MIAMI UNIVERSITY The Graduate School

Certificate for Approving the Dissertation

We hereby approve the Dissertation

of

Zachary St. John Taylor

Candidate for the Degree:

Doctor of Philosophy

______Susan M.G. Hoffman, Director

______David J. Berg, Reader

______Brian Keane, Reader

______Douglas B. Meikle

______Emily S. Murphree Graduate School Representative

ABSTRACT

GEOGRAPHICAL HETEROGENEITY AND LANDSCAPE-SCALE GENETIC PATTERNS IN GREAT LAKES POPULATIONS OF PEROMYSCUS

Zachary S. Taylor

Woodland mice of the genus Peromyscus are broadly distributed throughout North America, where they interact with a wide variety of landscape features, climates, and biological communities. Each of the central chapters of this dissertation examines genetic heterogeneity in a species of Peromyscus in relation to landscape features of the Great Lakes region, in order to illuminate the biogeographical constraints facing small mammals in this region. Chapter 1, General introduction Chapter 2, MtDNA genetic structure transcends natural boundaries in Great Lakes populations of deer mice (Peromyscus maniculatus gracilis), examines the genetic structure of deer mice to describe the effect of the Great Lakes on the colonization of northern Michigan from southern refugial sources after the end of the last glacial cycle. Analyses reveal a complex structure indicating occasional migration across Lake Michigan during the 10,000-year history of mouse habitation in the region. Chapter 3, Landscape fragmentation and geographical isolation define microsatellite genetic structure in Great Lakes populations of deer mice (Peromyscus maniculatus gracilis), describes the genetic structure of deer mice using nuclear microsatellite markers. In contrast to Chapter 2, which examines the Great Lakes as barriers to postglacial colonization, Chapter 3 considers the role of the lakes in promoting the differentiation of populations through genetic drift, after expansion from a common source. Chapter 4, Landscape-scale fragmentation and genetic structure in populations of the northern white-footed mouse (Peromyscus leucopus

noveboracensis), describes the genetic structure of white-footed mice along a transect from southern Ohio to northern Michigan. Because this transect covers a heterogeneous landscape and climatic gradient, habitat fragmentation and biogeographical range limitation are considered as possible determinants of genetic patterns. Chapter 5, Conclusion

GEOGRAPHICAL HETEROGENEITY AND LANDSCAPE-SCALE GENETIC PATTERNS IN GREAT LAKES POPULATIONS OF PEROMYSCUS

A DISSERTATION

Submitted to the Faculty of

Miami University

in partial fulfillment of

the requirements for the degree of

Doctor of Philosophy

Zoology

by

Zachary S. Taylor

Miami University

Oxford, Ohio

2010

Dissertation Director: Susan M.G. Hoffman

© Zachary S. Taylor 2010

Table of Contents

List of Tables ...... iv List of Figures ...... v CHAPTER 1 General Introduction ...... 1 Overview ...... 1 Background ...... 2 References ...... 8 CHAPTER 2 MtDNA genetic structure transcends natural boundaries in Great Lakes populations of deer mice (Peromyscus maniculatus gracilis) ...... 12 Introduction ...... 13 Materials and methods ...... 15 Results ...... 21 Discussion ...... 23 Acknowledgements ...... 27 References ...... 28 For 2 groups, with eastern populations only ...... 35 CHAPTER 3 Natural landscape fragmentation defines microsatellite genetic structure in Great Lakes populations of deer mice (Peromyscus maniculatus gracilis) ...... 42 CHAPTER 4 Landscape-scale fragmentation and genetic structure in populations of the northern white-footed mouse (Peromyscus leucopus noveboracensis) ...... 68 Abstract ...... 68 Introduction ...... 69 Materials and Methods ...... 71 Results ...... 74 Discussion ...... 76 Acknowledgments ...... 79 References ...... 80 CHAPTER 5. Conclusion ...... 95 References ...... 101

iii

List of Tables

Chapter 2 Table 1. Differentiation of populations and population divisions in Great Lakes Peromyscus maniculatus 33 Table 2. Differentiation of populations and population divisions in Great Lakes Peromyscus maniculatus 34

Chapter 3 Table 1. Microsatellite loci selected for analysis 61 Table 2. Molecular diversity indexes for populations of Peromyscus maniculatus 62

Chapter 4 Table 1. Microsatellite loci selected for analysis 85 Table 2. Standard molecular diversity indexes for populations of Peromyscus leucopus 86

iv

List of Figures

Chapter 2 Figure 1. Distribution of Peromyscus maniculatus gracilis haplogroups in the central Great Lakes region 35 Figure 2. Bayesian haplotype tree for Great Lakes Peromyscus maniculatus gracilis and related groups 36 Figure 3. Parsimony network for Great Lakes and North Carolina Peromyscus maniculatus 37 Mismatch analysis of Peromyscus maniculatus samples 38

Chapter 3 Figure 1. Trapping locations for Peromyscus maniculatus gracilis in the Great Lakes region 63 Figure 2. Spatial distribution of genetic structure for Peromyscus maniculatus gracilis 64 Figure 3. Clustering of Michigan populations of Peromyscus maniculatus gracilis by STRUCTURE 65

Chapter 4 Figure 1. Study area and trapping sites for Peromyscus leucopus 87 Figure 2. Habitat availability and genetic diversity for Peromyscus leucopus 88 Figure 3. Inter-population differentiation populations of Peromyscus leucopus 89 Figure 4. Clustering analyses using STRUCTURE 90

v

Dedication

To Victoria, Stella, June, and my parents, for your endless patience and support.

vi

Acknowledgements

I am indebted to my advisor, Susan Hoffman, for the opportunity to find my way through these exciting projects; and to David Berg, Brian Keane, Douglas Meikle, and Emily Murphree, for their help and patience. I could not have completed this dissertation without assistance from a large number of individuals and institutions, including Phil Myers, Rosa Moscarella, and Robbyn Abbitt, for extraordinary assistance vital to this dissertation. Many students and colleagues have contributed labor to the projects described herein, including Molly Steinwald, Cameron Beech, Justin Campbell, Brett Chatman, Darcy Dayhoff, Brian Dinh, Allison Dixon, and Kyle Westhafer. The staffs of Pigeon River State Forest, Seney National Wildlife Refuge, the Michigan State University Museum, and the University of Michigan Museum of Zoology were instrumental in sampling efforts over the years. Chris Wood, Xiaoyun Deng, and John Hawes of the Center for Bioinformatics and Functional Genomics at Miami University were always extremely knowledgeable and helpful with sequencing and genotyping. This work was completed with funding from the Department of Zoology and the Graduate School at Miami University, and from a Grant-in-Aid provided by the American Society of Mammalogists.

vii

CHAPTER 1

General Introduction

Overview

When natural or artificial barriers prevent animals from moving from one area to another, this interference can reduce the chances for long-term survival of affected populations or even species. In extreme cases, populations of animals trapped in a particular patch of habitat may become so small that they be extirpated. In less dramatic cases, the populations may survive, but because the gene pool is reduced they can eventually lose genetic diversity and become very different from populations in other habitat patches. Therefore, to assess the long-term viability of animal populations, one must ask what exactly constitutes a patch or a barrier. Older models based on the biogeography-influenced concept of islands of habitat in a sea of inhospitable matrix are now being replaced, as modern conservation-oriented studies have revealed that landscape composition affects animal movements in complex ways. Terrain that appears devoid of a particular species can in fact be permeable to varying extents, or even support some individuals at lower levels than the optimal habitat usually studied by population biologists. Likewise, apparently suitable habitat can vary in quality due to the presence of competitors or predators, or to more subtle climatic or environmental variables. While the complexity of population dynamics is difficult to study directly on a landscape scale, a variety of molecular markers and analytical techniques now exist that allow assessment of populations indirectly. By using a suite of techniques from phylogeography and population genetics, the impact on populations of both ancient events, such as species range expansions, and current constraints, such as habitat availability or geographical barriers, can be evaluated. Together, these studies can illuminate the relationships between populations and their environments, and allow better prediction of the fate of populations given known landscape constraints on gene flow and population size. This dissertation addresses how the genetic composition of small-mammal populations relates to ancient and present-day landscape composition and environmental variability. Two common, ecologically similar forest-dwelling mouse species, the white-

1

footed mouse (Peromyscus leucopus noveboracensis), and the woodland deer mouse (Peromyscus maniculatus gracilis) are examined separately, and used as model species to illuminate the effects of different types of geographical structure on the genetics of small mammals in the Great Lakes region. First, the role of the Great Lakes in influencing both ancient and recent genetic processes is examined using the woodland deer mouse. Then, the white-footed mouse is used to study the effects of wide-scale anthropogenic habitat fragmentation on population diversity.

Background

Peromsycus species in the Great Lakes region — The Great Lakes region presents a varied and dynamic biological landscape. During the last glacial maximum, the entire region was buried beneath snow and ice. The current biota of the region is thus largely the result of massive northward migrations of both plants and animals over the last 10,000 years, following the retreat of the glaciers. Because of its relatively central location in North America, this region appears to have functioned as a contact zone for intraspecific lineages of animals that survived the glacial periods in different glacial refugia. For example, two distinct lineages of short- tailed shrews (Blarina brevicauda) meet at the western edge of the Great Lakes region (Brant & Ortí 2003), as do two lineages of the black rat snake (Elaphe obsoleta; Burbrink et al. 2000), suggesting that both species migrated north from multiple refugia on either side of the Mississippi River (Soltis et al. 2006). Similar patterns have been observed in chipmunks (Tamias striatus; Rowe et al. 2004; 2006), white-footed mice (Rowe et al. 2006), and garter snakes (Thamnophis sirtalis; Placyk et al. 2007), though it is unlikely that the similar patterns reflect the use of the same specific refugia by each species (Rowe et al. 2006; Soltis et al. 2006). Two studies have provided some support for an overall similar pattern in deer mice (Lansman et al. 1983; Dragoo et al. 2006), but the sampling designs of those studies provided low resolution in the immediate vicinity of the lakes (Taylor & Hoffman 2010). Postglacial climate change and range expansion originally shaped populations of the species that now inhabit the Great Lakes region, but in the last few centuries anthropogenic changes have transformed the fauna in this area. Major deforestation,

2

beginning about 200 years ago, radically reduced the habitat available to woodland- dwelling small mammals, while more recently, increasing temperatures seem to be having significant ecological effects on mammals in the region (Myers et al. 2009). Deforestation has been pervasive in North America since European colonization, and continues today in areas of high human population density and agricultural demand (Botti & Moore 2006). Regional variation in habitat destruction has been exacerbated by abandonment of unproductive agricultural lands, which have developed into second- growth forest in many places, especially in the northern Midwestern states and in New England. Prospects for forest-dwelling wildlife have recently improved dramatically in these more-northern areas, and even once-rare large mammals such as moose, elk, and wolves are returning (Botti & Moore 2006). Unfortunately, warming temperatures may have the potential to offset some of the gains that have been made in habitat availability for forest-dwelling species. For instance, climate change over the last hundred years appears to be driving the northward expansion of small mammals, including white-footed mice (Peromyscus leucopus), eastern chipmunks (Tamias striatus), southern flying squirrels (Glaucomys volans), and common opossums (Didelphis virginiana), into the northern Midwest (Myers et al. 2009, Long 1996). In turn, these range expansions are correlated with a sharp decline in the abundance of historically more-northern mammal species, including woodland deer mice (Peromyscus maniculatus gracilis), southern red-backed voles (Myodes gapperi), woodland jumping mice (Napaeozapus insignis), least chipmunks (Tamias minimus), and northern flying squirrels (Glaucomys sabrinus). Perhaps because of the intervening barriers created by the Great Lakes, these recent changes have not occurred uniformly across the region. For instance, while the more southern white-footed mouse has expanded its range throughout both peninsulas of Michigan, the decline of the more northern woodland deer mouse in that state has been much more pronounced on the Lower Peninsula (Myers et al. 2009). The proximate reasons for diminishing populations of the northern mammal species are not fully understood, but some evidence suggests that in the case of the mice, milder winters have facilitated the decline of the deer mouse by allowing their ecological replacement by the white-footed mouse (Wolff 1985a; 1985b; Myers et al. 2009).

3

In the Great Lakes region, the white-footed mouse is represented by a single subspecies (Peromyscus leucopus noveboracensis; hereafter simply P. leucopus) while Peromyscus maniculatus is represented by two ecologically distinct subspecies: the woodland deer mouse (P. maniculatus gracilis; hereafter P. m. gracilis), and the prairie deer mouse (P. maniculatus bairdii, hereafter P. m. bairdii), the latter of which is discussed only tangentially in this dissertation. Both P. leucopus and P. m. gracilis are small forest-dwelling mice that are identical in size and nearly identical in appearance. These two species have overlapping ranges in the Great Lakes region, but most of the historic range of P. m. gracilis lies to the north of that of P. leucopus (Hall 1981). Although competition in nature is relatively difficult to demonstrate, Wolff (1985a; 1985b; 1996) built a strong case for competitive interactions between these species in a series of studies on P. leucopus and P. m. gracilis, a subspecies very closely related to P. m. gracilis (see Lansman et al. 1983; Taylor & Hoffman 2010). These species shared an essentially identical ecological niche, exhibited interspecific territoriality, and appeared to coexist in the Appalachian Mountains only because of differences in their responses to cold and to food limitation. That is, sustained winters seemed to favor P. m. nubiterrae, which hoarded more food and built more extensive nests, while shorter winters allowed P. leucopus to dominate (Wolff 1985a; 1985b). Although interactions between P. leucopus and P. m. gracilis in the Great Lakes region have not been studied as intensively, several research groups have shown that recent increases in the abundance of P. leucopus at the northern edge of its range in Wisconsin (Long 1996) and Michigan (Myers et al. 2009) have been accompanied by corresponding decreases in the abundance of P. m. gracilis. Peromyscus leucopus has also moved into areas where only P. m. gracilis had been found previously, and appears to be steadily increasing in those areas. Climate is not the only factor that has recently changed for native animals around the Great Lakes. Importantly, hardwood forest habitat has become more unevenly distributed throughout the region (Chapter 4, Figure 1), with higher densities in the north and very low densities in southern Michigan and in Ohio. Because P. m. gracilis is a forest specialist, fragmentation may contribute to its changing distribution by limiting this species to the northern Great Lakes region. Other regional species that are in decline, such as northern flying squirrels and woodland jumping mice, typically live at low

4

population densities and therefore are likely to depend on relatively large, contiguous forest habitat patches for survival. Those species and populations that persist in fragmented patches may still be affected; genetic evidence from mammals including fishers (Martes pennanti), wolverines (Gulo gulo), and brown bears (Ursos arctos) suggests that populations sizes and genetic diversity are lower in areas where human encroachment has fragmented habitat into fewer, smaller patches (Wisely et al. 2004; Cegelski et al. 2006; Paetkau et al. 1998).

Genetic Analyses — Events such as habitat fragmentation and migration leave genetic signals in the DNA of populations, which can be recovered using the techniques of molecular biology and analyzed using an ever-increasing array of sophisticated statistical methods. Two closely related, but distinct, disciplines have evolved around such studies; phylogeography rests on the analysis of the largely residual signals from ancient processes, such as species expansion and differentiation, while the discipline of population genetics is concerned with more contemporary processes such as gene flow and genetic drift. The separation of these two disciplines is in fact quite artificial, as the same processes underlie both; however, because the two different sets of tools are most suitable for different types of questions, they warrant some discussion.

Mitochondrial DNA Sequences and Phylogeography — A number of biologists studying animal mitochondrial DNA sequences in the 1980's discovered that these sequences have a number of very useful characteristics. Most notably, they are highly variable among and within species, but essentially monomorphic (and thus non- recombining) within an individual (Brown et al. 1979; Avise et al. 1979; Lansman et al. 1983). Because each haploid mitochondrial genome is inherited maternally as a single locus in most mammals (Hutchison et al. 1974; Avise et al. 1979), each individual animal contains one DNA sequence that can be seen as a node in a lineage that records its relationship to other animals, both within its species and among related species (Avise et al. 1983; Lansman et al. 1983). While DNA sequences from the nuclear chromosomes can be used to derive similar information, mitochondrial sequences remain the most popular for phylogeographic studies because they have shorter coalescence times, are

5

technically easier to obtain and yield more information per unit sequence than nuclear sequences (Avise 1998). One popular application of phylogeographic methods emerged with the realization that the evolutionary timescale of some mitochondrial DNA sequences overlaps the northward expansion of many species after the end of Pleistocene glaciations about 10,000 years ago (Avise 1998). The DNA sequences of animals in previously glaciated terrain can be seen as records of migration since the end of the most recent glacial cycles, and so a number of studies have used these sequences to ask how animals reacted to ancient climate change and habitat availability (reviewed in Soltis et al. 2006). These same methods are very useful for questions about the role of the Great Lakes, as natural barriers that formed at the end of the last glacial period, in restricting migration on an evolutionary timescale.

Nuclear Loci and Population Genetics —Small populations of animals behave differently from large populations. In particular, they are more vulnerable to the effects of random fluctuations in population size, which can in some cases lead to population extinction (Shaffer 1981). Small populations are also vulnerable to genetic drift, defined as random changes in allele frequencies that include the nonselective loss of alleles, because some individuals fail to reproduce due to chance events. In contrast, gene flow between populations has the effect of reducing drift by enlarging the genetically effective size of populations. Therefore, the genetic consequences of having small populations are most pronounced when those populations are also isolated. Population genetics provides laboratory and statistical methods for indirectly measuring aspects of populations that may be affected by drift, such as population size and migration. These methods have been developed to measure non-adaptive genetic differences that develop between populations due to genetic drift (Wright 1931; Nei et al. 1976). Microsatellites, also known as short tandem repeats (STR’s), or simple sequence repeats (SSR’s), are short, non-coding intergenic DNA sequences. Because of the repetitive structure of these sequences, they are difficult for polymerases to copy accurately during replication and are therefore prone to indel mutations. Further, because they do not tend to confer selective advantages or costs, large numbers of alleles can

6

accumulate, thus providing a means for differentiating individuals or populations. In addition, a lack of selection on microsatellites means that allelic distribution and abundance should reflect general trends in population size and gene flow rather than variation in selective pressures (Nei & Graur 1984). Analyses of molecular markers for population genetic purposes generally fall into one of two categories: diversity comparisons and differentiation analyses. Diversity comparisons are used primarily to derive information about relative current and historic population size. Common statistics of this type include the average number of alleles per locus in a population and the proportion of a population that is heterozygous at the loci examined. Measures of inter-population differentiation are used primarily to examine the extent to which populations are effectively separated. These include FST (Wright 1931), which is always a number between zero (no differentiation) and one (no shared alleles), and a number of related measures based on interpopulation variance in allele frequency and allele size (e.g., Slatkin 1995; Weir & Cockerham 1984). The distinct distributions of P. leucopus and P. m. gracilis around the Great Lakes make the two species useful for addressing the genetic effects of different types of habitat fragmentation. Because the range of P. m. gracilis is limited to northern forests but divided into separate landmasses by the Great Lakes, this species is especially appropriate for studying the importance of these natural barriers to population genetic processes. Different effects of the Great Lakes as barriers are addressed in two chapters: Chapter 2 uses mitochondrial DNA sequences and phylogeographic methods to ask how the lakes impacted the postglacial migration of P. m. gracilis thousands of years ago, while Chapter 3 uses microsatellite DNA markers to ask how the lakes affect migration between populations in the present day. On the other hand, because the range of P. leucopus traverses both heavily agricultural regions of Ohio and southern Michigan, and largely forested regions such as northern Michigan, its populations are useful for addressing the role of anthropogenic habitat fragmentation. Therefore, Chapter 4 of this dissertation uses techniques from population genetics and Geographical Information Systems (GIS) to address the role of anthropogenic habitat fragmentation and other factors in the population genetic structure of P. leucopus in Michigan and Ohio.

7

References

Avise, J. C., Lansman, R. A., Shade, R. O. 1979. The use of restriction endonucleases to measure mitochondrial DNA sequence relatedness in natural populations. I. Population structure and evolution in the genus Peromyscus. Genetics 92: 279-295.

Avise, J. C., Shapira, J. F., Daniel, S. W., Aquadro, C. F., Lansman, R. A. 1983. Mitochondrial DNA differentiation during the speciation process in Peromyscus. Molecular Biology and Evolution 1: 38-56.

Avise, J. C. 1998. The history and purview of phylogeography: a personal reflection. Molecular Ecology 7:371-379.

Botti, W. B., Moore, M. D. 2006. Michigan’s state forests: a century of stewardship. Michigan State University Press, East Lansing, Michigan.

Brown, W. M., George, M. Jr., Wilson, A. C. Rapid evolution of animal mitochondrial DNA. 1979. Proceedings of the National Academy of Sciences of the U.S.A. 76: 1967- 1971.

Burbrink, F. T., Lawson, R., Slowinski, J. B. 2000. Mitochondrial DNA phylogeography of the polytypic North American rat snake (Elaphe obsoleta): a critique of the subspecies concept. Evolution 54: 2107-2118.

Cegelski, C. C., Waits, L. P., Anderson, N. J., Flagstad, O., Strobeck, C., Kyle, C. J. 2006. Genetic diversity and population structure of wolverine (Gulo gulo) populations at the southern edge of their current distribution in North America with implications for genetic viability. Conservation Genetics 7: 197-211.

Dragoo, J. W., Lackey, J. A., Moore, K. E., Lessa, E. P., Cook, J. A., Yates, T. L. 2006. Phylogeography of the deer mouse (Peromyscus maniculatus) provides a predictive

8

framework for research on hantaviruses. Journal of General Virology 87: 1997-2003. doi: 10.1099/vir.0.81576-0.

Hall, E. R. 1981. The Mammals of North America. John Wiley & Sons, Inc., New York.

Hutchison, C. A. III, Newbold, J. E., Potter, S. S., Edgell, M. H. 1974. Maternal inheritance of mammalian mitochondrial DNA. Nature 251: 536-538.

Lansman, R. A., Avise, J. C., Aquadra, C. F., Shapira, J. F., Daniel, S. W. 1983. Extensive genetic variation in mitochondrial DNA’s among geographic populations of the deer mouse, Peromyscus maniculatus. Evolution 37: 1-16.

Long, C. A. 1996. Ecological replacement of the deer mouse, Peromyscus maniculatus, by the white-footed mouse, P. leucopus, in the Great Lakes region. Canadian Field- Naturalist 110: 271-277.

Myers, P., Lundrigan, B.L., Hoffman, S.M.G., Haraminac, A.P., Seto, S.H. 2009. Climate-induced changes in the small mammal communities of the northern Great Lakes Region. Global Change Biology 15: 1434-1454.

Nei, M., Fuerst, P.A., Chakraborty, R. 1976. Testing the neutral mutation hypothesis by distribution of single locus heterozygosity. Nature 262: 491-493.

Nei, M., Graur, D. 1984. Extent of protein polymorphism and the neutral mutation theory. Evolutionary Biology 17: 73-118.

Paetkau, D., Waits, L. P., Clarkson, P. L., Craighead, L., Vyse, E., Ward, R., Strobeck, C. 1998. Variation in genetic diversity across the range of North American brown bears. Conservation Biology 12: 418-429.

9

Placyk, J. S. Jr., Burghardt, G. M., Small, R. L., King, R. B., Casper, G. S., Robinson, J. W. 2007. Post-glacial recolonization of the Great Lakes region by the common gartersnake (Thamnophis sirtalis) inferred from MtDNA sequences. Molecular Phylogenetics and Evolution 43: 452-467.

Reed, D. H., Frankham, R. 2001. How closely related are molecular and quantitative measures of genetic variation? A meta-analysis. Evolution 55:1095-1103.

Reed, D. H., Frankham, R. 2003. Correlation between fitness and genetic diversity. Conservation Biology 17:230-237.

Rogers, A. R., Harpending, H. 1992. Population growth makes waves in the distribution of pairwise genetic differences. Molecular Biology and Evolution 9: 552-569.

Rowe, K. C., Heske, E. J. , Brown, P. W., Paige, K. N. 2004. Surviving the ice: northern refugia and postglacial colonization. Proceedings of the National Academy of Sciences of the USA 28: 10355-10359.

Rowe, K. C., Heske, E. J., Paige, K. N. 2006. Comparative phylogeography of eastern chipmunks and white-footed mice in relation to the individualistic nature of species. Molecular Ecology 15: 4003-4020.

Shaffer, M. L. 1981. Minimum population sizes for species conservation. BioScience 31: 131-134.

Slatkin, M. 1995. A measure of population subdivision based on microsatellite allele frequencies. Genetics 139: 457-462.

Soltis, D. E., Morris, A. B., McLachlan, J. S., Manos, P. S., Soltis, P. S. 2006. Comparative phylogeography of unglaciated eastern North America. Molecular Ecology 15: 4261-4293.

10

Taylor, Z. S., Hoffman, S. M. G. 2010. Mitochondrial DNA genetic structure transcends natural boundaries in Great Lakes populations of deer mice (Peromyscus maniculatus gracilis). Canadian Journal of Zoology 88: 404–415, doi: 10.1139/Z10-010.

Weir, B. S, Cockerham, C.C. 1984. Estimating f-statistics for the analysis of population structure. Evolution 38:1358-1370.

Wisely, S. M., Buskirk, S. W., Russel, G. A., Aubrey, K. B., Zielinski, W. J. 2004. Genetic diversity and structure of the fisher (Martes pennanti). J. Mammalogy 85: 640- 648.

Wolff, J. O. 1985a. Comparative population ecology of Peromyscus leucopus and Peromyscus maniculatus. Canadian Journal of Zoology 63: 1548-1555

Wolff, J. O. 1985b. The effects of density, food, and interspecific interference on home range size in Peromyscus leucopus and Peromyscus maniculatus. Canadian Journal of Zoology 63: 2657-2662.

Wolff, J. O. 1996. Coexistence of white-footed mice and deer mice may be mediated by fluctuating environmental conditions. Oecologia 108: 529-533

Wright, S. 1931. Evolution in Mendelian populations. Genetics 16:97-159.

11

CHAPTER 2

MtDNA genetic structure transcends natural boundaries in Great Lakes

populations of deer mice (Peromyscus maniculatus gracilis)

[Note: This chapter has been published as: Taylor, Z. S., and Hoffman, S. M. G. 2010. Mitochondrial DNA genetic structure transcends natural boundaries in Great Lakes populations of deer mice (Peromyscus maniculatus gracilis). Canadian Journal of Zoology 88: 404–415, doi: 10.1139/Z10-010.]

Abstract: The landscape of the Great Lakes region has been fragmented since the lakes formed starting about 20,000 years ago. Small mammals such as deer mice (Peromyscus maniculatus (Wagner, 1845)) inhabiting the region therefore face barriers to migration and gene flow, which could complicate ongoing range shifts related to climate change. We analyzed DNA sequences for 481 base pairs of the mitochondrial D-loop in order to compare mouse genetic structure with the fragmented landscape and geological history of the region. Phylogenetic analyses reveal two distinct lineages of mice in the Great Lakes region. The spatial distribution of these two groups is not congruent with the fragmentation of the landscape; rather, a western group is found from Minnesota through the western Upper Peninsula of Michigan, while an eastern group spans part of that peninsula as well as northern Michigan and southern Ontario. The genetic data suggest that the eastern clade colonized Michigan through Ontario from a source shared with southern Appalachian mice, but are less informative for the western clade. Together, these findings suggest that the Great Lakes are relatively porous barriers in the long term, but may still have implications for the response of small mammal communities to climate change.

12

Introduction

Anthropogenic climate change is impacting biological communities across many latitudes and ecosystems (IPCC 2001; Parmesan and Yohe 2003; Thomas et al. 2004). Because the ranges of many animals are constrained at higher latitudes by climatic tolerances (Walther et al. 2002; Parmesan et al. 2000), this rapid warming is likely to have broad-reaching effects on wildlife communities in temperate regions. In the vicinity of the Great Lakes, small mammal communities have changed markedly over the last hundred years as species such as the northern flying squirrel (Glaucomys sabrinus (Shaw, 1801)), and the least chipmunk (Tamias minimus Bachman, 1839), have experienced range contractions, while congeneric species have expanded into the region from the south (Myers et al. 2009). The presence of the lakes themselves presumably complicates these community rearrangements by limiting migration routes for both invaders and resident species. The potential for the Great Lakes to disrupt animal movements is greatest in Michigan, which is fragmented into a northern Upper Peninsula (UP), a southern Lower Peninsula (LP), and numerous islands (Fig. 1). This landscape emerged fully from the ice cover of the Laurentide ice sheet about 11,000 years ago (see Dyke et al. 2002). Since then, fluctuating water levels in lakes Michigan and Huron have periodically altered the connectivity of the islands and peninsulas, most notably during low water stages known as Chippewa (Lake Michigan) and Stanley (Lake Huron), about 8,000- 7,500 years ago (Hough 1966; Lewis et al. 2007). Until about 6000 years ago (Larsen 1985), when water approached its current level, a number of islands that are now far within Lake Michigan were likely part of the mainland LP, and the two peninsulas were themselves separated only by a narrow channel (Dietrich 1988). The woodland deer mouse, Peromyscus maniculatus gracilis (Le Conte, 1855), is among those species whose range shifts in Michigan are complicated by the presence of the lakes. It is one member of a closely related group of long-tailed, gracile P. maniculatus subspecies that inhabit similar niches throughout the forests of northern North America (Hall 1981). Except for a limited area in the southern and central Appalachian Mountains, the present-day ranges of these forest-dwelling deer mice were completely covered by the glaciers of the Wisconsin period; thus, the mice must have

13

colonized the region from the south after the Great Lakes formed. Peromyscus maniculatus gracilis now occupies transitional forests along the eastern U.S./Canada border, including the northern half of Michigan (Fig. 1). Because the historical southern range limit of P. m. gracilis passes through the middle of the LP of Michigan (Fig. 1), at approximately 44o N. latitude (Myers et al. 2005, 2009), while lakes surround the peninsula in all other directions, LP populations of this subspecies are effectively isolated. Gene flow between the LP and surrounding populations today would require that individuals traverse hundreds of kilometers of unsuitable habitat or more than 6 km over water. Therefore, if P. m. gracilis populations in the UP and LP have been separated to the degree suggested by geography, they should have diverged genetically since resettlement of the region after deglaciation, or at least since the end of the low-water stages that minimized over-water distances between the two peninsulas. Very recently, P. m. gracilis has been largely replaced on the Lower Peninsula (LP) of Michigan by the white-footed mouse (Peromyscus leucopus (Rafinesque 1812)) (Myers et al. 2005; Myers et al. 2009), so that the former is now rare outside a small region near 45o N latitude, on the border of Otsego and Cheboygan Counties. Lake Michigan islands in the Beaver Island group were also likely accessible from both peninsulas during the Chippewa stage of about 8000 years ago (Dietrich 1988), when they were colonized by trees (Kapp et al. 1969) and may have been colonized by animals including Peromyscus maniculatus (Hatt et al. 1948). As the islands have presumably been isolated since the end of the Chippewa stage, subsequent mouse colonization would require rafting on ice or debris, or anthropogenic transplantation. Two prior studies have used mitochondrial DNA to address the genetic structure of P. maniculatus across North America (Lansman et al. 1983; Dragoo et al. 2006), although the large scale of these studies precluded analysis of fine detail at the local scale. While these studies used different markers and are not in complete agreement, they provide a general picture of woodland P. maniculatus in eastern North America. First, one clade is found along the Appalachian Mountain chain, from North Carolina/Tennessee northeast to New Brunswick. A second clade overlaps the range of the first clade in the southern Appalachian Mountains and extends northwest to Ontario

14

(Lansman et al. 1983) and Michigan (Dragoo et al. 2006) in the eastern Great Lakes region. Finally, a third clade is found in the western UP (Lansman et al. 1983) and in Manitoba and far western Ontario (Dragoo et al. 2006). We are interested in the contact zone between the second and third of these groups in the Great Lakes region. We analyzed the D-loop of the mitochondrial control region of P. m. gracilis from populations in Ontario, the LP, the UP, and several islands in Lake Michigan, to address the questions of separate refugial origins and of subsequent gene flow between mice around the Great Lakes. We also included samples from an area of western North Carolina that represents a potential refugial source for Great Lakes populations of P. m. gracilis, as inferred from the study by Lansman and colleagues (1983); these samples were members of the closely related subspecies Peromyscus maniculatus nubiterrae Rhoads, 1896, which has been shown to be genetically indistinguishable from P. m. gracilis (Lansman et al. 1983; Dragoo et al. 2006). We tested four hypotheses that we developed on the basis of prior studies and known geological history. First, that mouse lineages from separate refugia occupied the Upper and Lower Peninsulas of Michigan, with the potential for some overlap between the two at the eastern end of the UP. Second, that P. m. gracilis in the northern Lower Peninsula is descended from mice in the southern Appalachian Mountains. Third, that mice followed a direct migration path from a southern Appalachian refugium, so that LP populations were ancestral to related populations in the Great Lakes region. Finally, we hypothesized that mice on Lake Michigan islands are descended from LP populations, due to high accessibility during the Chippewa low-water stand. We expect that the genetic structure of P. m. gracilis in the Great Lakes region and inferences about the permeability of these major landscape barriers to migration by small terrestrial animals will inform predictions about the impacts of barriers on populations faced with rapidly changing habitats.

Materials and methods

Sampling Animals were trapped from eight Michigan counties every year from 2002 to 2007 by our research group or by colleagues using the same techniques. Animals were

15

captured alive in standard Sherman traps. Traps were baited with whole oats, placed at 10m intervals in late afternoon and checked at dawn. Ear tissue for genetic analysis was removed with sterile dissecting scissors and stored in SET buffer (1% SDS, 10 mM Tris- HCl pH 7.5, 5 mM EDTA; Brinster et al. 1985) on ice until it could be frozen. Animal handling methods were developed in accordance with guidelines established by the American Society of Mammalogists (Gannon et al. 2007) and approved by the Institutional Animal Care and Use Committee of Miami University. We define a population as a single trapping site, or a small group of trapping sites within a 25 km diameter, that are on the same landmass and within the same county. Where a population included several trapping sites, the geographical center of these locations (obtained from the calculator at http: //www.geomidpoint.com) was used for distance calculations. Tissue samples from islands in the Beaver Island group were obtained from the collections of the University of Michigan Museum of Zoology, as were some samples from Alger County. These islands were treated as one location due to sample size limitations. Samples from North Fox Island were obtained from the collection of the Michigan State University Museum. Samples from two populations in Ontario were provided by Lawrence Pinto of Northwestern University. Populations studied are shown in Fig. 1, and exact trapping sites are given in Table S11. To place our populations within the genetic context established by prior studies, we obtained P. m. gracilis skulls from Clearwater County, Minnesota (from the Bell Museum of Natural History of the University of Minnesota), and P. m. nubiterrae skulls from Yancey County, North Carolina (from the University of Michigan Museum of Zoology)2. P. m. nubiterrae and P. m. gracilis are ecologically similar subspecies that have been shown to be genetically indistinguishable in mitochondrial DNA studies (Lansman et al. 1983; Dragoo et al 2006). DNA from museum skulls was isolated from maxilloturbinal bones using previously described techniques (Wisely et al. 2004), except that the smaller size of the Peromyscus skulls compared to the Mustela nigripes (Audubon & Bachman 1851) skulls used in that study often made it necessary to use bones from both nostrils.

1 Supplementary Table S1. 2 Supplementary Table S2. 16

Sample preparation and sequencing DNA was isolated from tissue using the e.Z.N.A. Tissue DNA Kit (Omega Bio- Tek, Norcross, Georgia). A 570 bp fragment of the mitochondrial D-loop, including part of the central conserved domain and most of the 3’ variable region, was amplified and sequenced using PeromtD-F4 (5’ TCTGGTTCTTACTTCAGGGCC 3’) and PeromtD-R (5’ GCATTTTCAGTGCTTTGCTTTATTG 3’). The polymerase chain reaction cycle used consisted of 94o C for 2 min; 40 cycles of 30 s at 94o C, 30 s at 50o C, and 1 min. at 72o C; 5 min. at 72o C. Products were purified by agarose gel electrophoresis and extracted with the QiaQuick Gel Extraction Kit (Qiagen, Valencia, California). Cycle sequencing was performed with the BigDye Terminator v3.1 kit and run on a 3130xl Genetic Analyzer (Applied Biosystems, Foster City, California). Forward and reverse sequences for each fragment were assembled and trimmed with Sequencher 4.8 (Gene Codes, Ann Arbor, Michigan) to remove ambiguous sequences from the ends. DNA from museum specimens was extracted and handled in a separate portion of the laboratory using designated reagents and equipment for all processes through PCR amplification.

Phylogenetic analyses Sequences were aligned with ClustalW (Larkin et al. 2007) as implemented in MEGA 4.0 (Tamura et al. 2007) and reduced to haplotypes with Collapse 1.2. In order to avoid method-specific tree artifacts, we constructed phylogenetic trees with neighbor- joining (NJ), maximum parsimony (MP), and Bayesian methods. Trees included all available Great Lakes sequences plus sequences from museum specimens representing North Carolina and Minnesota. For additional genetic context, we rooted our trees with a P. leucopus sequence generated during this study and included published sequences from congeners Peromyscus keeni Merriam, 1897 (GenBank EU140797) and Peromyscus polionotus (Wagner 1843) (EU140791) and additional subspecies Peromyscus maniculatus bairdii (Hoy and Kennicott, 1857) (EU140795), Peromyscus maniculatus pallescens (J. A. Allen 1896) (EU140794), and P. m. sonoriensis (Le Conte 1853) (EU140796). Modeltest v3.7 (Posada & Crandall 1998) was used to select a TrN + I + 

17

maximum likelihood model of molecular evolution. The model selected according to the Akaike information criterion (AIC) includes a matrix of substitution rates (6 substitution types), varying base frequencies (A = 0.3149, C = 0.2366, G = 0.1216, T = 0.3269), a proportion of invariable sites (I = 0.6964), and a gamma-shaped site rate distribution (shape parameter = 2.3181). NJ trees were constructed in PAUP* v4.0b10 (Swofford 2002) using Tamura-Nei distance and the gamma shape parameter from Modeltest, with 1,000 bootstrap replicates. Bayesian trees were constructed over 2.5. x 106 generations with MrBayes v3.1.2 (Ronquist & Huelsenbeck 2003), with model settings of Nst = 6, a proportion of invariable sites and a gamma-shaped rate distribution. MP trees were constructed in PAUP* using heuristic searches starting with stepwise addition trees, with 10 random sequence-addition replicates, and tree-bisection-reconnection (TBR) branch swapping. The maximum number of trees saved at each step was set to 100, due to computational limitations. The reliability of internal nodes was tested with 500 bootstrap replicates and a maximum trees setting of 500.

Population structure and genetic diversity Because most of our populations were in the geographically homogeneous UP, we used the simulated annealing approach implemented in SAMOVA v1.0 to identify functionally relevant population groups. This program is similar to the analysis of molecular variance (AMOVA) performed by Arlequin, but it first separates populations into a user-designated number of groups for which the proportion CT of all variance due to inter-group differences is maximized (Dupanloup et al. 2000). We performed the analysis for 2, 3, and 4 groups, with 100 starting conditions each, to explore the possibilities of differing types of genetic barriers. We also performed the analysis for 2 groups with western populations excluded, to test for structure within the eastern population group. Significance of  statistics was tested with 1,000 permutations of individuals within each level. We calculated pairwise ST with Arlequin v3.11 (Excoffier et al. 2005). For measures of genetic diversity, we estimated haplotype richness R, the number of haplotypes per population, with the program HP-Rare, which uses rarefaction to correct for sample size bias (Kalinowski 2006). Chippewa County was excluded from

18

estimation of R because of its limiting population size of 5. For combined analyses and for western haplotypes only, N = 9 was used for rarefaction, while for the eastern haplotypes only, the sample size for rarefaction was 11. Data for the smaller populations are presented as the total numbers of haplotypes present. The standard molecular diversity measures were calculated in Arlequin v3.1. Nucleotide diversity ( ) is the probability that two randomly chosen nucleotides are different, and k is the mean number of pairwise differences between all haplotypes in the sample. Because expected colonization pathways coincided with north-south transects between the two peninsulas, or east-west transects across the Upper Peninsula, we performed linear regression and correlation analyses of genetic diversity with latitude and longitude using the Analysis ToolPak for Excel (Microsoft, Redmond, Washington) to test our predictions of colonization routes.

Ancestral and derived haplotypes Minimum spanning haplotype networks were constructed using the statistical parsimony approach of TCS v1.21 (Clement et al. 2000). Ancestral and derived haplotypes were determined according to the outgroup weight assigned by TCS based primarily on level of connectivity within the network, where haplotypes with high outgroup weights are much more likely to be ancestral. In all cases, haplotypes with high outgroup weights occupied central positions within a radiation, while the effect of loops was minimal.

Population expansions Fu’s neutrality test (Fu 1997) and mismatch analyses were performed in Arlequin v3.11. The neutrality test was performed for the eastern and western clades separately, for evidence of recent expansions. The test for significance of FS values is based on comparison to random samples generated assuming neutrality and a population at equilibrium; FS is significant at the 5% level for P-values below 0.02, a unique property of this test identified by its author (Fu 1997). Significantly negative values of Fu’s FS indicate a recent spatial or demographic expansion. Mismatch analysis and comparison with a model of spatial expansion were performed in Arlequin for the eastern and western

19

clades separately, for the entire data set, and for the eastern clade and North Carolina group together. In analysis of the distribution of pairwise sequence mismatches, expansion is expected to increase the preservation of ancestral diversity, resulting in a single mode reflecting the time since the expansion (Rogers & Harpending 1992). We used a molecular clock approach based on mismatch distributions to generate rough estimates of the relative time since expansion or divergence. For the eastern clade populations, with and without the eastern clade haplotypes from North Carolina, we used the value of  estimated by Arlequin for the mode of the distribution and the 95% confidence limit of  for a time range for the expansion. For the weakly bimodal distribution seen in the mismatch comparison of the entire eastern clade and North Carolina clade, the mean pairwise difference between North Carolina and other populations was used as an approximation of the value of the second mode, to date the divergence of these groups. We used Nei’s corrected mean pairwise difference to reduce the distorting effects of ancestral polymorphisms. The same approach was used for the P. maniculatus data set as a whole, to date the original split between lineages. For the molecular clock calibration, we used an estimate of 500,000 years for the divergence of P. maniculatus and P. leucopus (Rowe et al. 2006), based on the appearance in the fossil record of morphological intermediates (Hibbard 1968). Although both the use of the fossil record for calibration and the assumption of constant mutation rates can introduce considerable error into these estimates, they are still useful as relative measures and for comparison with similar studies. Using 8 P. leucopus sequences obtained in this study, we found a corrected average of 70 pairwise differences between the two species (14.5% difference), yielding a mutation rate of 1.53 x 10-7 mutations/base pair/year/lineage. This value is comparable to values used for the control region of P. leucopus (Rowe et al. 2006). We used the simplifying assumption of two generations per year for P. maniculatus in this region, although real populations can have from one to three (Layne 1968). We estimated expansion or divergence time from the equation  = 2ut, where t is the expansion time in generations and u is the mutation rate per sequence (= sequence length x mutation rate per site) per generation (Rogers 1995).

20

Results

Sequencing and phylogenetic analyses We obtained 481 bp of unambiguous sequence for 198 individuals from the Great Lakes region, which resolved into 45 distinct haplotypes. Sequences of museum specimens from North Carolina and Minnesota added another 23 samples and 9 haplotypes. A representative of each haplotype was deposited in GenBank (accession numbers GQ249411-GQ249465). The maximum sequence divergence was 3.5% between pairs of Great Lakes samples and 4.2 % in all pairwise comparisons. All tested methods of phylogenetic analysis supported a structure consisting of two distinct clades in the Great Lakes region: a western clade consisting of 20 haplotypes from Minnesota and from throughout the UP, and an eastern clade of 27 haplotypes found in the central and eastern UP, Ontario, the LP, and the Lake Michigan islands (Fig. 2). The eastern clade included two additional haplotypes from North Carolina; the remaining North Carolina haplotypes formed a separate sister clade to the eastern group. Because of their geographical distance, Minnesota and North Carolina samples were excluded from subsequent analyses unless otherwise indicated.

Distribution of haplotypes The haplotype network defined three distinct groups of haplotypes. The two haplogroups found in the Great Lakes region were separated by a minimum of six mutational steps, corresponding geographically to an east/west split (Fig. 3). Western haplotypes were the only ones observed west of 86o W longitude, in the Gogebic and Menominee populations of the western UP. They also accounted for most of the samples in Delta County (central UP), and were found in smaller numbers throughout the eastern UP, in the Alger, Schoolcraft, and Mackinac populations, but were absent from Ontario, the LP, and the islands (Fig. 1). All western haplotypes had low outgroup weights, obscuring ancestral haplotypes, and most had very limited distributions. In contrast, two ancestral eastern haplotypes were abundant and widely distributed: SC105 ranged from Ontario to Schoolcraft County in the central UP, while DA04 was found in the central and eastern UP, as well as in the Beaver and Fox Island groups. The LP populations contained only 3 private (those not found elsewhere)

21

haplotypes; all of these were immediate derivatives of SC105. Island populations shared no haplotypes with the LP; instead, they contained both private haplotypes and a haplotype shared with the UP (DA04). The North Carolina haplogroup was separated by a minimum of 4 mutational steps from the eastern clade and by at least 6 mutational steps from the western haplogroup.

Population structure The SAMOVA determined the following groups of populations (Table 2): for 2 groups, the division corresponded to the split in dominance between western and eastern haplotypes. Populations in the western UP counties of Gogebic, Menominee and Delta were separated into one group, and those of all other UP counties, Ontario, the LP, and the Beaver Islands into the other, CT = 0.57. Increasing values of k resulted in the subdivision of the western population group: for 3 groups, Menominee County at the southwestern edge of the UP was identified as one group, with the other groups as described above for 2 groups, CT = 0.58, while for 4 groups, the LP samples were further separated from the eastern populations, CT = 0.55. When we removed the western populations from the analysis to test for division within the eastern population group, the

LP population was separated from the other eastern populations, but the resulting CT of 0.18 was not statistically significant. For all populations and clades in the Great Lakes region, diversity was high throughout the UP (average of 4.96 haplotypes/population) and lowest in the LP (R = 1.96), and on North Fox Island (R = 2.00; Table 1). Within the eastern clade, there was no relationship between diversity and longitude, which might be produced by movement across the UP, or between diversity and latitude. Haplotype richness tended to increase to the north but the regression was not significant (R2 = 0.46, P = 0.064, slope = 1.76). Within the western clade, both and k were lower to the north (for : R2 = 0.99, P = 0.0003, slope = -1.50). Within the three large populations of the western clade, haplotype richness increased to the east (R2 = 0.99, P = 0.045, slope = 0.86), but small sample sizes precluded this analysis for all western populations.

Population expansion and divergence

22

Fu’s neutrality test for the Great Lakes populations indicated a recent expansion of the eastern clade (FS = -11.55, P = 0.0040) and produced a nearly significant result for expansion of the western clade (FS = -5.76, P = 0.028). Minnesota samples were included with the western clade in this analysis for completeness, though the results were similar without them (FS = -5.20, P = 0.026). Mismatch analysis of the eastern clade revealed a unimodal distribution similar to that predicted by the spatial expansion model (1.19 < < 4.41; P = 0.32), consistent with an expansion 8,500 to 31,600 years BP (Fig. 4A). Including the two eastern clade haplotypes from North Carolina in the Eastern Clade data set produced a nearly identical distribution and expansion estimate of 8,600 to 32,200 years BP (1.20 <  < 4.50) for divergence of Great Lakes and North Carolina haplotypes of the eastern clade. Mismatch analysis of the eastern clade and the separate North Carolina clade produced a weakly bimodal distribution, with the first peak resulting from differences within the eastern clade and the second from differences between the two groups (Fig. 4C). The corrected number of pairwise differences between the eastern clade and the North Carolina clade was 4.12, corresponding to a split roughly 29,500 years BP. The North Carolina population was itself extremely diverse, in part reflecting the presence of two distinct lineages (the eastern Great Lakes clade and the North Carolina clade) in this population. Seven haplotypes were observed in the nine North Carolina samples, with an average of 7.56 pairwise differences, or over twice the value observed in the eastern Great Lakes populations. Finally, analysis of the entire P. maniculatus data set resulted in a bimodal distribution, with the second mode resulting from the differences between the eastern/North Carolina superclade and the western clade, and an average of 8.25 pairwise differences. This produces an estimate of 59,100 years BP for the split between these groups.

Discussion

Our analyses of mitochondrial D-loop sequences show two distinct lineages in Great Lakes populations of P. m. gracilis. Consistent with our first hypothesis, the two observed lineages do not sort strictly to the two peninsulas of Michigan; rather, the two groups meet sharply in the central UP. The eastern haplogroup dominates the eastern UP, the LP, the Lake Michigan islands, and Ontario, while the western group extends 23

westward from the central UP (Fig. 1). The close relationship between LP and eastern UP populations is consistent with the results of Dragoo et al. (2006), who also found closely related samples in both the LP and the eastern UP. Our finding of a distinct clade to the west of the LP is also consistent with that study, although the latter did not include samples from the western UP and therefore did not detect the genetic structure on the Upper Peninsula. Our results are also consistent with those of another phylogenetic study of Peromyscus (Lansman et al. 1983), in that we too found animals on the UP that are distinct from those on the LP. Together, these results suggest that the lakes have been sufficiently permeable to allow colonization of the LP and Lake Michigan islands from Ontario and the UP, while preventing panmixia of mitochondrial haplotypes across the lakes.

Refugial origins Our finding of eastern clade haplotypes in both North Carolina and most Great Lakes populations indicates a shared origin of these populations approximately 8600 to 32,200 years BP. Because our study area in the Great Lakes region was freed from the glaciers only about 11,000 years BP, the most obvious explanation for this relationship is that mice with refugial ancestors in the vicinity of the southern Appalachian Mountains colonized the Great Lakes region roughly 8600 to 11,000 years ago. This result is consistent with prior studies indicating that southern Appalachian Mountain haplotypes of Peromyscus maniculatus are in the same clade as eastern Great Lakes populations (Lansman et al. 1983; Dragoo et al. 2006) and are genetically intermediate between Ontario and northeastern Appalachian haplotypes (Lansman et al. 1983). However, we should note that identifying a specific refugium would require extensive sampling in the unglaciated United States, which is not possible given that little of the present distribution of woodland P. maniculatus is in unglaciated areas that might have served as refugia (see Fig. 1). Our inclusion of these North Carolina samples was intended to provide a relative genetic distance between southeastern deer mice, which are the best available representative of southern populations, and our eastern Great Lakes mice. Contrary to our expectations, the splits between the eastern clade and the other two clades appear to be not much older than the expansion of the eastern clade: on the

24

order of 30,000 years for the split between the eastern and North Carolina clades, and 59,000 years for the split between the eastern/North Carolina group and the western clade. Even given the many sources of uncertainty in these calculations, these dates are an order of magnitude more recent than would be expected of populations originating in separate refugia: two clades of Tamias striatus (Linneaus 1758) that colonized the western Great Lakes from separate refugia have an estimated divergence date of 200,000 years BP (Rowe et al. 2004). Therefore, our eastern and western clades of P. maniculatus may have diverged within the Wisconsin glacial period, before ultimately colonizing the Great Lakes region from different directions. However, because woodland deer mice are not found in the Midwest to the south of the western Great Lakes region (Hall 1981), identifying other possible refugial origins for any of these haplotypes would require DNA from fossilized remains.

Colonization of the Great Lakes region Based on current geography and on geological history, the Great Lakes themselves would be expected to restrict post-glacial migration paths for small, terrestrial animals, with the peninsulas of Michigan each being colonized separately by animals moving outward from the nearest mainland. If this were the case, the two peninsulas would be inhabited entirely by animals from different haplogroups. Instead, the overall pattern of haplotype distribution suggests that the UP was settled from both the west (via Wisconsin) and east (through Canada), while the LP was settled from the north or northeast. Because we expected a direct migration path from southern Appalachian refugia into Michigan, we hypothesized that the Lower Peninsula was an ancestral source of eastern haplotype populations in the area. However, three lines of evidence indicate that the LP populations are instead descended from UP ancestors. First, the haplotypic richness of the LP is much lower than it is to the north, as would be expected in the case of dispersal from the UP. Second, all of the haplotypes found on the LP are closely related to each other, implying recent descent from a common ancestor rather than the collapse of more diverse populations. Finally, the ancestral haplotype for all LP haplotypes is the most common haplotype on the UP. While a loss of diversity on the LP

25

due to genetic drift could have caused the first of these observed patterns, it is less likely to have caused the other two. Therefore, the eastern haplotype lineages probably arrived in the Lower Peninsula by way of the Upper Peninsula and Canada. Because the western clade is found only on the Upper Peninsula, we expect that it colonized the region from the mainland on its western end. Hypotheses for the post-glacial colonization of the Lake Michigan islands have focused on the Chippewa low-water stand of around 7,500 years BP, when the very shallow northern end of the Lake Michigan basin was largely empty. The Beaver Island group was presumably attached to the mainland LP and only weakly separated from the UP (Hatt 1948; Kapp et al. 1969; Bowen 2004). Although the origin of Beaver Island animals has received little study, colonization of the island has been studied for some tree species, which first arrived during the Chippewa period (Kapp et al. 1969), suggesting that small mammals could have also migrated to the island at that time. However, our genetic evidence is not consistent with an LP origin for island populations; populations from both the Beaver Island group and North Fox Island contain only haplotypes that are either unique or are shared with the UP. Thus, contrary to our fourth hypothesis, our genetic results suggest that the Beaver Island group was colonized by dispersers from the UP rather than from the LP, either during the Chippewa period or subsequently. North Fox Island, which is 10 km south of Beaver Island and was probably linked to it during the Chippewa low stand, may have been colonized from Beaver Island, or directly from the Upper Peninsula by means of rafting or a similar mechanism (Hatt et al. 1948).

Physical barriers and population structure As species invade new regions to occupy habitat made favorable by recently warming temperatures, the composition of invaded communities is changing. For instance, ecological replacement appears to be a factor in woodland deer mouse, northern flying squirrel, and least chipmunk population declines in the Great Lakes region (Long 1996; Myers et al. 2009). Similarly, the invasion of Scandinavian tundra by the red fox appears to be a component in the decline of arctic fox populations (Hersteinsson & Mcdonald 1992), although this decline may also be related to changing population dynamics of lemming prey (Gilg et al. 2008). Both invading and displaced species are

26

likely to be affected by barriers to dispersal, which inhibit expansion of the former and reduce habitat availability for the latter. Our study provides an example of a taxon whose past movements were relatively unconstrained by geography; on the relevant evolutionary time scale for Great Lakes populations of P. m. gracilis (~9,000 to 30,000 years), the lakes have apparently been a fairly porous barrier to their migration. However, dramatic changes in the local environment have occurred very rapidly over the last ~100 years. If the frequency of mouse travel over the lakes is low enough to be negligible over the short term, as seems likely given the observed genetic differences between peninsular and island mice, the lakes should impact P. m. gracilis populations in two important ways in the future. First, though warming temperatures are apparently allowing P. leucopus to ecologically replace P. m. gracilis populations on the peninsulas, limited access should protect island populations of P. m. gracilis from invasion. Second, the Lakes may effectively preventing demographic and genetic exchange between diminishing Lower Peninsula populations of P. m. gracilis and more robust populations on the UP. Therefore, even though these large lakes do not appear to have been severe barriers to the original colonization of the region by mice, they can be expected to interfere with rapid range shifts related to present-day climate change.

Acknowledgements

We thank B. Keane, P. Myers, D. Berg, R. Seidel, and two anonymous reviewers for helpful comments on versions of the manuscript. We are grateful to P. Myers, B. Lundrigan, R. Moscarella, and M. Steinwald for helpful conversations and assistance with sampling. The staffs of Seney National Wildlife Refuge and the Pigeon River Country State Forest, especially G. Corace, M. Tansy, T. Casselman and D. Mittlestat, generously shared their knowledge and provided assistance with sampling. We also thank L. Pinto, The University of Michigan Museum of Zoology, The Michigan State University Museum, and the Bell Museum of Natural History at the University of Minnesota for DNA, frozen tissue, and/or preserved museum specimens. This work was supported by Miami University and the Department of Zoology Summer Field Research Workshop.

27

References

Bowen, K.D., and Gillingham, J. C. 2004. Distribution of reptiles and amphibians on the islands of eastern Lake Michigan: summary and analysis. Mich. Acad. 36(2): 213-223.

Brinster, R.L., Chen, H.Y., Trumbauer, M.E., Yakle, M.K., and Palmiter, R.D. 1985. Factors affecting the efficiency of introducing foreign DNA into mice by microinjecting eggs. Proc. Natl. Acad. Sci. U.S.A. 82(13): 4438-4442.

Clement, M., Posada, D., and Crandall, K.A. 2000. TCS: a computer program to estimate gene genealogies. Mol. Ecol. 9(10): 1657-1659.

Dietrich, R.V. 1988. The geological history of Beaver Island. The Journal of Beaver Island History, 3: 59-77.

Dragoo, J.W., Lackey, J.A., Moore, K.E., Lessa, E.P., Cook, J.A., and Yates, T.L. 2006. Phylogeography of the deer mouse (Peromyscus maniculatus) provides a predictive framework for research on hantaviruses. J. Gen. Virol. 87(7): 1997-2003. doi: 10.1099/vir.0.81576-0.

Dupanloup, I., Schneider, S., and Excoffier, L. 2002. A simulated annealing approach to define the genetic structure of populations. Mol. Ecol. 11(2): 2571-2581. doi: 10.1046/j.1365-294X.2002.01650.x

Dyke, A. S., Andrews, J. T., Clark, P. U., England, J. H., Miller, G. H., Shaw, and Veillette, J. J. 2002. The Laurentide and Innuitian ice sheets during the Last Glacial Maximum. Quatern. Sci. Rev. 21: 9-31. doi: 10.1016/S0277-3791(01)00095-6.

Excoffier, L., Laval, G., and Schneider, S. 2005. Arlequin ver. 3.0: An integrated software package for population genetics data analysis. Evol. Bioinform. Online, 1: 47- 50.

28

Fu, Y-X. 1997. Statistical tests of neutrality of mutations against population growth, hitchhiking and background selection. Genetics, 147(2): 915-925.

Gannon, W.L., Sikes, R.S., and The Animal Care and Use Committee of the American Society Of Mammalogists. 2007. Guidelines of the American Society of Mammalogists for the use of wild mammals in research. J. Mammal. 88(3): 809-823. doi: 10.1644/06- MAMM-F-185R1.1.

Gilg, O., Sittler, B., and Hanski, I. 2009. Climate change and cyclic predator-prey population dynamics in the high Arctic. Glob. Chang. Biol. 15(11): 2634-2652. doi: 10.1111/j.1365-2486.2009.01927.x.

Hall, E.R. 1981. The Mammals of North America. John Wiley & Sons, Inc., New York.

Hatt, R.T., Van Tyne, J., Stuart, L.C., Pope, C.H., and Grobman, A.B. 1948. Island life: a study of the land vertebrates of the islands of eastern Lake Michigan. Cranbrook Inst. Sci. Bull. No. 27.

Hersteinsson, P., and Macdonald, D.W. Interspecific competition and the geographical distribution of red and arctic foxes Vulpes vulpes and Alopex lagopus. Oikos, 64: 505- 515.

Hibbard, C.W. 1968. Paleontology. In Biology of Peromyscus (Rodentia). Edited by J.A. King. Spec. Publ. Am. Soc. Mammal. No. 2, Stillwater, Okla. pp. 6-24.

Hough, J.L. 1966. Correlation of glacial lake stages in the Huron-Erie and Michigan basins. J. Geol. 74(1): 62-77.

Intergovernmental Panel on Climate Change (IPCC). 2001. Climate change 2001: synthesis report. Cambridge University Press, Cambridge, United Kingdom.

29

Kapp, R.O., Bushouse, S., and Foster, B. 1969. A contribution to the geology and forest history of Beaver Island, Michigan. Proceedings of the 12th Conference on Great Lakes Research, Ann Arbor, Mich. pp. 225-236.

Kalinowski, S.T. 2006. HP-Rare: A computer program for performing rarefaction on measures of allelic diversity. Mol. Ecol. Notes, 5(1): 187-189. doi: 10.1111/j.1471- 8286.2004.00845.x.

Lansman, R.A., J.C. Avise, C.F. Aquadro, J.F. Shapira, and S.W. Daniel. 1983. Extensive genetic variation in mitochondrial DNA’s among geographic populations of the deer mouse, Peromyscus maniculatus. Evolution, 37(1): 1-16.

Larkin, M.A., Blackshields, G., Brown, N.P., Chenna, R., Mcgettigan, P.A., Mcwilliam, H., Valentin, F., Wallace, I.M., Wilm, A., Lopez, R., Thompson, J.D., Gibson, T.J., and Higgins, D.G. 2007. Clustal W and Clustal X version 2.0. Bioinformatics, 23(21): 2947- 2948. doi: 10.1093/bioinformatics/btm404.

Larsen, C.E. 1985. Lake level, uplift, and outlet incision, the Nipissing and Algoma Great Lakes. In Quaternary evolution of the Great Lakes. Edited by P.F. Karrow and P.E. Calkin. Geological Association of Canada, St. John’s, pp. 64-77.

Layne, J.N. 1968. Ontogeny. In Biology of Peromyscus (Rodentia). Edited by J.A. King. Spec. Publ. Am. Soc. Mammal. No. 2, Stillwater, Okla. pp. 148-253.

Lewis, C.F.M., Heil, C.W., Jr., Hubeny, J.B., King, J.W., Moore, T.C., Jr., and Rea, D.K. 2007. The Stanley unconformity in Lake Huron basin: evidence for a climate-driven lowstand about 7900 14C BP, with similar implications for the Chippewa lowstand in Lake Michigan basin. J. Paleolimnol. 37(3): 435-452. doi: 10.1007/s10933-006-9049-y.

30

Long, C.A. 1996. Ecological replacement of the deer mouse, Peromyscus maniculatus, by the white-footed mouse, P. leucopus, in the Great Lakes region. Can. Field-Nat. 110(2): 271-277.

Myers, P., Lundrigan, B.L., and Koppel, R.V. 2005. Climate change and the distribution of Peromyscus in Michigan—Is global warming already having an impact? Pages 101-126 in Lacey EA and Myers P, editors. Mammalian diversification: from chromosomes to phylogeography. University of California Press, Berkeley.

Myers, P., Lundrigan, B.L., Hoffman, S.M.G., Haraminac, A.P., and Seto, S.H. 2009. Climate-induced changes in the small mammal communities of the northern Great Lakes Region. Glob. Chang. Biol. 15(6): 1434-1454. doi: 10.1111/j.1365-2486.2009.01846.x.

Parmesan, C., Root, T.L., and Willig, M.R. 2000. Impacts of extreme weather and climate on terrestrial biota. Bull. Am. Meteorol. Soc. 81(3): 443-450.

Parmesan, C., and Yohe, G. 2003. A globally coherent fingerprint of climate change impacts across natural systems. Nature (London), 421: 37-42. doi: 10.1038/nature01286.

Posada, D., and Crandall, K.A. 1998. Modeltest: testing the model of DNA substitution. Bioinformatics, 14: 817-818.

Rogers, A.R. 1995. Genetic evidence for a population explosion. Evolution, 49: 608- 615.

Rogers, A.R., and Harpending, H. 1992. Population growth makes waves in the distribution of pairwise genetic differences. Mol. Biol. Evol. 9: 552-569.

Ronquist, F., and Huelsenbeck, J.P. 2003. MRBAYES 3: Bayesian phylogenetic inference under mixed models. Bioinformatics, 19: 1572-1574.

31

Rowe, K.C, Heske, E.J., Brown, P.W., and Paige, K.N. 2004. Surviving the ice: northern refugia and postglacial colonization. Proc. Natl. Acad. Sci. U.S.A. 101: 10355-10359. doi: 10.1073/pnas.0401338101.

Rowe, K.C., Heske, E.J., and Paige, K.N. 2006. Comparative phylogeography of eastern chipmunks and white-footed mice in relation to the individualistic nature of species. Mol. Ecol. 15: 4003-4020. doi: 10.1111/j.1365-294X.2006.03063.x.

Swofford, D.L. 2002. PAUP*: phylogenetic analysis using parsimony (*and other methods), version 4.0b10. Sinauer Associates Inc., Sunderland, Massachusetts.

Tamura, K., Dudley, J., Nei, M., and Kumar, S. 2007. MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Mol. Biol. Evol. 24(8): 1596-1599. doi: 10.1093/molbev/msm092.

Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C., Erasmus, B.F.N., de Siqueira, M.F., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G.F., Miles, L., Ortega-Huerta, M.A., Peterson, A.T., Phillips, O.L., and Williams, S.E. 2004. Extinction risk from climate change. Nature (London), 427(6970): 145-148. doi: 10.1038/nature02121.

Walther, G.-R., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T.J.C., Fromentin, J.-M., Hoegh-Guldberg, O., and Bairlein, F. 2002. Ecological responses to recent climate change. Nature (London), 416(6879): 389-395. doi: 10.1038/416389a.

Wisely, S.M, Maldonado, J.E., and Fleische, R.C. 2004. A technique for sampling ancient DNA that minimizes damage to museum specimens. Conserv. Genet. 5(1): 105- 107. doi: 10.0123/B: COGE.0000014061.04963.da.

32

Tables Table 1. Standard molecular diversity indexes for populations of eastern and western haplotypes of Peromyscus maniculatus. N: sample size. R: haplotypic richness corrected by rarefaction, except where noted. : nucleotide diversity. k: mean number of pairwise differences between individuals in a population. Not included is one western haplotype individual from Schoolcraft population. Abbreviations used in Figure 1B are given in parentheses. Leelanau County is North Fox Island; Charlevoix includes the Beaver Islands; Cheboygan is in the LP; Ontario populations are in the general vicinity of Sault Ste. Marie; Clearwater County, shown in Figure 1A, is in Minnesota. All other populations are in the UP.

Population N R k 

Eastern Haplotypes 153 27* 2.93 +/- 1.54 0.0061 +/- 0.0036 Delta (DA) 4 2* 0.50 +/- 0.52 0.0010 +/- 0.0013 Alger (AL) 13 7.22 3.21 +/- 1.77 0.0067 +/- 0.0041 Schoolcraft (SC) 17 3.64 1.24 +/- 0.82 0.0026 +/- 0.0019 Mackinac (MA) 20 5.02 2.79 +/- 1.54 0.0058 +/- 0.0036 Chippewa (CP) 5 3* 3.00 +/- 1.87 0.0063 +/- 0.0046 Cheboygan (CH) 25 2.14 0.45 +/- 0.41 0.0009 +/- 0.0010 Charlevoix (CX) 27 3.88 2.03 +/- 1.18 0.0042 +/- 0.0027 Leelanau (LE) 15 2.00 2.38 +/- 1.37 0.0050 +/- 0.0032 Ontario 1 (ON1) 16 3.37 1.15+/- 0.78 0.0024 +/- 0.0018 Ontario 2 (ON2) 11 5.00 1.56 +/- 1.01 0.0033 +/- 0.0024 Western Haplotypes 59 20* 4.07 +/- 2.06 0.0085 +/- 0.0048 Clearwater, MN (CL) 14 2.00 1.48 +/- 0.95 0.0031 +/- 0.0022 Gogebic (GO) 9 2.18 2.28 +/- 1.38 0.0047 +/- 0.0032 Menominee (ME) 11 4.64 3.42 +/- 1.89 0.0071 +/- 0.0044 Delta (DA) 17 5.09 2.81 +/- 1.56 0.0059 +/- 0.0036 Alger (AL) 4 3* 1.67 +/- 1.22 0.0035 +/- 0.0030 Mackinac (MA) 3 2* 2.67 +/- 1.92 0.0056+/- 0.0050 Schoolcraft 1 1* N/A N/A ______*Rarefaction was not performed on haplotypic richness for small populations or for

33

haplogroup summaries.

34

Table 2. Differentiation of populations and population divisions in Great Lakes Peromyscus maniculatus. Divisions were determined by Spatial Analysis of Molecular Variance (SAMOVA) for 2, 3, or 4 groups. Using 5 groups led to further reduction of

CT.

Source of Variation Fixation P-value Percent of Variation Index

2 groups

Among all populations ST = 0.70 <0.0001 70.23

Among groups CT = 0.57 0.0068 56.64

Among populations within groups SC = 0.31 <0.0001 13.59 Within populations 29.77 3 groups

Among all populations ST = 0.70 <0.0001 70.04

Among groups CT = 0.58 0.0020 57.71

Among populations within groups SC = 0.29 <0.0001 12.34 Within populations 29.96 4 groups

Among all populations ST = 0.68 <0.0001 68.22

Among groups CT = 0.55 <0.0001 55.20

Among populations within groups SC = 0.29 <0.0001 13.02 Within populations 31.78 For 2 groups, with eastern populations only

Among all populations ST = 0.41 <0.0001 40.96

Among groups CT = 0.18 0.1085 17.80

Among populations within groups SC = 0.28 <0.0001 23.17 Within populations 59.04

35

Figure 1. Distribution of Peromyscus maniculatus gracilis haplogroups in the central Great Lakes region. A. Northeastern United States and southeastern Canada. Dashed line represents southern extent of Laurentide ice sheet, after Dyke et al. (2002). Black diamonds indicate additional samples included for context: Yancey County, North Carolina (YA) and Clearwater County, Minnesota (CL). Recent ranges of woodland subspecies Peromyscus maniculatus gracilis (light shading) and P. m. nubiterrae (dark shading) are drawn after Hall (1981). B. Study area with sampling locations and spatial haplotype structure. Circles represent populations as defined in the text, with haplotype composition indicated within. Grey in circles: eastern haplotypes; black: western haplotypes. Populations containing ancestral haplotypes DA04 () and SC105 () are indicated. Samples depicted in Lake Michigan are in the Beaver Island group, Charlevoix County, above, and on North Fox Island, Leelanau County. Population abbreviations are in Table 1.

36

Figure 2. Bayesian haplotype tree for Great Lakes Peromyscus maniculatus gracilis and related groups. Numbers above branches are partition probabilities from Bayesian analyses. Numbers below branches are bootstrap values from Maximum-Parsimony (MP), with Neighbor-Joining (NJ) bootstrap values in parentheses. Haplotypes are named after representative samples and indicate the origin population of that sample with the population abbreviations used in figure 1. Although the tree was rooted with Peromyscus leucopus for display purposes, a node between P. leucopus and all other sequences was not supported by any method. 37

Figure 3. Parsimony network for Great Lakes and North Carolina Peromyscus maniculatus, constructed using TCS. Each line represents 1 informative mutational step, with empty circles indicating presumed nodes not found in our samples. The size of each oval corresponds to the abundance of that haplotype, from N=1 (e.g., CX706) to N=27 (DA04). Spatial distribution of haplotypes DA05 () and SC105 () are shown in Figure 1.

38

Figure 4. Mismatch analysis of Peromyscus maniculatus samples. Gray: observed pairwise mismatches; white: mismatches predicted by spatial expansion model. a. eastern clade samples, n = 153, P = 0.326. b. western clade samples, n = 59, P = 0.2470. c. eastern clade and North Carolina samples, n = 160, P = 0.4280. d. all samples, n = 219, P = 0.3210.

39

Table S1. Trapping Sites (bold: geographical midpoint; parentheses: individual locations)

Alger: MP: 46.6288, -86.2110; (46.6543, -86.1848), (46.6629, -86.1629), (46.562, -86.2742), (46.636, -86.222) Delta Co.: (45.9348, -86.7543) Charlevoix: MP: 45.7563, -85.5569; (45.6163, -85.5508), (45.7244, -85.6719), (45.7925, -85.3636), (45.8369, -85.5867), (45.8113, -85.6116) Cheboygan Co.: MP: 45.2841, -84.4298; (45.2729, -84.4363), (45.2953, -84.4233) Chippewa Co.: MP: 46.4150, -84.8056; (46.4159, -84.8332), (46.4146, -84.8022), (46.4144, -84.7812) Gogebic Co.: MP: 46.2821, -89.2326; (46.2447, -89.2325), (46.3195, -89.2326) Leelanau Co.: (45.4818 -85.7778), estimated coordinates Mackinac Co.: MP: 46.0440, -84.6424; (45.9914, -84.5177), (46.0404, -84.5924), (46.0475, -84.7324), (46.0965, -84.7271) Menominee Co.: MP: 45.4614, -87.4606; (45.3707, -87.4133), (45.5003, -87.4137), (45.5132, -87.3932) Ontario 1,: (46.4910, -84.4823) Ontario 2: (47.2312, -84.6497) Schoolcraft Co.: (46.2175, -85.9697)

40

Table S2. Museum-sourced samples Origin Museum Sample Type Catalog # Yancey Co., NC UMMZ Skulls 85819, 85821, 85823, 85826, 85829, 85831, 85833, 85835, 85836 Alger Co., MI UMMZ Frozen tissue 166936, 166951, 166975, 166978, 166980, 166982, 166985, 166989, 166995, 167001, 167014, 167016, 167018, 167019, 167028, 165107 Charlevoix Co., MI UMMZ Frozen tissue 165124, 165717, 165740, 165869, 167333, 167344, 167347, 167348, 167349, 167350, 167350, 167352, 167354, 167354, 167355, 167356, 167358, 167359, 167362, 167364 Leelanau Co., MI MSUM Frozen tissue 37599, 37600, 37601, 37602, 37603, 37604, 37606, 37607, 37608, 37609, 37610, 37611, 37612, 37613, 37614 Clearwater Co., MN BMNH Skulls 18147, 18148, 18150, 18151, 18153, 18154, 18157, 18161, 18164, 18167, 18168, 18171, 18175 Abbreviations: UMMZ: University of Michigan Museum of Zoology BMNH: Bell Museum of Natural History at the University of Minnesota MSUM: The Michigan State University Museum

41

CHAPTER 3

Natural landscape fragmentation defines microsatellite genetic structure in Great

Lakes populations of deer mice (Peromyscus maniculatus gracilis)

ABSTRACT The Great Lakes impose high levels of natural fragmentation on local populations of animals in a way found in few other places within continental ecosystems. Besides being separated by major water barriers, populations of the woodland deer mouse (Peromyscus maniculatus gracilis) in Michigan encounter localized competition from a congeneric species and changing environmental conditions that have implications for the long-term distribution of this species in the region. I analyzed 10 microsatellite loci for 16 populations of P. m. gracilis distributed across 2 peninsulas and 6 islands in northern Michigan, in order to address the impacts of these various factors on the genetic structure of this highly vagile species. The results showed relatively high levels of genetic structure for this species and a significant correlation of interpopulation differentiation with separation by water, but little genetic structure and no isolation-by-distance within each of the land masses. Genetic diversity was generally high on both peninsulas, and was correlated to island size in the Beaver Island Archipelago. These results are consistent with genetic and demographic isolation of Lower Peninsula populations, despite a "soft" border to the south and human transportation networks in the region, a matter of concern given dramatic declines in P. m. gracilis abundance on the Lower Peninsula in recent years.

42

INTRODUCTION

Habitat fragmentation due to either natural or anthropogenic barriers can result in genetic drift by limiting both population size and gene flow between nearby populations (Wright 1931; Templeton et al. 1990; Gilpin 1991). Over time, this should result in inter- population differentiation that can be measured by variations in marker allele frequencies and diversity (Wright 1931). However, the robust dispersal ability of many small mammals tends to obscure the effects of landscape barriers on dispersal and genetic structure (McCullough & Chesser 1987; Mossman & Wasser 2001). In order to measure the effects of landscape barriers on the genetics of small mammal populations, one must look to areas where there are extreme geographic constraints on animal migration. The Great Lakes region, though centrally located within North America, was highly fragmented into peninsulas and islands when the lakes were formed by retreating glaciers, a process that ended about 11,000 years ago (see Dyke et al., 2002). The flora and fauna now inhabiting the region appear to have colonized it shortly thereafter, often from multiple southern refugia (Brant & Ortí 2003; Rowe et al. 2006; Taylor & Hoffman 2010). Among the colonizers were two distinct subspecies of deer mice (Peromyscus maniculatus). One subspecies is the woodland deer mouse (P. m. gracilis), a strict woodland specialist inhabiting boreal and mixed hardwood forests in extreme northern parts of the contiguous United States and in southern Canada, from the Midwest to the East Coast. The second subspecies, P. m. bairdii, is a morphologically, ecologically (Iverson et al. 1967; Dice 1933; Barbehenn & New 1957), and karyotypically (Singh & McMillan 1966) distinct prairie and farmland specialist that is not known to interbreed with P. m. gracilis in the wild. The range of P. m. bairdii is centered in the prairies of the American Midwest (Hall 1981), but its eastern edge partially overlaps the ranges of both P. m. gracilis and P. m. nubiterrae. Hooper (1942) suggested that this overlap is the result of a recent range expansion by P. m. bairdii as native forests in Michigan and Ohio were converted to farmland. No P. m. bairdii were trapped at any of the locations

43

included in this study, and for the purposes of this study it is considered to be a recent invader that does not significantly interact with P. m. gracilis. Studies of mitochondrial DNA sequences from the woodland deer mouse (P. m. gracilis) indicate that it most likely colonized the region from two refugia, resulting in the presence of two distinct lineages in modern populations (Lansman et al. 1983; Dragoo et al. 2006; Taylor & Hoffman 2010). Most of the populations in Michigan belong to an eastern lineage that probably colonized the region shortly after the glaciers retreated, about 9,000-11,000 years ago, crossing the lakes multiple times in the process (Taylor & Hoffman 2010). A second mitochondrial lineage dominates the westernmost end of Michigan’s Upper Peninsula, and also exists farther west in Minnesota. The zone marking the transition from one lineage to another does not correspond to any obvious geographical barrier (Fig. 1) and instead most likely represents contact between previously separated groups (Endler 1977; Dragoo et al., 2006; Taylor & Hoffman 2010). Therefore, the lakes do not appear to have played a significant role in structuring mouse populations during colonization. Although the Great Lakes appear to have been porous barriers on the timescale of postglacial colonization, their imposing presence in the landscape of Michigan suggests that they have the potential to act as significant genetic and demographic barriers for animal populations in the short term. The potential for the Great Lakes to deter migration is especially relevant for small mammals at the present time: many species in the region have experienced range expansions or contractions correlated to recent climate change (Long 1996; Jannett et al. 2007; Myers et al. 2009), so strong barriers could interfere with adaptive adjustments to range limits. The woodland deer mouse is among the species experiencing the most dramatic range shifts; its abundance and distribution on the LP appear to have declined sharply in the last hundred years (Myers et al. 2009). Because the southern range limit of this subspecies passes through the central LP (Hall 1981), gene flow to or from LP populations across land is highly unlikely. Therefore, the effectiveness of the Great Lakes as barriers to gene flow is relevant for the maintenance of genetic diversity and the long-term demographic persistence of these populations. Because postglacial colonization occurred on a very long timescale relative to the generation time of P. m. gracilis (thousands of years vs. 2-3 generations per year; Layne

44

1968), we should not expect that the limited role of the lakes as barriers to colonization would translate into a similarly limited role in population genetic dynamics; nuclear and mitochondrial DNA often present quite different patterns (Feulner et al 2004; Yang & Kenagy 2009). Nonetheless, North American field mouse (Peromsycus spp.) populations typically have low levels of nuclear differentiation (Mossman & Waser 2001; Vucetich et al. 2001; Anderson & Meikle 2010; Yang & Kenagy 2009), presumably because they readily disperse across apparently hostile terrain such as swampy areas or open fields (Cooke & Terman 1977; Cummings & Vessey 1994; Krohne & Hoch 1999), and can swim short distances (Sheppe 1965). Mossman & Waser (2001) found very low levels of microsatellite marker differentiation (FST ≤ 0.033) in populations of Peromyscus leucopus separated by up to 30 km, no effects due to the matrix type (woodland or farmland) separating the populations, no effect of geographic distance on differentiation, and very high levels of inter-population migration. On a much larger scale, Vucetich et al. (2001) found slightly higher, but still low, levels of differentiation in randomly amplified polymorphic DNA between populations of P. m. gracilis separated by approximately 750 km (FST ≤ 0.11) in mainland Michigan. Estimates of differentiation for island populations of P. m. gracilis have been somewhat higher (Vucetich et al. 2001;

0.07 < FST < 0.2), but comparable data has not been published for microsatellites. Therefore, I wanted to examine whether massive geographical barriers on the scale of the Great Lakes could impose genetic structure on this species, which is otherwise characterized by high levels of gene flow. I have analyzed microsatellite markers from Great Lakes populations of P. m. gracilis in order to address the roles of lakes in restricting post-colonization gene flow in this subspecies. I also address whether the phylogeographic break observed in the mitochondrial studies indicates a broader genetic distinction between groups or is restricted to mitochondrial haplotypes, due to their lower effective population size and relative lack of recombination (Lindell et al. 2008; Yang & Kenagy 2009).

MATERIALS AND METHODS

45

Sampling.— Tissue samples were obtained from 16 P. m. gracilis populations in northern Michigan (Figure 1; exact sampling locations are shown in the Appendix) using non- lethal methods described previously (Taylor & Hoffman 2010), from 2002 to 2009. Animals were handled in accordance with guidelines established by the American Society of Mammalogists (Gannon et al. 2007), using a protocol approved by the Institutional Animal Care and Use Committee of Miami University. Additional tissue samples for North Fox Island (Leelanau County), Alger County, and the Beaver Island group (Charlevoix County) were obtained from the collections of the Michigan State University Museum and the University of Michigan Museum of Zoology (Table S2).

Sample preparation and genotyping.—DNA was isolated from ear tissue using the e.Z.N.A. Tissue DNA Kit (Omega Bio-Tek, Norcross, Georgia). Microsatellites were amplified by the polymerase chain reaction (PCR) using Promega Go-Taq DNA polymerase and the Flexi buffer system. Primer sets were selected from the literature and amplified under published or experimentally determined conditions (Table 1). Typical

PCR reactions contained 20 ng template DNA, 1.5-2 mM MgCl2, and 0.2 mM each dNTP. The PCR cycle used consisted of 94o C for 2 min; 40 cycles of 30 s at 94o C, 30 s at 50-65o C, and 1 min. at 72o C; 5 min. at 72o C. Specific annealing temperatures and

MgCl2 concentrations are shown in Table 1. Forward primers labeled with the G5 dye set were obtained from Applied Biosystems (Foster City, California) and products were run on an Applied Biosystems 3130 or 3730 DNA Analyzer with the 600LIZ internal size standard (Applied Biosystems). Product peaks were identified manually using Peak Scanner v1.0 software (Applied Biosystems, Foster City, California), (Matschiner & Salzburger 2009).

Statistical analyses.—Genetic diversity was measured using allelic richness (AR), and the heterozygosity that was observed or expected under conditions of Hardy-Weinberg equilibrium (HO, HE). AR was standardized for sample size using the repeated sampling procedure implemented in HP-Rare (Kalinowski 2006). HO and HE were calculated in Arlequin v3.11 (Excoffier et al. 2005).

46

Data were examined for null alleles using the program Micro-Checker (Van Oosterhout et al. 2004). Where null alleles were detected at frequencies over 5% within a population, genotypes and allele frequencies were corrected using the method of Brookfield (1996). However, a side effect of correction for null alleles is the disruption of multilocus genotypes, because one cannot determine exactly which genotypes are homozygotes as opposed to heterozygotes incorporating a null allele (Van Oosterhout et al. 2004). Therefore, I used the original genotypes for clustering analyses, which require the information present in the unmodified multilocus genotypes. Population genetic analyses were conducted using the program GenePop version 4.0 (Rousset 2008). Significance of population differentiation, deviation from Hardy-Weinberg expectations (HWE), and linkage disequilibrium were determined using log likelihood ratio tests (Goudet et al. 1996) after adjusting for multiple comparisons using the sequential

Bonferroni correction (Rice 1989). I used pairwise estimates of FST () according to the method of Weir & Cockerham (1984), as well as the allele-size-dependent estimator ST, which incorporates a high-rate stepwise mutation model. Significance of differentiation was tested with log likelihood ratio tests (Goudet et al. 1996) in GenePop. The relationship between Lake Michigan as a barrier and genetic distance was examined using a Mantel test (Mantel, 1967) as implemented in Arlequin version 3.11 (Excoffier et al. 2005). Because minimum over-water distances were distributed in a clear multi-modal pattern (i.e., island-island distances were small while island-mainland distances were large), I represented over-water distances with the following categorical variables: (1) populations located on the same land mass; (2) populations separated by < 10 km of Lake Michigan (e.g., all comparisons between the UP and the LP or between islands) and (3) populations separated by ≥ 10 km of Lake Michigan (all island-mainland comparisons). A pairwise matrix of separation classes was constructed and used as the X matrix for comparison with FST values calculated by Arlequin. Isolation-by-distance within a given landmass was tested using a Mantel test with geographical distances as the X matrix. I examined the importance of over-water distance by combining categories two and three to produce a matrix in which the only categories were (1) not separated by Lake Michigan, and (2) separated by Lake Michigan.

47

For comparison with mitochondrial genetic structure, I used a Mantel test to compare a modified FST matrix to a pairwise distance matrix derived from previously described mitochondrial D-loop sequences (Taylor & Hoffman 2010). To allow direct comparison with that study, the Alpena, Beaver Island, and High Island populations were removed from this analysis. We also examined overall genetic structure using two Bayesian clustering algorithms. First, I examined the number of genetically supported populations using the program STRUCTURE, version 2.3.3 (Pritchard et al. 2000), which assigns individuals to a user-defined number of genetic clusters k based on similarities among multilocus genotypes. I performed three replicate runs with a burn-in of 100,000 and 10,000,000 subsequent iterations for each value of k from 1 to 10. I selected the number of populations that best fit the data by using the metric derived from the second-order rate of change of Ln P (D) (Evanno et al. 2005). I then performed ten long runs of 2,000,000 iterations with k fixed at 5, and used the results with the least negative value of Ln P(D) for assignment of individuals to clusters. I also used assignment tests implemented in the program GeneClass2 (Piry et al. 2004) to examine population structure and the possibility of contemporary gene flow. Higher levels of inter-population structure increase the correct assignment of individuals to their source populations (Paetkau et al. 1995), and potential immigrants can be identified as individuals with genotypes more likely to be found in non-source populations (Rannala & Mountain 1997).

RESULTS I genotyped 183 individuals from 16 populations at 11 loci. All loci were polymorphic in all population, with the lone exception of PO-09 on Squaw Island, which was fixed for a single allele. Linkage disequilibrium was significant for only one out of 880 pairwise comparisons between populations. By population, 13 of 176 tests resulted in significant deviations from HWE, spread across 7 populations. Four loci deviated in the Delta (UP) population, and 2 in the Webb Road (LP) population; no other population deviated at more than one locus. As heterozygote deficiencies were not consistent within populations, they are more likely to be due to null alleles than to an overall absence of HWE within any specific population.

48

Genetic diversity was high in all mainland populations, with H E = 0.80 and ĀR =

6.4 on the UP and H E = 0.82 and ĀR = 6.6 on the LP (Table 2). Diversity was much lower on the islands, averaging H E = 0.58 and ĀR = 3.6. Allelic richness for all island populations except Beaver Island 2 was significantly different from that of mainland populations (Table 2; ANOVA: P < 0.001). HE followed a similar trend, but only the Squaw and North Fox Island heterozygosities were significantly lower than the others

(Table 2; P < 0.001). Overall levels of differentiation were moderate and significant (ST

= 0.25; FST = 0.17; P < 0.0001). The lowest pairwise differences were significant under the adjusted threshold value of P = 0.00042. Differentiation was lowest among several population pairs within the peninsulas, such as between the Mackinac and Delta populations on the UP (ST = -0.016; FST = 0.041; P < 0.0001). Differentiation was generally high among the islands (ST = 0.58; FST = 0.27; P < 0.0001). On the Upper

Peninsula as a whole, genetic differentiation was low (overall ST = 0.044; FST = 0.036; P

< 0.0001). A similar trend was observed across the Lower Peninsula (ST = 0.014; FST = 0.082; P < 0.0001). Although differentiation between pairs of populations within the UP was often significant, values were typically low (average pairwise FST = 0.037) and showed no pattern of isolation-by-distance (r = 0.19; P = 0.27). Differentiation between populations separated by water was always statistically significant, and on average much higher (average pairwise ST = 0.27; FST = 0.19) than seen within the same land mass (Figure 2). Correlation between separation by water and genetic differentiation was highly significant, whether over-water distance was included in the separation categories (r = 0.61; P < 0.0001) or all populations separated by water were included in the same category (r = 0.63; P < 0.0001).

However, no correlation was observed between microsatellite FST values and mitochondrial genetic distances (r = 0.092; P = 0.0.36) produced for another study (Chapter 2; Taylor & Hoffman 2010). Similarly, when I considered whether populations on either side of a genetic break observed in that study (Figure 1) were more different than populations on the same side of the break, I did not observe any difference (r = - 0.13; P = 0.74)

49

Analyses using the STRUCTURE clustering algorithm (Pritchard et al. 2000) and the test developed by Evanno et al. (2005) indicated the presence of five distinct clusters within our sampling area (Figure 3). Populations on each peninsula clustered together, while the remaining clusters corresponded to individual or paired islands. Most individuals within a given population fall into the same predominant cluster; the small number of exceptions could indicate recent gene flow, especially from the Upper Peninsula (gray bars; cluster 2) into the Lower Peninsula and other populations. In the assignment tests performed in GeneClass2, 76.5% of individuals were assigned to the correct source population. Almost all mis-assignments (39/44) involved the assignment of an individual to another population on the same landmass. Five individuals were incorrectly assigned to a different landmass: 1 UP individual (GO) was assigned to an island (BE2), two individuals form the LP were assigned to the UP (OS1-Men; WEBB- CP), one from the LP was assigned to an island (WEBB-BE2) and one from Squaw Island was assigned to Beaver Island 2.

DISCUSSION

Natural barriers and genetic structure in Peromyscus species Populations of P. m. gracilis on Michigan's Lower Peninsula are geographically separated from surrounding populations in two ways. First, Lake Michigan and Lake Huron divide these populations from those in the Upper Peninsula and those in Ontario, respectively. Second, the southern range limit of this subspecies bisects the Lower Peninsula of Michigan latitudinally, roughly along the 45th parallel (Hall 1981; Myers et al. 2009), potentially blocking migration or gene flow around the shores of the Great Lakes. Therefore, between-peninsula genetic structure measures the effectiveness of both the Great Lakes to the north and the inhospitable terrain to the south as genetic barriers; if both function as barriers, the Lower Peninsula of Michigan should constitute a virtual island for populations of this deer mouse. All analyses of our microsatellite data do in fact indicate a genetic structure consistent with the isolation of P. m. gracilis populations on the Lower Peninsula from all other populations tested.

50

The intra-peninsular structure I find in Great Lakes populations of P. m. gracilis is consistent with a general lack of isolation-by-distance or geographically defined genetic structure in Peromyscus populations elsewhere (Mossman & Waser 2001; Vucetich et al. 2001; Yang & Kenagy 2009; Anderson & Meikle 2010). Gene flow among Peromyscus populations has been shown to be robust despite separations of hundreds of kilometers (Vucetich et al. 2001), or the intervention of putative barriers such as unsuitable habitat (Mossman & Waser 2001; Anderson & Meikle 2010) or major rivers (Mossman & Waser 2001). This lack of structure is consistent with the often- extraordinary vagility of these mice. Even though the typical dispersal of female Peromyscus between birth and breeding is likely around 250 m (Blair 1940; Neigel et al. 1991; Krohne et al. 1984), males tend to disperse slightly farther (Dice and Howard 1951; Keane 1990), and individuals of both sexes are known to cover large distances very quickly. Krohne et al. (1984) documented an individual traveling nearly 700 m over two nights; Maier (2002) found different tagged females 7 and 15 km from their original capture locations, after eight months or a single month, respectively. In this context, the genetic separation between the peninsulas provides a rare example of a geographical barrier inhibiting gene flow in these highly vagile mice. While I consistently find differentiation between deer mouse populations on different peninsulas, our results indicate much higher levels of differentiation among the Lake Michigan islands, even though the over-water distances separating many island pairs are less than the distance separating the two peninsulas. This finding is highlighted by the fact that including over-water distance in our categorical Mantel tests did not improve the correlation between FST and separation by water. Similar microsatellite genetic disjunction of populations separated by physical barriers has been demonstrated for wheat rust (Kolmer & Ordoñez 2007), jaguars separated by the Darien Straits or the Amazon River (Eizirik et al. 2001), and Columbia spotted frogs separated by mountain ridges (Funk et al. 2005). Populations of Peromyscus maniculatus and their parasites have also shown genetic structure in randomly amplified polymorphic DNA when completely separated by water (Vucetich et al. 2001; Landry and Lapointe 1999). However, other presumed physical barriers have shown no obvious relationship to genetic barriers, as in the case of P. leucopus separated by the Wabash River (Mossman

51

& Waser 2001) or prairie dog populations separated by mountains or rivers (Chesser 1983). Because the island populations are historically related to both UP and LP populations, as shown by mitochondrial DNA analyses (Taylor & Hoffman 2010), the very high levels of differentiation observed among island populations may indicate the islands are more isolated from each other (with respect to gene flow) than are the two peninsulas, and/or that island populations are more susceptible to the effects of genetic drift. Because a bridge links the two peninsulas, levels of human traffic between them are likely higher than traffic from the peninsulas to the islands, or between islands. Therefore accidental transportation of mice is more likely between the UP and LP than to or among the islands. Indeed, deer mice have been documented as stowaways on grain trucks in Colorado and are capable of surviving long trips (Baker 1994), so significant transportation across the bridge is possible. However, the number of incorrect assignments involving island and mainland populations (2) was identical to that of incorrect assignments between the two peninsulas. Combined with the higher levels of differentiation I observed between island populations, these results suggest that low effective population sizes could be more important than dispersal for island mice. I did find much lower levels of genetic diversity on the islands, and the very small size of some of the islands makes it unlikely that they can sustain large mouse populations.

Population genetic processes and genetic diversity

With few exceptions, our results indicate that individual populations are close to HWE with respect to the loci studied. The deviation from HWE of the entire sample set for each locus (Table 1) is expected in a region with strong genetic structure; lower heterozygosity in populations with fewer or different alleles results in a global heterozygote deficiency compared to HWE expectations and hence high FST values (Wright 1931). The loci selected do not exhibit linkage disequilibrium, as expected for independently segregating loci representing different segments of the P. m. gracilis genome. The overall genetic diversity reported here is comparable to levels reported in studies of other Peromyscus species (Mossman & Waser 2001; Yang & Kenagy 2009;

52

Anderson & Meikle 2010). I did find some evidence for null alleles, in the form of possible null homozygotes (un-genotyped individuals) and heterozygote deficiencies that were restricted to specific loci, rather than distributed throughout the genome (as would be expected for population-wide deviations from HWE). Null alleles have been observed at some of these loci in a study of P. leucopus (Anderson & Meikle 2010), but interspecific differences complicate direct comparisons between different studies. Our finding of low microsatellite diversity on the Lake Michigan islands is consistent with studies of mitochondrial DNA (Taylor & Hoffman 2010; Taylor & Hoffman in preparation) and allozymes (Meagher 1999) in island P. m. gracilis, and similar to patterns observed in mitochondrial DNA of garter snakes (Thamnophis sirtalis; Placyk et al. 2007) on the same islands. However, the relatively high levels of genetic diversity I found on the Lower Peninsula contrast markedly with the number of mitochondrial haplotypes observed in this area (Taylor & Hoffman 2010). This distinction may be important, because LP populations of P. m. gracilis have long been in decline (Myers et al. 2009), and low genetic diversity has been associated with reduced survivorship, depressed growth rates, and increased parasite load in Peromsycus species (Brewer et al. 1990; Jimenez et al. 1994; Meagher 1999). The differences between mitochondrial and nuclear genetic diversity here could indicate a reduction in effective population size that has not yet been severe enough to affect microsatellite diversity, which has a four-fold higher effective population size than mitochondrial DNA and should therefore be less sensitive to population decline. However, other factors could also explain the difference. For instance, male-biased dispersal, which is found in many Peromsycus species (Dice and Howard 1951; Keane 1990), could maintain nuclear genetic diversity at higher levels than found in the maternally inherited mitochondrial DNA.

Microsatellite genetic structure and the phylogeographic context

The genetic structure I have described in this study differs from the mitochondrial patterns in more ways than in those detailed above. First, the differentiation between mice on the two peninsulas is more pronounced in the present study. Second, and more

53

dramatic, is the absence of an east-west genetic structure (or phylogeographic break) on the Upper Peninsula that is comparable to the one observed in mitochondrial DNA analyses (Lansman et al. 1983, Dragoo et al. 2006; Taylor & Hoffman 2010). The lack of any statistical correlation between FST values from the two analyses underscores this point. Such cytonuclear discordance has been observed in a number of studies at both the interspecific and intraspecific levels of organization (Lindell et al. 2008; Yang & Kenagy 2009). One possible explanation for the results of the present study is that gene flow is slower at mitochondrial loci because smaller effective populations can sustain less diversity and are less likely to maintain introduced alleles, whereas male-biased dispersal speeds dissemination of nuclear loci (Yang & Kenagy 2009). However, unlike prior studies documenting intraspecific cytonuclear discordance (Lindell et al. 2008; Yang & Kenagy 2009), I describe a system in which landscape fragmentation and geographical range limits apparently restrict nuclear gene flow. In this case, the results paint two temporally distinct pictures of the development of genetic structure in Great Lakes populations of P. m. gracilis. Prior studies describe the post- glacial colonization of Ontario, the Lake Michigan islands, the Lower Peninsula of Michigan, and the eastern Upper Peninsula by a single mitochondrial lineage (Lansman et al. 1983, Dragoo et al. 2006; Taylor & Hoffman 2010). The second picture, based on microsatellite data, describes the post-colonization differentiation of populations of deer mice isolated on the peninsulas and islands of the Great Lakes region. The distinctiveness of these two pictures suggests that the functional properties of the relevant barriers are important. For instance, small numbers of cross-barrier colonization events over hundreds or even thousands of years would suffice to distribute a lineage across the Great Lakes, but higher levels of migration are likely necessary to prevent genetic differentiation. The differences I have observed suggest that the gene flow rates permitted by the Great Lakes fall somewhere in between, resulting in related but distinct populations on their various shores.

54

REFERENCES

Anderson CS, Meikle DB. 2010. Genetic estimates of immigration and emigration rates in relation to population density and forest patch area in Peromyscus leucopus. Conservation Genetics. Doi: 10.1007/s10592-009-0033-8

Baker AEM. 1994. Stowaway transport rates of house mice (Mus domesticus) and deermice (Peromyscus maniculatus). Proceedings of the 16th Vertebrate Pest Conference, Davis, California. pp. 106-112.

Barbehenn KR, New JG. 1957. Possible natural intergradation between prairie and forest deer mice. J. Mammal. 38(2): 210-218

Blair WF. 1940. A study of prairie deermouse populations in southern Michigan. American Midland Naturalist 24:273-305.

Brant SV, Ortí G. 2003. Phylogeography of the Northern short-tailed shrew, Blarina brevicauda (Insectivora: Soricidae): past fragmentation and postglacial recolonization. Mol. Ecol. 12:1435-1449.

Brewer, B. A., Lacy, R. C., Foster, M. L., Alaks, G. 1990. Inbreeding depression in insular and central populations of Peromyscus mice. Journal of Heredity 81:257-266.

Brookfield JFY. 1996. A simple new method for estimating null allele frequency from heterozygote deficiency. Molecular Ecology 5:453-455.

Chesser RK. 1983. Genetic variability within and among populations of the black-tailed prairie dog. Evolution 37:320-331.

55

Cooke JA, Terman CR. 1977. Influence of displacement distance and vision on homing behavior of the white-footed mouse (Peromyscus leucopus noveboracensis). Journal of Mammalogy 58:58-66.

Cummings JR, Vessey SH. 1994. Agricultural influences on movement patterns of white-footed mice (Peromyscus leucopus). American Midland Naturalist 132:209-218.

Dice LR. 1933. Fertility relationships between some of the species and subspecies of mice in the genus Peromyscus. Journal of Mammalogy 14:298-305.

Dice L, Howard W. 1951. Distance of dispersal by prairie deermice from birthplaces to breeding sites. Contributions of the Laboratory of Vertebrate Biology, University of Michigan 50:1-15.

Dragoo JW, Lackey JA, Moore KE, Lessa EP, Cook JA, Yates TL. 2006. Phylogeography of the deer mouse (Peromyscus maniculatus) provides a predictive framework for research on hantaviruses. Journal of General Virology 87:1997-2003.

Dyke AS, Andrews JT, Clark PU, England JH, Miller GH, Shaw J, Veillette JJ. 2002. The Laurentide and Innuitian ice sheets during the Last Glacial Maximum. Quaternary Science Reviews 21:9-31.

Endler, JA. 1977. Geographic variation, speciation, and clines (MPB-10). Princeton University Press, Princeton, New Jersey. Eizirik E, Kim J-H, Menotti-Raymond M, Crawshaw PG Jr., O'Brien SJ, Johnson WE. 2001. Phylogeography, population history and conservation genetics of jaguars (Panthera onca, Mammalia, Felidae). Molecular Ecology 10:65-79.

Evanno G, Regnaut S, Goudet J. 2005. Detecting the number of clusters of individuals using the software Structure: a simulation study. Molecular Ecology 14: 2611-2620.

56

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.

Feulner PGD, Bielfeldt W, Zachos FE, Bradvaravic J, Eckert I, Hartl GB. 2004. Mitochondrial DNA and microsatellite analyses of the genetic status of the presumed subspecies Cervus elaphus montanus (Carpathian red deer). Heredity 93:299-306.

Funk CW, Blouin MS, Corn PS, Maxell BA, Pilliod DS, Amish S, Allendorf FW. 2005. Population structure of Columbia spotted frogs (Rana luteiventris) is strongly affected by the landscape. Molecular Ecology 14:483-496.

Gannon WL, Sikes RS. The Animal Care and Use Committee of the American Society Of Mammalogists. 2007. Guidelines of the American Society of Mammalogists for the use of wild mammals in research. Journal of Mammalogy 88:809-823. doi:10.1644/06- MAMM-F-185R1.1.

Gilpin M. 1991. The genetic effective size of a population. Biological Journal of the Linnean Society 42:165-175.

Goudet J, Raymond M, de Meeüs T, Rousset F. 1996. Testing differentiation in diploid populations. Genetics 144: 1931-1938.

Hall ER. 1981. The Mammals of North America. John Wiley & Sons, Inc., New York.

Hooper ET. 1942. An effect on the Peromyscus maniculatus rassenkreis of land utilization in Michigan. Journal of Mammalogy 23:193-196.

Irwin DE. 2002. Phylogeographic breaks without geographic barriers to gene flow. Evolution 56:2383-2394.

57

Iverson SL, Seabloom RW, and Hnatiuk JM. 1967. Small-mammal distributions across the prairie-forest transition of Minnesota and North Dakota. American Midland Naturalist 78: 188-197.

Jannett FJ, Jr, Broschart MR, Grim LH, Schaberl JP. 2007. Northerly range extensions of mammalian species in Minnesota. American Midland Naturalist 158:168-176.

Jimenez JA, Hughes KA, Alaks G, Graham L, Lacy RC. 1994. An experimental study of inbreeding depression in a natural habitat. Science 266:271-273.

Kalinowski ST. 2006. HP-Rare: A computer program for performing rarefaction on measures of allelic diversity. Molecular Ecology Notes 5:187-189. doi:10.1111/j.1471- 8286.2004.00845.x.

Keane B. 1990. Dispersal and inbreeding avoidance in the white-footed mouse, Peromyscus leucopus. Animal Behaviour 40:143-152.

Kolmer JA, Ordoñez ME. 2007. Genetic differentiation of Puccina triticina populations in Central Asia and the Caucasus. Phytopathology 97:1141-1149.

Krohne DT, Dubbs BA, Baccus R. 1984. An analysis of dispersal in an unmanipulated population of Peromyscus leucopus. American Midland Naturalist 12:146-156.

Krohne T, Hoch GA. 1999. Demography of Peromyscus leucopus populations on habitat patches: the role of dispersal. Canadian Journal of Zoology 77:1247-1253.

Landry P-A, Lapointe F-J. 1999. The genetic heterogeneity of deer mouse populations (Peromyscus maniculatus) in an insular landscape. Researches on Population Ecology 41:263-268.

58

Lansman RA, J C Avise CF, Aquadro, JF Shapira, Daniel SW. 1983. Extensive genetic variation in mitochondrial DNA’s among geographic populations of the deer mouse, Peromyscus maniculatus. Evolution, 37:1-16.

Layne, J.N. 1968. Ontogeny. In Biology of Peromyscus (Rodentia). Edited by J.A. King. Special Publication of the American Society of Mammalogists, Number 2, Stillwater, Okla. pp. 148-253.

Lindell J, Menendez-de la Cruz FR, Murphy RW. 2008. Deep biogeographical history and cytonuclear discordance in the black-tailed brush lizard (Urosaurus nigricaudus) of Baja California. Biological Journal of the Linnean Society 94:89-104.

Long CA. 1996. Ecological replacement of the deer mouse, Peromyscus maniculatus, by the white-footed mouse, P. leucopus, in the Great Lakes region. Canadian Field- Naturalist 110:271-277.

Maier TJ. 2002. Long-distance movements by female white-footed mice, Peromyscus leucopus, in extensive mixed-wood forest. Canadian Field-Naturalist 116:108-111.

Mantel N. 1967. The detection of disease clustering and a generalized regression approach. Cancer Research 27:209-220.

Matschiner M, Salzburger W. 2009. TANDEM: integrating automated allele binning into genetics and genomics workflows. Bioinformatics Advance Access doi:101.1093/bioinformatics/btp303.

Meagher S. 1999. Genetic diversity and Capillaria hepatica (Nematoda) prevalence in Michigan deer mouse populations. Evolution 53:1318-1324.

Mossman CA, Waser PM 2001. Effects of habitat fragmentation on population

59

genetic structure in the white-footed mouse (Peromyscus leucopus). Canadian Journal of Zoology 79:285-295.

McCullough DA, Chesser RK. 1987. Genetic variation among populations of the Mexican prairie dog. Journal of Mammalogy 68:555-560.

Myers P, Lundrigan BL, Hoffman SMG, Haraminac AP, Seto SH. 2009. Climate- induced changes in the small mammal communities of the northern Great Lakes Region. Global Change Biology 15:1434-1454.

Neigel JE., Ball RM Jr., Avise JC. 1991. Estimation of single generation migration distances from geographic variation in animal mitochondrial DNA. Evolution 45:423- 432.

Paetkau D, Calvert W, Stirling I, Strobeck C. 1995. Microsatellite analysis of population structure in Canadian polar bears. Molecular Ecology 4:347-354.

Piry S, Alapetite A, Cornuet J-M, Paetkau D, Baudouin L, Estoup A. 2004. GENECLASS2: A software for genetic assignment and first-generation migrant detection. Journal of Heredity 95:536-539.

Placyk, J. S. Jr., Burghardt, G. M., Small, R. L., King, R. B., Casper, G. S., and J. W. Robinson. 2007. Post-glacial recolonization of the Great Lakes region by the common gartersnake (Thamnophis sirtalis) inferred from MtDNA sequences. Molecular Phylogenetics and Evolution 43: 452-467.

Pritchard JK, Stephens M, Donnelly P. 2000. Inference of population structure using multilocus genotypes. Genetics 155:945-959.

Rannala B, Mountain JL. 1997. Detecting immigration by using multilocus genotypes. Proceedings of the National Academy of Sciences, USA 94:9197-9201.

60

Rice WR. 1989. Analyzing tables of statistical tests. Evolution 43:223-225.

Rousset F. 2008. GenePop '007: a complete re-implementation of the GenePop software for Windows and Linux. Molecular Ecology Resources 8:103-106.

Rowe KC, Heske EJ, and Paige KN. 2006. Comparative phylogeography of eastern chipmunks and white-footed mice in relation to the individualistic nature of species. Molecular Ecology 15: 4003-4020.

Sheppe W. 1965. Island populations and gene flow in the deer mouse, Peromyscus leucopus. Evolution 19:480-495.

Singh RP, McMillan DB. 1966. Karyotypes of three subspecies of Peromyscus. Journal of Mammalogy 47:261-266.

Slatkin M. 1995. A measure of population subdivision based on microsatellite allele frequencies. Genetics 139: 457-462.

Taylor ZS, Hoffman SMG. 2010. Mitochondrial DNA genetic structure transcends natural boundaries in Great Lakes populations of deer mice (Peromyscus maniculatus gracilis). Canadian Journal of Zoology 88:404–415, doi: 10.1139/Z10-010.

Templeton, AR, Shaw K, Routman E, Davis SK. 1990. The genetic consequences of habitat fragmentation. Annals of the Missouri Botanical Garden 77:13-27.

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.

Vucetich LM, Vucetich JA, Joshi CP, Waite TA, Peterson RO. 2001.

61

Genetic (RAPD) diversity in Peromyscus maniculatus populations in a naturally fragmented landscape. Molecular Ecology 10:35-40.

Weir BS, and Cockerham CC. 1984. Estimating f-statistics for the analysis of population structure. Evolution 38:1358-1370.

Wright S. 1931. Evolution in Mendelian populations. Genetics 16:97-159.

Yang D-S, Kenagy GJ. 2009. Nuclear and mitochondrial DNA reveal contrasting evolutionary processes in populations of deer mice (Peromyscus maniculatus). Molecular Ecology 18:5115-5125.

62

Table 1. Microsatellite loci selected for analysis. References are indicated by superscripts: 1, Prince et al. 2002; 2, Schmidt 1999; 3, Chirhart et al. 2000. Global heterozygosities for each locus are indicated. ______

Locus [MgCl2] Temp. Size Range Alleles HE HO ______PO-091 1.5 mM 58o C 141-247 17 0.82 0.68 PO-211 2 mM 55o C 117-257 31 0.84 0.60 PO-261 2 mM 56o C 134-258 35 0.87 0.57 PO-351 1.5 mM 59o C 237-329 27 0.81 0.45 PO-401 1.5 mM 59o C 219-303 33 0.94 0.58 PO3-681 1.5 mM 58o C 230-288 27 0.92 0.73 PO3-851 2 mM 58o C 186-258 18 0.47 0.32 PLGT152 1.5 mM 58o C 240-272 14 0.84 0.60 PLGT-622 2 mM 53o C 131-189 20 0.88 0.64 Pml-01 3 2 mM 53o C 122-194 23 0.93 0.78 Pml-04 3 2 mM 50o C 190-296 18 0.90 0.73 ______

63

Table 2. Molecular diversity indexes for populations of Peromyscus maniculatus. N: sample size (individuals). Ā: average number of alleles per locus. ĀR: allelic richness corrected by rarefaction with a sample size of 7 individuals. ĀR T: Significance groupings according to Tukey's LSD; populations not sharing a letter designation are significantly different. HE: expected heterozygosity. HO: observed heterozygosity. ______

Population N Ā ĀR ĀR T HE HE T HO ______Upper Peninsula 80 8.11 6.38 0.80 0.74 Alger (AL) 19 10.63 6.92 A 0.79 AB 0.66 Chippewa (CP) 13 7.91 6.55 A 0.78 AB 0.75 Delta (DA) 14 8.91 6.75 A 0.81 AB 0.67 Gogebic (GO) 8 8.91 6.57 A 0.80 AB 0.72 Mackinac (MA) 9 7.27 6.53 A 0.82 AB 0.77 Menominee (ME) 4 4.73 N/D N/D 0.78 AB 0.82 Schoolcraft (SC) 13 8.45 6.62 A 0.81 AB 0.76 Lower Peninsula 29 8.24 6.55 0.82 0.67 Alpena (AP) 11 10.64 6.55 A 0.82 AB 0.67 Osmun 1 (OS1) 6 6.09 N/D N/D 0.83 A 0.74 Webb Road (WR) 12 8.00 6.38 A 0.79 AB 0.62 Islands 74 4.20 3.64 0.58 0.52 Beaver Is. 1 (BE1) 7 3.82 3.82 B 0.59 ABC 0.68 Beaver Is. 2 (BE2) 9 5.36 4.94 AB 0.74 ABC 0.66 High Is. (HI) 16 4.55 3.63 B 0.58 ABC 0.52 North Fox Is. (NFI) 20 4.73 3.49 B 0.61 BC 0.47 Whiskey Is. (WH) 11 3.09 2.81 B 0.47 C 0.42 Squaw Island (SQ) 11 3.64 3.17 B 0.51 C 0.39 ______

64

FIGURES

A.

DE AL

GO AP ME SC

WR,OS1

Squaw B. Whiskey

High

Beaver I

Beaver II

North Fox

Figure 1. Trapping locations for Peromyscus maniculatus gracilis in the Great Lakes region. A. Northeastern United States and southeastern Canada. Double line indicates the approximate location of phylogeographic split from Taylor and Hoffman (2010). Inset box indicates Lake Michigan Islands. B. Sampled islands in Lake Michigan.

65

0.45

0.4 A. 0.35

0.3

0.25 ST

F 0.2

0.15

0.1

0.05

0 No Open B. water water

0.45 0.4 0.35 0.3

0.25 ST

F 0.2 0.15 0.1 0.05 0 0 100 200 300 400 500 Distance (km)

Figure 2. Spatial distribution of genetic structure for Peromyscus maniculatus gracilis.

A. Pairwise FST as a function of terrestrial Euclidean distance. B. Pairwise FST values arranged by the presence or absence of intervening Lake Michigan.

66

1.00

0.80 Cluster 1 0.60 Cluster 2 Cluster 3 0.40 Cluster 4 Cluster 5 Cluster Assignment Cluster 0.20

0.00

HI

AL

CP DA SC AP

SQ

GO MA ME NFI

WR WH

BE1 BE2 OS1 UP LP Islands

Figure 3. Clustering of Michigan populations of Peromyscus maniculatus gracilis by STRUCTURE. Bars represent the average assignment of individuals in a population to the indicated clusters. Population abbreviations are given in Table1.

67

CHAPTER 4

Landscape-scale fragmentation and genetic structure in populations of the northern

white-footed mouse (Peromyscus leucopus noveboracensis)

Abstract

Much of the north central United States can be characterized by dramatic gradients in habitat availability, climate, and species distribution. Small mammals whose ranges traverse this area therefore face a variety of possible constraints on effective population size and its genetic consequences. I analyzed microsatellites markers and landscape characteristics in order to evaluate their relationship for populations of the white-footed mouse (Peromyscus leucopus) along a transect from southern Ohio to northern Michigan. Genetic diversity increased to the north, and was significantly correlated with habitat availability within a 0.5 km radius of population centers (r2 =

0.78; P = 0.0002). Inter-population differentiation measured as FST values did not show isolation-by-distance (r2 = 0.0040; P = 0.43), but did decrease towards the north (r2 = 0.26; P = 0.025). Thus, this study provides an unusual example of spatial structuring of Peromyscus populations within a mainland landscape. The observed gradients in diversity and inter-population differentiation were in the opposite direction to what would be expected if postglacial expansion or range constraints were determining gene flow; instead, the gradients were consistent with habitat availability being the major constraint on effective population size in this system.

68

Introduction

Habitat loss is one of the most pervasive threats to biodiversity around the world, in both developed and developing nations. Documented biological effects of habitat destruction and fragmentation range from the reproductive isolation of plants (Aguilar et al. 2006), to population decline in amphibians (Kolozsvary & Swihart 1999; Becker et al. 2007), to reductions in species richness in grassland birds (Herkert 1994), arthropods (e.g. Golden & Crist 1999; Gonzalez & Chaneton 2002) and other taxa (see Debinski & Holt 2000; Fahrig 2003). In areas where the native habitat is heavy forest, such as eastern North America, fragmentation can be extreme and visually dramatic, resulting in small forest stands surrounded by roads, agricultural land, or buildings. Inspired by island biogeographic theory (MacArthur & Wilson 1967), researchers have often treated the small remaining woodlands or other habitat patches as islands of habitat in a sea of uninhabitable matrix (Diamond 1975), with the expectation that patterns of species richness would parallel the well-established patterns drawn from studies of true islands (Darlington 1957; Diamond 1969; Simberloff & Wilson, 1969). Specifically, habitat patches are expected to support higher levels of species diversity as their size increases and their distance from other patches decreases. These predictions do appear to describe species richness fairly well in arthropods living in fragmented mainland habitats, though they apply less clearly to more vagile and generalist animals (reviewed in Debinski & Holt 2000). The predicted effects of decreased habitat availability and increasing isolation on species richness have clear parallels at the intraspecific level, where they should result in reduced genetic diversity and increased inter-population differentiation, respectively (Gilpin 1991). Such genetic effects of fragmentation have been demonstrated in habitat specialists with high resource demands, including fishers (Wisely et al. 2004), wolverines (Kyle & Strobeck 2001) and brown bears (Paetkau et al. 1998). The island biogeographic paradigm would appear especially apt for the white- footed mouse, represented in much of the American Midwest by the woodland subspecies Peromyscus leucopus noveboracensis (hereafter P. leucopus). In agriculturally important regions, including most of Ohio, Indiana, and Illinois, habitat conversion for farming and human habitation has removed the vast majority of the woodland habitat that housed

69

these mice prior to European colonization. The secondary forest found in small private woodlots and parks supports a number of woodland animal species, including P. leucopus, that are relatively rare in the surrounding agricultural fields or human residential areas (Getz 1961). Therefore, these species should show population genetic isolation related to overall habitat availability in these areas of highly fragmented habitat (Gilpin 1991; Hoffmann & Blows 1994; Kyle & Strobeck 2001). By contrast, mice should exhibit higher diversity and less differentiation in areas such as the northern edge of the Midwest, where much of the land has been allowed to return to forest following large-scale abandonment of unproductive agricultural holdings in the early 1900's (Botti & Moore 2006). While habitat availability patterns predict higher levels of genetic diversity and lower differentiation for P. leucopus at the northern edge of the Midwest, the reality is likely to be more complicated. Most importantly, studies of woodlot P. leucopus populations in the southern Ohio and Indiana have shown essentially no spatial structuring associated with rivers or with isolation-by-distance (Mossman & Waser 2001; Anderson & Meikle 2010), even though mouse populations on true islands exhibit significant structure congruent with landscape boundaries (Landry & Lapointe 1999; Vucetich et al. 2001; Taylor & Hoffman, in preparation). If these studies indicate true insensitivity to terrestrial habitat boundaries rather than merely reflecting robust migration over short distances, the island paradigm is not likely to be useful for this highly vagile species on the mainland (Lawlor 1998; Debinski & Holt 2000). The second complicating factor is that P. leucopus approaches its northern range limit in northern Michigan, possibly due to poor overwinter survivorship (Wolff 1996; Myers et al. 2004; 2009). If proximity to this range limit results in a reduced effective population size (the "core-periphery hypothesis"--see below), genetic diversity in northern Michigan should be lower than expected from habitat availability alone (Hoffmann & Blows 1994; Brown et al. 1995). I investigated the role of habitat availability in structuring the genetics of P. leucopus using microsatellite loci and landscape data analyzed with geographical information system (GIS) software. Whereas other studies of Peromyscus have examined genetic structure either at relatively local scales or without respect to overall habitat

70

availability (Mossman & Waser 2001; Root et al. 2003; Anderson & Meikle 2010, I analyzed genetic patterns along a 700-km transect from southwestern Ohio to northern Michigan. This is the first study of Peromyscus to address genetic variability at such a scale and with explicit reference to landscape variability and connectivity. I examined land cover type and distribution at three spatial scales based on known behavioral properties of these mice, in order to investigate the effects of habitat type on genetic diversity and inter-population differentiation. I expect the results to illuminate the applicability of island models to genetic structure in mice as well as for similar highly vagile animals.

Materials and Methods

Sampling.—Tissue samples were obtained from 12 P. leucopus populations in Michigan and Ohio (Figure 1; exact sampling locations are shown in Appendix 1), from 2002 to 2009. I trapped 8 of these populations directly, using non-lethal traps (H. B. Sherman Traps, Tallahassee, Florida) baited with whole oats that were placed in the evening and checked the following morning. A small piece of ear tissue was removed with sterile dissecting scissors from each animal caught, for marking and DNA extraction, before the animal was released. Animals were handled in accordance with guidelines established by the American Society of Mammalogists (Gannon et al. 2007), using a protocol approved by the Institutional Animal Care and Use Committee of Miami University. DNA samples from the Jericho, Bachelor, and Richardson populations in southwestern Ohio (Anderson & Meikle, 2010) were obtained from Christine Anderson (Miami University). DNA samples from the Carter Woods population in northern Ohio were obtained from Stephen H. Vessey (Bowling Green State University).

Sample preparation and genotyping. —We isolated DNA from ear tissue using the e.Z.N.A. Tissue DNA Kit (Omega Bio-Tek, Norcross, Georgia). I selected microsatellite primer sets from the literature and amplified the markers by the polymerase chain reaction (PCR) using Promega Go-Taq Flexi Polymerase system (Promega Corporation, Madison, Wisconsin) under empirically determined conditions (Table 1). Typical reactions contained 20 ng template DNA, 1.5-3.0 mM MgCl2, and 0.2 mM each dNTP in 71

15 l total volume. The PCR cycle consisted of 94o C for 2 min; 40 cycles of 30 s at 94o C, 30 s at 50-63.3o C, and 1 min. at 72o C; 5 min. at 72o C. Forward primers labeled with the G5 dye set were obtained from Applied Biosystems (Foster City, California) and products were run on an Applied Biosystems 3730 DNA Analyzer with the 600LIZ internal size standard (Applied Biosystems, Foster City, California). Product peaks were identified manually, using Peak Scanner v1.0 software (Applied Biosystems, Foster City, California). Samples that produced ambiguous or negative results on a first attempt were repeated; genotypes that remained ambiguous were repeated until identical duplicate genotypes were obtained; samples producing consistently ambiguous or negative genotypes after three repetitions were treated as null at that locus. I determined an error rate for genotyping by independently amplifying and genotyping 10% of the samples for each locus; null results attributable to poor amplification were not included in error rate estimates. Any identified errors were corrected by repeated genotyping, as described above.

Landscape Analyses. —Landscape analyses were performed in ArcGIS version 9 with the Spatial Analyst extension (ESRI, Redlands, California). The 2001 national land cover dataset (NLCD, Homer et al. 2004) was obtained for our study area in Michigan and Ohio from the National Map Seamless Server (http: //seamless.usgs.gov). The NLCD classifies the land cover of 30m x 30m pixels into one of 16 different cover types (exclusive of Alaska and coastal areas not included in this study). I considered categories 41 (deciduous forest) and 43 (mixed forest) to be suitable mouse habitat, and other types to be non-habitat or more resistant matrix, as described below. We calculated the habitat abundance and distribution in the neighborhood of our trapping locations at three scales likely to impact the genetics of local Peromyscus. I defined our smallest scale based on Rasmussen’s (1963) estimate that the effective population size for mice is approximated by the number of adults in a circle of radius 2s, where s is the standard deviation of female dispersal distances between birth and breeding, and thus an estimate of typical dispersal distances. For s, I used a value of 250 m, the midpoint of various estimates of typical dispersal distances for female Peromyscus, which ranged from 200m to 300m (Blair 1940; Neigel et al. 1991; Krohne et

72

al. 1984). Therefore, I measured habitat composition within a radius of 0.5 km, 5 km, or 50 km. At each spatial scale, I calculated five estimates of habitat abundance or distribution using ArcGIS v9.3 with the Spatial Analyst extension: proportion of habitat coverage; area: perimeter ratio (calculated as total habitat area divided by total habitat perimeter length); average patch size; patch density (number of patches per total area); and average nearest-neighbor distance between patches. For comparisons of genetic differentiation to the landscape, Euclidean distances between populations were calculated using the Geographical Distance Matrix Generator (Ersts 2010).

Statistical analysis.—I performed population genetic analyses in GenePop version 4.0 (Rousset 2008) unless otherwise noted. Data were evaluated for linkage disequilibrium and deviation from Hardy-Weinberg Equilibrium (HWE) using a log likelihood ratio test (Goudet et al. 1996). Significance of results was determined after applying a sequential Bonferroni correction for multiple comparisons (Rice 1989). The possibility of null alleles, which can lead to failed genotype determination and heterozygote deficiency, was examined using MicroChecker version 2.2.3 (Van Oosterhout et al. 2004) to analyze homozygote excess over the various allele size classes. When null alleles were found to be present at frequencies of over 5%, allele frequencies were adjusted using an estimator based on frequencies of heterozygote deficiency and null homozygotes (Brookfield 1996).

Interpopulation differentiation was measured using FST (Wright 1931) as approximated by Weir & Cockerham 1984), which estimates differentiation from variance in allele frequencies. For comparison with studies using stepwise mutation- based estimators, I also calculated ST , which measures the proportion of variance in allele size due to differences among populations under the assumption of a high-rate stepwise mutation model. The possibility of isolation-by-distance was examined using a

Mantel test (Mantel 1967), with FST as the y-variable, in Arlequin v3.11 (Excoffier et al., 2005) and geographical distance as the x-variable. To examine latitudinal variation in interpopulation differentiation, I used the Mantel test with the latitudinal midpoint of population pairs as the x-variable.

73

In addition to the F-statistic analyses described above, I used the Bayesian program Structure v2.3.3 (Pritchard et al. 2000) to assign individuals to populations based on genetic similarities. To determine the number of genetically relevant populations, I performed three runs of Structure at each value of k from 1 to 10, with 106 repetitions in each run and a burn-in of 100,000 repetitions. I evaluated the results using the k metric (Evanno et al. 2005), which is based on the second-order rate of change of the posterior probability of the data (Ln P(D)). I then fixed the value of k based on the mode of k, and performed 5 longer runs of 1.5 x 106 generations in order to evaluate assignment of individuals to clusters.

Genetic diversity was measured using allelic richness (AR), the average number of alleles per locus observed in a population, and also using observed and expected heterozygosity (HO, HE). A was standardized for sample size using the repeated sampling procedure implemented in HP-Rare (Kalinowski 2006). HO and HE were calculated in Arlequin v3.11 (Excoffier et al. 2005).

Results

One hundred thirty-seven individuals from 12 populations were trapped and sampled, with 5-20 animals per population and at least 10 animals in 9 of the 12 populations (Table 2). Mice were genotyped at 10 loci (Table 1), all of which were polymorphic in all populations. The overall rate for incorrect genotyping, measured from random sample repetition, was 0.7%. I did not detect any significant instances of linkage disequilibrium. However, significant deviations from Hardy-Weinberg equilibrium due to heterozygote deficiency were detected in 31 out of 120 population-locus comparisons. Fifteen of these deviations involved the loci PO-9 and PO-26, which exhibited null allele frequencies in excess of 20% and were therefore excluded from differentiation analyses. All habitat variables were highly correlated at the 5 km scale (r2 ≥ 0.67) and all but total average patch size were highly correlated at the 50 km scale. I chose area: perimeter ratio (APR) as a representative combined index of habitat availability and connectivity at these scales. At the 0.5 km scale, patch density and average nearest- neighbor distance (NND) showed low and non-significant correlation with habitat 74

availability, and were examined in addition to APR. Habitat availability varied widely over the study area (Figure 2A). All three measures (i.e., at the 0.5, 5, and 50 km scales) produced the lowest availability values at around 41o northern latitude and tended to increase to the north. Habitat availability at 0.5 km was correlated to that at 5 km (r2 = 0.62; P = 0.0025) and at 50 km (r2 = 0.45; P = 0.017). Genetic diversity was highest in northern Michigan and lowest in central and northern Ohio (Table 2). Allelic richness and expected heterozygosity were significantly positively correlated with each other (r2 = 0.85; P = 0.0001), with measures of habitat 2 availability, and with latitude (r = 0.46; P = 0.030. AR was not significantly correlated with observed heterozygosity (r2 = 0.24; P = 0.15), reflecting the presence of null alleles. Habitat availability was related to allelic richness at 5 km (r2 = 0.58; P = 0.011) and 50 km (r2 = 0.49; P = 0.023). At the 0.5 km scale, habitat availability and allelic richness were correlated (r2 = 0.56; P = 0.013), but the plot revealed a pronounced curve (Figure 2B) that could be linearized by log-transforming APR (r2 = 0.66; P = 0.004; 95% CI: 0.73 < slope < 2.75). Correlation of richness with average patch size was near significance (r2 = 0.34; P = 0.076), but I did not see any correlation between richness and NND (r2 = 0.0038; P = 0.88) at the 0.5 km scale.

Overall levels of differentiation were low but significant (FST = 0.024; ST =

0.012; P < 0.0001). I did not observe any pattern in ST values, which were low but highly variable, nor did I see any significant relationship between FST's and geographical distances between populations (Figure 3A; r2 = 0.0040; P = 0.43). However, interpopulation differentiation was correlated with latitude: FST values decreased as the latitudinal midpoint of each pair of sites moved to the north (Figure 3B; r = -0.51; P = 0.025). In other words, populations in the south were more differentiated from each other than were populations in the north. In the genetic clustering analyses, the mode of the k metric (Evanno et al. 2005) for Structure, which runs from k = 1 to k = 10, was at k = 2. The five subsequent longer runs at k = 2 produced nearly identical results by population; the run with the least negative value of Ln (P(D)) is shown (Figure 4). Individuals from the five northernmost populations were assigned predominantly to cluster 1, while more southern populations

75

were assigned predominantly to either cluster 1 or cluster 2, with no spatial pattern apparent among the southern populations.

Discussion

In my study of white-footed mice (P. leucopus), changes in both intrapopulation genetic diversity and inter-population differentiation were associated with spatial variation in habitat availability. In both cases, correlation is in the direction that would be expected if habitat availability were a major regulator of genetic drift in this species (Templeton et al. 1990; Hutchison & Templeton 1999), i.e., lower diversity and higher differentiation to the south. While this is in general agreement with the fundamental principles of population genetics (Wright 1931; Crow & Denniston 1988; Gilpin 1991), several factors could have counteracted the effects of habitat availability. The core-periphery hypothesis (CPH) predicts that populations at the periphery of a species' range will be smaller and less common, due to factors such as declining habitat suitability or the climatic tolerances of the species (Brown 1984; Hoffmann & Blows 1994; Brown et al. 1995). Extending these expectations to population genetics leads to the straightforward prediction that populations at the periphery of a species' range will have lower genetic diversity and be more highly differentiated from each other than will those near the center of its range. If this were true for P. leucopus in our study area, we would expect low diversity and high differentiation in northern Michigan, the opposite of what we found. However, the application of the CPH to the present study is complicated because the forest habitat of P. leucopus is much more dense at the periphery of its range, in northern Michigan, than it is towards the center, in southern Ohio (Figure 1). Therefore, if habitat availability were a major determinant of population size, we would expect the inverse of the core-periphery pattern. In either case, effective population size is expected to determine genetic diversity; the question is whether habitat availability or factors such as climatic tolerances are more important for determining effective population size near the range limit of the species. Our results indicate that the former is true for P. leucopus. A majority of CPH studies examined by Eckert et al. (2008), found the expected trends: 64% found a decrease in intrapopulation genetic diversity towards the periphery,

76

while 70% found increases in interpopulation differentiation towards the periphery. In most of these studies, however, the core-periphery axis corresponds to the expected directionality of postglacial expansion, as in the annual riverbank plant Corrigiola litoralis (Durka 1999) or the lizard Lacerta viridis (Böhme et al. 2006). Because both expansion and core-periphery could produce the same genetic results through different mechanisms, interpretation of these results can be difficult. (Eckert et al. 2008). Likewise, anthropogenic habitat fragmentation that can limit population size often follows the same directionality expected of the CPH. For instance, directional changes in diversity and differentiation of wolverines in the northwestern United States could result from the CPH mechanism (Kyle & Strobeck 2001), from postglacial expansion, from anthropogenic habitat fragmentation, or from a combination of all three factors. By contrast, because I found increased diversity in the northern part of our study area, neither postglacial expansion nor CPH appear to be primary determinants of genetic structure at the landscape scale in P. leucopus. Prior studies of P. leucopus have shown significant levels of inter-population differentiation at the local level, but no isolation-by-distance or spatially consistent distribution of genetic structure (Mossman & Waser 2001; Anderson & Meikle 2010). If these patterns applied uniformly across our study area, we would not expect the gradients in genetic diversity and differentiation described here. However, our results for the southern end of our study are consistent with results from other studies of P. leucopus in Indiana and Ohio (Mossman & Waser 2001; Anderson & Meikle 2010): all found relatively low but extremely variable levels of differentiation in heavily agricultural landscapes. Specific values for differentiation are also consistent across studies: the overall FST of 0.024 reported here is comparable to values of 0.041 (Anderson & Meikle 2010) and 0.033 (Mossman & Waser 2001) reported in the prior studies. However, ours is the first to show that patterns in differentiation and diversity change on a large scale in Peromyscus, with the more forested northern end of our study area showing less interpopulation differentiation and more diversity. I emphasize that this is not a case of isolation-by-distance. Rather, populations in the south tend to be more different from those in the north, regardless of their distance from each other (Figure 3B). Together with the high scatter I observed in the FST-distance curve (Figure 3A), this pattern is

77

consistent with a regional lack of migration-drift equilibrium in this system, and with a larger effect of drift than of gene flow (Hutchison & Templeton 1999). As additional support for this conclusion, I observed that measures of patch area and habitat availability could explain much of the variation in genetic diversity at the 0.5 km scale, while patch isolation (measured as nearest-neighbor distance) could not. Finally, I observed an apparent threshold of habitat connectivity (area: perimeter ratio), below which genetic diversity is highly dependent on connectivity (Figure 2B). Together, these results are consistent with a dependence of population structure on habitat availability and thereby on effective population size, rather than on landscape limits on migration and gene flow. This consistent relationship between habitat availability/connectivity at the local scale and genetic diversity is intriguing given the distinctive population biology of P. leucopus. That is, P. leucopus often exhibits a negative correlation between patch size and population density (Nupp & Swihart 1996; Wilder & Meikle 2005), such that small patches might be expected to sustain relatively high levels of genetic diversity and produce a non-linear effect on diversity at low levels of habitat connectivity. One explanation for the difference could be that the high densities attained in small patches do not translate to large effective population sizes because the high densities attained in these patches are offset by larger crashes (Lacy 1987; Wilder et al. 2005). Prior studies have suggested that high food availability in edges could drive higher reproductive rates (Wilder & Meikle 2005) to produce high summer and fall mouse densities. On the other hand, carrying capacity may not be similarly augmented in small patches: emigration from them is high (Anderson & Meikle 2010) and mouse densities are equal across patch sizes in the winter, when food is uniformly scarce in all patches (Wilder et al. 2005). These trends could eventually negate the genetic effects of periodically high population sizes in small patches. While I found that genetic diversity is related to habitat availability at all scales examined, the strong relationship between the two at the 0.5 km scale is particularly interesting for two reasons. First, it indicates that I do not need to look beyond the local environment to understand the genetic diversity of a population; this underscores the high relative importance of drift compared to dispersal in this system. Second, this correlation at the 0.5 km scale provides support for the idea that a genetically effective population in

78

small mammals can be defined on the basis of typical female dispersal (Rasmussen 1963), providing a useful method for delineating populations when they have no obvious natural borders; for P. leucopus, forest habitat within a 0.5 km radius would appear to be an appropriate definition for a genetic neighborhood. Finally, our results do roughly follow the expectations of island biogeography as applied to population genetics, in that habitat availability predicts both genetic diversity and interpopulation differentiation. Even in the absence of isolation-by-distance, this highly vagile species, which exhibits little geographically defined structure at a local scale can exhibit regional genetic variation congruent with habitat availability. The populations are indeed more diverse when habitat is more abundant in their immediate neighborhood (i.e., the island is large). Likewise, populations tend to differ more when the distance between them is more forbidding (the island is farther from a source population), as in the southern portion of our study area. The fact that most alternative general hypotheses for population genetic structure predict patterns in the opposite direction, with more diversity and less differentiation towards the south, makes a relatively clear case in favor of this explanation.

Acknowledgments

The authors would like to thank Christine Anderson, Douglas Meikle, and Stephen Vessey for contributing of DNA samples to this study; Robbyn Abbitt for assistance with GIS methodologies; the Edwin S. George Reserve, the University of Michigan Biological Station, and the staff of Pigeon River State Forest for permission to conduct research and additional assistance; David Berg, for many conversations and suggestions; Philip Myers, for helpful conversations and field contributions; Rosa Moscarella, Molly Steinwald, the Field Ecology class from the University of Michigan, and many others for assistance with field or laboratory work; The Miami University Zoology Summer Field Research Workshop and a Department of Zoology Dissertation Scholarship to Z. Taylor, for funding.

79

References

Aguilar, R., Ashworth L, Galetto L., and M. A. Aizen. 2006. Plant reproductive susceptibility to habitat fragmentation: review and synthesis through a meta-analysis. Ecology Letters 9: 968-980.

Anderson, C. S., and D. B. Meikle. 2010. Genetic estimates of immigration and emigration rates in relation to population density and forest patch area in Peromyscus leucopus. Conservation Genetics. Doi: 10.1007/s10592-009-0033-8

Becker, C. G., Fonseca, C. R., Haddad, C. F. B., Batista, R. F., and P. I. Prado. 2007. Habitat split and the global decline of amphibians. Science 318: 1775-1777.

Blair, W. F. 1940. A study of prairie deermouse populations in southern Michigan. American Midland Naturalist 24:273-305.

Böhme, M. U., Schneeweiß, N., Fritz, U., Schlege, M., and T. U. Berendonk. Small edge populations at risk: genetic diversity of the green lizard (Lacerta viridis viridis) in Germany and implications for conservation management. Conservation Genetics 8: 555- 563.

Botti, W. B., and M. D. Moore. 2006. Michigan’s state forests: a century of stewardship. Michigan State University Press, East Lansing, Michigan.

Brookfield, J. F. Y. 1996. A simple new method for estimating null allele frequency from heterozygote deficiency. Molecular Ecology 5:453-455.

Brown, J. H. 1984. On the relationship between abundance and distribution of species. American Naturalist 124: 255-279.

80

Brown, J. H., Mehlman, D. W., and G. C. Stevens. 1995. Spatial variation in abundance. Ecology 76: 2028-2043.

Crow, J. F., and C. Denniston. 1988. Inbreeding and variance effective population numbers. Evolution 42: 482-495.

Debinski, D. M., and R. D. Holt. 2000. A survey and overview of habitat fragmentation experiments. Conservation Biology 14: 342-355.

Darlington, P. 1957. Zoogeography. John Wiley, New York, New York, USA.

Diamond J. M. 1969. Avifaunal equilibria and species turnover rates on the Channel Islands of California. Proceedings of the National Academy of Sciences, U.S.A. 64: 57– 63.

Diamond, J. M. 1975. The island dilemma: lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation 7: 129-146.

Durka, W. 1999. Genetic diversity in peripheral and subcentral populations of Corrigiola litoralis L. (Illecebraceae). Heredity 83:476-484.

Eckert, C. G., Samis, K. E., and S. C. Lougheed. 2008. Genetic variation across species' geographical ranges: the central-marginal hypothesis and beyond. Molecular Ecology 17: 1170-1188.

Ersts, P. J. Geographic Distance Matrix Generator (version 1.2.3). American Museum of Natural History, Center for Biodiversity and Conservation. Available from http://biodiversityinformatics.amnh.org/open_source/gdmg. Accessed 6/23/2010.

81

Evanno G., Regnaut S., and J. Goudet. 2005. Detecting the number of clusters of individuals using the software Structure: a simulation study. Molecular Ecology 14: 2611-2620.

Excoffier, L., Laval, G., and S. Schneider. 2005. Arlequin ver. 3.0: An integrated software package for population genetics data analysis. Evolutionary Bioinformatics Online, 1: 47-50.

Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics 34: 487-515.

Gannon, W.L., Sikes, R.S., and The Animal Care and Use Committee of the American Society Of Mammalogists. 2007. Guidelines of the American Society of Mammalogists for the use of wild mammals in research. J. Mammal. 88: 809-823. doi: 10.1644/06- MAMM-F-185R1.1.

Gilpin, M. 1991. The genetic effective size of a population. Biological Journal of the Linnean Society 42:165-175.

Golden, D. M., and T. O. Crist. 1999. Experimental effects of habitat fragmentation on old-field canopy insects: community, guild and species responses. Oecologia 118: 371- 380.

Gonzalez, A., and E. J. Chaneton. 2002. Heterotroph species extinction, abundance and biomass dynamics in a n experimentally fragmented ecosystem. Journal of Animal Ecology 71: 594-602.

Goudet, J., Raymond, M., de Meeüs, T. and F. Rousset. 1996. Testing differentiation in diploid populations. Genetics 144: 1931-1938.

82

Getz, L. L. 1961. Notes on the local distribution of Peromyscus leucopus and Zapus hudsonius. American Midland Naturalist 65: 486-500.

Herkert J. R. 1994. The effects of habitat fragmentation on Midwestern grassland bird communities. Ecological Applications 4: 461-471

Hoffmann, A. A., and M. W. Blows. 1994. Species borders: ecological and evolutionary perspectives. Trends in Ecology and Evolution 9: 223-227.

Homer, C., Huang, C., Yang, L., Wylie, B., and M. Coan. 2004. Development of a 2001 National Landcover Database for the United States. Photogrammetric Engineering and Remote Sensing 70: 829-840.

Hutchison, D. W., and A. R. Templeton. 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.

Kalinowski, S. T. 2006. HP-Rare: A computer program for performing rarefaction on measures of allelic diversity. Molecular Ecology Notes 5: 187-189. doi: 10.1111/j.1471- 8286.2004.00845.x.

Kolozsvary, M. B., and R. K. Swihart. 1999. Habitat fragmentation and the distribution of amphibians: patch and landscape correlates in farmland. Canadian Journal of Zoology 77: 1288-1299.

Krohne D. T., Dubbs, B. A., and R. Baccus. 1984. An analysis of dispersal in an unmanipulated population of Peromyscus leucopus. American Midland Naturalist 112: 146-156.

Kyle, C. J., and C. Strobeck. 2001. Genetic structure of North American wolverine (Gulo gulo) populations. Molecular Ecology 10: 337-347.

83

Landry, P.-A. and F.-J. Lapointe. 1999. The genetic heterogeneity of deer mouse populations (Peromyscus leucopus) in an insular landscape. Research in Population Ecology 41: 263-268.

Lacy, R. C. 1997. Loss of genetic diversity from managed populations: interacting effects of drift, mutation, immigration, selection, and population subdivision. Conservation Biology 1: 143-158.

Lawlor, T. E. 1998. Biogeography of Great Basin mammals: paradigm lost? Journal of Mammalogy 79:1111-1130.

MacArthur, R. H., and E. O. Wilson. 1967. The Theory of Island Biogeography. Princeton University Press, Princetion, New Jersey.

Mantel, N. 1967. The detection of disease clustering and a generalized regression approach. Cancer Research 27: 209-220.

Mossman, C.A., and P. M. Waser. 2001. Effects of habitat fragmentation on population genetic structure in the white-footed mouse (Peromyscus leucopus). Canadian Journal of Zoology 79: 285-295.

Myers, P., Lundrigan, B. L., and R. V. Koppel. 2004. Climate change and the distribution of Peromyscus in Michigan—Is global warming already having an impact? Pages 101-126 in E. A. Lacey and P. Myers, editors. Mammalian diversification: from chromosomes to phylogeography. University of California Press, Berkeley.

Myers, P., Lundrigan, B. L., Hoffman, S. M. G., Haraminac, A.P., and S. H. Seto. 2009. Climate-induced changes in the small mammal communities of the northern Great Lakes Region. Global Change Biology 15: 1434-1454.

84

Neigel, J. E., Ball, R. M. Jr. and J. C. Avise. 1991. Estimation of single generation migration distances from geographic variation in animal mitochondrial DNA. Evolution 45: 423-432.

Nupp, T. E., and R. K. Swihart. 1996. Effect of forest patch area on population attributes of white-footed mice (Peromyscus leucopus) in fragmented landscapes. Canadian Journal of Zoology 74:467-472.

Paetkau, D., Waits, L. P., Clarkson, P. L., Craighead, L., Vyse, E., Ward, R., and C. Strobeck. 1998. Variation in genetic diversity across the range of North American brown bears. Conservation Biology 12: 418-429.

Preston, F.W. 1962. The canonical distribution of commonness and rarity. I. Ecology 43: 185-215

Pritchard, J. K., Stephens, M., and P. Donnelly. 2000. Inference of population structure using multilocus genotypes. Genetics 155:945-959.

Rasmussen, D. I. 1964. Blood group polymorphism and inbreeding in natural populations of the deer mouse Peromyscus maniculatus. Evolution 18: 219-229.

Rice, W.R. 1989. Analyzing tables of statistical tests. Evolution 43: 223-225.

Root, J. J., Black, W. C. IV, Calisher, C. H., Wilson, K. R., Mackie, R. S., Schountz, T., Mills, J. N., and B. J. Beaty. 2003. Analyses of gene flow among populations of deer mice (Peromyscus maniculatus) at sites near hantavirus pulmonary syndrome case- patient residences. Journal of Wildlife Diseases 39: 287-298.

Rousset, F. 2008. GenePop '007: a complete re-implementation of the GenePop software for Windows and Linux. Molecular Ecology Resources 8: 103-106.

85

Simberloff, D. and E. O. Wilson. 1969. Experimental Zoogeography of islands - colonization of empty islands. Ecology 50: 278-296

Templeton, A.R., Shaw, K., Routman, E., and S. K. Davis. 1990. The genetic consequences of habitat fragmentation. Annals of the Missouri Botanical Garden 77:13- 27.

Van Oosterhout, C., Hutchinson, W. F., Willis, D. P. M., and P. Shipley. 2004. Micro- checker: software for identifying and correcting null alleles. Molecular Ecology Notes 4: 725-727.

Vucetich, L. M., Vucetich, J. A., Joshi, C. P., Waite, T. A., and R. O. Peterson. 2001. Genetic (RAPD) diversity in Peromyscus maniculatus populations in a naturally fragmented landscape. Molecular Ecology 10: 35-40.

Weir, B.S, and C. C. Cockerham. 1984. Estimating f-statistics for the analysis of population structure. Evolution 38:1358-1370.

Wilder, S., and D. B. Meikle. 2005. Reproduction, foraging and the negative density- area relationship of a generalist rodent. Oecologia 144: 391-398.

Wilder, S. M., Abtahi, A. M., and D. B. Meikle. 2005. The effects of forest fragmentation on densities of white-footed mice (Peromyscus leucopus) during the winter. American Midland Naturalist 153:71-79.

Wisely, S. M., Buskirk, S. W., Russell, G. A., Aubry, K. B., and W. J. Zielinski. 2004. Genetic diversity and structure of the Fisher (Martes pennanti) in a peninsular and peripheral metapopulation. Journal of Mammalogy 85: 640-648.

Wolff, J. O. 1996. Coexistence of white-footed mice and deer mice may be mediated by fluctuating environmental conditions. Oecologia 108: 529-533.

86

Wright S. 1931. Evolution in Mendelian populations. Genetics 16:97-159.

87

Table 1. Microsatellite loci selected for analysis. References are indicated by superscripts: 1, Prince et al. 2002; 2, Schmidt 1999; 3, Chirhart et al. 2000. ______

Locus [MgCl2] Temp. Size Range Alleles ______PO-091 1.5 mM 55.0o C 135-191 19 PO-261 2.0 mM 62.0o C 139-241 23 PO-351 1.5 mM 60.2o C 243-327 30 PO3-681 1.5 mM 63.3o C 230-346 32 PO3-851 3.0 mM 50.8o C 182-234 20 PLGT152 2.0 mM 59.0o C 236-270 17 PLGT-622 2.0 mM 60.2o C 148-184 17 Pml-01 3 2.5 mM 55.0o C 146-178 17 Pml-04 3 2.0 mM 63.3o C 184-246 23 Pml-12 3 1.5 mM 57.0o C 134-174 18 ______

88

Table 2. Standard molecular diversity indexes for populations of Peromyscus leucopus, listed South to North. N: sample size. A: allelic richness. AR: allelic richness corrected by rarefaction for a sample size of 7. HE: expected heterozygosity. HO: observed heterozygosity. Abbreviations used in Figure 1 are given in parentheses. Allelic richness was not determined for the BAC or WEBB populations. ______

Population N A AR HE HO ______Bachelor Woods, OH (BAC) 7 6.5 N/D 0.84 0.57 Jericho, OH (JER) 12 7.9 6.71 0.85 0.56 Richardson, OH (RIC) 12 7.4 6.13 0.83 0.57 Mercer Co., OH (MR) 20 9.3 6.57 0.81 0.56 Carter Woods, OH (CAR) 13 7.4 6.32 0.83 0.65 George Reserve, MI (ESGR) 13 9.7 7.38 0.87 0.63 Missaukee County, MI (MIS) 11 8.3 7.50 0.86 0.56 Otsego County, MI (OT) 12 10.4 8.15 0.90 0.62 Fisherman Road, MI (FISH) 11 8.8 7.38 0.88 0.69 Osmun Road, MI (OS) 10 9.0 7.81 0.88 0.66 Webb Road, MI (WEBB) 5 5.6 N/D 0.88 0.50 UMBS, MI (UMBS) 8 7.7 7.38 0.87 0.67 ______

89

A. B. UMBS WEBB OS FISH OT

MIS

MI ESGR

CAR OH

MR

RIC JER BAC

Figure 1. Study area and trapping sites for Peromyscus leucopus. (A) Michigan, Ohio, and surrounding states, with inset showing study area. B. National land cover dataset showing forested (gray) and unforested regions in Michigan and Ohio, with trapping sites indicated by black circles.

90

A. 1.2

1

0.8

0.6

0.4

0.2

Proportion Suitable Habitat Proportion 0

39 40 41 42 43 44 45 46 B. 8.5 Latitude

8

7.5

7

6.5 Allelic Richness 6

5.5 0 50 100 150 200 Area:Perimeter Ratio

Figure 2. Habitat availability and genetic diversity for Peromyscus leucopus. A. Variation in habitat availability by latitude, in Michigan and Ohio, at 0.5 km (filled diamonds), 5 km (open squares), and 50 km (gray triangles). The y-axis indicates total area within the buffer zones assigned to categories 41 (deciduous forest) or 43 (mixed forest) as a proportion of the total area of each buffer zone. B. Allelic richness and land cover availability for P. leucopus populations.

91

A. 0.1

0.08

0.06

ST 0.04 F

0.02

0 B. -0.02 0 100 200 300 400 500 600 700

Distance (km)

0.1

0.08

0.06

ST 0.04 F 0.02

0

-0.02

39 40 41 42 43 44 45 46 Average Latitude

Figure 3. Inter-population differentiation populations of Peromyscus leucopus. A. 2 Linearized FST values compared to the Euclidean distance between sites, r = 0.004; P =

0.43. B. Comparison of FST with latitude, expressed as the midpoint between pairs of sites; r2 = 0.26; P = 0.025.

92

Figure 4. Clustering analyses using STRUCTURE. Bars indicate average assignment of individual genomes per population to cluster 1 (gray) or cluster 2 (black). Populations are arranged by increasing latitude, left to right. Abbreviations used are described in Table 2.

93

Appendix. Sample populations of Peromyscus leucopus in Michigan and Ohio. ______ID Name County State GPS ______BAC Bachelor Woods Butler OH 39.5126, -84.7077 JER Jericho Woodlot Butler OH 39.5317, -84.6882 RIC Richardson Butler OH 39.5380, -84.7941 MR St. Mary's Mercer OH 40.5237, -84.5754 WC Carter Woods Wood OH 41.3897, -83.5911 ESGR E. S. George Reserve Livingston MI 42.4568, -84.0140 MIS Missaukee/Roscommon Missaukee MI 44.2362, -84.8516 OT Otsego Otsego MI 45.1099, -84.4132 FISH Fisherman Road Cheboygan MI 45.2370, -84.4767 OS Osmun Road Cheboygan MI 45.2555, -84.4117 WR Webb Road Cheboygan OH 45.2729, -84.4363 UMBS University of Michigan Emmet MI 45.6455, -84.9734 Biological Station

94

CHAPTER 5. Conclusion

Habitat fragmentation and habitat destruction (together: "habitat conversion") are among the most important threats facing populations of plants and animals around the world (Sodhi et al. 2004; Hoekstra et al. 2005). Habitat fragmentation, or the breakup of large patches, can interfere with the exchange of individuals among populations (Dooley & Bowers 1998; Bender et al. 1998), while habitat destruction reduces the total area available for the maintenance of populations (Wood & Pullin 2002). These activities increase the risk to populations in a number of ways, such as by increasing the proportion of ecologically distinct edge habitat (Diamond 1975; Andren & Angelstam 1988; Skole & Tucker 1993), increasing the exposure of wild populations to threats including competitors (Coker & Capen 1995) and diseases (Deem & Emmons 2005; Greer & Collins 2007), and by restricting populations to marginal or relatively unfavorable habitat (Channell & Lomolino 2000). Most importantly, though, habitat conversion limits maximum population sizes, increasing the chance that catastrophic events, random fluctuations, or even natural population cycles will cause local extinctions or even threaten the existence of the species (Shaffer 1981). A more subtle effect of habitat fragmentation is to reduce the genetic diversity of affected populations, both by increasing the magnitude of genetic drift within a population and by reducing the ability of a population to recover diversity through interchange with other populations ("gene flow"). In extreme cases, low genetic diversity can lead to inbreeding depression and random loss of alleles due to drift, thereby helping to push species towards extinction (Roelke et al. 1993; Bouzat et al. 1998). In less severe cases, genetic drift in small, isolated populations could remove potentially adaptive genetic diversity and thus endanger the ability of populations to adapt to future environmental conditions. Unfortunately, potentially adaptive genetic diversity is very difficult to measure directly, and the question of whether neutral variation (which is relatively easy to measure) is valuable for understanding adaptive diversity remains extremely controversial (Reed & Frankham 2001). Although using genetic techniques to estimate the overall survival prospects of a population is problematic (Reed & Frankham 2001; 2003), genetic studies of natural

95

populations have more subtle, but less controversial applications. For instance, genetic studies can help delineate groups of morphologically similar animals for conservation purposes (e.g. King et al. 2006), or identify optimal stocks for replenishing failing populations (Tallmon et al. 2004). Most importantly for the purpose of this dissertation, genetic analyses provide an indirect means of assessing population or metapopulation (generically, a "population of populations") processes that are difficult or impossible to measure directly. Most commonly, they can be used to understand the evolution of species (phylogenetics), to track the recent evolution of intraspecific groups through space (phylogeography), or to examine the importance of habitat fragmentation for nearby populations (population genetics). The preceding chapters describe the application of techniques from population genetics and phylogeography to understanding the relationship between habitat fragmentation and the distribution of genetic structure in Great Lakes populations of Peromyscus mice. Using a variety of approaches has allowed examination of this relationship from a variety of perspectives. In Chapter 2, phylogeographic methodologies provide insight into the role of the Great Lakes in shaping postglacial colonization of the region by small terrestrial animals. Chapter 3 then applies a population genetics approach to understanding the ongoing role of the Great Lakes in the dynamics of regional populations of P.m. gracilis. Chapter 4 examines Peromyscus leucopus with respect to very recent anthropogenic habitat conversion across a large region of the Midwest. Together, these studies provide a thorough picture of Peromyscus genetics across two species, as they responded to natural and anthropogenic changes in a large geographical area, from thousands of years ago into the present. The natural habitat fragmentation resulting from the presence of the Great Lakes in the midst of the range of the woodland deer mouse (P. m. gracilis) has shaped its populations differently in the short and long term. Chapter 2 investigates how this subspecies colonized several disjunct peninsulas and a number of islands despite the interposition of between six and twenty kilometers of open lake water. Although P. m. gracilis as a subspecies is not currently threatened with extinction, the relatedness of populations throughout the Great Lakes region is relevant to the conservation of diversity, because populations in the Lower Peninsula have been in decline over the last

96

hundred years (Myers et al. 2004; 2009). Of more general concern, the ways in which these mice managed to colonize the region over the last ten thousand years provides insight into how small mammals can migrate across extremely challenging geographical barriers. This study also indicates how the Great Lakes could impact the ability of species to shift their distributions in the face of ongoing climate change. These results are consistent with the results of prior studies (Lansman et al. 1983; Dragoo et al. 2006) describing a contact zone between two lineages in the Great Lakes area, with the more common eastern lineage closely related to a different subspecies of P. maniculatus found in the southern Appalachian Mountains, and the other lineage related to populations farther west. However, my research fully describes the distribution of these lineages in this region, and provides information on migration pathways and specific roles for the lakes. The results suggest that the eastern lineage of deer mice migrated into Canada from a southern Appalachian source, around the eastern shore of Lake Huron, then moved southward onto the Upper Peninsula, and finally across Lake Michigan to the islands and the Lower Peninsula. In other words, the lakes acted variously as barriers (preventing gene flow across Lake Huron, between Canada and the Lower Peninsula) and filters, restricting exchange and limiting relatedness of populations, but not preventing it altogether: the mice must have moved over Lake Michigan to reach the islands, and likely did so to reach the Lower Peninsula. Expansion of a second, western lineage of deer mice may have been restricted in part by Lake Michigan, but information on this lineage is much more limited because it was found in fewer individuals distributed over a much smaller area than was the eastern lineage. While Chapter 2 elucidates the original colonization of the Great Lakes region by mice, Chapter 3 addresses the importance of the lakes to the population dynamics of P. m. gracilis in the region today. In a sense, this study consists of additional tests of one of the conclusions of Chapter 2, that even if the Great Lakes have been permeable to migration on an evolutionary time scale, such migration is probably very rare. Specifically, the results of Chapter 3 show that P. m. gracilis populations on the two peninsulas are significantly differentiated from each other, and that island populations are all quite distinct from each other and from the two peninsulas. These results indicate that the Great Lakes are in fact real barriers between Peromyscus populations on either side.

97

The occasional migration that allowed mice to colonize the islands and the LP appears to be just that: occasional migration on an evolutionary timescale of thousands of years. These results are consistent with the demonstrated ability of rare individual Peromyscus to disperse hundreds of meters in a single night (Krohne et al. 1984) or several kilometers over a period of months (Maier et al. 2002), and to utilize facilitated dispersal mechanisms (Baker 1994). The rarity of long-distance migrations mean that the unique genetic variants found only in the isolated mouse populations of the Lower Peninsula would be lost in the case of local extinction, and that other small animals at these sites likely face similar issues of fragmentation and population size limitation. This study also indirectly provides information about potential effects of human transportation systems on small mammals. Because a bridge connects the two major peninsulas of Michigan and a ferry system connects islands to the Lower Peninsula, and because P. maniculatus have been shown to travel in grain trucks in Colorado (Baker 1994), accidental transport of these mice is certainly possible. Indeed, my results suggest that limited gene flow is continuing to occur between the two peninsulas, and accidental transport is a likely explanation. Chapter 4 considers the importance of habitat availability to genetic diversity and gene flow in P. leucopus near the southern range limit of this species in the Great Lakes region. This chapter adds to the literature on this species by greatly expanding the scale used by other studies. It expands the field of inquiry beyond local inter-patch migration and differentiation to the population consequences of range limits and the influence of gradients on landscape characteristics such as habitat availability and connectivity. As is true with prior studies of Peromyscus, this study found no relationship between genetic differentiation among populations and their geographical distances from each other. However, because of its unusually large spatial scale, it was able to detect significant gradients in both genetic diversity and inter-population diversity that were correlated with local habitat density. These gradients showed a specific pattern (hereafter, "population genetic decline"), consisting of a decrease in genetic diversity and a simultaneous increase in inter-population differentiation along a north-south transect, which is expected to occur under a number of theoretical scenarios. Such a decline could be due to directional expansion from a small source population (as with postglacial expansion),

98

increased proximity to an environmental range limit, or a decrease in patch size or habitat availability. In this case, the genetic decline was associated with the decreased habitat availability that is found as P. leucopus populations occur farther north in the study area. Brown bears, fishers, and wolverines in North America (Paetkau et al. 1998; Wisely et al. 2004; Kyle & Strobeck 2001) have likewise shown population genetic decline with increasing fragmentation across their ranges, as have frogs in Sweden (Johansson et al. 2006), and elsewhere. However, the habitat fragmentation gradients associated with population genetic decline in these studies have also been confounded with one or more other possible causes of decline. In contrast, my study was able to make a stronger case for habitat fragmentation as a proximate cause of population genetic decline in this system, because the other likely explanations predict the opposite pattern of decline along the north-south transect. Since P. leucopus expanded into the region from the south (Rowe et al. 2006), genetic diversity should decline towards the north. This species also reaches its range limit in northern Michigan, which would be expected to limit population sizes and again result in genetic decline to the north. So even though, on the basis of prior studies, one could have reasonably expected no genetic pattern along the transect, habitat conversion does appear to have a determining effect on genetic patterns in this species. This conclusion should be viewed with some caution, however, because the signal of genetic decline is fairly weak and becomes apparent only over hundreds of kilometers, which implies a high degree of mobility for such a small animal. When Chapters 3 and 4 are compared, the degree of genetic separation between mouse populations imposed by the Great Lakes in the short term is seen to be much higher than that imposed by anthropogenic habitat fragmentation: island populations of P. m. gracilis separated by only a few kilometers of Lake Michigan (Chapter 3) are much more differentiated than are P. leucopus populations separated by hundreds of kilometers of farmland (Chapter 4). On the other hand, P. m. gracilis populations within the heavily forested Upper Peninsula (Chapter 3) show little interpopulation differentiation and no pattern of isolation-by-distance, similar to P. leucopus populations on the Lower Peninsula and in Ohio (Chapter 4). In conclusion, this dissertation leads to several important observations about populations of small mammals. Most notably, genetic patterns on a small scale cannot

99

necessarily be extrapolated to larger ones. While I have shown a lack of isolation-by- distance for populations of both mouse species, their levels of genetic diversity and interpopulation differentiation change with natural or anthropogenic habitat fragmentation. In addition, even highly vagile animals can be vulnerable to the effects of dramatic habitat fragmentation imposed by large swaths of hostile terrain or geographical barriers, so the impacts of these forces need to be considered for all organisms.

100

References

Andren, H., and Angelstam, P. 1988. Elevated predation rates as an edge effect in habitat islands: experimental evidence. Ecology 69: 544-547.

Baker, A.E.M. 1994. Stowaway transport rates of house mice (Mus domesticus) and deermice (Peromyscus maniculatus). Proceedings of the 16th Vertebrate Pest Conference, Davis, California. pp. 106-112.

Bender, D. J., Contreras, T. A., and Fahrig, L. 1998. Habitat loss and population decline: a meta-analysis of the patch size effect. Ecology 79: 517-533.

Bouzat, J. L., Cheng, H. H., Lewin, H. A., Westemeier, R. L., Brawn, J. D., and Paige, K. N. 1998. Genetic evaluation of a demographic bottleneck in the greater prairie chicken. Conservation Biology 12: 836-843.

Channell, R., and Lomolino, M. V. 2000. Dynamic biogeography and conseration of endangered species. Nature 402: 286-288.

Coker, D. R., and Capen, D. E. 1995. Landscape-level habitat use by brown-headed cowbirds in Vermont. Journal of Wildlife Management 59: 631-637.

Deem, S. L., and Emmons, L. L. 2005. Exposure of free-ranging maned wolves (Chrysocyon brachyurus) to infectious and parasitic disease agents in the Noel Kempff Mercado National Park, Bolivia. Journal of Zoo and Wildlife Medicine 36: 192-197.

Diamond, J. M. 1975. The island dilemma: lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation 7: 129-146.

Dooley, J. L. Jr., and Bowers, M. A. 1998. Demographic responses to habitat fragmentation: experimental tests at the landscape and patch scale. Ecology 79:969-980.

101

Dragoo, J. W., J. A. Lackey, K. E. Moore, E. P. Lessa, J. A. Cook, and Yates, T. L. 2006. Phylogeography of the deer mouse (Peromyscus maniculatus) provides a predictive framework for research on hantaviruses. Journal of General Virology 87: 1997-2003.

Greer, A. L., and Collins, J. P. 2007. Habitat fragmentation as a result of biotic and abiotic factors controls pathogen transmission throughout a host population. Journal of Animal Ecology 77: 364-369.

Hoekstra, J. M., Boucher, T. M., Ricketts, T. H., and Roberts, C. 2005. Confronting a biome crisis: global disparities of habitat loss and protection. Ecology Letters 8: 23-29.

Johansson, M. J., C. R. Primmer and Merila, J. 2006. History vs. current demography: explaining the genetic population structure of the common frog (Rana temporaria). Molecular Ecology 15: 975-983.

King, T. L., Switzer, J. F., Morrison, C. L., Eakles, M. S., Young, C. C., Lubinski, B. A., and Cryan, P. 2006. Comprehensive genetic analyses reveal evolutionary distinction of a mouse (Zapus hudsonius preblei) proposed for delisting from the US Endangered Species Act. Molecular Ecology 15: 4331-4359.

Kyle, C. J., and Strobeck, C. 2001. Genetic structure of North American wolverine (Gulo gulo) populations. Molecular Ecology 10: 337-347.

Krohne, D. T., Dubbs, B. A., and Baccus, R. 1984. An analysis of dispersal in an unmanipulated population of Peromyscus leucopus. American Midland Naturalist 12:146-156.

Lansman, R. A., J. C. Avise, C. F. Aquadra, J. F. Shapira, and Daniel, S. W. 1983. Extensive genetic variation in mitochondrial DNA’s among geographic populations of

102

the deer mouse, Peromyscus maniculatus. Evolution 37: 1-16.

Maier, T. J. 2002. Long-distance movements by female white-footed mice, Peromyscus leucopus, in extensive mixed-wood forest. Canadian Field-Naturalist 116: 108-111.

Myers, P., B. L. Lundrigan, and Koppel, R. V. 2004. Climate change and the distribution of Peromyscus in Michigan—Is global warming already having an impact? Pages 101-126 in E. A. Lacey and P. Myers, editors. Mammalian diversification: from chromosomes to phylogeography. University of California Press, Berkeley.

Myers, P., Lundrigan, B.L., Hoffman, S.M.G., Haraminac, A.P., and Seto, S.H. 2009. Climate-induced changes in the small mammal communities of the northern Great Lakes Region. Global Change Biology 15: 1434-1454.

Paetkau, D., Waits, L. P., Clarkson, P. L., Craighead, L., Vyse, E., Ward, R., and Strobeck, C. 1998. Variation in genetic diversity across the range of North American brown bears. Conservation Biology 12: 418-429.

Reed, D. H., and Frankham, R. 2001. Correlation between fitness and genetic diversity. Evolution 55: 1095-1103.

Reed, D. H., Frankham, R. 2003. Correlation between fitness and genetic diversity. Conservation Biology 17:230-237.

Roelke, M. E., Martenson, J. S., and O'Brien, S. J. 1993. The consequences of demographic isolation and genetic depletion in the endangered Florida panther. Current Biology 3: 340-350.

Rowe, K. C., Heske, E. J., and Paige, K. N. 2006. Comparative phylogeography of eastern chipmunks and white-footed mice in relation to the individualistic nature of species. Molecular Ecology 15: 4003-4020.

103

Shaffer, M. L. 1981. Minimum population sizes for species conservation. BioScience 31: 131-134.

Skole, D., and Tucker, C. 1993. Tropical deforestation and habitat fragmentation in the Amazon: satellite data from 1978 to 1988. Science 260: 1905-1910.

Sodhi, N. S., Koh, L. P., Brook, B. W., and Ng, P. K. L. Southeast Asian biodiversity: an impending disaster. Trends in Ecology and Evolution 19: 654-660.

Tallmon, D. A., Luikhart, G., and Waples, R. S. 2004. The alluring simplicity and complex reality of genetic rescue. Trends in Ecology and Evolution 19: 489-496.

Wisely, S. M., S.W. Buskirk, G. A. Russell, K. B. Aubry, and Zielinski, W. J. 2004. Genetic diversity and structure of the Fisher (Martes pennanti) in a peninsular and peripheral metapopulation. Journal of Mammalogy 85: 640-648.

Wood, B. C., and Pullin, A. S. 2002. Persistence of species in a fragmented urban landscape: the importance of dispersal ability and habitat availability for grassland butterflies. Biodiversity and Conservation 11: 1451-1468.

104