POPULATION GENETIC DIFFERENTIATION, MATING SYSTEM, AND EFFECTIVE POPULATION SIZE OF THE TULIPTREE (LIRIODENDRON TULIPIFERA L.) IN THE MID-ATLANTIC UNITED STATES

A Dissertation submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Biology

By

Ricardo Gutierrez Ozuna, M.Sc.

Washington DC December 19, 2017

Copyright 2017 by Ricardo Gutierrez Ozuna All Rights Reserved.

! ii ! POPULATION GENETIC DIFFERENTIATION, MATING SYSTEM, AND EFFECTIVE POPULATION SIZE OF THE TULIPTREE (LIRIODENDRON TULIPIFERA L.) IN THE MID-ATLANTIC UNITED STATES

Ricardo Gutierrez Ozuna, M.Sc.

Thesis Advisor: Matthew B. Hamilton, Ph.D.

ABSTRACT

Continuous natural habitats are increasingly transformed into anthropogenic landscapes—mosaics of natural habitats fragments surrounded by heterogeneous mixtures of different land use. Anthropogenic habitat transformation is particularly evident in the clearing of forests. Forests that have reached a great age with little or no anthropogenic disturbance (old- growth) are increasingly rare. At the same time, urbanization—one of the most readily apparent human-mediated effects on the landscape—continues to expand, and as a result, an increasing percentage of forests are becoming part of urban landscapes and ecosystems. The goal of this thesis is to investigate population genetic effects of anthropogenic landscapes on long-lived forest trees at different time scales: immediate, through the estimation of the mating system; short-time, through the estimation of effective population size using linkage disequilibrium based methods; and long-time, through the estimation of population genetic differentiation. To accomplish this goal, I researched the tuliptree (Liriodendron tulipifera L.), a tree in the family

Magnoliaceae native to Eastern North America that has a wide geographic distribution in the

! iii ! Southeast and mid-Atlantic United States and occurs in diverse habitats, including both old- growth remnants and forest patches embedded in urban areas where it reproduces and recruits naturally.

Chapter 1 describes the development and validation of novel L. tulipifera genomic microsatellite (SSR) genetic marker loci using high-throughput sequencing and bioinformatics methods. These new SSR markers facilitated empirical study of population genetic patterns in L. tulipifera. Chapter 2 compares the population genetic differentiation and the mating system among old-growth forests and urban forest patches. L. tulipifera population genetic differentiation in the mid-Atlantic U.S. was characterized by a pattern of isolation by distance

(IBD), suggesting that urbanization has not had a major impact on regional genetic differentiation patterns which have evolved over hundreds of generations in the past. Despite marked environmental differences between old-growth and urban forest, all populations shared a fully outcrossed mating system without any variation in self-fertilization that can accompany loss of pollinators. Chapter 3 compares the strength of random genetic drift estimated from the effective number of breeders (Nb) from seedlings and from mixed-age reproductively mature trees (Ne {mixed ages}). These two sampling methods showed similar and small effective population sizes, suggesting genetic drift is a major determinant of polymorphism within populations and of genetic differentiation among populations.

! iv !

To my mom.

! v ! ACKNOWLEDGMENTS

I am immensely grateful to my advisor, Dr. Matthew Hamilton, for his constant advice and guidance. I would also like to extend my gratitude to my thesis committee members: Dr.

Gina Wimp, Dr. Martha Weiss, and Dr. John Braverman, whose comments were deep and insightful, and who worked patiently to ensure I completed the process. I owe a debt of gratitude to Jack McGuire and Ana Carolina da Silva Antunes for their assistance in lab work.

I would like to express gratitude towards the Smithsonian Environmental Research

Center in Edgewater, Maryland; James Madison’s Montpelier in Orange, Virginia; Western

Shore Conservancy in Prince George County, Maryland; and Saddler's Woods Conservation

Association in Haddon Township, New Jersey for allowing me sampling specimens. I am greatly indebted to the Georgetown Environment Initiative and the Center for the Environment at

Georgetown University for funding my research. Many thanks go to the Consejo Nacional de

Ciencia y Tecnología (CONACYT) in Mexico for granting me a doctoral fellowship.

I am grateful from the bottom of my heart to the friends I made during this journey for their encouragement to keep going. My deep appreciation goes especially to Megan Wallen, Xin

Huang, Ewa Krzyszczyk, and Pasha Tabatabai. Last but not least, I would like to thank my family, in particular my mom, for her endless support and love.

! vi ! TABLE OF CONTENTS

INTRODUCTION ...... 1 ! CHAPTER 1. Identification and characterization of microsatellite loci in the tuliptree, Liriodendron tulipifera (Magnoliaceae) ...... 10

Introduction ...... 10

Methods and Results ...... 11

Microsatellite development ...... 11

Microsatellite data analysis ...... 13

Conclusion ...... 14 ! CHAPTER 2. Population genetic differentiation and mating system of a long-lived tree in old- growth forests and urban forest patches ...... 15

Introduction ...... 15

Materials and Methods ...... 22

Study species ...... 22

Field collections ...... 24

Genetic analysis ...... 24

Statistical analysis ...... 25

Genetic polymorphism ...... 25

Population genetic differentiation ...... 25

Pattern of genetic differentiation ...... 27

Mating system ...... 27

Results ...... 28

Genetic polymorphism ...... 28

Population genetic differentiation ...... 28

Pattern of genetic differentiation ...... 29

Mating system ...... 29

Discussion ...... 30

Genetic polymorphism ...... 30

! vii ! Population genetic differentiation ...... 31

Pattern of genetic differentiation ...... 33

Mating system ...... 34

CHAPTER 3. Small effective population sizes in a common, widespread, long-lived tree ...... 37

Introduction ...... 37

Materials and Methods ...... 43

Study species ...... 43

Field collections ...... 44

Genetic analysis ...... 45

Effective number of breeders and effective population size ...... 45

Life history simulations ...... 47

Results ...... 48

Discussion ...... 49 ! APPENDIX A: Tables ...... 56 ! APPENDIX B: Figures ...... 72 ! REFERENCES ...... 75

! viii ! LIST OF TABLES

Table 1.1. Characteristics of 23 polymorphic genomic microsatellite loci isolated from Liriodendron tulipifera ...... 56

Table 1.2. Locality and voucher information for the Liriodendron tulipifera samples used for the development of polymorphic microsatellite markers ...... 58

Table 1.3. Genetic properties by individual and pooled sampled locations of 23 polymorphic microsatellite markers developed in Liriodendron tulipifera ...... 59

Table 2.1. Site information for the Liriodendron tulipifera samples used ...... 60

Table 2.2. Microsatellite marker loci used to genotype Liriodendron tulipifera seedlings ...... 61

Table 2.3. Summary measures of genetic polymorphism for each sampling sampling site ...... 62

Table 2.4. Pairwise standardized genetic differentiation (F’ST) and Euclidean geographic distances in kilometers for all sampling site pairs ...... 63

Table 2.5. Analysis of molecular variance (AMOVA) for 384 seedlings from 8 sampling sites in 2 forest types ...... 64

Table 2.6. Estimates of L. tulipifera mating system parameters for each sampling site obtained using the maximum-expectation method implemented in MLTR and a Bayesian method implemented in BORICE ...... 65

Table 3.1. Sampling locations and total number of samples collected ...... 66

Table 3.2. Microsatellite marker loci used to genotype Liriodendron tulipifera seedlings and adult individuals ...... 67

Table 3.3. Effective of breeders (!") estimated from linkage disequilibrium in single age cohorts of Liriodendron tulipifera ...... 68

! ix ! Table 3.4. Effective population size estimates from linkage disequilibrium in single age (!") and mixed-age cohorts samples (!# $%&#'()*# ) of Liriodendron tulipifera ...... 69

Table 3.5. Correction factor point estimates, and two confidence intervals for correction factor point estimates used ...... 70

Table 3.6. Effective population size, Ne; effective number of breeders, Nb; Nb/Ne ratio; mean number of offspring per parent per time period, +; variance in reproductive success among adults in one time period, Vk; generation length (mean age of parents of a newborn cohort), G; estimated from eight life-history scenarios ...... 71

! x ! LIST OF FIGURES

Figure 2.1. Relationship between pairwise F’ST / (1 − F’ST) and natural logarithm of geographic distances (in km) for nine L. tulipifera sampling sites ...... 72

Figure 3.1. Comparison of effective population size estimates from linkage disequilibrium in single age (,-) and mixed-age cohorts samples (,. /01.2(34. ) of Liriodendron tulipifera ..... 73!

Figure 3.2. Median r^2 permute for r^2_{c} ...... 74

!

! xi ! INTRODUCTION

Across the globe, natural ecosystems are being converted to human-dominated landscapes at increasing rates, with potentially significant evolutionary consequences.

Anthropogenic habitat change is particularly evident in the clearing of forests. In the continental

United States, 95 percent of forests have been logged since the European settlement began

(Brown et al. 1998). Forests that have reached a great age with little or no anthropogenic disturbance (old-growth) are increasingly rare. In the United States less than one percent of eastern forests are considered as old-growth, existing only as small scattered remnants (Davies

1996). At the same time, urbanization continues to expand, and as a result, an increasing percentage of forests are becoming part of urban landscapes and ecosystems.

Urbanization is a dramatic anthropogenic environmental change; alterations in temperature (urban heat island effect), increased inputs of inorganic nutrients, and increased area of impervious surfaces are among its consequences. Urban areas frequently represent convergent environments, in which urban ecosystems in different geographic areas are more similar to one another than they are to their surrounding natural environments (reviewed in McKinney 2006;

Pickett et al. 2011). Because of this environmental convergence, urban areas can serve as valuable unplanned, somewhat replicated experiments that offer unique opportunities to study the direction, rate, and repeatability of evolutionary change (Donihue and Lambert 2015). Recent studies have shown that urbanization affects both nonadaptive and adaptive evolution (reviewed in Johnson and Munshi-South 2017). For example, urbanization causes loss of genetic polymorphism within and increases genetic differentiation among populations, which has been interpreted as a consequence of stronger genetic drift in urban areas. However, few direct measurements of genetic effective population size have been made in the context of urbanization.

! 1 ! Further, by transforming natural areas such as forests to impervious surfaces like roads, buildings, and parking structures, urbanization also creates barriers to dispersal for some organisms. Consequently, gene flow is often reduced among urban populations, thereby contributing to genetic differentiation among populations (Johnson and Munshi-South 2017).

Studies of evolution in urban environments have focused predominately on relatively short-lived animal species with low dispersal capabilities such as salamanders, lizards, and mice.

Considerably less is known about how urbanization affects the evolution of plants in general and long-lived trees in particular.

This dissertation was designed to evaluate population genetic impacts of urbanization on a long-lived tree. Although this project was designed and initiated in 2012, a more recent paper by Johnson et al. (2015) summarized five consensus predictions for possible effects of urbanization on plants. These predictions included a range of evolutionary processes that can be impacted by urbanization. These predictions are as follows: (1) the amount of genetic polymorphism will be different between urban and non-urban populations; (2) natural selection will differ between urban and non-urban populations due to environmental changes associated with urbanization; (3) genetic drift will be stronger in urban areas relative to rural areas because of smaller population sizes; (4) genetic differentiation between urban and non-urban populations should be proportional to the size of urban areas; and (5) insect-pollinated plants will have higher rates of self-pollination in urban areas due to reductions in the abundance and diversity of pollinators. An extensive literature search found no publications or studies testing all of these predictions with a long-lived forest tree.

I tested versions of these five predictions in the tuliptree (Liriodendron tulipifera L.). To test these hypotheses, I compared urban and old-growth populations in the highly developed

! 2 ! mid-Atlantic region of the United States. L. tulipifera is an excellent species for these studies because within its wide geographic distribution in the eastern United States, this species occurs in diverse habitats, including both old-growth remnant forests and forest patches embedded in urban areas where it reproduces and recruits naturally. The tuliptree is a pioneer species as well as a long-lived large canopy tree that may survive for as long as 300 years (Beck 1990). Further, it has a short-lived seed bank (Clark and Boyce 1964), so seeds and seedlings will experience recent environmental conditions. Following a chapter describing the identification and characterization of microsatellite (SSR) marker loci in L. tulipifera, tests of Johnson et al.'s

(2015) predictions 1, 4, 2, 5, and 3 will be described (in that order) below.

To provide means to test these predictions, Chapter 1 describes the development and validation of L. tulipifera genomic SSR markers. These novel loci avoided the potential limitations of previously reported SSR markers. Although there are numerous SSR markers available for L. tulipifera, most of them have been developed from expressed sequence tags

(EST) derived from transcribed regions (Xu et al. 2006, Yang et al. 2012, Xu et al. 2010, Zhang et al. 2015). EST-SSRs are located in or near functional genes, and as a consequence, are more likely to be affected by natural selection (Ellis and Burke 2007). In addition, genomic

(noncoding, nontranscribed) microsatellite loci have been reported for Liriodendron chinense and cross-amplified in L. tulipifera (Yao et al. 2008). However, cross-species amplification of

SSR loci might result in an ascertainment bias, which means that allele length is longest in the species from which the markers were developed. This leads to low levels of variation of homologous SSR loci in closely related species (Ellegren et al. 1995). The likely non-neutral evolution of EST-SSRs, and the potential ascertainment bias introduced by cross-species amplification might affect population genetic analyses involving estimations of intraspecific

! 3 ! genetic polymorphism, gene flow, and genetic distances (both within and among populations).

Therefore, in order to identify and characterize SSR marker loci that do not suffer from ascertainment bias and show high levels of polymorphism, we used high-throughput sequencing and bioinformatics methods. We fully describe the development of these marker loci in

Gutiérrez-Ozuna and Hamilton (2017). During this process, we also identified thousands of SSR loci and potentially amplifiable loci (PALs, with primer sequences), which we have made available as Supplementary Information in Gutiérrez-Ozuna and Hamilton (2017). Further, raw reads from L. tulipifera whole-genome sequencing were made publicly available at the NCBI's

Short Read Archive (SRA) under project accession number PRJNA331147, BioSample:

SAMN05417503, and may serve as a resource for the community.

Chapter 2 compares the population genetic differentiation and the mating system in L. tulipifera among four urban forest patches and four old-growth remnant forests sampled in the mid-Atlantic U.S. To test prediction 1, I obtained measures of genetic polymorphism in each sampling site, including the average number of different alleles per locus (ka), the average number of private alleles per locus (kp), and the multilocus average observed heterozygosity

(HO). I observed no differences in the levels of genetic polymorphism between urban and old- growth sites, which does not support prediction 1. In order to test predictions 2 and 4, I used the fixation index (FST) to quantify population genetic differentiation. Genetic differentiation was greater among old-growth forests than among urban forest patches, which initially could be interpreted as support for prediction 4 that there is genetic divergence between urban and non- urban populations. However, this is likely to be the result of closer geographic proximity of urban forests patches, rather than an effect of urbanization causing genetic divergence between urban and old-growth populations. Furthermore, prediction 2 should lead to a process of isolation

! 4 ! by adaptation (IBA), whereby local adaptation reduces gene flow among populations because of reduced establishment success of migrants with lower fitness from populations with different environmental conditions. IBA will lead to a pattern of isolation by environment (IBE).

However, I did not find support for prediction 2 because population genetic differentiation was characterized by a pattern of isolation by distance (IBD) instead of IBE. This result suggests that urbanization per se has not had a major impact on regional genetic differentiation patterns of L. tulipifera, which is better explained by neutral processes of gene flow declining with geographic distance.

In plants, the mating system characterizes and quantifies the distribution of mating unions in a population. Mating system is a component of gene flow, influencing gene exchanges within and among populations. While outcrossing facilitates population integrity, selfing acts as a barrier to gene flow. I estimated the mating system of each population to test prediction 5 that insect-pollinated plants will evolve higher rates of self-pollination in urban areas. Contrary to this prediction, I found that all populations shared a fully outcrossed mating system without any variation in rates of self-fertilization. This suggests that pollination is not altered in urban areas, despite marked environmental differences between old-growth and urban forests, consistent with studies showing that changes associated with urbanization are not detrimental to pollinators (e.g.,

Carper et al. 2014). L. tulipifera (as many other species in Magnoliaceae) is outcrossed by many different pollinators, including flies, beetles, honeybees, and bumblebees, which suggests that mating system does not have strong dependency on pollinators.

One possible interpretation to the lack of differences between urban and old-growth forests described in Chapter 2 is that the genetic measures used here have poor power to detect urbanization effects. In Chapter 2, I tested for population genetic effects of urbanization at two

! 5 ! different temporal scales. Because gene flow tends to homogenize allele frequencies while genetic drift tends to cause divergence due to fixation or loss of alleles within demes, the balance between these two evolutionary forces is a major determinant of population genetic differentiation. Any inference about processes that we make based on FST estimates, such as the effective number of migrants (Nem), assumes an equilibrium between these two opposite evolutionary forces, a process that can take a very long time to be reached. Therefore, it is not surprising that we did not observe genetic divergence between urban and old-growth populations, because urbanization is a relatively recent process compared to the genetic origin of the populations. It is likely that the differentiation patterns observed here evolved many generations in the past. This deep time evolutionary history may be one of the reasons why many studies of forest tree populations have not found significant consequences of habitat fragmentation and disturbance due to human activities such as logging, a result that has been called the “paradox of forest fragmentation genetics” (Kramer et al. 2008). In contrast, the other genetic measure used here, the mating system, can reach its equilibrium in a single reproductive event, which for L. tulipifera is one year. Disturbances causing deviations from random mating will lead to departure from Hardy-Weinberg expected genotype frequencies (HWE) in the offspring from that reproductive event. Nevertheless, the system will return to HWE in a single event of random mating (one reproductive season). In the case of the mating system, I used a tool that has the potential to detect immediate effects to urbanization; however, I still found no differences. This finding suggests that there are indeed no differences between urban and old- growth populations, rather than the mating system being insensitive to recent effects.

Chapter 3 tests prediction 3 that genetic drift is stronger in urban areas, using the effective number of breeders (,-), a measure of the strength of genetic drift in a single age

! 6 ! cohort for species with overlapping generations. I compared ,- between urban and old-growth populations which was estimated using a method based on two-locus disequilibrium. I observed that ,- estimates did not differ between urban and old-growth sites and were uniformly small, even in old-growth populations that could be considered as reservoirs of genetic polymorphism.

Genetic drift was strong in both forest types, which is inconsistent with prediction 3. Because ,- quantifies genetic drift over short timescales relative to ecological processes, it is unlikely that the lack of an urbanization effect on genetic drift was an artifact of the genetic measure used.

With pairs of loci experiencing free recombination, two-locus disequilibrium is expected to decay to nearly zero in few generations. Similar to the logic used for the mating system in

Chapter 2, this means that ,- as employed in this study has the potential to detect recent differences in the strength of genetic drift, if they exist. In addition, Chapter 3 compares effective population size estimates from two common sampling approaches, newborn seedlings (,-) and mixed-age reproductively mature trees (,. /01.2(34. ). Consistent with modeling results that showed Nb and Ne are expected to be very similar in this long-lived iteroparous species,

,. /01.2(34. estimates were also small and did not differ from ,- estimates. The small effective population sizes observed in this study suggest that genetic drift is a major determinant of polymorphism within populations and of genetic differentiation among populations, regardless of forest type.

In summary, I have tested for urbanization effects on population genetic patterns for a long-lived organism with overlapping generations, and high passive dispersal potential; most other recent tests for population genetic impacts of urbanization have focused on short-lived organisms with low dispersal capabilities. I have evaluated versions of all of Johnson et al.'s

! 7 ! (2015) predictions, and found no support for predictions 1, 3, 4, and 5. In addition, I found little support for prediction 2. The similar levels of genetic polymorphism observed across all populations do not support prediction 1 that the amount of genetic polymorphism will be different between urban and old-growth populations. Greater genetic differentiation among old- growth forests than among urban forest patches could initially be interpreted as support for prediction 4. However, this is likely to be the result of closer geographic proximity of urban forests patches. Finding IBD instead of IBE is inconsistent with prediction 2 of adaptation to urban environments. A fully outcrossing mating system shared among all populations does not support prediction 5 that insect-pollinated plants will have higher rates of self-pollination in urban areas. Uniformly small ,- estimates among populations suggest strong genetic drift in both forest types, which is inconsistent prediction 3 that genetic drift is stronger in urban areas.

In addition, the experimental approach used in this thesis was novel because it examined potential evolutionary effects of urbanization at three different temporal scales: (i) immediate, using mating system estimates; (ii) relatively short-time, using ,- estimates, an approach that has been rarely used in the context of urbanization; and (iii) long-time, estimating population genetic differentiation. In particular, mating system and ,- as employed in this study have the potential to detect effects of urbanization in the very recent past if they exist. This allowed me to rule out the alternative explanation that the lack of any urbanization effects was an artifact of the time to register changes of the genetic measures used. The lack of differences in mating system and ,- instead suggest that there are indeed no effects of urbanization on some key evolutionary processes that shape genetic polymorphism within and among populations of this long-lived tree.

! 8 ! Despite the increasing number of studies documenting effects of urbanization on patterns of genetic polymorphism within and among populations, it is currently not clear whether urban evolution is a universal phenomenon. This thesis highlights a series of population genetic phenomena that are going to be manifest differently in long-lived and short-lived organisms. For example, we already know that short-lived and self-compatible plants tend to have greater genetic differentiation at smaller scales than long-lived and outcrossing plants such as trees, and therefore the former are likely expected to exhibit stronger adaptation to local conditions. Based on differences in life history traits, population genetic theory can help to refine predictions on how urbanization can affect evolutionary processes. The Johnson et al. (2015) summary could potentially be updated with the data and information generated in this thesis about what we might expect for long-lived and short-lived organisms.

! 9 ! CHAPTER 1 Identification and characterization of microsatellite loci in the tuliptree, Liriodendron tulipifera (Magnoliaceae)1

Introduction

Liriodendron tulipifera L., commonly known as tuliptree, tulip poplar, or yellow poplar, is a pioneer tree in the family Magnoliaceae native to Eastern North America. It has a wide geographic distribution in the Southeast and mid-Atlantic United States and occurs in diverse habitats. In order to facilitate population genetic analyses of effective population size and population structure, we developed genomic microsatellite (SSR) markers without the potential limitations of previously reported SSRs. Liriodendron tulipifera SSRs have been developed from expressed sequence tags (EST; Xu et al. 2006, Xu et al. 2010, Yang et al. 2012, Zhang et al.

2015) located in or near functional genes, and as a consequence, they are more likely to be affected by natural selection (Ellis and Burke 2007). Liriodendron chinense (Hemsl.) Sarg. genomic (noncoding, nontranscribed) microsatellite loci have been cross-amplified in L. tulipifera (Yao et al. 2008). Cross-species amplification of microsatellite loci might result in ascertainment bias, where polymorphism is reduced when loci are employed in related species

(Ellegren et al. 1995). Preliminary tests of loci from Yao et al. (2008) carried out with 10 L. tulipifera individuals showed low polymorphism (results not shown). Non-neutral evolution or ascertainment bias can potentially impact the estimation of population genetic parameters.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 1!This chapter has been published as: Gutiérrez-Ozuna, R., and Hamilton, M.B. (2017). Identification and characterization of microsatellite loci in the tuliptree, Liriodendron tulipifera (Magnoliaceae). Applications in Plant Sciences, 5(8), 1700032.

! 10 ! Therefore, we identified and characterized polymorphic genomic microsatellite loci in L. tulipifera using Illumina next-generation sequencing and a bioinformatics pipeline.

Methods and Results

Microsatellite development

Total DNA from leaves of one L. tulipifera individual collected on the main campus of

Georgetown University in Washington, DC, USA was extracted using the DNeasy Plant Mini

Kit (QIAGEN, Valencia, California, USA). A genomic DNA library for Illumina paired-end sequencing was prepared from 4 µg of DNA following the PCR-free library prep kit from

Illumina (Illumina, San Diego, California, USA). DNA was sheared to 550 bp and sequenced as

150 bp paired-end reads on an Illumina HiSeq 2500 at Virginia Bioinformatics Institute at

Virginia Tech. We used PAL_FINDER_v0.02.04 (Castoe et al., 2012) to extract reads containing perfect microsatellites (uninterrupted and identical repeats). The reads were imported to PAL_FINDER and analyzed in two different ways: 1) as Illumina paired-end reads filtered to include ≥12 tri-, ≥10 tetra-, ≥8 penta-, and ≥6 hexa-nucleotide repeats, and 2) as FASTQ sequence files converted to FASTA format, treated as 454 single-end reads, and filtered to include ≥15 di-, ≥10 tri-, ≥8 tetra-, ≥6 penta-, and ≥4 hexa-nucleotide repeats. One potential advantage of using both methods is the development of loci with a broader range of amplified fragment sizes. In both cases, we identified microsatellite loci with flanking sequences suitable for PCR primer design or potentially amplifiable loci (PALs). Raw reads were deposited in

NCBI's Short Read Archive (SRA, Project: PRJNA331147, BioSample: SAMN05417503).

Summaries of reads containing microsatellite repeats and PALs (with primer sequences) detected

! 11 ! using both methods are available as Supplementary Information together with the published paper.

We selected a set of 77 PALs to empirically assess amplification using three individuals.

We amplified each locus in 25µL PCR reactions [1X OneTaq Standard reaction buffer, 160µM dNTPs, 0.2µM forward primer, 0.2µM reverse primer, 1.25 units OneTaq DNA polymerase

(New England BioLabs, Ipswich, Massachusetts, USA), 1µL template DNA (concentration was not determined), and ddH20 to 25µL]. Thermocycling conditions were 94°C (30s), followed by

30 cycles of denaturation at 94°C (30 s), annealing (30s, Table 1.1), and extension at 68ºC (30s), and a final extension of 68°C (5 min). Fifty-one primer pairs yielded products of the expected size without non-specific amplification and were then tested for polymorphism in seven individuals, by visualizing PCR products on 3% agarose gels. Of these 51 loci, 23 were polymorphic and used to genotype 12 to 20 individuals collected from three old-growth locations in the native range of L. tulipifera (Table 1.2). Because L. chinense, the other single species in

Liriodendron, has a restricted geographic distribution in China and Vietnam, we were not able to test for cross-species amplification. Since cross-amplification of genomic SSRs has limited success in plants (Merritt et al. 2015) and success declines as genetic divergence increases

(Barbará et al. 2007), we did not test for cross-amplification in other Magnoliaceae.

For fragment analyses, PCR products were fluorescently labeled either using primers tailed with a 5’ M13(-21) sequence following Schuelke (2000) or using primers with a 5’ fluorophore and amplified in multiplex (Table 1.1). In the tailed primer labeling method, two

PCR reactions were carried out using the same reverse primer. The first PCR used an M13(-21)- tailed locus-specific forward primer, while the second used a universal fluorescently labeled

M13(-21) as a forward primer. The products of the first PCR were purified using StrataPrep PCR

! 12 ! Purification Kit (Agilent Technologies, Santa Clara, California, USA), and then used as the template for the second PCR. Fluorescent products were electrophoresed on an ABI PRISM

3100 Genetic Analyzer, and amplicon sizes were estimated with either orange or red DNA size standard (MCLAB, San Francisco, California, USA) and GeneMapper software 3.7 (Applied

Biosystems, Foster City, California, USA) using the local southern sizing algorithm.

Microsatellite data analysis

Genotypes appeared diploid, displaying at most two alleles per locus per individual. Data were analyzed by sampled location and as a pooled population (Table 1.3). For each locus, number of alleles (k), observed heterozygosity (HO), expected heterozygosity under random

9 8 mating (HE), and polymorphism information content (PIC = 1 − 0:; 70 −

9(; 9 8 8 0:; =:0>; 270 7= ; Botstein et al. 1980) for the pooled population, were estimated using Cervus

3.0.3 (Kalinowski et al. 2007). We used Genepop 4.2 (Rousset 2008) to test deviation from

Hardy-Weinberg expected heterozygote frequency (HWE) using default values for Markov chain parameters, and to estimate the fixation index (F = (@A − @B) @A; Hamilton 2009) for the pooled population.

Population genetic parameters are listed for each sampled location showing loci exhibited

2-12 alleles, with almost all alleles common to each location (Table 1.3). Lack of population differentiation (DEF = 0.077 estimated using Genepop) justified pooling genotypes from the three locations. In the pooled population, observed and expected heterozygosities ranged from 0.233 to

0.865, and 0.272 to 0.876, respectively. Six loci showed significant deviations from HWE (Table

1.3) with deficits of heterozygotes which could have numerous causes. One hypothesis is non- random mating, which we tested using INEST (Chybicki and Burczyk 2009) with the individual

! 13 ! inbreeding model (IIM) run for 50,000 burn-in and 500,000 total cycles. The estimated average coancestry coefficient over all loci (f) was 0.041, with a 95% highest posterior density interval of

[0.000, 0.085], indicative of an outcrossing species. Another hypothesis for the deficit of heterozygosity is the presence of null alleles. Frequencies of null alleles estimated with INEST using IIM are listed in Table 1.3. The six loci with significant deficits of heterozygotes also showed evidence of null allele frequencies greater than zero.

When comparing models including null alleles (n), mating among relatives (f), and genotyping failures (b) using INEST, the full model (nfb) was found to best fit the data by the lowest deviance information criterion value (DIC = 5905.837), showing that all three parameters contributed to observed genotype frequencies. The nb model, without mating among relatives, exhibited the closest DIC value (5916.991), but the difference was greater than 10, indicating stronger support for the nfb model (Igor J. Chybicki, pers. comm.).

Conclusion

These 23 microsatellite markers do not suffer from ascertainment bias and show high levels of polymorphism. Taken together with thousands of PALs, this study provides resources for population genetic studies of L. tulipifera.

! 14 ! CHAPTER 2 Population genetic differentiation and mating system of a long-lived tree in old-growth forests and urban forest patches.

Introduction

Continuous natural habitats are increasingly transformed into anthropogenic landscapes—mosaics of natural habitats fragments surrounded by heterogeneous mixtures of different land use. Anthropogenic habitat transformation is particularly evident in the clearing of forests. The estimated cumulative loss of forests across the world over a period of 5,000 years is

1.8 billion hectares—an average net loss of 360,000 hectares per year (Williams 2002). In the continental United States, 95 percent of forests have been logged since the European settlement began (Brown et al. 1998). Since 1700, a quarter of the world’s forests have been converted to agriculture (Boakes et al. 2010). Forests that have reached a great age with little or no anthropogenic disturbance (old-growth) are increasingly rare. Less than one percent of Europe’s old-growth forests remains (Brown et al. 1998). Similarly, in the United States less than one percent of eastern forests are considered as old-growth, existing only as small scattered remnants often found in areas that either were protected or escaped harvesting for example due to their inaccessibility (Davis 1996). At the same time, urbanization continues to expand, and as a result, an increasing percentage of forests are becoming part of urban landscapes and ecosystems.

Urbanization is a dramatic anthropogenic environmental change—alterations in temperature (urban heat island effect), increased inputs of inorganic nutrients, and more impervious surface cover are only some of its consequences. Recent studies have shown that urbanization affects both nonadaptive and adaptive evolutionary processes (reviewed in Johnson and Munshi-South 2017). For example, it has been suggested that industrial pollution can elevate

! 15 ! mutation rates, e.g., in the herring gull, Larus argentatus (Yauk et al. 1996, 2000). Crested anoles, Anolis cristatellus, evolve longer limbs and more toepad lamellae (expanded subdigital scales used for clinging) in urban environments, which could increase locomotory performance on artificial surfaces (Winchell et al. 2016), representing evidence for adaptive phenotypic evolution in urban areas. Studies on urban evolution have focused predominately on relatively short-lived animal species. Considerably less is known about how urbanization affects the evolution of plants in general and long-lived trees in particular.

Johnson et al. (2015) summarized five consensus predictions for possible effects of urbanization on plants. These interrelated predictions included a range of evolutionary processes that can be impacted by urbanization. One example of these evolutionary processes is stronger genetic drift in cities as a result of reduced population sizes caused by urban encroachment or strong bottlenecks during colonization of urban areas. Johnson et al.’s (2015) prediction 1 suggests that genetic polymorphism in urban areas will be decreased as a consequence of these increased rates of genetic drift, and the results from the very few studies conducted on short- lived plants are consistent with this hypothesis. For example, the yellow toadflax, Linaria vulgaris, an herbaceous plant shows lower genetic polymorphism in urban populations than in non-urban ones (Bartlewicz et al. 2015). Considerably less is known about whether populations of long-lived trees exhibit lower genetic polymorphism in urban than in non-urban areas.

Additionally, old-growth forests are assumed to be potential reservoirs of genetic polymorphism for plants, yet there is little empirical evidence to support or contradict this view. However, if this assumption is correct, the levels of genetic polymorphism can be particularly different when comparing old-growth forests with urban forest patches.

! 16 ! Reduced rates of gene flow among urban areas is a common outcome of urbanization because the transformation of natural areas to impervious surfaces creates barriers to dispersal for some organisms (Johnson and Munshi-South 2017). This may in turn contribute to the genetic differentiation among urban areas caused by genetic drift. Johnson et al.’s (2015) prediction 4 suggests that there is genetic divergence between urban and non-urban plant populations due to decreased gene flow from non-urban to urban areas, in addition to increased genetic drift in urban areas. The degree of genetic divergence among populations is often quantified by estimators of FST (Wright 1942), which can be related to the combined processes of gene flow and genetic drift using different models. For example, under an infinite island model,

Wright (1931) showed that for diploid nuclear loci FST ≈ 1/(1+4Nem) for relatively low rates of gene flow (m). Island models are not spatially explicit and assume equal rates of gene flow among demes. As a result, all demes are equally differentiated. However, because levels of gene flow in nature tend to decline as a function of geographic distance, genetic differentiation among populations is expected to increase with geographic distance, a pattern known as isolation by distance or IBD (Wright 1943). IBD is widely used as a neutral null hypothesis in population genetic differentiation studies and can be tested by comparing levels of genetic differentiation and geographic distances between pairs of subpopulations.

Dispersal and gene flow may be asymmetric among populations, and a number of neutral models explicitly incorporating this variability have been proposed. For example, the isolation by resistance (IBR) model (McRae 2006) includes effects of irregular habitat assuming that wider or multiple dispersal pathways allow greater gene flow than a single narrow pathway. Using IBR,

McRae and Beier (2007) showed that landscape features previously considered as barriers do not block gene flow but rather change gene flow’s rate. In addition, asymmetric gene flow can be

! 17 ! due to other processes such as monopolization, where initial colonizers have a numerical advantage that reduces establishment success of later migrants, which in turn reduces gene flow among subpopulations over time (De Meester et al. 2012). Monopolization may result in a pattern of isolation by colonization (IBC). Under IBC, genetic structure reflects the temporal colonization history (founder effects), therefore there is not necessarily a relationship between genetic differentiation and geographical distances or environmental differences (Orsini et al.

2013). It is also possible that natural selection affects gene flow rate, for instance, through the process of isolation by adaptation (IBA). In this process, divergent local adaptation reduces gene flow among subpopulations because of reduced establishment success of lower fitness migrants from subpopulations with different environmental conditions, thereby increasing subpopulation divergence via genetic drift. IBA leads to a pattern of isolation by environment (IBE, Nosil et al.

2009).

Gene flow in plants is achieved by pollen and seed dispersal. However, gene flow by pollen exceeds gene flow by seed (Hamilton and Miller 2002). Gene flow by pollen is suggested to be more important in the maintenance of genetic polymorphism than gene flow by seed

(Ellstrand, 2014). Therefore, in animal-pollinated species, the pollinators and their movement through the landscapes can have a profound influence on genetic differentiation as well as on mating system. Studies on genetic differentiation and gene flow in populations of animal- pollinated plants in human-modified landscapes have largely focused on mixed agricultural- forest systems (Aguilar et al. 2008; Eckert et al. 2010). In these fragmented landscapes, pollen dispersal in insect-pollinated trees can extend several kilometers, for example, in Swietenia humilis (White et al. 2002) and Dinizia excelsa (Dick et al. 2003, reviewed in Dick et al. 2008).

Although there has been almost no research on gene flow among insect-pollinated trees in urban

! 18 ! landscapes, the very few studies comparing urban populations of trees have found similar findings to those from populations in mixed agricultural–forest landscapes. For example, Noreen et al. (2016) showed that insect-mediated gene flow persisted among individuals of the tropical tree species Koompassia malaccensis in an urban landscape, even across >2.5 km of impervious surface. These findings suggest that urban matrices may not be a barrier to gene flow for insect- pollinated trees. If gene flow happens equally along similar environments and across different environments within dispersal limits, this could lead to a pattern of IBD.

In contrast, gene flow can be strongest among similar environments (e.g., among urban sites), resulting in IBE. This pattern may arise by several mechanisms including response to natural selection in different environments, and environmentally mediated phenotypic plasticity.

Johnson et al.’s (2015) prediction 2 suggests that the ecological and environmental differences between urban and nonurban areas will cause divergent selection in these two environments.

Empirical support for this hypothesis has been found in short-lived plants such as the hawksbeard, Crepis sancta, where urbanization affects selection on seed dispersal-related traits

(Cheptou et al. 2008). In another example, Yakub and Tiffin (2017) using a common garden experiment, showed that plants of Virginia pepperweed, Lepidium virginicum, grown from seeds from urban areas bolted sooner, had fewer leaves, grew larger, had an extended time between bolting and flowering, and produced more seeds than plants grown from seeds from rural areas.

These traits have been suggested to allow plants to thrive in urban environments. Furthermore, urbanization leads to convergent environments, in which urban ecosystems in different geographic areas are more similar to one another than urban areas are to their surrounding natural environments (reviewed in McKinney 2006; Pickett et al. 2011). Based on this, parallel evolution of urban populations in response to natural selection in urban areas has also been

! 19 ! predicted. Studies that have found support for these adaptive predictions have been also conducted on short-lived plant species. For example, the white clover, Trifolium repens, consistently evolved decreased antiherbivore chemical defenses across four cities (Thompson et al. 2016). As mentioned before, an IBE pattern can also be caused by environmentally mediated phenotypic plasticity, as opposed to genetically based adaptive mechanisms. For example, within urban areas, earlier (and in a few cases later) flowering time compared to rural areas has been reported for plants, including several tree species (e.g., Hepper 2003, Jeong et al. 2011, Mimet et al. 2009, Neil and Wu 2006, Primack et al. 2004, Roetzer et al. 2000), a phenomenon that has been attributed to be a consequence of higher temperatures due to the urban heat island effect.

Shifts in flowering phenology may increase gene flow by pollen transfer among urban synchronized populations, as well as decrease gene flow among populations that differ in flowering phenology. This gene flow can lead to less genetic differentiation among urban sites compared to differentiation among old-growth forests (which may exhibit more diverse flowering phenology), and eventually can result in a pattern of IBE.

Gene flow tends to homogenize allele frequencies while genetic drift tends to cause divergence due to fixation or loss of alleles within demes. Any inference about processes that we make based on FST estimates, such as the effective number of migrants (Nem), assumes an equilibrium between these two opposite evolutionary forces, which can take a very long time (in generations) to be reached. Current genetic differentiation patterns can be the result of historical processes. Examples of historical differentiation patterns are the genetic signatures of Pleistocene glaciations in a large number of taxa (e.g., Shönswetter et al. 2002, Fraser et al. 2009, Harris and

Taylor 2009, Lanier et al. 2015), including many tree species in North America, such as

Liriodendron tulipifera (Sewell et al., 1996, Fetter 2011). Because the impacts of urbanization

! 20 ! on gene flow and genetic drift are relatively recent, they will not necessarily be readily observable in population genetic differentiation. In contrast, the mating system, a component of gene flow, might potentially be affected immediately after environmental changes associated with urbanization take place. Mating system can reach its equilibrium (Hardy-Weinberg expected genotype frequencies or HWE) in a single reproductive event, which in trees frequently takes one year (numerous tree species reproduce annually). Disturbances causing deviations from random mating will lead to departure from HWE in the offspring from that reproductive event.

However, the system will return to HWE if the population mates again at random. In plants, the mating system can be described by parameters such as outcrossing and selfing rates. Since plants are sessile, the mating system can vary in response to several factors that affect outcross pollination, such as the abundance, foraging behavior, and visitation rates of pollinators. Johnson et al.’s (2015) prediction 5 suggests that insect-pollinated plants in urban areas will have higher rates of self- pollination due to reductions in the abundance and diversity of pollinators. Similar to the case of studies of gene flow in trees, our knowledge on the impacts of anthropogenic activities on mating system of trees comes largely from studies in mixed agricultural–forest landscapes, many of which have shown that habitat fragmentation caused by deforestation decreases the outcrossing rate of several species (Aguilar et al. 2008, Eckert et al. 2010, Ferreira et al. 2013). The few studies estimating the mating system of tree populations in urban areas have documented increased selfing rates in smaller urban forest patches. For example, Wang et al. (2010) and Noreen et al. (2016) working with Chinese pine, Pinus tabulaeformis, and

Koompassia malaccensis, respectively. Yet, these studies have restricted their comparisons to tree populations in urban landscapes. There has been almost no research on comparing mating

! 21 ! system between urban forest patches and natural areas such as old-growth forests, which is one of the foci of this study.

In this study, I investigated population genetic effects of urbanization on forest tree populations using the long-lived tuliptree (Liriodendron tulipifera). Twelve polymorphic nuclear microsatellite marker loci were used to genotype single age cohort seedlings collected at nine geographic locations in the mid-Atlantic U.S. I estimated genetic polymorphism, genetic differentiation, and mating system to test versions of four of the five predictions proposed by

Johnson et al. (2015) on the effects of urbanization on plant evolution: Prediction 1, the amount of genetic polymorphism will be different between urban and non-urban populations. Prediction

2, natural selection will differ between urban and non-urban populations due to environmental changes associated with urbanization. Prediction 4, there is genetic differentiation between urban and non-urban populations. Prediction 5, insect-pollinated plants will have higher rates of self- pollination in urban areas due to reductions in the abundance and diversity of pollinators.

Materials and Methods

Study species

In this study, I utilized the tuliptree (Liriodendron tulipifera, L.), a deciduous, broadleaf tree in the family Magnoliaceae native to Eastern North America that is commercially valued for wood and as a honey tree (Beck 1990, Griffith 1991). It is both a pioneer species as well as a long-lived large canopy tree (life span is usually about 200 to 250 years, and some individuals may live up to 300 years, Beck 1990), reaching 100-150 feet (30.5!45.7 m) in height and 2-5 feet (0.6!1.5 m) DBH at maturity. L. tulipifera is self-compatible, with a mixed-mating system

(Brotschol et al. 1986, Griffith 1991, Houston and Joehlin 1989, Sewell 1992), pollinated (in

! 22 ! decreasing order of abundance) by flies (Muscidae), beetles (Coleoptera), honeybees (Apis spp.), and bumblebees (Bombus spp.; Beck and Della-Bianca 1981), and has wind-dispersed seeds that remain viable for up to four years under natural conditions in the forest litter (Clark and Boyce

1964). Therefore, seedlings are representative of recent ecological and evolutionary processes.

L. tulipifera is widespread and a component of numerous native eastern forests from southern New England, west through southern Ontario and Michigan, south to Louisiana and central Florida (Beck and Della-Bianca 1981), including the heavily urbanized mid-Atlantic region. Although several native species do not do well in cities and no longer occur there, within its range, L. tulipifera is found in both old-growth remnants and forest patches embedded in urban areas where it reproduces and recruits naturally, which makes it an excellent species model for this project. In addition, because the seedlings stage is the most sensitive phase of the life cycle to environmental changes (Harper 1977, Silvertown and Charlesworth 2001, Leck et al.

2008), seedlings are expected to be particularly sensitive to anthropogenic habitat disturbance such as urbanization.

Most old-growth forests in North America are not considered virgin forests (i.e. free from disturbance of modern humans) due to extensive encroachment, hunting, agriculture, and timber harvesting. In the case of eastern temperate forests, there are several definitions of old-growth, with no clear thresholds regarding disturbance frequency or age (White and Lloyd 1994, Leverett

1996). The old-growth sites analyzed in this study were chosen from the study of McGarvey et al. (2015), who used the operational definition of old-growth as forests with a stand age of ≥ 150 years, following Cogbill (1996). Furthermore, the metropolitan areas investigated here were originally mostly covered by forests in which L. tulipifera was commonly found (Ward 1881,

! 23 ! Adams et al. 2003). The urban forest patches in this study are remnants of the original native forests which were encroached by urban development.

Field collections

I collected one single age cohort of seedlings (age zero) in urban forest patches in large metropolitan areas and in old-growth forest remnants in the mid-Atlantic U.S. during the summers of 2015 and 2016. In addition, a suburban forest close to the north geographic range limit of this species was sampled (Southbury, Connecticut; Table 2.1). To ensure that seedlings were one single age cohort and had germinated in the year they were collected, sampled seedlings had both cotyledons still attached and the first set of true leaves. At each sampling site,

42-48 seedlings were collected haphazardly in an area of approximately 2,000 m2. The total sample size was 426 seedlings from 9 geographic locations, four old-growth, four urban, and one suburban (Table 2.1).

Genetic analysis

Twelve microsatellite markers previously developed specifically for L. tulipifera

(Gutierrez-Ozuna and Hamilton 2017) were used to genotype all sampled individuals. Three multiplex microsatellite sets (Table 2.2) were amplified in 6.75µL multiplex PCR reactions with the Type-it Microsatellite PCR Kit (QIAGEN, Valencia, California, USA). Each reaction contained 3.125µL Type-it Multiplex PCR Master Mix, 1.25 µL 10x primer mix (1-20µM of each primer, see Table 2.2 for primer-specific concentrations), 1.375 µL ddH2O, and 1 µL of

DNA template (concentration was not determined). Thermocycling conditions were 5 min at

95ºC, followed by 32 cycles of 30 s at 95ºC, 90 s at 57ºC, and 60 s at 72ºC, and finally 30 min at

60ºC to promote 3’ untemplated A nucleotide addition. Fluorescent PCR products were

! 24 ! electrophoresed on an ABI PRISM 3100 Genetic Analyzer, using either orange or red DNA size standard (MCLAB, San Francisco, California, USA). Alleles were binned and scored with the microsatellite plugin version 1.4.4 implemented in Geneious version 9.1.3

(https://www.geneious.com, Kearse et al. 2012) using the local southern sizing algorithm and were double-checked manually. To create expected allele size bins, I estimated expected fragment sizes from the microsatellite sequences reported in Gutierrez-Ozuna and Hamilton

(2017) with ±1 base pair estimation error.

Statistical analysis

Genetic polymorphism. To test prediction 1 that the amount of genetic polymorphism is different between urban and non-urban populations, for each sampling site, I estimated the following measures of genetic polymorphism averaged over the 12 loci using GenAlEx version

6.5 (Peakall and Smouse 2006, 2012): number of alleles (ka); effective number of alleles (ke =

H 8 1/( 0:; 70 )), where pi is the frequency of the ith allele and k is the number of alleles; number of private alleles per locus (kp); observed heterozygosity (HO = observed frequency of heterozygotes / n), where n is the number of genotyped individuals; and expected heterozygosity

H 8 (HE = 1 − 0:; 70 ). In addition, to gain insight into the mating system, the fixation index (F =

(HE - HO) / HE) was estimated for each sampling site.

Population genetic differentiation. Because FST depends on the distribution of allele frequencies (Charlesworth 1998, Hedrick 1999, Jakobsson et al. 2013), it may not range from 0.0 to 1.0 with more than two alleles and can be very small even when no alleles are shared between two sites. Therefore, to test prediction 4 that urban and old-growth populations have diverged, I employed a “standardized” measure of FST, called F’ST (Hedrick 2005). F’ST is calculated by

! 25 ! dividing observed FST for a given locus by the maximum theoretical FST based on the heterozygosity at that locus. Unlike FST, F’ST will be 1.0 at complete differentiation, even with high genetic polymorphism within sites, and will be zero when there is no differentiation. F’ST was estimated using GenoDive (Meirmans and Van Tienderen, 2005). To evaluate if the degree of genetic differentiation among sites was due to differences among habitats (old-growth vs. urban), I assessed the partition of total genetic polymorphism between forest types (old-growth vs urban), among and within sampling sites, and within individuals by carrying out a hierarchical analysis of molecular variance (AMOVA) on Euclidean pairwise distances among individuals implemented in GenAlEx. The expected distribution to test for statistical significance of variance components using permutation tests was based on 999 random permutations. In addition, the following five hierarchical fixation indices (F-statistics) were estimated using GenAlEx: 1) the proportion of genetic variation accounted for by allele frequency differences between forest

8 8 8 8 types, i.e., old-growth vs urban (FRT = I3 [I3 + I- + IL ]); 2) the proportion of genetic variation accounted for by allele frequency differences among sampling sites within a forest type

8 8 8 (FSR = I- [ I- + IL ]); 3) the proportion of genetic variation accounted for by allele frequency

8 8 8 8 8 8 differences among sampling sites (FST = [I3 + I- ] [I3 + I- + IL ]), where I3 is the variance

8 component due to differences between forest types, I- is the variance component due to

8 differences among sampling sites, and IL is the variance component due to differences among alleles within sampling sites; 4) deviation from Hardy-Weinberg expected genotype frequencies

(HWE) in a sampling site (FIS = [HS – HO] / HS); and 5) deviation from HWE in the total population (FIT = [HT – HO] / HT ), where HT is the expected heterozygosity in the total population, HS is the mean expected heterozygosity in sampling sites, and HO is the mean observed heterozygosity in sampling sites.

! 26 ! Pattern of genetic differentiation. Prediction 2 suggests that plant populations will experience divergent selection between urban and old-growth forests, which can eventually result in a pattern of IBE. To test the neutral null hypothesis of genetic differentiation among populations, IBD, the correlation between a matrix of linearized genetic distances F’ST / (1 −

F’ST) (Rousset 1997) and natural logarithm (ln) of geographic distances in kilometers (km) was examined. I tested the significance of the observed relationship by using a Mantel test (Mantel

1967) with a total of 10 000 random permutations used to generate an empirical null distribution for the data set as implemented in GenoDive.

Mating system. In order to test prediction 5 that insect-pollinated plants will have higher rates of self-pollination in urban areas, I estimated the mating system at the sampling site level using two methods. The first, according to the mixed mating model for many unlinked loci described by Ritland and Jain (1981), using MLTR program from Ritland (2002). The mating system parameters estimated were multilocus outcrossing rate (tm), single-locus outcrossing rate

(ts), mating among relatives (commonly referred to as biparental inbreeding, tm − ts) and multilocus paternity correlation (rp(m)) or proportion of full sibs among outcrossed seedlings.

These parameters were estimated using the maximum-expectation algorithm, and allowing for the presence of null alleles. Standard errors of the estimates were approximated as the standard deviation of 1000 bootstrap replicates. Furthermore, neighborhood size, i.e., the number of pollen donors, was estimated as 1/rp(m) (Ritland, 1989). In the second method, Bayesian estimates of population outcrossing rates (t-max) and fixation indices (F) were estimated with BORICE

(Koelling et al. 2012) based on a chain of 100 000 steps with a burn-in of the first 10000 steps.

! 27 ! Koelling et al. (2012) showed that with simulated data, BORICE provides accurate estimates of outcrossing rate when family sizes are small and maternal genotypes are unknown, as is the case here.

Results

Genetic polymorphism

The levels of genetic polymorphism were similar between urban and old-growth forests

(Table 2.3). The average number of alleles per locus ranged from 6.08 to 6.33 in urban forest patches and from 6.17 to 6.83 in old-growth forests. The average effective number of alleles varied from 3.59 to 4.07 in urban forests and from 3.39 to 4.47 in old-growth forests. Private alleles were infrequent, being absent in two urban populations (average frequency per locus ranged from 0 to 0.25). All sampling sites showed high values of heterozygosity. The average HE values varied from 0.67 to 0.70 in urban forests and from 0.66 to 0.72 in old-growth forests. All corresponding average HO values were slightly lower, ranging from 0.56 to 0.60 in urban forests and 0.58 to 0.60 in old-growth forests. The fixation index varied from 0.13 to 0.20 in urban forests and from 0.10 to 0.17 in old-growth forests.

Population genetic differentiation

Because observed heterozygosities were very high, maximum DEF was much less than one, e.g., global maximum DEF was 0.285 when considering the four urban and four old-growth sampling sites, and 0.301 when Southbury was included. Global D′EF for the eight sites (four urban and four old-growth) was 0.237, and 0.273 when Southbury was included. Pairwise D′EF ranged from 0.087 to 0.436 (Table 2.4). When sampling sites were grouped by forest type, D′EF

! 28 ! was higher when considering the four old-growth sites (0.242) compared to the four urban sites

(0.155). The average geographic distance among old-growth forests was 200.5 km and among urban forest patches was 26.4 km.

Pattern of genetic differentiation

Mantel's test showed a significant correlation between genetic distance and geographical distance when eight sampling sites (the four urban and four old-growth) were considered (r =

0.648, p < 0.001), and this correlation increased when Southbury was included (r = 0.759, p =

0.003; Figure 2.1). Analysis of molecular variance (AMOVA) based on D′EF values indicated that most of the genetic polymorphism (73%) occurred within individuals, also known as heterozygosity. The variation among individuals contributed 19% and the variation among sampling sites contributed 5% to the observed genetic polymorphism. Differences between urban and old-growth sites explained 2% of the total genetic polymorphism (Table 2.5). Results were largely the same when Southbury was included in the analysis.

Mating system

The fixation index (F) estimated with GenAlEx and BORICE were low in all sampling sites, ranging from 0.13 to 0.20, and from 0.06 to 0.18, respectively. FIS was low, being 0.15 when estimated using GenAlEx and 0.10 when estimated using BORICE. Estimates of the outcrossing rate were high in all sampling sites. Based on the results from MLTR, the multilocus outcrossing rate (tm) and single-locus outcrossing rate (ts) ranged from 0.800 to 1.000 and from

0.804 to 1.000, respectively (Table 2.6). The difference between the multilocus and the single- locus outcrossing rates (tm−ts) ranged from -0.004 to 0.008. It was positive in most sampling sites, suggesting that some mating among relatives occurred. Based on the multilocus paternity

! 29 ! correlation, the estimated number of pollen donors varied from 21.74 to 125.00. Using BORICE, the posterior distributions for outcrossing rate (t-max) were high, ranging from 0.69 to 0.91

(Table 2.6).

Discussion

Genetic polymorphism

The urban forest patches investigated here are remnants of former native forests that were once more extensive, but were encroached by urban development. Prediction 1 suggests that loss of natural habitat is expected to reduce population sizes, which will result in lower genetic polymorphism within urban populations when compared to non-urban populations. However, contrary to this prediction, urban forests and old-growth forests exhibited similar levels of genetic polymorphism. These results suggest that the ecological and environmental differences between old-growth and urban forests did not result, at least in the time since urbanization, in loss of genetic diversity in this long-lived tree in urban areas as observed in short-lived species such as the yellow toadflax, Linaria vulgaris (Bartlewicz et al. 2015). Additionally, this lack of differences in genetic polymorphism between urban and old-growth forests does not support the assumption and limited empirical evidence that old-growth forests are reservoirs of genetic polymorphism. Our results are in contrast to some studies of herbaceous plants that reported a positive correlation of genetic polymorphism and forest age. For example, the broad buckler- fern, Dryopteris dilatata (Reisch et al. 2007), and the white trillium, Trillium grandiflorum

(Vellend 2004). In contrast, similar to our results, some of the few studies on trees have found that old-growth forests do not harbor higher genetic polymorphism. For example, Mosseler et al.

(2003) did not find differences in some measures of genetic polymorphism such as the mean

! 30 ! number of alleles per locus and the mean observed heterozygosity in red spruce, Picea rubens, between old-growth and young forest stands. Another example is Zhu and Lou (2013), who compared Oriental thuja, Platycladus orientalis, old-growth (> 300 yrs.) populations and young

(< 100 yrs.) populations in both planted and natural forests. They did not find differences in genetic polymorphism including mean observed heterozygosity and allelic richness among old- growth populations (both natural and planted) and young natural populations. In another example, Chhatre and Rajora (2014) found that old-growth and second-growth natural populations of the eastern white pine, Pinus strobus, have similar levels of genetic polymorphism measured, for example, as mean number of alleles per locus, allelic richness, and mean observed heterozygosity.

Prediction 1 is based on one of the most commonly reported patterns of urban evolution in short-lived organisms, decreased genetic polymorphism within urban populations, which has been attributed to increased genetic drift. Thus, the lack of differences in genetic polymorphism between urban and old-growth forests observed in this study may be due to similar rates of genetic drift in these two environments. A version of this hypothesis was tested in Chapter 3.

Population genetic differentiation

Genetic differentiation was greater among old-growth forests than among urban forests, which could be interpreted as support for prediction 4 that urban and non-urban populations have diverged genetically. However, this could also be interpreted as a result of greater geographic distances among old-growth forests than among urban forest patches (see discussion below about the pattern of genetic differentiation). In addition, AMOVA indicated that most of the genetic polymorphism in L. tulipifera occurred both within individuals and within sampling sites, while

! 31 ! the differences among sites explained only a small portion of the total genetic polymorphism.

This result is similar to that found by Kovach (2009), who assessed L. tulipifera genetic polymorphism in populations from two physiographic regions in southeastern U.S. (mountains vs. coastal plain) using AFLPs, and showed that, while about 89% of the total polymorphism is maintained within populations, differences among populations (which spanned an area similar to the sampling area here) explained about 6%. These results are consistent with the general observed pattern that tree species maintain more polymorphism within rather than among populations, even when samples are taken from across a species' geographic range (e.g., Hamrick et al. 1992, Hamrick and Godt 1996, Loveless and Hamrick 1984). In this study, the lowest percentage of genetic polymorphism (2%) was explained by the differences between forest types

(urban vs. old-growth). This result was also similar to the study by Kovach (2009) wherein differences between mountains and coastal plain regions explained about 5% of the total polymorphism (although this percentage was not statistically different than zero).

Knowledge of the levels and patterns of genetic polymorphism are critical to directing conservation efforts. For example, most of the genetic polymorphism may occur within populations (little differentiation) or among populations (higher differentiation). Each of these situations requires different strategies for the conservation of genetic polymorphism. In this study, genetic differentiation among populations was low. Thus, the conservation of a particular forest remnant may be less important, because its genetic polymorphism will also be maintained elsewhere. On the contrary, if there had been high genetic differentiation among populations, it would have been required that every one of the differentiated populations be protected in order to conserve genetic polymorphism. This study shows the value of urban forest patches for the conservation of genetic polymorphism.

! 32 ! Pattern of genetic differentiation

The genetic differentiation among L. tulipifera populations in the mid-Atlantic U.S. was characterized by a pattern of isolation by distance (IBD). This hypothesis was supported by a significant Mantel test between a matrix of linearized D′EF and a matrix of geographical distance.

This finding does not support prediction 2 that plant populations will experience divergent selection between urban and nonurban environments, because this type of selection could eventually result in a pattern of isolation by environment (IBE) instead of IBD. Finding IBD in this study is not completely surprising because even insect-pollinated trees whose pollen dispersal can extend several kilometers in fragmented landscapes have limited dispersal. IBD is ubiquitous in nature. A significant correlation between genetic and geographic distance has been found in most of the studies in which IBD has been tested (Meirmans 2012). In contrast, IBE has been observed in fewer species. For example, Lee and Mitchell-Olds (2011) showed that water availability explains the discrete pattern of genetic divergence (east-west) that the mustard

Boechera stricta has in its native range. However, Lee and Mitchell-Olds (2011) did not completely rule out IBD, which played a fundamental role in the continuous pattern of divergence between north and south genetic groups of this mustard species.

Low gene flow and strong selection against genotypes adapted to other habitats are among the factors predicted to promote local adaptation (Kawecki and Ebert 2004). Consistent with this, most empirical evidence of local adaptation comes from herbaceous plants with low levels of gene flow and under strong selection. In contrast to gene flow in herbaceous plants, gene flow is strong and extensive in trees, as suggested for instance by the low levels of genetic differentiation among tree populations. For example, when measured as GST (a multi allelic analogue of FST among a finite number of demes; Nei 1973), population genetic differentiation ! 33 ! was on average for woody long-lived perennial species 0.084 and for annuals 0.355 (Hamrick et al. 1992). Because of its homogenizing influence, strong and extensive gene flow in trees can oppose not only the divergence caused by genetic drift but also can counteract all but the strongest local natural selection, preventing local adaptation. Therefore, adaptation to urban environments and parallel evolution of urban populations in response to natural selection in urban areas are likely not realistic for many tree species.

Mating system

Previous studies have shown that L. tulipifera mating system can vary among populations and geographic regions with different environments. For example, using allozymes, Brotschol et al. (1986) reported large outcrossing rate differences between mountain and coastal plain populations in North Carolina (t = 0.86 vs. t = 0.55, respectively). In contrast, in this study, populations in all sites were found to be almost completely or completely outcrossing. As the multilocus outcrossing rates were equal or very close to unity, the selfing rate can be considered very low to zero in each sampling site, regardless of urban forest type, which does not support prediction 5 that insect-pollinated plants will have higher rates of self- pollination in urban areas.

In flowering plants, a major factor favoring the evolution of outcrossing is inbreeding depression

(Lloyd 1979, Lande and Schemske 1985, Schemske and Lande 1985). Self-compatible trees, including species with mixed mating systems often display high inbreeding depression (Duminil et al. 2009, Ishida, 2006; Robertson et al. 2011, Rodger and Johnson 2013). Although one possible explanation for the fully or almost fully outcrossing mating system observed in all the populations is embryo abortion due to early inbreeding depression, Sewell (1992) showed through controlled pollinations that L. tulipifera is capable of seed germination from self- fertilized ovules. Sewell (1992) suggested that in L. tulipifera, inbreeding depression caused by a

! 34 ! pre-fertilization or post-pollen tube germination mechanism is unlikely because self and outcross pollen are equally competitive at the early stages of pollen tube germination. Thus, if L. tulipifera exhibits inbreeding depression, it must take place after the germination of the pollen tube. As inbreeding depression can occur after fertilization, sampling late stages in the life cycle may decrease the ability to detect mating system differences. For example, Kanashiro (1990) observed increases in heterozygosity in L. tulipifera from age class 0 (newly germinated seedlings) to age class 1 (saplings of 3-5 years old), and from age class 1 to age class 2 (saplings of 7-10 years old), which he interpreted as due to the elimination of homozygous genotypes because of lower fitness. In this study, to maximize the capture of germinated seeds from self- fertilization, I sampled seedlings of age class zero. This allowed me the estimation of mating system early in the life cycle before inbreeding depression could be fully manifest. Therefore, it is unlikely that the lack of differences among populations and between forest types is explained

(at least completely) by inbreeding depression early in the fertilization process.

All populations sharing a fully outcrossed mating system might suggest that there is no pollinator limitation in any of the sampling sites analyzed that could lead to reduction of outcross-fertilized ovules. This assumption might be supported by studies showing that several types of human disturbance, including urbanization, are not detrimental to pollinators. For example, in a meta-analysis, Winfree et al. (2009) found that although both abundance and species richness of unmanaged bees were significantly reduced by different anthropogenic disturbance types (including logging, agriculture, and fires), the effects were not strong. These effects were statistically significant only in study systems where extremely little natural habitat remains. In addition, Winfree et al. (2009) found that the abundance of managed honey bees was not associated with degree of anthropogenic disturbance. Furthermore, gene flow can even be

! 35 ! enhanced if urban areas support more abundant and diverse pollinator communities than old- growth forests. For example, Carper et al. (2014) observed higher bee abundance and species richness in suburban forests in a metropolitan area when compared to natural forests. These authors found that at the local scale, bee abundance and species richness were positively related to the abundance and richness of flowering species within forests, while at the landscape scale they were positively related to the proportion of developed open areas (e.g., yards and roadsides).

Additionally, our data suggest that some mating among relatives occurred in each of these sites.

This hypothesis is supported by most sites showing a positive difference between the multilocus and the single-locus outcrossing rates (tm−ts).

The time to equilibrium between gene flow and genetic drift that produces population genetic differentiation is long. Urbanization is a relatively recent process compared to the genetic origin of the populations. Therefore, it is not surprising that we did not observed genetic divergence between urban and old-growth populations. This is one potential explanation why many studies of forest trees have not found significant population genetic consequences of fragmentation (e.g., reduced genetic polymorphism, and increased genetic structure) due to recent human activities such as logging, which has been referred to as the “paradox of forest fragmentation genetics” (Kramer et al. 2008). In contrast, due to its dynamic nature and short time scale, mating system can be influenced by recent and rapid environmental changes brought on by urbanization. The estimation of the mating system allowed me to rule out the alternative explanation that the lack of any urbanization effects was an artifact of the time to register changes of the genetic measures used. The lack of differences in mating system instead suggests that there are indeed no effects of urbanization on some key ecological and evolutionary processes that shape genetic polymorphism within and among populations of this long-lived tree.

! 36 ! CHAPTER 3 Small effective population sizes in a common, widespread, long-lived tree

Introduction

The genetic effective population size (Ne), a concept introduced by Wright (1931), is the population genetic parameter that determines the rate of genetic drift in an idealized Wright–

Fisher population with discrete generations and sampling restricted only to the gamete pool

(Fisher 1930; Wright 1931). Ne is required to predict levels and dynamics of genetic polymorphism because allele frequency changes in finite populations are dictated by the balance between genetic drift and the other evolutionary forces. For example, the balance between the input of new mutations and genetic drift (4Neµ for diploid nuclear loci) determines the expected levels of genetic polymorphism. The balance between genetic drift and gene flow (4Nem for diploid nuclear loci in a finite island model) determines the expected levels of allele frequency divergence among subpopulations or demes (genetic differentiation). Also, the effectiveness of natural selection to maintain advantageous alleles or purge deleterious ones is determined by the product of Ne and the intensity of selection (Nes, where s represents the selection coefficient,

Charlesworth 2009, Hamilton 2009). Because of its role in these fundamental evolutionary processes, estimates of Ne are a powerful predictive tool in conservation genetics to maintain both genetic polymorphism and adaptive potential of populations under anthropogenic environmental change (Frankham 1995, Schwartz et al. 2007, Palstra and Ruzzante 2008, Hare et al. 2011).

Worldwide, urban centers are increasing in area and in human population size (Grimm et al. 2008). Urbanization results in severe environmental change, including alterations in

! 37 ! temperature (urban heat island effect), more impervious surface cover, and elevated pollution.

Furthermore, the similar environments associated with urbanization are predicted to have evolutionary consequences such as convergent adaptation and isolation by environment (IBE).

Recent studies have implicated urbanization as a cause of stronger genetic drift in short-lived animals such as mice and salamanders. This assumption comes from the most commonly reported patterns of urban evolution, decreased genetic polymorphism within and increased genetic differentiation among urban populations (reviewed in Johnson and Munshi-South 2017).

Similarly, in plants, stronger genetic drift in urban areas is one of the consensus predictions about how urbanization is expected to affect evolution in urban environments (see Johnson et al.

2015). However, few direct measurements of genetic effective population size (Ne) have been made in the context of urbanization (but see Unfried et al. 2012).

There are several genetic methods to estimate (Ne), which can be divided into temporal and single-sample methods. Temporal methods use variance in allele frequency between samples from the same population collected from at least two points in time. Single-sample methods can be based on neighborhood size (Wright 1943), heterozygote excess when Ne is ≤ 30 (Pudovkin et al. 1996, Balloux 2004, Zhdanova and Pudovkin 2008), coancestry (Nomura 2008), and linkage disequilibrium (hereafter LD; Sved 1971; Hill 1977, 1981; Waples 2006, 2008; Sved 1971; Sved et al. 2013). The logic behind Ne estimates using LD is that correlation between alleles at pairs of loci are not expected at unlinked loci in an infinitely large population. However, in a finite population allelic correlations arise just by chance. The smaller Ne, the larger the expected LD

(Hill 1977, 1981). Methods to estimate Ne based on LD have advantages over other methods. For example, with microsatellite data, Ne estimates based on LD tend to have higher precision than

! 38 ! estimates based on the temporal method, if the latter is based on samples taken few generations apart (Waples and Do 2010).

Methods to estimate Ne based on LD assume a Wright-Fisher population model with discrete generations and semelparity. However, actual biological populations depart from this model, for instance, because of iteroparity (termed polycarpy for plants) and overlapping generations, as is the case of long-lived plant species. In age-structured populations, each cohort of individuals experiences genetic drift in different ways and at different times during the life cycle. For example, through: i) age-specific mortality, because each age class has a different expected probability of surviving into the next age class via random survival (lx), and ii) expected fecundity at age x (bx), because the parents are individuals of different age classes contributing unequally to the pool of gametes (Felsenstein 1971; Jorde and Ryman 1995). In populations with overlapping generations, analysis of individuals belonging to a single age cohort provides estimates of the effective size in a given age cohort, known as the effective number of breeders or Nb (Waples 1989). Nb expresses the strength of genetic drift in one reproductive event. Both Ne and Nb are influenced by the same factors, however, they measure drift over different temporal scales. Nb mainly reflects local demographic and evolutionary processes that occur on a seasonal or year to year basis, in contrast to the generational basis of

Ne. Understanding and estimating Nb is important for several reasons. Estimates of Nb may be more suitable to elucidate evolutionary and ecological processes occurring at short timescales than estimates of generational Ne (see Waples and Antao 2014). An example is in understanding the impacts of urbanization in trees where Nb measures contemporary genetic drift.

Estimates of Nb are also important because they may be used to predict generational Ne when the relationship between these two parameters is known. For example, in populations with

! 39 ! discrete generations and complete reproductive synchrony, Nb and Ne are equal (Hare et al.

2011). In contrast, Nb and Ne can be very different under other types of life history (Waples

1990a, 1990b; Waples et al. 2013). In organisms with age-structure, most of our knowledge on the relationship between Nb and Ne has come from work on semelparous fish species, often economically important Salmoniformes (e.g., Waples 1990a, 1990b, 2002, 2005). In contrast, so far distinction between Nb and Ne has rarely been made in plants, where Nb is often estimated as generational Ne and as neighborhood size (e.g., Schaal 1980). Further, while young-of-the-year sampling is commonly carried out in fish and other vertebrates for Nb estimates, such samples are rarely obtainable for tree species. Instead, what is commonly done is to sample mixed-age individuals, principally reproductively mature adult trees. Waples and Do (2010) hypothesized that LD in mixed-age individuals should approximately estimate Ne if individuals are sampled at random from a number of consecutive age cohorts roughly equal to the generation length. Some studies such as Robinson and Moyer (2013), and Waples et al. (2014) have found support for this hypothesis by simulating age-structured genetic data for several iteroparous species and applying a method based on LD to estimate Ne. These studies also found that differences between Nb and generational Ne values can be explained by differences in simple life history traits. However, these studies were focused on relatively short-lived species. For example, in Waples et al.

(2014), the bottlenose dolphin (Tursiops truncatus) had the highest adult lifespan (AL, maximum number of years during which an individual can reproduce), which was 27 years. Less is known about the relationship between Nb and Ne in long-lived, iteroparous organisms such as numerous tree species where maximum age is much longer (Condit et al. 1995; Chambers et al. 1998) and, as a consequence, the range of age cohorts present among the breeders is usually larger, and mixed-age samples are likely taken from non-consecutive age cohorts.

! 40 ! Nb/Ne ratio in iteroparous species, including some plants, is highly correlated with two simple life-history traits frequently influenced by growth rate, age at reproductive maturity (", first age with bx > 0) and adult lifespan (AL, maximum age ! " + 1; Waples et al. 2013). Studies have shown that urbanization has effects on growth rates of trees that could potentially alter " and AL, and thereby impact Nb/Ne. For example, Searle et al. (2012) showed increased growth in

Quercus rubra L. urban-grown seedlings relative to those grown at rural sites. In a major analysis of urban and rural trees around the world, Pretzsch et al. (2017) showed that urban trees in the boreal and subtropical zones grow faster than their rural counterparts. In contrast, urban trees in temperate zones grew significantly less than rural ones. Increased growth in cities is presumably due to the urban heat island effect (higher temperatures), that may stimulate photosynthetic activity and extend the growing season. Higher CO2 concentrations, larger atmospheric N-deposition, and lower ozone concentration in urban areas compared to the surrounding landscapes might also foster growth rates. Accelerated growth may mean earlier ages at maturity, more rapid ageing, and shortened lifespan, which in turn can have effects on Nb,

Ne, and their relationship.

Although estimates of Ne and Nb based on LD have recently expanded, especially in vertebrates, they remain rare in plants. The few examples include American eelgrass, Vallisneria americana (Hydrocharitaceae; Lloyd et al. 2012); rice, Oryza sativa (Poaceae; Morais Júnior et al. 2017); and palms, e.g., Chamaedorea alternans (Arecaceae; Peñaloza-Ramírez et al. 2016).

Even fewer empirical studies have focused in long-lived plants, including cycads, e.g., several species of Zamia (Zamiaceae; Meerow et al. 2012), and Cycas (Cycadaceae; Feng et al. 2016,

2017); and trees, e.g., Euptelea pleiospermum (Eupteleaceae; Wei and Jiang 2011), Fagus crenata (Fagaceae; Inanaga et al. 2016), and Platanus racemosa (Platanaceae; Johnson et al.

! 41 ! 2016) but none of them to directly compare the strength of genetic drift in urban and non-urban habitats.

This study estimated the genetic effective population size in both urban and old-growth populations of the long-lived tuliptree, Liriodendron tulipifera, in the mid-Atlantic U.S. Twelve polymorphic nuclear microsatellite marker loci were used to genotype single age cohort seedlings collected at nine geographic locations, as well as reproductively mature mixed-age trees sampled at three of the same sites. The observed average LD between pairs of loci was used to estimate ,- for seedlings and the effective population size for the mixed-age adult trees

(,. /01.2(34. ). To test the consensus prediction that genetic drift will be stronger in urban areas

(Johnson et al.’s [2015] prediction 3), I compared ,- between urban and old-growth populations.

To evaluate sampling approaches, ,- and ,. /01.2(34. were compared in the context of demographic models that predict the relationship between Nb and generational Ne for long-lived iteroparous species such as L. tulipifera. This latter comparison is critical to the interpretation of the effective population size estimates and is especially relevant to studies of tree species where seedling sampling is often difficult or impossible. It was observed that ,- estimates did not differ between urban and old-growth sites and were uniformly small even in old-growth populations that could be considered as reservoirs of genetic polymorphism. In addition,

,. /01.2(34. estimates were also small and did not differ from ,- estimates, consistent with modeling results that showed Nb and Ne are expected to be very similar in this long-lived iteroparous species.

! 42 ! Materials and Methods

Study species

The tuliptree (Liriodendron tulipifera, L.) is a diploid (Liang et al. 2011), deciduous, broadleaf tree in the family Magnoliaceae native to Eastern North America that is commercially valued for wood and as a honey tree (Beck 1990). It is both a pioneer species as well as a long- lived large canopy tree whose lifespan is usually about 200 to 250 years (some individuals may live up to 300 years; Beck 1990, Busing 1995), with age at first reproduction (seed production) reported as 15 to 20 years (Olson 1969; Renshaw and Doolittle 1958; Lusk et al 2007). L. tulipifera reaches 100-150 feet (30.5!45.7 m) in height and 2-5 feet (0.6!1.5 m) DBH at maturity. It is a prolific seeder, with a common seedfall of 741,000 to 1,482,000/ha (300,000 to

600,000/acre; Beck 1990). Seeds are wind-dispersed and remain viable for up to four years under natural conditions in the forest litter (Clark and Boyce 1964), indicating that L. tulipifera does not form a persistent soil seed bank. L. tulipifera is widespread and a component of numerous native eastern forests from southern New England, west through southern Ontario and Michigan, south to Louisiana and central Florida (Beck and Della-Bianca 1981), including the heavily urbanized mid-Atlantic region. Although several native species do not do well in cities and no longer occur there, within its range, L. tulipifera is found in both old-growth remnants and forest patches embedded in urban areas where it reproduces and recruits naturally, which makes it an excellent species model for this project.

! 43 ! Field collections

Sampling was carried out during the summers of 2015 and 2016. To estimate effective number of breeders (,-), one single age cohort of seedlings (age zero) were collected from nine geographic locations (four urban forest patches, four old-growth forest remnants, and one suburban forest) in the mid-Atlantic U.S. (Table 3.1). The metropolitan areas investigated here were originally mostly covered by forests in which L. tulipifera was commonly found (Ward

1881; Maryland Geological Survey 1929; USDA-NRCS 1998; Adams et al. 2003). The urban forest patches in this study are remnants of the original native forests which were encroached by urban development. Most old-growth forests in North America are not considered virgin forests

(i.e. free from disturbance of modern humans) due to extensive encroachment, hunting, agriculture, and timber harvesting. In the case of eastern temperate forests, there are several definitions of old-growth, with no clear thresholds regarding disturbance frequency or age

(White and Lloyd 1994, Leverett 1996). The old-growth sites analyzed in this study were chosen from the study of McGarvey et al. (2015), who used the operational definition of old-growth as forests with a stand age of ≥ 150 years, following Cogbill (1996). The suburban site sampled in this study was located close to the north geographic range limit of this species (Southbury,

Connecticut). At each sampling site, 42-48 seedlings were collected haphazardly in an area of approximately 2,000 m2. To ensure that seedlings were one single age cohort and had germinated in the year they were collected, sampled seedlings had both cotyledons still attached and the first set of true leaves. To obtain estimates from mixed-age samples (,. /01.2(34. ), 12-20 adult trees were haphazardly sampled from three of the nine locations where seedlings were collected

(Table 3.1). The total sample size was 426 seedlings, and 52 adult trees.

! 44 ! Genetic analysis

Twelve microsatellite markers previously developed specifically for L. tulipifera

(Gutiérrez-Ozuna and Hamilton 2017) were used to genotype all sampled individuals. Three multiplex microsatellite sets (Table 3.2) were amplified in 6.75µL multiplex PCR reactions with the Type-it Microsatellite PCR Kit (QIAGEN, Valencia, California, USA). Each reaction contained 3.125µL Type-it Multiplex PCR Master Mix, 1.25 µL 10x primer mix (1-20µM of each primer), 1.375 µL ddH2O, and 1 µL of DNA template (concentration was not determined).

Thermocycling conditions were 5 min at 95ºC, followed by 32 cycles of 30 s at 95ºC, 90 s at

57ºC, and 60 s at 72ºC, and finally 30 min at 60ºC to promote 3’ untemplated A nucleotide addition. Fluorescent PCR products were electrophoresed on an ABI PRISM 3100 Genetic

Analyzer, using either orange or red DNA size standard (MCLAB, San Francisco, California,

USA). Alleles were binned and scored with the microsatellite plugin version 1.4.4 implemented in Geneious version 9.1.3 (https://www.geneious.com, Kearse et al. 2012) using the local southern sizing algorithm and were double-checked manually to create expected allele size bins,

I estimated expected fragment sizes from the microsatellite sequences reported in Gutiérrez-

Ozuna and Hamilton (2017) with ±1 base pair estimation error.

Effective number of breeders and effective population size

The microsatellite genotype data were used to estimate ,- and ,. empirically per sampling location with the LD-based method. ,- estimates were obtained using exclusively seedlings of age class zero. In addition, because it has been shown that generational Ne is best estimated using random samples of reproductively mature adults (Robinson and Moyer 2013),

,. estimates were obtained using exclusively reproductively mature mixed-age trees

! 45 ! (,. /01.2(34. ). Although commonly referred to as methods based on LD, they rely on the correlation of alleles at independent (not physically linked) loci. Effective population size is estimated from observed average disequilibrium between several pairs of loci in multiple individuals sampled from one population. With pairs of loci experiencing free recombination, two-locus disequilibrium is expected to decay to nearly zero in few generations. This observed disequilibrium, measured by O8, is caused not only by genetic drift but also by other factors including physical linkage of loci, mutation, population structure, natural selection, finite sample size, within-locus disequilibrium (e.g., excess of homozygosity) and genotyping artifacts such as null alleles. Therefore, to estimate effective population size reflecting the strength of genetic drift alone, it is required estimates of O8 that avoid or that have been corrected for the disequilibrium

8 contributed by the factors mentioned above (O LPQQ.LR.2 ).

8 Ne and Nb can be estimated from O LPQQ.LR.2 ) using the general relationship:

; , STOS, = S Equation 1 . - V QW(LPQQ.LR0P9SX3LRPQ

The software SpEED-Ne (Hamilton et al. 2018) was used to estimate r2 and the correction factor in Equation 1, permuting among loci but not within loci (maintaining allelic pairs). I used three confidence intervals: 1) 95% confidence interval for the correction factor determined by permutation, 2) jackknife over individuals, and 3) jackknife over individuals with an inverse hyperbolic tangent transformation (see Jones et al. 2016). The distinction between the latter two has to do with jackknife performance when there is a large number of non-independent comparisons. Because these are all pairwise comparisons, the number of non-independent comparisons grows rapidly and as the sample sizes gets bigger. Jones et al. (2016) argued that

! 46 ! this post hoc transformation corrects for non-independent comparisons and improves the coverage performance of the estimated confidence interval.

Life history simulations

To examine how life history traits affect Nb, Ne, and their relationship (Nb/Ne) in a long- lived, iteroparous organism, I conducted computer simulations under a specific model implemented in the computer program AgeNe (Waples et al. 2011). This program uses a discrete-time, age-structured, and deterministic model. It assumes a closed population of constant census size, where individuals of age x produce an average of bx offspring (age-specific fecundity) and then survive to age x + 1 with probability sx (age-specific survival). Both bx and sx are independent of events in previous time periods. The fraction of individuals in a cohort surviving to age x is lx. AgeNe tracks only individuals that survive to their first birthday, so fecundities are scaled to result in a stable population that produces a fixed number (N1) of individuals per cohort that survive to age 1. Given a specified life history table and a value for

N1, AgeNe estimates the total population size, the adult population size, and numbers in each age class (Nx), as well as Ne and Nb. Because the required detailed demographic data to use AgeNe is not available for L. tulipifera, the survival and fecundity of another long-lived tree species, the hoop pine, Araucaria cunninghami, were used. This table was originally reported as stage-based data by Enright and Watson (1991), and given in Appendix S2 of Waples et al. (2013) as an age- based life table, which generates a Type III survivorship curve. Trees have Type III survivorship curves (Harcombe 1987), where mortality is a decreasing function of age (Pinder et al. 1978).

Based on the hoop pine life history table, eight life-history scenarios (A-H) were designed differing either in age-specific survival probabilities (A, B), age at reproductive

! 47 ! maturity (C, D), or maximum age (E-H). In scenario A, survival probability was the lowest in age 1 (s1 = 0.05) and increased from age 2 onwards. To examine how increased mortality in early years affects Nb, Ne, and their relationship (Nb/Ne) in a long-lived tree, age 2 survival probability was decreased from 0.26 to 0.05 in scenario B. In this scenario, survival probabilities were the lowest in both ages 1 and 2 (s1 = 0.05 and s2 = 0.05), increasing from age 3 onwards. Waples et al. (2013) showed that two frequently related life history traits, age at reproductive maturity (", first age with fecundity greater than zero) and adult lifespan (AL, maximum age ! " + 1), are highly correlated with Nb/Ne in iteroparous, relatively short-lived species. To examine how " affects Nb, Ne, and Nb/Ne in long-lived trees, this life history trait was delayed from 6 years to 16 and 31 years in scenarios C and D respectively, keeping age-specific survival probabilities unchanged from original. Further, because AL is related to maximum age in addition to ", the effects of maximum age were examined, shortening it from 400 years in scenario E to 300 years in scenario F, 200 years in scenario G, and 100 years in scenario H.

Results

Effective number of breeders (,-) for each of the nine sampling sites (four urban, four old-growth, and one suburban) are presented in Table 3.3. There was a twofold variation in the median ,- from 10.73 to 112.94. However, ,- confidence intervals indicated no differences between these estimates across sampling sites. In addition, there were no differences in median

,- between urban and old-growth sites. Furthermore, at the three sites in which the two ways of sampling were performed (single age cohort of seedlings, and mixed-age samples of adult trees),

,- and ,. /01.2(34. estimates were not different from each other (Table 3.4). The median

,. /01.2(34. varied from 45.77 to 64.61. The overlapping of the confidence intervals of ,- and

! 48 ! ,. /01.2(34. suggested no differences between these estimators (Table 3.4 and Figure 3.1).

Table 3.5 shows correction factor point estimates, and two confidence intervals for these correction factor estimates: 1) jackknife over individuals (Figure 2), and 2) jackknife over individuals with an inverse hyperbolic tangent transformation, which were very consistent.

The results of the simulations of populations using life-history data are presented in Table

3.6. Ne decreased from 15.9 in Scenario A (95% mortality the first year) to 3.1 in Scenario B

(95% mortality in both the first and second years). Similarly, Nb decreased from 16.7 in Scenario

A to 3.3 in Scenario B. The ratio Nb/Ne was close to one in both scenarios A and B. When age at first reproduction was delayed, Ne decreased from 14.6 in scenario C (16 years) to 12.9 in

Scenario D (31 years). A similar pattern was observed in Nb, which decreased from 14.3 in

Scenario C to 11.3 in Scenario D. In both scenarios C and D, Nb/Ne ratio was close to one. When maximum age was shortened, Ne increased steadily, ranging from 12.9 in Scenario E (maximum age of 400 years) to 31.2 in Scenario H (maximum age of 100 years). Likewise, Nb increased from Scenario E to Scenario H. However, the increments were not as pronounced as those observed in Ne, ranging from 11.3 to 14.3. The ratio Nb/Ne was close to one in scenarios E and F, and decreased to 0.76 in scenario G and 0.46 in scenario H.

Discussion

The most commonly reported patterns of urban evolution, decreased genetic polymorphism within and increased genetic differentiation among urban populations, have been attributed to increased genetic drift (Johnson and Munshi-South 2017). However, few direct measurements of the strength of genetic drift in urban and non-urban habitats via estimates of genetic effective population size have been made. In this study, ,- estimates did not differ

! 49 ! between urban forest patches and old-growth forests. Additionally, ,- estimates were uniformly small among populations, even in old-growth populations that could be considered as reservoirs of genetic polymorphism. Therefore, contrary to the prediction that the environmental changes associated with urbanization cause stronger genetic drift, this evolutionary force was strong in both forest types.

The absence of population genetic consequences of anthropogenic disturbance on forest tree species has also been documented in other partially deforested systems such as mixed agricultural-forest landscapes. For example, habitat fragmentation is expected to lead to similar population genetic patterns to those predicted for urban evolution, decreased genetic polymorphism within and increased genetic differentiation among plant populations due to increased genetic drift, elevated mating among relatives, and reduced gene flow (Young et al.,

1996). However, empirical studies have shown that not all habitat loss and fragmentation events result in a decrease of genetic diversity in forest tree populations (reviewed in Young et al. 1996;

Lowe et al. 2005, 2015). The scarcity of significant consequences of habitat fragmentation and disturbance in forest tree populations due to human activities such as logging, has been referred to as the “paradox of forest fragmentation genetics” (Kramer et al. 2008).

One of the mechanisms that have been proposed to explain this paradox is the long-lived nature of trees, because most forest fragmentation is recent and current populations of long-lived trees may simply pre-date it (Kramer et al. 2008; Lowe et al. 2015). Urbanization is a relatively recent phenomenon as well. However, this study avoids the paradox by analyzing seedlings instead of adult trees. In plants, the seedling stage is the most sensitive phase of the life cycle to environmental changes (Harper 1977, Silvertown and Charlesworth 2001, Leck et al. 2008).

Therefore, seedlings are expected to be particularly sensitive to anthropogenic habitat ! 50 ! disturbance such as urbanization. Furthermore, L. tulipifera has a short-lived seed bank, and as a consequence, seedlings are representative of recent ecological and evolutionary processes. By focusing on seedlings, I had the potential to detect recent processes acting on genetic drift. In addition, ,- quantifies genetic drift over short timescales relative to ecological processes.

Therefore, the lack of a difference in ,- is best explained by lack of an urbanization effect, rather than as a consequence of temporal averaging of drift over many generations.

Extensive gene flow is a mechanism that potentially explains the observed absence of genetic consequences of urbanization in seedlings. Noreen et al. (2016) have shown that insect- mediated gene flow persisted among individuals of the tropical tree species Koompassia malaccensis in an urban landscape, even across >2.5 km of impervious surface. Several of

Noreen et al. (2016) results are similar to findings from insect-pollinated trees in mixed agricultural–forest landscapes in which pollen dispersal can extend several kilometers in fragmented landscapes, e.g., in Swietenia humilis (White et al. 2002) and Dinizia excelsa (Dick et al. 2003). Extensive pollen movement could swamp out the localized effects of urbanization, if they indeed exist.

It is noteworthy that L. tulipifera, a common species with large fecundities exhibited estimates of both ,- and ,. /01.2(34. on order of ten or one hundred. Small effective population sizes have also been observed in other species. For example, in marine fish, Ne has been shown to be three to six orders of magnitude less than population census size (N), or even smaller (e.g. Hauser et al. 2002, Turner et al. 2002). Because many of these species have very large census sizes and high fecundities, small Ne estimates have been treated with skepticism and are potentially explained by downward biased caused by aspects of sampling methods or

! 51 ! 8 estimation methodology. For example, Waples (2006) suggested that estimates of Ne based on O for unlinked loci are strongly biased downwards and he proposed a regression fit estimator based on simulated data. Small ,.. estimates can also be the result of not taken into account processes and factors that could increase the total amount of observed disequilibrium (in addition to the disequilibrium caused by genetic drift). For example, admixture can occur among genetically differentiated populations and increase the degree of LD when individuals from these subpopulations with differentiated allele frequencies end up mixing in the sample, which can in turn downwardly bias ,.. Another potential factor to explain small ,. estimates is the presence of null alleles. For example, Hamilton et al. (2018) using simulated data showed that the widely employed Waples (2005) Ne estimator has very large variance when null alleles are present, which could yield biased ,. estimates. Lastly, another factor that could produce small Ne estimates is age-structure. For example, in age-structured populations, young-of-the-year samples produce Nb estimates. If age structure is not taken into consideration, these estimates could be interpreted as generational Ne estimates instead of Nb estimates, and these two parameters can be very different under certain types of life history. For example, Nb can be smaller than Ne, as in the case of several Salmoniformes. Waples (1990a, 1990b) showed that in an overlapping generation model, with semelparous and asynchronous reproduction (salmon model), Ne is approximately equal to g*Nb, where g is the generation length (average age of parents). In contrast, Nb can be greater than Ne in iteroparous organisms. For example, by analyzing 63 iteroparous species of animals and plants, Waples et al. (2013) found that Nb > Ne in over half the species.

! 52 ! The effective population sizes observed in this study are truly small because I accounted for each of the above-mentioned factors that can produce small estimates. For example, given that there is evidence of null alleles in some of the loci used in this study (Gutiérrez-Ozuna and

Hamilton 2017), I did not use the widely employed Waples (2006) Ne estimator implemented in the program LDNe (Waples and Do 2008), which has been shown to be subject to troubles with null alleles (see Hamilton et al. 2018). Instead, the permutation approach implemented in the software SpEED-Ne was used to estimate within-locus disequilibrium caused by the combination of finite sampling and null alleles. Further, although admixture LD caused by occurrence of more than one subpopulation can downwardly bias ,., Waples and England (2011) used computer simulations to show that unless the migration rate is very low (which leads to highly differentiated subpopulations), there is not much of an impact on estimates of ,.. In this particular case, there is low population genetic differentiation among L. tulipifera sampling sites.

For example, Gutiérrez-Ozuna and Hamilton (manuscript in preparation) estimated L. tulipifera population genetic differentiation in the mid-Atlantic U.S., finding that global D′EF was 0.273, which suggests that admixture is not a major factor contributing to the total amount of observed disequilibrium in this study. Finally, the distinction between Nb and Ne was drawn. To estimate

Nb, an extra effort to sample single age cohorts was made. To achieve this, I collected seedlings germinated the same year of collection. In addition, exclusively mixed-age samples of adult trees were used to estimate Ne (,. /01.2(34. ).

It is helpful to understand the conditions under which Nb might be much less than generational Ne in order to know if a given empirical estimate represents ,. or ,-. As mentioned above, in some iteroparous organisms such as fish and other vertebrates, Nb and Ne can be very different. Contrastingly, this study predicts based on simulations that Nb and Ne values are

! 53 ! similar in long-lived trees. The simulations manipulating life history traits generated Nb/Ne ratios close to one across most scenarios examined. Nb and Ne only diverged increasingly as maximum age was reduced to one half or to one quart of its original value, which increased Ne values and yielded Nb/Ne ratios of 0.76 and 0.46, respectively. This was because if age at maturity is kept unchanged, shortening maximum age decreases adult lifespan (maximum age ! " + 1), and this latter life history trait has been also shown to be highly correlated with Nb/Ne (Waples et al.

2013). The lack of disparity between Ne and Nb observed in most scenarios suggests that the strength of genetic drift in long-lived trees is expected to be similar per reproductive event and per generation.

The relationship between Nb and Ne has rarely been examined empirically in very long- lived species with overlapping generations and iteroparity such as trees. Here, Nb, by definition, was estimated analyzing seedlings belonging to a single age cohort. This way of sampling is not practical in many tree species because identifying seedlings based on phenotypic characteristics can be challenging, especially if they are a small size or highly affected by herbivory. Seedling identification is particularly difficult in some environments such as tropical forests due to the high number of species. The most common sampling approach in trees is sampling mixed-age reproductively mature individuals. However, consistent with our modeling results, empirical

,. /01.2(34. estimates did not differ from ,- estimates, suggesting that the two sampling approaches give similar information about the strength of genetic drift in long-lived trees.

Understanding the relationship between Nb and Ne is important, for example, to predict adaptation to urban environments because most of the population genetic theory on adaptation and genetic drift-natural selection balance is based on the concept of generational Ne. For

! 54 ! example, natural selection will dictate allele frequencies if 4Nes >> 1, the selection coefficient (s) is large and Ne is not very small. In contrast, if 4Nes << 1, either s is extremely small or genetic drift is very strong because Ne is very small (as is the case of this study). In either of the two latter scenarios, genetic drift will dictate the allele frequencies (Kimura 1983). In this study, empirical estimates of ,. /01.2(34. were small and did not differ from ,- estimates in the sites where the two ways of sampling were performed, consistent with modeling results that showed that Nb and Ne are expected to be very similar in long-lived trees. In addition, ,- estimates were small across all populations, which might suggest that generational Ne is also small in each site.

Therefore, it might be predicted that only genes with major effects and under very strong natural selection will experience adaptation not only in urban but also in old-growth populations.

Most of our knowledge on population genetic effects of urbanization come from short- lived organisms such as mice, salamanders, and herbaceous plants. This study adds to this literature by extending the age distribution. In addition, this study is novel because it examines urbanization effects in tree populations quantifying directly the strength of genetic drift through measurements of genetic effective population size (Ne), rather than other approaches that have been used. Uniformly small estimates of both ,- and ,. /01.2(34. across all populations documented here suggest that genetic drift is going to be a substantial force in terms of predicting levels of genetic polymorphism through time and also generating population genetic differentiation in L. tulipifera.

! 55 ! APPENDIX A: Tables

Table 1.1. Characteristics of 23 polymorphic genomic microsatellite loci isolated from Liriodendron tulipifera.

Observed Fluorescent Repeat Ta Locus Primer sequences (5’—3’) amplicon labeling Identifiera motif (ºC) size (bp) method

Lt006 F: GTGGTAACGCATGAGATGGC AAG12 416-437 58 M13-labeled KX869968 R: TGAGCTTTCCATCAGGTGAGC 1 Lt011 F: GCATGGACATGGTGTAACCC AAT12 223-250 58 6-FAM KX869967 R: TTCCATGTGTGCCCTACTGC HWI- Lt014 F: CTCACATGAACAACAAGGAAACC AAT16 108-126 58 M13-labeled 1KL163:137:H7DJKADXX:1:2116:14 R: CTGGGATCTGCACTGATTGG 211:16050 1:N:0:CTTGTA HWI- Lt023 F: CGTGGCCAGCATCTTGTAGC AAC16 113-140 58 M13-labeled 1KL163:137:H7DJKADXX:1:2214:45 R: CAAGAACAACGAGGCAAACAGC 79:5389 1:N:0:CTTGTC HWI- Lt025 F: AGTTGGGAATTGGGACAAGG ATC13 105-120 57 M13-labeled 1KL163:137:H7DJKADXX:1:2114:75 R: TTCAGTGCTCCAGTTTTCACG 09:71461 1:N:0:CTTGTA HWI- Lt032 F: GCTCCTACAAACATCAAAAGC AAG13 95-122 50 M13-labeled 1KL163:137:H7DJKADXX:1:2109:25 R: CAAAACCCATTTCGTGTTCC 16:54908 1:N:0:CTTGTA 1 HWI- Lt035 F: CACAGAGCTTGGTGCTTTACG AAT16 89-119 56 NED 1KL163:137:H7DJKADXX:1:1115:14 R: AAGTCCATGTTCCACTCATTCG 496:36717 1:N:0:CTTGTA 3 HWI- Lt036 F: TTGAAGTTTGAATCCCCATCC ATC13 107-143 50 VIC 1KL163:137:H7DJKADXX:1:1109:19 R: TGATTGGGCCATGTTAATCG 109:85000 1:N:0:CTTGTA 3 HWI- Lt043 F: TCATCTTCCTTTGGTTTGCC AAG12 119-149 50 6-FAM 1KL163:137:H7DJKADXX:1:2103:58 R: TGGGGATTTGACAGAGAACG 25:78158 1:N:0:CTTGTA HWI- Lt052 F: TGGTCCCGAGATTGTTCACC AAC14 94-115 57 M13-labeled 1KL163:137:H7DJKADXX:1:2106:81 R: TCTTACCCACCACAACCATCG 32:35499 1:N:0:CTTGTA 2 HWI- Lt054 F: CCTCGTAGTGTTGATCGTTGC AAT16 88-121 57 NED 1KL163:137:H7DJKADXX:1:1214:10 R: TCAACCTTCCACCAGTGTCC 945:60086 1:N:0:CTTGTA 3 HWI- Lt059 F: CTGCCCCTTCAAATTCTTGG AGG12 93-111 55 NED 1KL163:137:H7DJKADXX:1:2102:16 R: AATTGCGTGAAGCTCAGACC 887:46302 1:N:0:CTTGTA HWI- Lt060 F: CCTACTCTTCCGGAGTTTCG AAT16 126-150 57 M13-labeled 1KL163:137:H7DJKADXX:1:1215:81 R: GGGATGGGGTGGAATATAAGC 50:67169 1:N:0:CTTGTA 3 HWI- Lt061 F: TTGGCGGATGATTGAGAGC AAG14 122-146 55 PET 1KL163:137:H7DJKADXX:1:2205:15 R: TTAATGCCGTGGGTTCTGC 636:87476 1:N:0:CTTGTA 2 Lt064 F: AAGGATGACTTTCACTGAGG AAT13 120-156 53 6-FAM HWI- R: CATTGGGACTTTATTTCTCTCC 1KL163:137:H7DJKADXX:1:1203:17

! 56 ! 908:71099 1:N:0:CTTGTA HWI- Lt066 F: TCTGGCCCTTGATACTGTGG ATC12 134-152 57 M13-labeled 1KL163:137:H7DJKADXX:1:1109:17 R: CCCACTTGGGTGTTTCAGG 69:97275 1:N:0:CTTGTA 2 HWI- Lt068 F: AAACTCCCTAACAAGGGTCTCC AAT16 89-131 55 VIC 1KL163:137:H7DJKADXX:1:2211:40 R: ACCACAACACAGAAACAATGGG 01:84575 1:N:0:CTTGTA 1 HWI- Lt070 F: TTCTCCGCCATCGTCTTACC ATC12 112-133 57 6-FAM 1KL163:137:H7DJKADXX:1:2211:16 R: CTAATGAACGGTCGGGATGG 362:43312 1:N:0:CTTGTA HWI- Lt075 F: CATCGCCATTGTTTTCTCTGC AAT13 119-134 55 M13-labeled 1KL163:137:H7DJKADXX:1:1201:45 R: TAACTGCCTGCGTATCATCC 54:96377 1:N:0:CTTGTA 1 HWI- Lt077 F: TATCCAAACGGCCCTTAACC AAT14 84-105 55 HEX 1KL163:137:H7DJKADXX:1:1214:22 R: GATCACAAAACTCCCACATGC 79:95306 1:N:0:CTTGTA 2 HWI- Lt079 F: GGGAGACTCGGCTTTAATCGCC ATC12 76-94 61 PET 1KL163:137:H7DJKADXX:1:2215:19 R: GAGTGGAGTAGCGGGACAGG 866:60074 1:N:0:CTTGTA HWI- Lt080 F: GGCCTGAATTCCTTTGTTCCC AAT12 130-151 57 M13-labeled 1KL163:137:H7DJKADXX:1:1107:42 R: CCCTCAATGTACACGCTTGC 32:34637 1:N:0:CTTGTA HWI- Lt081 F: ATGGATTCCGGCAAGTCTCC AAT17 96-123 57 M13-labeled 1KL163:137:H7DJKADXX:1:2115:20 R: TGAGGAAGAGAACAAACAGGGG 835:65097 1:N:0:CTTGTA

Note: Ta, annealing temperature. In column Fluorescent labeling method 1, 2 or 3 indicate PCR multiplex sets. a Numbers are either GenBank accession numbers or Illumina sequence identifiers associated with NCBI's Short Read Archive Project: PRJNA331147, BioSample: SAMN05417503.

! 57 ! Table 1.2. Locality and voucher information for the Liriodendron tulipifera samples used for the development of polymorphic microsatellite markers.

No. of Location County, State Latitude Longitude Voucher no.a individuals sampled

Frog Canyon, Anne Arundel, 38.884284 -76.552695 MARY1021991 20 Smithsonian Maryland Environmental Research Center

James Madison’s Orange, 38.226667 -78.179444 MARY1021990 20 Landmark Forest Virginia

Saddler’s Woods Camden, New 39.900722 -75.057750 MARY1021989 12 Jersey

a Voucher samples for each location were deposited in the Norton-Brown Herbarium of the University of Maryland, College Park, Maryland, USA (MARY).

! 58 ! Table 1.3. Genetic properties by individual and pooled sampled locations of 23 polymorphic microsatellite markers developed in Liriodendron tulipifera.

Locus Montpeliera Frog Canyona Saddler’s Woodsa Pooled locations

N k HE HO N k HE HO N k HE HO N k HE HO PIC F NullIIM Lt006 10 6 0.789 0.800 10 2 0.100 0.100 10 3 0.689 0.700 30 6 0.637 0.533 0.591 0.166 0.074 Lt011 20 6 0.726 0.600 19 7 0.834 0.789 12 5 0.779 0.833 51 9 0.789 0.725 0.749 0.081 0.039 Lt014 10 6 0.768 0.600 10 5 0.800 0.800 10 6 0.816 0.700 30 7 0.821 0.700 0.782 0.149 0.078 Lt023 10 5 0.505 0.500 10 5 0.616 0.600 10 6 0.779 0.800 30 6 0.767 0.633 0.715 0.177 0.075 Lt025 10 4 0.658 0.900 10 2 0.268 0.300 10 3 0.426 0.300 30 4 0.462 0.500 0.420 -0.085 0.056 Lt032 10 5 0.700 0.700 10 5 0.711 0.900 10 5 0.779 0.700 30 8 0.809 0.767 0.773 0.053 0.036 Lt035 19 7 0.750 0.263*** 12 6 0.815 0.167*** 11 4 0.571 0.273* 42 10 0.826 0.238*** 0.793 0.714 0.402† Lt036 20 8 0.856 0.850 20 8 0.856 0.750 12 6 0.812 0.917 52 9 0.850 0.827 0.822 0.027 0.017 Lt043 20 6 0.805 0.750 20 5 0.703 0.300*** 12 7 0.862 0.667** 52 7 0.807 0.558*** 0.771 0.311 0.098† Lt052 9 4 0.699 0.667 10 6 0.684 0.700 10 5 0.805 0.900 29 7 0.782 0.759 0.744 0.030 0.036 59

Lt054 19 7 0.764 0.579 20 7 0.782 0.700 10 6 0.826 0.100*** 49 11 0.806 0.531*** 0.773 0.344 0.171† Lt059 20 5 0.583 0.650 20 4 0.674 0.500* 12 5 0.717 0.917 52 6 0.694 0.654 0.646 0.059 0.030 Lt060 10 5 0.774 0.600 10 5 0.695 0.500 10 7 0.789 0.700 30 9 0.775 0.600 0.733 0.228 0.101 Lt061 20 4 0.458 0.450 20 3 0.555 0.050*** 12 3 0.301 0.333 52 6 0.492 0.269*** 0.432 0.455 0.130† Lt064 20 5 0.750 0.600 17 7 0.693 0.176*** 12 5 0.725 0.417 49 10 0.825 0.408*** 0.791 0.508 0.238† Lt066 10 4 0.695 0.600 10 4 0.574 0.600 10 2 0.521 0.500 30 4 0.598 0.567 0.519 0.053 0.064 Lt068 20 12 0.837 0.900 20 10 0.873 0.850 12 10 0.783 0.833 52 13 0.876 0.865 0.855 0.013 0.014 Lt070 20 3 0.578 0.450 19 5 0.629 0.632 12 3 0.663 0.667 51 5 0.632 0.569 0.557 0.102 0.059 Lt075 10 2 0.395 0.500 10 2 0.268 0.100 10 2 0.100 0.100 30 4 0.272 0.233 0.254 0.145 0.168 Lt077 20 4 0.742 0.700 19 6 0.725 0.842 12 6 0.641 0.750 51 7 0.765 0.765 0.717 0.000 0.033 Lt079 20 4 0.391 0.400 20 3 0.188 0.200 12 3 0.420 0.417 52 4 0.320 0.327 0.297 -0.021 0.025 Lt080 10 3 0.542 0.200* 10 5 0.774 0.200** 10 4 0.616 0.300* 30 6 0.654 0.233*** 0.587 0.647 0.438† Lt081 10 6 0.811 0.600 10 5 0.774 0.700 10 6 0.763 0.800 30 7 0.809 0.700 0.766 0.137 0.065

Note: N, number of individuals genotyped; k, number of alleles; HE, expected heterozygosity under random mating; HO, observed heterozygosity; *p<0.05, **p<0.01, ***p<0.001, for hypothesis test of Hardy-Weinberg expected genotype frequencies; PIC, polymorphism information content; F, fixation index; NullIIM, estimate of null allele frequency given the individual inbreeding model (IIM); † 95% highest posterior density interval does not include zero.

a See Table 1.2 for locality and voucher information.

! ! Table 2.1. Site information for the Liriodendron tulipifera samples used.

Forest Sampling site County, State Latitude Longitude n type Belt Woods Prince George, Maryland 38.9036915 -76.7598648 48 Frog Canyon Old- Anne Arundel, Maryland 38.8842837 -76.5526953 48 Montpelier growth Orange, Virginia 38.216053 -78.16418 48 Saddler's Woods Camden, New Jersey 39.9007314 -75.0577526 48 British Embassy Washington, D.C. 38.921773 -77.063099 48 Johns Hopkins Baltimore, Maryland 39.3287183 -76.6241754 48 Urban Rock Creek Washington, D.C. 38.952632 -77.040443 48 Georgetown Washington, D.C. 38.908537 -77.077736 48 Southbury Suburban Southbury, Connecticut 41.499565 -73.179114 42

Note: n, number of genotyped individuals.

! 60 ! Table 2.2. Microsatellite marker loci used to genotype Liriodendron tulipifera seedlings.

Primer concentration (each Locus Multiplex set 5’ fluorophore primer, µM)

Lt011 1 6-FAM 20 Lt035 1 NED 20 Lt036 3 VIC 2 Lt043 3 6-FAM 5 Lt054 2 NED 2 Lt059 3 NED 2 Lt061 3 PET 2 Lt064 2 6-FAM 10 Lt068 2 VIC 2 Lt070 1 6-FAM 1 Lt077 1 HEX 1 Lt079 2 PET 2

! 61 ! Table 2.3. Summary measures of genetic polymorphism for each sampling sampling site (with standard errors in parentheses),

including forest type, number of genotyped individuals (n), average number of different alleles per locus (ka); average number of

effective alleles per locus (ke), average number of private alleles per locus (kp); multilocus average observed heterozygosity (HO),

multilocus average expected heterozygosity (HE), and the fixation index (F).

a b c d Site Forest type n ka ke kp HO HE F Belt Woods 48 6.75 (0.79) 4.21 (0.58) 0.08 (0.08) 0.60 (0.06) 0.72 (0.03) 0.17 (0.08) Frog Canyon 48 6.83 (0.66) 4.47 (0.58) 0.08 (0.08) 0.58 (0.06) 0.71 (0.05) 0.15 (0.07) Old-growth Montpelier 48 6.17 (0.73) 3.39 (0.38) 0.08 (0.08) 0.59 (0.05) 0.66 (0.04) 0.10 (0.04) Saddler's Woods 48 6.75 (0.60) 3.65 (0.35) 0.25 (0.13) 0.58 (0.06) 0.68 (0.05) 0.13 (0.07)

6 British Embassy 48 6.17 (0.56) 4.07 (0.52) 0.00 (0.00) 0.56 (0.07) 0.68 (0.05) 0.20 (0.07) 2

Johns Hopkins 48 6.33 (0.64) 3.91 (0.53) 0.17 (0.11) 0.59 (0.07) 0.67 (0.06) 0.15 (0.06) Urban Rock Creek 48 6.33 (0.67) 3.68 (0.38) 0.17 (0.11) 0.59 (0.06) 0.70 (0.03) 0.16 (0.07) Georgetown 48 6.08 (0.63) 3.59 (0.38) 0.00 (0.00) 0.60 (0.03) 0.69 (0.02) 0.13 (0.04) Southbury Suburban 42 4.83 (0.47) 2.42 (0.26) 0.00 (0.00) 0.45 (0.06) 0.53 (0.05) 0.13 (0.06)

Note: a ( & ke = 1/( %)* $% ) b HO = observed frequency of heterozygotes / n c ( & HE = 1 − %)* $% d F = (HE - HO) / HE where pi is the frequency of the ith allele.

! ! Table 2.4. Pairwise standardized genetic differentiation (F’ST, below diagonal) and Euclidean geographic distances in kilometers (above diagonal) for all sampling site pairs.

Belt Frog Saddler's British Johns Rock Site Montpelier Georgetown Southbury Woods Canyon Woods Embassy Hopkins Creek

Belt Woods — 18.1 144.1 183.5 26.3 48.7 24.9 27.5 419.2

Frog Canyon 0.150 — 158.6 171.1 44.4 49.8 42.9 45.5 408.2

Montpelier 0.213 0.221 — 327.1 123.8 182.0 127.5 121.9 560.5

Saddler's 0.329 0.235 0.264 — 203.8 148.5 200.3 205.6 238.1

6 Woods 3

British 0.250 0.192 0.254 0.241 — 59.0 4.0 1.9 436.9 Embassy Johns 0.274 0.259 0.320 0.341 0.087 — 55.1 60.9 378.5 Hopkins

Rock Creek 0.235 0.205 0.266 0.297 0.106 0.152 — 5.9 433.1

Georgetown 0.133 0.238 0.251 0.345 0.193 0.230 0.118 — 438.8

Southbury 0.387 0.419 0.376 0.419 0.436 0.423 0.382 0.424 —

! ! Table 2.5. Analysis of molecular variance (AMOVA) for 384 seedlings from 8 sampling sites in 2 forest types.

Fixation Source df SS MS Est. Var. % p-value indices

Between forest 1 63.954 63.954 0.090 2% FRT = 0.20 0.001 types

Among sampling 6 175.526 29.254 0.251 5% FST = 0.74 0.001 sites

Among individuals 376 1951.469 5.190 0.902 19% FIS = 0.210 0.001

Within individuals 384 1300.000 3.385 3.385 73% FIT = 0.269 0.001

Total 767 3490.949 4.629 100%

Note: df, degrees of freedom; SS, sum of squares; MS, mean squares; Est. Var., estimated of variance; %, percentage of total variation; p-value based on 999 permutations. FRT refers to between forest type to total, FST refers to sampling sites to total, FIS refers to among individuals within sampling site, FIT refers to within individuals to total. FSR (sampling sites within forest type) was 0.055 (p-value =0.001).

! 64 ! Table 2.6. Estimates of L. tulipifera mating system parameters for each sampling site obtained using the maximum-expectation method implemented in MLTR and a Bayesian method implemented in BORICE. Estimates from MLTR include multilocus

outcrossing rate (tm), singlelocus outcrossing rate (ts), multilocus correlation of paternity (rp(m)), and number of pollen donors (1/rp(m); with standard deviations in parentheses). Estimates from BORICE include outcrossing rate (t-max) and the fixation index (F; with credible intervals, 2.5 and 97.5 percentiles, in parentheses).

Population MLTR BORICE Sampling site type tm ts tm-ts rp(m) 1/ rp(m) t-max F Belt Woods 0.800 (0.113) 0.804 (0.092) -0.004 (0.035) 0.000 (0.000) (—) 0.86 (0.73-0.95) 0.07 (0.02-0.14)

Frog Canyon 0.993 (0.022) 0.986 (0.003) 0.007 (0.020) 0.008 (0.002) 125.00 0.85 (0.71-0.95) 0.08 (0.02-0.15) Old-growth Montpelier 0.910 (0.009) 0.905 (0.003) 0.005 (0.007) 0.046 (0.007) 21.74 0.85 (0.71-0.95) 0.08 (0.02-0.15) 6 5 Saddler's Woods 0.999 (0.000) 0.997 (0.001) 0.002 (0.000) 0.029 (0.005) 34.48 0.81 (0.63-0.95) 0.09 (0.02-0.19)

Georgetown 0.967 (0.006) 0.959 (0.002) 0.008 (0.005) 0.018 (0.001) 55.56 0.81 (0.67-0.92) 0.10 (0.04-0.17)

British Embassy — (—) — (—) — (—) — (—) (—) 0.80 (0.61-0.89) 0.13 (0.06-0.22) Urban Johns Hopkins 0.942 (0.008) 0.937 (0.001) 0.005 (0.007) 0.025 (0.002) 40.00 0.91 (0.75-0.97) 0.06 (0.01-0.12)

Rock Creek — (—) — (—) — (—) — (—) (—) 0.84 (0.69-0.94) 0.08 (0.02-0.17)

Southbury Suburban 1.000 (0.000) 1.000 (0.003) 0.000 (0.000) 0.023 (0.007) 43.48 0.69 (0.51-0.83) 0.18 (0.08-0.29)

Note: Dashes indicate that parameters were not estimated due to software limitations.

! ! Table 3.1. Sampling locations and total number of samples collected.

Sampling site Population type Sample size

Seedlings

1) Montpelier, VA 48

2) Belt Woods, MD 48 Old-growth 3) Frog Canyon (SERC), MD 48

4) Saddler’s Woods, NJ 48

5) Georgetown, DC 48

6) British Embassy, DC 48 Urban 7) Rock Creek Park, DC 48

8) Johns Hopkins, Baltimore, MD 48

9) Southbury, CT Suburban 42

Adult trees

1) Montpelier, VA 20

2) Frog Canyon (SERC), MD Old-growth 20

3) Saddler’s Woods, NJ 12

! 66 ! Table 3.2. Microsatellite marker loci used to genotype Liriodendron tulipifera seedlings and adult individuals.

Primer concentration (each Locus Multiplex set 5’ fluorophore primer, µM)

Lt011 1 6-FAM 20 Lt035 1 NED 20 Lt036 3 VIC 2 Lt043 3 6-FAM 5 Lt054 2 NED 2 Lt059 3 NED 2 Lt061 3 PET 2 Lt064 2 6-FAM 10 Lt068 2 VIC 2 Lt070 1 6-FAM 1 Lt077 1 HEX 1 Lt079 2 PET 2

! 67 ! Table 3.3. Effective of breeders (!") estimated from linkage disequilibrium in single age cohorts of Liriodendron tulipifera. Results of delete-one jackknifing over individuals. (!" = % $ ). Correction factor is median r^2 permute for r^2_{c}. Estimates are & '()*+'',*-.+/$01*-+' allele frequency weighted (AFW).

Sampling !" confidence Sampling site Population type Median !" approach interval Belt Woods Seedlings 48.63 39.98 - 60.86 Montpelier Seedlings 47.8 39.18 - 62.01 Old-growth Frog Canyon Seedlings 112.94 72.20 - 168.52 Saddler's Woods Seedlings 27.3 23.51 - 30.59 Johns Hopkins Seedlings 39.51 32.74 - 54.95 Rock Creek Seedlings 91.9 65.57 - 140.99 Urban British Embassy Seedlings 25.94 22.59 - 28.54 Georgetown 2015 Seedlings 16.53 14.88 - 20.58 Southbury Suburban Seedlings 10.73 9.81 - 11.68

! 68 ! Table 3.4. Effective population size estimates from linkage disequilibrium in single age (!") and mixed-age cohorts samples (!, 2.3,4)15, ) of Liriodendron tulipifera. Results of delete-one % jackknifing over individuals. (! = $ ). Correction factor is median r^2 , & '()*+'',*-.+/$01*-+' permute for r^2_{c}. Estimates are allele frequency weighted (AFW).

Effective population Effective population size Sampling site Sampling approach size estimate confidence interval

Seedlings 47.8* 39.18 - 62.01 Montpelier Adults 64.61 29.46 - 88.64 Seedlings 112.94* 72.20 - 168.52 Frog Canyon Adults 61.74 24.21 - 126.30 Seedlings 27.3* 23.51 - 30.59 Saddler's Woods Adults 45.77 14.90 - 40.59

* Estimates correspond to effective number of breeders (!").

! 69 ! Table 3.5. Correction factor point estimates (r^2 correction factor is median r^2 permute for r^2_{c}), and two confidence intervals for correction factor point estimates used in this study: 1) jackknife over individuals, and 3) jackknife over individuals with Jones et al. (2016) inverse hyperbolic tangent transformation. Estimates are allele frequency weighted (AFW).

Delete-one jackknife over Delete-one jackknifing over Sampling r^2_{comp} AFW individuals with Jones et al. (2016) Sampling site individuals percentile confidence approach point estimate transformation normal distribution interval confidence interval Belt Woods Seedlings 0.037449 0.036072 - 0.038932 0.036274 - 0.038624 British Embassy Seedlings 0.04531 0.044140 - 0.047217 0.043996 - 0.046624 Georgetown 2015 Seedlings 0.049662 0.045691 - 0.051899 0.047351 - 0.051973 7 0 Johns Hopkins Seedlings 0.037974 0.035602 - 0.039717 0.036504 - 0.039444 Southbury Seedlings 0.064968 0.062455 - 0.067904 0.061736 - 0.068199 Rock Creek Seedlings 0.034749 0.033486 - 0.036205 0.033458 - 0.036040 Seedlings 0.034519 0.032922 - 0.036054 0.033037 - 0.036001 Montpelier Adults 0.068742 0.067344 - 0.074898 0.064586 - 0.072895 Seedlings 0.034466 0.033493 - 0.036132 0.033390 - 0.035543 Frog Canyon Adults 0.082675 0.079915 - 0.091042 0.077787 - 0.087559 Seedlings 0.040997 0.039687 - 0.042969 0.039511 - 0.042483 Saddler's Woods Adults 0.110112 0.111041 - 0.125195 0.101268 - 0.118938

! ! Table 3.6. Effective population size, Ne; effective number of breeders, Nb; Nb/Ne ratio; mean number of offspring per parent per time period, !; variance in reproductive success among adults in one time period, Vk; generation length (mean age of parents of a newborn cohort), G; estimated from eight life-history scenarios.

Parameters Scenario A Scenario B Scenario C Scenario D Scenario E Scenario F Scenario G Scenario H

Ne 15.9 3.1 14.6 12.9 16.2 17.0 19.2 31.2

Nb 16.7 3.3 14.3 11.3 17.1 16.5 14.5 14.3 7

1 Nb/Ne 1.05 1.06 0.98 0.88 1.06 0.97 0.76 0.46

G 169.124 169.24 181.332 197.407 169.124 157.648 128.903 76.178

!"##$"% 1.604 1.838 1.598 1.598 1.598 1.599 1.603 1.615

!&'()*'# 2.000 2.00 2.00 2.00 2.00 2.000 2.000 2.000

+!"##$"% 190.552 1124.394 223.054 281.623 186.071 192.928 219.366 223.968

+!&'()*'# 42552.676 218416.00 49544.113 61188.523 41718.258 36990.645 26884.301 9760.308

! ! APPENDIX B: Figures

0.9

0.8

0.7

0.6 ) ST 0.5 F'

− 0.4 / (1

ST 0.3 F' 0.2

0.1

0 0 1 2 3 4 5 6 7 ln (geographic distance)

Figure 2.1. Relationship between pairwise F’ST / (1 − F’ST) and natural logarithm of geographic distances (in km) for nine L. tulipifera sampling sites (Mantel r = 0.759, p = 0.003). The dashed line represents a linear regression line fit to the data.

! 72 ! 180

160

140

120

100

80

60

40

20 Effective population sizepopulationEffective estimates

0 Montpelier Frog Canyon Saddler's Woods

Nb Ne

Figure 3.1. Comparison of effective population size estimates from linkage disequilibrium in single age (!") and mixed-age cohorts samples (!# $%&#'()*# ) of Liriodendron tulipifera. / Results of delete-one jackknifing over individuals. (! +,-! = - ). r^2 # " 0 12(3411#35%46-7)3541 correction factor is median r^2 permute for r^2_{c}. Estimates are allele frequency weighted (AFW). Error bars represent confidence intervals from jackknifing over individuals.

! 73 ! 0.08

0.07

0.06

0.05

0.04

0.03 r^2_{comp} AFW AFW r^2_{comp} 0.02

0.01

0 Southbury Montpelier Belt Woods Belt Rock Creek Rock Frog Frog Canyon Johns Johns Hopkins British British Embassy Saddler's Woods Saddler's ! 2015 Georgetown Figure 3.2. Median r^2 permute for r^2_{c}. Estimates are allele frequency weighted (AFW). Error bars represent confidence intervals from jackknifing over individuals.

! 74 ! REFERENCES

Adams, M.B., Loughry, L., and Plaugher, L. (2003). Experimental forests and ranges of the USDA Forest Service. U.S. For. Serv. General Technical Report NE-321.

Aguilar, R., Quesada, M., Ashworth, L., Herrerias-Diego, Y., and Lobo, J. (2008). Genetic consequences of habitat fragmentation in plant populations: susceptible signals in plant traits and methodological approaches. Molecular Ecology, 17, 5177–5188.

Balloux, F. (2004). Heterozygote excess in small populations and the heterozygote-excess effective population size. Evolution, 58, 1891–1900.

Barbará, T., Palma-Silva, C., Paggi, G.M., Bered, F., Fay, M.F., and Lexer, C. 2007. Cross- species transfer of nuclear microsatellite markers: Potential and limitations. Molecular Ecology, 16, 3759–3767.

Bartlewicz, J., Vandepitte, K., Jacquemyn, H., and Honnay, O. (2015). Population genetic diversity of the clonal self-incompatible herbaceous plant Linaria vulgaris along an urbanization gradient. Biological Journal of the Linnean Society, 116: 603–613.

Beck, D.E. (1990). Liriodendron tulipifera L. Yellow-Poplar. In: Burns, R.M.; Honkala, B.H., technical coordinators. Silvics of North America. Volume 2. Hardwoods. Agric. Handb. 654. Washington, DC: U.S. Department of Agriculture, Forest Service, 406–416.

Beck, D.E., and Della-Bianca, L. (1981). Yellow-poplar: Characteristics and Management. USDA Agricultural Handbook 583.

Boakes, E.H., Mace, G.M., McGowan, P.J.K., and Fuller, R.A. (2010). Extreme contagion in global habitat clearance. Proceedings of the Royal Society B: Biological Sciences, 277 (1684), 1081-1085.

Botstein, D., White, R.L., Skolnick, M., and Davis, R.W. (1980). Construction of a genetic linkage map in man using restriction fragment length polymorphisms. The American Journal of Human Genetics, 32, 314–331.

! 75 ! Brotschol, J.V., Roberds, J.H. and Namkoong, G. (1986). Allozyme variation among North Carolina populations of Liriodendron tulipifera L. Silvae Genetica, 35, 131–138.

Brown, L.R., Flavin, C., and French, H. (1998). State of the World 1998: A Worldwatch Institute Report on Progress toward a Sustainable Society, edited by Linda Starke. W.W. Norton.

Busing, R. (1995). Disturbance and the population dynamics of Liriodendron tulipifera: simulations with a spatial model of forest succession. Journal of Ecology, 83(1), 45-53.

Carper, A.L., Adler, L.S., Warren, P.S., and Irwin, R.E. (2014). Effects of suburbanization on forest bee communities. Environmental Entomology, 43, 253–262.

Castoe, T.A., Poole, A.W., de Koning, A.P., Jones, K.L., Tomback, D.F., Oyler-McCance, S.J., Fike, J.A., Lance, S.L., Streicher, J.W., Smith, E.N., and Pollock, D.D. (2012). Rapid microsatellite identification from Illumina paired-end genomic sequencing in two birds and a snake. PLoS ONE, 7, e30953.

Chambers, J., Higuchi, N., and Schimel, J.P. (1998). Ancient trees in Amazonia. Nature, 391, 135–136.

Charlesworth, B. (1998). Measures of divergence between populations and the effect of forces that reduce variability. Molecular Biology and Evolution, 15, 538–543.

Charlesworth, B. (2009). Fundamental concepts in genetics: effective population size and patterns of molecular evolution and variation. Nature Reviews Genetics, 10, 195–205.

Cheptou, P.O., Carrue, O., Rouifed, S., and Cantarel, A. (2008). Rapid evolution of seed dispersal in an urban environment in the weed Crepis sancta. Proceedings of the National Academy of Sciences USA, 105, 3796–3799.

Chhatre, V.E., and Rajora, O.P. (2014). Genetic divergence and signatures of natural selection in marginal populations of a keystone, long-lived conifer, eastern white pine (Pinus strobus) from northern Ontario. PLoS One, 9(5), e97291.

! 76 ! Chybicki, I.J., and Burczyk, J. (2009). Simultaneous estimation of null alleles and inbreeding coefficients. Journal of Heredity, 100, 106–113.

Clark, B.F., and Boyce, S.G. (1964). Yellow–poplar seed remain viable in the forest litter. Journal of Forestry, 62, 564–567.

Cogbill, C.V. (1996). Black growth and fiddlebutts: The nature of old-growth red spruce. Pp. 113-125, In Davis, M.B. (Ed.). Eastern Old-growth Forests: Prospects for Rediscovery and Recovery. Island Press, Washington, DC.

Condit, R., Hubbell, S.P., and Foster, R.B. (1995). Mortality Rates of 205 Neotropical Tree and Shrub Species and the Impact of a Severe Drought. Ecological Monographs, 65, 419-439.

Davis, M.B. (1996). Old growth in the East: a survey. Island Press, Washington D.C., USA.

De Meester L, Gomez A, Okamura B, and Schwenk K. (2002). The monopolization hypothesis and the dispersal-gene flow paradox in aquatic organisms. Acta Oecologica, 23, 121–135.

Dick, C.W., Etchelecu, G., and Austerlitz, F. (2003). Pollen dispersal of tropical trees (Dinizia excelsa: Fabaceae) by native insects and African honeybees in pristine and fragmented Amazonian rainforest. Molecular Ecology, 12, 753-764.

Donihue, C.M., and Lambert, R.M. (2015). Adaptive evolution in urban ecosystems. AMBIO, 44, 194–203.

Eckert, C.G., Kalisz, S., Geber, M.A., Sargent, R., Elle, E., Cheptou, P.O., Goodwillie, C., Johnston, M.O., Kelly, J.K., Moeller, D.A., Porcher, E., Ree, R.H., Vallejo-Marín, M., and Winn, A.A. (2010). Plant mating systems in a changing world. Trends in Ecology and Evolution, 25, 35–43.

Ellis, J.R., and Burke, J.M. (2007). EST-SSRs as a resource for population genetic analyses. Heredity, 99(2), 125-132.

Ellegren, H., Primmer, C.R., and Sheldon, B. (1995). Microsatellite ‘evolution’: directionality or bias? Nature Genetics, 11, 360–362.

! 77 ! Ellstrand, N.C. (1992). Gene Flow by Pollen: Implications for Plant Conservation Genetics. Oikos, 63, 77-86.

Ennos, R. A. (1994). Estimating the relative rates of pollen and seed migration among plant populations. Heredity, 72, 250–259.

Enright, N.J., and Watson, A.D. (1991). A matrix population model analysis for the tropical tree, Araucaria cunninghamii. Australian Journal of Ecology, 16, 507-520.

Felsenstein, J. (1971). Inbreeding and variance effective numbers in populations with overlapping generations. Genetics, 68, 581–597.

Feng, X., Liu, J., Chiang, Y.C., and Gong, X. (2017). Investigating the Genetic Diversity, Population Differentiation and Population Dynamics of Cycas segmentifida (Cycadaceae) Endemic to Southwest China by Multiple Molecular Markers. Frontiers in Plant Science, 8, 839.

Feng, X.Y., Zheng, Y., and Gong, X. (2016). Middle-upper Pleistocene climate changes shaped the divergence and demography of Cycas guizhouensis (Cycadaceae): Evidence from DNA sequences and microsatellite markers. Scientific Reports, 6, 27368.

Ferreira, P.A., Boscolo, D., and Viana, B.F. (2013). What do we know about the effects of landscape changes on plant–pollinator interaction networks? Ecological Indicators 31, 35–40.

Fetter, K.C. (2011). Migration, adaptation, and speciation: A post-glacial history of the population structure, phylogeography, and biodiversity of Liriodendron tulipifera L.(Magnoliaceae). Unpublished master’s thesis, University of North Carolina at Chapel Hill.

Fisher R.A. (1930). The genetical theory of natural selection. Oxford University Press.

Frankham, R.D. (1995). Effective population size/adult population size ratios in wildlife: a review. Genetics Research, 66, 95–107.

! 78 ! Fraser, C. I., Spencer, H.G., and Waters, J.M. (2009). Glacial oceanographic contrasts explain phylogeography of Australian bull kelp, Molecular Ecology, 18(10), 2287–2296.

Griffith, R.S. (1991). Liriodendron tulipifera. In: Fire Effects Information System, [Online]. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory (Producer).

Grimm, N.B., Faeth, S.H., Golubiewski, N.E., Redman, C.L., Wu, J., Bai, X., and Briggs, J.M. (2008). Global Change and the Ecology of Cities. Science, 319, 756–760.

Gutiérrez-Ozuna, R., and Hamilton, M.B. (2017). Identification and characterization of microsatellite loci in the tuliptree, Liriodendron tulipifera (Magnoliaceae). Applications in Plant Sciences, 5(8), 1700032.

Hamilton, M.B. (2009). Population Genetics. Wiley-Blackwell.

Hamilton, M.B., Tartakovsky, M., and Battocletti, A. (2018). SPEED-NE: Software to simulate

and estimate genetic effective population size (Ne) from linkage disequilibrium observed in single samples. Molecular Ecology Resources, 00, 1–15.

Hamrick, J.L., and Godt, M.J.W. (1996). Effects of life history traits on genetic diversity in plants. Philosophical Transactions of the Royal Society B: Biological Sciences, 351, 1291–1298.

Hamrick, J.L., Godt, M.J.W., and Sherman-Broyles, S.L. (1992). Factors influencing levels of genetic diversity in woody plant species. New Forests, 6, 95–124.

Harcombe, P.A. (1987). Tree Life Tables: Simple birth, growth, and death data encapsulate life histories and ecological roles. Bioscience, 37 (8), 557–568.

Hare, M.P., Nunney, L., Schwartz, M.K., Ruzzante, D.E., Burford, M., Waples, R.S., Ruegg, K., and Palstra, F. (2011). Understanding and estimating effective population size for practical application in marine species management. Conservation Biology, 25, 438–449.

Harper, J.L. (1977). Population biology of plants. Academic Press.

! 79 ! Harris, L. N., and Taylor, E. B. (2010). Pleistocene glaciations and contemporary genetic diversity in a Beringian fish, the broad whitefish, Coregonus nasus (Pallas): inferences from microsatellite DNA variation. Journal of Evolutionary Biology, 23 (1), 72–86.

Hauser, L., Adcock, G.J., Smith, P.J., Bernal Ramirez, J.H., and Carvalho G.R. (2002). Loss of microsatellite diversity and low effective population size in an overexploited population of New Zealand snapper (Pagrus auratus). Proceedings of the National Academy of Sciences USA, 99, 11742-11747.

Hedrick, P.W. (1999). Perspective: highly variable loci and their interpretation in evolution and conservation. Evolution, 53, 313–318.

Hedrick, P.W., (2005) A standardized genetic differentiation measure. Evolution, 59, 1633–1638.

Hepper, F.N. (2003). Phenological records of English garden plants in Leeds (Yorkshire) and Richmond (Surrey) from 1946 to 2002. An analysis relating to global warming. Biodiversity and Conservation, 12, 2503–2520.

Hill, W.G. (1977). Correlation of gene frequencies between neutral linked genes in finite populations. Theoretical Population Biology, 11, 239-248.

Hill, W.G. (1981). Estimation of effective population size from data on linkage disequilibrium. Genetics Research, 38, 209–216.

Houston, D.B., and Joehlin, K.A. (1989). Are pollination bags needed for controlled pollination programs with Yellow-poplar? Silvae Genetica, 38, 137–140.

Inanaga, M., Koyama, Y., Ida, H., Okada, M., Nakanishi, A., Takahashi, M., and Tomaru, N. (2016). Pollen dispersal patterns and population persistence in a small isolated population of Fagus crenata. Tree Genetics and Genomes, 12, 69.

Ishida, K. (2006). Maintenance of inbreeding depression in a highly self-fertilizing tree, Magnolia obovata Thunb. Evolutionary Ecology, 20, 173–191.

! 80 ! Jakobsson, M., Edge, M.D., and Rosenberg, N.A. (2013). The relationship between FST and the frequency of the most frequent allele. Genetics, 193(2), 515–528.

Jeong, J.H., Ho, C.H., Linderholm, H.W., Jeong, S.J., Chen, D., and Choi, Y.S. (2011). Impact of urban warming on earlier spring flowering in Korea. International Journal of Climatology, 31, 1488–1497.

Jha, S. (2015). Contemporary human-altered landscapes and oceanic barriers reduce bumble bee gene flow. Molecular Ecology, 24, 993–1006.

Johnson, M.G., Lang, K., Manos, P., Golet, G.H., and Schierenbeck, K.A. (2016). Evidence for genetic erosion of a California native tree Platanus racemosa via recent ongoing introgressive hybridization with an introduced ornamental species. Conservation Genetics 17, 593–602.

Johnson, M.T.J., and Munshi-South, J. (2017). Evolution of life in urban environments. Science, 358, eaam8327.

Johnson, M.T.J., Thompson, K., and Saini, H. (2015). Plant evolution in the urban jungle. American Journal of Botany, 102, 1951–1953.

Jones, A.T., Ovenden, J.R., and Wang Y.G. (2016). Improved confidence intervals for the linkage disequilibrium method for estimating effective population size. Heredity, 117, 217–223.

Jorde, P.E., and Ryman, N. (1995). Temporal allele frequency change and estimation of effective size in populations with overlapping generations. Genetics, 139, 1077–1090.

Kalinowski, S.T., Taper, M.L., and Marshall, T.C. (2007). Revising how the computer program CERVUS accommodates genotyping error increases success in paternity assignment. Molecular Ecology, 16, 1099–1106.

Kanashiro, M. (1990). The demographic genetics of an Appalachian stand of Liriodendron tulipifera L. Unpublished PhD thesis, North Carolina State University.

! 81 ! Kawecki, T.J., Ebert, D. (2004). Conceptual issues in local adaptation. Ecology Letters, 7, 1225– 1241.

Kearse, M., Moir, R., Wilson, A., Stones-Havas, S., Cheung, M., Sturrock, S., Buxton, S., Cooper, A., Markowitz, S., Duran, C., Thierer, T., Ashton, B., Mentjies, P., and Drummond, A. (2012). Geneious Basic: an integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinformatics, 28 (12), 1647–1649.

Kimura, M., (1953). “Stepping stone” model of population. Annual Report of the National Institute of Genetics, 3, 63–65.

Kimura, M. (1983). The Neutral Theory of Molecular Evolution. Cambridge University Press, Cambridge.

Kovach, K.E. (2009). Assessment of genetic variation of Acer rubrum L. and Liriodendron tulipifera L. populations in unmanaged forests of the southeast United States. Unpublished master’s thesis, Virginia Polytechnic Institute and State University.

Kramer, A.T., Ison, A.J., Ashley, M.V., and Howe, H.F. (2008). The paradox of forest fragmentation genetics. Conservation Biology, 22, 878–885.

Lande, R., and Schemske, D.W. (1985). The evolution of self-fertilization and inbreeding depression in plants. I. Genetic models. Evolution, 39, 24–40.

Lanier, H.C., Massatti, R., He, Q., Olson, L.E., and Knowles, L.L. (2015). Colonization from divergent ancestors: glaciation signatures on contemporary patterns of genomic variation in Collared Pikas (Ochotona collaris). Molecular Ecology, 24 (14), 3688–3705.

Leck, M.A., Parker, V.T., and Simpson, R.L. (Eds.). (2008). Seedling Ecology & Evolution. Cambridge University Press, Cambridge, UK.

Lee, C.R., and Mitchell-Olds, T. (2011). Quantifying effects of environmental and geographical factors on patterns of genetic differentiation. Molecular Ecology, 20, 4631–4642.

! 82 ! Leverett, R. (1996). Definitions and history. Pp.3-17. In M.B. Davis (Ed.), Eastern Old-Growth Forests. Island Press, Washington, D.C.

Lloyd, D.G. (1979). Some reproductive factors affecting the selection of self-fertilization in plants. American Naturalist, 113, 67–79.

Lloyd, M.W., Burnett, R.K., Jr., and Engelhardt, K.A.M., and Neel, M.C. (2012). Does genetic diversity of restored sites differ from natural sites? A comparison of Vallisneria americana (Hydrocharitaceae) populations within the Chesapeake Bay. Conservation Genetics, 13, 753-765.

Loveless, M.D., and Hamrick, J.L. (1984). Ecological determinants of genetic structure in plant populations. Annual Review of Ecology, Evolution, and Systematics, 15, 65–95.

Lowe, A.J., Boshier, D., Ward, M., Bacles, C.F.E., and Navarro, C. (2005). Genetic resource impacts of habitat loss and degradation; reconciling empirical evidence and predicted theory for neotropical trees. Heredity, 95, 255–273.

Lowe, A.J., Cavers, S., Boshier, D., Breed, M.F., and Hollingsworth, P. (2015). The resilience of forest fragmentation genetics—no longer a paradox—we were just looking in the wrong place. Heredity, 115, 97–99.

Lusk, J.J., Swihart, R.K., and Goheen, J.R. (2007). Correlates of interspecific synchrony and interannual variation in seed production by deciduous trees. Forest ecology and management, 242, 656– 670.

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

Maruyama, T. (1970). Effective number of alleles in a subdivided population. Theoretical Population Biology, 1, 273–306.

Maryland Geological Survey. (1929). Baltimore County. Baltimore, MD. John Hopkins University Press.

! 83 ! McGarvey, J.C., Thompson, J.R., Epstein, H.E., and Shugart, H.H. (2015). Carbon storage in old-growth forests of the Mid-Atlantic: toward better understanding the eastern forest carbon sink. Ecology, 96 (2), 311–317.

McKinney, M.L. (2006). Urbanization as a major cause of biotic homogenization. Biological Conservation, 127, 247e260.

McRae, B.H. (2006). Isolation by resistance. Evolution, 60, 1551–1561.

McRae, B.H., and Beier, P. (2007). Circuit theory predicts gene flow in plant and animal populations. Proceedings of the National Academy of Sciences USA, 104, 19885–19890.

Meerow, A.W., Francisco-Ortega, J., Calonje, M., Griffiths, M. P., Ayala-Silva, T., Stevenson, D.W., and Nakamura, K. (2012). Zamia (Cycadales: Zamiaceae) on Puerto Rico: Asymmetric genetic differentiation and the hypothesis of multiple introductions. American Journal of Botany, 99 (11), 1828–1839.

Meirmans, P.G. (2012). The trouble with isolation by distance. Molecular Ecology, 21, 2839– 2846.

Meirmans, P.G., and Van Tienderen, P.H. (2004). GENOTYPE and GENODIVE: two programs for the analysis of genetic diversity of asexual organisms. Molecular Ecology Notes, 4, 792–794.

Merritt, B.J., Culley, T.M., Avanesyan, A., Stokes, R., and Brzyski, J. (2015). An empirical review: Characteristics of plant microsatellite markers that confer higher levels of genetic variation. Applications in Plant Sciences, 3, 1500025.

Mimet, A., Pellissier, V., Quenol, H., Aguejdad, R., Dubreuil, V., and Roze, F. (2009). Urbanisation induces early flowering: evidence from Platanus acerifolia and Prunus cerasus. International Journal of Biometeorology, 53, 287–298.

Morais Júnior, O.P., Breseghello, F., Duarte, J.B., Morais, O.P., and Rangel, P.H.N. (2017). Effectiveness of recurrent selection in irrigated rice breeding. Crop Science, 57 (6), 3043–3058.

! 84 ! Mosseler. A., Major, J.E., and Rajora, O.P. (2003). Old-growth red spruce forests as reservoirs of genetic diversity and reproductive fitness. Theoretical and Applied Genetics, 106, 931– 937.

Munshi-South J, Zolnik CP, and Harris SE. (2016). Population genomics of the Anthropocene: urbanization is negatively associated with genome-wide variation in white-footed mouse populations. Evolutionary Applications, 9, 546-64.

Nei, M. (1973). Analysis of the gene diversity in subdivided populations. Proceedings of the National Academy of Sciences USA, 70, 3321–3323.

Neil, K., and Wu, J. (2006). Effects of urbanization on plant flowering phenology: a review. Urban Ecosystems, 9, 243–257.

Nomura, T. (2008). Estimation of effective number of breeders from molecular coancestry of single cohort sample. Evolutionary Applications, 1, 462–474.

Noreen, A.M., Niissalo, M.A., Lum, S.K., and Webb, E.L. (2016). Persistence of long-distance, insect-mediated pollen movement for a tropical canopy tree species in remnant forest patches in an urban landscape. Heredity, 117 (6), 472–480.

Nosil, P., Funk, D.J., and Ortiz-Barrientos, D. (2009). Divergent selection and heterogeneous genomic divergence. Molecular Ecology, 18, 375–402.

Olson, D.F., Jr. (1969). Silvical characteristics of yellow-poplar (Liriodendron tulipifera L.). USDA Forest Service, Research Paper SE-48. Southeastern Forest Experiment Station, Asheville, NC.

Orsini, L., Vanoverbeke, J., Swillen, I., Mergeay, J., and De Meester, L. (2013). Drivers of population genetic differentiation in the wild: isolation by dispersal limitation, isolation by adaptation and isolation by colonization. Molecular Ecology, 22, 5983–5999.

Palstra, F.P., and Ruzzante, D.E. (2008). Genetic estimates of contemporary effective population size: what can they tell us about the importance of genetic stochasticity for wild population persistence? Molecular Ecology, 17, 3428–3447.

! 85 ! Peakall, R., and Smouse, P.E. (2006). GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6, 288–295.

Peakall, R., and Smouse, P.E. (2012). GenAlEx 6.5: genetic analysis in Excel. Population genetic software for teaching and research-an update. Bioinformatics, 28, 2537-2539.

Pickett, S. T. A., Cadenasso, M. L., Grove, J. M., Boone, C. G., Groffman, P. M., Irwin, E., Kaushal, S.S., Marshall, V., McGrath, B.P., Nilon, C.H., Pouyat, R.V., Szlavecz K., Troy A., and Warren, P. (2011). Urban ecological systems: Scientific foundations and a decade of progress. Journal of Environmental Management, 92 (3), 331-362.

Pinder, J.E., III, Wiener, J.G., and Smith, M.H. (1978). The Weibull distribution: a new method of summarizing survivorship data. Ecology, 59, 175–179.

Primack, D., Imbres, C., Primack, R.B., Miller-Rushing, A.J., and Del Tredici, P. (2004). Herbarium specimens demonstrate earlier flowering times in response to warming in Boston. American Journal of Botany, 91, 1260–1264.

Pretzsch, H., Biber, P., Uhl, E., Dahlhausen, J., Schütze, G., Perkins, D., Rötzer, T., Caldentey, J., Koike, T., Con, T.V., Chavanne, A., Toit, B.D., Foster K., and Lefer, B. (2017). Climate change accelerates growth of urban trees in metropolises worldwide. Scientific Reports, 7 (1), 15403.

Pudovkin, A.I., Zaykin, D.V., and Hedgecock, D. (1996). On the potential for estimating the effective number of breeders from heterozygote-excess in progeny. Genetics, 144, 383– 387.

Reisch, C., Mayer, F., Rüther, C. and Nelle, O. (2008). Forest history affects genetic diversity— Molecular variation of Dryopteris dilatata (Dryopteridaceae) in ancient and recent forests. Nordic Journal of Botany, 25, 366-371.

Renshaw, J.F., and Doolittle, W.T. (1958). Silvical characteristics of yellow-poplar. USDA Forest Service, Station Paper 89. Southeastern Forest Experiment Station, Asheville, NC.

! 86 ! Ritland K. (1989). Correlated matings in the partial selfer Mimulus guttatus. Evolution, 43, 848– 859.

Ritland K. (2002). Extensions of models for the estimation of mating systems using n independent loci. Heredity, 88, 221–228.

Ritland, K., and Jain, S.K. (1981). A model for the estimation of outcrossing rate and gene frequencies using n independent loci. Heredity, 47, 35–52.

Robertson, A.W., Kelly, D., and Ladley, J.J. (2011). Futile selfing in the trees Fuchsia excorticata (Onagraceae) and Sophora microphylla (Fabaceae): inbreeding depression over 11 years. International Journal of Plant Sciences, 172, 191–198.

Robinson, J.D., and Moyer, G.R. (2013). Linkage disequilibrium and effective size when generations overlap. Evolutionary Applications, 6 (2), 290-302.

Rodger, J.G., and Johnson, S.D. (2013). Self-pollination and inbreeding depression in Acacia dealbata: Can selfing promote invasion in trees? South African Journal of Botany, 88, 252–259

Roetzer, T., Wittenzeller, M., Haeckel, H., and Nekovar, J. (2000). Phenology in central Europe - differences and trends of spring phenophases in urban and rural areas. International Journal of Biometeorology, 44, 60–66.

Rousset, F. (1997). Genetic differentiation and estimation of gene flow from f-statistics under isolation by distance. Genetics, 145, 1219–1228.

Rousset, F. (2008). GENEPOP’007: A complete re-implementation of the GENEPOP software for Windows and Linux. Molecular Ecology Resources, 8, 103–106.

Schaal, B.A. (1980). Measurement of gene flow in Lupinus texensis. Nature, 284, 450–451.

Schemske, D.W., and Lande, R. (1985). The evolution of self-fertilization and inbreeding depression in plants. II. Empirical observations. Evolution, 39, 41–52.

! 87 ! Schönswetter, P., Tribsch, A., Barfuss, M., and Niklfeld, H. (2002). Several Pleistocene refugia detected in the high alpine plant Phyteuma globulariifolium Sternb & Hoppe (Campanulaceae) in the European Alps. Molecular ecology, 11 (12), 2637–2647.

Schuelke, M. (2000). An economic method for the fluorescent labeling of PCR fragments. Nature Biotechnology 18, 233–234.

Schwartz, M.K., Luikart, G., and Waples, R.S. (2007). Genetic monitoring as a promising tool for conservation and management. Trends in Ecology and Evolution, 22, 25–33.

Searle, S.Y., Turnbull, M. H., Boelman, N. T., Schuster, W.S.F., Yakir, D., and Griffin, K.L. (2012). Urban environment of New York City promotes growth in northern red oak seedlings. Tree Physiology, 32 (4), 389–400.

Sewell, M.M. (1992). Mating System and Molecular Variation in Liriodendron tulipifera L. (Magnoliaceae). Unpublished PhD thesis, University of North Carolina at Chapel Hill.

Sewell, M.M., Parks, C.R., and Chase, M.W. (1996). Intraspecific chloroplast DNA variation and biogeography of North American Liriodendron L. (Magnoliaceae). Evolution, 50, 1147–1154.

Silvertown, J., and Charlesworth, D. (2001). Introduction to Plant Population Biology, 4th Edition. Wiley-Blackwell.

Sved, J.A. (1971). Linkage disequilibrium and homozygosity of chromosome segments in finite populations. Theoretical Population Biology, 2, 125–141.

Sved, J.A., Cameron, E.C., and Gilchrist, A.S. (2013). Estimating Effective Population Size from Linkage Disequilibrium between Unlinked Loci: Theory and Application to Fruit Fly Outbreak Populations. PLoS ONE, 8 (7), e69078.

Thompson, K.A., Renaudin, M., Johnson, M. (2016). Urbanization drives the evolution of parallel clines in plant populations. Proceedings of the Royal Society B: Biological Sciences, 283, 20162180.

! 88 ! Turner, T.F., Wares, J.P., and Gold, J.R. (2002). Genetic effective size is three orders of magnitude smaller than adult census size in an abundant, estuarine-dependent marine fish (Sciaenops ocellatus). Genetics, 162, 1329–1339.

Unfried, T.M., Hauser, L., and Marzluff, J.M. (2013). Effects of urbanization on Song Sparrow (Melospiza melodia) population connectivity. Conservation Genetics, 14, 41–53.

USDA-NRCS. (1998). Soil Survey of City of Baltimore, Maryland. USDA. Natural Resources Conservation Service, Washington, D.C.

Vellend M. (2004). Parallel effects of land-use history on species diversity and genetic diversity of forest herbs. Ecology, 85 (11), 3043–3055.

Wang, H.F., Sork, V.L., Wu, J.G., and Ge, J.P. (2010). Effect of patch size and isolation on mating patterns and seed production in an urban population of Chinese pine (Pinus tabulaeformis Carr.). Forest Ecology and Management, 260, 965–974.

Waples, R.S. (1989). A generalized approach for estimating effective population size from temporal changes in allele frequency. Genetics, 121, 379–391.

Waples, R.S. (1990a). Conservation genetics of Pacific salmon. II. Effective population size and the rate of loss of genetic variability. Journal of Heredity, 81, 267–276.

Waples, R.S. (1990b). Conservation genetics of Pacific salmon. III. Estimating effective population size. Journal of Heredity, 81, 277–289.

Waples, R.S. (2002). Effective size of fluctuating salmon populations. Genetics, 161, 783-791.

Waples, R.S. (2005). Genetic estimates of contemporary effective population size: To what time periods do the estimates apply? Molecular Ecology, 14, 3335–3352.

Waples, R.S. (2006). A bias correction for estimates of effective population size based on linkage disequilibrium at unlinked gene loci. Conservation Genetics, 7, 167–184.

! 89 ! Waples, R.S., and Antao, T. (2014). Intermittent breeding and constraints on litter size:

consequences for effective population size per generation (Ne) and per reproductive cycle

(Nb). Evolution, 68, 1722–1734.

Waples, R.S., Antao, T., and Luikart, G. (2014). Effects of overlapping generations on linkage disequilibrium estimates of effective population size. Genetics, 197 (2), 769–780.

Waples, R.S., and Do, C. (2008). LDNE: a program for estimating effective population size from data on linkage disequilibrium. Molecular Ecology Resources, 8, 753–756.

Waples R.S., and Do, C. (2010). Linkage disequilibrium estimates of contemporary Ne using SNPs and highly polymorphic molecular markers: an evaluation of precision and bias. Evolutionary Applications 3, 244–262.

Waples, R.S., Do, C., and Chopelet, J. (2011). Calculating Ne and Ne/N in age-structured populations: a hybrid Felsenstein-Hill approach. Ecology, 92, 1513-1522.

Waples R.S., England P.R. 2011. Estimating contemporary effective population size on the basis of linkage disequilibrium in the face of migration. Genetics 189, 633–644.

Waples, R.S., Luikart, G., Faulkner, J.R., and Tallmon, D.A. (2013). Simple life-history traits explain key effective population size ratios across diverse taxa. Proceedings of the Royal Society B: Biological Sciences, 280, 20131339.

Waples, R.S., and Yokota, M. (2007). Temporal estimates of effective population size in species with overlapping generations. Genetics, 175, 219–233.

Ward, L.F. (1881). Guide to the flora of Washington and vicinity. U.S. Natl. Museum Publ. No. 22, 264 p.

Wei X., and Jiang M. (2012). Contrasting relationships between species diversity and genetic diversity in natural and disturbed forest tree communities. New Phytologist, 193, 779- 786.

! 90 ! White, D.L.; and Lloyd, F.T. (1994). Defining old growth: implications for management. Paper presented at the Eighth Biennial Southern Silvicultural Research Conference, Auburn, AL, Nov. l–3, 1994.

White G.M., Boshier D.H., Powell W. 2002. Increased pollen flow counteracts fragmentation in a tropical dry forest: an example from Swietenia humilis Zuccarini. Proceedings of the National Academy of Sciences USA, 99, 2038–2042.

Williams, M. (2002). Deforesting the earth: from prehistory to global crisis. Chicago, USA, University of Chicago Press.

Winchell K.M., Reynolds R.G., Prado-Irwin S.R., Puente-Rolón A.R., and Revell L.J. (2016). Phenotypic shifts in urban areas in the tropical lizard Anolis cristatellus. Evolution, 70, 1009–1022.

Winfree, R., Aguilar, R., Vazquez, D.P., LeBuhn, G., and Aizen, M. (2009). A meta-analysis of bees' responses to anthropogenic disturbance. Ecology, 90, 2068–2076.

Wright, S., (1922). Coefficients of inbreeding and relationship. The American Naturalist, 56, 330–338.

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

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

Wright, S., (1951). The genetical structure of populations. Annals of Eugenics, 15, 323–354.

Xu, M., Li, H., and Zhang, B. (2006). Fifteen polymorphic simple sequence repeat markers from expressed sequence tags of Liriodendron tulipifera. Molecular Ecology Notes, 6, 728– 730.

Xu, M., Sun, Y., and Li, H. (2010). EST-SSRs development and paternity analysis for Liriodendron spp. New Forests, 40, 361–382.

Yakub, M., and Tiffin, P. (2017). Living in the city: urban environments shape the evolution of a native annual plant. Global Change Biology, 23, 2082–2089.

! 91 ! Yang, A., Zhang, J., Yao, X., and Huang, H. (2011). Chloroplast microsatellite markers in Liriodendron tulipifera (Magnoliaceae) and cross-species amplification in L. chinense. American Journal of Botany, 98, e123–e126.

Yang, A.H., Zhang, J.J., Tian, H., and Yao, X.H. (2012). Characterization of 39 novel EST-SSR markers for Liriodendron tulipifera and cross-species amplification in L. chinense (Magnoliaceae). American Journal of Botany, 99, e460–e464.

Yao, X., Zhang, J., Ye, Q., and Huang, H. (2008). Characterization of 14 novel microsatellite loci in the endangered Liriodendron chinense (Magnoliaceae) and cross-species amplification in closely related taxa. Conservation Genetics, 9, 483–485.

Yauk C.L., and Quinn J.S. (1996). Multilocus DNA fingerprinting reveals high rate of heritable genetic mutation in herring gulls nesting in an industrialized urban site. Proceedings of the National Academy of Sciences USA, 93, 12137–12141.

Yauk C.L., Fox G.A., McCarry B.E., and Quinn, J.S. (2000). Induced minisatellite germline mutations in herring gulls (Larus argentatus) living near steel mills. Mutation Research, 452, 211–218.

Young, A., Boyle, T., and Brown, T. (1996). The population genetic consequences of habitat fragmentation for plants. Trends in Ecology and Evolution, 11, 413–418.

Zhang, X., Carlson, A., Tian, Z., Staton, M.E., Schlarbaum, S.E., Carlson, J., and Liang, H. (2015). Genetic characterization of Liriodendron seed orchards with EST-SSR markers. Journal of Plant Science and Molecular Breeding, 4(1).

Zhdanova, O.L., and Pudovkin, A.I. (2008). Nb_HetEx: a program to estimate the effective number of breeders. Journal of Heredity, 99, 694–695.

Zhu L. and Lou, A. (2013). Old-Growth Platycladus orientalis as a Resource for Reproductive Capacity and Genetic Diversity. PLoS ONE, e56489.

! 92 !