throughout the range to test the alternate hypothesis that, rather than being associated with human disturbance, hybridization occurs as a regional hybrid zone along the eastern United States, as in some natural hybrid complexes (Ruegg 2008, Mullen 2008). Secondly, we investigate direction of gene flow, e.g., if one species is more likely to be the female parent of a hybrid than the other. Asymmetrical hybridization has been suggested in empirical studies and theoretical models of hybridization, and may be caused by differences in relative abundance, phenology, and partial one-way genetic incompatibilities (Burgess et al., 2005; Currat et al., 2009; Hamzeh et al., 2007; Lexer et al., 2005; Milne, Abbott, 2008). “Pollen swamping,” when one species contributes most of the potential pollen pool, may result in dilution of the other species‟ nuclear gene pool, and eventual organelle capture, which is well documented in sympatric oak species (Petit et al., 2004b; Valbuena-Carabaña et al., 2005). Based on our previous work, we propose that the direction of hybridization is asymmetrical and that this asymmetry is consistent across all habitats and in later hybrid generations (F2 and BC1). Alternatively, asymmetry may only exist for F1 hybrids or only in certain habitats due to differences in abundance of the two species and differences in pollen and seed dispersal. Asymmetrical gene flow has significant conservation implications for the speed and direction of gene flow across space, and the degree to which the native gene pool is retained in the hybrid population (Currat et al., 2009; Metcalf et al., 2008). Lastly, we investigate spatial dynamics within hybrid populations. In trees, related individuals are often clustered due to local recruitment of offspring (Loveless and Hamrick 1984, Vekemans and Hardy 2004, Petit & Hampe 2006), but microspatial

103

clustering of hybrids is a largely unexplored area. Because butternut has a heavy, gravity and rodent dispersed seed, which usually exhibit very restricted dispersal in temperate trees (Vekemans, Hardy, 2004), we propose that when hybrids occur in a population, they occur in clusters. In summary, we propose three specific hypotheses: (1) rates of hybridization are different in the three habitat types we examine, (2) J. ailantifolia is more likely to be the seed parent than J. cinerea in all locations, and (3) hybrids exhibit spatial aggregation within the populations where they occur. We use a larger set of markers than our previous study for greater confidence in our hybrid assignments. We discuss the implications of our findings in light of conservation management of this rare tree, and compare our results to other studies of hybridization between native and introduced species.

6.2 Materials and methods

6.2.1 Species Juglans cinerea (butternut, also known as white walnut) is a wind pollinated, outcrossing tree species native to eastern North American riparian ecosystems. Butternut grows well in flood plains but also occurs on limestone slopes and open sites with fertile, well-drained soil. A shade and drought intolerant species, butternut tends to be sparsely distributed even under favorable conditions. Butternut has experienced severe decline in the 20th century, primarily due to the disease butternut canker (caused by an introduced pathogenic fungus, Sirococcus clavigignenti-juglandacearum), and is now a rare tree

104

protected under several national and state threatened species lists (Fleguel, 1996b; Schultz, 2003). Butternut is also found in rural and suburban landscapes, in small woodlots or in yards, either deliberately planted or as natural recruits that are protected for the tree‟s edible nuts. Individual trees typically live ~70 years and are reproductive at 10 to 15 years (Fleguel, 1996b; Rink, 1990). The large ovoid fruit (5-10 cm in length and 3-6 cm in width) consists of a nut encased in a thick husk (Davis, 1966; Millikan et al., 1990). Depending on site conditions and tree health, a tree may produce between none and several hundred fruits per season, with heavy fruit set occurring every two to three years, also depending on site conditions (Davis, 1966; Ostry, Pijut, 2000; Reed, Davidson, 1958). Flowering and pollination occurs concurrently with bud break in April, and nuts develop throughout summer, although developing nuts will abort under drought or stress (McDaniel, 1956). After seed fall (October or November), butternuts are dispersed by squirrels, large birds, and gravity. The fruits float so water dispersal is possible. Japanese walnut (most cultivated varieties are called heartnut due to the distinctive nut shape) is native to riparian and mountain forests of Japan, and has similar reproductive biology to butternut (Ashworth, 1969; Reed, Davidson, 1958; Sargent, 1894). The first introduction to North America is unknown, but likely occured in the mid to late 1800s, after which Japanese walnut was widely planted on farms, along roadsides, and in orchards throughout eastern North America, in at least 30 states and 8 provinces (Neilson, 1930; Reed, Davidson, 1958). Cultivars of Japanese walnut (Japanese heartnuts) produce nuts with thin shells and a distinctive heart-shaped nut, in contrast to the thick, ridged shell of butternuts. Heartnut cultivars tend to be smaller trees than

105

butternut cultivars. Japanese walnuts are also distinguished from butternut by twig and bud characteristics, a smaller fruit size, an inflorescence that typically sets five to seven fruits in contrast to the two or three fruits of butternut, and a more spreading crown (Reed, Davidson, 1958). Additionally, some Japanese walnut cultivars have moderate to high levels of tolerance to butternut canker, although infections may occur (Ostry, 1997). Some populations of Japanese heartnuts have naturalized, particularly in woodlots, pastures, and abandoned fields. Trees with a mixture of characteristics of the two species as well as vigorous growth and remarkable reproductive output have been reported in many locations (Ashworth, 1969; Gellatly, 1966; Ostry, Woeste, 2004). A suite of morphological characters was recently developed to identify hybrid individuals (Ross-Davis et al., 2008a), but these guidelines do not enable identification of all hybrids, and will be unreliable for generations beyond the F1 due to independent assortment of traits. We recently developed both nuclear and chloroplast markers for use in identifying hybrid individuals, as well as their maternal ancestry (Hoban et al., 2008; McCleary et al., 2009).

6.2.2 Collections We collected samples from botanic gardens, arboreta, germplasm repositories, and nurseries that were identified as J. cinerea (N=113), J. ailantifolia (N=181) or hybrid (N=16). This collection was the reference set for hybrid assignment (see below, Hybrid analysis section). We also collected leaf or twig samples from 1301 trees throughout the range, from 52 locations (Table 1, Figure 1), some of which were populations while other locations were composed of just a few scattered trees. For all of these trees, we recorded 106

diameter at breast height (DBH) and GPS location. Tissue samples were kept cool during transport, and then at -20 C until DNA extraction.

Figure 6.1. Sampling locations of trees used in this chapter. Each dot represents at least one tree.

107

TABLE 6.1 POPULATIONS USED IN THIS STUDY

Site

State/ Province

Habitat

BB Barre CNNF CNF DG BValley Hoosier MCNP MarkTwain1 MarkTwain2 Ozarkother Ozarks1 Ozarks2 Ozarks3 G Washington GMNF_North GMNF_South SFNF Shenandoah North Shenandoah South Bernheim forest Allegheny_Dew_drop Allegheny_Riverway Allegheny_Golden Allegheny_Mayberg JR Jerricho Allegheny_Cook Blackville GilbertIsland Hartman Lake SACT Forested Mass Forest KeswickRidge Iowa NY NottawasgaLake Ontario_S-tech Putney Waupaca Whitewater

ON VT WI TN ON TN IN KY MO MO MO MO MO MO WV VT VT AR VA VA KY PA PA PA PA WV VT PA NB NB WI CT MA NB IA NY ON ON VT WI WI

forest forest forest forest forest forest forest forest forest forest forest forest forest forest forest forest forest forest forest forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest fragmented forest

N

JC

29 22 28 1 26 168 6 68 13 31 18 26 24 18 14 20 10 39 13 19 1 22 27 51 10 22 27 19 41 39 1 3 1 33 19 1 24 7 24 20 40

29 21 28 1 26 163 6 65 13 31 18 26 24 18 14 20 9 37 13 19 1 20 27 45 9 22 21 19 41 39 1 1 1 33 16 1 24 7 23 20 40

Habitat

N

JC

anthropogenic anthropogenic anthropogenic anthropogenic anthropogenic anthropogenic anthropogenic anthropogenic anthropogenic

27 3 33 51 19 2 1 1 63

9

JA

F1

F2

BCJC

BCJA

mix 1

1

2

1

2

2

1 1

1

1 1

4

1 1 1

2

3

1

2

1

1

1

1

TABLE 6.1 (Continued)

Site Allegheny_Yard_trees Bernheim farm Indiana Mass MM PA extra ND UW SA

State/ Province PA KY IN MA CT PA IN WA CT

108

7 18 4

1 7

JA

3 1

4

F1

F2

BCJC

9 1 16 6 3 1 1

3 1

2 1 4 8 5

1 3 2

2 11 2

24

10

6

3

9

5 2 1

BCJA

mix 4

NC Finger Lakes Total

NC NY

anthropogenic unknown location

21 3 1249

4 3 1045

5

7

14

78

28

2

1

2

36

10

38

Counts in this table are after removal of clones

6.2.3 Genotyping Samples were genotyped at twelve nuclear microsatellite loci and at least two (and in most cases three) species-specific chloroplast markers. Microsatellite loci showed varying diagnostic capability, from near species specificity to moderate differences in allele frequency. These microsatellites have been previously tested in butternut for null alleles, Hardy Weinberg Equilibrium (HWE) and linkage disequilibrium (Hoban et al., 2008). We did not test for HWE for these data because admixture due to interspecifc hybridization is a major violation of the assumptions for equilibrium. Before further analysis, we used the identity analysis feature of CERVUS (Kalinowski et al., 2007) to identify duplicate genotypes in our samples. In natural populations, some butternuts occur as a cluster of several trunks (personal observation), which may result from stump sprouting, in which case all stems will have identical genotypes. In germplasm repositories, botanical gardens, and nurseries, grafting is a common method of propagating cultivated varieties, a practice that can occasionally result in identical genotypes with different names (Hoban, unpublished data). Identical genotypes were removed from analysis, as they bias the allele frequencies used to define a species (see NEWHYBRIDS analysis below).

6.2.4 Hybrid analysis We used NEWHYBRIDS (Anderson, Thompson, 2002) to assign a „hybrid status‟ to each individual. This software uses a Bayesian approach to generate allele frequencies 109

for each species and then assign individuals to one of the following categories: J. cinerea, J. ailantifolia, F1 hybrid, F2 hybrid, first generation backcross to J. cinerea (BC1 to J.c.), and backcross to J. ailantifolia (BC1 to J.a.). We used a cutoff of 0.75 probability in a given category for making an assignment, although the vast majority of assignments were made with >0.90 probability. Individuals that were not assigned to any single category with greater than 0.75 were assigned to a „mix‟ hybrid category. These individuals are likely offspring of an advanced hybrid cross. Program settings: Jeffrey‟s prior on pi and theta, burn-in: 100,000, MCMC steps: 500,000. We did not use the z option (use of a reference „training‟ set), because in our previous investigation we found that samples from arboreta, botanic gardens and nurseries that are reputed to be a pure species may occasionally be hybrids. Therefore, inferences from NEWHYBRIDS are based entirely on the genetic data. The choice of prior may affect some hybrid assignments when a large number of rare alleles are present in either or both species. In particular, the Jeffrey‟s prior may “provide apparent sharpness to hybrid assignment (Anderson, 2003).” We therefore recalculate hybrid assignment using the uniform prior.

6.2.5 Habitat assignment Each location was assigned one of the following habitat types: continuous forest, fragmented forest, or anthropogenic landscape (Table 1). While forests occur on a continuum between these habitat sites, our assignment was based on the following. Continuous forest sites were characterized by large (hundreds to thousands of hectares) tracts of forest with no farms or other development, although roads or hiking trails were present. These sites, as with most forests in Eastern North America, have been logged to various extents, and in some cases briefly used for agriculture, so this designation reflects 110

contemporary conditions. Fragmented forest sites are small (< 30 hectares), and typically occur in a matrix of agricultural and residential land. Locations on the edge of a continuous forest where many trees occurred along roads or the edge of fields were also classified as fragmented. Anthropogenic sites include suburban and rural yard trees, as well as trees along fence rows or roadsides in agricultural landscapes, in which no continuous or fragmented forest existed. Small woodlots (> 1 hectares) were considered anthropogenic. In anthropogenic landscapes, some trees were likely planted while others were obvious natural recruits.

6.2.6 Tests between habitat type To test the hypothesis that the habitat type (continuous forest, fragmented forest, and anthropogenic) results in a difference in rates of hybridization, we used a Fisher‟s exact test to compare the habitat types for counts of non-J. cinerea individuals (the sum of all hybrid categories, the „mix‟ class, and J. ailantifolia). To test the hypothesis that, among the hybrid individuals, habitat type results in a different proportion of each hybrid category (i.e. one habitat may be more conducive to F1 hybrids), we used a Fisher‟s exact test to compare counts in each category, in each habitat.

6.2.7 Maternal ancestry Our previous study indicated that J. ailantifolia is usually the female (seed) parent of hybrids, but our previous sample size was not large enough to demonstrate a significant difference. To test the null hypothesis that the two species are equally likely to serve as seed parent, we used a Fisher‟s exact test to compare the number of hybrids with the J. cinerea chloroplast to the number of hybrids with the J. ailantifolia 111

chloroplast across habitats, across hybrid types (F1, F2,BC1to J. ailantifolia, BC1to J. cinerea or Mix), and across all hybrids in total. All tests were performed in R (R-CDT, 2005).

6.2.8 Spatial analysis Within each continuous or fragmented forested population in which we found more than one hybrid (N=6), we measured the pair-wise geographic distance between all non-J. cinerea individuals (hereafter, dnot-jc) and the pair-wise geographic distance between all J. cinerea individuals (hereafter, djc). For the four populations in which more than two hybrids were found (so we could establish a mean dnot-jc), we compared the distribution of dnot-jc to the distribution of djc using t-tests. For the two populations in which only two hybrids were found (and thus only one dnot-jc was available), we used a z test to compare dnot-jc to the distribution of djc. Geographic distances were calculated using SPAGEDI (Hardy, Vekemans, 2002). If dnot-jc is significantly smaller than djc within a given site, e.g. hybrids are significantly closer to each other than J. cinerea individuals are to each other, we conclude that hybrids are spatially clustered.

6.3 Results

6.3.1 Genetic diversity and number of hybrids We genotyped a total of 1611 trees. Of these, 210 were identified as duplicate genotypes (102 J. ailantifolia reference, 9 hybrid reference, 47 J. cinerea reference, and 52 naturally occurring trees). These were removed, leaving 1401 unique genotypes for

112

analysis: 79 J. ailantifolia reference, 7 hybrid reference, 66 J. cinerea reference, and 1249 naturally occurring trees. A total of 356 alleles were identified (maximum per locus=69, minimum=18, mean=29.7). Mean observed heterozygosity across all individuals and loci was 0.748. Marker loci showed strong allele frequency differences between the species, and three were nearly diagnostic (WGA_82, B121, and B264). Among the 1249 naturally occurring individuals examined, we identified 1045 J. cinerea (JC), 14 J. ailantifolia (JA) , 78 F1 hybrids, 28 F2 hybrids, 36 BC1 to JC, 10 BC1 to JA, and 38 „mixed‟ hybrids (Table 1). Within sites in which more than three individuals were collected, the extent of admixture (non- J. cinerea individuals) ranged from 0 to 89%.

6.3.2 Comparison between habitats The forested habitat was represented by 20 locations (N=593), the fragmented habitat by 21 locations (N=402), and the anthropogenic habitat by 10 locations (N=221). Five forested sites, seven fragmented sites, and all but one anthropogenic site showed some hybrid individuals. Considering only sites where hybrids were found, and in which more than three individuals were collected, admixture (proportion of non-native individuals) ranged from 5 to 10% in the continuous forested sites; from 4 to 22% in the fragmented forested sites, and from 65 to 89% in the anthropogenic habitat. A Fisher‟s exact test revealed that habitats were significantly different for admixture, with forested sites having the least (2.0%), then fragmented sites (4.6%) and then anthropogenic sites (43.6%) (p<<0.001) (Table 2). Overall, habitat type did not influence the proportion of different classes of hybrids (p=0.204), but the forested sites 113

were marginally significantly different from the anthropogenic sites for hybrid classes (p=0.045), due to fewer F1 hybrids and more „mixed‟ hybrids in forested sites than the other two categories (Table 3).

TABLE 6.2 COUNTS OF J. CINEREA AND NON-J. CINEREA IN EACH HABITAT TYPE

Count of Non J. cinerea 12 21 171

Count of J. cinerea Forested Fragmented Anthropogenic

593 432 221

Proportion of Non J. cinerea 0.020 0.046 0.436

TABLE 6.3 COUNT IN EACH HYBRID CLASS, IN EACH HABITAT TYPE

Forested Fragmented Anthropogenic

F1

F2

1 9 68

2 4 22

BCJC BCJA 4 4 28

0 0 10

mix 5 3 30

We found a strong asymmetry in seed parents for hybrid individuals. A significantly higher number of hybrids (89.1%) had the J. ailantifolia chloroplast (p<<0.001) (Table 4). Additionally, the hybrid classes were significantly different for the proportion of J. cinerea to J. ailantifolia seed parents (p=0.028). This overall difference was due to a highly significant difference between the F1 and the F2 hybrid classes 114

(p=0.005); 94.7% of F1 hybrids and 73.1% of F2 hybrids had the J. ailantifolia chloroplast (Table 4). No other hybrid classes showed significant differences. Considering only hybrid individuals, habitat types were also significantly different for the proportion of J. cinerea to J. ailantifolia seed parents (p<0.001), and pair-wise comparisons revealed that all three habitat types differed in this proportion (Table 5). Specifically, 63.6% of hybrids in the forested habitat had the J. cinerea chloroplast, 20% of hybrids in the fragmented habitat had the J. cinerea chloroplast, and 5.9% of the hybrids in the anthropogenic habitat had the J. cinerea chloroplast. Spatial analysis of the six populations in which more than one hybrid was found revealed that in two populations hybrids were significantly clustered (distance between non-J. cinerea was significantly smaller than distance between J. cinerea), while in four populations they were not. Both populations in which spatial clustering was shown were fragmented populations. Of the four that did not show clustering, three were continuous forested and one was fragmented (Table 6).

TABLE 6.4 OCCURRENCES OF J. CINEREA AND J. AILANTIFOLIA CHLOROPLAST IN EACH HYBRID CLASS

JC chl JA chl F1 F2 BCJC BCJA mix Total

4 7 3 0 6 20 115

72 19 31 10 32 164

Proportion with JC chl 0.053 0.269 0.097 0.000 0.158 0.109

TABLE 6.5 OCCURRENCES OF J. CINEREA AND J. AILANTIFOLIA CHLOROPLAST IN HYBRIDS IN EACH HABITAT

Continuous Forest Fragmented Anthropogenic

JC chl

JA chl

9 4 7

144 16 4

Proportion with JC chl 0.059 0.200 0.636

TABLE 6.6 MEAN PAIR-WISE DISTANCE BETWEEN J. CINEREA INDIVIDUALS AND BETWEEN NON J. CINEREA INDIVIDUALS IN SIX POPULATIONS

Continuous Forested Population mean d J. cinerea mean d non-J. cinerea P value, one-tailed

SFNF 0.072 0.080 0.435

HS 0.008 0.012 0.910

MC 0.082 0.082 0.494

Fragmented DD 0.011 0.028 0.984

JVT 0.003 0.001 0.005*

GR 0.002 0.001 <0.001*

6.4 Discussion Previous work in both plant and animal taxa has shown that many introduced and native taxa hybridize, and that the degree of hybridization (i.e. proportion of hybrids in the population) may vary in space and time (Hails, Morley, 2005; Levin et al., 1996; Wolf et al., 2001). Hybridization may occur rapidly and in all populations (Metcalf et 116

al., 2008; Rubidge, Taylor, 2004), or may be limited, such as to F1 hybrids (Mallet, 2005; Milne, Abbott, 2008), or to a habitat mosaic (Fitzpatrick, Shaffer, 2007b; Watano et al., 2004). We investigated hybridization between a native and an introduced forest tree to establish range-wide patterns of hybridization, investigate the influence of habitat on hybridization rates, and determine the primary direction of gene flow. This is one of the first demonstrations of hybrid dynamics between a native and introduced forest tree in North America, and to our knowledge the largest in spatial scale. By investigating across the entire range, and by specifically focusing on habitat types, we show that hybridization is not advancing in a regional hybrid „front,‟ but rather as „bubbles‟ within suitable habitat types, and that habitat significantly influences both extent and direction of hybridization, and possibly spatial clustering of hybrids.

6.4.1 Comparison between habitats We demonstrated that hybridization rates differ between continuous forest, fragmented forest, and anthropogenic habitat. In particular, rates of hybridization were much higher in anthropogenic sites. Considering only sites with more than one tree collected, all anthropogenic sites showed large rates of hybridization: the minimum admixture (percentage of non-J. cinerea trees) within an anthropogenic site was 67%, while no forested or fragmented site showed more than 22% admixture. We explain high rates of hybridization in anthropogenic locations as follows. First, introductions of J. ailantifolia have historically been in rural areas and small towns, in back yards, along fences, and along roads (Gellatly, 1966; Hoover, 1919; Morris, 1917; Neilson, 1930; Theiss, 1933). J. cinerea is also a popular farm tree, and is frequently protected by farmers if growing along field edges or fence rows (Nielsen et al., 2003; Ostry, Pijut, 117

2000), so the two species often co-occur in this habitat. As phenology for the two species overlap and both species are heterodichogamous, many offspring formed (as seeds) on lone J. ailantifolia trees will be hybrids. The open nature of anthropogenic habitat, which is suited to the early colonizing nature of both species, will facilitate recruitment of hybrids along fences and later into the fields if the farmland is abandoned. Additionally, farmers, who have long recognized the vigor and reproductive capacities of hybrids, often preferentially allow hybrid recruits to grow (Hoover, 1919; Morris, 1917; Reed, Davidson, 1958). In contrast to the anthropogenic habitat, we observed that both continuous and fragmented forested sites show little hybridization. The patterns of chloroplast data offer a likely explanation. Higher representation of the J. ailantifolia chloroplast in hybrids, and the highest degree of asymmetry in F1 hybrids, shows that most hybrid seed results initially from J. cinerea pollination of a J. ailantifolia flower. As seed dispersal is generally quite limited in temperate trees (Hardy et al., 2006; Petit et al., 2005), hybrids formed on a J. ailantifolia mother tree will likely recruit locally in the anthropogenic habitat. Movement of seed out of this habitat, even into neighboring forest, may be rare. This could explain why fragmented habitats show approximately twice as much hybridization as forested sites (4.5% compared to 2%), as they are nearer to anthropogenic sites. In contrast to anthropogenic habitats, most hybrids in forested environments have the J. cinerea chloroplast, suggesting that when hybridization events do occur in forests, they are primarily due to J. ailantifolia pollen dispersal. These events are likely to be rare due to the greater proportion of butternuts (and butternut pollen) in the forest.

118

Our spatial analysis, although at a limited number of sites, supports this scenario. Both populations that contained clustered hybrids were fragmented forests, and the clusters of hybrids were on the boundary between forest and field. In one site, in the interior of the fragment, we identified an isolated hybrid with the J. cinerea chloroplast. In comparison, in the continuous forested habitats, hybrids were never significantly clustered, which is consistent with rare dispersal of J. ailantifolia pollen into the forest. Another possible, although perhaps less parsimonious explanation, is fitness differences between habitats, i.e. that hybrid individuals are at a fitness disadvantage to J. cinerea in forested sites. Habitat specific hybrid fitness has been previously reported in both plants and animals (Mercer et al., 2006; Tauleigne-Gomes, Lefèbvre, 2005). We can speculate on several possible fitness differences between J. cinerea and hybrids that would depend on habitat. First, it may be that the vigorous growth of hybrids requires greater water and light resources than are available in some forested locations. It is also possible that hybrid seeds are subject to greater predation in the forest, as butternut has a very thick shell, Japanese walnut has a thin shell, and hybrid individuals may have a range of shell thickness (Ashworth, 1969; Gellatly, 1966). Thirdly, the greater competition in forested sites (a significant physiological stress) may expose a genetic incompatibility that does not appear in individuals not subject to competition (Hochwender, Fritz, 1999; Johansen-Morris, Latta, 2008). Lastly, it is also possible that most hybrids exhibit hybrid breakdown during early stages of development (Bomblies et al., 2007; Burke, Arnold, 2001; Turelli, Orr, 2000), and that the high rates of hybridization in open habitats are simply a product of much greater reproductive output, increasing the likelihood of at least some surviving hybrids. However, we suggest that,

119

while these factors may contribute to the observed patterns, the differing patterns of hybridization between habitats can be explained by the much higher proportion of J. ailantifolia initially in anthropogenic habitats, and dispersal limitation out of these habitats.

6.4.2 Comparison to other work Previous studies in short lived species have approached the importance of habitat in hybrid population dynamics. Fitzpatrick and Schaffer (2007) have shown that habitat type greatly influences the degree to which hybridization occurs between California and barred tiger salamanders (Ambystoma californiense and A. tiggrinum mavortium, respectively). In this system, hybrids achieve a fitness advantage over the native in permanent ponds, growing to greater size and producing more offspring. Habitat also influences the fitness and success of hybrids in other species, including those in Helianthus and Iris (Martin et al., 2006; Rieseberg et al., 2003). However, Conesa et al (2008) suggested that habitat does not play a role in recurring hybridization between two Viola species. It is unlikely that habitat will always influence hybridization rates, but it is increasingly evident that habitat often does so, either by altering the fitness environment or changing dispersal patterns. Future work is clearly needed to understand the degree to which dispersal vs. fitness differences alter hybrid dynamics. Fitness measures of hybrids with reference to J. cinerea and J. ailantifolia, under different conditions, and studies of the genetics of tolerance to butternut canker in both species and their hybrids, will help elucidate the possible role of selection. Quantifying gene dispersal in J. cinerea will also contribute to this future direction.

120

Our results are consistent with past observations in both plant and animal taxa that hybridization is often asymmetric (Hamzeh et al., 2007; Metcalf et al., 2008; Milne, Abbott, 2008). While asymmetry may be due to reproductive incompatibilities (Landry et al., 2007; Tiffin et al., 2001), we suggest that an equally parsimonious explanation in our system is habitat-specific relative abundance and dispersal patterns. A recent model of introgression suggested that gene flow should be primarily from the native to the introduced species (Currat et al., 2009), in absences of selection, as the native (typically, at least at first) is more numerous. This model predicts that the genetic impact of the introduced species will be minimal. The recognition of frequent asymmetry in gene flow has important implications for predicting how hybrid populations will change over time, particularly the degree to which native genes will be retained (Allendorf et al., 2001; Hails, Morley, 2005; Mallet, 2005). Selection need not necessarily be invoked as a mechanism by which habitat can facilitate hybridization.

6.4.3 Butternut canker and hybridization We and others have previously suggested that hybrids may exhibit a fitness advantage due to tolerance of the butternut canker disease. While the butternut canker infection is usually lethal to butternuts, hybrid offspring display a wide range tolerance to butternut canker, which may increase the likelihood that some hybrid seedling will survive to maturity (Orchard, 1984; Ostry, Woeste, 2004; Schlarbaum et al., 2004). However, we observed several hybrids that were heavily diseased (Hoban, unpublished data), suggesting that tolerance is not universal in hybrids. This agrees with observations from Orchard (1984, 1982), who showed that butternuts, hybrids, and J. ailantifolia individuals exhibit a large range of disease tolerance. Tolerance may be due in part to 121

remarkable growth of early generation hybrids, which are able to callous over the canker wound, preventing spread of the canker (Orchard, 1984). Hybrid vigor is likely to disappear in advanced generations (Facon et al., 2005; Johansen-Morris, Latta, 2008). On the other hand, some tolerance may be inherited from J. ailantifolia as a polygenic trait. If so, the disease may provide positive selection for later generation hybrids with larger blocks of the J. ailantifolia genome. A future direction of this project is to better quantify and understand the extent of tolerance in J. ailantifolia, J. cinerea, and hybrids. Whether or not the disease gives hybrids a fitness advantage, our results make it clear that the disease has not resulted in a widespread replacement of J. cinerea with hybrids. In all forested locations, J. cinerea outnumber hybrids, in spite of the vigor of hybrids, the length of time since introduction of J. ailantifolia (five to ten generations), the success of backcrosses, and the disease pressure. However, future predictions are difficult, as thresholds may exist after which hybridization quickly becomes widespread, or disappears. Furthermore, if populations of butternut continue to decline due to disease and other factors, the hybrid proportion may increase. Future exploration of these possibilities may be pursued with mathematical models of population and disease dynamics.

6.4.4 Predictions and future directions The identification of the importance of habitat type and the dispersal limitation of hybrids has important implications for hybrid population dynamics over time (Rhymer, Simberloff, 1996; Wolf et al., 2001). For butternut, most hybridization is occurring in „bubbles‟ of anthropogenic habitat, and the spatial expansion of a hybrid population out of this landscape may proceed slowly, or not at all. In forested habitats, we propose that 122

rare immigration of J. ailantifolia relative to the vast majority of J. cinerea pollen will continue to result in a dynamic balance of creation and dilution of hybrid genotypes. However, it is apparent that no population, even those in large tracts of continuous forest, is „immune‟ from hybridization events, due to rare immigration by pollen. Prediction of future rare hybridization events will be difficult due to unknown pollen dispersal curves, which presents difficulties for conservation management goals that are aimed at a zerotolerance policy for hybridization. On the other hand, hybridization is uncommon in most forested sites. The molecular tools that we have developed will enable researchers to verify the ancestry of collections both for ex-situ conservation and further study of tolerance of the disease. Several future directions are apparent. First, an important question is the parentage source of hybrid individuals in forested habitats, particularly the widely separated individuals, as we identified very few J. ailantifolia in any population. At this point we do not know if hybrids within a forest are the product of one pollen parent or several. The small census size typical of butternut populations (usually <100) should make paternity exclusion analysis possible, but more molecular markers (20+) may be necessary for highly confident assignments, as some J. ailantifolia cultivars are closely related. Such knowledge would shed light on limits to hybrid seed dispersal, as well as realized long distance pollen dispersal. Our extensive collection of named varieties of J. ailantifolia may reveal whether a particular cultivar is often the parent of these hybrids. Further, the development of more markers, especially nuclear diagnostic markers, will increase the certainty of the hybrid classes. This study suggests a greater degree of

123

backcross to J. cinerea than to J. ailantifolia, but this is still a small sample, with some margin of uncertainty. We also note no population in Wisconsin or Canada showed evidence of hybridization. This pattern may be due to environmental conditions, as J. ailantifolia is much less cold tolerant than J. cinerea, which means that both introductions of J. ailantifolia and subsequent hybrid survival may be greatly reduced in these areas. Analysis of additional Wisconsin and Canadian samples could verify whether hybridization is truly absent in these locations. If hybrids do suffer greater mortality in colder regions, this may need to be considered in planned hybrid breeding programs. As mentioned earlier, common garden experiments and a greater knowledge of dispersal will allow us to determine the degree to which selection and dispersal contribute to the influence of habitat on hybridization dynamics. Finally, our work clearly demonstrates that frequent hybridization between two species separated for ~2-10 million years occurs, that hybrids are vigorous and fertile, and that advanced generation hybrids grow to maturity. Many forest tree species of the northern hemisphere now native to Europe or east Asia will likely hybridize successfully with North American species in the same genus, despite millions of years of allopatry (He, Santamour, 1983; Wen, 1999), a key consideration for the genetic consequences of past, current, and future introductions. While presenting a great conservation challenge, this also presents opportunities for studies of speciation. We find it remarkable that hybridization in forest trees has received so little notice in the literature devoted to the origin of species.

124

Overall, our results show that habitat is a key consideration in the dynamics of hybrid populations. In butternut, as well as other species, anthropogenic habitats may be locations where the probability of hybridization is highest. However, we did not observe a clear continuum of hybridization, as continuous forests only had slightly less hybridization than fragmented forests, indicating that there may be a threshold between fragmented forests and anthropogenic. If conservation objectives are to limit the degree of hybridization, conservation management approaches (e.g., hybrid removal) may differ between the two habitats. Further, we show that habitat not only affects the extent of hybridization but also other population dynamics including the direction of gene flow, which will determine the degree to which native alleles are retained in hybrid populations. In addition, we suggest that the influence of habitat does not necessitate fitness differences, but may be due instead to dispersal limitation and relative census size of the two species. The influence of habitat on dispersal may supersede the influence of habitat on selection by preventing hybrids from establishing in the first place. If hybridization is frequently dispersal limited, as we suggest, the preservation and restoration of large blocks of continuous, natural landscapes may form a barrier against genetic invasion, and in situ preservation of these natural areas should receive high priority.

125

CHAPTER 7: CONCLUSION

In this dissertation, I have explored several ways in which populations respond to ecological changes caused by both natural and anthropogenic processes. Specifically, I have quantified the relative effects of large-scale contemporary and historical ecological change on population size and connectivity (Chapter 4), the contrasting effects of riparian and upland habitat on recruitment and colonization (Chapter 5), and the facilitative effect of anthropogenic landscapes on interspecific gene flow (Chapter 6). To do so, I used techniques from spatial and molecular ecology, to first identify signatures of genetic change (interpopulation gene flow, admixture, genetic drift, diversity, and dispersal of relatives), and then correlate them to environmental variables. My work has encompassed large and small spatial and temporal scales, from Holocene-era range shifts to contemporary patterns in seed dispersal. Throughout my research, I have grounded my questions in current and often controversial topics in ecology and population genetics, as well as pressing issues in population and conservation biology. Some of my conclusions concur with decades of past research, while others challenge past conclusions and present new, exciting research directions. Here I will summarize my main conclusions, and describe how they advance the fields of population, conservation, and molecular ecology.

126

My results from Chapter 4 suggest that large-scale contemporary population losses, increasingly common in forests due to biotic and abiotic stress, should be considered in the light of past events. Overlapping generations, long life span, and high dispersal abilities may provide a buffer against genetic loss during rapid environmental change, helping populations to retain heterozygosity and avoid inbreeding, even though local population sizes and migration rates may be altered (RobledoArnuncio et al., 2005; Savolainen et al., 2007). My data indicate that this buffer has limits, as genetic bottlenecks and differentiation clearly increase towards the northern edges of the range. The limits of a species‟ ability to retain diversity as population census falls clearly merits more detailed investigation. This potential for genetic buffering should not be restricted to trees, as some animals (particularly some of the charismatic megafauna that are the focus of much conservation work) also exhibit these characteristics (Archie et al., 2008; Okello et al., 2008). Declines in many species, due to past climate shifts or epidemic disease, may equal or exceed those that are observed today (Johansson et al., 2006; Storz, Beaumont, 2002; Taylor et al., 2007). An approach similar to the one I have taken will enable researchers to determine the relative impact of contemporary change, as well as to predict the impact of further decline. More specifically, carefully designed sampling schemes, and comparative statistical analysis, including recently developed Bayesian methods (Faubet, Gaggiotti, 2008; Foll, Gaggiotti, 2006), will contribute to this research direction. I did not incorporate other contemporary changes such as logging and overall forest fragmentation into my analysis, but this would be possible using a similar approach.

127

I introduced Chapter 4 by stating that the knowledge generated therein should improve the predictive power of models of future population responses. My results suggest that key parameters include the temporal length as well as the magnitude of size reduction, as well as connectivity to other populations. The long-term effects of changes in these parameters is still relatively unknown, particularly how they interact. Past research has suggested that short, intense bottlenecks and long, shallow bottlenecks will have different effects (Garza, Williamson, 2001; Williamson-Natesan, 2005), but these models may be biologically irrelevant as they often do not incorporate heterogeneity in population size and distribution. A challenging but important question to explore in the future would be, “What contemporary population size reduction would produce the same genetic impact as postglacial colonization founder events?” Mathematical models parameterized with genetic data from this and other studies will contribute to this question. Another important area of future study will be to model not only the loss of diversity due to population bottlenecks but also the recovery from these bottlenecks, which has been a neglected area of research. Knowledge of the time and the gene flow needed to make up for loss of diversity, and relationship between census and effective population size, will help determine whether conservation plans such as reintroductions will be effective, and should provide a better understand how genetic drift affects small populations of trees. An additional direction in predictive modeling research is to further explore the level of connectivity needed to avoid bottlenecks. A common rule-of-thumb in conservation genetics has been that at least one migrant per generation (OMPG) is

128

required to avoid loss of genetic diversity in small populations. While violations of the many simplifying assumptions of this rule have been explored in animals (Vucetich, Waite, 2000; Wang, 2004), the OMPG rule has not been extensively explored in plants, particularly those with overlapping generations. Based on my results, I would suggest that lower levels of migration may be allowable, as even isolated butternut populations have retained substantial fractions of the total diversity in the species. A final direction of work for these predictive models will be to better understand postglacial spread, including the speed of colonization and size of colonizing populations. The butternut project is well suited to this exploration, and is similar to much previous work in both Europe and North America (McLachlan et al., 2005; Petit et al., 2002; Tollefsrud et al., 2009). Future work for this investigation should include validation of the presumed extreme ecological marginality of the range edge using both environmental data (to produce habitat suitability metrics for „typical‟ and „marginal‟) and Forest Inventory Analysis (to quantify surveyed population size and distribution). This is important to verify that the range edge truly results in smaller and more widely separated populations, as the abundant center hypothesis has received some criticism (Kirkpatrick, Barton, 1997; Sagarin et al., 2006). It would also be instructive to compare the range edge conditions for butternut (a species with moderate to low population densities) to dominant hardwood taxa (e.g., Quercus, Acer, Fagus). Another focus of my research was to explore and validate frequently used tools in population genetics by comparing results from different approaches. In this investigation, I used a recently developed Bayesian method designed to identify the

129

environmental variables that are most influential in determining population differentiation. This method was in agreement with classical model simplification as well as two Bayesian clustering methods, which were also concordant with each other. While the estimates of FST from this new method were slightly inflated, they were highly correlated to traditional methods. This new method will likely supersede classical model comparison in future studies when many predictor variables must be considered simultaneously, e.g., the multiple spatial and temporal scales of both natural and anthropogenic processes. I also used, as others have, several methods to distinguish between signatures left by recent and historical bottlenecks. A further exploration of the signatures left by different kinds of bottlenecks would be very useful, as it is possible that more information, such as length of the bottleneck and contribution of later migration vs. mutation, may be extracted from the genetic data. Even without implementation of mathematical simulations, it is apparent that there are limits to the decline that a species can tolerate, without lasting genetic change, as demonstrated by my results for range shifts. Even if these limits were known, I would not suggest that populations be allowed to reach the absolute minimum population size, as the stochastic forces that contribute to population viability are inherently difficult to predict. However, in some species, a balanced evaluation of contemporary decline in light of a species‟ past may reveal that more targeted conservation activities may be sufficient. Moderate conservation actions for some species could allow resource allocation to serve a greater number of species. Further, the population sizes needed for ex-situ conservation of some species may be considerably smaller than generally recommended (Allendorf, Ryman, 2002; Frankham et al., 2002) and still meet

130

conservation needs. Similarly, some in situ populations may retain conservation value even though they are very small. Even extremely isolated individuals may be contributing to recruitment (Ahmed et al., 2009). In particular, based on my results, I suggest, as others have recently, that the minimum viable population size for such species may be < 50 individuals, although this is well below some recommendations. A better understanding of the extent of human influence on population ecology, stability, and survival may reveal conservation compromises that preserve natural population processes to some extent, but also integrate with human landscapes. A future goal of conservation biology will likely be to determine what characteristics enable species to coexist in a radically altered landscape, especially at small population sizes. Chapter 4, while optimistic in some respects, also reiterates one of the great challenges of global climate change: to preserve genetic diversity in species experiencing range shifts, contractions, or expansions. My results show that the northernmost populations of butternut, and perhaps those of many other relatively rare temperate species, have yet to recover from bottlenecks from range shifts that occurred more than ten thousand years ago. Lower diversity and high differentiation, like that seen in northern populations of butternut, are likely to be found in species that colonized quickly after the ice age but were not numerous enough, or did not have sufficient long distance gene flow, to restore genetic connectivity to populations after the shift. While paleoecology shows that many species have successfully shifted ranges (Prentice et al., 1991) as well as established new community structure in radically altered environments (Williams, Jackson, 2007), the current rate of climate change is faster than the rate of change after the last glacial period. A reduced gene pool may impede further

131

movement north, while pests, pathogens, and drought cause decline in populations that do not move. Further, even if tree populations are able to track climate change by shifting ranges, the genetic consequences of future range shifts may last hundreds of generations. Lastly, the reduced diversity in these populations may impede species‟ ability to coexist in current and future no-analogue climate communities, as recent work suggests that within species diversity is needed not only for adaptation but also for community coexistence of many competing species (Clark, 2010). This somewhat dismal picture further emphasizes the need for further work to disentangle the effects of ecological change multiple spatial scales. Such investigations will continue to provide insight into the speed with which populations establish equilibrium in their current range, and the time and spatial scales over which drift develops. My work in Chapter 5, by examining natural regeneration and seed dispersal, will complement large-scale studies such as Chapter 4, and provide a connection between range-wide and within-population processes. This chapter demonstrated how habitat, in particular the frequency and type of disturbance, can alter population processes such as dispersal, recruitment, and colonization. My main finding was that certain habitats, in this case upland habitats, may facilitate more frequent colonization and extinction of local populations, while riparian habitats contribute to more stable populations, with local, frequent recruitment. An emphasis on the location and distribution of suitable locations for seedling establishment is not a new direction in ecology (Howe, 1989; Janzen, 1970), but it has only very recently been considered in studies of spatial genetic structure (Sezen et al., 2009). This transition in focus for spatial genetics accompanies a move from studies focused on common, typically high-density

132

species (e.g., Quercus, Fraxinus, (Dutech et al., 2005; Heuertz et al., 2003)) to studies that explore SGS in naturally uncommon trees, as well as species with different modes of seed dispersal (e.g., bird dispersal, Sezen et al., 2009), different population dynamics (e.g., colonization, Pardini and Hamrick, 2008), and integration with human altered landscapes (e.g., Kyndt et al., 2009). As with the previous chapter, one of my goals was to generate results that can improve the predictive value of models of population response to future environmental perturbations (e.g., selective harvest, habitat degradation) at a range of spatial and temporal scales. The effect of habitat on strength and patterns of SGS suggests that population processes such as dispersal and recruitment may rapidly change as the habitat changes, an important consideration in a world with increasingly fragmented and degraded habitats. Studies such as mine not only highlight the population level consequences of habitat degradation but also suggest that habitat management regimes may contribute towards or detract from population stability by altering recruitment and colonization dynamics. An immediate important lesson is that habitat and species management (e.g., restoration, relocation, and reintroduction) must have short-term goals such as preserving current individuals as well as long-term goals such as frequency and location of recruitment. In general, a greater understanding of the interaction between habitat and population processes should enable more appropriate, targeted conservation management efforts. Studies such as mine should also generate a greater understanding of the mechanisms by which small populations persist in dynamic landscapes (Magurran, 2009), which may include colonization and local extinction as a natural process.

133

Allowing local population extinction may seem anathema for forest conservation biologists, but my results suggest that this may be a normal, and unpreventable, aspect of natural history for some organisms. In any case, a greater understanding of population colonization and extinction as a natural process in trees, similar to current knowledge in shorter-lived species (Hanski, Gaggiotti, 2004; Orsini et al., 2008), is clearly needed. An immediate avenue of interest will be to explore potential source-sink dynamics in butternut. The frequent colonization of upland sites by a drought-intolerant species may actually have two explanations. It is possible that upland populations are relatively self-sustaining metapopulations as described in Chapter 5, but it is also possible that they are sink populations that are sustained by frequent dispersal out of a source riparian habitat (Dias, 1996; Vignieri, 2005). In either case, it will be necessary to understand the factors that may lead to local extinction of upland metapopulations. Such work could help predict the consequences of degradation in each habitat type. If compromises must be made in sacrificing some habitats, this information will help inform conservation managers about which decisions are less destructive. My work joins the increasing tendency of SGS studies to integrate ecology, population dynamics, and genetics (Born et al., 2008b). My work suggests that colonization events and co-dispersal of siblings be major determinants of patterns of spatial genetic structure. This is an important departure from previous directions in fine scale spatial genetic studies, in which an isolation-by-distance process was assumed, based on distance between parent and offspring (Fenster et al., 2003; Heuertz et al., 2003; Rousset, 1997). I suggest that when populations exhibit metapopulation dynamics, occur in suboptimal habitat, or have small population size, models of isolation-by-distance

134

based on parent-offspring distances may be insufficient to explain SGS. Future improved models of the development of SGS may incorporate degree of sibling co-dispersal (related to seed dispersal agent and site suitability), frequency and location of recruitment (related to but perhaps not explicitly tied to disturbance regime), density and distribution of adult trees, and population stability/ colonization dynamics. Due to habitat change (Diamond et al., 1989; Kinnison et al., 2007; Meldgaard et al., 2003), the development and maintenance of spatial genetic structure may be a more complex picture than previously assumed (Born et al (2008); Sezen et al (2009). Rare species will likely differ from common species in the mechanisms that shape local distribution of diversity, through population processes such as different recruitment and colonization dynamics. A better understanding of the population biology of rare species, how they differ from common species, and adaptations they have developed to prevent extinction in spite of their rarity, will be crucial to their management. Another avenue of research will be to understand how SGS is reshaped when a once common species becomes rare. Some recent investigations have moved in this research direction (Johnson et al., 2009; Slavov et al., 2009). In any case, both the design and interpretation of SGS studies should carefully distinguish between isolation by distance and colonization/ extinction processes. As with my other studies, in this chapter I emphasized the implementation and evaluation of both traditional and emerging population genetics methods. First, my approach in this study is a new direction in spatial genetic studies. By incorporating a priori, mechanistic hypotheses, I show that natural history can contribute to the entire study, from selection of study sites to determination of the appropriate analysis, as well as

135

interpretation of molecular results. Further, to my knowledge, no intraspecific study of SGS in trees has incorporated enough populations to establish statistical significance of the habitat or process of interest. I suggest that somewhat smaller numbers of samples per population and larger numbers of populations should facilitate new research efforts in the field of spatial genetics, including the statistical support of the influence of various ecological processes. Additionally, I described a huge variance across populations within a species, and I suggest that as studies incorporate more populations, the variation in SGS within a species may frequently exceed the difference observed between some species. I show that a description of the variance complements point estimates and leads to a better understanding of the degree to which environmental factors may interact with natural history and reproductive biology. This variance also suggests that while some populations may have the dispersal capabilities to track climate change, others may not. Lastly, I suggest that knowledge of the dynamics of dispersal and recruitment can be applied to ex-situ conservation sampling of individuals by describing a minimum distance at which sampling will likely exclude highly related individuals. In Chapter 6, I also investigate how habitat can influence dispersal, but in a different way, with regards to the movement of introduced genetic material. One of my main findings was that extensive admixture (proportion of non-native individuals) is only present in highly anthropogenic impacted sites, such as agricultural landscapes. I suggest that hybridization occurs as “bubbles” within suitable habitat rather than an advancing “hybrid front.” My results suggesting the importance of habitat are consistent with previous work in both plant and animal systems, for both

136

natural and anthropogenically induced hybrid zones, but they depart in the explanation of why certain habitats facilitate hybridization and others do not, as I stress the importance of dispersal instead of selection. It is difficult from my results to determine whether hybrids are under positive selection in these environments, or whether they are restricted due to limited dispersal. However, the patterns of chloroplast representation strongly suggest limited effective J. ailantifolia pollen flow into continuous forested and fragmented forested sites. Selection has frequently been described as a mechanism for hybridization associated with particular habitats, such as exploitation of a new resource (Rieseberg et al., 2003). However, I suggest that if dispersal is limiting the advance of hybrids into certain habitats, it is not necessary to invoke selection in butternut, and perhaps dispersal will also be a valid explanation in many hybrid systems. Interestingly, it seems that habitat plays a role both in hybrid zones that have persisted for a long time (natural) and those that are more recent (anthropogenically induced) (Ellstrand, 2003; Lewontin, Birch, 1966). The importance of habitat and direction of introgression has clear implications for the speed of hybrid „invasions‟ and the ultimate outcome for the retention of native genetic material. I suggest, based on my results, that hybrids will likely remain restricted to anthropogenic habitats for some time. I base this conclusion on the fact that J. ailantifolia have existed in substantial numbers in these landscapes for nearly 150 years, and perhaps longer (Anagnostakis, 2009; Neilson, 1930; Sargent, 1894), and have still failed to contribute substantial genetic material to forested areas. It is possible that the dynamics of butternut canker may alter this outcome, by reducing the number of butternuts, and thus the amount of butternut pollen. For instance, if butternuts

137

are completely or mostly eliminated from the native range due to disease, they may or may not be replaced by hybrids. For the present time, however, widespread replacement of butternuts by hybrids is unlikely to be a major concern in forested locations. I did not explore the possibility of temporal variation in selection, however, and a temporally shifting fitness dynamic for hybrids suggests that the factors limiting hybridization may not remain constant. Some proposed conservation programs for butternut include use of J. ailantifolia and hybrids to provide disease tolerance to backcrossed „mostly‟ J. cinerea individuals. This method of repeated backcrossing of the native species to an Asian congeneric species is similar to a program for the American chestnut (Castanea dentata), which was nearly eliminated from the native range by disease (Chestnut blight, Cryphonectria parasitica) in the early 20th century. However, I would caution against this approach, as disease tolerance may be polygenic, giving F1 hybrids an advantage that may dissipate in advanced generations. Further, hybridization may also disrupt the gene complexes that suit butternut to its native habitat, including northern locations where hybrids and J. ailantifolia may not be able to persist. Furthermore, our field work has identified hybrid individuals with heavy disease infections, and others (Anderson, personal communication) have suggested that environmental conditions may be as important as genetics for infection. My results strongly suggest that quantifying rates of hybridization is not sufficient to generate an understanding of hybrid dynamics, nor to allow predictions. While hybridization rates across the range are substantial, the high rates of hybridization are located in specific habitats, and hybrid dynamics over time will be substantially

138

different in different habitats. Management decisions about hybrids can incorporate this information. For example, collection of ex-situ conservation material from anthropogenic landscapes is likely to be of hybrid origin. Additionally, I strongly emphasize the preservation of large natural areas as a potential barrier against hybrid „invasion‟ in this and other systems. The influence of habitat on dispersal may supersede the influence of habitat on selection by preventing hybrids from establishing in the first place. This system presents many future directions of research. An interesting finding was that continuous forested and fragmented forests show very similar rates of hybridization. The low rates of hybridization in both habitats is rather surprising, considering that many of these forests certainly regenerated from previous agricultural land, where J. ailantifolia was likely present in at least some cases. Some of these habitats show evidence of previous farming activities, but in these regenerated forests hybridization rates are still low. The fitness differences of hybrids in various environments may be a partial example, and is a clear avenue of research. While common garden experiments can contribute to this knowledge, measures of hybrid fitness will be most relevant if they are subject to competition, resource fluctuations, temperature changes, and other processes that butternut may be adapted to in forested environments, and field experiments will also be useful. As stated above, a large gap in knowledge is the genetic and environmental basis of tolerance to butternut canker, which is being explored by other butternut researchers. Knowledge of fitness differences may be incorporated into computer population models to generate a range of predicted outcomes

139

of hybridization in this system, and more broadly, to determine whether dispersal or selection is more important in butternut and in other systems. The question of how both habitat and disease interact to change the dynamics of hybrid populations is a largely unexplored area in global change biology, but one with increasing relevance. The combination of epidemic disease, habitat change, and genetically compatible introduced species is a circumstance that is likely to increase in frequency for both plants and animals under future climate warming and anthropogenic habitat change. Additionally, conservation biologists may be increasingly faced with decisions such as whether to incorporate non-native genetic material in native species to provide tolerance to epidemic disease or abiotic changes in the environment. The wellknown detrimental consequences of hybrids (including reduced fitness or increased fitness), will have to be balanced with the possible benefits. A better understanding of the genetic basis of hybrid vigor and any fitness gains will be necessary to evaluate whether these benefits will persist for many generations. Further, currently little is known about the possible cascading ecosystem consequences caused by hybrid replacement of a native species. The loss of keystone native species may substantially alter ecosystems (Schlarbaum, 1997), but the potential ecosystem consequences of loss of rare species is another avenue of conservation research. I will close here with a connection between the various research topics I have approached. My research into several areas of conservation biology demonstrates that butternut is faced with several conservation challenges, as well as opportunities. Indeed, conservation of butternut may be a model example of the complex situation that many threatened species will soon face. Northern populations are faced with both climate

140

change and a reduced gene pool, while southern populations are faced with increasing isolation and population decline. Simultaneously, population processes at a local level may be easily altered by habitat change, as well as unknown influence of hybridization. Perhaps the most immediate challenge is at the local level, to maintain habitat that can facilitate natural population processes such as recruitment. In particular, the potential source-sink dynamics of riparian and upland sites deserves future study, to determine what facilitates local population persistence. The next challenge may be to determine, through mathematical modeling, the size of populations that will be too small before irreversible genetic loss occurs. Both past and present genetic diversity provide benchmarks for this modeling. While fitness consequences of hybridization, particularly with reference to the disease, also contribute to conservation plans, I suggest that this research direction comes after the other two. The spatial restriction of hybrid populations suggests that they are expanding slowly, if at all. Further, large-scale introductions of partly J. cinerea, canker resistant individuals may require greater resources than are currently available. Integration of results from all three chapters also suggests other implications for current global change biology. The habitat mediated limitation of hybrid dispersal (Chapter 6), and the relatively strong spatial genetic structure observed in most butternut populations (Chapter 5) both suggest limited dispersal for this species. Therefore, further movement north during climate change may be difficult for butternut, and many similar species, as dispersal is typically short distance, and dispersal may be further limited in an already occupied, and also changing, habitat. It is certain that movement north will be remarkably different from colonization after glacial retreat, in which conditions (high

141

winds, large scale flooding) may have facilitated long distance movement of propagules into unoccupied habitat. In addition, as shown in Chapter 4, movement may be further restricted by the reduced gene pool observed in northern populations, which reduces species‟ ability to adapt as well as to coexist. This is a bleak picture of the inability for species to colonize northward. However, movement into suitable, local microclimates may be an alternate mechanism by which species can persist (Jaramillo-Correa et al., 2009; Kurt et al., 2009; Robledo-Arnuncio et al., 2005; Vandergast et al., 2007). In my work I have highlighted the utility of a comparative framework to quantify how populations respond to environmental influences. I think one of the most useful aspects of this work is my examination of these processes from an observational rather than a manipulative framework, utilizing natural processes as „experiments in action.‟ The inherent stochasticity of such processes only strengthens the reliability of any conclusions, as the signal must be quite strong to overcome such noise. This emphasizes one of the great strengths of molecular ecology, the use of genetic data to reveal information about ecological processes in situ, including size reductions, population stability and persistence, colonization/ dispersal, recruitment, and migration. My dissertation has explored the connections between the abiotic and biotic environment and the individuals, populations, species, and communities that both inhabit and compose this environment. These interactions may be complex, but may be disentangled with a comparative, hypothesis driven framework. I summarize my main findings as: (1) Holocene-era population fluctuations due to environmental change may exceed those of contemporary population decline, (2) contemporary habitat may alter population dynamics including recruitment and colonization, and (3) habitat may be a

142

potent barrier to gene flow between species. My results stress that molecular ecologists must consider the effects of short and long time scale environmental changes in order to predict future population responses. A greater understanding of the ways in which biogeographic and ecological processes interact across space and time should facilitate conservation management, including prioritization of the most important directions, during a time of both global and local environmental change.

143

APPENDIX A: SUPPLEMENTAL TABLES

TABLE A. 1: PROPORTION OF BOTTLENECKED VS. NON BOTTLENECKED POPULATIONS IN EACH CATEGORY

Contemporary decline Low, high M ratio Heterozygote excess Mode shift

7/10 vs. 6/19 1/10 vs. 5/19 2/10 vs. 5/19

p value 0.943 0.292 0.500

Ecological marginality Core, p value periphery 4/14 vs. 0.092 9/15 3/14 vs. 0.500 4/15 2/14 vs. 0.358 4/15

* significant

144

Range shift North, south 10/13 vs. 3/16 2/13 vs. 4/16 4/13 vs. 3/16

p value 0.003* 0.569 0.376

TABLE A. 2 T-TESTS FOR CATEGORIES UNDER EACH PROCESS.

Contemporary decline

Ecological marginality

Range shift

Variable Number alleles

Low

High

p

Core

Periph

p

North

South

p

9.255

10.254

0.040*

10.253

9.588

0.162

9.042

10.618

0.008*

Allelic richness

7.142

7.585

0.079

7.704

7.178

0.034*

6.941

7.831

0.001*

Rare alleles

4.000

6.895

0.046*

6.143

5.667

0.405

4.077

7.375

0.038*

Heterozygosity

0.810

0.819

0.251

0.822

0.809

0.139

0.800

0.829

0.009*

FST

0.051

0.042

0.129

0.039

0.051

0.035*

0.055

0.038

0.004*

RST

0.054

0.039

0.081

0.037

0.051

0.078

0.058

0.033

0.006*

M-ratio

0.651

0.695

0.029*

0.694

0.667

0.148

0.640

0.713

0.001*

* significant

145

TABLE A. 3 T-TEST PROBABILITIES USING ALTERNATE (100KM) DEFINITION OF THE PERIPHERY.

Model 1- Ecological marginality Periphery Core p Number alleles 9.905 9.922 0.487 Allelic richness 7.390 7.564 0.309 Rare alleles 5.773 6.286 0.364 Heterozygosity 0.813 0.821 0.308 FST 0.046 0.042 0.289 RST 0.043 0.047 0.393 M-ratio 0.681 0.675 0.395

146

TABLE A. 4 T-TEST PROBABILITIES FOR POPULATION CONTRASTS UNDER EACH MODEL, AFTER HOLM‟S CORRECTION.

Contemporary decline

Ecological marginality

Range shift

Low

High

P

Core

Periph

P

Above

Below

P

Number alleles

9.255

10.254

0.240

10.253

9.588

0.324

9.042

10.618

0.024*

Allelic richness

7.142

7.585

0.316

7.704

7.178

0.238

6.941

7.831

0.007*

Rare alleles

4.000

6.895

0.230

6.143

5.667

0.405

4.077

7.375

0.038*

Heterozygosity

0.810

0.819

0.251

0.822

0.809

0.556

0.800

0.829

0.018*

FST

0.051

0.042

0.258

0.039

0.051

0.210

0.055

0.038

0.020*

RST

0.054

0.039

0.243

0.037

0.051

0.390

0.058

0.033

0.024*

M-ratio

0.651

0.695

0.203

0.694

0.667

0.444

0.640

0.713

0.006*

*significant

147

TABLE A. 5 T-TESTS FOR CATEGORIES OF EACH PROCESS, AFTER REMOVAL OF THE DISJUNCT POPULATIONS, 27-29

Contemporary decline

Ecological marginality

Range shift

Low

High

P

Core

Periph

P

Above

Below

P

Number alleles

9.571

10.254

0.121

10.253

9.856

0.302

9.200

10.614

0.013*

Allelic richness

7.501

7.585

0.385

7.704

7.397

0.140

7.132

7.831

0.012*

Rare alleles

4.286

6.895

0.092

6.143

6.250

0.482

4.300

7.375

0.056

Heterozygosity

0.822

0.818

0.360

0.822

0.816

0.302

0.805

0.828

0.047*

FST

0.042

0.042

0.496

0.039

0.046

0.140

0.050

0.038

0.030*

RST

0.047

0.039

0.271

0.037

0.046

0.189

0.054

0.033

0.034*

M-ratio

0.672

0.695

0.142

0.694

0.684

0.354

0.651

0.713

0.005*

*significant

148

TABLE A. 6 MINIMAL ADEQUATE MODEL TO EXPLAIN NUMBER OF ALLELES, ALLELIC RICHNESS, HETEROZYGOSITY, M, RST, FST, AND NUMBER OF RARE ALLELES.

Number alleles (Intercept) Dice R2: 0.163 Allelic richness (Intercept) Dedge Dice Tdisease R2: 0.490 Rare alleles (Intercept) Dice R2: 0.096 Heterozygosity (Intercept) Dice Dedge Tdisease R2: 0.373 FST (Intercept) Dice Dedge Tdisease R2: 0.503 RST (Intercept) Dice R2: 0.1886 M-ratio (Intercept) Dice Tdisease R2: 0.353

149

Estimate 1.00E+01 1.88E-03

p <0.001 0.030

Estimate -4.97E+01 1.51E-03 2.39E-03 2.92E-02

p 0.020 0.089 <0.001 0.008

Estimate 6.10E+00 4.25E-03

p <0.001 0.102

Estimate -1.71E+00 8.85E-05 6.79E-05 1.29E-03

p 0.071 0.004 0.091 0.010

Estimate 1.36E+00 -5.12E-05 -3.93E-05 -6.73E-04

p 0.004 <0.001 0.039 0.005

Estimate 4.28E-02 -3.02E-05

p <0.001 0.019

Estimate -2.43E+00 1.82E-04 1.59E-03

p 0.199 0.005 0.103

TABLE A. 7 MINIMAL ADEQUATE MODELS AFTER HOLM‟S CORRECTION.

Number alleles (Intercept) Dice R2: 0.163 Allelic richness (Intercept) Dedge Dice Tdisease R2: 0.490 Rare alleles (Intercept) Dice R2: 0.096 Heterozygosity (Intercept) Dice Dedge Tdisease R2: 0.373 FST (Intercept) Dice Dedge Tdisease R2: 0.503 RST (Intercept) Dice R2: 0.1886 M-ratio (Intercept) Dice Tdisease R2: 0.353

150

Estimate 1.00E+01 1.88E-03

p <0.001 0.240

Estimate -4.97E+01 1.51E-03 2.39E-03 2.92E-02

p 0.18 0.445 0.011 0.096

Estimate 6.10E+00 4.25E-03

p <0.001 0.306

Estimate -1.71E+00 8.85E-05 6.79E-05 1.29E-03

p 0.426 0.064 0.364 0.110

Estimate 1.36E+00 -5.12E-05 -3.93E-05 -6.73E-04

p 0.060 0.010 0.273 0.070

Estimate 4.28E-02 -3.02E-05

p <0.001 0.190

Estimate -2.43E+00 1.82E-04 1.59E-03

p 0.199 0.065 0.206

TABLE A. 8 MINIMAL ADEQUATE MODELS, WITH NEW BRUNSWICK POPULATIONS REMOVED.

Number alleles (Intercept) Dice R2: 0.100 Allelic richness (Intercept) Dedge Dice Tdisease R2: 0.357 Rare alleles (Intercept) Dedge Dice Dedge: Dice R2: 0.285 He (Intercept) Dedge Dice Tdisease R2: 0.311 FST (Intercept) Dedge Dice Tdisease R2: 0.365 RST (Intercept) Dice R2: 0.081 M-ratio (Intercept) Dice Tdisease R2: 0.232

151

Estimate 9.974 0.002

p <0.001 0.116

Estimate -66.7 0.003 0.003 0.038

p 0.014 0.053 0.002 0.007

Estimate 6.704 -0.003 0.014 <0.001

p <0.001 0.704 0.011 0.023

Estimate -2.463 <0.001 <0.001 0.002

p 0.037 0.049 0.005 0.007

Estimate 1.677 <0.001 <0.001 -0.001

p 0.004 0.032 0.002 0.005

Estimate 0.042 <0.001

p <0.001 0.159

Estimate -2.538 <0.001 0.002

p 0.217 0.020 0.121

TABLE A. 9 COEFFICIENTS OF DETERMINATION AND P VALUES FOR ALL LINEAR MODELS.

Variable Number alleles Allelic richness Rare alleles Heterozygosity FST RST M-ratio * significant

Contemporary decline 2 R p 0.080 0.139 0.104 0.088 0.069 0.169 0.029 0.376 0.086 0.122 0.156 0.100 0.117 0.070

Ecological marginality R2 p 0.027 0.394 0.169 0.027* 0.002 0.802 0.115 0.071 0.191 0.018* 0.068 0.173 0.067 0.176

152

Range shift R2 0.163 0.305 0.096 0.156 0.278 0.188 0.281

p 0.030* 0.002* 0.102 0.034* 0.003* 0.019* 0.003*

TABLE A. 10 COEFFICIENTS OF DETERMINATION AND P VALUES FOR ALL LINEAR MODELS AFTER HOLM‟S CORRECTION.

Contemporary decline 2 Variable R p Number alleles 0.080 0.417 Allelic richness 0.104 0.528 Rare alleles 0.069 0.338 Heterozygosity 0.029 0.376 FST 0.086 0.488 RST 0.156 0.500 M-ratio 0.117 0.49

Ecological marginality R2 p 0.027 0.788 0.169 0.162 0.002 0.802 0.115 0.355 0.191 0.126 0.068 0.692 0.067 0.528

*significant

153

Range shifts R p 0.163 0.090 0.305 0.014* 0.096 0.102 0.156 0.068 0.278 0.018* 0.188 0.076 0.281 0.015* 2

TABLE A. 11 GESTE BAYESIAN MODEL COMPARISON. CUMULATIVE PROBABILITIES OVER ALL MODELS (LEFT) AND COEFFICIENTS OF SINGLE HIGHEST PROBABILITY MODEL (RIGHT) FOR SEVEN AND FOR FIVE PREDICTOR VARAIBLES, FOR THE RESPONSE VARIABLE FST.

155 1

Predictor variables sample size

Pr1 0.037

α0 (Constant)

mean2 -2.89

connectivity

0.144

α6 (Dice)

-0.337

latitude longitude Dedge Dice Tdisease

0.306 0.078 0.187 0.433 0.114

σ²

0.356

mode2 95% CI -2.88 [-3.12 ; -2.66] [-0.572 ; -0.327 0.119] 0.310 0.179 ; 0.571

Predictor variables connectivity latitude Dedge Dice Tdisease

Pr 0.152 0.410 0.192 0.358 0.113

mean2 -2.88 0.318 0.369

mode2 -2.89 0.331 0.311

α0 (Constant) α2 (latitude) σ²

95% CI [-3.11 , -2.65] [0.075, 0.552] [0.188, 0.582]

In Bayesian statistics, the highest probability is most likely, and can be compared to lesser probabilities using a Bayes‟ factor, or the ratio of the probabilities of the models. 22 This is the mean and the mode of the coefficients of the highest probability model, across the 50,000 MCMC runs.

TABLE A. 12 RESULTS OF MANTEL TEST ON ALL POPULATIONS, AND FIVE POPULATION SUBSETS.

Set of populations All populations Continuous range South of ice boundary Core Low impact of disease Main BAPS cluster

FST R2 0.128 0.069 0.072 0.067 0.132 0.079

155

p <0.001 <0.001 <0.001 0.005 0.008 0.001

RST R2 0.067 0.084 0.098 0.032 0.041 0.062

p <0.001 <0.001 0.001 0.046 0.085 0.001

BIBLIOGRAPHY

Abuín M, Díaz R, Alfonsín J, et al. (2004) PCR-RFLP Analysis of cpDNA in the Genus Juglans. In: V International Walnut Symposium. Acta Horticulturae, Sorrento, Italy. Adams JM, Piovesan G, Strauss S, Brown S (2002) The case for genetic engineering of native and landscape trees against introduced pests and diseases. Conservation Biology 16, 874-879. Ahmed S, Compton SG, Butlin RK, Gilmartin PM (2009) Wind-borne insects mediate directional pollen transfer between desert fig trees 160 kilometers apart. PNAS 106, 20342-20347. Allendorf FW, Leary RF, Spruell P, Wenburg JK (2001) The problems with hybrids: setting conservation guidelines. TRENDS In Ecology and Evolution 16. Allendorf FW, Ryman N (2002) The role of genetics in population viability analysis. In: Population Viability Analysis (eds. Beissinger SR, McCullough DR). University of Chicago Press, Chicago. Altschul SF, Madden TL, Schaffer AA, et al. (1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Research 25, 3389-3402. Anagnostakis SA (2009) The butternut that is not a butternut. Connecticut Woodlands 73, 30-31. Anderson E (1948) Hybridization of the habitat. Evolution 2, 1-9. Anderson EC (2003) Users guide to the program NewHybrids v 1.1. Anderson EC, Thompson EA (2002) A model-based method for identifying species hybrids using multilocus genetic data. Genetics 160, 1217-1229. Anderson RL (1996) Butternut canker. In: Proceedings, Southern Appalachia biological control initiative workshop, pp. 26- 27, Asheville, NC. Anderson RL, LeMadeleine LA (1978) The distribution of butternut decline in the eastern United States. USDA, Forest Service, NE Area, State and Private Forestry Report S-3-78, 4.

156

Aradhya MK, Potter D, Simon CJ (2004) Origin, Evolution, and Biogeography of Juglans: A Phylogenetic Perspective. In: V International Walnut Symposium. Acta Horticulturae, Sorrento, Italy. Archie EA, Maldonado JE, Hollister-Smith JA, et al. (2008) Fine-scale population genetic structure in a fission–fusion society. Molecular Ecology 17. Ashworth FL (1969) Butternuts, siebold (Japanese) walnuts, and their hybrids. In: Handbook of North American nut trees (ed. Jaynes RA), pp. 224-231. The Northern Nut Growers Association, Knoxville, TN. Auble GT (1995) Riparian ecosystems. In: Our living resources : a report to the nation on the distribution, abundance, and health of U.S. plants, animals, and ecosystems (ed. LaRoe ET). U.S. Dept. of the Interior, National Biological Service, Washington, DC. Austerlitz F, Garnier-Gere PH (2003) Modeling the impact of colonisation on genetic diversity and differentiation of forest trees: interaction of life cycle, pollen flow and seed long-distance dispersal. Heredity 90, 282. Bizoux J-P, DaÏNou K, Bourland N, et al. (2009) Spatial genetic structure in Milicia excelsa (Moraceae) indicates extensive gene dispersal in a low-density windpollinated tropical tree. Molecular Ecology 18, 4398-4408. Bomblies K, Lempe J, Epple P, et al. (2007) Autoimmune response as a mechanism for a Dobzhansky-Muller-type incompatibility syndrome in plants. PLoS Biology 5, e236. Born C, Hardy OJ, Chevallier M-H, et al. (2008a) Small-scale spatial genetic structure in the Central African rainforest tree species Aucoumea klaineana: a stepwise approach to infer the impact of limited gene dispersal, population history and habitat fragmentation. Molecular Ecology 17, 2041-2050. Born C, Kjellberg F, Chevallier M-H, et al. (2008b) Colonization processes and the maintenance of genetic diversity: insights from a pioneer rainforest tree, Aucoumea klaineana. Proceedings of the Royal Society B: Biological Sciences 275, 2171-2179. Boysen B (2009) Butternut: an urban species at risk? Ontario Arborist January/ February Bradbury IR, Bentzen P (2007) Non-linear genetic isolation by distance: implications for dispersal estimation in anadromous and marine fish populations. Marine Ecology Progress Series 340, 245-257. Brede N, Sandrock C, Straile D, et al. (2009) The impact of human-made ecological changes on the genetic architecture of Daphnia species. PNAS 106, 4758-4763. 157

Brockerhoff EG, Liebhold AM, Jactel H (2006) The ecology of forest insect invasions and advances in their management. Canadian Journal of Forest Research 36. Brown JW, Groot PJVCD, Birt TP, et al. (2007) Appraisal of the consequences of the DDT-induced bottleneck on the level and geographic distribution of neutral genetic variation in Canadian peregrine falcons, Falco peregrinus. Molecular Ecology 16, 327-343. Bruckerhoff DN (2005) Improving black walnut stands. Kansas State University Agricultural Experiment Station Publication L-718. Burgess KS, Morgan M, Deverno L, Husband BC (2005) Asymmetrical introgression between two Morus species (M. alba, M. rubra) that differ in abundance. Molecular Ecology 14, 3471-3483. Burke JM, Arnold ML (2001) Genetics and the fitness of hybrids. Annual Review of Genetics 35, 31-52. Campbell LG, Snow AA, Ridley CE (2006) Weed evolution after crop gene introgression: greater survival and fecundity of hybrids in a new environment. Ecology Letters 9, 1198-1209. Canham C, Denslow J, Platt W, et al. (1990) Light regimes beneath closed canopies and tree-fall gaps in temperate and tropical forests. Canadian Journal of Forest Research 20. Cassel A, Tammaru T (2003) Allozyme variability in central, peripheral and isolated populations of the scarce heath (Coenonympha hero: Lepidoptera, Nymphalidae); implications for conservation. Conservation Genetics 4, 83-93. Casteele TVD, Galbusera P, Matthysen E (2001) A comparison of microsatellite-based pairwise relatedness estimators. Molecular Ecology 10, 1539-1549. Cavers S, Degen B, Caron H, et al. (2005) Optimal sampling strategy for estimation of spatial genetic structure in tree populations. Heredity 95, 281-289. Clark James S (1998) Why trees migrate so fast: Confronting theory with dispersal biology and the paleorecord. The American Naturalist 152, 204-224. Clark JS (2010) Individuals and the Variation Needed for High Species Diversity in Forest Trees. Science 327, 1129-1132. Clark SL, Brosi SL, Schlarbaum SE, Grissino-Mayer HD (2008) Dendrochronology of two butternut (Juglans cinerea) populations in the southeastern United States. Forest Ecology and Management 255, 1772-1780. Corander J, Waldmann P, Sillanpaa MJ (2003) Bayesian analysis of genetic differentiation between populations. Genetics 163, 367 - 374. 158

Cottrell JE, Munro RC, Tabbener HE, et al. (2003) Comparison of fine-scale genetic structure using nuclear microsatellites within two British oakwoods differing in population history. Forest Ecology and Management 176, 287-303. Coyne JA, Orr HA (2004) Speciation Sinauer Associates, Sunderland, MA. Craft KJ, Ashley MV (2007) Landscape genetic structure of bur oak (Quercus macrocarpa) savannas in Illinois. Forest Ecology and Management 239, 13-20. Craft KJ, Ashley MV, Koenig WD (2002) Limited hybridization between Quercus lobata and Quercus douglasii (Fagaceae) in a mixed stand in central coastal California. American Journal of Botany 89, 1792–1798. Currat M, Ruedi M, Petit RmJ, Excoffier L, Wolf J (2009) The Hidden Side of Invasions: Massive Introgression by Local Genes. Evolution 62, 1908-1920. Darwin C (1859) On the Origin of Species Murray's, London. Davis C (1966) The Butternut. Annual Report of the Northern Nut Growers Association 57, 101- 102. Davis MB (1983) Quaternary history of deciduous forests of Eastern North America and Europe. Annals of the Missouri Botanical Garden 70, 550-563. Diamond JM, Ashmole NP, Purves PE (1989) The Present, Past and Future of HumanCaused Extinctions [and Discussion]. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 325, 469-477. Dias PC (1996) Sources and sinks in population biology. Trends in Ecology & Evolution 11, 326-330. Dick C, Hardy O, Jones F, Petit R (2008) Spatial Scales of Pollen and Seed-Mediated Gene Flow in Tropical Rain Forest Trees. Tropical Plant Biology 1, 20-33. Doligez A, Baril C, Joly HI (1998) Fine-Scale Spatial Genetic Structure with Nonuniform Distribution of Individuals. Genetics 148, 905-920. Driesche RV, Lyon S, Blossey B, Hoddle M, Reardon R (2002) Biological Control of Invasive Plants in the Eastern United States, USDA Forest Service Publication FHTET-2002-04. Duncan RS, Chapman CA (1999) Seed Dispersal and Potential Forest Succession in Abandoned Agriculture in Tropical Africa. Ecological Applications 9, 998-1008. Dutech C, Sork VL, Irwin AJ, Smouse PE, Davis FW (2005) Gene flow and fine-scale genetic structure in a wind-pollinated tree species, Quercus lobata (Fagaceaee). Am. J. Bot. 92, 252-261.

159

Eckert CG, Samis KE, Lougheed SC (2008) Genetic variation across species' geographical ranges: the central-marginal hypothesis and beyond. Molecular Ecology 17, 1170-1188. Ellstrand NC (2003) Current Knowledge of Gene Flow in Plants: Implications for Transgene Flow. Philosophical Transactions: Biological Sciences 358, 11631170. Endler JA (1977) Geographic variation, speciation, and clines. Princeton University Press., Princeton, NJ. England PR, Osler GHR, Woodworth LM, et al. (2003) Effects of intense versus diffuse population bottlenecks on microsatellite genetic diversity and evolutionary potential. Conservation Genetics 4, 595-604. Epperson BK, Li T-Q (1997) Gene dispersal and spatial genetic structure. Evolution 51. Excoffier L, Foll M, Petit R (2009) Genetic consequences of range expansions. Annual Review of Ecology, Evolution, and Systematics 40, 481-501. 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. Facon B, Jarne P, Pointier JP, David P (2005) Hybridization and invasiveness in the freshwater snail Melanoides tuberculata: hybrid vigour is more important than increase in genetic variance. Journal of Evolutionary Biology 18, 524-535. Faubet P, Gaggiotti OE (2008) A new Bayesian method to identify the environmental factors that influence recent migration. Genetics, genetics.107.082560. Fenster CB, Vekemans X, Hardy OJ (2003) Quantifying Gene Flow from Spatial Genetic Structure Data in a Metapopulation of Chamaecrista fasciculata (Leguminosae). Evolution 57, 995-1007. Fernández-M. JF, Sork VL (2007) Genetic Variation in Fragmented Forest Stands of the Andean Oak Quercus humboldtii Bonpl. (Fagaceae). Biotropica 39, 72-78. Fischer M, Matthies D (1997) Mating structure and inbreeding and outbreeding depression in the rare plant Gentianella germanica (Gentianaceae). Am. J. Bot. 84, 1685-. Fitzpatrick BM, Shaffer HB (2004) Environment-dependent admixture dynamics in a tiger salamander hybrid zone. Evolution 58. Fitzpatrick BM, Shaffer HB (2007a) Hybrid vigor between native and introduced salamanders raises new challenges for conservation. PNAS 104.

160

Fitzpatrick BM, Shaffer HB (2007b) Introduction history and habitat variation explain the landscape genetics of hybrid salamanders. Ecological Applications 17, 598-608. Fjellstrom R, Parfitt D (1994) Walnut (Juglans spp.) genetic diversity determined by restriction fragment length polymorphisms. Genome 37. Fjellstrom RG, Parfitt DE (1995) Phylogenetic analysis and evolution of the genus Juglans (Juglandaceae) as determined from nuclear genome RFLPs. Plant Systematics and Evolution 197, 19-32. Fleguel RV (1996a) A Literature Review of Butternut and the Butternut Canker Ontario Ministry of Natural Resources, Eastern Ontario Model Forest, Information Report No. 20. Fleguel RV (1996b) A literature review of butternut and the butternut cankerInformation Report No. 20. Ontario Ministry of Natural Resources, Eastern Ontario Model Forest, Kemptville, Ontario, Canada. Fleguel RV (1996c) A literature review of butternut and the butternut cankerInformation Report No. 20. Ontario Ministry of Natural Resources, Eastern Ontario Model Forest. Foll M, Gaggiotti O (2006) Identifying the environmental factors that determine the genetic structure of populations. Genetics 174, 875-891. Fralish JS (2002) The central hardwood forest: its boundaries and physiographic provinces. In: 13th Central Hardwood Forest conference eds. Van Sambeek JW, Dawson JO, Ponder F, Jr., Loewenstein EF, Fralish JS). USDA: Forest Service, Urbana, IL. Frankham R (1995) Conservation Genetics. Annual Review of Genetics 29, 305-327. Frankham R (2005) Genetics and extinction. Biological Conservation 126, 131-140. Frankham R, Ballou J, Briscoe D (2002) Introduction to Conservation Genetics Cambridge University Press. Franklin JF, Spies TA, Pelt RV, et al. (2002) Disturbances and structural development of natural forest ecosystems with silvicultural implications, using Douglas-fir forests as an example. Forest Ecology and Management 155, 399-423. Fraser DJ, Bernatchez L (2001) Adaptive evolutionary conservation: towards a unified concept for defining conservation units. Molecular Ecology 10, 2741-2752. Fritz RS (1999) Resistance of hybrid plants to herbivores: Genes, environment or both? Ecology 80, 382-391. Gaggiotti OE, Bekkevold D, Jørgensen HBH, et al. (2009) Disentangling the effects of evolutionary, demographic, and environmental factors influencing genetic

161

structure of natural populations: Atlantic herring as a case study. Evolution 63, 2939-2951. Gamache I, Jaramillo-Correa JP, Payette S, Bousquet J (2003) Diverging patterns of mitochondrial and nuclear DNA diversity in subarctic black spruce: imprint of a founder effect associated with postglacial colonization. Molecular Ecology 12, 891-901. Gapare WJ, Aitken SN, Ritland CE (2005) Genetic diversity of core and peripheral Sitka spruce (Picea sitchensis (Bong.) Carr) populations: implications for conservation of widespread species. Biological Conservation 123, 113-123. Garza JC, Williamson EG (2001) Detection of reduction in population size using data from microsatellite loci. Molecular Ecology 10, 305-318. Gaskin JF, Wheeler GS, Purcell MF, Taylor GS (2009) Molecular evidence of hybridization in Florida's sheoak (Casuarina spp.) invasion. Molecular Ecology 18, 3216-3226. Gaston KJ (1996) Species-range-size distributions: patterns, mechanisms and implications. Trends in Ecology & Evolution 11, 197-201. Gellatly JU (1966) Heartnuts - outstanding selections and some of their best hybrids. Annual Report of the Northern Nut Growers Association 57, 103- 110. González-Martínez SC, Krutovsky KV, Neale DB (2006) Forest-tree population genomics and adaptive evolution. New Phytologist 170, 227-238. Goudet J (1995) Fstat: a computer program to calculate F-statistics. Journal of Heredity 86, 485-486. Greenberg CH, McNab WH (1998) Forest disturbance in hurricane-related downbursts in the Appalachian mountains of North Carolina. Forest Ecology and Management 104, 179-191. Grivet D, Sork VL, Westfall RD, Davis FW (2008) Conserving the evolutionary potential of California valley oak (Quercus lobata): a multivariate genetic approach to conservation planning. Molecular Ecology 17, 139-156. Guillot G, Leblois R, Coulon A, Frantz AC (2009) Statistical methods in spatial genetics. Molecular Ecology 18, 4734-4756. Gunnell K, Tada M, Hawthorne F, Keeley E, Ptacek M (2008) Geographic patterns of introgressive hybridization between native Yellowstone cutthroat trout (Oncorhynchus clarkii bouvieri) and introduced rainbow trout (O. mykiss) in the South Fork of the Snake River watershed, Idaho. Conservation Genetics 9, 49-64.

162

Hails RS, Morley K (2005) Genes invading new populations: a risk assessment perspective. TRENDS In Ecology and Evolution 20, 245-252. Hain F (2006) New Threats to Forest Health Require Quick and Comprehensive Research Response. Journal of Forestry 104, 182-186. Halpern CB (1989) Early Successional Patterns of Forest Species: Interactions of Life History Traits and Disturbance. Ecology 70, 704-720. Hampe A, Masri LE, Petit RJ (2010) Origin of spatial genetic structure in an expanding oak population. Molecular Ecology 19, 459-471. Hamrick JL (2004) Response of forest trees to global environmental changes. Forest Ecology and Management 197, 323-335. Hamzeh M, Sawchyn C, Perinet P, Dayanandan S (2007) Asymmetrical natural hybridization between Populus deltoides and P. balsamifera (Salicaceae). Canadian Journal of Botany 85, 1227-1232. Hanski I, Gaggiotti OE (2004) Ecology, Genetics and Evolution of Metapopulations Academic Press. Hardy OJ (2009) Direct monitoring of long-distance pollen and seed dispersals: How fat is the tail? Heredity 103, 437-438. Hardy OJ, Maggia L, Bandou E, et al. (2006) Fine-scale genetic structure and gene dispersal inferences in 10 Neotropical tree species. Molecular Ecology 15, 559571. Hardy OJ, Vekemans X (2002) spagedi: a versatile computer program to analyse spatial genetic structure at the individual or population levels. Molecular Ecology Notes 2, 618-620. Hardy OJ, Vekemans X (2009) User's Manual: Spagedi 1.3: a program for spatial pattern analysis of genetic diversity. Haygood R, Ives AR, Andow DA (2003) Consequences of recurrent gene flow from crops to wild relatives. Proceedings: Biological Sciences 270, 1879-1886. He S-A, Santamour FS, Jr. (1983) Isoenzyme Verification of American-Chinese Hybrids of Liquidambar and Liriodendron. Annals of the Missouri Botanical Garden 70, 748-749. Henry P, Miquelle D, Sugimoto T, et al. (2009) In situpopulation structure and ex situ representation of the endangered Amur tiger. Molecular Ecology 18, 3173-3184. Heuertz M, Hausman J-F, Hardy OJ, et al. (2004) Nuclear microsatellites reveal contrasting patterns of genetic structure between western and southeastern European populations of the common ash (Fraxinus excelsior L.). Evolution 58, 976-988.

163

Heuertz M, Vekemans X, Hausman JF, Palada M, Hardy OJ (2003) Estimating seed vs. pollen dispersal from spatial genetic structure in the common ash. Molecular Ecology 12, 2483-2495. Hewitt G (2000) The genetic legacy of the Quaternary ice ages. Nature 405, 907-913. Hoban S, Anderson R, Mccleary T, Schlarbaum S, Romero-Severson J (2008) Thirteen nuclear microsatellite loci for butternut (Juglans cinerea L.). Molecular Ecology Resources 8, 643-646. Hoban S, McCleary, T., Schlarbaum, S., Romero-Severson, J. (2009) Geographically extensive hybridization between the forest trees American butternut and Japanese walnut. Biology Letters 5, 324-327. Hochwender CG, Fritz RS (1999) Fluctuating asymmetry in a Salix hybrid system: the importance of genetic versus environmental causes. Evolution 53, 408-416. Holm S (1979) A Simple Sequentially Rejective Multiple Test Procedure. Scandinavian Journal of Statistics 6, 65-70. Hoover MH (1919) Farms by the Side of the Road. In: Northern Nut Growers Report of the Proceedings of the 10th Annual Meeting pp. 23- 32, Battle Creek Michigan. Hopkin A, Innes L, Harrison K (2001) Distribution of butternut canker (Sirococcus clavigignenti-juglandacearum) in Eastern Canada. In: Frontline Express. Canadian Forest Service, Great Lakes Forestry Centre, Saul Ste. Marie, Ontario. Howe HF (1989) Scatter-and clump-dispersal and seedling demography: hypothesis and implications. Oecologia 79, 417-426. Hu L-J, Uchiyama K, Shen H-L, Ide Y (2010) Multiple-scaled spatial genetic structures of Fraxinus mandshurica over a riparian–mountain landscape in Northeast China. Conservation Genetics 11, 77-87. Hutchison DW (2003) Testing the central/peripheral model: Analyses of microsatellite variability in the Eastern Collared Lizard (Crotaphytus collaris). American Midland Naturalist 149, 148-162. Hutchison DW, Templeton AR (1999) Correlation of pairwise genetic and geographic distance measures: Inferring the relative influences of gene flow and drift on the distribution of genetic variability. Evolution 53, 1898-1914. Jansen PA, Forget PM (2001) Scatterhoarding and tree regeneration. In: Nouragues: dynamics and plant–animal interactions in a Neotropical rainforest. (eds. Bongers F, Charles-Dominique P, Forget PM, The´ry M). Kluwer Academic Publishers, Dordrecht, The Netherlands. Janzen DH (1970) Herbivores and the Number of Tree Species in Tropical Forests. The American Naturalist 104, 501.

164

Jaramillo-Correa JP, Beaulieu J, Khasa DP, Bousquet J (2009) Inferring the past from the present phylogeographic structure of North American forest trees: seeing the forest for the genes. Canadian Journal of Forest Research 39, 286-307. Jaynes RA (1979) Chestnuts. In: Nut Tree Culture in North America (ed. Jaynes RA), pp. 111-147. Northern Nut Growers Association, Hamden, CT. Johansen-Morris AD, Latta RG (2008) Genotype by environment interactions for fitness in hybrid genotypes of Avena barbata. Evolution 62, 573-585. Johansson M, Primmer CR, Merila J (2006) History vs. current demography: explaining the genetic population structure of the common frog (Rana temporaria). Molecular Ecology 15, 975-983. Jones AG, Stockwell CA, Walker D, Avise JC (1998) The molecular basis of a microsatellite null allele from the white sands pupfish. Journal Of Heredity 89, 339- 342. Jones FA, Hamrick JL, Peterson CJ, Squiers ER (2006) Inferring colonization history from analyses of spatial genetic structure within populations of Pinus strobus and Quercus rubra. Molecular Ecology 15, 851-861. Jones FA, Hubbell SP (2006) Demographic spatial genetic structure of the Neotropical tree, Jacaranda copaia. Molecular Ecology 15, 3205-3217. Jones KC, Levine KF, Banks JD (2002) Characterization of 11 polymorphic tetranucleotide microsatellites for forensic applications in California elk (Cervus elaphus canadensis). Molecular Ecology Notes 2, 425-427. Jordan M, Morris D, Gibson S (2008) The influence of historical landscape change on genetic variation and population structure of a terrestrial salamander (Plethodon cinereus). Conservation Genetics. Jump AS, Penuelas J (2006) Genetic effects of chronic habitat fragmentation in a windpollinated tree. PNAS 103, 8096-8100. Kalinowski S, Taper M, Marshall T (2007) Revising how the computer program CERVUS accommodates genotyping error increases success in paternity assignment. Molecular Ecology 16, 1099-1106. Kimura M, Weiss GH (1964) The stepping stone model of population structure and the decrease of genetic correlation with distance. Genetics 49, 561-576. Kinnison M, Hendry A, Stockwell C (2007) Contemporary evolution meets conservation biology II: impediments to integration and application. Ecological Research 22, 947-954.

165

Kirkpatrick M, Barton NH (1997) Evolution of a Species' Range. The American Naturalist 150, 1-23. Klapproth J, Johnson J (2009) Understanding the science behind riparian forest buffers. Virginia Cooperative Extension Publication 420-152. Knowles LL, Carstens BC, Keat Marcia L (2007) Coupling genetic and ecological-niche models to examine how past population distributions contribute to divergence. Current Biology 17, 940-946. Kurt EG, David JH, Kelly RZ (2009) When cold is better: climate-driven elevation shifts yield complex patterns of diversification and demography in an alpine specialist (American pika, ochotona princeps). Evolution 63, 2848-2863. Kyndt T, Assogbadjo AE, Hardy OJ, et al. (2009) Spatial genetic structuring of baobab (Adansonia digitata, Malvaceae) in the traditional agroforestry systems of West Africa. Am. J. Bot. 96, 950-957. Landry CR, Hartl DL, Ranz JM (2007) Genome clashes in hybrids: insights from gene expression. Heredity 99, 483-493. Latch EK, Dharmarajan G, Glaubitz JC, Rhodes OE (2006) Relative performance of Bayesian clustering software for inferring population substructure and individual assignment at low levels of population differentiation. Conservation Genetics 7, 295 - 302. Lawton JH (1993) Range, population abundance, and conservation. Trends in Ecology & Evolution 8, 409-413. Ledig FT (1988) The conservation of diversity in forest trees: Why and how should genes be conserved? BioScience 38, 471-479. Legendre P, Fortin M-J (1989) Spatial Pattern and Ecological Analysis. Vegetatio 80, 107-138. Lesica P, Allendorf FW (1995) When Are Peripheral Populations Valuable for Conservation? Conservation Biology 9, 753-760. Levin DA (1981) Dispersal Versus Gene Flow in Plants. Annals of the Missouri Botanical Garden 68, 233-253. Levin DA (1988) Consequences of Stochastic Elements in Plant Migration. The American Naturalist 132, 643-651. Levin DA (2002) Hybridization and extinction: in protecting rare species, conservationists should consider the dangers of interbreeding, which compound the more well-known threats to wildlife. American Scientist 90, 254-261. Levin DA, Francisco-Ortega J, Jansen RK (1996) Hybridization and the Extinction of Rare Plant Species. Conservation Biology 10, 10-16. Lewis PO, Zaykin D (2001) Genetic data analysis: computer program for the analysis of allelic data.

166

Lewontin RC, Birch LC (1966) Hybridization as a Source of Variation for Adaptation to New Environments. Evolution 20, 315-336. Lexer C, Buerkle CA, Joseph JA, Heinze B, Fay MF (2006) Admixture in European Populus hybrid zones makes feasible the mapping of loci that contribute to reproductive isolation and trait differences. Heredity 98, 74-84. Lexer C, Fay MF, Joseph JA, Nica MS, Heinze B (2005) Barrier to gene flow between two ecologically divergent Populus species, P. alba (white poplar) and P. tremula (European aspen): the role of ecology and life history in gene introgression. Molecular Ecology 14, 1045-1057. Liepelt S, Bialozyt R, Ziegenhagen B (2002) Wind-dispersed pollen mediates postglacial gene flow among refugia. PNAS 99, 14590-14594. Little EL, Jr. (1971) Atlas of United States trees, volume 1, conifers and important hardwoods: UDA Miscellaneous Publication 1146, 9 p., 200 maps. Loiselle BA, Sork VL, Nason J, Graham C (1995) Spatial Genetic Structure of a Tropical Understory Shrub, Psychotria officinalis (Rubiaceae). American Journal of Botany 82, 1420-1425. Lönn M, Prentice HC (2002) Gene diversity and demographic turnover in central and peripheral populations of the perennial herb Gypsophila fastigiata. Oikos 99, 489498. Loveless MD, Hamrick JL (1984) Ecological determinants of genetic structure in plant populations. Annual Review of Ecology and Systematics 15, 65-95. Lowe A, Harris S, Ashton P (2004) Ecological genetics: Design, analysis and application Blackwell Publishing Ltd., Malden, MA. Luikart G, Allendorf FW, Cornuet J-M, Sherwin WB (1998) Distortion of allele frequency distributions provides a test for recent population bottlenecks. Journal Of Heredity 89, 238-247. Luikart G, Cornuet J-M (1998) Empirical evaluation of a test for identifying recently bottlenecked populations from allele frequency data. Conservation Biology 12, 228-237. Lynch M, Ritland K (1999) Estimation of Pairwise Relatedness With Molecular Markers. Genetics 152, 1753-1766. Lynch M, Walsh B (1998) Genetics and analysis of quantitative traits Sinauer Associates. Magni CR, Ducousso A, Caron H, Petit RJ, Kremer A (2005) Chloroplast DNA variation of Quercus rubra L. in North American and comparison with other Fagaceae. Molecular Ecology 14. 167

Magurran AE (2009) Threats to Freshwater Fish. Science 325, 1215-1216. Mallet J (2005) Hybridization as an invasion of the genome. TRENDS In Ecology and Evolution 20, 229-237. Manel S, Schwartz MK, Luikart G, Taberlet P (2003) Landscape genetics: combining landscape ecology and population genetics. Trends in Ecology & Evolution 18, 189-197. Manning WE (1978) The classification within the Juglandaceae. Annals of the Missouri Botanical Garden 65, 1058-1087. Marquardt PE, Epperson BK (2004) Spatial and population genetic structure of microsatellites in white pine. Molecular Ecology 13, 3305-3315. Marsico TD, Hellmann JJ, Romero-Severson J (2009) Patterns of seed dispersal and pollen flow in Quercus garryana (Fagaceae) following post-glacial climatic changes. Journal of Biogeography 36, 929-941. Martin NH, Bouck AC, Arnold ML (2006) Detecting adaptive trait introgression between Iris fulva and I. brevicaulis in highly selective field conditions. Genetics 172, 2481-2489. Maruyama T, Kimura M (1980) Genetic variability and effective population size when local extinction and recolonization of subpopulations are frequent. PNAS 77, 6710-6714. Maudet C, Miller C, Bassano B, et al. (2002) Microsatellite DNA and recent statistical methods in wildlife conservation management: applications in Alpine ibex [Capra ibex (ibex)]. Molecular Ecology 11, 421-436. Mayr E (1992) A Local Flora and the Biological Species Concept. American Journal of Botany 79, 222-238. McCleary T, Robichaud R, Nuanes S, et al. (2009) Four Cleaved Amplified Polymorphic Sequence markers for the detection of the Juglans ailantifolia chloroplast in putatively native J. cinerea populations. Molecular Ecology Resources 9, 525527. McDaniel JC (1956) The pollination of Juglandaceae varieties- Illinois observations and review of earlier studies. Annual Report of the Northern Nut Growers Association 47, 118- 132. McDaniel JC (1979) Other walnut including butternuts, heartnuts and hybrids. In: Nut Tree Culture in North America (ed. Jaynes RA), pp. 98-127. Northern Nut Growers Association, Hamden, CT. McLachlan JS, Clark JS, Manos PS (2005) Molecular indicators of tree migration capacity under rapid climate change. Ecology 86, 2088-2098. 168

Meldgaard T, Nielsen E, Loeschcke V (2003) Fragmentation by weirs in a riverine system: A study of genetic variation in time and space among populations of European grayling ( Thymallus thymallus ) in a Danish river system. Conservation Genetics 4, 735-747. Mercer KL, Shaw RG, Wyse DL (2006) Increased Germination of Diverse Crop-Wild Hybrid Sunflower Seeds. Ecological Applications 16, 845-854. Metcalf JL, Siegle MR, Martin AP (2008) Hybridization Dynamics between Colorado's Native Cutthroat Trout and Introduced Rainbow Trout. J Hered 99, 149-156. Millikan DF, Stefan SJ, Rigert KS (1990) Selection and preservation of Butternut, Juglans cinerea L. Annual Report of the Northern Nut Growers Association 81, 22- 25. Milne RA, Abbott RJ (2008) Reproductive isolation among two interfertile Rhododendron species: low frequency of post-F1 hybrid genotypes in alpine hybrid zones. Molecular Ecology 17, 1108-1121. Morin R, Beaulieu J, Deslauriers M, Daoust G, Bousquet J (2000) Low genetic diversity at allozyme loci in Juglans cinerea. Canadian Journal of Botany 78, 1238- 1243. Morris RT (1917) Notes on an excursion to Merribrooke, the country place of Dr. Robert T Morris, at Stamford, Conn., September 15, 1917 Annual Report of the Norther Nut Growers Association 8. Muir G, Lowe AJ, Fleming CC, Vogl C (2004) High nuclear genetic diversity, high levels of outcrossing and low differentiation among remnant populations of Quercus petraea at the margin of its range in Ireland. Annals of Botany 93, 691697. Mullen SP, Dopman EB, Harrison RG (2008) Hybrid zone origins, species boundaries, and the evolution of wing-pattern diversity in a polytypic species complex of North American admiral butterflies (Nymphalidae: limenitis). Evolution 62, 14001417. Naiman and RJ, Decamps H (1997) The ecology of interfaces: Riparian zones. Annual Review of Ecology and Systematics 28, 621-658. Naiman RJ, Fetherston KL, McKay SJ, Chen J (1998) Riparian forests. In: River ecology and management: lessons from the Pacific coastal ecoregion (eds. Naiman RJ, Bilby RE, Kantor S). Springer. Nair VMG, Kostichka CJ, Kuntz JE (1979) Sirococcus clavigignenti-juglandacearum: An Undescribed Species Causing Canker on Butternut. Mycologia 71, 641-646.

169

Neilson JA (1930) Some notes on the Japanese walnut in North America. Northern Nut Growers Report of the Proceedings of the 21st Annual Meeting, 39- 46. Nesom G (2000) Plant Guide- Butternut (Juglans cinerea L.). USDA NRCS. Nicese FP, Hormaza JI, McGranahan GH (1998) Molecular characterization and genetic relatedness among walnut (Juglans regia L.) genotypes based on RAPD markers. Euphytica 101, 199-206. Nielsen C, Cherry M, Boysen B, et al. (2003) COSEWIC status report on the butternut Juglans cinerea in Canada (ed. Committee on the Status of Endangered Wildlife in Canada. Ottowa C). Nilsson C, Svedmark M (2002) Basic principles and ecological consequences of changing water regimes: Riparian plant communities Environmental Management 30. Oddou-Muratorio S, Demesure-Musch B, Raphael P, Gouyon P-H (2004) Impacts of gene flow and logging history on the local genetic structure of a scattered tree species, Sorbus torminalis L. Crantz. Molecular Ecology 13, 3689-3702. Okello JBA, Wittemyer G, Rasmussen HB, et al. (2008) Effective population size dynamics reveal impacts of historic climatic events and recent anthropogenic pressure in African elephants. Molecular Ecology 17, 3788-3799. Omar S, Lemonnier B, Jones N, et al. (2000) Antimicrobial activity of extracts of eastern North American hardwood trees and relation to traditional medicine. Journal of Ethnopharmacology 73, 161-170. Orchard LP (1984) Butternut canker: host range, disease resistance, seedling-disease reactions and seed-borne transmission. Ph.D. Diss., University of Wisconsin. Orsini L, Corander J, Alasentie A, Hanski I (2008) Genetic spatial structure in a butterfly metapopulation correlates better with past than present demographic structure. Molecular Ecology 17, 2629-2642. Ostry M (1997) Sirococcus clavigignenti-juglandacearum on Heartnut (Juglans ailantifolia var. cordiformis). Plant Disease 81, 1461. Ostry M, Ellingson B, Seekins D, Ruckheim W (2002) The need for silvicultural practices and collection of butternut germplasm for species conservation. In: Proceedings, 13th Central Hardwood Forest Conference eds. Van Sambeek JW, Dawson JO, Ponder F, Loewenstein EF, Fralish JS). USDA, Forest Service, North Central Research Station, Urbana IL. Ostry M, Woeste K (2004) Spread of butternut canker in North America, host range, evidence of resistance within butternut populations and conservation genetics. 170

Black walnut in a new century, proceedings of the 6th Walnut Council research symposium. USDA Forest Service General Technical Report NC-243. Ostry ME, Mielke ME, Skilling DD (1994) Butternut-- Strategies for managing a threatened tree. General Technical Report NC-165. USDA Forest Service, North Central Forest Experiment Station, St. Paul, MN. Ostry ME, Pijut PM (2000) Butternut: an underused resource in North America. Hort Technology 10, 302- 306. Pardini EA, Hamrick JL (2008) Inferring recruitment history from spatial genetic structure within populations of the colonizing tree Albizia julibrissin (Fabaceae). Molecular Ecology 17, 2865-2879. Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes 6, 288-295. Pearson RG (2006) Climate change and the migration capacity of species. Trends in Ecology & Evolution 21, 111-113. Peterson CJ (2000) Damage and recovery of tree species after two different tornadoes in the same old growth forest: a comparison of infrequent wind disturbances. Forest Ecology and Management 135, 237-252. Petit RJ, Aguinagalde I, de Beaulieu J-L, et al. (2003) Glacial refugia: hotspots but not melting pots of genetic diversity. Science 300, 1563-1565. Petit RJ, Bialozyt R, Garnier-Gere P, Hampe A (2004a) Ecology and genetics of tree invasions: from recent introductions to Quaternary migrations. Forest Ecology and Management 197, 117-137. Petit RJ, Bodenes C, Ducousso A, Roussel G, Kremer A (2004b) Hybridization as a Mechanism of Invasion in Oaks. New Phytologist 161, 151-164. Petit RJ, Brewer S, Bordács S, et al. (2002) Identification of refugia and post-glacial colonisation routes of European white oaks based on chloroplast DNA and fossil pollen evidence. Forest Ecology and Management 156, 49-74. Petit RJ, Duminil J, Fineschi S, et al. (2005) Comparative organization of chloroplast, mitochondrial and nuclear diversity in plant populations. Molecular Ecology 14, 689-701. Petit RJ, Hu FS, Dick CW (2008) Forests of the past: A window to future changes. Science 320, 1450-1452. Petit RJ, Mousadik Ae, Pons O (1998) Identifying populations for conservation on the basis of genetic markers. Conservation Biology 12, 844- 855.

171

Petit RmJ, Hampe A (2006) Some evolutionary consequences of being a tree. Annual Review of Ecology, Evolution, and Systematics 37, 187-214. Pielou EC (1992) After the ice age: the return of life to unglaciated North America Oxford University Press. Pollegioni P, Major A, Bartoli S, et al. (2004) Application of Microsatellite and Dominant Molecular Markers for the Discrimination of Species and Interspecific Hybrids in Genus Juglans. In: V International Walnut Symposium. Acta Horticulturae, Sorrento, Italy. Prentice IC, Bartlein PJ, Thompson W, III (1991) Vegetation and Climate Change in Eastern North America Since the Last Glacial Maximum. Ecology 72, 2038-2056. Preston FW (1962) The canonical distribution of commonness and rarity, part I. Ecology 43, 185-215. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155, 945-959. Queller D, Goodnight K (1989) Estimating relatedness using genetic markers. Evolution 43. R-CDT (2005) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rajora OP, Pluhar SA (2003) Genetic diversity impacts of forest fires, forest harvesting, and alternative reforestation practices in black spruce (Picea mariana). TAG Theoretical and Applied Genetics 106, 1203-1212. Raymond M, Rousset F (1995a) GENEPOP (Version 1.2): Population genetics software for exact tests Journal Of Heredity 86, 248-249. Raymond M, Rousset F (1995b) GENEPOP (Version 1.2): Population genetics software for exact tests and ecumenicism. Journal of Heredity 86, 248-249. Reed CA, Davidson J (1958) The Improved Nut Trees of North America, and How to Grow Them Devin-Adair Company, New York. Rhymer JM, Simberloff D (1996) Extinction by hybridization and introgression. Annual Review of Ecological and Systematics 27, 83-109. Rieseberg LH (1997) Hybrid Origins of Plant Species. Annual Review of Ecology and Systematics 28, 359-389. Rieseberg LH, Raymond O, Rosenthal DM, et al. (2003) Major ecological transitions in wild sunflowers facilitated by hybridization. Science 301, 1211-1216. Rink G (1990) Juglans cinerea L. Butternut. In: Silvics of North America. Vol 2. Hardwoods. Agriculture Handbook 654 (eds. Burns RM, Honkala BH), pp. 386390. U.S. Forest Service, Washington, DC.

172

Robichaud R, Glaubitz J, Rhodes O, Woeste K (2006a) A robust set of black walnut microsatellites for parentage and clonal identification. New Forests 32, 179-196. Robichaud RL, Glaubitz JC, Rhodes OE, Woeste K (2006b) A robust set of black walnut microsatellites for parentage and clonal identification. New Forests 32, 179-196. Robledo-Arnuncio JJ, Collada C, Alía R, Gil L (2005) Genetic structure of montane isolates of Pinus sylvestris L. in a Mediterranean refugial area. Journal of Biogeography 32, 595-605. Rogers DL, Millar CI, Westfall RD (1999) Fine-Scale Genetic Structure of Whitebark Pine (Pinus albicaulis): Associations with Watershed and Growth Form. Evolution 53, 74-90. Romano S (2010) Our current understanding of the Upper Mississippi River System floodplain forest. Hydrobiologia 640, 115-124. Rosenberg NA, Li LM, Ward R, Pritchard JK (2003) Informativeness of Genetic Markers for Inference of Ancestry. The American Journal of Human Genetics 73, 14021422. Ross-Davis A, Huang Z, McKenna J, Ostry M, Woeste K (2008a) Morphological and molecular methods to identify butternut (Juglans cinerea) and butternut hybrids: relevance to butternut conservation. Tree Physiology 28, 1127-1133. Ross-Davis A, Ostry M, Woeste KE (2008b) Genetic diversity of butternut (Juglans cinerea) and implications for conservation. Canadian Journal of Forest Research 38, 899-907. Rossetto M, Jones R, Hunter J (2004) Genetic effects of rainforest fragmentation in an early successional tree (Elaeocarpus grandis). Heredity 93, 610-618. Rousset F (1997) Genetic differentiation and estimation of gene flow from F-statistics under isolation-by-distance. Genetics 145, 1219-1228. Rozen S, Skaletsky HJ (2000) Primer3 on the WWW for general users and for biologist programmers. In: Bioinformatics Methods and Protocols: Methods in Molecular Biology. (eds. Krawetz S, Misener S), pp. 365-386. Humana Press, Totowa, NJ. Rubidge EM, Taylor EB (2004) Hybrid zone structure and the potential role of selection in hybridizing populations of native westslope cutthroat trout (Oncorhynchus clarki lewisi) and introduced rainbow trout (O. mykiss). Molecular Ecology 13, 3735-3749. Ruegg K (2008) Genetic, morphological, and ecological characterization of a hybrid zone that spans a migratory divide. Evolution 62, 452-466.

173

Saccheri I, Kuussaari M, Kankare M, et al. (1998) Inbreeding and extinction in a butterfly metapopulation. Nature 392, 491-494. Sagarin RD, Gaines SD (2002) The 'abundant centre' distribution: to what extent is it a biogeographical rule? Ecology Letters 5, 137-147. Sagarin RD, Gaines SD, Gaylord B (2006) Moving beyond assumptions to understand abundance distributions across the ranges of species. Trends in Ecology & Evolution 21, 524-530. Santamour FS (1970) Hybrid Vigour in Seedlings of Re-created London Plane. Nature 225, 1159-1160. Sargent CS (1894) Forest flora of Japan. Notes on the forest flora of Japan Houghton, Mifflin and Co., Boston. Savolainen O, Pyhajarvi T, Knurr T (2007) Gene flow and local adaptation in trees. Annual Review of Ecology, Evolution, and Systematics 38, 595-619. Schlarbaum SE, Anderson RL, Ostry ME, et al. (2004) Butternut conservation: understanding aspects of canker resistance for the development of breeding orchards. In: Forest Genetics and Tree Breeding in the Age of Genomics: Progress and Future, IUFRO Joint Conference of Division 2. North Carolina State University, Charleston, SC. Schlarbaum SE, Hebard F, Spaine PC, Kamalay JC (1997a) Three American tragedies: chestnut blight, butternut canker, and Dutch elm disease. In: Proceedings, Exotic pests of eastern forests (ed. Britton KO), pp. 45- 54. Tennessee Exotic Pest Plant Council, Nashville, TN. Schlarbaum SE, Hebard F, Spaine PC, Kamalay JC (1997b) Three American tragedies: chestnut blight, butternut canker, and Dutch elm disease. . In: Exotic pests of eastern forests, pp. 45- 54. Exotic Pest Plant Council, Nashville, TN. Schmalz T, Bergdahl D (2006) Impact of Sirococcus clavigignenti-juglandacearum on health of butternut Phytopathology 96, S182. Schultz J (2003) Conservation Assessment for Butternut or White Walnut (Juglans cinerea L.). USDA Forest Service, Eastern Region. Schwartz M, McKelvey K (2009) Why sampling scheme matters: the effect of sampling scheme on landscape genetic results. Conservation Genetics 10, 441-452. Schwartz MK, Mills LS, Ortega Y, Ruggiero LF, Allendorf FW (2003) Landscape location affects genetic variation of Canada lynx (Lynx canadensis). Molecular Ecology 12, 1807-1816. Schweitzer JA, Martinsen GD, Whitham TG (2002) Cottonwood hybrids gain fitness traits of both parents: a mechanism for their long-term persistence. American Journal of Botany 89, 981-990. 174

Segelbacher G, Höglund J, Storch I (2003) From connectivity to isolation: genetic consequences of population fragmentation in capercaillie across Europe. Molecular Ecology 12, 1773-1780. Sezen UU, Chazdon RL, Holsinger KE (2009) Proximity is not a proxy for parentage in an animal-dispersed Neotropical canopy palm. Proceedings of the Royal Society B: Biological Sciences 276, 2037-2044. Slatkin M (1993) Isolation by distance in equilibrium and non-equilibrium populations. Evolution 47, 264-279. Slavov GT, Leonardi S, Burczyk J, et al. (2009) Extensive pollen flow in two ecologically contrasting populations of Populus trichocarpa. Molecular Ecology 18, 357-373. Smith JF, Doyle JJ (1995) A Cladistic Analysis of Chloroplast DNA Restriction Site Variation and Morphology for the Genera of the Juglandaceae. American Journal of Botany 82, 1163-1172. Smouse PE, Peakall R (1999) Spatial autocorrelation analysis of individual multiallele and multilocus genetic structure. Heredity 82, 561-573. Smulders MJM, Cottrell JE, Lefèvre F, et al. (2008) Structure of the genetic diversity in black poplar (Populus nigra L.) populations across European river systems: Consequences for conservation and restoration. Forest Ecology and Management 255, 1388-1399. Snow AA, Moran-Palma P, Rieseberg LH, Wszelaki A, Seiler GJ (1998) Fecundity, phenology, and seed dormancy of F1 wild-crop hybrids in sunflower (Helianthus annuus, Asteraceae). Am. J. Bot. 85, 794-. Sokal RR, Rohlf FJ (1969) Biometry, pp. 715 - 717. W H Freeman and Company, New York. Soltis DE, Morris AB, McLachlan JS, Manos PS, Soltis PS (2006) Comparative phylogeography of unglaciated eastern North America. Molecular Ecology 15, 4261-4293. Soto A, Lorenzo Z, Gil L (2007) Differences in fine-scale genetic structure and dispersal in Quercus ilex L. and Q. suber L.: consequences for regeneration of mediterranean open woods. Heredity 99, 601-607. Sousa WP (1984) The Role of Disturbance in Natural Communities. Annual Review of Ecology and Systematics 15, 353-391. Stanford AM, Harden R, Parks CR (2000) Phylogeny and Biogeography of Juglans (Juglandaceae) Based on matK and ITS Sequence Data. American Journal of Botany 87, 872-882. 175

Stockwell CA, Hendry AP, Kinnison MT (2003) Contemporary evolution meets conservation biology. Trends in Ecology & Evolution 18, 94-101. Storz JF, Beaumont MA (2002) Testing for Genetic Evidence of Population Expansion and Contraction: An Empirical Analysis of Microsatellite DNA Variation Using a Hierarchical Bayesian Model. Evolution 56, 154-166. Tabacchi E, Correll DL, Hauer R, et al. (1998) Development, maintenance and role of riparian vegetation in the river landscape. Freshwater Biology 40, 497-516. Tamura N, Hayashi F (2008) Geographic variation in walnut seed size correlates with hoarding behaviour of two rodent species. Ecological Research 23, 607-614. Tauleigne-Gomes C, Lefèbvre C (2005) Natural hybridisation between two coastal endemic species of Armeria (Plumbaginaceae) from Portugal. 1. Populational in situ investigations. Plant Systematics and Evolution 250, 215-230. Taylor SS, Jamieson IG, Wallis GP (2007) Historic and contemporary levels of genetic variation in two New Zealand passerines with different histories of decline. Journal of Evolutionary Biology 20, 2035-2047. Theiss LE (1933) Nut trees used as landscaping. In: Northern Nut Growers Report of the Proceedings of the 24st Annual Meeting, Downington, PA. Tiffin P, Olson MS, Moyle LC (2001) Asymmetrical crossing barriers in angiosperms. Proceedings of the Royal Society: Biological Sciences 268, 861-867. Tkacz B, Moody B, Villa Castillo J (2007) Forest Health Status in North America. The Scientific World 7. Tollefsrud MM, Sonstebo JH, Brochmann C, et al. (2009) Combined analysis of nuclear and mitochondrial markers provide new insight into the genetic structure of North European Picea abies. Heredity 102, 549-562. Trapnell DW, Schmidt JP, Hamrick JL (2008) Spatial Genetic Structure of the Southeastern North American Endemic, Ceratiola ericoides (Empetraceae). Journal Of Heredity 99, 604-609. Turelli M, Orr HA (2000) Dominance, Epistasis and the Genetics of Postzygotic Isolation. Genetics 154, 1663-1679. Twedt DJ, Wilson RR (2002) Supplemental Planting of Early Successional Tree Species During Bottomland Hardwood Afforestation, Gen. Tech. Rep. SRS-48. Asheville, NC: U.S. Department of Agriculture, Forest Service, Southern Research Station. pg. 358-364.

176

Valbuena-Carabana M, Gonzalez-Martinez SC, Hardy OJ, Gil L (2007) Fine-scale spatial genetic structure in mixed oak stands with different levels of hybridization. Molecular Ecology 16, 1207-1219. Valbuena-Carabaña M, González-Martínez SC, Sork VL, et al. (2005) Gene flow and hybridisation in a mixed oak forest (Quercus pyrenaica Willd. and Quercus petraea (Matts.) Liebl.) in central Spain. Heredity 95, 457-465. Van Mantgem PJ, Stephenson NL, Byrne JC, et al. (2009) Widespread increase of tree mortality rates in the western United States. Science 323, 521-524. 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. Van Sambeek JW, Ostry ME, Zaczek JJ (2002) Survival and grwoth of deep-planted, inleaf grafts in a germplasm repository of canker-resistant butternut. In: Proceedings, 13th Central Hardwood Forest Conference (eds. Van Sambeek JW, Dawson JO, Ponder F, Loewenstein EF, Fralish JS). USDA, Forest Service, North Central Research Station, Urbana, IL. Vandergast AG, Bohonak AJ, Weissman DB, Fisher RN (2007) Understanding the genetic effects of recent habitat fragmentation in the context of evolutionary history: phylogeography and landscape genetics of a southern California endemic Jerusalem cricket (Orthoptera: Stenopelmatidae: Stenopelmatus). Molecular Ecology 16, 977-992. Vaughan SP, Cottrell JE, Moodley DJ, Connolly T, Russell K (2007) Distribution and fine-scale spatial-genetic structure in British wild cherry (Prunus avium L.). Heredity 98, 274-283. Vekemans X, Hardy OJ (2004) New insights from fine-scale spatial genetic structure analyses in plant populations. Molecular Ecology 13, 921-935. Vellend M (2004) Parallel Effects of Land-Use History on Species Diversity and Genetic Diversity of Forest Herbs. Ecology 85, 3043-3055. Vellend M, Orrock. JL (2009) Ecological and genetic models of diversity: Lessons across disciplines. In: The Theory of Island Biogeography Revisited. eds. Losos J, Ricklefs RE), pp. 439-461. Princeton University Press. Vendramin GG, Fady B, González-Martínez SC, et al. (2008) Genetically depauperate but widespread: the case of an emblematic Mediterranean pine. Evolution 62, 680-688.

177

Victory ER, Glaubitz JC, Rhodes OE, Jr., Woeste KE (2006) Genetic homogeneity in Juglans nigra (Juglandaceae) at nuclear microsatellites. American Journal of Botany 93, 118-126. Vignieri SN (2005) Streams over mountains: influence of riparian connectivity on gene flow in the Pacific jumping mouse (Zapus trinotatus). Molecular Ecology 14, 1925-1937. Vucetich JA, Waite TA (2000) Is one migrant per generation sufficient for the genetic management of fluctuating populations? Animal Conservation 3, 261-266. Wang J (2002) An Estimator for Pairwise Relatedness Using Molecular Markers. Genetics 160, 1203-1215. Wang J (2004) Application of the one-migrant-per-generation rule to conservation and management. Conservation Biology 18, 332-343. Watano Y, Kanai A, Tani N (2004) Genetic structure of hybrid zones between Pinus pumila and P. parviflora var. pantephylla (Pinaceae) revealed by molecular hybrid index analysis. American Journal of Botany 91, 65-72. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution 38, 1358-1370. Wen J (1999) Evolution of eastern Asian and eastern North American disjunct distributions in flowering plants. Annual Review of Ecology and Systematics 30, 421-455. Whitmore TC (1989) Canopy Gaps and the Two Major Groups of Forest Trees. Ecology 70, 536-538. Wiegand KM (1935) A Taxonomist's Experience with Hybrids in the Wild. Science 81, 161-166. Williams JW, Jackson ST (2007) Novel climates, no-analog communities, and ecological surprises. Frontiers in Ecology and the Environment 5, 475-482. Williamson-Natesan E (2005) Comparison of methods for detecting bottlenecks from microsatellite loci. Conservation Genetics 6, 551-562. Wolf DE, Takebayashi N, Rieseberg LH (2001) Predicting the risk of extinction through hybridization. Conservation Biology 15. Yao X, Ye Q, Kang M, Huang H (2007) Microsatellite analysis reveals interpopulation differentiation and gene flow in the endangered tree Changiostyrax dolichocarpa with fragmented distribution in central China. New Phytologist 176, 472-480. Zalapa JE, Brunet J, Guries RP (2007) Hybridization and introgression patterns between native red elm (Ulmus rubra Muhl.) and exotic, invasive Siberian elm (Ulmus pumila L.) examined using species-specific microsatellite markers [abstract]. In: CONGEN3: The Third International Conservation Genetics Symposium, p. 20. 178

Zellmer AJ, Knowles LL (2009) Disentangling the effects of historic vs. contemporary landscape structure on population genetic divergence. Molecular Ecology 18, 3593-3602. Zhang Q, Liu Y, Sodmergen (2003) Examination of the Cytoplasmic DNA in Male Reproductive Cells to Determine the Potential for Cytoplasmic Inheritance in 295 Angiosperm Species. Plant & Cell Physiology 44, 941-951. Zheng H, Wu Y, Ding J, et al. (2004) Invasive Plants of Asian Origin Established in the United States and Their Natural Enemies. USDA Forest Service General Technical Publication FHTET-2004-05.

179