REPRODUCTIVE ECOLOGY and GENETICS of PLANT INVASIONS: a CASE STUDY of LYTHRUM SALICARIA (PURPLE LOOSESTRIFE) by Christopher M. B

REPRODUCTIVE ECOLOGY AND GENETICS OF PLANT INVASIONS: A CASE

STUDY OF LYTHRUM SALICARIA (PURPLE LOOSESTRIFE)

Christopher M. Balogh

A thesis submitted in conformity with the requirements

for the degree of Doctor of Philosophy

Department of Ecology and Evolutionary Biology

University of Toronto

REPRODUCTIVE ECOLOGY AND GENETICS OF PLANT INVASIONS: A CASE

STUDY OF LYTHRUM SALICARIA (PURPLE LOOSESTRIFE)

Christopher M. Balogh

Doctor of Philosophy

Department of Ecology and Evolutionary Biology

University of Toronto

2018

ABSTRACT

Biological invasions provide a valuable context for studying contemporary evolution.

Because reproduction is a key process determining invasion success, investigations of the reproductive ecology and genetics of populations are particularly insightful. My Ph.D. thesis is comprised of five inter-related studies on Lythrum salicaria (purple loosestrife), a tristylous plant from Eurasia that has invaded wetlands in North America during the past

150 years. The specific questions I addressed, using computer simulations, surveys of natural populations, and glasshouse and field experiments, concerned the mechanisms governing the frequencies of mating types in populations, the characteristics and functional significance of partial self-incompatibility, the influence of floral morph structure and demography on mating and fertility, and the occurrence and significance of inbreeding depression for invading populations.

Stochastic computer models incorporating morph-specific self-compatibility and tetrasomic inheritance revealed patterns of morph-frequency variation and asymmetric morph loss consistent with data from L. salicaria, including a survey I conducted in

Ontario of 114 populations that revealed stochastic loss of the S-morph from small populations. Glasshouse experiments involving controlled self- and cross-pollination demonstrated that ~34% of L. salicaria plants exhibited partial self-incompatibility, that this trait was weakly heritable, and was stable in expression over two years. Progeny testing of open-pollinated families in six populations and isolated plants in a field experiment revealed high rates of inter-morph mating demonstrating that L. salicaria is robust to demographic variation associated with colonization. Cumulative estimates of inbreeding depression (δ) in a four-year experiment, three under field conditions, revealed

δ = 0.48 and 0.68, depending on how multiplicative fitness was estimated. The values were consistent with the outcrossed mating system of populations and high enough to oppose the spread of self-fertilization, unless pollinators or mates limit fertility, but lower than is frequently reported in other outcrossing plant species.

In summary, this research provides novel information on the mechanisms that maintain tristyly in invasive populations of L. salicaria, despite the occurrence of partial self- incompatibility. High rates of inter-morph mating promoted by reliable bee-mediated pollinator service limit opportunities for the breakdown of the polymorphism, an evolutionary transition that has commonly occurred in other invasive tristylous species.

iii

ACKNOWLEDGEMENTS

My PhD thesis would not have been possible without extensive help from many individuals and organizations. First, I would like to thank my PhD supervisor Spencer

Barrett, whose knowledge on plant life history and reproductive biology, highly attuned writing skills, and apparently endless energy and enthusiasm for science have been absolutely necessary for my thesis completion. I have benefited immensely from his guidance in my graduate training and will not forget his contributions to my growth professionally as a scientist and personally. I also thank my committee members James

Thompson and John Stinchcombe for providing invaluable feedback on my project over the last 5+ years, as well as Peter Kotanen, Helen Rodd, and Art Weis who have served on my defence committee and Stephen Wright who assisted with my appraisal. I thank my external examiner Stephen Weller for his thoughtful feedback and questions on my thesis and during my defence. Finally, I thank Andrew Stephenson and the members of the

Stephenson lab for introducing me to plant mating systems and the questions which abound in them.

Additionally, I thank Christopher Eckert and Robert Colautti at Queens University whose works on the invasion genetics of Lythrum salicaria were pivotal to my studies on this system. I thank the other members of the Barrett Lab including Ramesh Arunkumar,

Stuart Campbell, Julia Charlebois, Joana Costa, Daisy Crowson, Nicolay Cunha, Josh

Hough, Zoe Humphries, Wan Jin, Joana Rifkin, Andrew Simpson, David Timerman,

Haoran Xue and Wei Zhou who have provided intellectually stimulating discussion as well as useful guidance at various time during my thesis research at the University of

Toronto. I also thank members of the Department of Ecology and Evolutionary Biology

particularly Megan Greischar, Eddie Ho, Tyler Kent, Julia Kreiner, Jason Laurich, Vanessa

Nielsen, Rebecca Schalkowski, Sarah Steele, Matt Stata, Stefanie Sultmanis, Alison

Wardlaw, Corlett Wood and Amanda Xerub for their advice and support. I would not have been able to gather the data required to complete my thesis without endless hours of assistance from a multitude of undergraduate assistants. I am especially grateful to Cole

Brookson, Jamy Fu, Sahil Gupta, Dowon Lee, Teresa Maddison, Tobias Mankis, Baily

McCollough, Fernanda Pazin, Shirley Qiu, and Carol Wong for all of their help with data collection in the field, glasshouse and laboratory. I thank Bill Cole for providing advice and logistical support in the glasshouse and in the field as well as Andrew Petrie and Bruce

Hall for their assistance with horticulture and the growing of thousands of Lythrum salicaria plants in the glasshouse. I thank Stephen Schneider and John Jensen at the

Koffler Scientific Reserve in providing assistance with setting up my experiments, maintenance and troubleshooting in the field. Without these individuals I would not have been able to properly conduct the large-scale field experiments that I required for making this thesis a success.

The doctoral research that comprises this PhD thesis was funded by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada to Spencer Barrett, as well as by student scholarships from the Department of Ecology and Evolutionary Biology,

University of Toronto.

Finally, I thank my parents Robert and Karla Balogh for their unwavering support during my time at the University of Toronto. I also thank my wife Jennifer Balogh for her love, patience and emotional support especially during the long stretches I was away from

Pennsylvania. I also thank her for assistance in putting the final thesis togther over the past few months. Finally, I thank coffee – coffee made everything possible.

TABLE OF CONTENTS

ABSTRACT ...... II ACKNOWLEDGEMENTS ...... IV TABLE OF CONTENTS ...... VII LIST OF TABLES ...... X LIST OF FIGURES ...... XI LIST OF APPENDICES ...... XIV CHAPTER 1 GENERAL INTRODUCTION ...... 1

Invasion genetics and reproductive systems ...... 1 Uniparental reproduction and colonization...... 3 Tristyly and colonization ...... 9 Colonization and inbreeding depression...... 15 Study system: Lythrum salicaria ...... 18 Thesis objectives ...... 22

CHAPTER 2 THE INFLUENCE OF PARTIAL SELF- INCOMPATIBILITY AND AUTOTETRAPLOIDY ON FLORAL MORPH FREQUENCIES IN FINITE TRISTYLOUS POPULATIONS ...... 27

Abstract ...... 27 Introduction ...... 28 Material and Methods ...... 34 Results ...... 36 Discussion ...... 47

CHAPTER 3 STOCHASTIC PROCESSES DURING INVASION: THE INFLUENCE OF POPULATION SIZE ON STYLE-MORPH FREQUENCY VARIATION IN LYTHRUM SALICARIA (PURPLE LOOSESTRIFE) ...... 60

Abstract ...... 60 Introduction ...... 61 Materials and Methods ...... 65

vii

Results ...... 72 Discussion ...... 81

CHAPTER 4 GENETIC AND ENVIRONMENTAL INFLUENCES ON PARTIAL SELF-INCOMPATIBILITY IN LYTHRUM SALICARIA (LYTHRACEAE) ...... 90

Abstract ...... 90 Introduction ...... 91 Materials and Methods ...... 96 Results ...... 105 Discussion ...... 116

CHAPTER 5 THE INFLUENCE OF FLORAL MORPH RATIOS AND LOW PLANT DENSITY ON MATING AND FERTILITY IN A TRISTYLOUS COLONIZING SPECIES ...... 126

Abstract ...... 126 Introduction ...... 127 Materials and Methods ...... 130 Results ...... 145 Discussion ...... 150

CHAPTER 6 AN EXPERIMENTAL STUDY OF INBREEDING DEPRESSION AND THE EFFECTS OF COMPETITION IN AN INVASIVE PLANT...... 158

Abstract ...... 158 Introduction ...... 159 Materials and Methods ...... 164 Results ...... 175 Discussion ...... 185

CHAPTER 7 GENERAL CONCLUSIONS AND FUTURE DIRECTIONS ...... 193

Overview ...... 193

viii

Final Remarks ...... 201

LITERATURE CITED ...... 203 APPENDIX TO CHAPTER 2 ...... 243 APPENDIX TO CHAPTER 5 ...... 276 APPENDIX TO CHAPTER 6 ...... 298

LIST OF TABLES

Table 2.1. Effects of simulated model terms on the outcome of models ...... 41

Table 2.2. Floral morph frequencies (L-, M-, S-morph), number of sampled populations and plants (N), and number of dimorphic (Nd) and monomorphic

(Nm) populations from surveys of Lythrum salicaria conducted over the past 150 years ...... 54

Table 3.1. Heterogeneity G-test comparing the number of non-trimorphic and trimorphic populations between size classes in the 2013 study of Lythrum salicaria in Ontario ...... 75

Table 3.2. Heterogeneity G-tests comparing the proportion of non-trimorphic and trimorphic populations of Lythrum salicaria in Ontario ...... 76

Table 4.1. Location of the four population of Lythrum salicaria in Toronto, Ontario, Canada ...... 98

Table 4.2. Effects of family, environment and their interaction on the expression of partial self-incompatibility in cloned individuals of Lythrum salicaria ...... 109

Table 4.3. The number of L- and M-morph progeny in selfed families of the M-morph of Lythrum salicaria...... 112

Table 5.1. The floral morph structure and morph ratios of the six populations of Lythrum salicaria in Ontario that were used in progeny testing ...... 135

Table 5.2. Patterns of mating in six populations of Lythrum salicaria based on progeny testing of open-pollinated seed families ...... 140

LIST OF FIGURES

Figure 1.1. The long-styled, mid-styled, and short-styled morphs (L-, M-, and S-morphs, respectively) of a tristylous species ...... 11

Figure 1.2. A de Finetti diagram representing the morph-frequency space that can be occupied by populations of a tristylous species ...... 13

Figure 1.3. Images of Lythrum salicaria from various locations worldwide ...... 22

Figure 2.1. The mean number of generations required for each of the tristyly alleles to be lost from simulated populations with different parameters ...... 37

Figure 2.2. The mean frequency in which simulated populations lost alleles ...... 38

Figure 2.4. The influence of population size (columns) and frequency of morph-specific partial selfing on the evolution of dimorphic population structures ...... 43

Figure 3.1. The location of Lythrum salicaria populations sampled for style morph ratios in Ontario, Canada during summer 2013 ...... 68

Figure 3.2. Histograms of population sizes in Lythrum salicaria, Ontario, Canada...... 69

Figure 3.3. de Finetti plots of population morph frequencies in populations of Lythrum salicaria sampled from Ontario, Canada ...... 73

Figure 3.4. The relation between morph evenness and population size in Lythrum salicaria populations sampled in Ontario, Canada, during summer 2013 ...... 80

Figure 4.1. Variation in the proportion of plants setting different quantities of A) fruit and B) seed after cross- and self-pollination ...... 106

Figure 4.2. Morph-specific variation in partial self-incompatibility in Lythrum salicaria based on experimental pollination conducted under glasshouse conditions on 338 plants ...... 107

Figure 4.3. Relations between partial self-incompatibility expressed in 2013 versus 2014 in a sample of Lythrum salicaria plants pollinated in each year under glasshouse conditions ...... 108

Figure 4.4. Reaction norms of partial self-incompatibility for 12 cloned genotypes of Lythrum salicaria grown under wet and dry conditions in a glasshouse ...... 111

Figure 4.5. Relation between parental and offspring values of partial self- incompatibility following self-pollination of Lythrum salicaria plants grown under glasshouse conditions ...... 114

Figure 4.6. Morph-specific differences in partial self-incompatibility in L- and M-morph offspring segregating from self-fertilization of the M-morph of Lythrum salicaria ...... 115

Figure 5.1. Locations and morph ratios in populations of Lythrum salicaria sampled in Ontario, Canada for mating analyses ...... 134

Figure 5.2. Estimated frequency of inter-morph mating and 95% confidence intervals in six populations of Lythrum salicaria of varying floral morph structure ...... 146

Figure 5.3. Index of pollen limitation (PL) based on open- and hand- pollinated fruit and seed set (see Methods) for isolated plants of Lythrum salicaria in a field experiment...... 148

xii

Figure 5.4. Reproductive success of open-pollinated flowers of Lythrum salicaria in a field experiment...... 149

Figure 6.1. The competition and breeding treatments used in the inbreeding depression experiment ...... 166

Figure 6.2. Images of the inbreeding depression experiment on Lythrum salicaria ...... 167

Figure 6.3. Mean and 95% confidence intervals (bars) of trait values in early life in selfed- and outcrssed-individuals of Lythrum salicaria ...... 176

Figure 6.4. The inbreeding depression (δ) mean and 95% confidence intervals (bars) for early-life traits of Lythrum salicaria ...... 177

Figure 6.5. The mean and 95% confidence intervals (bars) for trait values of selfed and outcrossed progeny in Lythrum salicaria from 2014 to 2017 ...... 179

Figure 6.6. Inbreeding depression (δ) mean and 95% confidence intervals (bars) of life-history traits in Lythrum salicaria ...... 180

Figure 6.7. Inbreeding depression (δ) in average growth rate (AGR) experienced by Lythrum salicaria plants in the non-linear growth models ...... 182

Figure 6.8. The relative performance (RP) of Lythrum salicaria plants represented as the average growth rate (AGR) between competitive environments (‘none’, selfed, outcrossed) ...... 183

Figure 6.9. Multiplicative inbreeding depression (δ) in each year and cumulatively at the end of the four-year experiment on Lythrum salicaria with 95% confidence intervals (bars) ...... 185

xiii

LIST OF APPENDICES

Appendix 2 Table 1. Effects of model terms on the number of generations to allele loss in simulated tristylous populations ...... 243

Appendix 2 Table 2. Optimized models of the frequency of allele loss in simulated tristylous populations ...... 247

Appendix 2 Table 3. Optimized models for the frequency in which trimorphism is lost from populations resulting in dimorphism ...... 258

Appendix 2 Table 4. Frequency of dimorphic population structures evolving from trimorphism under different simulation parameters ...... 261

Appendix 2 Table 5. Comparison between simulation results in this study and that of Heuch (1980) for allele loss from tristylous populations ...... 265

Appendix 2 Table 6. Comparison of model genotype frequencies against deterministically predicted genotpye frequencies at equilibrium in tristylous populations ...... 266

Appendix 2 Table 7. Optimized model terms describing the frequency of each morph in populations of different size and with different partially selfing morphs ...... 268

Appendix 2 Table 8. Morph frequencies in trimorphic populations of different size, selfing rates and with different selfing morphs ...... 271

Appendix 2 Fig. 1. Frequency of floral morphs in tristylous populations with morph-specific partial selfing ...... 275

Appendix 5 Table 1. The identification number, genotype, frequency at equilibrium and floral morph for individuals of Lythrum salicaria ...... 276

xiv

Appendix 5 Table 2. The number of maternal families of each genotype sampled from populations and the progeny of floral morphs and total progeny per genotype...... 278

Appendix 5 Table 3. Progeny morph ratios obtained from maternal parents in six natural populations of Lythrum salicaria of varying floral morph structure ...... 282

Appendix 5 Table 4. Progeny ratios of isolated plants of Lythrum salicaria in a field experiment at the Koffler Scientific Reserve ...... 294

Appendix 5 Fig. 1. Locations and floral morphs of 57 isolated plants in two adjacent fields of Lythrum salicaria at the Koffler Scientific Reserve ...... 297

Appendix 6 Table 1. The likelihood-ratio significance values for harvest data collected from the glasshouse experiment on Lythrum salicaria in 2014 including germination, survival, flowering, flowering time and final inflorescence mass...... 298

Appendix 6 Table 2. The AIC values for nonlinear model types to which each year of growth data for Lythrum salicaria was fit ...... 303

Appendix 6 Fig. 2. The multiplicative depiction of relative performance (RP) of plants of Lythrum salicaria in the different competitive environments of the study ...... 306

CHAPTER 1

GENERAL INTRODUCTION

Invasion genetics and reproductive systems

The field of invasion genetics uses biological invasions as models for understanding processes in ecology and evolution. Invasion genetics was initiated at the 1964 meeting

“The Genetics of Colonizing Species” in Asilomar, California (USA) and the proceedings were subsequently published and became an influential volume edited by Baker and

Stebbins (1965). Over 30 luminaries in the fields of evolution and ecology including R. W.

Allard, T. Dobzhansky, R. C. Lewontin, E. W. Mayr, and E. O. Wilson gave talks at this meeting. The participants presented work in which they used colonizing species and biological invasions to address a range of fundamental questions including the role of stochastic forces and selection during colonization, patterns of genetic diversity in colonizing populations, and the life history and reproductive traits that commonly occur in successful colonizers (Baker & Stebbins 1965).

The influence of the Baker and Stebbins’ volume continues today through wide citations and an anniversary meeting at Asilomar in 2014. This meeting led to a second volume celebrating the legacy of the Baker and Stebbins volume that provided a synthesis of contemporary studies on the genetics of biological invasions (Barrett et al. 2017).

Today, invasion genetics provides a valuable framework for investigating the importance of reproduction for plant colonization and the genetics of populations. My thesis focuses on the role of reproductive systems in the invasive spread of tristylous Lythrum salicaria

(purple loosestrife) and addresses a range of questions concerning the maintenance of the

1 CHAPTER 1: GENERAL INTRODUCTION 2 floral polymorphism tristyly in colonizing populations, the potential role of self- fertilization in colony establishment and the importance of inbreeding depression to the maintenance of outcrossing and future population growth.

Low density is a ubiquitous feature of plant invasions. Individuals may be isolated after colonizing events and exist as outliers (Levin 1995), or they may form small populations after propagules establish in an area. Isolation or small population size may produce Allee effects in invasive species resulting in limitations on successful reproduction (Allee 1931; Byers & Meagher 1992; Steffan-Dewenter & Tscharntke 1999).

Under these conditions, the reproductive system plays a key role in determining whether plant invaders can successfully establish and persist at low density (Sakai et al. 2001;

Pannell 2015). Despite the initial constraints of low density, invasive species can often spread rapidly and undergo subsequent adaptation to novel conditions if sufficient genetic variation is present in populations. Indeed, clinal variation in life-history traits is reported along geographical gradients in various environmental factors in invasive species (Maron et al. 2004; Montague et al. 2008; Colautti & Barrett 2010a). In some cases, common garden and transplant studies have shown that these changes are adaptive and therefore represent examples of ‘contemporary evolution’ (Colautti & Barrett 2013). The reproductive system represents a key life-history trait for successful invasions and as a result there has been considerable interest in the reproductive diversity of colonizing species (reviewed in Baker 1965; Barrett et al. 2008; Barrett 2011; Pannell 2015).

Flowering plants display a wide range of reproductive systems. Although approximately 90% of all plant species are hermaphroditic (Renner & Ricklefs 1995), patterns of mating can vary greatly depending on whether plants are obligately outcrossing

CHAPTER 1: GENERAL INTRODUCTION 3 or are self-fertile (Goodwillie et al. 2005; Barrett & Harder 2017). Because self- compatible hermaphroditic parents are capable of reproducing via self-fertilization

(selfing), various degrees of inbreeding exist in plant populations (Lloyd 1979, 1992;

Lloyd & Schoen 1992). However, selfing may also be costly in largely outcrossing populations as it increases homozygosity in offspring, exposes deleterious recessive alleles, and can result in reduced fitness owing to inbreeding depression (Charlesworth &

Willis 2009; Igić & Busch 2013). The interplay between mating system, genetic diversity and natural selection are important in influencing the success of propagules in forming new colonies and opportunities for rapid evolution during biological invasion. Therefore, understanding mating patterns in colonizing populations represents one of the key research problems in the evolutionary ecology of biological invasions.

Uniparental reproduction and colonization

Plant species have evolved floral strategies which govern the relative amounts of cross- and self-fertilization. One of the most widespread mechanisms maintaining high levels of cross-fertilization is self-incompatibility (SI): a physiological-genetic system that enables individuals to reject their own pollen or pollen tube growth while allowing pollen from other individuals to effect fertilization (Bateman 1952; de Nettancourt 1977; Franklin-

Tong 2008). Self-incompatibility has multiple origins and is widely conserved across plant lineages. For example, three major families (Solanaceae, Scrophulariaceae, Rosaceae) possess a homologous, gametophytically-based SI system, whereas the Brassicaceae possess a sporophytic SI system which does not share homology with gametophytic systems (McCubbin & Kao 2000; Igić & Kohn 2001; Steinbachs & Holsinger 2002).

Despite SI being reported in 71 families and 250 genera representing at least 60% of all

CHAPTER 1: GENERAL INTRODUCTION 4 angiosperm species (Allen & Hiscock 2008), it is frequently lost resulting in the acquisition of self-compatibility (SC) through loss-of-function mutations (Good-Avila et al. 2008; Mable 2008). The breakdown of SI to SC and the evolution of uniparental reproduction is one of the most common reproductive transitions in plant evolution. It is particularly well exemplified in Solanum in which there have been numerous independent transitions from SI to SC, but in no case is there an example of the reverse transition, and this is general pattern in angiosperms (Igić et al. 2008). Basic models for the evolution of

SC from SI usually assume a single self-compatibility mutation spreading in a population, which is certainly supported in some cases (Stone 2002). However, SI is not always completely expressed in some species and can show considerable quantitative variation

(Good-Avila & Stephenson 2002). Therefore, the spread of self-compatibility via a single mutation may not always accurately reflect the processes which occur in nature during the loss of SI.

In the vast majority of SI species that have been studied experimentally, by controlled self and cross-pollinations possess, individuals that possess the capacity for self-fertilization have been detected. This phenomenon has been described in the literature using multiple terms including leaky self-incompatibility, pseudo-self-compatibility, imperfect self-incompatibility or partial self-incompatibility (PSI). Throughout this thesis I will use the latter term which can be defined as the occurrence of selfed seed set of varying amounts by individuals within a largely SI species. Levin (1996) used the term pseudo-self-fertility in his review of this condition, but I prefer PSI in this thesis as it is more widely used in the literature and identifies that the phenomenon results from a weakening of SI expression and is not associated with fertility-sterility criterion. Evidence

CHAPTER 1: GENERAL INTRODUCTION 5 exists that PSI is determined by genetic modifiers and can respond to selection. For example, in Nicotiana sanderae individuals were detected which produced various levels of expression of PSI (East 1934), and selection experiments in Nemesia strumosa resulted in lines with either high or low levels of PSI (Robacker & Ascher 1978). More recent molecular work has identified a segregating allele responsible for PSI in Petunia hybrida

(Flaschenriem & Ascher 1979). Although there is evidence for a genetic basis to PSI, other evidence also suggests that the level of seed set after self-pollination by self- incompatible plants can be influenced by environmental factors. For example, heat treatments induced the expression of PSI in Lillium longiflorum and Trifolium pretense, both of which exhibited increased quantities of pollen-tube growth from self-pollination after heat treatment (Hopper et al. 1967; Kendall & Taylor 1969).

The timing of pollination with respect to the blooming period also appears to influence the expression of PSI. For example, isolated individuals of Lythrum salicaria set more seed later in the season than earlier in the season (Stout 1923). In Beta vulgaris more seed was set at the end of the season after self-pollination than earlier in the season (Owen

1942). The presence of prior fruit set or early pollination in a flower may also alter the expression of PSI. For example, individuals of Solanum carolinense which failed to set fruit early in the season were more likely to set fruit after self-pollination later in the season than plants which set fruit early in the season (Travers et al. 2004), and flowers of

Brassica campestris and Campanula rapunculoides which do not receive pollen in the first days of anthesis were more likely to exhibit PSI just before they senesced than those which received pollen shortly after anthesis (Richards & Thurling 1973; Vogler et al.

1998). The majority of studies that have investigated PSI expression have been conducted

CHAPTER 1: GENERAL INTRODUCTION 6 on agricultural and horticultural species and relatively little is known about the potential ecological and evolutionary consequences of PSI in wild populations of plants.

Studies on the evolutionary and ecological roles of PSI have investigated several hypotheses about the function of PSI in wild species. In-depth investigations of

Campanula rapunculoides, a North American weed indicate: 1) a genetic component to

PSI and significant parent/offspring regression of PSI expression (Good-Avila &

Stephenson 2002); 2) a developmental component to PSI manifested as increased expression with flower age (Stephenson et al. 2000) and; 3) a potential ecological role of

PSI in reproductive assurance because of the higher self-seed set in isolated plants when pollinator service was limited (Good-Avila & Stephenson 2002). Similar studies in

Solanum carolinense have uncovered that PSI is associated with particular S-alleles, and that inbreeding depression induced by herbivores may oppose a transition to full self- compatibility in the species (reviewed in Mena-Ali & Stephenson 2007).

The most detailed genetic analysis of PSI was recently conducted on

Leavenworthia alabamica by Baldwin and Schoen (2017). They investigated the heritability of PSI by examining the number of pollen tubes after self-pollination in the style in four populations. They found that a continuous distribution of PSI was evident in all populations, with up to 90% of plant exhibiting PSI, and their heritability estimates ranged from 0.39-0.57. A positive correlation between PSI and self-fertilization was detected in a meta-analysis with a largely bimodal, though distinctly quantitative expression of PSI (Raduski et al. 2012). These findings suggest that under the appropriate demographic circumstances PSI may provide reproductive assurance in species that experience frequent colonizing episodes. Moreover, heritable PSI expression may

CHAPTER 1: GENERAL INTRODUCTION 7 sometimes provide the necessary genetic variation to enable evolutionary transitions from predominant outcrossing to high rates of selfing, although this possibility has not been investigated in any detail.

The relations between colonization and reproductive systems are one of the most persistent themes in evolutionary biology. Darwin (1876) observed that although “nature generally abhors perpetual self-fertilization”, an absence of mates may sometimes favour a transition to self-fertilization (Darwin’s reproductive assurance hypothesis). Reproductive assurance explains the observation that SC species are much more common on islands compared to continental regions than SI species (Grossenbacher et al. 2017) and also among successful colonizers (Sakai et al. 2001; Barrett 2011). This is often referred to as

‘Baker’s law’ (Baker 1955, 1967), which states that SC individuals should have an advantage in establishment after long-distance dispersal because a single colonist is capable of founding a population without the requirement of mates. Despite the intuitive appeal of Baker’s law, several authors have challenged its generality (e.g. Dornier et al.

2008; Cheptou & Massol 2009) by extending the scope of his arguments to other demographic scenarios, such as metapopulations (see Pannell & Barrett 1998); and including the joint evolution of mating and dispersal in arguments.

Empirical tests of Baker’s law commonly involve either measurements of the frequency of SI and SC in native vs introduced plant communities, or the frequency of SI and SC in sister taxa with different range sizes or invasion success. For example,

Rambuda and Johnson (2004) measured the capacity for autonomous reproduction in 17 invasive species in South Africa and found that 72% were capable of autonomous self- pollination, a higher frequency than occurs in the native flora. Van Kleunen et al. (2008)

CHAPTER 1: GENERAL INTRODUCTION 8 investigated the capacity for autonomous self-pollination in ten species pairs of Iridaceae introduced to various locations around the world by the horticultural trade, and in which one species from each pair became naturalized and the other did not. The authors found that the naturalized member of each species pair exhibited a higher capacity for autonomous self-pollination than the non-naturalized member of each pair.

Whereas these studies and others like them support the key concepts of Baker’s law this support is not universal. A meta-analysis of invasive plant traits conducted by

Sutherland (2004) proposed that traits or environments associated with self-compatibility, such as life history or habitat, may be the actual drivers of biological invasions and not self-compatibility per se. However, several recent syntheses (see Pannell 2015; Pannell et al. 2015; Grossenbacher et al. 2017) have concluded that Baker’s basic proposition about the advantages of SC in establishment at low density are generally supported by theoretical and comparative evidence. Nevertheless, there are relatively few studies that provide direct experimental evidence demonstrating that self-compatible plants or plants capable of autonomous selfing are favoured by reproductive assurance when mates or pollinators are limiting.

An experiment in Campanula rapunculoides exposed individuals to varying levels of pollinator service and detected higher selfing rates in plants exposed to pollinators every fourth day versus every day (Good-Avila et al. 2001), a pattern consistent with reproductive assurance. Similarly, Moeller and Geber (2005) placed Clarkia xantiana individuals with variation in traits associated with self-pollination into low and high- density arrays and found that plants with traits related to selfing tended to set more seed in low density conditions than plants with trait variation related to outcrossing. Given the

CHAPTER 1: GENERAL INTRODUCTION 9 frequent association between colonizing success and uniparental reproduction among invasive plants, future experimental studies on invasive species with variable levels of self-compatibility, including PSI, would be valuable to investigate whether the capacity to self-fertilize at low density is favoured when individuals are isolated, or populations are small in size. One of the goals of this thesis is to investigate the nature and potential reproductive consequences of partial self-incompatibility in a predominantly outcrossing invasive plant species.

Tristyly and colonization

In the majority of SI species, mating types are morphologically indistinguishable and an individual’s mating type cannot be determined without extensive crosses and/or genotyping. This barrier, as well as the large number of mating types typically found in SI species, has deterred extensive studies of the ecology and evolution of SI in natural populations (de Nettancourt 1977). In contrast, species with heteromorphic incompatibility systems possess mating types which are morphologically distinguishable as a result of differences in the relative positioning of sexual organs within flowers. As a consequence, there is an extensive literature on the relative frequency of mating types in natural populations of species with this form of incompatibility (reviewed in Barrett 1993; Barrett

& Shore 2008). There are two types of heteromorphic incompatibility – distyly and tristyly – and heteromorphic floral polymorphisms are reported from 28 angiosperm families and have originated by convergent evolution numerous times (Darwin 1877;

Ganders 1979; Barrett 1992; Weller 2009). Whereas distyly occurs in virtually all of these families, tristyly is restricted to six families of which only three (Lythraceae, Oxalidaceae,

Pontederiaceae) have been examined in detail (Barrett 1993).

CHAPTER 1: GENERAL INTRODUCTION 10

Populations of tristylous plants usually possess three mating types, the long-, mid- and short-styled morphs, hereafter referred to as the L-, M-, and S-morphs, which differ in the reciprocal placement of anthers and stigmas in each mating type (Fig. 1.1). Tristyly is most commonly governed by two diallelic loci (S and M), with the S-locus epistatic to the

M-locus (Weller 1976; Lewis & Jones 1992; Gettys & Wofford 2008). Most tristylous species possess a trimorphic incompatibility system in which compatible pollinations only occur when pollen is transferred between anthers and stigmas of similar height (depicted as arrows in Fig. 1.1), whereas incompatible pollinations, including self-pollination, occur when pollen is transferred from anthers to stigmas at different levels and these result in no or limited seed set (Darwin 1877; Barrett & Cruzan 1994). Trimorphic incompatibility governs mating patterns in tristylous populations and when it is strongly expressed causes intermorph (disassortative) mating. This outcrossed mating system should result in negative frequency-dependent selection and equal morph ratios in tristylous populations.

A notable feature of trimorphic incompatibility is that it often varies in expression.

The variation ranges from species with rigid incompatibility (Weller 1992), to those that possess cryptic trimorphic incompatibility (Cruzan & Barrett 2016), to those that are fully self-compatible (Ornduff 1972). Of particular significance is the occurrence of morph- specific partial self-incompatibility, in which there is variation among the floral morphs in the strength of trimorphic incompatibility. Most commonly, the M-morph exhibits significantly higher level of PSI than the remaining two morphs, the S-morph exhibits very little PSI, and the L-morph exhibits a low level of PSI (Darwin 1877; Barrett 1977;

Barrett & Anderson 1985; Barrett & Cruzan 1994; Colautti et al. 2010a; Puentes et al.

2013). This curious pattern of incompatibility expression raises a number of interesting

CHAPTER 1: GENERAL INTRODUCTION 11 questions concerning the proximate genetic mechanism governing this variation and the potential ecological significance of polymorphism in self-compatibility. Given the ease of identifying incompatibility types under field conditions, tristylous species represent useful model systems for investigating the extent to which variation in incompatibility may influence mating patterns and hence the style morph structure of populations.

Figure 1.1. The long-styled, mid-styled, and short-styled morphs (L-, M-, and S-morphs, respectively) of a tristylous species. Triangles represent stigmas, ovals represent anthers and the genotypes determining each style morph are provided for the diploid and autotetraploid case, assuming that selfing enables genotypes homozygous at the S locus.

Arrows represent compatible pollinations resulting in seed set; other pollen stigma combinations are usually incompatible. Modified from Barrett & Shore (2008).

CHAPTER 1: GENERAL INTRODUCTION 12

Field surveys of style morph frequencies in heterostylous populations have a long and venerable history (reviewed in Barrett 1993). Mathematical and computer simulation models have focused on the different morph-ratio equilibria which could potentially occur in tristylous species under various reproductive and demographic scenarios (Fisher 1941a;

Fisher & Mather 1943; Finney 1952, 1983; Heuch 1979a; Barrett et al. 1983, 1989;

Morgan & Barrett 1988; Husband & Barrett 1992a; Barrett & Arroyo 2012). In both types of analysis, morph ratios can be represented visually by de Finetti diagrams (Fig.1.2) and mathematically through the index of evenness (Equation 1.1). Beginning with Fisher

(1941a) these studies have established that with equal fitness among the morphs, symmetrical inter-morph mating, and large population size, equal style morph ratios should be the stable equilibrium (isoplethy).

Empirical surveys have shown that style morph ratios in natural populations are frequently not equal (anisoplethy; reviewed in Heuch 1979a, b; Barrett 1993). Simulations demonstrate that anisoplethy can result from unequal fitness between morphs owing to differences in pollen production and male fertility (Barrett et al. 1983) or by asymmetrical mating including assortative mating in populations (Barrett & Hodgins 2006; Hodgins &

Barrett 2008). Anisoplethy may also occur because of stochastic processes, including drift and founder events, and non-equilibrium conditions in clonal populations with limited sexual recruitment, in which case populations may slowly approach isoplethy over much longer periods of time (Morgan & Barrett 1988; Eckert & Barrett 1995). The study of morph ratios in populations of tristylous species has provided fertile ground for investigating the relative importance of natural selection and stochastic forces in governing the maintenance of an adaptive polymorphism.

CHAPTER 1: GENERAL INTRODUCTION 13

Figure 1.2. A de Finetti diagram representing the morph-frequency space that can be occupied by populations of a tristylous species. Each symbol represents a population. A)

The population in the centre of the triangle contains equal morph ratios (isoplethy); B) An anisoplethic trimorphic population with a preponderance of M-morph individuals; C) a dimorphic population in which the S-morph is absent; D) a monomorphic population containing the L-morph only.

The index of morph evenness (E) for a tristylous populations

1−(푓(퐿)2+푓(푀)2+푓(푆)2) 1.1 퐸 = (2⁄3)

CHAPTER 1: GENERAL INTRODUCTION 14

This index is based on a combination of Simpson’s index of diversity (1949) and Nei’s formula for calculating the gene diversity at multiple loci between populations (1973). The equation uses the frequencies of the L-, M-, and S-morphs (f(L), f(M), and f(S), respectively) to produce a quantitative measure of isoplethy. The index is a continuous variable that equals one at isoplethy and zero if only one morph is present in a population

(Barrett et al. 1989).

Isoplethic and anisoplethic trimorphic populations are not the only morph structures that can occur in tristylous species. Whereas negative frequency-dependent selection is predicted to maintain the three morphs in a population, morph loss may result from in situ genetic drift in small populations (Heuch 1980), or through founder events in which the alleles determining a morph are absent in founding propagules initiating a new population (Colautti et al. 2010a). Weak morph-specific incompatibility may also cause deviations from disassortative mating and this can potentially result in morph loss through weakened negative frequency-dependent selection (Eckert & Barrett 1992).

Although it might be naively assumed that the style morphs would be lost at equal rates from isoplethic populations by stochastic processes this is in fact not the case. The probability of morph loss is governed by the genetic architecture of tristyly, specifically differences in the frequencies of allele at the S- and M-loci, and the occurrence of epistasis operating between the loci. Because the dominant S-allele is only carried by genotypes of the S-morph (SsMM, SsMm, Ssmm), its frequency is significantly lower than the remaining alleles at the two loci and therefore the S-morph is lost most frequently due to founder events or genetic drift (Heuch 1980; Heuch & Lie 1985; Barrett et al. 1989; Eckert &

Barrett 1992; Husband & Barrett 1992a). Conversely, the L-morph (ssmm) is determined

CHAPTER 1: GENERAL INTRODUCTION 15 by recessive alleles (s, m) which are at significantly higher frequencies because they can be present in M- (ssmm, ssMm) and S-morph (Ssmm, SsMm) genotypes. Therefore, the L- morph is rarely lost from populations because it can be restored by crosses between genotypes carrying recessive alleles.

As a result of the inheritance pattern for tristyly, trimorphic populations exhibit a characteristic signature of stochastic morph loss with the order of loss S->M->L-morph.

Surveys of tristylous species generally provide empirical support for these predictions

(reviewed in Barrett 1993). Therefore, in tristylous species that experience frequent population bottleneck, owing to drift and founder events associated with biological invasion, I expect to see population morph ratios that deviate from isoplethy and populations that lack the S-morph. In this thesis, I investigate the extent to which stochastic forces have played a role in affecting morph-frequency variation in an invasive plant species.

Colonization and inbreeding depression

One of the major factors that may constrain colonizing success is inbreeding depression.

Progeny resulting from selfing or biparental inbreeding are often less fit than those that result from outcrossing between unrelated individuals (Charlesworth & Charlesworth

1987; Keller & Waller 2002). Darwin (1876) initiated the first comprehensive experimental studies of inbreeding depression and today the harmful consequences of inbreeding are widely recognized. In addition, inbreeding depression is a key parameter in models of mating-system evolution in plants (Lloyd 1980; Lande & Schemske 1985).

Because small population size is a ubiquitous feature of biological invasions, it is

CHAPTER 1: GENERAL INTRODUCTION 16 important to consider the role that inbreeding depression resulting from biparental inbreeding or occasional selfing might play during colonization, particularly in predominantly outcrossing species.

Plants in small populations or those growing at low density may experience mate limitation through insufficient pollinator service. This phenomenon often results in lower fertility of open-pollinated plants in small populations compared with large populations

(Bierzychudek 1981; Larson & Barrett 1999; Waites & Ågren 2004). In introduced species, pollinators may visit flowers with which they have not co-evolved and may be unable to provide effective pollination (Barrett 1980; Devaux et al. 2014). In extreme cases where a population is founded by a single individual, selfing is the only possible means by which a population might establish (Smyth & Hamrick 1984; Jennersten 1988;

Oostermeijer et al. 1992; Terrie et al. 1992). In each of these ecological circumstances there may be selection for reproductive assurance favouring plants capable of self- fertilization. However, the offspring produced via self-fertilization may not perform as well as those that could potentially arise from cross-fertilization thus constraining population growth rate and possibly leading to local extirpation of populations. Such an effect is likely to be especially pronounced in outcrossing species because of the greater genetic load in these populations.

Self-fertilization has profound influence on the genetics of predominantly outcrossing populations. Alleles in self-fertilized offspring are more likely to be homozygous than expectations at Hardy-Weinberg equilibrium (Fisher 1941b), and increased homozygosity exposes deleterious recessive alleles that are the major cause of inbreeding depression (Charlesworth & Willis 2009). Deleterious alleles may also become

CHAPTER 1: GENERAL INTRODUCTION 17 fixed through drift in small populations and combined with the loss of any heterozygote advantage, and increased variance in quantitative characters, inbreeding may move populations away from trait optima (Lande & Schemske 1985; Charlesworth &

Charlesworth 1987).

Darwin (1876) found that species which outcross express more inbreeding depression than those which regularly self-fertilize. This pattern has been supported in dozens of subsequent experimental studies and is thought to be caused by the purging of strongly deleterious alleles in selfing populations (Barrett & Charlesworth 1991; reviewed in Crnokrak & Barrett 2002). Thus, while local inbreeding in autogamous colonists may have limited effects on fitness components, predominantly outcrossing populations may be expected to suffer greater fitness reductions after inbreeding. To my knowledge there have been no detailed investigations of inbreeding depression in outcrossing plant invaders and hence one of the goals of this thesis is to investigate this issue.

Although the genetic basis of inbreeding depression is now reasonably well understood (see Charlesworth & Willis 2009), the potentially myriad roles of environmental factors in modulating its intensity are less well studied. Adverse environmental conditions (often referred to as ‘stressful environments’), in which organism performance is reduced relative to benign environments, has been hypothesized to induce changes in the expression of inbreeding depression (Agrawal & Whitlock 2010).

However, there has been considerable debate as to whether inbreeding depression should be more intense in ‘stressful’ versus benign environments (Armbruster & Reed 2005).

Adverse conditions can arise from abiotic and biotic factors but until recently most plant studies have tended to focus on abiotic factors.

CHAPTER 1: GENERAL INTRODUCTION 18

In a pioneering study, Dudash (1990) demonstrated that inbreeding depression was expressed at different intensities in Sabatia angularis in glasshouse, garden, and field conditions. Differences in survival rates between the self- and outcrossed-progeny were most intense under field conditions and least intense in the glasshouse, suggesting that the increased importance of biotic factors under field conditions enhanced inbreeding depression, although variation among growing conditions in abiotic factors could not be completely ruled out. Similarly, inbreeding depression increased in Impatiens capensis when self-fertilized individuals were grown in mixed stands with cross-fertilized individuals (Schmitt & Ehrhardt 1990). This result was thought to have occurred because of the pre-emption of resources from less vigorous self-fertilized plants by more vigorous outcrossed plants, a phenomenon described as ‘dominance and suppression’. Finally, inbreeding depression may alter species responses to herbivore pressure, as found in

Solanum carolinense (Kariyat et al. 2011, Kariyat et al. 2013a, b). In these studies of S. carolinense, self-fertilized progeny exhibited less effective physical and chemical defences against herbivore attack than cross-fertilized individuals. A growing number of studies (reviewed in Johnson et al. 2015) indicate that biotic selection imposed by antagonists can magnify differences in fitness between selfed and outcrossed progeny with consequences for reproductive system evolution in plants. One of the objectives of this thesis is to investigate the influence of biotic factors, specifically competition, on the intensity of inbreeding depression in an outcrossing invader.

Study system: Lythrum salicaria

Lythrum salicaria (Lythraceae) is an autotetraploid, tristylous, herbaceous, perennial native to Eurasia. Plants possesses square-shaped stems with lower leaves that are either

CHAPTER 1: GENERAL INTRODUCTION 19 arranged opposite each other or are presented in whorls of three leaves, whereas upper leaves and bracts exhibit an alternate arrangement (Mal et al. 1992). The maximum height of plants is approximately 2.5 meters, but in most populations plants average 1-1.5m

(Thompson et al. 1987). Flowers are borne on cymes which emerge from axillary buds of leaf-like bracts in a terminal inflorescence which open over the course of 10-14 days

(Hickey & King 1981). Because of the showy purple-pink floral displays in L. salicaria

(Fig. 1.3), the species is easily spotted in the field and therefore populations can be readily surveyed for morph frequencies and demographic parameters.

Flowers of L. salicaria are predominantly bee-pollinated, particularly by Apis mellifera and Bombus spp., although other pollinators such as small bees, flies, butterflies, wasps and occasional hummingbirds visit flowers for pollen/and or nectar (Thompson et al. 1987; King & Sargent 2012). The main flowering time for this species in North

America is June – October (Montague et al. 2008). Of note is a pollen-colour polymorphism between the anther levels: long-level anthers produce green pollen whereas mid- and short-level anthers usually produce yellow pollen (Darwin 1877; Barlow 1923).

It has been proposed that these pollen colours may be an adaptation preventing insects from distinguishing between the floral morphs and limiting inter-morph cross-pollination

(Lunau 1996). Pollen of L. salicaria also exhibits a size polymorphism between anther levels much like other trimorphic species but without the strong degree of size trimorphism that is evident in some species (Mulcahy & Caporello 1970; Dulberger 1992).

Experimental studies exploiting the pollen-size polymorphism of L. salicaria have recently been employed to evaluate Darwin’s hypothesis (Darwin 1877) that tristyly functions to promote inter-morph legitimate cross-pollination (Costa et al. 2017). In this

CHAPTER 1: GENERAL INTRODUCTION 20 study, the authors did demonstrate that tristyly promoted inter-morph pollen flow and found support for Darwin’s hypothesis.

The maximum age of plants has not been established by demographic studies but individuals often persist for at least 12 or more years (Thompson et al. 1987; S.C.H.

Barrett pers. observ.). Most root stocks only spread for 0.5m, producing numerous flowering stems, and the species possesses no mechanisms of extensive clonal growth; reproduction is therefore entirely by seed (Yakimowski et al. 2005). Seedlings of L. salicaria can reach flowering within 8-10 weeks after germination (Shamsi & Whitehead

1974), each seed capsule produces between 80-130 seeds, and plants can produce thousands of seed capsules per year (Halkka & Halkka 1974). It has been calculated that a mature plant may produce as many as 2,700,000 seeds over its lifetime (Thompson et al.

1987). However, seed and fruit set can be negatively affected by small population size and isolation leading to a negative relation between population size and the intensity of pollen limitation of maternal fertility (Ågren 1996). Recent evidence based on flow cytometry indicates variation in ploidy in the Eurasian range: whereas most Western

European populations and all surveyed North American populations are autotetraploid, populations containing diploid, triploid, and hexaploid individuals occur in Turkey and

Israel (Kubátová et al. 2008).

Lythrum salicaria is amphibious and colonizes a wide variety of wetland habitats including low-lying marshes and pastures, stream and lake edges, and roadside ditches

(Mal et al. 1992). The species has a preference for open sites and occurs only rarely in shaded habitats. Lythrum salicaria was introduced to the mid-Eastern seaboard of North

America in the 1800s and in the intervening 150-200 years has spread northward and

CHAPTER 1: GENERAL INTRODUCTION 21 westward following corridors of human activity (Thompson et al. 1987). During the invasion of eastern North America, L. salicaria has undergone adaptive evolution forming conspicuous clines in life-history traits in response to growing season length, particularly for height and flowering time (Montague et al. 2008; Colautti & Barrett 2010, 2011, 2013;

Colautti et al. 2010b). These clines parallel those reported in the Eurasian range (Olsson &

Ågren 2002; Bastlová et al. 2004).

Extensive surveys of style morph ratios in L. salicaria populations in Europe and eastern North America indicate significant differences between the two continents. In

France and Portugal, ~95% of L. salicaria populations (n = 198 populations) are trimorphic and trimorphic populations possess a mean evenness of 0.9 (Eckert et al.

1996a; Costa et al. 2016). In contrast, a survey of populations in eastern North America (n

= 102) indicated that ~25% of populations were dimorphic and that virtually all dimorphic populations were composed of the L- and M-morph, and populations with >50 individuals had a mean evenness of 0.83 (Eckert & Barrett 1992). The North American survey revealed a similar pattern to that reported from Swedish island populations (n = 66) in which ~20% of populations lacked a floral morph, although one major distinction between

North America and Sweden is that virtually all dimorphic populations in Sweden contained only the L- and S-morph (Ågren & Ericson 1996). Genetic drift and founder events in the introduced range and more extensive gene flow in France and Portugal have been proposed as explanations for these contrasting patterns of morph-frequency variation.

CHAPTER 1: GENERAL INTRODUCTION 22

Figure 1.3. Images of Lythrum salicaria from various locations worldwide. A) The L- morph; B) Isolated plants on a beach in Estonia; C) An invasive population in a field in

Ontario; D) Inflorescences of a multi-stemmed individual in a marsh at the Koffler

Scientific Reserve. Image A courtesy of Joana Costa and Images B and D courtesy of

Spencer C.H. Barrett.

Thesis objectives

The overall objective of my thesis is to perform a comprehensive study of the reproductive ecology and genetics of an invasive plant species. My research approach involves a combination of computer simulations, population surveys, glasshouse and field

CHAPTER 1: GENERAL INTRODUCTION 23 experiments and progeny tests. In this section, I briefly outline the five research chapters in the thesis. Each chapter was written as a stand-alone publication and therefore of necessity there is some level of repetition between them. I indicate the publication or submission details for published or submitted chapters.

CHAPTER 2 - The influence of partial self-incompatibility and autotetraploidy on floral

morph frequencies in finite tristylous populations.

In tristylous species, inter-morph mating and negative frequency dependent selection are recognized as the major forces maintaining isoplethic morph ratios (Fisher 1941a; Finney

1952; Heuch 1979a). However, genetic drift, founder events, unequal reproductive output by floral morphs and asymmetrical mating among morphs can result in anisoplethic morph frequencies (Heuch 1979b; Eckert & Barrett 1992; Barrett & Hodgins 2006). Furthermore, different ploidy levels are predicted to alter the relative strength of selection and genetic drift in populations via the altered effective population sizes of alleles (Wright 1969). In this chapter, I developed computer simulations of tristylous populations with different morph-specific self-fertilization, population size and ploidy level, factors which have not been simultaneously investigated before in stochastic simulations. Morph-specific partial selfing affected allele loss asymmetrically and autotetraploid populations experienced lower frequencies of recessive allele loss than diploid populations. Selfing in the L-morph resulted in the highest loss of trimorphism, whereas selfing in the S-morph resulted in the lowest rate of loss. Finally, partial selfing by the M- and S-morphs frequently affected the identity of morphs in dimorphic populations whereas selfing in the L-morph did so only rarely. This article was submitted to the Journal of Evolutionary Biology in December

2017 with S.C.H. Barrett as co-author and is being revised following review.

CHAPTER 1: GENERAL INTRODUCTION 24

CHAPTER 3 - Stochastic processes during invasion: the influence of population size on style-morph frequency variation in Lythrum salicaria (purple loosestrife)

An earlier survey of L. salicaria from eastern North American revealed that ~25% of invasive populations lacked the S-morph (Eckert & Barrett 1992). In this study, I sampled populations in the same region of Ontario as were visited in 1992 to test whether 25 years year of ongoing invasion and increased population density across the landscape have resulted in more trimorphic populations as a result of gene flow. I found that ~25% of populations lacked the S-morph indicating that stochastic processes are still a major feature of the ongoing invasion. This chapter was published in the International Journal of

Plant Sciences (Balogh, C.M. & Barrett, S.C.H. 2016. Stochastic processes during invasion: The influence of population size on style-morph frequency variation in Lythrum salicaria (purple loosestrife). International Journal of Plant Sciences 177: 409-418).

CHAPTER 4 - Genetic and environmental influences on partial self-incompatibility in

Lythrum salicaria (Lythraceae)

In many self-incompatible species, some individuals set variable amounts of seed after self-pollination through the phenomenon of partial self-incompatibility (PSI). Despite the common occurrence of PSI in self-incompatible species, few studies have investigated the ecological and evolutionary consequences of this phenomenon (but see Good-Avila et al.

2008; Tsuchimatsu et al. 2010). In L. salicaria, PSI is especially evident in the M-morph

(Darwin 1877; Stout 1923; O’Neil 1994; Colautti et al. 2010a). In this chapter, I investigated the year-to-year repeatability and broad sense heritability of PSI, as well as the influence of environmental conditions on the expression of PSI in plants originating

CHAPTER 1: GENERAL INTRODUCTION 25 from four populations. I found that PSI expression was evident in all populations and exhibited significant morph-specific repeatability among years and between parents and offspring, but that the expression of PSI under different growing conditions remained relatively stable. This article was submitted to the International Journal of Plant Sciences in November 2017 with Spencer C.H. Barrett as a co-author and was accepted pending revision.

CHAPTER 5 - The influence of floral morph ratios and low plant density on mating and

fertility in a tristylous colonizing species

During biological invasions, individuals may experience low density and/or small population size, which may result in reduced pollinator service, mate limitation and Allee effects (Allee 1931; Ashman et al. 2004). The ability to self-fertilize, may provide an opportunity for isolated individuals to reproduce when mates are rare (Baker 1955, 1967).

However, pollinator-mediated gene flow among isolated individuals may serve to ameliorate the effects of isolation and enable inter-morph mating to be maintained. In this chapter, I estimated intra-morph mating rates from progeny morph ratios in six wild populations of L. salicaria of varying morph structure and size. I also conducted a field experiment investigating the mating and fertility of isolated individuals. I detected high rates of inter-morph mating in the M- and S-morph from wild populations and from isolated individuals in the field experiment demonstrating the importance of bee-mediated gene flow. This work was submitted to Botany in January 2018 with Spencer C.H. Barrett as a co-author.

CHAPTER 1: GENERAL INTRODUCTION 26

CHAPTER 6 - An experimental study of inbreeding depression and the effects of

competition in an invasive, perennial species.

Inbreeding depression is a key evolutionary process affecting the fitness of populations and the scope for mating-system evolution. However, few studies have investigated inbreeding depression in invasive plants, particularly those that are largely outcrossing. In this chapter, I exploited the occurrence of partial self-incompatibility in L. salicaria to compare selfed and outcrossed families in a multi-year study in the glasshouse (1 year) and field (3 years) to investigate the intensity of inbreeding depression among life-history traits and years, and cumulatively over the duration of the experiment. The experimental design included a competition treatment to allow an examination of the extent to which the breeding history of neighbours (selfed or outcrossed) influenced plant performance. I found significant inbreeding depression for key life-history traits in most years with an average cumulative inbreeding over the four years of δ = 0.48 and 0.68. However, there was no influence of the competition treatment on plant performance. The magnitude of inbreeding depression in L. salicaria appears to be sufficient to prevent the spread of self- compatibility in populations and this probably explains why tristyly is maintained in the species and no fully self-compatible populations have been reported.

CHAPTER 2

THE INFLUENCE OF PARTIAL SELF-INCOMPATIBILITY AND

AUTOTETRAPLOIDY ON FLORAL MORPH FREQUENCIES IN FINITE

TRISTYLOUS POPULATIONS

Abstract

The floral polymorphism tristyly provides opportunities for investigating the interplay between natural selection and genetic drift in finite populations. Negative frequency- dependent selection should result in 1:1:1 floral morph ratios (isoplethy) in equilibrium populations. However, a variety of deterministic and stochastic processes can result in skewed morph ratios (anisoplethy) or morph loss from trimorphic populations. The largest body of morph-ratio data for a tristylous species is for Lythrum salicaria, which maintains predominantly trimorphic populations across its native and introduced ranges, but dimorphic populations in North America and Sweden from which the short- and mid- styled morphs are absent, respectively. Here, I use a Monte Carlo sampling protocol and varying population sizes to examine the influence of two features of L. salicaria, morph- specific partial self-incompatibility and polyploidy, on the dynamics of allele loss from the loci governing tristyly and floral morph loss (L-, M-, and S-morphs) from populations.

Morph-specific partial selfing affected allele loss asymmetrically and autotetraploid populations experienced lower frequencies of recessive allele loss than diploid populations. Selfing in the L-morph resulted in the highest loss of trimorphism, whereas selfing in the S-morph resulted in the lowest rate of loss. Finally, partial selfing by the M- and S-morphs frequently affected the identity of morphs in dimorphic populations whereas selfing in the L-morph did so only rarely. My simulations indicate that morph-specific

27 CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 28 selfing and autotetraploidy interacting with population bottlenecks and founder events can help to explain observed patterns of anisoplethy and morph loss in populations of L. salicaria.

Introduction

Tristyly is a complex floral polymorphism that has evolved independently in six angiosperm families and promotes outbreeding in plant populations. Populations of tristylous species are typically comprised of three floral morphs that differ reciprocally in stigma and anther height (Darwin 1877; Ganders 1979; Barrett 1993). Because of conspicuous style-length trimorphism the floral morphs are referred to as the long-, mid-, and short-styled morphs (hereafter L-, M-, and S-morphs). In many tristylous species, floral polymorphism is associated with a physiologically-mediated trimorphic incompatibility system that prevents self- and intra-morph mating (heteromorphic self- incompatibility, Barrett & Cruzan 1994). In species with heteromorphic incompatibility, compatible pollinations occur only between stigmas and anthers of equivalent height and thus fertile unions result from crosses between floral morphs (disassortative mating). As a result of this mating system, negative frequency-dependent selection operates on floral morph frequencies giving rise to equal morph ratios (isoplethy) in equilibrium populations

(Fisher 1941a). However, a variety of deterministic and stochastic processes have the potential to cause skewed morph ratios (anisoplethy) in tristylous populations (e.g. Barrett et al. 1983, 1989; Weller 1986; Eckert & Barrett 1992, 1995; Weller et al. 2016; reviewed in Barrett 1993). Theoretical models can help to identify the potential factors causing morph-ratio bias and the loss of morphs from tristylous populations.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 29

Studies on the genetics of tristyly indicate that the most common form of diploid inheritance in the three most thoroughly investigated families (Lythraceae, Oxalidaceae and Pontederiaceae) involves two diallelic loci (S, M) with the S locus epistatic to the M locus (Fisher & Mather 1943; Weller 1976, 1992; Eckert & Barrett 1993; Gettys &

Wofford 2008). Plants carrying a dominant S-allele are of the S-morph (SsMM, SsMm,

Ssmm) and plants which are homozygous recessive at the S-locus are either of the M- morph if they carry a dominant M-allele (ssMM, ssMm), or the L-morph (ssmm) if they are homozygous recessive at both loci (Lewis & Jones, 1992). The loci can be closely linked as in Eichhornia paniculata (Barrett 1993; Arunkumar et al. 2017) and diploid species of

Oxalis section Ionoxalis (Weller 1976) or unlinked as in Pontederia cordata (Gettys &

Wofford 2008), Lythrum salicaria (Fisher & Mather 1943), and Decodon verticillatus

(Eckert & Barrett 1993). Some tristylous species including O. alpina, O. tetraphylla

(Weller 1976; Weller et al. 2016) and L. salicaria (Fisher & Mather 1943) exhibit autopolyploidy and produce complex progeny segregation ratios when parental plants are duplex or triplex at a given locus. Autopolyploid species may also possess a higher effective population size of alleles than diploids and as a result, populations should experience less genetic drift than diploid populations (Wright 1969). The unequal frequency of alleles at loci governing tristyly (see Heuch & Lie 1985), and the predicted

1:1:1 morph ratios caused by disassortative mating, make tristyly an outstanding plant sexual polymorphism for investigating the countervailing effects of stochastic processes and negative frequency-dependent selection on phenotypic evolution.

The tristylous polymorphism has inspired a variety of theoretical models that have improved understanding of the evolution and ecology of plant sexual polymorphisms. In

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 30 contrast to species with homomorphic self-incompatibility, in which extensive crossing studies and/or genotyping are required to identify the numerous mating types that occur in populations (de Nettancourt 1977; Lane & Lawrence 1993), mating-type frequencies in trimorphic populations are easy to determine visually. As a result, morph ratios can be examined with relative ease using large-scale survey data from natural populations.

Ronald Fisher carried out the first theoretical analysis of equilibrium morph ratios in tristylous populations (Fisher 1941a, 1943). He demonstrated that isoplethy could be maintained across generations by negative frequency-dependent mating. However, his analysis did not exclude the possibility of different equilibria that might arise through other hypothetical mechanisms of inheritance (and see Finney 1952, 1983; Moran 1962;

Spieth & Novitski 1969; Spieth 1971). Later, Heuch (1979a) derived a general theorem for tristylous populations and demonstrated that isoplethy was the only possible equilibrium in large populations with disassortative mating, provided that there are no fitness differences among the morphs.

More recent extensions of models of tristyly have incorporated additional features of the reproductive systems of tristylous species. These have included self-fertilization

(hereafter ‘selfing’) (Heuch 1979b), clonal propagation (Eckert & Barrett 1992, 1995), differential male fertility owing to pollen production differences among the morphs

(Barrett et al. 1983), morph-specific differences in assortative mating (Barrett et al. 2004;

Barrett & Hodgins 2006), or a combination of selfing and inbreeding depression

(Charlesworth 1979). The results of these models have often indicated that anisoplethic morph ratios and even morph loss or morph fixation occur deterministically, particularly as a result of deviations from symmetrical disassortative mating among the morphs due to

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 31 selfing and assortative mating (Barrett et al. 1989; Eckert & Barrett 1992). A key finding of deterministic models is that although isoplethic populations are a general result of disassortative mating, the frequency of alleles at the two tristyly loci at equilibrium are unequal because of dominance and epistasis (Fisher 1941a; Finney 1952, 1983). As a result, the genetic architecture of tristyly plays an important role in governing morph ratios and morph loss from finite populations because of stochastic process such as genetic drift and founder events (Heuch 1980; Morgan & Barrett 1988; Barrett et al. 1989; Husband &

Barrett 1992b; Eckert & Barrett 1992, 1995; Eckert et al. 1996a; Balogh & Barrett 2016).

Specifically, the S-morph is most frequently lost from finite populations because the dominant S-allele is present at a lower frequency than the other three alleles (M, m, and s) at equilibrium. Conversely, the L-morph is lost least often and the M-morph has an intermediate probability of loss through stochastic processes.

Among tristylous species, patterns of morph-frequency variation have been most extensively studied in Lythrum salicaria (Lythraceae). Population survey data from this species has been used to address a variety of questions on the relative importance of stochastic processes and natural selection in governing morph-frequency variation (Heuch

1979a, b; Eckert & Barrett 1992; Ågren & Ericson 1996; Eckert et al. 1996a; Balogh &

Barrett 2016; Costa et al. 2016). Lythrum salicaria is native to Eurasia but has been introduced to many other temperate regions and it has become a serious invasive species in some (e.g. North America; Thompson et al. 1987; Colautti & Barrett 2013). In common with other tristylous species with trimorphic incompatibility, large equilibrium populations of L. salicaria are expected to maintain isoplethic morph-frequencies if disassortative mating predominates and the three morphs have equivalent fitness (Heuch 1979a; Barrett

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 32

1993). However, at the northern European range edge (Sweden) and in Ontario, Canada,

~21 and ~25%, respectively, of populations surveyed contained only two floral morphs

(Eckert & Barrett 1992; Ågren & Ericson 1996; Balogh & Barrett 2016). In northern

Europe dimorphic populations are most frequently comprised of the L- and S-morphs whereas in Ontario populations most commonly lack the S-morph. In contrast, ~95% of populations in central and southern Europe are trimorphic (Heuch 1979a, b; Eckert et al.

1996a; Costa et al. 2016). These regional differences in patterns of morph-frequency variation and morph loss raise the question of what mechanism(s) are responsible.

Two specific features of L. salicaria may be important in influencing morph- frequency variation within and among populations. Controlled pollination studies in L. salicaria have revealed that populations generally exhibit partial trimorphic incompatibility with the frequency of self-compatible plants varying among the floral morphs (reviewed in Colautti et al. 2010a; Chapter 4). Specifically, incompatibility is often weakly expressed in the M-morph, whereas plants of the S-morph are usually strongly self-incompatible. It has been proposed that partial self-incompatibility may influence mating patterns and also play a role enabling self-compatible individuals to found populations (Colautti et al. 2010a; Balogh & Barrett 2016). A second feature of L. salicaria is that the species possesses both diploid and polyploid populations.

Significantly, all North American and most European populations are autotetraploid

(Kubátová et al. 2008). One of the goals of this study is to examine whether interactions between partial self-incompatibility and tetrasomic inheritance may influence allele and morph loss from finite tristylous populations.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 33

Here, I investigate through stochastic simulations the influence of partial self- incompatibility and autotetraploidy on the persistence of the alleles determining tristyly and the likelihood of morph loss from populations. Although my simulations have relevance for most non-clonal tristylous species they are motivated by specific features of the reproductive system of L. salicaria. I specifically addressed three questions. 1) How does morph-specific partial selfing affect patterns of allele loss? Selfing might be expected to result in an increased frequency of alleles determining the selfing morph and a concomitant decrease in alleles carried by outcrossing morphs. Higher levels of selfing are also expected to decrease the time to loss of those alleles. 2) What is the influence of autotetraploidy on the loss of alleles and morphs from tristylous populations with partial self-incompatibility? Earlier work indicated no differences between diploid and autotetraploid populations in morph loss using deterministic and stochastic models when rates of selfing were equivalent among the morphs (Eckert & Barrett 1992). However, here I was interested in exploring the extent to which stochasticity may affect the probability and time until allele loss from diploid and autotetraploid tristylous population when selfing rates were morph-specific. 3) What are rates of allele and morph loss in tristylous populations affected by morph-specific selfing and ploidy level? I expect that these two features will alter the dynamics of alleles in finite populations in comparison with predominantly outbreeding diploid populations and differentially affect rates of allele and morph loss. I investigated these questions using Monte Carlo sampling procedures for a range of population sizes.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 34

Material and Methods

Simulation design

I modelled the dynamics of finite diploid and autotetraploid populations using Monte-

Carlo simulations in the programming language R version 3.3.2 ‘sincere pumpkin patch’

(R Core Team 2016). I represented each population as a vector of length N, where N is the population size based upon model parameters (see below). I initiated each simulation by producing the predicted equilibrium representations of the nine possible genotypes (for the diploid case) or the 25 possible genotypes (in the autotetraploid cases) following earlier deterministic predictions (see Heuch & Lie 1985). I simulated each generation by sampling female parents, with replacement, to serve as the maternal parents and with the number of parents selected through the logistic growth function defined by growth rate r and carrying capacity k. In cases with no mate limitation (e.g. where only one floral morph persisted), populations remained at k after initiation; if a population lost two floral morphs, then the rate of selfing in the remaining morph and the r of the population interacted to produce the final population size, or population extinction occurred if rates of selfing were too low to maintain individuals. Each of the selected parents engaged in inter-morph mating or selfing at rate s. Inter-morph mating in each floral morph occurred at rate 1- s; in these cases, a selected paternal parent of a different floral morph for each maternal parent was selected. After choosing the maternal and paternal parents the simulation used diploid or autotetraploid Mendelian segregation ratios at unlinked S- and M-loci to produce offspring genotypes.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 35

Parameters in simulated populations

I simulated a variety of population sizes (N = 9, 24, 42, 72, 102, 201) which I chose based on earlier theoretical and empirical studies of tristylous species, particularly L. salicaria

(Heuch 1980; Eckert & Barrett 1992; Eckert et al. 1996a; Balogh & Barrett 2016) and started each simulation at the isoplethic equilibrium. I set the initial N and k of each population equal to the parameters I modelled and set r = 0.5 for all simulations. I performed 900 simulation runs for t = 200 generations in each size class in which I assigned s = 0 for all floral morphs. I then performed 300 simulations for t = 200 generations at each population size and ploidy level with the s value of the focal morph assigned values from 0.2 to 1.0 using increments of 0.2 while I maintained s = 0 in the remaining morphs. I sampled genotype frequencies at every fifth generation. I measured the proportion of populations in which each allele was the first allele lost along with the mean number of generations until each allele was first lost in those populations. Next, I measured the proportion of extant populations which were tristylous at generation t = 100 and 200 at each parameter level. I also reported the mean frequency of each morph in all populations and morph-frequency in trimorphic populations at generation t = 100. Finally,

I estimated the number of dimorphic populations which contained each set of possible morphs (L- and M-morphs, L- and S-morphs, and M- and S-morphs; hereafter LM-, LS-, and MS-dimorphic populations) at generation t = 100.

Analysis

I maintained statistical independence for all model intercepts by assigning 300 separate simulations where s = 0 in all morphs as the intercept response values for regressions

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 36 when each of the floral morphs was selfing. At each of these size classes, and with one morph at a time with a non-zero selfing rate, I produced saturated models with the output data predicted by (s + s2 + s3) * ploidy, and performed the “step” function, which iteratively adds and removes terms from models until a combination of terms with the lowest Bayesian Information Criterion (BIC) is found. I produced plots which showed all significant terms in the optimized model and included only terms in which the model could provide predictions (i.e. only across independent variable values where the model produced outputs). I applied a generalized linear version of this model to binomial responses, such as proportion of which populations are trimorphic vs non-trimorphic, loss of a particular allele, or allele persistence, and applied linear models to continuous response variables such as generations to allele loss or morph-frequency. I log- transformed the time to loss prior to analysis so that data would match the assumptions of normality. I also compared the proportion of trimorphic and non-trimorphic populations at generations t = 100 and t = 200 with generalized-linear models for populations with each set of parameters. In the results section I present a subset of population sizes which best illustrate overall patterns, and plots of all population sizes along with raw mean values from each model, are provided in the supplemental materials.

Results

The effect of morph-specific self-fertilization on allele and morph loss

The identity of the selfing morph strongly influenced both the number of generations to allele loss and the identity of the alleles that were lost from the loci governing tristyly.

Increased selfing in the L-, M- and S-morphs reduced the time until allele loss for the S-

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 37 and M-, S- and m-, and s-, M-, and m-alleles, respectively (Fig. 2.1, Appendix 2 Table 1).

The degree of selfing in the S-morph weakly reduced the time to loss of the M- and m- alleles and strongly reduced the time to s-allele loss. Selfing in the L- and M-morphs correlated positively with the frequency of loss of the S-allele, whereas selfing in the S- morph positively correlated with the frequency of loss of the s-allele (Fig. 2.1, Appendix 2

Table 2). Intermediate selfing rates in the L- and S-morphs resulted in the highest frequency of loss of the M-allele and intermediate selfing in the M- and S-morphs resulted in the highest frequency of loss of the m-allele. In each of these cases, high and low selfing rates caused the infrequent loss of the M- and/or m-alleles. These results demonstrate strong asymmetry in the likelihood of allele loss based upon the identity of the selfing morph in a population.

Figure 2.1. The mean number of generations required for each of the tristyly alleles to be lost from simulated populations with different parameters. A) S-allele, B) M-allele, C) m-

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 38 alleles, D) s-allele, for different population sizes (columns) and amounts of morph-specific partial selfing. Estimates were produced from back-transformed models using log- transformed values. At higher levels of selfing, alleles are generally lost more rapidly. For recessive alleles (s- and m-alleles), the number of generation to loss is slower in autotetraploid than diploid populations. In small populations, allele loss is faster overall.

When the L-morph was the selfing morph, the m-allele was lost more slowly from autotetraploid populations than from diploid populations. Similarly, selfing in the M- and

S-morph caused differential rates of m- and s-allele loss from autotetraploid and diploid populations.

Figure 2.2. The mean frequency in which simulated populations lost alleles. Rows; A): no alleles lost, B) S-allele lost, C) M-allele lost, D) m-allele lost, E) s-allele lost, for different population sizes (columns) following morph-specific partial selfing. Higher rates of

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 39 selfing generally resulted in a higher frequency of allele loss. Partial selfing in the L- and

M-morphs resulted in increased loss rates for the M- and S- and m- and S-alleles, respectively, whereas selfing in the S-morph increased the loss rates for the m-, M-, and s- alleles. Autotetraploid populations lost recessive m- and s-alleles less frequently than diploid populations and the M- and S-alleles were lost more frequently than diploid populations.

In all floral morphs the selfing rate was positively correlated with the frequency with which tristyly was lost from populations. However, selfing in the L-morph resulted in the highest rate of loss in trimorphism and selfing in the S-morph resulted in the lowest rate (Fig. 2.3, Appendix 2 Table 3). Within the models, there were significant differences at the P < 0.05 level for the frequency of non-trimorphic populations present at generations t = 100 and t = 200 for approximately half of the model parameters. A generalized linear model containing squared rate of selfing (s2), population size (N), ploidy level, and selfing morph demonstrated that the proportion of simulations with P < 0.05 was influenced by s2 and the interaction of N with s2, but not by the identity of the selfing morph or ploidy level (Table 2.1). These results indicate that the identity of the selfing morph strongly influences the frequency at which tristyly is lost and that morph loss is often ongoing after generation t = 100.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 40

Figure 2.3. The influence of generation time, rows; A) t = 100, B) t = 200, and population size (columns) on the evolution of non-trimorphic simulated populations with different amounts of morph-specific selfing. Smaller populations were non-trimorphic more frequently than larger populations and higher rates of selfing resulted in a higher frequency of non-trimorphic population. Selfing in the L-morph resulted in the highest frequency of non-trimorphic populations whereas non-trimorphic populations were least common when the S-morph selfed. In approximately half of the model parameters, there are more non-trimorphic populations at generation t = 200 than at t = 100 indicating that the process of morph loss is often ongoing.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 41

Table 2.1. Effects of simulated model terms on the outcome of models. Terms investigated were ploidy, selfing morph identity (L-, M-, S-morph), population size (N), and the selfing rate (s) on the proportion of model parameters which are significant.

Likelihood ratio P- Parameter df Χ2 value

Ploidy 0.230 1 0.631

Morph 1.594 2 0.451

N 3.521 1 0.061

s2 8.850 1 0.003*

Ploidy:Morph 0.552 2 0.759

Ploidy:N 0.142 1 0.707

Morph:N 1.414 2 0.493

Ploidy:s2 0.649 1 0.420

Morph:s2 0.724 2 0.696

N:s2 4.769 1 0.029*

Ploidy:Morph:N 0.440 2 0.803

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 42

Ploidy:Morph:s2 0.190 2 0.909

Ploidy:N:s2 0.315 1 0.575

Morph:N:s2 0.323 2 0.851

Ploidy:Morph:N:s2 0.012 2 0.994

The identity of the selfing morph influenced the morph structure of dimorphic populations. The selfing rate of the L-morph was negatively correlated with the frequency of MS-dimorphic populations when population size was small (N<42) but at larger population sizes had no significant influence on the frequencies of dimorphic population morph structures (Fig. 2.4, Appendix 2 Table 4). In contrast, the selfing rate of the M- morph positively influenced the frequency of LM-dimorphic populations and negatively influenced the frequency of LS-dimorphic populations. Increased rates of selfing in the S- morph decreased frequencies of LM-dimorphic populations and increased the frequency of

LS- and MS-dimorphic populations. These results indicate a strong influence of the selfing rate of the M- and S-morphs and a weak influence of the selfing rate of the L-morph on the morph composition of dimorphic populations.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 43

Figure 2.4. The influence of population size (columns) and frequency of morph-specific partial selfing on the evolution of dimorphic population structures. Rows; frequencies of

A) LM-, B) LS-, and C) MS-dimorphic populations formed from trimorphic populations.

Partial selfing in the M-morph increased the frequency of LM-dimorphic populations, selfing in the S-morph increased the frequency of LS-dimorphic populations, and selfing in the L- and M-morphs produced increased frequencies of MS-dimorphic populations.

Autotetraploid populations evolved into MS-dimorphic populations less often than diploid populations when the S-morph was the selfing morph. Selfing in the L-morph only weakly affected the frequency of dimorphic population structures.

The effects of ploidy level on allele and morph loss

The number of generations to allele loss and the frequency of allele loss from the tristyly loci differed significantly between diploid and autotetraploid populations. This relation strongly depended on the identity of the selfing morph and allele dominance. When the M-

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 44 or the S-morph was the selfing morph in a population, the recessive m-allele persisted for a shorter time in diploid than autotetraploid populations. When the S-morph was the selfing morph the time to loss of the recessive s-allele was shorter in diploid than autotetraploid populations (Fig. 2.1, Appendix 2 Table 1). There were no consistent differences between ploidy levels in the number of generations taken for loss of alleles at the tristyly loci when the L-morph was the selfing morph. The frequency of allele loss also differed between ploidy levels depending on the selfing morph and dominance of particular alleles. When the M- or the S-morph was the selfing morph within a population, the dominant S- and M-alleles, respectively, experienced a higher rate of loss from autotetraploid than diploid populations (Fig. 2.2, Appendix 2 Table 2). Alternatively, the recessive m-allele was lost more frequently from diploid than autotetraploid populations when the M- and S-morphs selfed, and the recessive s-allele was lost more frequently from diploid than autotetraploid populations when the S-morph selfed. There were no differences between ploidy levels in the frequency of allele loss when the L-morph was the selfing morph. These results demonstrate that the effects of ploidy depend on the identity of the selfing morph and the dominance of alleles at the tristyly loci, with dominant alleles more often lost from autotetraploid populations and recessive alleles lost more often from diploid populations.

The number of generations in which tristyly was lost from populations differed between diploid and autotetraploid populations with partial self-incompatibility and this was influenced by which morph was partially selfing. Trimorphism was lost in autotetraploid populations in which M- and S-morph plants selfed more often than diploid populations (Fig. 2.3, Appendix 2 Table 3). When the L-morph partially selfed, however,

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 45 ploidy did not influence the number of generations in which trimorphism was lost. This indicates that the identity of the partially selfing morph plays a key role in determining whether or not ploidy level affects the loss of tristyly.

The types of dimorphic populations that originated through morph loss from tristylous populations also differed between ploidy levels. When the S-morph was partially selfing, diploid populations evolved MS-dimorphism more often than autotetraploid populations and autotetraploid populations evolved LS-dimorphism more often than diploid populations (Fig. 2.4, Appendix 2 Table 4). In contrast, when the M-morph was the selfing morph, autotetraploid populations were more likely to evolve LM-dimorphism than diploid populations and diploid populations were more likely to evolve MS- dimorphism than autotetraploid populations. However partial selfing in the L-morph did not produce differences in the frequency of dimorphic population structures between ploidy levels. These results further support my earlier findings that the identity of the partially selfing morph is crucial in determining whether or not diploid and autotetraploid populations experience morph loss differently.

Interactions between self-fertilization and ploidy on allele and morph loss

Interactions between the selfing rate of a morph and ploidy level affected the number of generations to loss and the frequency of allele loss at the tristyly loci. Increasing the selfing rate in the S-morph decreased differences in the number of generations until loss of the dominant M-allele between diploid and autotetraploid populations (Fig. 2.1, Appendix

2 Table 1). Conversely, increased selfing rates in the S-morph increased the differences between ploidy levels in the number of generations to loss of the dominant S- and

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 46 recessive s-alleles. Diploid populations in which the L-morph selfed at a low rate lost the dominant S-allele more slowly and the dominant M-allele more rapidly than autotetraploid populations; however, at higher rates of selfing, diploid populations lost the dominant S- allele more rapidly and the dominant M-allele more slowly than in autotetraploid populations. In populations where the S-morph selfed, the frequency of dominant M-allele loss was greater between diploid and autotetraploid populations at higher selfing rates whereas the difference between ploidy levels in the frequency of recessive m-allele loss was highest at intermediate selfing rates (Fig. 2.2, Appendix 2 Table 2). When the M- morph was the selfing morph, intermediate rates of selfing resulted in the greatest difference between ploidy levels in the frequency of loss of dominant S- and recessive m- alleles. The selfing rate of the L-morph did not consistently interact with ploidy level to affect the frequency of allele loss from the tristyly loci. Once again, the identity of the selfing morph is important in defining the interaction between ploidy and selfing rate in the number of generations to allele loss and frequency of allele loss in tristylous populations.

Despite the importance of interactions between selfing rate and ploidy level on allele loss, the joint influence of these parameters had relatively minor influences on the number of generations to the loss of trimorphism from populations. Indeed, there was no interaction between ploidy level and selfing rates on the number of generations until the loss of trimorphism but only on the types of dimorphic morph structures that evolved (Fig.

2.3, Appendix 2 Table 3). When the S-morph was the selfing morph, the difference between ploidy levels in the frequency of LS-dimorphic populations increased as selfing rate increased. Similarly, increased selfing in the M-morph resulted in a greater difference

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 47 between diploid and autotetraploid populations in the frequency of LM-dimorphic populations (Fig. 2.4, Appendix 2 Table 4). However, partial selfing in the L-morph did not result in significant interactions between selfing rate and ploidy in affecting the frequency of dimorphic population types. These results indicate that ploidy and selfing rate do not exhibit interactions that influence the number of generations until tristyly is lost. The two parameters only interact when the M- and S-morphs self and primarily influence the frequency of dimorphic population structures. Such an effect was not evident when the L-morph was the partially selfing morph.

Discussion

My theoretical studies of morph-frequency dynamics in tristylous populations with variation in selfing rate and ploidy level have revealed several novel findings. Regardless of ploidy level, the number of generations before allele loss and the frequency of allele loss from the two loci governing tristyly were asymmetric among the morphs because of the genetic architecture of tristyly (Figs. 2.1 and 2.2). With selfing in the L- and M- morphs, the timing and frequency of allele loss affected two alleles (S and M in the L- morph and S and m in the M-morph), whereas three alleles (M, m, s) were similarly affected in the S-morph. Autotetraploid populations lost recessive alleles more slowly than diploid populations, with changes in the frequency of allele loss between ploidy levels only apparent after morph-specific selfing (Figs. 2.1 and 2.2). Selfing in the L-morph resulted in the highest probability of morph loss whereas trimorphism was most often maintained with selfing in the S-morph (Fig. 2.3). The identity of the selfing morph altered the frequency in which different dimorphic population types evolved by morph loss. In the

M- and S-morphs, LM- and LS-dimorphic populations, respectively, were most common,

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 48 whereas selfing in the L-morph had a very minor influence on the relative frequencies of dimorphic population structures (Fig. 2.4). Below I contrast my findings with earlier theoretical work on the tristylous genetic polymorphism and identify patterns in the large body of empirical data on morph-frequency variation in L. salicaria that can be explained by my results.

Comparison of results with earlier models

My simulations confirmed several results from earlier models of tristyly, but also provided new insights concerning the effects of partial selfing on morph and genotype equilibria.

Without selfing in my models, I obtained very similar frequencies of allele loss to earlier studies of small (N≥30) tristylous populations (see Heuch 1980; Appendix 2 Table 5).

Furthermore, when I simulated larger populations (e.g. N=201) with no selfing, genotype frequencies conformed to earlier predictions of deterministic equilibria (Fisher 1941a;

Heuch & Lie 1985; Appendix 2 Table 6). My models differed from earlier theoretical work on tristylous systems by combining deterministic parameters such as morph-specific partial selfing and intermorph mating in populations of variable size but additionally I considered the role of genetic drift on morph-frequency dynamics. Significantly, I also monitored allele and morph loss over generations providing novel results on morph and allele dynamics in non-equilibrium populations. Previous models only considered morph and/or allele frequencies at the end point of the simulations conducted. I was interested in investigating dynamics in non-equilibrium populations because my earlier studies indicated that many invasive populations of L. salicaria were small and likely of a young age due to ongoing invasion (Balogh & Barrett 2016). I detected the asymmetric loss of alleles and differences in the number of generations for alleles to be lost from populations.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 49

My incorporation of morph-specific partial selfing builds on the deterministic results obtained by Heuch (1979b) and provides new information on the stochastic loss of alleles from populations under non-equilibrium conditions. My study is therefore novel in considering the likely influence of the colonization process on allele and morph-frequency dynamics in tristylous populations.

Population genetic models of autopolyploid species predict that because populations with tetrasomic inheritance possess a larger number of heterozygous states than diploid populations, genotype frequency equilibrium is reached more slowly in autotetraploid than diploid populations (Bever & Felber 1992). Thus, alleles more frequently found in a homozygous state (e.g. m, s alleles) should be lost more slowly in autotetraploid than diploid populations owing to the effects of finite population size.

Earlier models have investigated differences in allele loss in diploid and autopolyploid tristylous populations and have found no differences in the frequency with which the tristyly alleles are lost (Heuch 1980; Eckert & Barrett 1992). However, with morph- specific selfing I found clear evidence of asymmetries in the frequencies in which specific alleles were lost from tristylous populations. Both diploid and autotetraploid populations exhibited these asymmetries but the number of generations required (Fig. 2.1) and the frequency (Fig. 2.2) of allele loss differed between ploidy levels. Therefore, differences in the effective population size of alleles at the tristyly loci between diploid and autotetraploid populations (see Wright 1969) are accentuated in populations with morph- specific selfing. Based on the geographical distribution of cytolyses in L. salicaria (see

Kubátová et al. 2008) it would appear that virtually all populations that have been sampled for floral morph ratios (see Table 2.2) are autotetraploid. My models predicted that with a

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 50 nonzero rate of morph-specific selfing there should be differences between diploid and tetraploid populations in the types of dimorphic populations that evolve by allele and morph loss. Future surveys of morph-frequency variation in L. salicaria should include diploid populations to investigate these predictions.

Although my models provided insight into the consequences of morph-specific partial selfing and ploidy in tristylous populations, they did not consider several factors that could influence morph-frequency variation. The absence of inbreeding depression in my models, and also in those of Heuch (1979b), resulted in the expected increase in the frequency of the selfing morph (Appendix 2 Fig. 1, Appendix 2 Tables 7 and 8), as predicted by Fisher’s (1941b) automatic selection hypothesis. However, experimental studies on inbreeding depression in L. salicaria (O’Neil 1994; Chapter 6) have revealed some fitness differences between selfed and outcrossed progeny, and thus future work should include this parameter. Several earlier models of tristyly (e.g. Charlesworth 1979;

Barrett et al. 1983; Heuch & Lie 1985) included selfing and inbreeding depression and indicated that fitness differences between progeny types can influence population morph ratios at equilibrium.

My models also did not include variation in maternal fertility (seed set) owing to pollen limitation and this might also be valuable to explore since there is evidence of pollen limitation of seed set in L. salicaria. For example, pollen limitation varied with population size in L. salicaria in Sweden with small population experiencing the most reduced seed set (Ågren 1996). Moreover, evidence for morph-specific pollen limitation involving the L-morph was reported in range margins populations from northern Sweden

(Ågren & Ericson 1996). In a model of the evolution of morph ratios in tristylous

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 51

Narcissus triandrus, variation in maternal fertility influenced morph-frequencies, particularly when deviations from symmetrical disassortative mating occurred (Barrett &

Hodgins 2006). To what extent differences in fertility of the morphs in L. salicaria might influence their frequency in finite populations with partial selfing requires further theoretical work.

Finally, my models only considered tristylous populations in which a single floral morph selfed, with the remaining morphs engaged in disassortative mating. This approach was pursued for two reasons. First, there is empirical evidence for morph-specific differences in partial self-incompatibility in tristylous species (reviewed in Barrett &

Cruzan 1994), including L. salicaria in which plants of the M-morph are often self- compatible whereas those of the S-morph are usually strongly self-incompatible (Colautti et al. 2010a; Chapter 4). I wanted to explore how such differences may potentially influence the loss of alleles and floral morphs from populations. Second, although introducing variable selfing rates among the floral morphs to models of tristyly would certainly be worthwhile, the range of potential outcomes on morph-frequency dynamics given the complex genetic system governing tristyly were considered to be beyond the scope of the present study.

Relevance to natural populations

Several of the model results that I obtained help to explain observed patterns of morph- frequency variation obtained from field surveys of L. salicaria populations. During the past century and a half at least 83885 individuals of L. salicaria have been scored for floral morph from a total of 557 populations in the native and introduced ranges of the

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 52 species (Darwin 1864; East 1932; Haldane 1936; Schoch-Bodmer 1938; Fisher & Mather

1943; Høeg 1944; Halkka & Halkka 1974; Heuch 1979a; Gilbert & Lee 1980; Eckert &

Barrett 1992; Andersson 1994; Anderson & Ascher 1995; Ågren & Ericson 1996; Eckert et al. 1996a; Balogh & Barrett 2016; Costa et al. 2016; summarized in Table 2.2). This impressive body of morph-frequency data is the most extensive for any heterostylous species. Almost two-thirds of the individuals of L. salicaria sampled were from European populations and most of the remaining one-third was from Ontario. To date only18 populations have been sampled elsewhere on the North American continent (Gilbert &

Lee 1980; Anderson & Ascher 1995), despite the prevalence of L. salicaria across the

USA and Canada (Thompson et al. 1987), therefore there is considerable geographical scope for additional sampling.

A fundamental question about tristylous species is whether or not populations conform to the classic Fisherian isoplethic equilibrium of equal morph-frequencies. Many populations in both the native and introduced range of L. salicaria do indeed conform to this expectation, but average morph-frequencies from large samples of populations almost always deviate from isoplethy (Table 2.2). In most cases the L-morph is the most frequent morph, although it is not clear what mechanism(s) are responsible for its small numerical superiority (see Heuch 1979a). My models indicated that with L-morph specific selfing

(and no inbreeding depression) the L-morph would experience an elevated frequency in populations. However, controlled pollination studies suggest this may be unlikely as the

M-morph exhibits the weakest self-incompatibility system not the L-morph (Waites &

Ågren 2006; Colautti et al. 2010a; Chapter 4). Nevertheless, eight populations monomorphic for the L-morph have been recently been reported from Ontario, Canada

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 53

(Balogh & Barrett 2016) and progeny tests of open-pollinated families from one of these populations was almost exclusively comprised of L-morph plants, a pattern consistent with high rates of selfing and/or intramorph mating.

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 54

Table 2.2. Floral morph frequencies (L-, M-, S-morph), number of sampled populations and plants (N), and number of dimorphic (Nd) and monomorphic (Nm) populations from surveys of Lythrum salicaria conducted over the past 150 years.

Weighted averages were calculated based upon estimated population sizes. Authors of surveys from which data were gathered are listed on the far right. The method that each author used to present their data is displayed next to the morph frequencies: A = pooled, B = weighted mean, C = trimorphic only, D = raw count.

Floral morph N N Location L M S Nd Nm Author Year pops plants

Native Range

Finland 0.341 0.364 0.294 A* 21 10802 0 0 Kuusvuori 1960

Halkka & Finland 0.361 0.331 0.307 B 16 1823 0 0 1974 Halkka

France 0.360 0.330 0.310 B 102 14621 5 0 Eckert et al. 1996

Germany 0.401 0.381 0.218 B 2 501 0 0 von Ubisch 1925

Germany 0.389 0.317 0.294 C 1 1260 0 0 Heuch 1979

Iberian 0.378 0.333 0.290 B† 96 NA 4 1 Costa et al. 2016

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 55

Peninsula

Norway 0.326 0.340 0.334 B 2 1209 0 0 Høeg 1944

Sweden 0.373 0.354 0.272 D 1 2001 0 0 Andersson 1994

Ågren & Sweden 0.388 0.243 0.363 B 66 10498 14 0 1996 Erickson

Schoch- Switzerland 0.364 0.330 0.306 B 8 6169 0 0 1938 Bodmer

UK 0.377 0.344 0.280 B 2 393 0 0 Darwin 1864

UK 0.362 0.318 0.320 B 5 2365 0 0 Haldane 1936

Fisher & UK 0.339 0.252 0.409 D 1 301 0 0 1943 Mather

Regional Total 0.366 0.323 0.310 323 51943 23 1

Adventive Range

Gilbert & Australia 0.439 0.329 0.232 D 1 82 0 0 1980 Lee

CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY 56

Gilbert & Canada, BC 0.402 0.488 0.110 D 1 82 0 0 1980 Lee

Eckert & Canada, ON 0.414 0.238 0.348 B 102 15415 20 3 1992 Barrett

Balogh & Canada, ON 0.322 0.362 0.316 B 114 4038 21 9 2016 Barrett

USA, MA 0.329 0.371 0.300 D 1 407 0 0 East 1934

Anderson & USA, MN 0.348 0.338 0.314 B 16 11918 2 3 1995 Ascher

Regional Total 0.391 0.271 0.338 235 31942 43 15

Grand total 0.379 0.296 0.324 558 83885 66 16

* As cited in Halkka and Halkka 1974; † Number of plants sampled not reported

57 CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY

The lower average weighted mean frequency of the M-morph in invasive Ontario populations (L-morph = 0.39, M- morph = 0.27, S-morph = 0.34; N = 235 populations) could possibly be explained by higher selfing and inbreeding depression of this morph because of its weaker self-incompatibility system. Weber et al. (2013) predicted that the low frequency and ultimate loss of the M-morph in tristylous populations of Oxalis alpina may result from morph-specific selfing and inbreeding depression and provided evidence for this in some populations. . Estimates of selfing rates in invasive populations of L. salicaria in Ontario would be useful in evaluating whether selfing rates play any role in either increasing (e.g. L-morph) or decreasing (M-morph) particular floral morph frequencies.

My simulation results have relevance for the evolution of dimorphic from trimorphic populations in L. salicaria. Survey data from Europe indicates that with the exception of populations sampled by Ågren & Ericson (1996) at an archipelago at the range margin in northern Sweden, of which 14 were dimorphic, most other surveys recorded no dimorphic populations or a very small number (Table 2.2). In contrast, two large-scale surveys in Ontario revealed a much higher level of dimorphism with approximately 19% of the 216 populations surveyed comprised of two morphs (Eckert &

Barrett 1992; Balogh & Barrett 2016). In Sweden 71% of populations were LS-dimorphic and in Ontario 67% of the populations were LM-dimorphic. My model indicates that with partial selfing of the M-morph, populations missing the S-morph are most commonly produced (Fig. 2.4). With increasing amounts of selfing the probability of S-allele loss increases and is always substantially higher than m-allele loss, which peaks at intermediate levels of selfing (Fig. 2.2). Thus, selfing in the M-morph combined with stochastic forces

58 CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY such as founder events and drift (see Eckert & Barrett 1992; Balogh & Barrett 2016) can potentially explain the observed patterns of LM-dimorphism in Ontario.

It seems unlikely based on my models that the occurrence of LS-dimorphism in northern Sweden is associated with morph-specific selfing because populations of this particular morph structure most commonly resulted from high selfing of the S-morph.

However, all controlled pollination studies on L. salicaria have indicated that the S-morph has the strongest self-incompatibility system (O’Neil 1994; Mal et al. 1999; Waites &

Ågren 2006; Colautti et al. 2010a; Chapter 4), a pattern also seen in other tristylous species (reviewed in Barrett & Cruzan 1994). It also seems unlikely that the absence of this morph from populations in northern Sweden results from morph-specific selfing and inbreeding depression. Ågren & Ericson (1996) found no consistent differences in the fitness of open-pollinated families of the floral morphs of L. salicaria from this region.

The causes of the pattern of LS-dimorphism in Sweden are unclear but may be associated with the significantly higher frequency of S-morph genotypes carrying recessive m versus dominant M alleles and thus the higher frequency of the m than the M allele carried by this morph (see Fisher 1941a; Heuch & Lie 1985). Contingencies associated with colonization may increase the probability of S-morph genotypes with recessive alleles being represented among the founders of new populations. This process would be exacerbated if propagules descended from these populations then initiate new populations (and see Ågren

& Ericson 1996). In colonizing populations of L. salicaria both stochastic and deterministic processes probably interact to shape morph-frequency variation.

Earlier surveys of morph-frequency variation in L. salicaria populations have often included relatively large population (N>100 plants) and may have avoided small

59 CHAPTER 2: SIMULATION OF PARTIAL SELF-INCOMPATIBILITY populations (N≤50). In the future, it would be useful in assessing the morph structure of dimorphic populations to increase the sampling of small populations in order to evaluate my model results. These models and those of Heuch (1980) and Barrett et al. (1989) indicate that morph loss will only occur from tristylous populations with <30-40 individuals, as long as populations maintain high levels of disassortative mating. Indeed, the few surveys that have included a significant number of small populations support these results and indicate that tristyly is often maintained in surprisingly small populations

(Halkka & Halkka 1974; Eckert et al. 1996a; Balogh & Barrett 2016; Costa et al. 2016).

Despite the colonizing life history of L. salicaria, tristyly appears to be remarkably stable to both the deterministic and stochastic processes that are known to cause the destabilization of the polymorphism in other tristylous species (reviewed in Weller 1992;

Barrett 1993). Future surveys of floral-morph representation in L. salicaria populations accompanying the world-wide spread of the species would be useful for testing models of morph-frequency dynamics, especially in small non-equilibrium populations.

Supplementary material for this chapter can be found in appendix 2, pp. 243-275.

CHAPTER 3

STOCHASTIC PROCESSES DURING INVASION: THE INFLUENCE OF

POPULATION SIZE ON STYLE-MORPH FREQUENCY VARIATION IN LYTHRUM

SALICARIA (PURPLE LOOSESTRIFE)

Abstract

During biological invasion, the genetic diversity of populations may be reduced by founder events and genetic drift. The floral polymorphism tristyly provides an exceptional opportunity to investigate the influence of stochastic forces on the maintenance of genetic polymorphism because small population size leads to a characteristic signature of morph loss from populations. Here, I investigate the relations between population size and morph-frequency variation in invasive populations of tristylous L. salicaria in Ontario,

Canada. I also compare my results to a similar survey conducted 25 years ago in the same region. I surveyed the size and morph ratios of 114 L. salicaria populations in 2013. I calculated the relations between population size and style morph absence, population size and style morph evenness, and the number of populations lacking particular style morphs.

For comparison of the patterns of morph-frequency variation between surveys, I used a sample of populations (1988/9: 51 populations; 2013: 101 populations) with similar size distributions. My survey confirmed that smaller populations were more likely to lack a style morph than larger populations and morph ratios were less even in smaller populations. In dimorphic populations, the S-morph was absent most often and the L- morph was least often missing, a pattern consistent with the stochastic theory of asymmetrical morph loss for tristylous species. There were no significant differences between the 1988/9 and 2013 surveys in the frequency of populations missing style

60 CHAPTER 3: POPULATION SURVEY 61 morphs, or the relations between population size and style morph evenness. Despite an increase in abundance of L. salicaria in Ontario during the past 25 years, genetic drift and founder events still play a dominant role in governing patterns of morph-frequency variation.

Introduction

Founder events and genetic drift are pervasive forces during the repeated episodes of colonization that characterize biological invasions. These stochastic processes have the potential to both limit and reduce genetic diversity in colonizing populations. The diversity of genetically based mating types in small populations of self-incompatible species is especially relevant because when mating types are lost through founder events and genetic drift, populations may experience pollen limitation of fertility or fail to reproduce sexually

(Allee 1931; Byers & Meagher 1992; Vekemans et al. 1998; Ashman et al. 2004; Young

& Pickup 2010; Barrett 2011). Though mate limitation may favour self-fertile individuals during establishment following long-distance dispersal (Baker 1955; Stebbins 1957), to date there is mixed evidence that self-fertilization is commonly selected to relieve mate limitation during biological invasion (Sutherland 2004; van Kleunen & Johnson 2007;

Pannell 2015). Other landscape level processes may aid in overcoming mate limitation, particularly as invasions mature in a region and populations become more abundant and less reproductively isolated. The increased connectivity resulting from pollen and seed dispersal may often serve to restore missing mating types and increase the genetic diversity of populations. Evidence for temporal changes in genetic diversity as biological invasions mature has not been investigated in colonizing species.

CHAPTER 3: POPULATION SURVEY 62

The genetic polymorphism tristyly provides a valuable system for investigating the role of stochastic processes in colonizing populations. Populations of tristylous species are generally composed of the three mating types or morphs (long-, mid-, and short-styled morphs; hereafter L-, M-, and S-morphs) distinguished primarily by style length and the position of anthers within a flower (Darwin 1877; Ganders 1979; Barrett 1993). Floral trimorphism is commonly associated with a trimorphic incompatibility system in which the style morphs are self- and intramorph-incompatible; the only compatible pollinations are intermorph and occur between anthers and stigmas of equivalent height. Large tristylous populations at equilibrium are expected to contain equal frequencies of the style morphs (isoplethy) when all three morphs are completely outcrossing and possess equal fitness (Fisher 1941a; Fisher and Mather 1943; Heuch 1979a, b). The principal mechanism maintaining 1:1:1 morph ratios in tristylous populations is negative frequency-dependent selection resulting from phenotypic disassortative mating (Barrett et al. 1987; Eckert et al.

1996b). However, surveys of style morph frequencies in tristylous species frequently indicate deviations from this equilibrium expectation. Biased morph ratios (anisoplethy) or morph absence from populations of tristylous species can result from a variety of stochastic (e.g. Barrett et al. 1989; Husband & Barrett 1992b; Eckert et al. 1996a) and deterministic forces (e.g. Barrett et al. 1983; Weller 1986; Barrett et al. 2004). Colonizing events and periods of small population size are commonly associated with stochastic processes and therefore successful tristylous invaders provide useful model systems for investigating the influence of demographic factors on morph-frequency variation.

Over the past century, purple loosestrife (Lythrum salicaria, Lythraceae), a tristylous wetland perennial, has become one of the most widespread invasive species in

CHAPTER 3: POPULATION SURVEY 63

North America. The species is native to Eurasia but was introduced to the eastern seaboard during the past 150 years, spreading over much of the northern regions of eastern North

America and more recently to central and western North America (Thompson et al. 1987;

Mal et al. 1992; Colautti & Barrett 2013). Studies of L. salicaria have a long and venerable history beginning with the seminal work of Darwin (1864, 1877) who characterized the general morphological features of tristyly and performed controlled pollinations to establish the compatibility relationships of style morphs. Later, Fisher and

Mather (1943) determined the genetic basis of tristyly in L. salicaria through controlled crosses, and subsequent surveys of style morph frequencies in European populations found general support for Fisher’s theoretical prediction of 1:1:1 style morph ratios in populations at equilibrium (Haldane 1936; Schoch-Bodmer 1938; Halkka & Halkka 1974;

Ågren & Ericson 1996). More recently, investigations of invasive L. salicaria populations in eastern North America have sought to determine the influence of finite population size on morph-frequency variation (Eckert & Barrett 1992), and to compare morph frequency variation in the introduced versus native range of the species (Eckert et al. 1996a). These surveys indicate striking differences between continents, with patterns of variation in introduced populations consistent with those expected if stochastic processes were playing a more important role in governing variation in style-morph frequencies in comparison with native populations.

A particularly attractive feature of the tristylous polymorphism is that theoretical studies demonstrate that stochastic processes produce a characteristic signature of asymmetric morph loss because of the inheritance of tristyly (Heuch 1980; Barrett et al.

1989; Eckert & Barrett 1992). Genetic studies in the three most widely studied families

CHAPTER 3: POPULATION SURVEY 64

(Lythraceae, Oxalidaceae, Pontederiaceae) indicate that the polymorphism is governed by two diallelic loci (S, M) with the S locus epistatic to the M locus (Fisher & Mather 1943;

Weller 1976; Lewis & Jones 1992; Gettys & Wofford 2008). Because the dominant S- allele that governs the expression of short styles is only carried by the S-morph (S---), it is at a lower frequency in equilibrium populations than the remaining other three alleles (s,

M, m) at the loci governing tristyly. As a result, when all morphs possess equal fitness, the

S-morph is more susceptible to stochastic loss from populations than the other style morphs. In contrast, the L-morph (genotype ssmm) should be rarely lost from populations because the recessive s and m alleles are commonly present in genotypes of all three style morphs and therefore exist at higher frequencies in equilibrium populations (Heuch & Lie

1985). Thus, if genetic bottlenecks and drift play a prominent role in the population biology of a tristylous species, as a result of repeated colonizing events, there should be a characteristic pattern in which the S-morph is absent most often, the L-morph is rarely absent, and the M-morph is absent at a rate intermediate between the S- and L-morphs

(e.g. probability of morph absence: S-morph > M-morph > L-morph). Empirical support for these theoretical predictions has come from extensive surveys of invasive Lythrum salicaria populations in Ontario, Canada (Eckert & Barrett 1992, Eckert et al. 1996a) and in populations of the annual colonizer Eichhornia paniculata (Barrett et al. 1989) in Brazil and the Caribbean. Both species exhibit dimorphic populations that are predominantly composed of the L- and M-morphs.

Here, I measure style-morph frequency variation in L. salicaria populations in

Ontario, Canada in relation to their size. I asked a series of questions related to the influence of stochastic processes on morph evenness and morph loss from populations: 1)

CHAPTER 3: POPULATION SURVEY 65

Is morph absence in populations related to their size? I predicted that style morphs would be missing from small populations more often than large populations. 2) Does the evenness of morph ratios change with population size? I predicted greater morph evenness in larger populations than small populations. 3) Among the three style morphs, is the S- morph most commonly absent from non-trimorphic populations, as predicted because of stochastic loss of the S-allele governing the phenotype of this morph? 4) Are there differences in the overall patterns of morph-frequency variation in comparisons between the 1988/9 and 2013 surveys? Lythrum salicaria has increased in abundance and expanded its range dramatically in Ontario since Eckert and Barrett (1992) measured morph frequencies ~25 years ago (Thompson et al. 1987; Colautti & Barrett 2013; S.C.H. Barrett unpubl observ). The change in abundance led us to hypothesize that there may be a greater connectivity among populations and more opportunities for gene flow than in 1988/9 and this could result in a reduced frequency of non-trimorphic populations and greater evenness in morph frequencies.

Materials and Methods

Study species

Lythrum salicaria is an outcrossing, showy, insect-pollinated herb that was introduced to eastern North America from its native Eurasian range during the late eighteenth century

(Thompson et al. 1987). The species possesses a trimorphic incompatibility system limiting opportunities for self and intramorph mating; however, occasional individuals, particularly of the M-morph, are pseudo-self-compatible (PSC) enabling fertile seed to be produced from self- and intra-morph mating (Colautti et al. 2010a). Lythrum salicaria

CHAPTER 3: POPULATION SURVEY 66 colonizes a variety of wetland habitats, particularly roadside ditches, marshes, and low- lying pastures, exclusively through seed dispersal (Mal et al. 1992; Yakimowski et al.

2005). Seedlings of L. salicaria can reach flowering within 8-10 weeks after germination

(Shamsi & Whitehead 1974) and individuals are perennial and can survive up to 12 years and perhaps longer (S.C.H. Barrett pers. observ.). The species forms an extensive seed bank which allows populations to regenerate quickly after disturbance (Yakimowski et al.

2005). Herbarium specimens indicate population expansion north and south along the eastern North American seaboard and northwest into southern Ontario during the twentieth century (Thompson et al. 1987; Blossey et al. 2001). Subsequently, rapid northward expansion of the geographical range of L. salicaria in Ontario has occurred over the last

50 years (Colautti & Barrett 2013; S.C.H. Barrett pers. observ). The species is commonly associated with active transportation corridors and human settlement, with most of the dispersal northwards in Ontario associated with human activities. In Ontario, populations flower between late June and mid-September, with peak flowering in most populations occurring in early to mid-August (Montague et al. 2008).

Population surveys

In 2013 I estimated style morph frequencies in 114 populations of L. salicaria in Ontario

(Fig. 3.1). My sampling area largely overlapped the geographic region sampled by Eckert and Barrett (1992) in their 1988/9 survey of 102 populations; however, it was not possible to resample the specific populations investigated by Eckert and Barrett (1992) because geographic coordinates were not available and many easily identified populations were extirpated. I defined a population as a group of three or more plants separated from other such groups by at least ~0.5 km; the vast majority of populations that I sampled were

CHAPTER 3: POPULATION SURVEY 67 separated by more than 1 km. Because my study focused primarily on the role of stochastic forces on morph-frequency variation, I intentionally sampled mostly populations of ~100 or fewer individuals because these are likely to be most susceptible to genetic drift and populations above this size are predominantly trimorphic (see Eckert &

Barrett 1992). The identification of genets in L. salicaria is straightforward because they develop into clumps with the root stock being the main organ of perennation (Shamsi &

Whitehead, 1974). In populations smaller than 100 individuals I counted all flowering individuals in the population whereas in populations larger than 100 individuals I randomly sampled ~100 plants. I estimated census size in populations larger than 100 individuals by counting the number of individuals in 2 transects that intersected at 90 degrees near the centre of the population. I estimating the area covered by each population and calculated census size from the area and approximate density.

After comparing the 1988/9 and 2013 data with a Wilcoxon sum-rank test, as expected I found that my sampling scheme resulted in a smaller median population size in the 2013 survey than the 1988/9 survey (W = 3341, P < 0.001; 1988/9 median = 125.5;

2013 median = 29.5; Fig. 3.2A and B). To obtain comparable population size distributions for comparisons between years, I selected a subset of populations containing < 130 individuals from each survey. This procedure resulted in no significant difference in median population size according to a Wilcoxon sum-rank test (W = 2909, P > 0.25;

1988/9: median = 31, n = 52 populations; 2013; median = 28, n = 101 populations; Fig.

3.2C and D). In most cases, subsequent comparisons between the two surveys used the subset of populations with these equivalent size distributions.

CHAPTER 3: POPULATION SURVEY 68

Figure 3.1. The location of Lythrum salicaria populations sampled for style morph ratios in Ontario, Canada during summer 2013. Trimorphic, dimorphic, and monomorphic populations depicted as triangles, squares, and circles, respectively. Light grey lines represent major roadways whereas black lines represent provincial borders.

CHAPTER 3: POPULATION SURVEY 69

Figure 3.2. Histograms of population sizes in Lythrum salicaria, Ontario, Canada. A)

1988/9, complete survey, B) 2013, complete survey, C) 1988/9, populations < 130 individuals, and D) 2013, populations < 130 individuals. Solid lines represent mean population size whereas the dashed lines represent median population size. A and B possess different average population sizes (W = 3341, P < 0.001; estimated difference between survey medians = 76.0, 95% CI: lower = 40, upper = 135), whereas C and D are similar in average population size (W = 2909, P > 0.25; estimated difference between survey medians = 4.0, 95% CI: lower = -3.0, upper = 15.0). Neither survey exhibits a normal distribution for population size.

CHAPTER 3: POPULATION SURVEY 70

Style morph absence and population size

I examined data from each survey and survey subset for differences in morph frequency distribution or evenness using de Finetti diagrams plotted with the R package ‘ggtern’

(Hamilton 2014). Points close to the centre of these plots represent populations with equal morph ratios, points along edges of the triangle are dimorphic and missing the morph labelled at that edge, and points at the apices of the triangle are monomorphic for style morph (see Eckert & Barrett 1992).

I tested the association between population size and morph absence in the 2013 survey using a 2 x 3 heterogeneity G-test. My test compared the number of populations with three morphs (trimorphic populations) to the number of populations lacking at least one morph (non-trimorphic populations) in each of three size classes (3-25, 26-50, and >

50 individuals). To investigate whether the likelihood of morph absence differed between the 1988/9 and 2013 surveys, population size classes, or the combination of survey and population size class I used a 2 x 4 heterogeneity G-test which compared the number of trimorphic and non-trimorphic populations present in each of two population size classes

(3-25, 26-130 individuals) from the 1988/9 and 2013 surveys. In these analyses I used two population size classes rather than three because the 1988/9 survey contained only nine populations in the 26-50 size class and inclusion of this category reduced statistical power.

I also conducted a series of 2 x 2 heterogeneity G-tests comparing the number of trimorphic and non-trimorphic populations across various combinations of size classes and survey years to investigate the individual effects of survey year and population size class on the frequency of trimorphic versus non-trimorphic populations.

CHAPTER 3: POPULATION SURVEY 71

Relation between population size and evenness in 1988/9 and 2013

I tested the effect of population size on morph evenness in the 2013 survey (all populations) using linear regression in R version 3.1.1 ‘Sock it to me’ (R Core Team

2014); R was also used for all further analyses. Population size was heavily skewed and a few large populations overpowered the smaller populations. Therefore, I used the natural logarithm of population size to mitigate the effects of large populations on the regression. I calculated a morph evenness index for each population using the trimorphic equilibrium equation (Barrett et al. 1989, equation 1.1), which provides the evenness of a population on a scale from 1 (isoplethy) to 0 (a single morph remaining).

I tested for differences in the relation between population size and morph evenness between the 1988/9 and 2013 surveys using the combined data sets of the two subsamples.

I constructed three linear regression models using the ‘lm’ function in R. The most complex model explained the response variable (population evenness) as a function of the natural logarithm of population size, survey year, and the interaction of population size and survey year, the second model contained only the natural logarithm and survey year as terms, and the most simple model contained only the natural logarithm of population size as a predictor. I compared the three models by F-tests using the ANOVA function in R which determined the significance of each survey year’s intercept and the interaction between survey slopes relative to simpler models. These data possessed similar distribution shapes but also contained minor violations of model assumptions of normality and homoscedasticity. Therefore I compared population evenness between years in the 3-

25 and the 26-130 individual size classes through a Kruskal-Wallis sum-rank test to validate my results.

CHAPTER 3: POPULATION SURVEY 72

Asymmetric morph loss

I identified the number of populations lacking each style morph in the two surveys, and in the subsamples from each survey, to determine if the 2013 survey data exhibited a similar pattern of morph absence to the 1988/9 survey. Formal statistical tests were not possible because the absence of multiple morphs from some populations resulted in non- independence and small sample sizes, thus preventing us from analysing combinations of morph absences.

Results

Style morph absence and population size

Among the 114 populations that I sampled in 2013, 84 (74%) were trimorphic, 21 (18%) were dimorphic and 9 (8%) were monomorphic. Of 21 dimorphic populations, 12 contained only L- and M-morphs, 6 contained L- and S-morphs, and 3 contained M- and

S-morphs. Of the 9 monomorphic populations, 8 contained only L-morph individuals and

1 contained only M-morph individuals. Although my survey included a higher representation of small populations than the 1988/9 survey, the proportion of populations that were trimorphic was not significantly different between the two complete surveys

(1988/9 = 77%; 2013 = 74%; G = 0.414, df = 1, P > 0.50). The majority of populations in the 1988/9 and 2013 surveys clustered around the centre of the De Finetti diagram (Figs.

3.3A and B, respectively). In the subsets of smaller populations from each survey, I observed as expected that a higher percentage of populations were missing a morph than in the full surveys, and points (populations) were also more scattered around the centre of the De Finetti diagram (Figs. 3.3C and D).

CHAPTER 3: POPULATION SURVEY 73

Figure 3.3. de Finetti plots of population morph frequencies in populations of Lythrum salicaria sampled from Ontario, Canada. A) 1988/9, complete survey, B) 2013, complete survey, C) 1988/9, populations < 130 individuals, and D) 2013, populations < 130 individuals. Triangles, squares, and circles represent trimorphic, dimorphic, and monomorphic populations, respectively. A point located farther from the L-, M- or S- edge of the triangle contains a high frequency of the L-, M-, or S-morph, respectively; whereas a point on the edge of the triangle lacks the morph labelled on that edge. Numbers near the triangle apices represent the number of monomorphic populations with that morph present in the sample. In each study, the S-morph is absent more often than the M-morph, and the

M-morph is absent more often than the L-morph.

CHAPTER 3: POPULATION SURVEY 74

The proportion of non-trimorphic populations differed significantly between population size classes in the complete 2013 survey (G = 7.97, df = 2, P < 0.02; Table

3.1). A lower proportion of large populations lacked a morph than smaller populations (3-

25 individuals: 39% missing a morph; 26-50: 22%; > 50: 12%) indicating a negative relation between population size and probability of morph absence. The 2 x 4 heterogeneity G-test, which investigated the differences between populations in each survey year and each population size category, revealed that the proportion of non- trimorphic populations in different population size classes and survey years was significantly different (G = 10.77, df = 3, P < 0.02; Table 3.2A). However, the additional 2 x 2 heterogeneity G-tests, which investigated differences between the survey years or population size classes, did not detect significant differences in the overall proportion of non-trimorphic populations between surveys (G = 0.97, df = 1, P > 0.3; Table 3.2B), or between the surveys within each population size class (G = 1.18, df = 1, P > 0.25; Table

3.2D and G = 0.10, df = 1, P > 0.70; Table 3.2E). Further G-tests detected significant differences in the proportion of non-trimorphic populations present in the population size classes overall (G = 9.5, df = 1, P < 0.01; Table 3.2C), and in the population size classes within each survey (G = 5.05, df = 1, P < 0.03; Table 3.2F and G = 4.76, df = 1, P < 0.03;

Table 3.2G). G-tests comparing survey years (Tables 3.2B, D and E) do not support a difference between the surveys in the proportion of populations lacking a morph but G- tests comparing population size classes (Tables 2C, F and G) demonstrated significant differences in the proportion of non-trimorphic populations in each size class. These results indicate that in both of the surveys smaller populations are more likely to be non- trimorphic than larger populations as predicted.

CHAPTER 3: POPULATION SURVEY 75

Table 3.1. Heterogeneity G-test comparing the number of non-trimorphic and trimorphic populations between size classes in the

2013 study of Lythrum salicaria in Ontario. Significant P-values are bolded.

Population count:

Population Observed (expected) size % Non- Non-trimorphic Trimorphic G df P trimorphic

3-25 19 (13) 30 (36) 39%

25-50 7 (8) 25 (24) 22% 7.97 2 < 0.02

> 50 4 (9) 29 (24) 12%

CHAPTER 3: POPULATION SURVEY 76

Table 3.2. Heterogeneity G-tests comparing the proportion of non-trimorphic and trimorphic populations of Lythrum salicaria in Ontario. Stats compare the surveys, population size classes, and combination of surveys and size classes in each survey year.

Significant P-values are bolded.

Population count:

Observed (Expected) % Non-trimorphic

Non-trimorphic Trimorphic G df P

A. Complete model year: population size interaction included

1988/9: 3-25 13 (8) 12 (17) 52%

2013: 3-25 19 (15) 30 (34) 39% 10.77 3 < 0.02 1988/9: 26-130 6 (8) 21 (19) 22%

2013: 26-130 10 (16) 42 (36) 19%

B. Compare survey year pooled population size

1988/9 19 (16) 33 (36) 37% 0.97 1 > 0.3

CHAPTER 3: POPULATION SURVEY 77

2013 29 (32) 72 (70) 29%

C. Compare population size pooled year

3-25 32 (32) 42 (50) 43% 9.5 1 < 0.01 26-130 16 (25) 63 (54) 20%

D. Compare years populations of 3-25 individuals

1988-9 13 (11) 12 (14) 52% 1.18 1 > 0.25 2013 19 (21) 30 (28) 39%

E. Compare years populations of 26-130 individuals

1988/9 6 (5) 21 (22) 22% 0.1 1 > 0.70 2013 10 (11) 42 (41) 19%

F. 1988/9 large vs small populations

3-25 13 (9) 12 (16) 52% 5.05 1 < 0.03

CHAPTER 3: POPULATION SURVEY 78

26-130 6 (10) 21 (17) 22%

G. 2013 large vs small populations

3-25 19 (14) 30 (35) 39% 4.76 1 < 0.03 26-130 10 (15) 42 (37) 19%

CHAPTER 3: POPULATION SURVEY 79

Relation between population size and evenness in 1988/9 and 2013

The linear regression of population evenness (E) over the natural logarithm of population size indicated a positive, significant relation between these two variables in the complete

2013 survey (F = 12.7, df = 1, 112, P < 0.001; Fig. 3.4A). The small r2 value from this relation (r2 = 0.102) indicates that although greater style morph evenness tends to occur in larger populations, other unmeasured factors besides population size also contribute towards the evenness of morphs in populations.

In the reduced data sets excluding larger populations I also detected a significant, positive relation between morph evenness and the natural logarithm of population size from the 1988/9 (F = 21.7, df = 1, 50; P < 0.0001; Fig. 3.4B) and 2013 (F = 8.83; df = 1,

99; P < 0.01; Fig. 3.4C) surveys. The relations in each survey did not differ significantly in their slopes (F = 0.775, df = 1, P > 0.35) or intercepts (F = 0.018, df = 1, P > 0.85). The

Kruskal-Wallis sum-rank test validated these findings as there were no significant differences in morph evenness between surveys in either population size class (3-25 individual size class: Kruskal-Wallis Χ2 = 0.40, df = 1, P > 0.50; 26-130 individuals:

Kruskal-Wallis Χ2 = 0.81, df = 1, P > 0.35). These results indicate that for smaller populations of L. salicaria the relation between population size and morph evenness has not changed significantly between 1988/9 and 2013.

CHAPTER 3: POPULATION SURVEY 80

Figure 3.4. The relation between morph evenness and population size in Lythrum salicaria populations sampled in Ontario, Canada, during summer 2013. A morph evenness index of

1 represents a population in which the 3 morphs are equally frequent, whereas values below 1 represent populations with unequal morph ratios or morph loss. A) All populations from the 2013 survey (E = 0.54 + 0.006 * log(N)), B) 1988/9 survey for populations containing < 130 individuals (E = 0.382 + 0.119 * log(N)), and C) 2013 survey for populations containing < 130 individuals (E = 0.49 + 0.0836 * log(N)). The triangles, squares, and circles represent trimorphic, dimorphic, and monomorphic

CHAPTER 3: POPULATION SURVEY 81 populations, respectively. Lines represent a regression of the evenness index (E) as predicted by the natural logarithm of population size (log N). Regressions for the complete

2013 survey and the restricted 1988/9 and 2013 surveys are shown. All three regressions are significant; however, low r2 values indicate that 90% of the variance in population evenness is not explained by population size; A) r2 = 0.102, P < 0.001; B) r2 = 0.303, P <

0.0001; C) r2 = 0.0819, P < 0.01).

Asymmetric morph loss

In 2013, data on morph representation from the complete survey of populations revealed that the S-morph was absent from populations most often (21 times), followed by the M- morph (14 times), and the L-morph was least often absent from populations (4 times).

Data from the restricted sample of smaller populations from 2013 was near identical, except that the S-morph was absent 20 times. These results parallel data from Eckert and

Barrett (1992) in which the S-morph was absent 18 times and the M- and L-morphs were absent 7 and 1 times, respectively. In the restricted 1988/9 sample the S-morph was absent

14 times and frequencies of absence of the L- and M-morphs were identical in both the complete and restricted samples.

Discussion

The primary goal of my study was to investigate the role of stochastic forces during biological invasion using patterns of style-morph frequency variation in invasive populations of L. salicaria. Theoretical models of the influence of finite population size on the maintenance of the tristylous genetic polymorphism make specific predictions about the expected patterns of morph-frequency variation and morph loss (Heuch 1980; Barrett

CHAPTER 3: POPULATION SURVEY 82 et al. 1989; Eckert & Barrett 1992). My results were generally consistent with the predictions of these models and indicate that repeated colonizing events and small population size play an important role in governing morph-frequency variation during invasion.

An earlier survey of invasive populations of L. salicaria in the same region (Eckert

& Barrett 1992) provided us with an opportunity to examine whether patterns of morph- frequency variation have changed during 25 years of ongoing invasion. The increased abundance of populations across the landscape has the potential to increase genetic connectivity among populations which in turn may limit the intensity of stochastic processes acting on morph evenness and loss (and see Eckert et al. 1996a). However, my comparisons of data collected in 1988/9 and 2013 revealed few differences in the patterns of morph-frequency variation between the surveys. The similarity in results indicates that despite the overall differences in invasion age between the two surveys, stochastic processes continue to play an important role during this biological invasion, especially in small populations. I begin my discussion by considering the diverse historical, demographic and genetic factors that may explain the results of my study and then consider why the patterns of morph-frequency variation differ between native and introduced populations of L. salicaria.

The pervasive influence of stochastic forces in invasive populations

Despite 25 years of ongoing invasion and an increase in the abundance of L. salicaria in many disturbed habitats of southern and central Ontario, the patterns of morph-frequency observed in 1988/9 and 2013 were remarkably similar. The relative frequencies of non-

CHAPTER 3: POPULATION SURVEY 83 trimorphic populations in the two samples were not significantly different (1988/9 = 23%;

2013 = 26%), and the specific style morphs that were absent from dimorphic populations were essentially the same in both surveys, with the S-morph and L-morph most and least often absent, respectively, results consistent with stochastic theory (Heuch 1980). In both surveys I also found a positive relation between morph evenness and population size with larger populations containing morph ratios closer to the isoplethic equilibrium. Despite the concordance between the findings of the two surveys, it is worth considering other explanations that may have caused the overall similarities in results.

The two independent surveys of morph-frequency variation in L. salicaria conducted 25 years apart both involved a large number of populations distributed across the same parts of southern and central Ontario. A more optimal sampling strategy would have been to revisit the entire set of 102 populations originally surveyed by Eckert and

Barrett (1992) in 1988/9. However, GPS coordinates were not available from the earlier survey and some easily located populations from the first survey were not located and were therefore presumed destroyed. My sample therefore mainly involved L. salicaria populations that were not represented in the original survey. Because of the increased abundance of L. salicaria in Ontario, many of these ‘new’ populations are likely to have been established since the 1988/9 survey. Also, the survey I conducted in 2013 focused mainly on populations of <100 individuals because of my interest in looking for signatures of stochasticity and, as a result, the median population size was significantly smaller than the sample collected in 1988/9. Because of this difference, I conducted comparisons of population samples from the two surveys with similar population size distributions and median population sizes. It is possible that this subsampling procedure contributed to us

CHAPTER 3: POPULATION SURVEY 84 not finding a difference between the samples, although comparisons involving the complete samples from both surveys gave similar results. Future work investigating the influence of invasion age on patterns of morph-frequency variation in L. salicaria should either involve the same sample of populations among years, or involve a large random sample of populations across the region of interest.

My initial assumption in comparing the two surveys was that the increased abundance of L. salicaria in Ontario during the 25 years between samples should have reduced the influence of stochastic forces on morph-frequency variation, even in smaller populations. Higher overall morph evenness values and fewer populations missing style morphs may have been expected. Such patterns could potentially occur through two mechanisms. First, the degree of isolation of many populations along roadsides and other transport corridors should be reduced through time, with new populations colonizing areas that were previously unoccupied. This process could influence the genetic connectivity of populations through pollen and seed mediated gene flow and buffer populations against morph loss even in small populations, as has been observed in France (Eckert et al. 1996a) and Spain (Costa et al. 2016). Second, populations surviving during the 25-year period might increase in size and this would serve to reduce the influence of stochastic processes on morph loss. Indeed, my data on the relation between population size and morph representation in populations support this possibility.

However, both of these initial assumptions about the demography and genetics of populations may be false. Although there are undoubtedly more populations of L. salicaria occupying disturbed habitats in Ontario today than 25 years ago, including some that are very large, the majority of populations are still small and many are relatively isolated. It is

CHAPTER 3: POPULATION SURVEY 85 therefore unclear how frequently gene flow plays a role in converting populations from stylar dimorphism to trimorphism. Moreover, although I have not studied changes in population size of L. salicaria in Ontario over many years in detail, my field observations indicate that population growth is not an inevitable feature of invasive populations. Rather, many populations remain relatively small owing to restrictions on suitable habitat, with some becoming reduced in size or becoming locally extirpated either through control measures or chance environmental disturbance (e.g. urban development, road widening).

Significantly, Eckert et al. (1996b) found no relation between the magnitude of the increase in population size and changes in morph evenness in a sample of 24 populations sampled over a 5-year period, suggesting that morph-frequency change even with population growth may be a relatively slow process owing to demographic factors associated with perenniality and variation in amounts of sexual recruitment. Therefore, the dynamics of the invasion process in a particular region can be highly heterogeneous with ongoing colonization and population turnover taking place over many decades so long as there are unoccupied sites across the landscape. In a relatively long-lived perennial such as

L. salicaria, non-equilibrium morph frequencies may persist for long time periods (see

Eckert et al. 1996b), and thus a greater duration between sampling intervals than my 25- year interval may be needed to observe significant changes in the patterns of morph- frequency variation in an invaded region.

Style-morph absence from populations in Ontario versus Europe

Several investigations of style-morph frequency variation in L. salicaria populations in

Europe provide an opportunity to compare patterns between native and introduced populations. The overall results of European surveys generally indicate that populations

CHAPTER 3: POPULATION SURVEY 86 differ significantly from those in Ontario. Approximately 95% of European populations that have been surveyed are tristylous (Haldane 1936; Schoch-Bodmer 1938; Halkka and

Halkka 1974; Andersson 1994; Ågren & Ericson 1996; Eckert et al. 1996a; Costa et al.

2016), compared with only ~75% of populations in Ontario (Eckert & Barrett 1992, this study). Two large-scale surveys of style-morph frequency variation in France (Eckert et al.

1996a; n = 102 populations) and Iberian Peninsula (Costa et al. 2016; n = 96 populations) are worth highlighting, as both used the similar sampling approaches used in the two

Ontario surveys. In each European survey, only five populations of L. salicaria were found that lacked a style morph. The large difference (~20%) in frequency of non- trimorphic populations between the native and introduced range of L. salicaria therefore raises the question of what factor(s) might account for these different geographical patterns.

The difference in the degree of non-trimorphism between Ontario and European populations of L. salicaria may arise from differences in metapopulation dynamics between the regions. Computer simulations suggest that gene flow on the order of m ≥

0.05 between populations in a metapopulation has the potential to maintain tristyly in small populations by restoring absent morphs to dimorphic populations (Eckert et al.

1996a). Fossils of L. salicaria in Eurasia have been dated to 10,000 years before present

(Graham 2013), indicating that this species has had a much longer time to saturate suitable habitats in its European range relative to the much shorter time period that the species has been present in Ontario (Thompson et al. 1987). Thus, according to this observation, native European metapopulations have had a significantly longer time to reach drift- migration equilibrium. Some level of ongoing colonization is obviously a feature of all

CHAPTER 3: POPULATION SURVEY 87 species (Lewontin 1965), regardless of whether one considers native or introduced populations; however, it seems probable that native populations of L. salicaria are largely tristylous because gene flow limits the persistence of non-trimorphic populations over long time scales. Estimates of morph evenness are significantly higher for trimorphic populations in France compared to those in Ontario (Eckert et al. 1996a), further supporting the view that long term gene flow in native populations reduces the influence of stochastic forces on morph-frequency variation.

Another factor that may contribute to the difference in frequency of non-trimorphic populations between native and introduced populations of L. salicaria concerns the types of landscapes in which L. salicaria occurs. Large areas of Europe, particularly in France and the Iberian Peninsula, are composed of agricultural land, thus providing L. salicaria with numerous dispersal routes along roads and drainage ditches and a high level of habitat connectivity (Eckert et al. 1996a; Costa et al. 2016). Although agriculture activities are a dominant feature of landscapes in southern Ontario where the majority of my populations were sampled, a portion of the area I sampled was on the Canadian Shield, where mixed deciduous and coniferous forests dominate. In forested regions of central

Ontario, L. salicaria populations are less abundant across the landscape and are largely restricted to open roadside ditches and disturbed wetlands. This pattern of distribution is less likely to permit gene flow among populations, potentially enhancing the significance of founder effects and genetic drift, and contributing to the lower levels of evenness and higher percentage of non-trimorphism in Ontario populations.

It is often difficult to determine whether genetic drift or founder events are the specific cause of morph loss from tristylous populations. However, several of the patterns

CHAPTER 3: POPULATION SURVEY 88 revealed in surveys of native versus introduced populations of L. salicaria suggest that the relative importance of these two stochastic processes may differ between Europe and

Ontario. In their comparison of native versus introduced populations, Eckert et al. (1996a) found that the probability of morph evenness and morph loss was more strongly associated with population size in the native compared to the introduced range. Only very small populations of less than 10 individuals in France lost morphs, a pattern also found in the

Iberian Peninsula by Costa et al. (2016), where all non-trimorphic populations were less than 15 individuals. These results are consistent with the operation of genetic drift. In contrast, non-trimorphic populations in the introduced range can occasionally be quite large (e.g. Ontario: N = 100-500; Eckert & Barrett 1992). In my survey, dimorphic populations in Ontario possessed a median size of 25 individuals in contrast to seven in dimorphic populations from Sweden (Ågren & Ericson 1996). Moreover, among the eight monomorphic populations of the L-morph revealed by my survey, four were relatively large (N = 21-33 individuals), and above the threshold value of 15 below which theoretical studies indicate that genetic drift will play a dominant role in causing morph loss (Heuch

1980). Indeed, these models demonstrate that tristyly can be maintained for up to 150 generations if population sizes are above 20 individuals. The occurrence of relatively large non-trimorphic populations in Ontario suggests that unlike the native range, founder events may play a more important role than genetic drift in explaining morph absence from populations of L. salicaria.

During invasion, colonizing plants often lack compatible mates and weak self- incompatibility (pseudo-self-fertility) may therefore provide reproductive assurance to founding individuals (Levin 1996). Variation in the expression of pseudo-self-fertility

CHAPTER 3: POPULATION SURVEY 89 among floral morphs may favour some morphs during founder events, potentially contributing to morph absence from populations. Beginning with Darwin’s early work on

L. salicaria (summarized in Darwin 1877), controlled crosses have consistently revealed a small number of self-fertile individuals, especially of the M-morph and to a lesser extent the L-morph (reviewed in Colautti et al. 2010a). These individuals are capable of setting seed from self- and intramorph pollinations. As a result of the inheritance of tristyly

(Barlow 1923; Fisher & Mather 1943), plants of the M-morph following selfing always segregate M-morph progeny and in most cases L-morph progeny; whereas L-morph plants produce only L-morph progeny. S-morph progeny are not produced in either case because the dominant S-allele is restricted to genotypes of the S-morph. Thus, it is quite plausible that genetic drift played no role in the origin of the 12 dimorphic L-M morph populations and eight monomorphic L-morph populations in my survey. Rather, these could have arisen from founder events favouring individuals of the M- and L-morph with leaky self- incompatibility, a process consistent with Baker’s law (Baker 1955; Pannell 2015). Future studies comparing native and introduced populations of L. salicaria would be valuable to determine whether the invasion process may have selected for an increased prevalence of pseudo-self-fertility in North American populations.

CHAPTER 4

GENETIC AND ENVIRONMENTAL INFLUENCES ON PARTIAL SELF-

INCOMPATIBILITY IN LYTHRUM SALICARIA (LYTHRACEAE)

Abstract

Angiosperms are commonly classified as self-incompatible or self-compatible. This dichotomy has influenced research on the ecological and demographic consequences of colonization because of the predicted benefits of self-compatibility for establishment at low density. Some individuals of self-incompatible species, however, exhibit partial self- incompatibility (PSI) meaning that they set variable amounts of seed following self- pollination. Here, I investigate genetic and environmental components of PSI in tristylous

Lythrum salicaria as a context for understanding colonization and the floral morph structure of invasive populations. I surveyed variation in PSI using experimental self- and cross-pollinations on plants (n=338) grown under glasshouse conditions. I compared the stability in expression of PSI over two years (n=80), and by pollinating selected clones

(n=12) grown under wet and dry conditions. I compared the compatibility status of mid- styled parents (n=14) and their selfed offspring to determine if there was a genetic component to PSI, and whether the expression of PSI, measured as fruit set after self- pollination, differed between segregating long- and mid-styled plants. Approximately 34% of plants set seed following self-pollination. The mid-styled morph produced more fruit following self-pollination in comparison with the remaining morphs. An index of self- compatibility (ISC) exhibited significant repeatability for individual plants over two successive flowering seasons. There was no systematic influence of wet and dry growing conditions at flowering on PSI, but significant genetic differences among clones in overall

90 CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 91 expression of PSI were evident. A heritable component to PSI was confirmed by parent- offspring regression and compatibility values were significantly higher in mid- versus long-styled progeny. My study demonstrated a weak but significant genetic component to morph-specific variation in PSI. The capacity of the M-morph to set seed after selfing may enable individuals to found populations and contribute to the occurrence of dimorphic populations missing the short-styled morph in the invasive range of L. salicaria.

Introduction

Hermaphroditism provides opportunities for both cross- and self-fertilization and their relative amounts represent a key axis of mating-system variation in flowering plants with profound ecological, genetic and evolutionary consequences (Darwin 1876; Lloyd 1992;

Igić & Busch 2013; Wright et al. 2013; Barrett & Harder 2017). A major determinant of outcrossing is self-incompatibility, a genetically-based physiological mechanism (de

Nettancourt 1977; Franklin-Tong 2008), that serves to limit the harmful consequences of self-fertilization resulting from inbreeding depression (Charlesworth & Charlesworth

1987). Self-incompatibility systems differ in physiology, genetics, and molecular control

(e.g., homomorphic versus heteromorphic, sporophytic versus gametophytic), and it has been estimated that these self-rejection mechanisms occur in at least 40% of angiosperms from over 100 families, making them the most common and phylogenetically dispersed anti-selfing mechanisms in plants (Allen & Hiscock 2008; Igić et al. 2008). However, self- incompatibility frequently breaks down resulting in plants that exhibit different degrees of self-compatibility (de Nettancourt 1977; Barrett 1988; Mable et al. 2005). This shift from self-incompatibility to self-compatibility is commonly associated with the transition from outcrossing to selfing and has been documented in numerous families (Stebbins 1974;

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 92

Goodwillie 1999; Goldberg et al. 2010). The evolution of self-compatibility may be especially important when ecological and demographic conditions limit cross-pollen dispersal among plants.

The early stages of colonization associated with biological invasions represent an important ecological and demographic context for evaluating the costs and benefits of plant mating systems (reviewed in Brown & Burdon 1987; Barrett 2011; Pannell 2015).

The capacity for self-fertilization should be particularly beneficial in species that experience recurrent colonizing episodes as it allows reproduction under the low-density conditions typical of establishment following dispersal (Lloyd 1980). In contrast, colony foundation at low density may be restricted if self-incompatibility is strongly expressed because of the requirements for cross-compatible mates and pollen vectors that may be absent from small populations. The idea that self-compatibility is beneficial for establishment at low density after long-distance dispersal, e.g. on oceanic islands, is known as Baker’s law (Baker 1955; 1967). Despite some controversies concerning the generality of Baker’s law (Busch 2011; Cheptou 2012; reviewed in Pannell et al. 2015), there is now considerable evidence in favour of the benefits of self-compatibility for colonization under a range of demographic scenarios (Rambuda & Johnson 2004; Van

Kleunen et al. 2008; Pannell 2015; Grossenbacher et al. 2017). Notably, much of the discussion on plant mating and colonization tends to contrast the simple dichotomy of self- incompatibility and self-compatibility.

Although the compatibility status of plants is commonly viewed as a binary trait

(but see Cheptou et al. 2002; Good-Avila et al. 2008), increasing evidence indicates variation in the expression of self-incompatibility within and among plant populations. For

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 93 example, in a survey of self-incompatibility expression in 1238 species conducted by

Raduski et al. (2012), ~25% exhibited 0.2 – 0.8 of the fruit set following self-pollination that they would have set after cross-pollination. This variation in expression of self- incompatibility can be manifested in numerous ways from sporadic self-compatible individuals in otherwise strongly self-incompatible populations (Shore & Barrett 1986;

Tsuchimatsu et al. 2010) to wide quantitative variation in the capacity of plants to set seed following self-pollination (Levin 1996; Stone et al. 2006; Good-Avila et al. 2008). In addition, in some fully self-fertile species, incompatibility is cryptically expressed and can only be detected by studies of cross- and self-pollen tube growth and/or the use of genetic markers (Bateman 1956; Weller & Ornduff 1977; Cruzan & Barrett 1993). Weak or

‘leaky’ self-incompatibility occurs because diverse genetic, developmental, and environmental mechanisms influence the expression of this anti-selfing mechanism.

The occurrence of self-fertile individuals in species that retain a functioning self- incompatibility system has been referred to as pseudo-self-fertility (de Nettancourt 1977;

Levin 1996) or pseudo-self-compatibility (Baldwin & Schoen 2017). Some progress has been made in understanding the genetic basis of pseudo-self-fertility. In several species, self-compatibility results from unlinked polygenic modifiers that alter the strength of self- incompatibility, whereas in others mutations of alleles at the S-locus are involved (Ascher

1984; Good-Avila & Stephenson 2008; Baldwin & Schoen 2017). Moreover, crossing studies indicate that pseudo-self-compatibility can respond to artificial selection

(Lundqvist 1968; Henny &Ascher 1976; Flaschenriem & Ascher 1979; Dana & Ascher

1985; Bixby & Levin 1996) and may provide the necessary standing genetic variation for the selection of full self-compatibility. Environmental and developmental factors are also

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 94 known to modify the expression of self-incompatibility. For example, high temperatures and increasing flower age are commonly associated with a weakening of incompatibility and have often been exploited to produce self-seed (Asher & Peloquin 1966; de

Nettancourt 1977; Stephenson et al. 2000; Levin 1996). Plasticity in the expression of self- incompatibility (sensu Travers et al. 2004; Good-Avila et al. 2008) may complicate genetic analysis but may be ecologically significant for populations exposed to fluctuating environmental conditions.

Variation in the expression of self-incompatibility is widely studied in heterostylous plants owing to the ease with which these plants can be identified in natural populations. Controlled self- and cross-pollinations in both distylous and tristylous species have often revealed differences between the floral morphs in the strength of heteromorphic incompatibility (reviewed in Barrett & Cruzan 1994). The underlying physiological and genetic mechanisms responsible for morph-specific differences in incompatibility are poorly understood but may be associated with the contrasting patterns of pollen tube growth and inhibition evident in the floral morphs (Bawa & Beach 1983; Scribailo &

Barrett 1991). A particularly notable example is reported in several tristylous taxa

(Eichhornia azurea – Bianchi et al. 2000; Pontederia spp. – Barrett 1977; Glover &

Barrett 1983; Barrett & Anderson 1985; Puentes et al. 2013; Lythrum salicaria – Darwin

1877; Stout 1923; O’Neil 1994; Colautti et al. 2010a) in which the mid-styled morph

(hereafter M-morph) has significantly weaker self-incompatibility than the long- and short-styled morphs (hereafter L- and S-morphs). Colautti et al. (2010a) proposed that the weak expression of trimorphic incompatibility in the M-morph of L. salicaria could

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 95 increase its colonizing potential compared to the L- and S-morphs by facilitating colony foundation following dispersal.

Here, I investigate variation in the expression of trimorphic incompatibility in

Lythrum salicaria, with a particular focus on the M-morph. Throughout this chapter, I refer to the occurrence of plants that set variable amounts of seed following self- pollination in an otherwise self-incompatible species as exhibiting ‘partial self- incompatibility’ (hereafter PSI) in preference to ‘pseudo-self-compatibility’ or ‘pseudo- self-fertility’ because at this stage several of the criteria for pseudo-self-fertility outlined by Levin (1996) have yet to be established in L. salicaria. Through controlled self- and cross-pollinations conducted under glasshouse conditions, my study addressed the following questions: 1) what is the frequency of PSI in invasive populations of L. salicaria and does the M-morph exhibit the weakest expression of self-incompatibility? Based on previous studies of L. salicaria (reviewed in Colautti et al. 2010a), I predicted higher quantities of fruit and seed set following self-pollinations of the M-morph in comparison with the L- and S-morphs. 2) Given previously known environmental influence on PSI

(see Levin 1996; Good-Avila & Stephenson 2008), how stable is the expression of PSI between years among genotypes which express PSI to different degrees? I conducted controlled self- and cross-pollinations on plants in two consecutive years and compared fruit and seed set. 3) To what extent does the expression of PSI change with growing conditions at flowering? In an experiment with replicated clones of selected genotypes grown under wet and drought stress conditions I evaluated the plasticity of PSI, and also the extent to which genotypes maintained their differences in PSI. 4) Is there evidence of a heritable component to variation in PSI, and are morph-dependent differences in the

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 96 expression of PSI maintained in selfed offspring? I compared parental and offspring expression of PSI in genotypes of the M-morph that varied in PSI and tested the hypothesis that incompatibility should be weaker in M- than L-morph progeny, owing to morph-limited expression of PSI.

Materials and Methods

Study species

Lythrum salicaria is a largely outcrossing, showy, autotetraploid, insect-pollinated, amphibious perennial native to Eurasia. It colonizes a range of wetland habitats, including freshwater marshes, flooded pastures and roadside ditches, and has been introduced to several temperate regions of the world, in some of which it is an invasive species. Lythrum salicaria was introduced to the eastern seaboard of North America in the early 1800s and has since expanded its range southward to Georgia, northward to various US states and

Canadian provinces, and westward to the Pacific coast (Thompson et al. 1987). In eastern

North America there is evidence that populations of L. salicaria maintain considerable quantitative genetic variation for life-history traits that vary clinally with latitude and that populations have evolved local adaption to growing season length (Montague et al. 2008;

Colautti & Barrett 2010, 2013; Colautti et al. 2010b). Populations of L. salicaria reproduce exclusively from seed although individuals can produce large clumps ~1m in diameter (Mal et al. 1992; Yakimowski et al. 2005). Plants begin flowering ~12 weeks after germination and produce numerous inflorescences with weakly zygomorphic flowers that are in anthesis for 2-3 days (Thompson et al. 1987). Individuals can be long-lived, flowering for at least 12 years (S.C.H. Barrett pers. obs.).

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 97

Lythrum salicaria exhibits the floral polymorphism tristyly, with most populations composed of the long-, mid-, and short-styled floral morphs (hereafter L-, M-, S-morphs).

In common with many tristylous species, the reciprocal arrangement of stigma and anther heights is associated with a physiologically-controlled trimorphic incompatibility system that limits self and intramorph mating (Darwin 1877; Barrett & Cruzan 1994). Compatible crosses involve cross-pollinations between anthers and stigmas of equivalent height

(‘legitimate pollinations’). However, experimental evidence from controlled pollinations of L. salicaria indicates more weakly expressed trimorphic incompatibility in some plants, especially in the M-morph in which self-pollinations with pollen from long-level anthers can result in fruit and seed set (reviewed in Colautti et al. 2010a).

Sampling and cultivation of plants

I collected 120 maternal seed families of L. salicaria from four geographically separated locations in the Greater Toronto Area (population designations: HUM-1, CDV, RRV,

DON) during October – November, 2012 (Table 4.1). Each of the populations sampled was large, tristylous, and comprised of approximately several hundred individuals based on visual estimates. In early spring 2013, I germinated seed and grew six offspring per maternal parent in a pollinator-free glasshouse at the University of Toronto, St. George campus. I transplanted seedlings into 15.24 cm (6 inch) pots containing Promix BX potting media and placed these on flooded benches in the glasshouse. I maintained glasshouse temperatures between 20-25°C with supplemental light for 12 hours per day and provided 20-20-20 fertilizer biweekly following the manufacturer’s instructions. I over-wintered plants in the glasshouse from November – February by cutting back above ground stems to ~5cm above the soil surface, reducing temperatures to 10-15°C, and

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 98 applying fertilizer only once each month. I also repotted plants with Promix BX media 2-3 times/year. All experiments described below are based on the plants and glasshouse conditions described above, unless stated.

Table 4.1. Location of the four population of Lythrum salicaria in Toronto, Ontario,

Canada. Also displayed is the number of families sampled and the frequency of individuals per population which set at least one fruit after self-pollination.

Number Frequency Population Latitude Longitude of setting name families self-fruit

Humber Bay N 43.62270000 W 79.47388889 36 0.33

Cedarvale Park N 43.68973611 W 79.42416667 45 0.40

Rouge River N 43.81380000 W 79.15527778 39 0.30 Park

East Don River N 43.78075000 W 79.37000000 35 0.35

Frequency of partial self-incompatibility

In May 2013, I randomly selected 338 plants evenly distributed among the four populations on which to investigate the frequency of occurrence of self-compatible plants.

Because there was no significant difference among populations in the frequency of plants expressing partial self-incompatibility (Table 4.1; generalized linear model likelihood

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 99 ratio; Χ2 = 2.19, df = 3, P > 0.53), I pooled data from plants sampled from the four populations in this study. On a single inflorescence of each plant, I pollinated five pairs of flowers on their first day of anthesis. I accomplished this by removing selected anthers with forceps from a donor plant and rubbing the anthers against recipient stigmas to deposit abundant pollen. I self-pollinated one flower per pair using pollen from mid-level anthers in the L- and S-morphs or long-level anthers of the M-morph. I chose these anther levels because previous pollination experiments on L. salicaria and other tristylous species have established that they produce more seed following self-pollination than the alternate anther levels within a flower (Darwin 1877; Barrett & Cruzan 1994; O’Neil 1994; Colautti et al. 2010a). I cross-pollinated the second flower in each pair with pollen from a haphazardly chosen compatible pollen donor using the anther level corresponding to the recipient stigma level (‘legitimate pollination’). I removed the mid-level anthers from all flowers that were cross-pollinated on L- and S-morph plants and long-level anthers from flowers cross-pollinated on M-morph plants to prevent self-pollen contamination. I performed all self-pollinations before cross-pollinations and cleaned forceps in alcohol between treatments to avoid pollen contamination of crosses. I marked the base of the corolla tube of each flower with either red or yellow paint to identify whether flowers received self- or cross-pollen, respectively. After six weeks, I recorded fruit set on each plant and harvested all mature fruit. I allowed fruits to dry for one week in open Eppendorf tubes before I counted the number of seeds per fruit under an Olympus SZ61 dissecting microscope. From this experiment, I established the frequency and extent of partial self- incompatibility among individuals in my sample of L. salicaria plants.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 100

After plotting the frequency distribution of partial self-incompatibility in my sample of L. salicaria (Fig. 4.1), I examined the extent to which plants set fruit after self- pollination using a generalized linear model with a binomial residual distribution. All statistical analyses described in this article were conducted in R version 3.3.2 “Sincere

Pumpkin Patch” (R Development Core Team, 2016). In all linear models, I extracted F- statistics and model significance using the “summary” function in R, and in all ANOVA, generalized linear models, and mixed models, I obtained either the sum of squares or the likelihood-ratio Χ2 and significance with the type III sum of squares in the package ’car‘

(Fox & Weisberg 2011). I tested the effects of morph on the number of fruit set after self- fertilization with the main effects of morph and outcross fruit set as predictor variables. I investigated differences between the number of self-fertilized seed set per fruit in plants setting at least one self-pollinated fruit using a linear model of the natural logarithm of average seed set per fruit in each plant, to match the assumption of normality in the seed counts, with morph and outcross fruit set as predictive factors.

Stability of partial self-incompatibility between years

Having established that a significant number of L. salicaria plants were partially self- incompatible, I next investigated the stability of this trait in individual plants by comparing self-pollinated fruit and seed set between years. In April 2014, I selected 20 L-,

40 M-, and 20 S-morph individuals which set at least one self-pollinated fruit following hand pollination in the previous year. I chose twice as many individuals of the M-morph for this experiment because more individuals of this morph, as expected, set a higher number of fruit after self-pollination than the L- or S-morphs. In May 2014, I repeated the same pollination protocol as 2013 on this subset of 80 plants, except that I pollinated 10

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 101 pairs of flowers rather than five. All fruits were harvested when mature and I recorded fruit set and mean seed per fruit per plant as in 2013.

I examined the relation between outcrossed fruit set and self-fertilized fruit set through a generalized linear model with a Poisson residual distribution on the count of fruit set from self-fertilization versus the count of fruit set from cross-fertilization. I detected a significant relation between these variables (Χ2 = 7.0125, df = 1, P < 0.01) indicating an overall effect of the number of outcrossed fruit set on the number of self- fertilized fruit set per plant. I also tested for an effect of mean seed set per fruit after selfing and outcrossing in each plant; however, I did not detect a significant relation between these variables (F = 0.06616, df = 1, residual df = 37, P > 0.75).

I determined the proportion of plants in 2014 that set fruit after self-fertilization and compared this value with the 2013 percent fruit set using a heterogeneity G-test. I controlled for the positive correlation between self-fertilized and cross-fertilized fruit set by using an index of self-compatibility (hereafter ISC) for fruit set as my response variable. I calculated this index as the number of self-fertilized fruit divided by the number of cross-fertilized fruit for each plant, with the few values greater than one reduced to one so that all ISC fruit set values ranged between zero and one. I then examined the correlation between 2013 ISC fruit set and 2014 ISC fruit set, as well as the correlation between mean self-fertilized seed and cross-fertilized seed per fruit for each plant using linear regression.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 102

Environmental modification of partial self-incompatibility

To investigate the extent to which different environmental conditions may influence the expression of PSI, I selected 15 individuals of the M-morph in 2014-5 comprising a range of variation in this trait. My experiment contrasted different moisture regimes because L. salicaria plants experience both flooded and terrestrial conditions and my field observations in Ontario indicated that summer drought commonly impacts plants growing in roadside ditches.

From each of the 15 genotypes, I produced 14 stem cuttings from ~10 cm long axillary shoots and rooted the stems in perlite substrate in a plant propagation chamber with regular misting in the glasshouse facility. After approximately four weeks, at which time cuttings had produced roots, I transplanted them into 15.24 cm (6 inch) pots filled with Promix media and grew them under glasshouse conditions as described above. In

April of 2016, I selected 12 of the original 15 genotypes with 10 clones and assigned half to a “dry” treatment and half to a “wet” treatment. I established 24 plastic trays, half of which involved a “wet” treatment and half of which experienced a “dry” treatment, and arranged them in a checkerboard pattern across a glasshouse bench. I randomly placed five plants into each tray with the restriction that I placed clones of each genotype into trays matching their pre-assigned condition (wet versus dry) and I only placed a maximum of one clone of each genotype into a given tray. I maintained a depth of 3.5 cm of water in all trays until inflorescence initiation, at which time I reduced the water in the dry trays to approximately 705ml per plant each week while continuing to provide 3.5 cm of standing water in the wet trays. Plants in the dry treatment were noticeably stressed and showed wilting between waterings. From mid-May to early July, I self- and cross-pollinated five

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 103 pairs of flowers on each plant using my standard protocol described above. In mid-July to early August I collected mature fruit and measured fruit set and seed set per capsule.

I tested for the relation between self-fertilized fruit set and mean seed set per fruit per plant versus cross-fertilized fruit set and mean seed per fruit per plant as discussed above. I detected a positive correlation between self- and cross-fruit set in my generalized linear model with Poisson residual distribution (Χ2 = 17.674, df = 1, P < 0.0001), but I detected no significant relation between the mean self- and cross-seed set per fruit in plants setting at least one self-fertilized seed in the linear regression (F = 2.393, df = 1, residual df = 69, P > 0.1). Therefore, I chose to use the ISC for fruit set and mean self- seed set per capsule per plant as response variables in my analyses. I performed two-way

ANOVAS to detect the effects of environmental treatment (wet versus dry), plant genotype and their interaction with ISC fruit set and mean self-seed set per fruit per plant.

Relations between parent and offspring partial self-incompatibility

To investigate whether offspring from partially self-incompatible plants of the M-morph also exhibited similar values of PSI as their parents, and whether segregating L- and M- morphs within selfed families differed in compatibility status, I examined the relations between parent and offspring values of PSI. In April 2016, I selected 14 genotypes of the

M-morph included in the preceding experiment. I germinated seed from self-fertilized fruits from each of the parents and obtained a total of 344 offspring, of which 288 flowered (mean number of plants per family: 19.4 range 6-24). In June, I transplanted individuals into 10.16 cm (4 inch) plastic pots and placed them on a flooded glasshouse bench. When these families commenced flowering, I recorded the style morph of each

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 104 plant and pollinated five pairs of flowers using the same protocol as described above. I measured fruit and seed set of each plant when fruit matured during August to September.

I found a significant relation between self-fertilized fruit per plant and cross- fertilized fruit per plant (Χ2 = 326.24, df = 1, P < 0.0001). I also detected a positive relation between self- and outcross-seed set per fruit per plant (F = 54.73, df = 1, residual df = 144, P < 0.0001). Therefore, I used the ISC of fruit set and a similar index based on mean self-seed per fruit (ISC seed set), which I obtained by dividing the mean seed set per capsule from selfing by the mean seed set per capsule from outcrossing per plant.

The results I obtained from self-pollinating offspring from plants of the M-morph

(see Fig. 4.6) indicated that the values of PSI differed markedly between L- and M-morph progeny within families. I therefore separately examined the relation between the parental

ISC and the responses of the L- and M-morph offspring in all further analyses. I first performed a generalized linear mixed-model with binomial residuals on the likelihood that plants set one self-fertilized fruit, with the parental ISC for fruit set designated as a fixed effect and progeny family mean as a random effect. I then tested the effect of parental ISC on offspring ISC for fruit set, using linear mixed model regressions of ISC over parental

ISC for fruit set as a fixed factor, with progeny mean in each family as a random factor. I also tested the effects of parental ISC for mean seed set per fruit against progeny ISC for mean seed set per fruit. This was conducted using a linear mixed model with the ISC for seed set per fruit in offspring predicted by the mean seed set per fruit in the parent as a fixed effect, and the family mean for ISC for seed set as a random variable in the model.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 105

Results

Frequency of partially self-incompatible plants

Following self-pollination, approximately a third of all individuals sampled from among the four L. salicaria populations set at least one fruit. The frequency of individuals with

PSI ranged from 0.30–0.40 among the populations. Most plants set four or five fruit following cross-pollination (mean = 4.4) and a relatively high, and normally distributed, number of seeds per fruit (mean = 75). However, after self-fertilization, plants set a significantly smaller number of fruit and seed (mean fruit per plant = 0.8, mean seed per fruit: 11.1; Fig. 4.1). Plants which set more cross-pollinated fruit were more likely to set self-fertilized fruit (Χ2 = 17.3, df = 1, P < 0.0001). Individuals of the M-morph were more likely to set fruit than individuals of the S-morph, and marginally more likely to set fruit than individuals of the L-morph (Fig. 4.2A). The M-morph set significantly more self- pollinated fruit than the L- or the S-morph (Fig. 4.2B). However, I did not detect a significant difference among the floral morphs in the mean number of seeds set per fruit in plants setting at least one self-fertilized fruit (Fig. 4.2C).

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 106

Figure 4.1. Variation in the proportion of plants setting different quantities of A) fruit and

B) seed after cross- and self-pollination of five flower pairs of Lythrum salicaria. The data are from experimental pollinations conducted under glasshouse conditions on 338 plants from four populations.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 107

Figure 4.2. Morph-specific variation in partial self-incompatibility in Lythrum salicaria based on experimental pollination conducted under glasshouse conditions on 338 plants.

A) Proportion of plants setting fruit after self-pollination. The M-morph set fruit with a significantly higher probability than the S-morph. B) Number of fruit set per plant after self-pollination. The M-morph set significantly more fruit after self-pollination than the L- or S-morph. C) Mean number of self-fertilized seed per fruit in plants which set at least one fruit after self-pollination. There was no significant difference among morphs in number of seed set per fruit.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 108

Stability of partial self-incompatibility between years

Lythrum salicaria plants that I pollinated in 2014 were more likely to set fruit than the total sample of plants that were used in controlled pollinations in 2013 (G = 31.4, df = 1, P

< 0.0001). I found a significant relation in the ISC for fruit set per plant between 2013 and

2014 (F = 8.97, df = 1, residual df = 63, P < 0.01; Fig. 4.3A); however, there was no significant relation between the mean number of seed set per fruit between years (F =

4.02, df = 1, residual df = 32, P > 0.05; Fig. 4.3B).

Figure 4.3. Relations between partial self-incompatibility expressed in 2013 versus 2014 in a sample of Lythrum salicaria plants pollinated in each year under glasshouse conditions. A) Index of self-compatibility (ISC) for fruit set in 2013 vs 2014. The index of self-compatibility was calculated as the number of self-fertilized fruit divided by the number of cross-fertilized fruit (see text for details). B) Mean seed per fruit in plants that set at least one self-fertilized fruit in each year.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 109

Environmental modification of partial self-incompatibility

In my experiment, the ISC for fruit set and the mean number of seeds per fruit differed significantly among cloned genotypes in both the wet (Fig. 4.4A) and dry conditions (Fig.

4.4B). However, there was no significant main effect of environmental treatment or an interaction between genotype and environment for either measure (Table 4.2). Individual genotypes varied in their response to the treatments but in the majority of plants the values obtained were not significantly different between environmental treatments, and there was no predictable general increase or decrease in PSI under water stress conditions.

Table 4.2. Effects of family, environment and their interaction on the expression of partial self-incompatibility in cloned individuals of Lythrum salicaria. A) Effects of variables on the index of self-compatibility (ISC) for fruit set; B) effects of variables on the mean number of seed set per self-fertilized fruit per plant.

A. Index of self-compatibility for fruit set

Sum of Squares df P-value

(Intercept) 1.5625 1 < 0.0001

Family 6.5866 1 < 0.0001

Environment 0.0201 1 > 0.60

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 110

Family by Environment 1.3488 1 > 0.20

Residuals 7.7159 5

B. Seed set per fruit

Sum of Squares df P-value

(Intercept) 4812.9 1 < 0.0001

Family 20784 1 < 0.0001

Environment 32.2 1 > 0.75

Family by Environment 5583.2 1 > 0.10

Residuals 27411.5 5

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 111

Figure 4.4. Reaction norms of partial self-incompatibility for 12 cloned genotypes of

Lythrum salicaria grown under wet and dry conditions in a glasshouse. A) Mean and 95% confidence intervals for the index of self-compatibility (ISC) for fruit set. Cloned genotypes possessed significantly different values of partial self-incompatibility; however, the environment and environment by genotype effects were not significant. B) Mean seed set per fruit in plants setting at least one fruit. There was a significant effect of genotype, but no significant effect of environment or genotype by environment.

Relations between parent and offspring partial self-incompatibility

I detected L- and M-morph progeny in 12 of the 14 selfed families and only M-morph progeny in the remaining two families. The overall ratio of L- to M-morph progeny in the

12 segregating heterozygous families was not significantly different from the 3:1 ratio expected after self-fertilization in M-morph plants which are simplex for the dominant M- allele (Χ2 = 0.16, df = 1, P > 0.65, Table 4.2A; Fisher & Mather 1943). This does not deviate from the expectations of populations at isoplethic equilibrium because ~78-89% of

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 112

the M-morph individuals are expected to be simplex at the M-locus, depending on the rate

of double reduction at the M-locus (Heuch & Lie 1985).

Table 4.3. The number of L- and M-morph progeny in selfed families of the M-morph of

Lythrum salicaria. A) Twelve families segregated L- and M-morph, progeny. Progeny

morph ratios did not significantly diverge from the expected 3:1 ratio for the M- and L-

morphs (Χ2 = 0.16, df = 1, P > 0.65). B) Two families only produced M-morph progeny

suggesting multiple dominant alleles at the M-locus for those parents.

A. Families segregating L- and M-morph progeny

Family L-morph M-morph

223 3 10

535 6 17

555 6 11

596 8 14

645 8 15

687 1 14

843 2 12

469 7 17

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 113

661 6 18

793 4 18

816 7 17

861 6 18 total 64 181

B. Families producing only M-morph progeny

Family L-morph M-morph

210 0 21

472 0 6 total 0 27

I found a positive relation between the proportion of M-morph progeny which set

fruit after self-fertilization and the parental ISC for fruit set (Χ2 = 5.3, df =1, P < 0.05; Fig.

4.5A). But the relation between the parental ISC and the proportion of L-morph progeny

which set fruit was not significant (Χ2 = 0.0, df = 1, P > 0.95; Fig. 4.5B). Similarly, I

detected a significant relation between the parental ISC for fruit set and the offspring ISC

for fruit set of the M-morph but not in the L-morph (M-morph: Χ2 = 6.9, df = 1, P < 0.01,

L-morph: Χ2 = 0.09, df = 1, P > 0.75; Fig. 4.5B). There was no significant relation

between the ISC for seed set in the parental versus offspring generations for either the L-

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 114 or M-morph (M-morph: Χ2 = 1.9, df = 1, P > 0.15; L-morph: Χ2 = 3.2, df = 1, P > 0.05;

Fig. 4.5C).

Figure 4.5. Relation between parental and offspring values of partial self-incompatibility following self-pollination of Lythrum salicaria plants grown under glasshouse conditions.

I obtained all families through self-fertilization of M-morph plants which expressed varying levels of partial self-incompatibility. M-morph offspring are depicted in black, whereas L-morph offspring are depicted in grey. A) Relation between the proportion of offspring setting fruit after self-pollination and the parental index of self-compatibility

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 115

(ISC) for fruit set. There was a positive correlation between the proportion of M-morph offspring per family which set fruit and the parental ISC for fruit set. This pattern was not evident in L-morph progeny. B) Relation between the offspring ISC in fruit set after self- pollination and the parental ISC for fruit set. There was a strong positive correlation for this measure of partial self-incompatibility between M-morph progeny and parents, but no significant correlation between L-morph progeny and their parents. C) Relation between offspring ISC for mean seed set per fruit in plants setting at least one self-fertilized fruit and the parental ISC for mean seed set per fruit. There was no significant correlation between parents and offspring for either the L- or M-morph

The overall average expression of PSI based on the ISC for fruit set and seed set per plant was significantly higher in self-fertilized progeny of the M-morph compared to L- morph progeny (Fig. 4.6).

Figure 4.6. Morph-specific differences in partial self-incompatibility in L- and M-morph offspring segregating from self-fertilization of the M-morph of Lythrum salicaria. Values

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 116 plotted are grand means and 95% confidence intervals from 12 selfed families that segregated the L- and M-morphs comprising 64 L- and 181 M-morph plants. A) Index of self-compatibility for fruit set in L- and M-morph progeny. B) Index of self-compatibility for mean seed per fruit in each plant when the plant set at least one fruit after self- fertilization.

Discussion

Controlled self- and cross-pollinations on a large sample of L. salicaria plants indicated that approximately one third exhibited partial self-incompatibility, setting at least one fruit with seed following self-pollination (Fig. 4.1). Using these partially self-incompatible plants I conducted several experiments under glasshouse conditions which revealed the following major findings: 1) Despite considerable variation from year-to-year in the expression of PSI, there was significant repeatability of self-fertilized fruit set values for individual plants (Fig. 4.3); 2) Using replicated clones given contrasting growing conditions during flowering, I found no evidence that drought stress modified the expression of PSI; however, there were significant differences among clonal groups across conditions indicating genetic variation for PSI (Fig. 4.4; Table 4.2). 3) Parent-offspring analyses confirmed the presence of a low level of heritable variation in PSI and demonstrated morph-specific expression, with M-morph progeny on average setting more fruit following self-pollination than L-morph progeny. I now discuss my findings focusing on the potential factors influencing PSI variation in L. salicaria and their ecological and evolutionary significance in the context of colonization, population morph structure and the evolution of self-compatibility.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 117

Mechanisms governing variation in partial self-incompatibility

Controlled self- and cross-pollination studies are widely employed to investigate the compatibility status of plants. Measurements of the percent of pollinated flowers setting fruit and the number of seeds per fruit are commonly used to evaluate the strength of incompatibility. I used both of these measurements but in general found that fruit set provided a better indicator of the strength of incompatibility than seed set per fruit. In some comparisons (e.g. among floral morphs, Fig. 4.2C), there was no significant difference in seed number per fruit following cross- and self-pollination whereas fruit set differed significantly among the morphs. These two measures of fertility are governed by distinct reproductive processes and therefore may not always be correlated with one another. Seed number variation per fruit can be strongly influenced by the abortion of developing embryos and seeds owing to early-acting inbreeding depression (Husband &

Schemske 1996). In contrast, empirical evidence indicating that fruit set is affected by inbreeding depression to the same degree is less common. This may have led Raduski et al. (2012) to state “we are not aware of any evidence that suggests a measurable effect of inbreeding depression on fruit set (p. 1277)”. However, fruits in which few developing embryos survive are often aborted (Stephenson 1981), and thus quantities of fruit and seed set are influenced by inbreeding depression but perhaps to different degrees. Lythrum salicaria is primarily outbreeding and it is probable that some inbreeding depression occurs during seed development and this may have potentially influenced the quantities of fruit and seed set that I obtained from controlled self-pollinations.

In several of my comparisons I used an index of self-compatibility. Similar relative indices have been used to assess variation in compatibility status in the literature on self-

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 118 incompatibility (see Lloyd 1965; Bawa 1974; Lloyd & Schoen 1992; Raduski et al. 2012).

I used this index in an effort to control for the positive correlation I detected between self- and cross-fertilized fruit set that was evident for individual plants. The basis of this association is unclear but may simply reflect variation among plants in their overall vigor, as reflected in their ability to set fruit from cross-pollination. Several earlier studies have shown that the number of developing fruit on a plant can influence incompatibility expression (see Good-Avila & Stephenson 2008 and references therein). In most cases, the presence of prior outcrossed fruit tended to reduce self-fertilized seed set (Vogler et al.

1998; Stephenson et al. 2003; Travers et al. 2004) whereas in contrast higher outcrossed fruit was associated with increased PSI in my study. Elevated expression of PSI in the absence of fruit production probably occurs because floral longevity is prolonged when no resources are allocated to fruit, thus enabling slower self-pollen tubes to fertilize ovules

(Good-Avila & Stephenson 2008). In my experiments, all self- and cross-pollinations on plants were conducted on the same day using pairs of one-day flowers on the same inflorescence. Therefore, it seems unlikely that floral longevity and prior fruit development played any significant role in governing the patterns of self- and cross-fruit set that I recorded.

In common with previous pollination studies of L. salicaria (reviewed in Colautti et al. 2010a; see Fig. 4.2) I detected a higher expression of PSI in the M-morph than in the L- and S-morphs. This pattern was reflected in a higher proportion of M-morph plants setting self-fertilized fruit and also in the number of self-fertilized fruit per plant, but not in the number of seeds per self-fertilized fruit (Fig. 4.2). The morph-specific differences in PSI that were evident in my population samples were also maintained in my parent-offspring

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 119 comparisons of PSI involving plants of the M-morph heterozygous at the M locus. Among segregating families, PSI was more strongly expressed in the M-morph in comparison with the L-morph (Fig. 4.6). This pattern was also found between parents and offspring in

Pontederia cordata, a species in which the M-morph possesses significantly weaker incompatibility than the L- and S-morphs (Barrett & Anderson 1985). These authors proposed that morph-specific differences in PSI may arise from pleiotropic effects of genes that directly control the expression of tristylous characters (e.g. style length, anther height, pollen size), although differences in PSI among genotypes of the M-morph as I observed in L. salicaria seem likely to also involve modifier genes, perhaps linked to the dominant M allele(s) at the M locus.

Self-compatibility in L. salicaria occurs most frequently in pollinations involving pollen from the long-level anthers (Darwin 1877), which is a factor held in common with other tristylous species that exhibit weak incompatibility expression in the M-morph (e.g. see Barrett 1977; Glover & Barrett 1983; Barrett & Anderson 1985; Bianchi et al. 2000;

Puentes et al. 2013). Self-pollinations with pollen from the alternate short-level anthers of the M-morph are largely incompatible (Darwin 1877; O’Neil 1994). Because pollen from the two anther levels within a flower is genetically identical, exhibiting the same range of pollen genotypes, the difference in its ability to self-fertilize ovules is probably determined by specific aspects of the developmental environment of stamens and their influences on pollen and pollen-tube growth.

Insights into the mechanisms governing differential incompatibility expression in tristylous species come from studies of pollen-tube growth following self-pollination of

Pontederia sagittata (Scribailo & Barrett 1991), which exhibit PSI in the M-morph

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 120

(Glover & Barrett 1983). Small-sized pollen from short-level anthers of the M-morph failed to reach ovaries probably because of insufficient storage reserves; whereas in contrast, larger-sized pollen from long-level anthers grew much longer distances down mid styles past ovaries and may not have been competent to respond to ovular signals in the same manner as legitimate (compatible) pollen tubes. However, it seems unlikely that in L. salicaria pollen size alone can explain the differences in siring ability between small- and large-sized pollen following self-pollination of the M-morph. Pollen from long-level anthers is much larger than the remaining pollen sizes and there is a high degree of overlap in size between pollen from short- and mid-level anthers, the latter being the source of compatible pollen (Costa et al. 2017; Fig. 4.2). Future studies of pollen-pistil interactions in L. salicaria would be valuable to determine whether pollen-size variation and the dynamics of pollen-tube growth play an important role in causing PSI in the M-morph.

In contrast to rigid self-incompatibility, in which self-rejection is usually strongly expressed under most environmental conditions, PSI can be modified by diverse environmental and developmental factors (de Nettancourt 1977; Levin 1996; Good-Avila et al. 2008 and references therein). As a result, non-genetic factors can result in considerable variation in the expression of PSI both within and between plants. This variation was evident in my comparison of PSI for plants pollinated in successive years in the glasshouse. Despite the significant repeatability in PSI values that I measured, there was considerable scatter in the data obtained over the two years with some individuals setting high counts of fruit set in 2013 but low counts in 2014 (Fig. 4.3A). This variation may have been associated with differences between years in glasshouse conditions,

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 121 particularly temperature and humidity, despite my efforts to minimize these sources of variation.

To investigate ecologically relevant changes in growing conditions that may alter the expression of incompatibility I performed a glasshouse experiment with 12 genotypes exhibiting a wide range of PSI values. Just prior to inflorescence initiation, clones of these genotypes were either maintained in flooded growing conditions or were subjected to water stress. I specifically chose these treatments because my field observations in Ontario have indicated that whereas some populations grow under permanently flooded conditions others, particularly those in roadside ditches, experience drought in August during flowering. However, I found no systematic change in the expression of PSI under drought as I may have expected if stress impaired the functioning of incompatibility causing higher quantities of selfed seed set. In fact, PSI was remarkably stable to the different growing conditions both for genotypes with high and low indices of self-compatibility (Fig. 4.4). In general, the degree to which PSI was expressed in individual clonal groups reflected their genotypic identity, as opposed to the wet and dry environments in which they were pollinated, reflecting genetic variation for PSI among the sample of plants in the experiment. My failure to induce changes in PSI may of course reflect my choice of environments and it would be premature to conclude that the compatibility status of genotypes is impervious to environmental manipulation, especially given the results of my repeatability study.

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 122

Partial self-incompatibility, colonization, and floral morph ratios

My initial sample of 338 L. salicaria plants originated from seed families collected from four populations in the Metropolitan region of Toronto, Ontario, Canada. I had no a priori expectation that there would be significant differences among the populations in the frequency of plants with PSI. Indeed, this is what I found with approximately 30–40% of plants in each population exhibiting PSI. Lythrum salicaria is common in this region with numerous populations and gene flow among them may limit opportunities for genetic divergence in their reproductive systems. A more extensive survey of PSI by Colautti et al. (2010a) in 12 populations of L. salicaria spanning much of the latitudinal range in eastern North America found significant variation among populations in PSI (mean frequency of PSI = 0.30; range 0–0.75). However, there was no obvious geographical pattern in the distribution of PSI providing no evidence that selfing potential increased towards the northern range edge associated with the known invasion history of the species.

Floral morph ratios in populations of tristylous species are largely governed by the interaction of stochastic processes and negative frequency-dependent selection. Extensive surveys of several hundred populations of L. salicaria in native European and introduced

North American populations have revealed difference between the continents in floral morph structure (Halkka & Halkka 1974; Eckert & Barrett 1995; Ågren and Erickson

1996; Eckert et al. 1996a; Balogh & Barrett 2016; Costa et al. 2016). Whereas tristylous populations predominate in Europe (~95%), dimorphic populations are more common in the invasive North American range (~25%). Significantly, the vast majority of these dimorphic populations are comprised of the L- and M-morph. Stochastic simulations of finite size indicate that genetic drift results in asymmetric morph loss with the S-morph

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 123 being lost most often and the L-morph least often from tristylous populations (Heuch

1980; Barrett 1993; Eckert & Barrett 1995). This pattern is a consequence of the genetic system controlling tristyly in which the dominant S-allele is only carried by the S-morph and is therefore at a lower frequency in equilibrium populations than the other three alleles

(s, M, m) at the loci (S, M) governing the polymorphism (Heuch & Lie 1985). Genetic drift is a pervasive feature of the population biology of colonizing species and thus provides a satisfying explanation for the observed pattern of style length dimorphism in invasive populations of L. salicaria, and other tristylous species (e.g. Eichhornia paniculata –

Barrett et al. 1989; Husband & Barrett 1992a) that experience frequent population bottlenecks.

Genetic drift has undoubtedly played a role in the loss of the S-morph from invasive populations of L. salicaria; however, an alternative process of founder events involving M-morph individuals with PSI may also be involved (Eckert & Barrett 1992;

Colautti et al. 2013; Balogh & Barrett 2016). Owing to tetrasomic inheritance in L. salicaria (Fisher & Mather 1943), 78-89% of plants in equilibrium populations should be simplex at the M-locus, depending on the rate of double reduction at this locus (Heuch &

Lie 1985). Upon selfing, the offspring of these plants segregate both M- and L-morphs in a

3:1 ratio. Of the 14 genotypes used in my parent-offspring analysis, 12 segregated both L- and M-morph progeny and the overall ratio was not significantly different from the expected 3:1 ratio, although individual families were not large enough to rule out other heterozygous tetrasomic genotypes. Nevertheless, founder events involving any of the 12 segregating parental genotypes would be sufficient to establish dimorphic populations which lack the S-morph. Most importantly my current experimental work on selfed and

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 124 outcrossed progenies of the M-morph exhibiting PSI indicates significant inbreeding depression under glasshouse and field conditions (Chapter 6). It is therefore likely that at least some dimorphic populations of L. salicaria in the invasive range may have originated from colonization events involving genotypes of the M-morph with PSI.

Evolutionary significance of partial self-incompatibility

Variation in PSI within populations may provide the necessary genetic substrate for the selection of complete self-compatibility (Levin 1996), perhaps leading to the evolution of autogamy. To my knowledge, there is no evidence for the complete loss of trimorphic incompatibility in L. salicaria, although sporadic self-compatible homostylous variants have been reported (Stout 1925; C.G. Eckert pers. com.). Elsewhere in the genus Lythrum, small-flowered, self-compatible homostylous species occur and have likely evolved from tristylous ancestors (Charlesworth 1979; Weller 1992), although this would need to be confirmed by phylogenetic studies. Thus, although L. salicaria is a highly successful invasive species and recurrent colonizing episodes are a characteristic feature of its population biology, trimorphic incompatibility appears to be remarkably resilient to evolutionary breakdown, as occurs in many other tristylous taxa, including several that are highly invasive (reviewed in Weller 1992).

Several factors may help to explain this apparent paradox. First, although several lines of evidence from my investigations indicate that PSI has a genetic component, the heritability of this trait appears to be quite low and the amount of standing genetic variation in populations may not be sufficient for a sustained response to selection. A study of artificial selection of PSI would be valuable to test this hypothesis and to

CHAPTER 4: PARTIAL SELF-INCOMPATIBILITY 125 determine if fully self-compatible lines of L. salicaria can be established. Second, several features of the life history of L. salicaria may serve to ameliorate the demographic stochasticity and Allee effects associated with colonization that could potentially limit reproductive success. These include plant longevity, extended mass flowering, high pollinator visitation rates and fecundity. Although small populations of L. salicaria certainly experience some pollen limitation (Ågren 1996), this may rarely be sufficiently chronic to favour persistent selection for the complete breakdown of trimorphic incompatibility. Thirdly, selection for self-compatibility is strongly influenced by the intensity of inbreeding depression (Lande & Schemske 1985; Lloyd 1992). A 4-year experimental study of differences in fitness between outcrossed and selfed offspring of L. salicaria indicates that the magnitude of inbreeding depression may be sufficient to limit spread of mating-system modifier genes in populations (Chapter 6). Finally, PSI may provide a ‘best-of-both-worlds’ mating strategy in which some selfing at low density enables colony foundation, but that predominant outcrossing eventually replaces selfing once population growth occurs and pollinator service becomes reliable (see Pannell 2015).

In many tristylous populations of L. salicaria, PSI may have relatively little influence on mating patterns, particularly if selfed pollen tubes are less competitive than outcrossed, as occurs in most species with PSI (Levin 1996). But when periodic low-density conditions favour reproductive assurance, partial self-incompatibility may be adaptive thus explaining why this feature of trimorphic incompatibility appears to be a pervasive feature of L. salicaria populations.

CHAPTER 5

THE INFLUENCE OF FLORAL MORPH RATIOS AND LOW PLANT DENSITY ON

MATING AND FERTILITY IN A TRISTYLOUS COLONIZING SPECIES

Abstract

Sexual reproduction in heterostylous populations may be vulnerable to demographic conditions because of the small number of mating types in populations. Here, I investigate mating and fertility under natural and experimental conditions in tristylous Lythrum salicaria, an invasive species that exhibits a wide range of floral morph ratios and demographic contexts. I grew 147 open-pollinated seed families from six populations with different morph structures to estimate inter-morph mating (d). In a field experiment, I used progeny ratios from 47 spatially isolated individuals to estimate d, and measured the intensity of pollen limitation experienced by the morphs. The M- and S-morph experienced high rates of d, regardless of population size or morph ratio. Estimates for the

L-morph revealed low levels of intra-morph mating in three dimorphic and two trimorphic populations but near complete intra-morph mating in a monomorphic population. Despite high levels of inter-morph mating in the field experiment the morphs experienced significant pollen limitation of fruit and seed set but this did not differ in intensity among the morphs. My field experiment demonstrates that although plant isolation was associated with pollen limitation of seed set, ‘long-distance’ bee-mediated pollen flow served to maintain inter-morph mating. Tristyly in L. salicaria is remarkably robust to demographic variation associated with colonization.

126 CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 127

Introduction

In populations of self-incompatible plants, the ecological and demographic context in which mating occurs and the availability of cross-compatible mates play key roles in determining the number and genetic quality of offspring. In animal-pollinated species, small population size and low density can limit reproductive success owing to a lack of pollen vectors resulting in pollen limitation of seed set (Ågren 1996; Larson & Barrett

2000; Ashman et al. 2004). These conditions can induce Allee effects on sexual reproduction and may favour the selection of self-fertilization (Allee 1931; Lloyd 1980;

Barrett 2011; Pannell 2015). Despite the immobility of plants and the potential vulnerability of their reproductive ecology to the number and identity of neighbours, I have only limited understanding of how the social context in which plants occur influences both their mating and fertility.

Mating in heterostylous populations can be particularly vulnerable to demographic influences. Unlike species with homomorphic incompatibility, in which populations usually contain numerous mating types, heterostylous populations most commonly contain only two or three cross-compatible mating types (Darwin 1877; Ganders 1979; Barrett &

Shore 2008). The mating types (floral morphs) differ in sex-organ length and are usually self- and intra-morph incompatible. Negative frequency-dependent selection operating in large populations of heterostylous species should result in equal frequencies of the floral morphs (isoplethy); however, a variety of ecological and demographic factors can cause deviations from isoplethy with potential consequences for mating and fertility (Ishihama et al. 2003; Wang et al. 2005; Stehlik et al. 2006). Deviations from isoplethy are especially likely in heterostylous species with well-developed colonizing ability in which founder

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 128 events and restrictions on sexual recruitment may contribute to biased morph ratios (e.g.

Eichhornia crassipes, Barrett & Forno 1982; Oxalis pes-caprae, Ferrero et al. 2015). The principal goal of this study is to investigate the influence of morph ratio bias and low plant density on mating and fertility in an invasive heterostylous species. My study system for this investigation is Lythrum salicaria which exhibits a wide range of population sizes and floral morph structures.

Mating patterns in heterostylous populations are rarely investigated as it is generally assumed that the floral morphs are self-incompatible and all mating is between the floral morphs (inter-morph or disassortative mating). However, in several heterostylous species, incompatibility is only partially expressed, and in some cases can be absent altogether or unlinked to floral morph type (reviewed in Barrett & Cruzan 1994).

Mating system estimates in heterostylous species that do not possess conventional heteromorphic incompatibility have revealed a wide range of mating patterns including variable selfing rates and significant levels of intra-morph (assortative) mating (e.g.

Amsinckia – Ganders 1975; Eichhornia paniculata – Barrett & Husband 1990; Decodon verticillatus – Eckert & Barrett 1994a; Narcissus triandrus – Hodgins & Barrett 2006,

2008; Oxalis alpina – Weber et al. 2013; Luculia pinceana – Zhou et al. 2015). In such cases, deviations from symmetrical disassortative mating can result in biased morph ratios

(reviewed in Barrett & Hodgins 2006). Although most estimates of mating in heterostylous populations have used allozymes or microsatellite markers, a few studies have used the style-length loci themselves as genetic markers to estimate inter- and intra- morph mating by progeny testing open-pollinated seed families (e.g. Ganders 1975;

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 129

Barrett 1979; Barrett et al. 1987). I used this approach here because L. salicaria exhibits tetrasomic inheritance complicating the use of allozymes or microsatellite markers.

A neglected aspect of the population biology of colonizing species concerns the consequences of spatial isolation for mating and fertility. Most studies in reproductive ecology focus on populations of moderate to large size and the ecology and genetics of

‘plant outliers’– plants that occur outside population boundaries– are not well characterized (reviewed in Levin 1995). In particular, the extent to which spatial isolation may be buffered by ‘long distance’ pollen flow, either from larger populations or from other isolated individuals, is rarely investigated, aside from studies investigating gene flow using genetic markers in tropical forest trees where conspecific individuals are usually spatially isolated and at low density (Nason et al. 1998; White et al. 2002; Ashley 2010).

If cross-pollination is compromised by low density through unreliable pollinator service increased rates of self-pollination may be likely. However, the extent to which pollen limitation influences mating and fertility will depend on the compatibility status of individuals in populations (Lloyd & Schoen 1992), and comparative evidence indicates that self-incompatible species are more vulnerable to pollen limitation than self- compatible species (Larson & Barrett 2000). One of the objectives of this study is to investigate whether morph-specific differences in the expression of incompatibility may differentially affect the reproductive performance of L. salicaria plants occurring at low density.

Here, I investigate mating patterns and fertility in tristylous Lythrum salicaria, an invasive herb that exhibits a wide range of population sizes and floral morph ratios in its adventive range of Ontario, Canada (Eckert &Barrett 1992; Balogh & Barrett 2016). I

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 130 examined two demographic contexts, natural population variation in morph ratios and a field experiment involving plant isolation, to examine the following specific questions: 1)

To what extent are levels of inter-morph (disassortative) mating affected by the floral morph structure of populations? Specifically, I contrasted progeny morph ratios sampled from parents in trimorphic and dimorphic populations with expectations for inter- and intra-morph mating to assess whether floral morph loss from populations is associated with deviations from inter-morph mating. In addition, I also progeny tested a small monomorphic population comprised of the L-morph in which I observed seed set to determine whether mating was predominantly intra-morph and/or whether ‘long-distance’ pollen flow from other populations could account for seed production. 2) What is the influence of low plant density on mating and fertility? I performed an experiment with spatially isolated plants of the three floral morphs in two old fields separated by forest and ponds at a field station to determine whether fruit and seed set was pollen limited and, if so, whether pollen limitation varied among the morphs. Using progeny tests of the isolated plants, I also assessed the extent to which inter-morph gene flow may alleviate any pollen limitation.

Materials and Methods

Study system

Lythrum salicaria (Lythraceae) is a showy-flowered, herbaceous, amphibious perennial native to Eurasia. It has been introduced to North America where it colonizes a wide range of wetland habitats including freshwater marshes, river edges, low-lying pastures and roadside ditches. Flowers of L. salicaria are visited by diverse pollinators, particularly

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 131

Bombus spp. and Apis mellifera, but also small pollen-collecting bees, wasps, butterflies and occasional hummingbirds (Thompson et al. 1987; Brown et al. 2002; King & Sargent

2012). Populations of L. salicaria are commonly tristylous and the species possesses a trimorphic incompatibility system which generally prohibits self- and intra-morph fertilization. However, populations commonly exhibit partial self-incompatibility, with the

M-morph displaying the weakest incompatibility and the S-morph the strongest, based on controlled self- and cross-pollinations (Darwin 1877; Colautti et al. 2010a; Chapter 4). At present, it is unclear whether variation in partial self-incompatibility influences mating patterns in populations of L. salicaria. Despite a considerable literature on the reproductive ecology of L. salicaria, there have been no studies on the mating system of the species.

Plants of L. salicaria produce very large numbers of small, easily dispersed seeds

(thousands to 2.5 million per plant) resulting in prolific colonizing ability (Thompson et al. 1987). The species is perennial, with some individuals living over 12 years, but reproducing exclusively by seed (Yakimowski et al. 2005). As a consequence, the species displays a wide range of population sizes from isolated individuals and very small colonies to large, monospecific, high density stands containing many thousands of plants (Eckert &

Barrett 1992). Variation in population size in L. salicaria is associated with pollen limitation of seed set with smaller populations experiencing more intense pollen limitation than larger populations (Ågren 1996). Extensive surveys of style-morph ratios in populations of L. salicaria in Europe and eastern North America indicate significant differences between the two continents. In France and the Iberian Peninsula, ~95% of L. salicaria populations (n = 198 populations) are trimorphic (Eckert et al. 1996a; Costa et

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 132 al. 2016). In contrast, surveys of populations in eastern North America (n = 216) indicate that ~25% of populations are dimorphic with the vast majority of dimorphic populations composed of the L- and M-morphs (Eckert & Barrett 1992; Balogh & Barrett 2016).

Occasional populations containing a single floral morph are also reported from Ontario

(Balogh & Barrett 2016), raising the question of whether these populations are self- sustaining through sexual reproduction and to what degree intra-morph incompatibility limits mating and fertility.

The inheritance of alleles determining style morph in Lythrum salicaria was first enumerated by Fisher and Mather (1943), based on controlled crosses among the floral morphs. They demonstrated that two unlinked loci (S and M), with the S-locus epistatic to the M-locus, govern inheritance and that because L. salicaria is an autotetraploid, the loci exhibit tetrasomic inheritance. Fisher (1941a) calculated the gamete frequencies produced by autotetraploid individuals of different genotypes and predicted the genotype frequencies one would expect to find in a population at isoplethic equilibrium. Fisher and

Mather (1943) extended this work to include the phenomenon of double reduction, in which sister chromatids are inherited by the same gamete at various rates (Darlington

1929; Bever & Felber, 1992). The frequencies of double reduction at the M- and S-loci of

L. salicaria have been estimated to be approximately 10% and 2.5%, respectively (Fisher

1949; Fyfe 1953). Furthermore, rather than possessing homozygous/heterozygous allelic states as in diploid organisms, autotetraploid L. salicaria can possess between zero to four copies of a dominant or recessive allele, resulting in a total of 25 possible genotypes at the

S and M loci (Fisher 1941a; Appendix 5 Table 1). At equilibrium, by far the most common genotypes are nulliplex or simplex (i.e. possessing zero or one dominant allele at a locus),

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 133 whereas only seven percent of the individuals possess duplex genotypes (which possess two dominant alleles at a locus) or higher (Heuch & Lie 1985). As outlined below, the occurrence of tetrasomic inheritance and double reduction in L. salicaria needs to be taken into account when inferring maternal genotypes for progeny testing.

Progeny testing of maternal seed families from natural populations

In summer 2015, I sampled an average of 24.5 (range 4 - 34) maternal seed families from six L. salicaria populations in southern and central Ontario (Fig. 5.1). The populations were originally located in summer 2013 as part of an extensive survey of floral morph ratios in L. salicaria populations in Ontario (Balogh & Barrett 2016). In each population, I estimated census number (population size) and floral morph ratios using methods previously described in detail in Balogh and Barrett (2016) (Table 5.1). The populations were chosen to include the complete range of population morph structures (trimorphic – 2 populations; dimorphic – 3 populations; monomorphic – 1 population). At peak flowering in each population (July – August), I used aluminium forestry tags with unique identifiers to label the floral morphs of plants to be progeny tested. In late October - early November

2015, I revisited the six populations and harvested six mature fruit from each tagged plant.

The fruits were sampled from a single inflorescence with a special effort made to include fruits from throughout the inflorescence (bottom to top) to maximize the range of

‘pollination environments’ experienced by each plant. I allowed fruits to dry in open

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 134

Eppendorf tubes at 6 degrees Celsius until April 2016 to break seed dormancy.

Figure 5.1. Locations and morph ratios in populations of Lythrum salicaria sampled in

Ontario, Canada for mating analyses. One population is monomorphic for the L-morph

(153), two are dimorphic for the L- and M-morphs (68, 100), one is dimorphic for the L- and S-morphs (84), and two are trimorphic (92, 135). Map tile from Stamen (Stamen

Design, available from http://maps.stamen.com; accessed 11 February 2018 via GGMAPS in R v 3.3.2, Kahle and Wickam 2013).

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 135

Table 5.1. The floral morph structure and morph ratios of the six populations of Lythrum salicaria in Ontario that were used in progeny testing. I also present P-values of Χ2 tests comparing morph ratios to the isoplethic equilibrium.

P-value of Χ2

Population Structure L M S n plants deviation from

isoplethy

153 L 1 0 0 8 NA

68 L, M .49 .51 0 187 > 0.70

100 L, M .59 .41 0 47 > 0.25

84 L, S .33 0 .66 108 < 0.01

92 L, M, S .16 .27 .56 73 < 0.001

135 L, M, S .48 .05 .48 30 < 0.05

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 136

In April 2016, I sowed the seed from each fruit into Promix BX soil (Promix

Company, Rivière-du-Loup QC, Canada) and in mid-May transplanted seedlings into SC7 sized Ray Leach “Cone-tainers” which are 3.81 cm wide and 13.97 cm deep (Stuewe and

Sons, Inc., Tangent OR, USA) filled with Promix BX soil. The mean number of seedlings grown to flowering for each maternal family was 17 (range: 2 - 37) and depended on germination success, which was generally high for most families. The sample sizes for families and plants per floral morph for each population are presented in Table 5.1. I added a 20-20-20 fertilizer with micronutrients on a biweekly basis following the manufacturer’s instructions. I recorded the date of flowering and floral morph of each plant.

Estimation of inter-morph mating in natural populations

I estimated morph-specific rates of inter-morph mating (= disassortative mating or d) in each population using the morph ratios of open-pollinated seed families weighted by the number of offspring produced by each of the genotypes within each morph (Appendix 5

Table 2). I estimated d in my study (and its inverse, 1-d, intra-morph mating = assortative mating) rather than outcrossing (and selfing) because the genetics of tristyly make it difficult or impossible (depending on maternal genotype) to distinguish assortative mating from selfing. I adapted my methods from those detailed in Barrett et al. (1987), where only the diploid case with linkage between the M- and S-loci was considered, and I extended these methods for the tetrasomic case with unlinked M- and S-loci, as found in L. salicaria (Fisher and Mather 1943). I represented genotype frequencies as elements in a data vector in R where each element contained the frequency of that genotype in the population or progeny. I assigned which element defined each genotype using the

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 137 genotype numbers in Appendix 5 Table 1. I defined j as the paternal genotype, i as the maternal genotype, crossfreq as a 25x25 matrix initially containing zeros which expresses the probability of cross-fertilization between genotype i and genotype j, and Fi or Fj as the frequency in a population of a given genotype i or j. I set the variable mat.morph to represent all genotypes, x, which define the floral morph expressed by genotype i – these are genotype 1 for L-morph parents, genotypes 2-5 for the M-morph, and genotypes 6-25 for S-morph – when i equals one of the genotypes in each of these categories, then mat.morph will represent i and the other genotypes in that morph. I then used the equation:

1 1 5.1 푐푟표푠푠푓푟푒푞푖,푗 = 퐹푗 ∗ ∗ 퐹푖 ∗ ; (1−∑푥=푚푎푡.푚표푟푝ℎ 퐹𝑖) (∑푥 = 푚푎푡.푚표푟푝ℎ 퐹𝑖) to calculate the mating frequency between each compatible genotype in the population.

Equation 5.1 only defines inter-morph mating and therefore the frequencies of intra-morph mating equal zero on the crossfreq matrix. I calculated the frequency with which each genotype produced a given gamete following Fisher and Mather (1943) using the estimated double reduction rates of α = 0.10 at the M-locus and α = 0.025 for the S-locus

(Fisher 1949; Fyfe 1953). I summed these values following procedures detailed in Fisher and Mather (1943) to establish a 25x25x25 array (progfreq) in which dimensions one and two represent paternal and maternal parents and the third dimension represents the probability that the parents produced a given offspring genotype. I used progfreq and crossfreq to determine the expected number of offspring from each genotype after inter- morph mating. I created a third matrix, progeny, which is a 25x25 matrix, to represent the output of each progeny genotype k from each maternal genotype i crossed with all paternal parents of genotype j. I filled this matrix using the equation:

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 138

5.2 푝푟표푔푒푛푦[푖, 푘] = ∑(푓표푟 푎푙푙 푝푎푡푒푟푛푎푙 푔푒푛표푡푦푝푒푠 푗) 푐푟표푠푠푓푟푒푞[푖, 푗] ∗ 푝푟표푔푓푟푒푞[푗, 푖, 푘]; using the expected predicted genotypes, I then calculated the expected number of each floral morph in the progeny. To calculate the progeny ratio expected after intra-morph mating, I repeated the steps above except I defined the genotypes x of mat.morph to be all genotypes of the same morph as i.

I inferred the genotypes of parental plants using a combination of floral morph identity and progeny ratios. By definition, L-morph plants are nulliplex at the M- and S- loci and could therefore be assigned to genotype 1. For each family of the M- and S- morph, I performed a Χ2 test comparing the progeny morph ratios expected for ‘simple’

M- and S-morph genotypes, which consist of M-morph plants simplex at the M-locus

(genotype 2) or S-morph plants nulliplex at the M-locus (genotype 6). I also compared the observed progeny morph ratios in M- and S-morph plants to more ‘complex’ expectations, which consist of M-morph plants duplex at the M-locus (genotype 3) or S-morph plants simplex at the M-locus (genotype 7). I chose these genotypes for my comparisons because prior studies (Fisher 1941a; Fisher & Mather 1943; Heuch & Lie 1985) indicate that genotypes 2, 3, 6, and 7 represent ~95% of the M- and S-morph plants in autotetraploid, tristylous populations at isoplethic equilibrium. My simulations of dimorphic and non- equilibrium populations demonstrated that triplex and higher dominant genotypes were rare with complete inter-morph mating and none of the observed progeny morph ratios suggested triplex maternal genotypes. I assigned a genotype to each plant based upon which prediction was the best fit to the data according to the Χ2 test (Appendix 5 Table 3).

In total, 109 plants had genotypes that were clearly assigned; 64, 29, and 16 of the L-, M-

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 139 and S-morph. I estimated genotype frequencies in the populations using these genotypes, and then using the identities assigned from the first run of this test, I repeated the Χ2 comparisons to ensure that the progeny ratios did not deviate from the predicted genotypes to any substantial degree.

I pooled the progeny of families of each floral morph in each population to estimate the frequency of inter-morph mating (Table 5.2). I obtained the predicted frequency of L-, M-, and S-morph progeny from each genotype in the L- and M-morphs after inter-morph or intra-morph mating (Xb and Sb respectively, where b represents the progeny floral morphs L-, M-, and S-morph).I then entered these values along with the

2 observed progeny morph counts (obsb) into the following equation, which outputs the Χ fit based on d values:

2 2 ((푆푏 + (푋푏 − 푆푏) ∗ 푑) − 표푏푠푏) 5.3 훸푑 = ∑(푓표푟 푏 푤ℎ푒푟푒 표푏푠 >0) ; 푏 표푏푠푏

I used a modified version of this equation for families of the S-morph, with n as the total number of progeny produced by the S-morph maternal parents and subscripts L, M, and S representing the L-, M-, and S-morphs respectively:

2 2 2 (1/4 ∗ (푛) ∗ (1 +푑) − (표푏푠퐿 + 표푏푠푀)) (푛 ∗ (3 −푑)/4 −표푏푠푆) 5.4 훸푑 = + ; (표푏푠퐿 + 표푏푠푀) (표푏푠푆)

I translated these equations into a script for the “mle” function in the R package “stat4” in

R version 3.3.2 (R Core Team, 2016) which uses the log-likelihood values of the Χ2 to predict the mean and 95% confidence intervals of d based on the observed and predicted morph ratios. I present the predicted offspring which I used for the MLE function in Table

5.2.

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 140

Table 5.2. Patterns of mating in six populations of Lythrum salicaria based on progeny testing of open-pollinated seed families.

Provided are the expected morph ratios produced by the floral morphs in each population after intra-morph and inter-morph mating as well as the observed progeny morph ratios. All predicted morph ratios were estimated using the assumption that double reduction equaled 0.10 at the M-locus and 0.025 at the S-locus and were weighted by the progeny produced per genotype in each floral morph.

Intra-morph Inter-morph mating mating Observed progeny expectation expectation

Population Structure Morph L M S L M S L M S n

153 L L 50 0 0 0 0 0 48 1 1 50

68 L, M L 384 0 0 195 189 0 194 190 0 384

68 L, M M 105 293 0 202 196 0 202 195 1 398

100 L, M L 277 0 0 123 154 0 108 169 0 277

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 141

100 L, M M 50 145 0 98 98 0 88 106 1 195

84 L, S L 119 0 0 55 5 59 68 14 37 119

84 L, S S 18 12 86 43 15 57 53 11 51 115

92 L, M, S L 103 0 0 34 35 34 44 37 22 103

92 L, M, S M 23 90 0 24 46 43 19 44 50 113

92 L, M, S S 27 34 178 45 76 118 62 72 105 239

135 L, M, S L 188 0 0 95 8 84 107 19 62 188

135 L, M, S M 6 14 0 8 7 5 12 8 0 20

135 L, M, S S 43 0 124 81 4 82 86 8 99 193

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 142

Mating and fertility of isolated plants

In the summer of 2014 I selected 19 plants of each floral morph for use in a field experiment at the Koffler Scientific Reserve (KSR) near Newmarket, Ontario, Canada

(44.0309 N, 79.5389 W). The plants were previously screened for partial self- incompatibility in the glasshouse and exhibited the typical pattern in which plants of the

M-morph set variable numbers of fruit after self-pollination and the L- and S-morphs set few if any fruit (Chapter 4). This treatment provided the highest likelihood of detecting differences in reproductive output between morphs owing to PSI. I grew these individuals in the glasshouse in 21.6 cm diameter by 21.6 cm height (8.5 by 8.5 inches) pots in Promix

BX soil with Osmocote slow-release fertilizer (The Scotts Miracle-Gro Company,

Marysville OH, United States) added as per the manufacturer’s instructions until they were transferred to the field.

I moved the 57 plants to KSR in early July 2014 and placed them individually 50-

85m apart along mowed paths in two large old fields. I transferred plants in these pots into

35.5 cm diameter x 26.3 cm height (14” x 10.4” inches) pots that were sunk into the soil for stability and surrounded the outer pots with clear plastic to maintain moisture levels. I evenly assigned plants of the L-, M-, and S-morph among the 57 locations distributed across the two fields and randomly assigned individuals of the appropriate morph to each location (Appendix 5 Fig. 1). Before the experiment commenced, I removed any inflorescences that had developed prior to the plants being placed in the field.

On each plant I tagged two inflorescences and marked ~20 flowers on each with yellow or red paint at their base depending on treatment. A subset of flowers on each

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 143 inflorescence were marked with yellow paint and allowed to be naturally pollinated (mean per inflorescence: 20.4 flowers range: 5-45). On one inflorescence of half of the plants I additionally conducted hand-pollinations using a mixed compatible pollen donors (hand- pollination: mean per inflorescence 14.2 flowers, range: 6–20) and marked these flowers with red paint to test for pollen limitation. In late August, I removed all flowers from the plants before moving them back to a glasshouse at the University of Toronto. Ten plants did not survive the summer and are not considered further. Once fruit were mature I recorded the number of flowers in each treatment that had produced fruit and these were then harvested. Overall, I harvested 1216 open-pollinated fruit and 280 hand-pollinated fruit; a mean of 25.9 open-pollinated fruit per plant on 47 plants (15 L-, 16 M-, 16 S- morph) and 12.2 hand-pollinated fruit per plant on 23 plants (7 L-, 8 M-, and 8 S-morph).

I recorded the number of seeds produced in each fruit. Based on the marked open- and hand-pollinated flowers on inflorescences which I hand-pollinated, I calculated the percent fruit set, the mean number of seeds set per fruit, and a multiplicative index of percent fruit set times mean seed set per fruit after open- and hand-pollination. I then calculated an index of pollen limitation (PL) following (see Larson & Barrett 2000):

5.5 푃퐿 = 1 – 푝표; 푝푠

where po and ps are the measure of fruit or seed set in marked open-pollinated and hand- pollinated flowers for each plant and PL is an index ranging from 0 – 1 representing no pollen limitation (where there was no difference in the measure of fertility between open- pollinated and hand-pollinated flowers) to complete pollen limitation (where open- pollinated flowers set no seed) in the response variable. I set all negative values from this

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 144 index to 0 because these results most likely resulted from experimental or statistical error

(see Young & Young 1992; Burd 1994). I compared these pollen limitation indices for each output variable between the three floral morphs using ANOVA. Finally, because I did not hand cross-pollinate every plant (see above), I collected open-pollinated fruit and seed set per fruit from all marked plants and inflorescences in the field (~40 marked flowers per plant). I tested the potential effects of hand pollination on resource allocation to open-pollinated fruits, as proposed in Zimmerman and Pyke, (1988), by comparing the percent flowers setting fruit and the mean seeds per fruit between shoots on plants that received hand-pollination and those that did not [see Ågren & Ericson (1996) for a similar treatment on L. salicaria]. There was no significant difference in reproductive measures between these treatments (percent fruit set: P > 0.10; mean seed per fruit: P > 0.25) indicating that resource reallocation was not a significant factor in my experiment. For all plants in the study I measured the percent of marked open-pollinated flowers that set fruit and the mean seed set per open-pollinated fruit in each of the floral morphs and compared these values among the morphs using ANOVA.

I inferred mating between plants in the experiment using progeny testing of L- and

M-morph families. I chose these morphs because unambiguous inferences concerning minimum levels of inter-morph mating can be obtained from genotypes of these morphs but not the S-morph. I germinated seeds from 9 L-morph and 9 M-morph families, obtaining an average of 18 progeny from each plant (range: 8 - 30). When the 18 families flowered, I recorded the floral morph of each plant. I performed heterogeneity Χ2 tests on morph ratios of families grown from each parental morph to detect if there were differences among individual plants in progeny ratios. I checked L-morph families for M-

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 145 and S-morph progeny and M-morph families for S-morph progeny, which in both cases unambiguously indicate inter-morph mating.

Results

Inter-morph mating in natural populations

Of the 147 open-pollinated families from six natural populations of L. salicaria used to estimate inter-morph mating, I unambiguously identified the parental genotypes of 109 using Χ2 tests (Appendix 5 Table 3). My maximum likelihood estimates of inter-morph mating in trimorphic and dimorphic populations in general were not substantially different from expectations predicted by complete inter-morph mating and there was no significant difference between the M- and S-morph in estimated values of d (Fig. 5.2). However, there were several cases where deviations from inter-morph mating were evident. In populations 84, 92, and 135 there was a significant excess of L-morph progeny from L- morph plants resulting in mean estimated values of d = 0.72 CI = 0.61 – 0.83; d = 0.83 CI

= 0.73 – 0.93; and d = 0.81 CI = 0.72 – 0.92 in each population, respectively, with the

95% CI not overlapping 1.0 in any of these cases. In population 100, I detected an excess of M-morph progeny in L-morph families resulting in an estimate of d slightly greater than

1 (d = 1.09, CI = 1.02 – 1.17). In populations 84 and 92, the 95% CI of d for the L-morph did not overlap with the 95% CI for the M- and S-morphs. In the monomorphic L-morph population (153), 48 of the 50 progeny were of the L-morph, a result consistent with a high degree of intra-morph mating and/or selfing (d = 0.04, CI = -0.15 – 0.23). In populations 68 and 100, both of which lacked the S-morph, I found a single S-morph progeny in one maternal family in each population indicating gene flow from nearby

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 146 populations. Similarly, the occurrence of a single M- and S-morph progeny in population

153 also indicates gene flow from a neighbouring population.

Figure 5.2. Estimated frequency of inter-morph mating and 95% confidence intervals in six populations of Lythrum salicaria of varying floral morph structure. The estimates were calculated through maximum likelihood using progeny ratios predicted after inter-morph and intra-morph mating. In the majority of cases, the estimate of inter-morph mating was not significantly different from 1.0. However, in populations 84, 92, and 135 L-morph plants experienced a low level of selfing and/or intra-morph mating and in population 153 the L-morph reproduced almost exclusively through selfing and/or intra-morph mating.

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 147

Mating and fertility of plants at low density

Progeny tests of open-pollinated families of L- and M-morph plants provided unequivocal evidence of inter-morph mating. I found that 23% and 11% of the total progeny produced by L- and M-morph plants (N= 200 and 127, respectively; Appendix 5 Table 4) were of the S-morph, and 24% of L-morph progeny were of the M-morph. Because genotypes of the L- and M-morph cannot produce S-morph plants, and the L-morph cannot produce the

M-morph, except by inter-morph mating, these offspring must have arisen from pollen dispersal among plants in the fields. I also detected significant heterogeneity in the progeny morph ratios in families from L- and M-morph parents (L-morph: Χ2 = 40, df =

16, P < 1x10-3, M-morph: Χ2 = 30, df = 16, P < 0.05) reflecting variation in the paternity of offspring produced in each plant and possibly some variation in the genotypes of M- morph parents represented in the experiment.

Isolated plants in my field experiment experienced significant pollen limitation for all three measures of fertility. The index of pollen limitation for fruit set and seed set per fruit of open-pollinated flowers was 0.25 and 0.56, respectively, with little variation among morphs (Fig. 5.3A and B). The pollen limitation index based on the multiplicative measure of seed set was 0.59 and there were no significant differences among the floral morphs in the intensity of pollen limitation (Fig. 5.3C). There was also no evidence of differences among the floral morphs in the percent of open-pollinated flowers setting fruit,

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 148 the mean seed set per fruit, or the per plant measure of total seed set (Fig. 5.4).

Figure 5.3. Index of pollen limitation (PL) based on open- and hand-pollinated fruit and seed set (see Methods) for isolated plants of Lythrum salicaria in a field experiment. A)

Percent fruit set; B) mean seed per fruit; C) percent fruit set multiplied by mean seed per fruit set. The floral morphs did not differ in percent fruit set, mean seeds per fruit, or total seeds per plant. However, all values were significantly greater than zero indicating significant pollen limitation for all three indices of pollen limitation.

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 149

Figure 5.4. Reproductive success of open-pollinated flowers of Lythrum salicaria in a field experiment as measured by: A) percent fruit set, B) mean seed per fruit C) percent fruit set multiplied by seed set. The values include individuals that did or did not receive supplemental hand-pollination. There were no significant differences in measures of female fertility among the floral morphs.

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 150

Discussion

The primary goal of this investigation was to elucidate the influence of floral morph ratios and low plant density on mating and fertility in invasive L. salicaria. Two features of this species motivated my study. First, because L. salicaria has prolific colonizing ability, populations display a wide range of both population size and floral morph composition

(Eckert & Barrett 1992; Ågren & Erickson 1996; Eckert et al. 1996b). Among my study populations, population sizes ranged from 8-187 individuals with a wide range of morph frequencies (Table 5.1). Second, controlled pollination studies have consistently revealed that L. salicaria exhibits partial trimorphic incompatibility, with plants of the M-morph more often self-compatible and those of the S-morph most strongly self-incompatible

(Darwin 1865; Stout 1923; O’Neil 1994; Mal 1999; Colautti et al. 2010a; Chapter 4). This variation in demography and incompatibility expression raised the question of whether these features might influence patterns of mating and fertility.

Using progeny tests of maternal families and measure of fruit and seed set in two demographic contexts – natural populations and a field experiment – I evaluated the sensitivity of several reproductive parameters to morph ratio bias and spatial isolation. I found that inter-morph mating predominated in all five polymorphic (trimorphic, dimorphic) populations, regardless of their floral morph ratios. Also, in my field experiment on plant isolation inter-morph mating was frequent because of extensive bee- mediated pollen dispersal among plants. Although individuals in the experiment suffered moderate pollen limitation (mean PL= 0.59 for the multiplicative measure of seed set), there was no evidence that its intensity varied among the floral morphs owing to morph- specific partial self-incompatibility. Earlier studies on pollen limitation in L. salicaria

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 151 provided mixed results concerning whether morph-specific effects were evident and which morphs were susceptible (O’Neil 1992; Ågren & Erickson 1996; Waites & Ågren 2004).

In the ecological contexts that I investigated, I found no strong evidence that partial self- incompatibility in the M-morph of L. salicaria plays a significant role in affecting mating patterns, or in alleviating pollen limitation at low density. I now consider the implications of my results for the reproductive ecology of invasive populations of L. salicaria and in explaining the patterns of floral morph ratios reported in native and introduced populations.

Mating in natural populations of varying morph composition

Using the progeny test method, I found that the vast majority of mating in trimorphic and dimorphic populations of L. salicaria involved inter-morph cross-fertilization and, concomitantly, that selfing and intra-morph mating occurred at very low levels (Fig. 5.2).

This indicates that trimorphic incompatibility functions effectively to promote disassortative mating, regardless of the pollen loads delivered to stigmas by insect pollinators, which may contain a significant amount of intra-morph pollen (Mulcahy &

Caporello 1970; Waites & Ågren 2004; Costa et al. 2017). The resilience of inter-morph mating, despite deviations from isoplethy (three of five polymorphic populations had anisoplethic morph ratios – populations 84, 92, and 135, see Table 5.1), should promote negative frequency-dependent selection and over time populations are predicted to proceed from biased morph ratios to a phenotypic equilibrium involving isoplethy (Fisher

1941a; Heuch & Lie 1985). However, both theoretical and empirical studies of the dynamics of morph ratios in populations of tristylous species (e.g. Morgan & Barrett 1988;

Eckert & Barrett 1992, 1995) indicate that the time that this takes can be quite protracted,

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 152 despite increases in population size and persistent frequency-dependent mating. These studies indicate that the composition of founding genotypes, population structure, features of life history and patterns of sexual recruitment can each slow progress to the isoplethic equilibrium.

Whereas progeny tests of the M- and S-morph in polymorphic populations generally revealed high rates of inter-morph mating (Fig. 5.2), the small excess of L- morph progeny from L-morph parents in populations 84, 92, and 135 indicated a low rate of intra-morph mating and/or selfing, which cannot be distinguished using the progeny test method used here. This result may help to explain the empirical observation that the L- morph is the most common morph in tristylous populations, based on extensive surveys of natural populations of L. salicaria (Overall average frequencies from surveys: L-morph =

0.379, M-morph = 0.296, S-morph = 0.338; population count N = 558, plant count n =

83885; Heuch 1979a; Anderson & Ascher 1995; Ågren & Ericson 1996; Eckert et al.

1996a; Balogh & Barrett 2016; Costa et al. 2016; Chapter 2). Intra-morph and/or self- mating in the L-morph produces only L-morph progeny because this morph is homozygous recessive at the S- and M-loci controlling tristyly. In contrast, all other tristylous genotypes segregate two or three floral morphs from intra-morph mating. In a deterministic model of morph-frequency dynamics, Heuch (1979b) calculated that with rates of intra-morph mating of 8.4–11% the average observed frequency of the L-morph in

European populations could be obtained. In extreme cases, such as tristylous Narcissus triandrus in which intra-morph mating is permitted, it has been shown that high rates of assortative mating in the L-morph cause this morph to predominate in populations throughout the geographical range of the species (Barrett et al. 2004; Hodgins & Barrett,

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 153

2008). Thus, low levels of selfing or assortative mating as observed in my study could explain the generally higher frequency of the L-morph in tristylous populations of L. salicaria.

Progeny tests of a small number of L-morph families in the single monomorphic population in my study revealed that all progeny except for two were of the L-morph. This result indicates that the observed seed produced in this population arises almost exclusively from selfing and/or intra-morph mating and suggests that partial self- incompatibility in the L-morph can play a role in maintaining populations in which cross- compatible floral morphs are absent. Although monomorphic populations of L. salicaria are infrequent, a survey of 114 invasive populations in Ontario revealed the occurrence of eight populations that were composed exclusively of the L-morph and one population that contained only the M-morph (Balogh & Barrett 2016). Partial self-incompatibility may function to maintain monomorphic L-morph populations whereas monomorphic M-morph populations are likely to segregate L- and M-morph progeny (Fisher & Mather 1943).

However, occasional long-distance pollen flow events have the potential to convert monomorphic to polymorphic populations and the finding of single mid- and short-styled offspring in families from population 153 provides evidence for this process. Depending on the spatial isolation of populations and the degree of long-distance gene flow, monomorphic floral morph structure may be quite transient and this may account for its general rarity in L. salicaria.

Partial self-incompatibility in the M-morph was not a significant factor influencing mating patterns in the present study but may provide reproductive assurance in this morph during colonizing events. Previous studies of morph-frequency variation found that ~25%

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 154 of L. salicaria populations in Ontario were dimorphic containing the L- and M-morphs

(Eckert & Barrett 1992; Balogh & Barrett 2016). Selfing and/or assortative mating of M- morph founders with partial self-incompatibility may be responsible for the origin of some of these populations as the vast majority of genotypes of the M-morph are heterozygous and would segregate L- and M-styled plants (Heuch & Lie 1985). However, there is no evidence that the frequency of partial self-incompatibility in plants of the M-morph of L. salicaria varies across the geographical gradient of invasion in Ontario (Colautti et al.

2010a), as one might predict if self-compatibility provided reproductive assurance at low density in isolated populations at the range front of the biological invasion (Baker 1955;

Barrett 2011; Pannell 2015). Nevertheless, it is too early to discount the role of partial self- incompatibility in enabling the establishment of colonies following dispersal of M-morph colonists and more studies of mating in small dimorphic populations would be valuable to address this question.

My data suggest that following the proposed origin of dimorphic populations, inter-morph mating may increase in frequency leading to isoplethic L- and M-morph ratios. Indeed, this may have occurred in dimorphic populations 68 and 100 in which morph ratios are close to being equal in frequency. Significantly, of the 41 dimorphic populations so far reported from surveys in Ontario, 26 possess isoplethic morph ratios

(Eckert & Barrett 1992; Balogh & Barrett 2016) indicating that tristyly can still function effectively in promoting high rates of disassortative mating despite morph loss. Mating patterns in colonizing populations of L. salicaria may be quite dynamic, changing rapidly from the establishment phase to consistent inter-morph mating if compatible mates arise in progeny or by gene flow. Thus, partial self-incompatibility in the M-morph may provide

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 155 reproductive assurance for only a few generations during the early establishment phase in dimorphic populations. In contrast, in populations comprised of the L-morph partial self- incompatibility may enable sexual reproduction to continue for many generations until migrants of the M and S-morph arrive by seed or pollen flow.

Plant isolation and ‘long-distance’ pollen flow

The patterns of inheritance in tristylous populations provide an opportunity for the detection of long-distance pollen flow using open-pollinated families because maternal seed families of the L- and M-morph cannot produce plants of the S-morph without crossing with an S-morph paternal individual. Similarly, the L-morph cannot produce M- morph progeny after intra-morph mating (Fisher & Mather 1943). Simulation models indicate that rates of gene flow of approximately m = 0.05/generation can maintain trimorphism in over 95% of populations with N > 10 individuals over the course of 100 years (Eckert et al. 1996a). In my study LM-dimorphic populations 68 and 100 experienced a minimum migration rate, based on the frequency of S-morph progeny, of m

= 0.001 and 0.002, respectively, and the L-morph monomorphic population exhibited a minimum m = 0.042 based on the presence M- and S-morph progeny. These estimates of m using the tristyly loci should be considered minimum estimates that likely underestimate the true level of gene flow among individuals and populations. A major future challenge given the autotetraploid status of invasive populations of L. salicaria will be to develop hypervariable molecular markers to estimate mating parameters and gene flow.

Most long-distance pollen flow studies that have screened for genetic markers in progeny have focused on forest trees. However, several relevant studies have been

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 156 conducted in herbaceous plants (e.g. Raphanus sativa, Ellstrand & Marshall 1985;

Solanum tuberosum, Skogsmyr 1994; Primula elatior, Van Rossum et al. 2011) and have provided evidence for mating events at ~151 m (Primula elatior) to over one kilometre

(Raphanus sativa) from paternal parents. My results on inter-morph mating from the plant isolation experiment indicated that moderate rates of pollen dispersal at distances of 50-85 meters were common and largely mediated by Bombus spp., which I frequently observed visiting flowers of isolated plants. In the natural populations of L. salicaria that I investigated I recorded rare instances of gene flow events exceeding 1 km based on the distance to the nearest populations to the focal populations used in progeny testing. Once again I commonly observed Bombus visiting flowers in these populations for nectar and pollen. Long-distance Bombus foraging from 200 m to 1.5 km has been reported in the literature (Osborne et al. 1999; Walther-Hellwig & Frankl 2000; Knight 2005) and given that this taxon is a major visitor to L. salicaria populations my results on gene flow distances are not especially surprising.

Isolated flowering individuals of L. salicaria that I did not observe near my focal populations may have also been the source of pollen responsible for the rare gene flow events I recorded in populations 63, 100, and 153. This possibility points to the need for detailed surveys of isolated individuals in studies of inter-population gene flow in plants, a concern raised by Levin (1995) in emphasizing the role of plant isolates in acting as gene flow bridges between populations. In future, high resolution genetic markers and detailed mapping approaches would be useful for thoroughly investigating the frequency and landscape-level scale of gene flow in invasive populations of L. salicaria, particularly from isolated plant outliers that commonly occur in this species.

CHAPTER 5: POLLEN LIMITATION AND MATING PATTERNS 157

Supplemental material for this chapter is present in appendix 5, pp. 276-275.

CHAPTER 6

AN EXPERIMENTAL STUDY OF INBREEDING DEPRESSION AND THE EFFECTS

OF COMPETITION IN AN INVASIVE PLANT

Abstract

Inbreeding depression and its interaction with environmental conditions, including competition, are likely to play an important role during biological invasion. But few studies have investigated the fitness of selfed and outcrossed offspring in invasive plants or quantified how competition might affect the intensity of inbreeding depression. Here, I address these questions experimentally by comparing selfed and outcrossed progeny of the invasive, tristylous, wetland perennial Lythrum salicaria (Lythraceae) over four growing seasons, including three under field conditions. I compared progeny without and with intraspecific competition from selfed or outcrossed neighbours and examined the influence of competition and breeding treatment on fitness correlates by measuring a range of life- history traits. These included: proportion of seeds germinating, days to germination, survival, proportion of plants flowering, time to flowering, vegetative mass, and inflorescence number and mass. The resulting data were analysed for each trait and using functions from time series estimates of growth and two multiplicative estimates of fitness.

I detected varying intensities of inbreeding depression for a range of traits including inflorescence mass, a measure of reproductive output, in three out of the four years of the experiment. Cumulative inbreeding depression over the four years averaged δ = 0.48 and

0.68, depending on the method of multiplicative fitness used. Unexpectedly, the competition treatments did not significantly affect plant performance and the magnitude of inbreeding depression. The detection of inbreeding depression for several key life-history

158 CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 159 traits in L. salicaria is predicted given the largely outcrossing mating system of populations and suggests that biparental inbreeding in small colonizing populations is likely to have significant effects on demographic parameters such as population growth.

Introduction

The amount of inbreeding is a key determinant of fitness in plant populations and can influence their demography and growth rate. When a species consists of small isolated patches, individuals may often outcross with related individuals (biparental inbreeding), or if self-compatible they may reproduce via self-fertilization. These forms of inbreeding increase homozygosity across the genome leading to the expression of deleterious recessive alleles normally sheltered in the heterozygous state in outcrossing populations.

This process causes inbreeding depression (δ) and a loss of fitness (Lande & Schemske

1985; Charlesworth & Charlesworth 1987; Barrett & Kohn 1991; Charlesworth & Willis

2009). Populations that are small and isolated are a ubiquitous feature of colonizing species and individuals on the edge of an invasion front may undergo serial bottlenecks causing the loss of heterozygosity (Shigesada & Kawasaki 1997; Excoffier et al. 2009).

Thus, the extent to which inbreeding may influence the fitness of populations of colonizing species is a key question in invasion biology and for contemporary evolution.

Inbreeding depression is not a static property of individuals or populations but rather may vary in intensity depending on environmental conditions (Dudash 1990) and the extent of inbreeding in populations (Lande & Schemske 1985). It has often been suggested that adverse environmental conditions should reduce the performance of inbred offspring more strongly than outbred offspring, thus increasing the strength of inbreeding

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 160 depression (Cheptou et al. 2000a, b; Kariyat et al. 2011). However, this hypothesis does not find consistent support. For example, Sandner and Matthies (2006) detected inconsistent changes in inbreeding depression in Silene vulgaris across environments and

Nason and Ellstrand (1995) detected no impact of drought conditions on inbreeding depression in Raphanus sativus. A meta-analysis of numerous plant and animal taxa detected that ‘stress’ conditions increased inbreeding depression by 69%, but that this response was not universal across all studies (Armbruster & Reed 2005). A key issue in evaluating the influence of adverse growing conditions on the intensity of inbreeding depression is what form of stress is being evaluated and whether this involves abiotic or biotic challenges. For example, variation in the importance of density-dependent biotic interactions has been shown to influence the strength of inbreeding depression (Yun &

Agrawal 2014). In plants, such effects are sometimes expressed as dominance and suppression under competitive conditions in which outcrossed progeny pre-empt resources from selfed progeny reducing the growth of selfed progeny and causing elevated inbreeding depression (Schmitt et al. 1987; Schmitt & Ehrhardt 1990). Because of the immobility of plants, the influence of inbreeding on their fitness is likely to be particularly sensitive to the density and breeding history of neighbours.

Biological invasions can act as natural experiments allowing the investigation of evolutionary processes over contemporary time. Invasive species frequently encounter novel environments during range expansion leading to natural selection and local adaptation (Maron et al. 2004; Dlugosch & Parker 2008; Colautti & Barrett 2013).

However, serial bottlenecks and founder events are also an intrinsic feature of the invasion process and are often accompanied by increased inbreeding, a loss of genetic diversity and

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 161 reduced fitness (Barton & Charlesworth 1984; McCommas & Bryant 1990). These stochastic processes may cause the fixation of deleterious alleles on the invasion front

(Peischl et al. 2013), which in turn can alter the demographic performance of populations

(Pujol et al. 2009; Barringer et al. 2012). Despite the diverse processes that can potentially influence the fitness of colonizing populations there have been surprisingly few attempts to quantify inbreeding depression in invasive species (but see Parisod 2005; Facon et al.

2011; Mullarckey et al. 2013; Rosche et al. 2017), and even fewer have investigated how inbreeding depression may vary with environment conditions (Garcia-Serrano et al. 2008), despite the possibility that this might play a role in evolution during biological invasion

(Schrieber & Lachmuth 2017). One of the goals of this study is to investigate the influence of competition in modulating inbreeding depression in a successful plant invader.

Lythrum salicaria (Lythraceae) is an autotetraploid, outcrossing, herbaceous, perennial native to wetlands, ditches and river edges in Eurasia. Recently, the species has become highly invasive, especially in eastern North America where it has spread rapidly over the past 150 years since its introduction to the eastern seaboard of the U.S.A.

(Thompson et al. 1987). Invasive populations of L. salicaria vary considerably in size and density, from small isolated populations to very large monospecific standards comprised of many thousands of plants, often at high density (Eckert & Barrett 1992; Chapter 3).

Plants do not reproduce by clonal growth and all regeneration in populations is by seed

(Yakimowski et al. 2005). The maximum age of plants has not been established by demographic studies but individuals often persist for at least 12 or more years (Thompson et al. 1987; S. C. H. Barrett pers. observ.). Flowers of L. salicaria are predominantly bee- pollinated, particularly by Apis mellifera and Bombus spp., although other pollinators such

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 162 as butterflies, wasps, and occasional hummingbirds visit flowers for pollen/and or nectar

(Thompson et al. 1987; King & Sargent 2012; Chapter 4). Local adaptation in flowering time and plant stature has been demonstrated in both the native and invasive range of L. salicaria based on the length of the growing season (Olsson & Ågren 2002; Bastlová et al.

2004; Colautti et al. 2010b). Because Lythrum salicaria is still undergoing invasion to new areas in North America the species is a desirable study system for addressing the role of inbreeding depression and competition in an outcrossing invasive species.

The contemporary evolution of local adaptation in invasive populations in eastern

North America has undoubtedly been enhanced by considerable amounts of quantitative genetic variation for key life-history traits (Colautti & Barrett 2011). The maintenance of this variation is promoted by the largely outcrossed mating system of populations.

Lythrum salicaria is tristylous and possesses a trimorphic incompatibility system that enforces high rates of disassortative mating (Chapter 5). However, in most populations of

L. salicaria partially self-incompatible individuals have been detected following controlled hand-pollination (Chapter 4). It is unclear to what extent the occurrence of standing genetic variation for partial self-incompatibility in L. salicaria influences mating patterns in large populations; however, in small isolated populations where reproductive assurance may be important, selfing and intramorph mating seem likely (Chapter 5).

Regardless of the function of partial self-incompatibility, the occurrence of some degree of self-compatibility in L. salicaria facilitates the generation of selfed offspring and I exploited this feature of the species to enable my investigation of inbreeding depression.

In this chapter, I investigate the expression of inbreeding depression in L. salicaria by comparing the performance of selfed and outcross families. The only previous study of

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 163 inbreeding depression in L. salicaria only investigated germination and seedling traits and was conducted over a five-week period reporting values of δ ranging from 0.44-0.64

(O’Neil 1994). In contrast, in this study I extend the time period in which inbreeding depression was measured to encompass several growing seasons. My study addressed the following specific questions. 1) How strong is inbreeding depression in L. salicaria and what is the cumulative expression of inbreeding depression over four growing seasons? I investigated this question by measuring a range of vegetative and reproductive traits under both glasshouse and field conditions. I predicted that owing to the outcrossed mating system of L. salicaria, inbreeding depression would be evident in the experiment but two factors may reduce its magnitude in comparison with other outcrossing, perennial species that have been investigated. First, frequent colonizing episodes, bottlenecks and biparental inbreeding in L. salicaria may reduce genetic load and hence the strength of inbreeding depression (Lande & Schemske 1985). Second, theoretical and empirical studies of inbreeding depression in diploid versus autotetraploid populations (Lande & Schemske

1985; Bever & Felber 1992; Husband & Schemske 1997) suggest reduced inbreeding depression owing to the masking of deleterious load. 2) Does the presence of intraspecific selfed and outcrossed competitors influence the magnitude of inbreeding depression in life-history traits? I addressed this question by adding either a selfed or an outcrossed competitor to selected pots in the experiment and determining their influence on a focal selfed or outcrossed plants in the same pot. There were two predictions from the competition treatments: i) outcrossed focal plants should experience a greater reduction in performance when competing against outcrossed than selfed offspring; ii). Selfed focal plants should experience a greater reduction in performance when competing against

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 164 outcrossed than selfed offspring. These predictions would be detected as an interaction between competitive environment and breeding treatment in the subsequent analysis. 3)

When during the season does inbreeding depression most strongly impact the growth rate of individuals? I evaluated this by examining changes in plant height during the growing season (June – September), fitting the data to several nonlinear growth-rate functions each year, and comparing the average growth rates between treatments in each year.

Materials and Methods

Source material

The selfed and outcrossed families of L. salicaria that are the basis of this study were obtained from 29 maternal plants that set more than 10 seeds following self-pollination in the study of partial self-incompatibility detailed in Chapter 4. The plants originated from four populations (HUM, CDV, RRV, and DON) all of which are trimorphic and each contained ~1000 individuals. The populations were all located within the Greater Toronto

Area (see Table 4.1; Chapter 4). The 29 families used in the experiment were obtained from self- and cross-pollinations of 5 L-, 17 M- and 7 S-morph parents.

Comparison of fitness components under glasshouse conditions

In early April 2014, I filled five 200-cell germination trays with Promix-BX potting mix and planted 10 self- and 10 cross-fertilized seeds per family blocked across these trays. I germinated the seeds under vented plastic covers on a bench in a glasshouse at the Earth

Sciences Centre, University of Toronto, where I maintained the temperature between 15-

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 165

25C until germination. I recorded whether each seed germinated, the time of germination, and the survival of each seedling after germination.

On May 8-11, after seedlings had developed true leaves and around the time when

L. salicaria begins to grow in southern Ontario, I assigned three self- and cross-fertilized progeny (hereafter S and X, respectively) from the 29 families which produced at least six offspring per breeding treatment to serve as focal plants in all stages of the inbreeding depression experiment. I assigned each of the selected seedlings to one of three

‘competition environments’ by transplanting individuals: 1) singly in a pot; 2) in a pot with a selfed competitor; 3) in a pot with an outcrossed competitor. Thus, I produced a total of six treatments (S, X, XS, XX, SX, SS) where the first letter represents the breeding treatment of the focal plant and the second letter (or lack thereof) represents the breeding treatment (or absence) of the competitor (Fig. 6.1). The non-focal plant in pots with two plants originated from a different family than the focal plant. I used 12.5 cm (5” standard) pots with Promix-BX media and randomly blocked plants by family into six blocks on two flooded (~2 cm depth) benches in the glasshouse (Fig. 6.2A). I maintained water in the benches at all times and added water-soluble fertilizer (14N-14P-14K) every two weeks following the manufacturer’s instructions. Throughout the summer, I measured plant height from the soil surface approximately every two weeks and recorded the flowering date of each plant as days since transplanting was completed (May 11). When plants ceased flowering in early October, corresponding to the end of the growing season, I recorded each plant’s survival since transplant, flowering, height of each individual measured from the soil surface to the tallest point, length of the longest inflorescence on each individual, the basal stem diameter from two positions at 90-degree angles from each

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 166 other (with the two measures averaged), the number of vegetative stems at the base of the plant, and the number of inflorescences on each plant. I then harvested the above ground tissue from each plant, dried it in a 50C degree drying oven for 30 days, and measured the dry vegetative and total inflorescence mass of each plant.

Figure 6.1. The competition and breeding treatments used in the inbreeding depression experiment. In each pot is a focal individual with and without a S (selfed) or X

(outcrossed) competitor. Focal plants are always on the left in this illustration but were arranged haphazardly in the experiment.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 167

Figure 6.2. Images of the inbreeding depression experiment on Lythrum salicaria. (A)

Earth Sciences glasshouse and (B-D) Koffler Scientific Reserve over four growing seasons

(2014-7). A) July 2014, flowering; B) Early October 2015, post flowering; C) Early

August 2016, flowering; D) Late June 2017, pre-flowering.

Comparison of fitness components under field conditions

In the spring of 2015, I transported all pots in the glasshouse experiment by van to the

Koffler Scientific Reserve (44.02610 N, 79.53549 W) in Newmarket, Ontario, Canada where they were placed into a common garden field experiment for three flowering

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 168 seasons (2015-7; Fig. 6.2B-D). Pots were randomized and sunk into saturated soil in a disturbed freshwater marsh dominated by Typha latifolia and some wild L. salicaria plants. The experimental plot measured 4.5 by 10 m and was cleared of above ground vegetation at planting and at the beginning of each growing season, and two subsequent times in July and September in each growing season, to facilitate the location of treatment plants. The plants were visited twice per week during the growing season and the day of first flowering for each plant was estimated. In 2015 and 2017 the site was flooded owing to natural rainfall and watering was not required; however, in 2016 drought conditions prevailed and I added approximately 30-40 gallons of water to the plants and soil around them on a weekly basis. On four occasions during the growing seasons of 2015 and 2016 and on three occasions in 2017, I measured plant height from soil surface to the tallest point, the number of vegetative stems at the base, and length of the longest inflorescence.

In early October of each year when plants ceased growth, but had not yet dropped their leaves, I harvested the above-ground biomass (inflorescence and vegetative tissue), dried, and weighed these tissues, as described in the preceding section. I also noted whether or not each plant survived or flowered in each of the years. After the completion of the experiment in October 2017, I removed all pots from the field and subsequently destroyed all plants.

Statistical analysis of germination and end-of-season life-history trait data in each year

I used the end of season trait data from the glasshouse and field as approximations of individual fitness components. I then used values of these traits in each year to determine multiplicative fitness correlates in each treatment. I performed all statistical analyses in R version 3.3.2 ‘Sincere Pumpkin Patch’ (R Development Core Team, 2016). I estimated the

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 169 overall correlations between each trait that I measured at the end of each year (2014-17), measured the correlations between traits in each of the six breeding and competitive treatment combinations, and visualized the correlations between these values in the package ‘corrplot’ (Wei & Simko, 2017) (Appendix 6 Fig. 1). Significant correlations occurred between traits overall; however, the correlations differed significantly between the six treatments. As a result, I was not able to select a single trait correlate of plant biomass at the end of each year, nor could I use one easily measured variable as a surrogate for another. Therefore, each year, I used plant survival, whether or not a plant flowered, date of flowering, and total inflorescence mass as the end-of-year fitness correlates. I used plant height as the proxy for plant vigour in the non-linear functions of growth (see below).

I analysed germination data using mixed models with binomial (for germination and survival) or Poisson (for days to germination) residual distributions with breeding treatment (selfed or outcrossed) as a fixed variable and family and germination tray as random variables. I analysed the end-of-year data for each of the four years following mixed-modelling protocols with family as a random variable and three structures for the fixed variables: breeding treatment, competitive environment, or these variables plus their interaction. I performed mixed models for untransformed inflorescence mass in 2014, log- transformed (to meet model assumptions) inflorescence mass in 2015 to 2017, and the log- transformed number of days from the 2014 final transplant date (May 11, see earlier) to flowering date in each year to maintain a standard for potential comparison between years.

I applied a generalized linear mixed model with binomial variables for survival and flowering in each year.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 170

I used the package estimated marginal means (‘emmeans’, Lenth 2018) to obtain the mean and 95% confidence interval estimates of each trait in each set of treatments. I also used the R package ‘emmeans’ to calculate various relative performance (RP) metrics

– I estimated the relative performance between inbred and outcrossed progeny, inbreeding depression (δ), as δ = 1 - ws /wo (ws = selfed progeny performance and wo = outcrossed progeny performance) if wo > ws or δ = ws /wo – 1 if ws > wo (following Ågren & Schemske

1993). I tested reduction in performance caused by competition environment with an analogous set of metrics: 1 - wscomp /wnone if wnone > wscomp, 1 - wocomp /wnone if wnone > wocomp, and 1 - wocomp/wscomp if wscomp> wocomp (where wnone, wscomp, and wocomp represent the performance of all focal plants without competitors, with self-fertilized competitors, and with cross-fertilized competitors, respectively) and negative values if performance is in the opposite direction: wnone /wscomp – 1 if wnone < wscomp, wnone /wocomp – 1 if wnone < wxcomp, and wscomp /wocomp – 1 if wocomp < wscomp. In both of these metrics, non-significant differences in performance are equal to 0 whereas significant differences are different from zero.

Competitive treatments rarely caused a significant change in plant performance, but in the cases where competition had a significant effect I calculated inbreeding depression independently for plants with no competitor, a selfed competitor, and an outcrossed competitor, and statistically compared the difference in these ratios using log-transformed data if the original data was continuous, following Johnston and Schoen (1994).

Analysis of growth via non-linear time series

The time at which inbreeding depression affects growth rate can alter competitive performance between selfed and outcrossed progeny in a multiplicative fashion. In particular, the expression of inbreeding depression in early-life vigour may exacerbate

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 171 later differences in plant fitness correlates via dominance and suppression, which occurs when outcrossed progeny pre-empt resources and increase the expression of inbreeding depression in selfed plants (Schmitt et al. 1987; Schmitt & Ehrhardt 1990). I fitted all nonlinear growth curves using the R package ‘nlme’ (Pinheiro et al. 2018) and controlled for multiple measures on individuals by defining each plant as a ‘group’ random factor term, which directed the model to estimate fixed model terms based upon each individual’s growth curve. I standardized measurement dates as days from May 11 in each year, as discussed earlier. These controls allowed comparisons within years between treatments and between years for model behaviour.

I fitted the mean height of all plants (y) at each measurement time (t) in 2014-2016 to each of four asymptotic non-linear growth models presented by Paine et al. 2012. I did not collect data at enough individual time points during 2017 to produce a nonlinear model. These four models consist of different parameters; the ‘monomolecular model’

(Richards 1959):

−푒푥푝푙푟푐∗푡푖푚푒 5.1 푦(푡) = 퐴푠푦푚푝 + (푅0 − 퐴푠푦푚푝) ∗ 푒푥푝 ;

In which Asymp represents y at high values of t, R0 determines y when t is 0, and lrc determines the growth rate in the model; the ‘logistic model’ (Hunt 1981, Zeide 1993):

퐴푠푦푚푝 5.2 푦(푡) = 푥푚𝑖푑−푡𝑖푚푒; 1+푒푥푝 푠푐푎푙

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 172

In which Asymp defines y at high values of t, xmid is the time t at which y value is ½

Asymp, and scal is a parameter relating the model y values to t; the ‘Gompertz model’

(Gompertz 1825; Winsor 1932):

−푏2∗푏3 푡𝑖푚푒 5.3 푦(푡) = 퐴푠푦푚푝 ∗ 푒푥푝 ;

In which Asymp is equal to y when t is large, b2 is a parameter defining y when t = 0, and b3 scales the model relative to t and the ‘four-part-logistic model’ (Hunt 1981; Zeide

1993):

(퐵−퐴) 5.4 푦(푡) = 퐴 + 푥푚𝑖푑−푡𝑖푚푒; 1+푒푥푝 푠푐푎푙푒

In which A determines y when t is small, B determines the maximum value of y when t is large, xmid is the value of t at which y is exactly halfway between A and B, and scal adjusts positioning of the model relative to t. I fit the periodic height measurements from each year to each of these four models and measured the AIC of each fit. I analysed each year using the growth curve that possessed the lowest AIC.

I produced mixed models fitting the growth curve with the lowest AIC for each year following three formats of fixed variables: breeding treatment by growth curve parameters, competitive environment by growth curve parameters, and these factors plus their interaction by parameters. If the models did not converge with breeding and competitive environment treatment attached to each model parameter, I removed the breeding and competitive environment covariate from those parameters. I measured the significance of the breeding and competition terms on model parameters using the ‘nlme’ package’s marginal values ANOVA. I also used the R package ‘emmeans’ to calculate the

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 173 relative performance of nonlinear model parameters between breeding and competitive treatments following the same protocols as I did for the harvest traits.

I analysed the intensity and timing of growth rate differences between models using the average growth rate (AGR) from model curves calculated independently in each of the breeding and competitive environmental treatments. I independently modelled the growth curves from each breeding treatment, competitive environment, and the six combinations of these variables in each of the years studied. I used the distribution of means, variances, and co-variances in parameters from each subset of the model to produce 1000 simulated growth curves from each distribution of parameter sets and then extracted the 95% confidence intervals of model terms. Using the first derivative of the growth curves and the 95% CI of model terms, I produced estimates of AGR across model time and its 95% confidence interval (centimetres of growth per day) from each breeding treatment, competitive environment, and the interaction of these terms continuously from t

= 0 to t = 150 days and estimated the inbreeding depression in average growth rate across these time periods.

Comparisons of multiplicative fitness function

Inbreeding depression depends on multiplicative interactions between life-history traits across the lifetime of an organism. In perennial organisms, year-to-year survival and reproductive output determine an organism’s lifetime reproductive fitness and may vary from year-to-year depending upon changes in environmental conditions (Johnston &

Schoen 1994; Husband & Schemske 1996). I calculated two multiplicative fitness functions for selected trait data – one was based on family-level fitness components

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 174 whereas for the other I simulated organisms and assigned each with fitness correlates by resampling from the observed data. I used two models because small family size prevented me from investigating the interactions between breeding treatment and competitive environment in the family-based measure, whereas a simulated distribution enabled me to investigate breeding and competitive effects, and their interaction. In the family-based model, I multiplied the mean germination percent expressed by each family in 2014 by the mean proportion of plants surviving, proportion of plants flowering, and the mean inflorescence mass for plants producing inflorescences from each of the years. I added these terms for each treatment in each family to produce a ‘cumulative reproductive success’ value. I then calculated an index of relative performance between selfed and outcrossed multiplicative fitness correlates within each family - if wo > ws, I calculated the inbreeding depression for those plants as 1 – ws/wo; however, if ws > wo, I calculated the relative performance index as wo/ws – 1, which generated a negative inbreeding depression value symmetrical to the positive inbreeding depression value. I measured the mean inbreeding depression and standard error using family mean inbreeding depression values.

Calculating the multiplicative fitness function via resampling required additional specialization in model production. To produce a protocol for individuals in the resampling simulation I needed to maintain a similar simulated population size (n) to the size of the observed populations. Therefore, the resampling fitness function in R simulated

246 self- and outcrossed-seeds which, on average, provided the same sample size of adult plants as in the study (n = ~176 in 2014). I started with these simulated seeds and sampled germination success, survival, flowering success, and inflorescence mass for each simulated plant from the distributions in the observed data for each treatment and for each

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 175 year of data. If a simulated plant failed to flower in a year it received NA (not applicable) for inflorescence mass; if an individual failed to survive it neither flowered nor produced stems in subsequent years. I simulated 5000 runs using this protocol. I calculated multiplicative fitness for each of the six treatments in each year as proportion germinated x proportion which survived x proportion flowering x mean inflorescence mass. I additionally produced a cumulative inflorescence mass for each plant over the four simulated years as the ‘cumulative output’ function. I then examined the influence of breeding treatment, competitive treatment, and these values plus their interaction on the expression of relative performance, as described for the end-of-season data.

Results

Comparisons under glasshouse conditions (2014)

Outcrossed seed were 10% more likely to germinate than self-fertilized seeds (Χ2 = 11, df

= 1, P < 0.01; Fig. 6.3) but there were no significant effects of breeding treatment on days to germination or on the survival of seedlings that germinated (days to germination: Χ2 =

0.09, df = 1, P > 0.75; survival of seedlings: Χ2 = 1.71, df = 1, P > 0.15). Thus, germination frequency experienced significant, but relatively weak, inbreeding depression

(δ = 0.12, 95% CI = 0.05-0.20, Fig. 6.4) whereas days to germination and survival after germination were not influenced by breeding treatment.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 176

Figure 6.3. Mean and 95% confidence intervals (bars) of trait values in early life in selfed- and outcrssed-individuals of Lythrum salicaria. Left: binomial traits (germination frequency and survival after germination); only germination frequency was significantly different between breeding treatments. Right: days to germination (‘Day’; notice the difference in the y-axis scale relative to the left panel) was not significantly different between the treatments.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 177

Figure 6.4. The inbreeding depression (δ) mean and 95% confidence intervals (bars) for early-life traits of Lythrum salicaria. Traits are germination frequency, survival after germination and days until germination, ‘Day’. There was weak but significant inbreeding depression for seed germination but not for survival after germination or days to germination.

There was no significant inbreeding depression in first year survival, or in whether plants flowered (survival: Χ2= 0.01, df = 1, P > 90; proportion flowering: Χ2 = 2.05, df = 1,

P > 0.10; Appendix 6 Table 1, Figs. 6.5, 6.6). However, log-transformed time to flowering was 10% shorter in outcrossed than selfed families and inflorescence mass was 72% greater in outcrossed than selfed plants (flowering time; Χ2 = 11.01, df = 1, P < 0.001; inflorescence mass; Χ2 = 19.21, df = 1, P < 0.0001). I also detected a significant effect of competition environment on inflorescence mass in 2014 between plants with selfed and outcrossed competitors compared to plants with no competitor. Inflorescence mass in

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 178 plants of both breeding treatments without a competitor was 40% greater than inflorescence mass in plants with a selfed competitor and 70% greater than the inflorescence mass of plans with an outcrossed competitor, respectively (Χ2 = 10.86, df =

2, P < 0.01). No other response variables were affected by competition and there was no significant interaction effect of breeding by competition for inflorescence mass (results shown in Appendix 6 Table 1).

Comparison of traits under field conditions (2015-2017)

From 2015 through 2017, I periodically discarded focal plants that possessed selfed or outcrossed competitors if they became indistinguishable, owing to the production of multiple stems and intertwining between the rhizomes of competitors in pots, and thus sample size decreased between years (n = 176, 174, 121 in 2015, 2016, 2017). Outcrossed plants in 2016 were 37% more likely to flower than selfed plants and outcrossed inflorescences in 2016 and 2017 were 82% and 97% heavier in mass than selfed inflorescences, respectively, regardless of whether they were in competitive treatments or not (proportion of plants flowering in 2016; Χ2 = 8.21, df = 1, P < 0.01; inflorescence mass in 2016; Χ2 = 5.00, df = 1, P < 05, inflorescence mass in 2017: Χ2 = 7.05, df = 1, P < 0.01; table Appendix 6 Table 1). Breeding treatment did not significantly affect other response variables. I detected that plants of either breeding treatment with no competitor were 85% more likely to survive than plants with an outcrossed competitor in 2017 (Χ2 = 8.2, df = 1,

P < 0.05); however, survival of plants with selfed competitors was not significantly different from survival of plants with no competitors, or from plants with outcrossed competitors and there was no significant interaction between breeding and competitive

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 179

treatment for survival (Appendix 6 Table 1). -

Figure 6.5. The mean and 95% confidence intervals (bars) for trait values of selfed and outcrossed progeny in Lythrum salicaria from 2014 to 2017. Values are depicted for survival, proportion of plants flowering (‘flowering’), flowering time, and inflorescence mass (note the differences in y-axis scales for flowering time and inflorescence mass). The largest difference in means was found in inflorescence mass for 2014, 2016, and 2017.

Other mean values are relatively similar to each other and show no evidence of significant inbreeding depression.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 180

Figure 6.6. Inbreeding depression (δ) mean and 95% confidence intervals (bars) of life- history traits in Lythrum salicaria. Traits are survival, proportion flowering [‘flowering’], flowering time, and inflorescence mass in each year of the experiment (2014 – 2017).

There was marginally significant inbreeding depression in survival in 2017 but not in other years. Traits with significant inbreeding depression are indicated with an asterisk (*).

Analysis of growth via non-linear time-series

The growth data for L. salicaria grown under glasshouse conditions in 2014 best fit a

Gompertz curve with an AIC of 10774.12 (Appendix 6 Table 2). The optimal nonlinear model parameters were: Asymp = 121.69 (se = 3.18), b2 = 4.28 (se = 0.07), b3 = 0.97 (se =

0.0009). I was able to successfully fit breeding treatment and competitive treatment to

Asymp and b2 in the 2014 model: the Asymp values in outcrossed plants were 18% greater

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 181 than in selfed plants (F = 24.03, df = 1, P > 1.0x10-5) but b2 did not differ significantly between breeding treatments (F-value = 0.00, df = 1, P > 0.95). The competitive environment did not significantly affect nonlinear growth parameters in 2014 (Asymp: F =

1.18, df = 2, P > 0.30; b2: F = 2.02, df = 2, P > 0.10). Inbreeding depression in Asymp was

δ = 0.17, 95% CI: 0.11, 0.24).

In 2015-2016, plant growth most closely conformed to the Logistic model based on

AIC (Appendix 6 Table 2). The model terms in 2015 were Asymp = 36.95 (se = 1.18), xmid = 59.79 (se = 0.61), scal = 12.57 (se = 0.88) and the 2016 terms were Asymp = 91.79

(se = 0.252), xmid =39.96 (se = 0.74), scal = 29.13 (se = 1.12). I was able to successfully test the effects of breeding treatment on the Asymp parameter of each model, but the models did not converge with treatments as covariates to other model terms. I found a significant effect of breeding treatment on the asymptote for 2016 with outcrossed plants having an Asymp 29% greater than selfed plants (F = 23.16, df = 1, P < 1.0x10-5); however, there was no difference in asymptotes between breeding treatments in 2015 (F =

3.36, df = 1, P > 0.05). There was a 23% higher Asymp term for plants with no competitor relative to those with an outcrossed competitor in 2016 (F = 4.9, df = 2, P <

0.01) but not in other competitive environments or in 2015 (Asymp in 2015: F = 1.46, df =

2, P > 0.20). I did not detect a significant effect of the interaction between breeding treatment and competitive environment on the Asymp value in 2016 (Asymp from the interaction of breeding treatment and competitive environment: F = 0.38, df = 2, P > 0.65).

Inbreeding depression in Asymp from 2015 and 2016 was δ = 0.10, 95% CI: 0.0, 0.21, and

δ = 0.23, 95% CI: 0.14, 0.31, respectively. These patterns indicate that growth curves differ in the field and glasshouse and possess different asymptotes.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 182

Inbreeding depression in AGR during 2014 and 2016 was significantly above zero from day 0.85 to 76.75 and 0 to day 85.7, respectively, whereas in 2015 this ratio was only significantly different between day 61.6 and 76.15 (Fig. 6.5). The ratios of competitive values differed between the plants with an outcrossed competitor and plants with no competitor for only short periods of time in each year (between days 52.40-53.05 and

54.35-55.90 in 2014, 39.10-47.65 in 2015, and 52.90-99.20 in 2016), and only between the selfed competitor and no competitor treatments from days 57.90 – 75.50 in 2016 (Fig.

6.6). There was no interaction between the breeding and competition treatments for inbreeding depression in AGR.

Figure 6.7. Inbreeding depression (δ) in average growth rate (AGR) experienced by

Lythrum salicaria plants in the non-linear growth models. The grey shading represents the

95% confidence interval of estimates of δ. In 2014 and 2016 the outcrossed plants exhibited a significantly higher AGR relative to selfed plants early to mid-season (0.85 to

76.75 and 0 to 85.7 days, respectively), whereas in 2015 average growth rate was only significantly higher for a short time window mid-season (61.6 to 76.15 days). The

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 183 difference in the first two years caused a difference in asymptote (Asymp) whereas the difference in 2015 was not significant.

Figure 6.8. The relative performance (RP) of Lythrum salicaria plants represented as the average growth rate (AGR) between competitive environments (‘none’, selfed, outcrossed). I calculated the relative performance (where AGR1 and AGR2 represent the first and second competitive environment labelled to the right of each row in the plot) as

RP = 1 - AGR1/AGR2 when AGR2 > AGR1, or AGR2/AGR1 – 1 if AGR2 < AGR1. There were some differences in average growth rate between outcrossed and no competitor for short times in all years, and one short period in 2016 where the AGR of selfed over ‘none’ was significantly different from zero.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 184

Multiplicative function

The two metrics of multiplicative inbreeding depression were generally congruent (Fig.

6.7). The family-based metric of mean inbreeding depression in 2014 when plants were grown under glasshouse conditions was δ = 0.50, 95% CI: 0.35, 0.66, which contrasted with the first year under field conditions (2015) when inbreeding depression did not differ significantly from zero (δ = -0.03, 95% CI: -0.31, 0.26). However, inbreeding depression became significant again in 2016 and 2017 (δ =0.40, 95% CI: 0.15, 0.67; δ = 0.44, 95%

CI: 0.20, 0.69, respectively, Fig. 6.7). Overall cumulative inbreeding depression from family-based estimates over all four years of the experiment was δ = 0.48, 95% CI: 0.28,

0.69.

Inbreeding depression via the resampling method gave similar results and additionally confirmed that the competition treatment did not strongly influence plant performance. The sampling based multiplicative inbreeding depression values for each year’s output matched generally, with the exception that the 2015 data had a much greater variance (δ = 0.54, 95% CI: 0.23, 0.91). Cumulative inbreeding depression via sampling at the end of the experiment was slightly higher than the family-based values, though not significantly so (δ = 0.68, 95% CI: 0.51, 0.74). The values for multiplicative per-year and cumulative relative performance in all environments overlapped in all cases except for the cumulative value of plants with a selfed competitor versus those with no competitor (RP =

0.39, 95% CI: 0.03, 0.65, Appendix 6 Fig. 2). These values indicate that there was no consistent overall effect of competitive treatment on inbreeding depression in the experiment.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 185

Figure 6.9. Multiplicative inbreeding depression (δ) in each year and cumulatively at the end of the four-year experiment on Lythrum salicaria with 95% confidence intervals

(bars). The two values plotted were calculated based on family mean inbreeding depression and a resampling method from the observed distributions of data (see

Methods). Overall, the two approaches gave similar values in each year and, with the exception of 2015, inbreeding depression was of similar magnitude among years.

Discussion

In this chapter I report on a study of the performance of self- and cross-fertilized progeny of the invasive outcrossing plant L. salicaria. I detected a small amount of inbreeding depression in seed germination (δ = 0.12), and inconsistent inbreeding depression in survival and proportion of plants flowering. In contrast, at the end of three out of the four growing seasons in the experiment reproductive output, approximated as inflorescence tissue mass, exhibited inbreeding depression values of δ = 0.44, 0.45, and 0.49 in 2014,

2016 and 2017, respectively. Overall, cumulative inbreeding depression based on several

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 186 key life-history traits was δ = 0.48 or 0.68, depending on whether I applied a family-based or resampling-based multiplicative inbreeding depression function. My analysis of relative growth rates of selfed and outcrossed families indicated significant inbreeding depression in all three years in which plant size measurements were made, although the time in the growing season when this was manifested differed in one of the three years. Finally, I found no consistent effects of the competitive environment (selfed or outcrossed competitor) on the magnitude of inbreeding depression. Below, I discuss the implications of my findings for the invasion biology of L. salicaria and whether inbreeding depression may affect the mating system of invading populations. I also discuss some of the limitations of my experiment and how future work on inbreeding depression in L. salicaria might be improved over the present study.

Inbreeding depression during biological invasion

Few studies have investigated inbreeding depression in invasive species despite the key role that this important population-genetic parameter likely plays in the growth and spread of invaders. Biological invasions are punctuated by frequent founder events and population bottlenecks (Barrett & Husband 1990; Novak & Mack 2005; Barrett et al.

2017), which can expose deleterious recessive alleles to selection and purging (Lande &

Schemske 1985; Kirkpatrick & Jarne, 2000) and/or result in their fixation resulting in reduced fitness (Keller & Waller 2002). Despite the potential of these processes to induce a lag phase for an invasion or enable heterosis after secondary contact between invasive populations (Sakai et al. 2001, Roman & Darling 2007; Keller et al. 2014), very few studies have used experimental approaches to measure inbreeding depression or genetic load in invasive populations.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 187

Lythrum salicaria is tristylous and largely outcrossing owing to the successful functioning of trimorphic incompatibility in most invasive populations (Chapter 5).

Outcrossing populations are predicted to maintain a significant genetic load of deleterious recessive alleles, which when exposed through selfing should result in significant inbreeding depression (Lande & Schemske 1985; Charlesworth & Willis 2009). My experiments confirmed this prediction as inbreeding depression was evident across a range of life-history traits and among the four years of the experiment. Based on the cumulative measures of inbreeding depression calculated over the entire experiment (Fig. 6.7), selfed offspring of L. salicaria were on average roughly half as fit as outcrossed offspring, a pattern generally consistent with an earlier study on seedling growth traits (O’Neil 1994), and consistent with the strength of selection predicted to oppose the evolution of selfing in the absence of selection for reproductive assurance (Fisher 1941b; Lloyd 1979). This level of inbreeding depression is significantly lower than has been reported in other outcrossing perennial angiosperms (e.g. Kohn 1988; Sakai et al. 1989; Eckert & Barrett 1994; Lande et al. 1994; Vogler et al. 1999; Dorken et al. 2002) raising the question of why this may occur.

One obvious answer to the cause of the relatively moderate inbreeding depression in L. salicaria is that my experiment only lasted for four growing seasons and plants of the species may live for much longer. It is quite possible that studies of cumulative inbreeding depression over the entire life span of L. salicaria would reveal a higher intensity of inbreeding depression, as there is good evidence that inbreeding depression intensifies through the life history and therefore may be more strongly expressed at later life stages

(Husband & Schemske 1996). However, it is worth pointing out that there is no evidence

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 188 that year-to-year measures of inbreeding depression intensified over the course of the experiment. As yet, no studies to my knowledge have attempted to investigate cumulative inbreeding depression over the entire lifetime of long-lived perennials using experimental approaches. This is presumably because of the logistical and resource demands that such a long-term experiment would entail. However, genetic marker-based estimates of inbreeding depression (see Ritland 1990; Koelling et al. 2012) can be used in such a situation and these have been usefully applied to long-lived plant species (see for example

Eckert & Barrett 1994b; Dorken et al. 2002). Studies such as these provide a powerful way to examine inbreeding depression in the particular environment in which natural populations occur and avoid the problem of context dependency, a weakness of Darwin’s

(1876) experimental approach which has been a shortcoming in many comparisons of fitness in selfed and outcrossed progeny.

A second possible reason that might explain why inbreeding depression was not especially severe in my experiment is because L. salicaria is a highly successful colonizing species and bottlenecks and periods of small population size are a common feature of its population biology. These demographic events, as well as the possibility of bouts of biparental inbreeding in small populations, could serve to reduce genetic load and hence the intensity of inbreeding depression (Lande & Schemske 1985; Kirkpatrick &

Jarne, 2000; Pujol et al. 2009; Verhoeven et al. 2011; Peischl et al. 2013; Marchini et al.

2016). Comparisons of inbreeding depression between the native and invasive range of L. salicaria, between large and stable versus small and transient populations, or between populations with different invasion histories in the introduced range, could provide insight

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 189 on the role of demographic factors in shaping the intensity of inbreeding depression in the species.

The third potential factor that might be influence the intensity of inbreeding depression in L. salicaria is the autotetraploid nature of invasive populations in North

America (Kubátová 2008). There are theoretical arguments (Lande & Schemske 1985,

Bever & Felber 1992) and empirical evidence (Husband & Schemske 1997) that the strength of inbreeding depression in autotetraploids is likely to be less than in diploids because autotetraploid populations should experience slower progress to homozygosity than diploids per generation and therefore inbreeding will expose fewer recessive deleterious alleles (Lande & Schemske 1985; Bever & Felber 1992). Because both diploid and autotetraploid populations of L. salicaria occur in the native European range

(Kubátová 2008) it would be possible to evaluate the extent to which autopolyploidy might reduce the strength of inbreeding depression in L. salicaria.

Timing and cumulative effects of inbreeding depression

The timing of inbreeding depression is a key factor for understanding its overall expression and I addressed this topic with nonlinear growth curves. Nonlinear growth curves and estimates of relative growth rate have been used in models of plant performance, particularly involving the physiological responses of agricultural crops to genetic or environmental adversity (Chen et al. 2014; Campbell et al. 2015), examination of the relations between metabolism versus size (Muller-Laundau et al. 2006), the presence of trade-offs between growth rate and life history traits, including survival and reproduction, after disturbance (Rose et al. 2009) or growth rate versus herbivore defences

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 190

(Paul-Victor et al. 2010; Züst et al. 2011). However, to my knowledge this approach has not been used to study differences between selfed and outcrossed families in plant populations.

My analyses revealed complex patterns that showed variation among years, with two showing consistent curves (2014, 2016) and another (2015) in which the timing and magnitude of inbreeding depression in average growth rate was quite different (Fig. 6.5).

It is not clear what mechanisms were responsible for this variation, especially since years

2015 and 2016 were both under field conditions. However, results in general for 2015, the first year of field conditions, differed from the remaining two years in the field in showing only a relatively short widow of time in which outcrossed progeny outperformed selfed progeny. In 2014 and 2016 inbreeding depression was evident early to mid-season, which may be an indication that earlier-acting growth of outcrossed plants may be magnified and result in the large differences evident in end-of-season inflorescence mass. Future studies of inbreeding depression could usefully apply these methods to understand the timing of the effects of dominance and suppression between competing selfed and outcrossed progeny (Schmitt et al. 1987; Schmitt & Ehrhardt 1990). Also, these approaches in conjunction with genetic mapping studies (see Charlesworth & Willis 2009) could be used to better quantify the time at which recessive deleterious alleles are expressed during plant growth potentially enabling the identification of loci governing inbreeding depression.

Environment by inbreeding depression interactions

Unexpectedly, my experiment revealed no consistent effects of competition on the magnitude of inbreeding depression in L. salicaria. In contrast, earlier plant studies have

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 191 reported significant competitive effects on inbreeding depression (e.g. Schmitt et al. 1987;

Cheptou et al. 2000 a, b). Although there were a few sporadic significant differences in trait values between competitive environments (i.e. inflorescence mass in 2014, survival in

2017, cumulative multiplicative fitness; Appendix 6 Table 1, Fig. 2), these differences usually only occurred between a subset of the competitive treatments and also lacked an interaction term between competition environment and breeding treatment. Other work of this type has exposed focal plants to a set of competitors (e.g. Schmitt et al. 1987; Schmitt

& Earnhardt 1990; Wolfe 1993; Cheptou & Schoen 2003) whereas in my study I only used a single competitor against the focal plant within a pot (Fig. 6.1). Under field conditions this design may not have provided sufficient power for the detection of competitive differences between selfed and outcrossed plants. One feature of the design may have been especially important in this respect. For practical considerations, I left all pots sunk into the ground for the entire duration of the field experiment (3 years). Therefore, plants in all pots were able to grow roots through the bottom of the pots into the surrounding soil and exploit available belowground resources and it seems likely that plants in the competitive treatments were not exposed to the full severity of competitive conditions that they would have faced in the more constrained below-ground environment of a single pot. Also,

Lythrum salicaria in the invasive range frequently forms dense monospecific stands

(Thompson et al. 1987; Eckert & Barrett 1992). Under these circumstances plant competition will come from numerous neighbours occurring at a higher density than in the present study. Future research on the effects of competition on inbreeding depression in L. salicaria could usefully implement experimental designs in which a much higher density of competing plants is used than in the current study.

CHAPTER 6: INBREEDING DEPRESSION AND COMPETITION 192

Inbreeding depression and the maintenance of tristyly

Genes promoting self-fertilization can spread in outcrossing populations as a result of

‘automatic selection’ unless inbreeding depression is severe (Fisher 1941b; Lloyd 1979;

Lande & Schemske 1985). The threshold value of inbreeding depression preventing the spread of selfing varies depending on a variety of genetic, demographic and reproductive factors (Lloyd 1980; 1992; Uyenoyama et al. 1993; Goodwillie et al. 2005). But generally, if outcrossed offspring are more than twice as fit as selfed offspring, outcrossing in animal-pollinated species should be maintained as long as pollinator service is reliable.

Investigations reported in Chapter 5 indicate that standing genetic variation in partial self- incompatibility appears to be a general feature of most L. salicaria populations. However, there is no evidence from the literature that any population of L. salicaria is fully self- compatible or has transitioned to high selfing rates as a result of the breakdown of tristyly, despite this transition having occurred elsewhere in the genus Lythrum (see Weller 1992).

Therefore, although demographic conditions associated with biological invasion might favour the evolution of selfing from outcrossing, this transition has apparently not occurred. This finding strongly suggests that the maintenance of tristyly in L. salicaria populations occurs because any inbred offspring that do result from selfing are strongly selected against owing to the general superiority of outcrossed offspring. Cumulative values of inbreeding depression over the four years of this study are generally consistent with this hypothesis.

Supplemental material for this chapter is present in Appendix 298-306

CHAPTER 7

GENERAL CONCLUSIONS AND FUTURE DIRECTIONS

Overview

In my thesis, I investigated a range of problems concerned with the reproductive ecology and genetics of Lythrum salicaria (Lythraceae), a tristylous wetland plant that has become a serious invasive species in eastern North America. I chose this species because it is abundant in Ontario where most of my research was conducted, easily grown under field and glasshouse conditions, and is amenable to a range of experimental manipulations including controlled crosses and the cloning of individual genotypes. My thesis is comprised of five research chapters that were investigated using a range of approaches including: computer simulations, field and glasshouse experiments and extensive surveys of natural populations. In this final chapter of my thesis I briefly summarize in sequence the main findings and conclusions of the five chapters, raise unresolved issues and suggest avenues of further investigation in invasive species.

Modeling the evolutionary dynamics of tristylous populations

A combination of deterministic and stochastic processes may produce skewed morph ratios (anisoplethy) or the loss of floral morphs from tristylous populations. For example, deviations from strict disassortative mating including selfing and assortative mating

(Heuch 1979b; Hodgins & Barrett 2006) or founder events and genetic drift (Heuch 1980;

Eckert & Barrett 1992) can result in anisoplethy or morph loss. In chapter 2 of my thesis, I used a Monte-Carlo computer simulation to determine how morph-specific partial self-

193 GENERAL CONCLUSIONS 194

incompatibility and polyploidy affected the dynamics of allele loss from the loci governing tristyly and hence floral morph loss from populations. The details of these models incorporated features of the reproductive and genetic system of L. salicaria. I discovered that morph-specific partial selfing affected allele loss asymmetrically, which is consistent with predictions based on equilibrium allele frequencies (Fisher 1941a; Heuch

& Lie 1985). I also found that in the presence of morph-specific partial self- incompatibility, autotetraploid populations experienced lower frequencies of recessive allele loss than diploid populations. Selfing in the L-morph resulted in the highest loss of trimorphism, whereas selfing in the S-morph resulted in the lowest rate of loss. Finally, partial selfing by the M- and S-morphs frequently affected the identity of morphs in dimorphic populations whereas selfing in the L-morph did so only rarely. I used these theoretical results to interpret empirical data on morph-frequency variation from natural populations of L. salicaria and in some cases this provided satisfying explanations for the observed patterns

The main limitation of this work is that I did not explicitly include inbreeding depression in my models and as Chapter 6 indicates inbreeding depression is likely to be an important feature of the population biology of L. salicaria. Future stochastic models would be more biologically realistic if they incorporated dynamic inbreeding depression, with the possibility of purging, as well as the heritability of selfing rates. Another potential extension of the models would be to include meta-population structure in stochastic populations (see Eckert et al. 1996a). For simplicity, in the present study all simulations were initiated from populations that were at the isoplethic (1:1:1) equilibrium, despite the

GENERAL CONCLUSIONS 195

fact that this is unlikely to be the case in nature owing to founder events and genetic drift

(Chapter 3). Using a metapopulation framework it would be possible to model how non- equilibrium morph frequencies in populations across the landscape would influence the evolution of dimorphic populations from trimorphic populations.

Morph frequencies after 25 years of ongoing invasion

An earlier study of morph-frequency variation in invasive populations of tristylous L. salicaria in Ontario detected that ~23-% of populations were dimorphic and the vast majority of these populations were missing the S-morph (Eckert & Barrett 1992). This finding was consistent with founder events and genetic drift giving rise to the stochastic loss of the S-allele, which was confirmed by computer simulations. Chapter 3 reports a study in which I surveyed 114 populations in the same region ~25 years later to determine whether, over the intervening period, gene flow (Eckert et al. 1996a) and increased population densities across the landscape associated with on-going invasion had reduced the number of dimorphic populations and those lacking the S-morph. Contrary to this expectation, I found that 26% of populations were dimorphic and most lacked the S- morph, a result not significantly different from the 1988/9 survey. This finding indicates that, despite the increasing maturity of the invasion in Ontario, stochastic processes are still a dominant feature of ongoing spread, with the origin of small populations associated with the stochastic loss of alleles from them.

One of the shortcomings of my survey was that I was unable to locate the specific populations that were included in the Eckert and Barret (1992) survey. GPS co-ordinates were not recorded in 1992, and in some cases although the actual location of the

GENERAL CONCLUSIONS 196

population was unambiguous populations were extirpated. As a result, the vast majority of the populations that I surveyed were different from those included in the 1992 survey. In future it would be valuable to follow specific populations over longer time periods to investigate morph-frequency dynamics and assess how demographic stochasticity affects morph frequencies in situ. This has been carried out over a 5-year interval for L. salicaria populations in Ontario to measure the influence of frequency-dependent selection on morph frequencies (Eckert et al. 1996b), but not over longer time periods. Several other future surveys of morph-frequency variation could also be conducted on L. salicaria. For example, with the exception of Ontario there is a paucity of surveys from other parts of the invasive range, especially from western North America where the species has been introduced for a much shorter time. Data on the early stages of invasion would be especially useful to identify the roles of founder events versus genetic drift in affecting morph-frequency variation.

The nature and evolution of partial self-incompatibility

For convenience, angiosperms are often classified as self-incompatible or self-compatible.

However, some individuals of self-incompatible species exhibit partial self-incompatibility

(PSI) because they set small amounts of seed following self-pollination (Levin 1996). In

Chapter 4, I surveyed the extent of variation in PSI in L. salicaria using experimental self- and cross-pollinations of plants grown under glasshouse conditions from four populations.

I quantified the stability in expression of PSI in plants over two consecutive growing seasons and by cloning plants with a range of PSI values and growing them in two contrasting environments quantified variation in reactions norms for PSI. I also compared

GENERAL CONCLUSIONS 197

the compatibility status of L- and M-morph offspring from selfed mid-styled parents heterozygous at the M-locus to determine if PSI expression was morph specific.

Approximately 34% of plants set seed following self-pollination and the mid-styled morph exhibited higher levels of self-compatibility than the remaining morphs. Plants of the M- morph exhibited significant repeatability of PSI across years and within clones, and there was a weak heritable component to PSI in parent-offspring regressions of compatibility values. I also detected evidence for morph-specific difference in the levels of PSI between the L- and M-morph progeny of M-morph parents. My results demonstrate a small genetic component to morph-specific variation in PSI with environmental factors also contributing to variation in this trait.

A primary motivation for investigating PSI in L. salicaria was to determine the amount of variation in populations and whether or not self-compatibility has a heritable basis and could therefore respond to selection. My studies indicate that PSI has a genetic component, although it appears to be relatively small. Future studies on PSI in L. salicaria could address the following two key questions: 1) Is PSI manifested by differences in pollen-tube growth between self- and inter-morph pollinations? If pollen-tube growth after self-pollination is significantly slower than pollen-tube growth after inter-morph pollinations, PSI may have little functional significance for the mating system of trimorphic populations that receive reliable pollinator service and outcrossed pollen.

Partial self-incompatibility may only be ecologically important in low-density situations where selfing may provide reproductive assurance. 2) Is it possible to obtain fully self- compatible plants of L. salicaria through artificial selection? The apparent absence of fully

GENERAL CONCLUSIONS 198

self-compatible populations of L. salicaria may be because the amount of standing genetic variation for PSI within populations is relatively small. Artificial selection among PSI plants would be valuable to determine the magnitude of selection response that can be achieved over several generations. A weak response would provide an explanation as to why there is no current evidence for the breakdown of trimorphic incompatibility in L. salicaria.

Mating and fertility under demographic stress

Sexual reproduction in heterostylous populations may be especially vulnerable to demographic conditions because of the small number of mating types in populations. In

Chapter 5, I investigated the mating patterns and female fertility of L. salicaria under natural and experimental conditions. I grew 147 open-pollinated seed families from six populations with different morph structures and estimated inter-morph mating (d) based on progeny morph ratios. In a field experiment, I isolated 47 individuals of all three morphs, used progeny ratios to estimate d, and measured the intensity of pollen limitation experienced by plants. The M- and S-morph experienced high rates of d, regardless of population size or morph ratio. Despite significant pollen limitation of fruit and seed set in the field experiment, I found no evidence of differences among the morphs. Estimates of d in the L-morph revealed low levels of intra-morph mating in three populations and near complete intra-morph mating in a monomorphic population. The field experiment demonstrated that although plant isolation caused pollen limitation of seed set, ‘long- distance’ bee-mediated pollen flow served to maintain inter-morph mating, which may enable population outliers (sensu Levin 1995) to participate in the evolution of the

GENERAL CONCLUSIONS 199

invasion process. My study demonstrates that tristyly in L. salicaria is remarkably robust to demographic variation and can maintain outcrossing in the face of challenges associated with colonization.

One of the most significant discoveries in this chapter was the finding that a small monomorphic population of L. salicaria comprised of the L-morph set seed through selfing and/or intra-morph mating. Although the fate of this population is at present unknown, it would be of considerable interest to conduct long-term studies on this population, and other monomorphic populations described in Chapter 3. Two potential scenarios can be predicted and would certainly be worth investigating. The first is that inbreeding depression, perhaps combined with demographic stochasticity, would ultimately cause the local extirpation of these monomorphic populations. Under this model such monomorphic populations are largely ephemeral. As revealed in Chapter 6, moderate inbreeding depression occurs in L. salicaria and hence this scenario is certainly plausible and may explain the overall rarity of monomorphic population in the species despite frequent colonization events. The second scenario predicts that inter-population gene flow through seeds and/or pollen could introduce missing morphs to populations. If this occurs, opportunities for disassortative mating and frequency-dependent selection could restore stylar polymorphism and the functioning of tristyly.

Work in this chapter identified bee-mediated ‘long-distance’ gene flow via pollen through progeny tests of open-pollinated families in natural populations and isolated plants in the field experiment. These processes probably occur in invasive populations that are not highly isolated and cause changes to mating patterns over time from selfing at

GENERAL CONCLUSIONS 200

population initiation to increased outcrossing with population growth. This general notion of mating system change was recently predicted by Pannell (2015) for self-compatible colonizing plants but may also apply to those like L. salicaria that are partially self- incompatible. Future studies of the demography and genetics of small invading populations would be valuable for understanding the factors that determine the life and death of colonizing populations.

Inbreeding depression and invasion

Despite the burgeoning literature on inbreeding depression over the past few decades and the rapid growth of invasion biology there have been relatively few experimental studies of inbreeding depression in invasive plants. In Chapter 6, I investigated the effects of self- and cross-fertilization on progeny performance in L. salicaria over four growing seasons, including three under field conditions. Two different multiplicative estimates based on key life-history traits yielded relatively similar inbreeding depression (δ) values of 0.48 and

0.68, and there was considerable variation among specific life-history traits in the intensity of inbreeding depression. The experimental design I used in my study involved contrasting competitive environments, but unexpectedly I found no consistent influence of competition on inbreeding depression. The detection of inbreeding depression for several key life-history traits in L. salicaria is not unexpected given the largely outcrossing mating system of the species. My results suggest that inbreeding is likely to have significant effects on demographic parameters such as population growth in small colonizing populations and, as discussed above, could potentially lead to a decline in mean fitness and the local extirpation of populations.

GENERAL CONCLUSIONS 201

My study of inbreeding depression exploited the occurrence of partial self- incompatibility in L. salicaria to enable the production of selfed progeny. One issue with this approach is that if early-acting inbreeding depression caused the abortion of selfed embryos prior to seed maturation, the resulting measures of δ would not include this source of inbreeding depression. This could have resulted in an underestimate of the true value inbreeding depression. One way to circumvent this problem in such studies is to avoid self-fertilization altogether and to investigate fitness decline using pedigrees based on relatedness obtained from a formal breeding design (see Lynch & Walsh 1998). Future research on inbreeding depression might consider this approach. As discussed in Chapter

6, there are several other improvements that could be implemented in future experimental studies of inbreeding depression in L. salicaria. In particular, the failure to show that competition influenced the fitness of selfed and outcross progeny in any consistent way is perplexing and should motivate future investigations that more realistically simulate the competitive conditions of natural populations, including both intraspecific and interspecific competition.

Final Remarks

Over the past few decades a growing literature has appeared concerned with the ecology and environmental impacts of Lythrum salicaria. This interest has developed because the species has become a rapidly expanding and pernicious invasive species that has colonized a wide range of seasonally flooded habitats and has degraded important wetland ecosystems (Mal et al. 1992; Blossey et al. 2001; Schooler et al. 2005). In common with the field of invasion biology in general, much of the focus on L. salicaria has been on

GENERAL CONCLUSIONS 202

ecological problems with less attention paid to genetic and evolutionary questions.

Building on earlier Ph.D. thesis research at the University of Toronto on L. salicaria by

Christopher Eckert (1993) and Robert Colautti (2010), my thesis research has contributed several novel findings on the reproductive ecology and genetics of L. salicaria. I hope this work will stimulate future studies on this fascinating plant.

LITERATURE CITED

Agrawal AF, Whitlock MC. 2010. Environmental duress and epistasis: How does stress

affect the strength of selection on new mutations? Trends in Ecology & Evolution

25: 450–458.

Ågren J. 1996. Population size, pollinator limitation, and seed set in the self-incompatible

herb Lythrum salicaria. Ecology 77: 1779–1790.

Ågren J, Ericson L. 1996. Population structure and morph-specific fitness differences in

tristylous Lythrum salicaria. Evolution 50: 126–139.

Ågren J, Schemske DW. 1993. Outcrossing rate and inbreeding depression in two annual

monoecious herbs, Begonia hirsuta and B. semiovata. Evolution 47: 125–135.

Allee WC. 1931. Animal aggregations: a study in general sociology. Chicago: University

of Chicago Press.

Allen AM, Hiscock SJ. 2008. Evolution and phylogeny of self-incompatibility systems in

angiosperms In: Franklin-Tong V, ed. Self-incompatibility in flowering plants.

Berlin:, Springer-Verlag: 73–102.

Anderson NO, Ascher PD. 1995. Style morph frequencies in Minnesota populations of

Lythrum (Lythraceae). Sexual Plant Reproduction 8: 105–112.

Andersson S. 1994. Unequal morph frequencies in populations of tristylous Lythrum

salicaria (Lythraceae) from southern Sweden. Heredity 72: 81–85.

203 LITERATURE CITED 204

Armbruster P, Reed DH. 2005. Inbreeding depression in benign and stressful

environments. Heredity 95: 235–242.

Arunkumar R, Wang W, Wright SI, Barrett SCH. 2017. The genetic architecture of tristyly

and its breakdown to self-fertilization. Molecular Ecology 26: 752–765.

Ascher PD. 1984. Self-incompatibility In: Sink KC, ed. Petunia: monographs on

theoretical and applied genetics. Berlin: Springer-Verlag: 92-110.

Ascher PD, Peloquin SJ. 1966. Effect of floral aging on the growth of compatible and

incompatible pollen tubes in Lilium longiflorum. American Journal of Botany 53:

99–102.

Ashley M V. 2010. Plant parentage, pollination, and dispersal: How DNA microsatellites

have altered the landscape. Critical Reviews in Plant Sciences 29: 148–161.

Ashman TL, Knight TM, Steets JA, Amarasekare P, Burd M, Campbell DR, Dudash MR,

Johnston MO, Mazer SJ, Mitchell RJ, Morgan MT, Wilson WG. 2004. Pollen

limitation of plant reproduction: ecological and evolutionary causes and

consequences. Ecology 85: 2408–2421.

Baker HG. 1955. Self-compatibility and establishment after ‘long-distance’ dispersal.

Evolution 9: 347–349.

Baker HG. 1965. Characteristics and modes of origin in weeds In: Baker HG, Stebbins

GL, eds. The genetics of colonizing species. New York: Academic Press, 147–168.

LITERATURE CITED 205

Baker HG. 1967. Support for Baker’s law - as a rule. Evolution 21: 853–856.

Baker HG, Stebbins GL (Eds.). 1965. The genetics of colonizing species. New York:

Academic Press.

Baldwin SJ, Schoen DJ. 2017. Genetic variation for pseudo-self-compatibility in self-

incompatible populations of Leavenworthia alabamica (Brassicaceae). New

Phytologist 213: 430–439.

Balogh CM, Barrett SCH. 2016. Stochastic processes during invasion: the influence of

population size on style-morph frequency variation in Lythrum salicaria (purple

loosestrife). International Journal of Plant Sciences 177: 409–418.

Barlow N. 1923. Inheritance of the three forms in trimorphic species. Journal of Genetics

13: 133–146.

Barrett SCH. 1977. The breeding system of Pontederia rotundifolia L., a tristylous

species. New Phytologist 78: 209–220.

Barrett SCH. 1979. The evolutionary breakdown of tristyly in Eichhornia crassipes

(Mart.) Solms (Water Hyacinth). Evolution 33: 499–510.

Barrett SCH. 1980. Sexual reproduction in Eichhornia crassipes (water hyacinth). II .

Seed production in natural populations. Journal of Applied Ecology 17: 113–124.

Barrett SCH. 1988. The evolution, maintenance, and loss of self-incompatibility systems

In: Doust JL, Doust LL, eds. Plant reproductive ecology: patterns and strategies.

LITERATURE CITED 206

Oxford: Oxford University Press, 98–124.

Barrett SCH (Ed.). 1992. Evolution and function of heterostyly. Monographs on

theoretical and applied genetics. Berlin: Springer-Verlag.

Barrett SCH. 1993. The evolutionary biology of tristyly In: Futuyma D, Antonovics J, eds.

Oxford surveys in evolutionary biology. Volume 9. Oxford: Oxford University

Press, 283–326.

Barrett SCH. 2011. Why reproductive systems matter for the invasion biology of plants In:

Richardson DM, ed. Fifty years of invasion ecology: the legacy of Charles Elton.

Oxford: Oxford University Press, 195–210.

Barrett SCH, Anderson JM. 1985. Variation in expression of trimorphic incompatibility in

Pontederia cordata L. (Pontederiaceae). Theoretical and Applied Genetics 70:

355–362.

Barrett SCH, Arroyo MTK. 2012. Variation in floral morph ratios in tristylous Oxalis

squamata (Oxalidaceae): an Andean alpine endemic. Botany 90: 1180–1185.

Barrett SCH, Brown AHD, Shore JS. 1987. Disassortative mating in tristylous Eichhornia

paniculata (Pontederiaceae). Heredity 58: 49–55.

Barrett SCH, Charlesworth D. 1991. Effects of a change in the level of inbreeding on the

genetic load. Nature 352: 522–524.

Barrett SCH, Colautti RI, Dluglosh KM, Rieseberg LH. 2017 (Eds.). Invasion genetics:

LITERATURE CITED 207

the Baker and Stebbins legacy.Wiley, London.

Barrett SCH, Colautti RI, Eckert CG. 2008. Plant reproductive systems and evolution

during biological invasion. Molecular Ecology 17: 373–83.

Barrett SCH, Cruzan MB. 1994. Incompatibility in heterostylous plants In: Williams EG,

Knox RB, Clarke AE, eds. Genetic control of self-incompatibility and reproductive

development in flowering plants. The Netherlands: Kluwer Academic Publishers,

189–219.

Barrett SCH, Forno IW. 1982. Style morph distribution in new world populations of

Eichhornia crassipes (Mart.) Solms-Laubach (Water Hyacinth). Aquatic Botany

13: 299–306.

Barrett SCH, Harder LD. 2017. The ecology of mating and its evolutionary consequences

in seed plants. Annual Review of Ecology, Evolution, and Systematics 48: 135–

157.

Barrett SCH, Harder LD, Cole WW. 2004. Correlated evolution of floral morphology and

mating-type frequencies in a sexually polymorphic plant. Evolution 58: 964–975.

Barrett SCH, Hodgins KA. 2006. Floral design and the evolution of asymmetrical mating

systems In: Harder LD, Barrett SCH, eds. The ecology and evolution of flowers.

Oxford, UK: Oxford University Press, 239–255.

Barrett SCH, Husband BC. 1990. Variation in outcrossing rates in Eichhornia paniculata:

the role of demographic and reproductive factors. Plant Species Biology 5: 41–55.

LITERATURE CITED 208

Barrett SCH, Kohn JR. 1991. Genetic and evolutionary consequences of small population

size in plants: Implications for conservation In: Falk DA, Holsinger KE, eds.

Genetics and conservation of rare plants. Oxford, New York: Oxford University

Press, 3–30.

Barrett SCH, Morgan MT, Husband BC. 1989. The dissolution of a complex genetic

polymorphism: the evolution of self-fertilization in tristylous Eichhornia

paniculata (Pontederiaceae). Evolution 43: 1398–1416.

Barrett SCH, Price SD, Shore JS. 1983. Male fertility and anisoplethic population

structure in tristylous Pontederia cordata (Pontederiaceae). Evolution 37: 745–

759.

Barrett SCH, Shore JS. 2008. New insights on heterostyly: comparative biology, ecology,

and genetics In: Franklin-Tong VE, ed. Self-incompatibility in flowering plants –

evolution, diversity, and mechanisms. Berlin: Springer-Verlag, 3–32.

Barringer BC, Kulka EA, Galloway LF. 2012. Reduced inbreeding depression in

peripheral relative to central populations of a monocarpic herb. Journal of

Evolutionary Biology 25: 1200–1208.

Barton NH, Charlesworth B. 1984. Genetic revolutions, founder effects, and speciation.

Annual Review of Ecology and Systematics 15: 133–164.

Bastlová D, CCizkova H, Bastl M, Kvet J. 2004. Growth of Lythrum salicaria and

Phragmites australis plants originating from a wide geographical area: response to

LITERATURE CITED 209

nutrient and water supply. Global Ecology and Biogeography 13: 259–271.

Bateman AJ. 1952. Self-incompatibility systems in angiosperms: I. Theory. Heredity 6:

285–309.

Bateman AJ. 1956. Cryptic self-incompatibility in the wallflower: Cheiranthus cheiri L.

Heredity 10: 257–261.

Bawa K. 1974. Breeding systems of tree species of a lowland tropical commumity.

Evolution 28: 85–92.

Bawa KS, Beach JH. 1983. Self-incompatibility systems in the Rubiaceae of a tropical

lowland wet forest. American Journal of Botany 70: 1281–1288.

Bever JD, Felber F. 1992. The theoretical population genetics of autopolyploidy In:

Antonovics J, Futuyma D, eds. Oxford surveys in evolutionary biology. Volume 8.

Oxford: Oxford University Press, 185–217.

Bianchi M, Vesprini J, Barrett SCH. 2000. Trimorphic incompatibility in Eichhornia

azurea (Pontederiaceae). Sexual Plant Reproduction 12: 203–208.

Bierzychudek P. 1981. Pollinator limitation of plant reproductive effort. American

Naturalist 117: 838–840.

Bixby PJ, Levin DA. 1996. Response to selection on autogamy in Phlox. Evolution 50:

892–899.

Blossey B, Skinner LC, Taylor J. 2001. Impact and management of purple loosestrife

LITERATURE CITED 210

(Lythrum salicaria) in North America. Biodiversity and Conservation 10: 1787–

1807.

Brown ADH, Burdon JJ. 1987. Mating systems and colonizing success in plants In: Gray

AJ, Crawley MJ, Edwards PJ, eds. Colonization, succession and stability. Oxford:

Blackwell Scientific Publications, 115–131.

Brown BJ, Mitchell RJ, Graham SA. 2002. Competition for pollination between an

invasive species (purple loosestrife) and a native congener. Ecology 83: 2328–

2336.

Burd M. 1994. Bateman’s principle and plant reproduction: the role of pollen limitation in

fruit and seed set. The Botanical Review 60: 83–139.

Busch JW, Joly S, Schoen DJ. 2011. Demographic signatures accompanying the evolution

of selfing in Leavenworthia alabamica. Molecular Biology and Evolution 28:

1717–1729.

Byers DL, Meagher TR. 1992. Mate availability in small populations of plant species with

homomorphic sporophytic self-incompatibility. Heredity 68: 353–359.

Campbell MT, Knecht AC, Berger B, Brien CJ, Wang D, Walia H. 2015. Integrating

image-based phenomics and association analysis to dissect the genetic architecture

of temporal salinity responses in rice. Plant Physiology 168: 1476–1489.

Charlesworth D. 1979. The evolution and breakdown of tristyly. Evolution 33: 486–498.

LITERATURE CITED 211

Charlesworth B. 2001. The effect of life history and mode of inheritance on neutral

genetic variability. Genet. Res 77: 153–166.

Charlesworth D, Charlesworth B. 1987. Inbreeding depression and its evolutionary

consequences. Annual Review of Ecology and Systematics 18: 237–268.

Charlesworth D, Willis JH. 2009. The genetics of inbreeding depression. Nature Reviews

Genetics 10: 783–796.

Chen D, Neumann K, Friedel S, Kilian B, Chen M, Altmann T, Klukas C. 2014.

Dissecting the phenotypic components of crop plant growth and drought responses

based on high-throughput image analysis. The Plant Cell 26: 4636–4655.

Cheptou P-O. 2012. Clarifying Baker’s law. Annals of Botany 109: 633–641.

Cheptou P-O, Berger A, Blanchard A, Collin C, Escarre J. 2000a. The effect of drought

stress on inbreeding depression in four populations of the mediterranean

outcrossing plant Crepis sancta (Asteraceae). Heredity 85: 294–302.

Cheptou P-O, Imbert E, Lepart J, Escarre J. 2000b. Effects of competition on lifetime

estimates of inbreeding depression in the outcrossing plant Crepis sancta

(Asteraceae). Journal of Evolutionary Biology 13: 522–531.

Cheptou P-O, Massol F. 2009. Pollination fluctuations drive evolutionary syndromes

linking dispersal and mating system. American Naturalist 174: 46–55.

Cheptou P-O, Schoen DJ. 2003. Frequency-dependent inbreeding depression in

LITERATURE CITED 212

Amsinckia. American Naturalist 162: 744–53.

Colautti RI. 2010. Evolution of local adaptation during plant invasion: purple loosestrife

(Lythrum salicaria – Lythraceae) in eastern North America. Ph.D. thesis,

University of Toronto.

Colautti RI, Barrett SCH. 2010. Natural selection and genetic constraints on flowering

phenology in an invasive plant. International Journal of Plant Sciences 171: 960–

971.

Colautti RI, Barrett SCH. 2011. Population divergence along lines of genetic variance and

covariance in the invasive plant Lythrum salicaria in eastern North America.

Evolution 65: 2514–2529.

Colautti RI, Barrett SCH. 2013. Rapid adaptation to climate facilitates range expansion of

an invasive plant. Science 342: 364–366.

Colautti RI, White NA, Barrett SCH. 2010a. Variation of self-incompatibility within

invasive populations of purple loosestrife (Lythrum salicaria L.) from eastern

North America. International Journal of Plant Sciences 171: 158–166.

Colautti RI, Eckert CG, Barrett SCH. 2010b. Evolutionary constraints on adaptive

evolution during range expansion in an invasive plant. Proceedings of the Royal

Society of London Ser. B 277: 1799–1806.

Costa J, Castro S, Loureiro J, Barrett SCH. 2016. Variation in style morph frequencies in

tristylous Lythrum salicaria in the Iberian Peninsula: the role of geographical and

LITERATURE CITED 213

demographic factors. Annals of Botany 117: 331–340.

Costa J, Castro S, Loureiro J, Barrett SCH. 2017. Experimental insights on Darwin’s

cross-promotion hypothesis in tristylous purple loosestrife (Lythrum salicaria).

American Journal of Botany 104: 616–626.

Crnokrak P, Barrett SCH. 2002. Purging the genetic load: a review of the experimental

evidence. Evolution 56: 2347–2358.

Cruzan MB, Barrett SCH. 1993. Contribution of cryptic incompatibility to the mating

system of Eichhornia paniculata (Pontederiaceae). Evolution 47: 925–934.

Cruzan MB, Barrett SCH. 2016. Postpollination discrimination between self and outcross

pollen covaries with the mating system of a self-compatible flowering plant.

American Journal of Botany 103: 568–576.

Dana MN, Ascher PD. 1985. Pseudo-self compatibility (PSC) in Petunia integrifolia.

Journal of Heredity 76: 468–470.

Darlington CD. 1929. Chromosome behaviour and structural hybridity in the

Tradescantiae. Journal of Genetics 21: 207–286.

Darwin C. 1864. On the sexual relations of the three forms of Lythrum salicaria. Journal

of the Proceedings of the Linnean Society. Botany. 8: 169–196.

Darwin C. 1876. The effects of cross- and self-fertilisation in the vegetable kingdom.

London: John Murray.

LITERATURE CITED 214

Darwin C. 1877. The different forms of flowers on plants of the same species. London:

John Murray.

Devaux C, Lepers C, Porcher E. 2014. Constraints imposed by pollinator behaviour on the

ecology and evolution of plant mating systems. Journal of Evolutionary Biology

27: 1413–1430.

Dlugosch KM, Parker IM. 2008. Founding events in species invasions: genetic variation,

adaptive evolution, and the role of multiple introductions. Molecular Ecology 17:

431–449.

Dorken ME, Friedman J, Barrett SCH. 2002. The evolution and maintenance of monoecy

and dioecy in Sagittaria latifolia (Alismataceae). Evolution 56: 31–41.

Dornier A, Munoz F, Cheptou P-O. 2008. Allee effect and self-fertilization in

hermaphrodites: reproductive assurance in a structured metapopulation. Evolution

62: 2558–2569.

Dudash MR. 1990. Relative fitness of selfed and outcrossed progeny in a self-compatible,

protandrous species, Sabatia angularis L. (Gentianaceae): a comparison in three

environments. Evolution 44: 1129–1139.

Dulberger R. 1992. Floral polymorphisms and their functional significance in the

heterostylous syndrome In: Barrett SCH, ed. Evolution and function of heterostyly.

Monographs on theoretical and applied genetics. Berlin, New York: Springer-

Verlag, 41–84.

LITERATURE CITED 215

East EM. 1932. Further observations on Lythrum salicaria. Genetics 17: 327–334.

East E. 1934. Norms of pollen-tube growth in incompatible matings of self-sterile plants.

Proceedings of the National Academy of Sciences USA 20: 225–230.

Eckert CG. 1993. The evolutionary maintenance of a sexual polymorphism in clonal

Decodon verticillatus (Lythraceae). Ph.D. thesis, University of Toronto.

Eckert CG. 2002. Effect of geographical variation in pollinator fauna on the mating system

of Decodon verticillatus (Lythraceae). International Journal of Plant Sciences 163:

123–132.

Eckert CG, Barrett SCH. 1992. Stochastic loss of style morphs from populations of

tristylous Lythrum salicaria and Decodon verticillatus (Lythraceae). Evolution 46:

1014–1029.

Eckert CG, Barrett SCH. 1993. The inheritance of tristyly in Decodon verticillatus

(Lythraceae). Heredity 71: 473–480.

Eckert CG, Barrett SCH. 1994a. Post-pollination mechanisms and the maintenance of

outcrossing in self-compatible, tristylous, Decodon verticillatus (Lythraceae).

Heredity 72: 396–411.

Eckert CG, Barrett SCH. 1994b. Inbreeding depression in partially self-fertilizing

Decodon verticillatus (Lythraceae): population genetic and experimental analyses.

Evolution 48: 952–964.

LITERATURE CITED 216

Eckert CG, Barrett SCH. 1995. Style morph ratios in tristylous Decodon verticillatus

(Lythraceae): selection vs. historical contingency. Ecology 76: 1051–1066.

Eckert CG, Manicacci D, Barrett SCH. 1996a. Genetic drift and founder effect in native

versus introduced populations of an invading plant, Lythrum salicaria

(Lythraceae). Evolution 50: 1512–1519.

Eckert CG, Manicacci D, Barrett SCH. 1996b. Frequency-dependent selection on morph

ratios in tristylous Lythrum salicaria (Lythraceae). Heredity 77: 581–588.

Ellstrand NC. 1992. Gene flow by pollen: implications for plant conservation genetics.

Oikos 63: 77–86.

Ellstrand NC, Marshall DL. 1985. Interpopulation gene flow by pollen in wild radish,

Raphanus sativus. American Naturalist 126: 606–616.

Excoffier L, Foll M, Petit RJ. 2009. Genetic consequences of range expansions. Annual

Review of Ecology, Evolution, and Systematics 40: 481–501.

Facon B, Hufbauer RA, Tayeh A, Loiseau A, Lombaert E, Vitalis R, Guillemaud T,

Lundgren JG, Estoup A. 2011. Inbreeding depression is purged in the invasive

insect Harmonia axyridis. Current Biology 21: 424–427.

Ferrero V, Barrett SCH, Castro S, Caldeirinha P, Navarro L, Loureiro J, Rodríguez-

Echeverría S. 2015. Invasion genetics of the Bermuda buttercup (Oxalis pes-

caprae): complex intercontinental patterns of genetic diversity, polyploidy and

heterostyly characterize both native and introduced populations. Molecular

LITERATURE CITED 217

Ecology 24: 2143–2155.

Finney DJ. 1952. The equilibrium of a self-incompatible polymorphic species. Genetica

26: 33–64.

Finney DJ. 1983. Genetic equilibria of species with self-incompatible polymorphisms.

Biometrics 39: 573–585.

Fisher RA. 1941a. The theoretical consequences of polyploid inheritance for the mid style

form of Lythrum salicaria. Annals of Eugenics 11: 31–38.

Fisher RA. 1941b. Average excess and average effect of a gene substitution. Annals of

Eugenics 11: 53–63.

Fisher RA. 1943. Allowance for double reduction in the calculation of genotype

frequencies with polysomic inheritance. Annals of Eugenics 12: 169–171.

Fisher RA. 1949. The linkage problem in a tetrasomic wild plant, Lythrum salicaria.

Hereditas 35: 225–233.

Fisher RA, Mather K. 1943. The inheritance of style length in Lythrum salicaria. Annals

of Eugenics 12: 1–23.

Flaschenriem DR, Ascher PD. 1979. Pollen tube expression of pseudo-self-compatibility

(PSC) in Petunia hybrida. Theoretical and Applied Genetics 54: 97–101.

Fox J, Weisberg S. 2011. An R companion to applied regression. Thousand Oaks, CA:

Sage Publications.

LITERATURE CITED 218

Franklin-Tong VE. 2008. Self-incompatibility in flowering plants: evolution, diversity,

and mechanisms. Berlin: Springer.

Fyfe VC. 1953. Double reduction at the mid locus in Lythrum salicaria. Heredity 7: 285–

292.

Ganders FR. 1975. Mating patterns in self-compatible distylous populations of Amsinckia

(Boraginaceae). Canadian Journal of Botany 53: 773–779.

Ganders FR. 1979. The biology of heterostyly. New Zealand Journal of Botany 17: 607–

635.

Garcia-Serrano H, Escarre J, Cano L, Sans FX. 2008. Comparing the effect of habitat on

the magnitude of inbreeding depression in the Mediterranean native Senecio

malacitanus and the alien S. inaequidens: consequences for invasive ability.

Botany 86: 63-75.

Gettys LA, Wofford DS. 2008. Genetic control of floral morph in tristylous pickerelweed

(Pontederia cordata L.). Journal of Heredity 99: 558–563.

Gilbert N, Lee SB. 1980. Two perils of plant population dynamics. Oecologia 46: 283–

284.

Glover DE, Barrett SCH. 1983. Trimorphic incompatibility in Mexican populations of

Pontederia sagittata Presl. (Pontederiaceae). New Phytologist 95: 439–455.

Goldberg EE, Kohn JR, Lande R, Robertson KA, Smith SA, Igić B. 2010. Species

LITERATURE CITED 219

selection maintains self-incompatibility. Science 330: 493–495.

Gompertz B. 1825. On the nature of the function expressive of the law of human

mortality, and on a new mode of determining the value of life contingencies.

Philosophical Transactions of the Royal Society of London 115: 513–583.

Good-Avila SV, Frey F, Stephenson AG. 2001. The effect of partial self-incompatibility

on the breeding system of Campanula rapunculoides L. (Campanulaceae) under

conditions of natural pollination. International Journal of Plant Sciences 162:

1081–1087.

Good-Avila SV, Mena-Ali JJ, Stephenson AG. 2008. Genetic and environmental causes

and evolutionary consequences of variations in self-fertility in self incompatible

species In: Franklin-Tong V, ed. Self-incompatibility in flowering plants:

evolution, diversity, and mechanisms. Berlin: Springer-Verlag, 33–51.

Good-Avila SV, Stephenson AG. 2002. The inheritance of modifiers conferring self-

fertility in the partially self-incompatible perennial, Campanula rapunculoides L.

(Campanulaceae). Evolution 56: 263–272.

Goodwillie C. 1999. Multiple origins of self-compatibility in Linanthus section

Leptosiphon (Polemoniaceae): phylogenetic evidence from internal-transcribed-

spacer sequence data. Evolution 53: 1387–1395.

Goodwillie C, Kalisz S, Eckert CG. 2005. The evolutionary enigma of mixed mating in

plants: occurrence, theoretical explanations, and empirical evidence. Annual

LITERATURE CITED 220

Review of Ecology, Evolution, and Systematics 36: 47–79.

Graham SA. 2013. Fossil records in the Lythraceae. Botanical Review 79: 48–145.

Grossenbacher DL, Brandvain Y, Auld JR, Burd M, Cheptou P-O, Conner JK, Grant AG,

Hovick SM, Pannell JR, Pauw A, Petanidou T, Randle AM, Rubio de Casas R,

Vamosi J, Winn A, Igic B, Busch JW, Kalisz S, Goldberg EE. 2017. Self-

compatibility is over-represented on islands. New Phytologist 215: 469–478.

Haldane JBS. 1936. Some natural populations of Lythrum salicaria. Journal of Genetics

32: 393–397.

Halkka O, Halkka L. 1974. Polymorphic balance in small island populations of Lythrum

salicaria. Annals Botanici Fennici 11: 267–270.

Hamilton N. 2014. ggtern: An extension to ggplot2, for the creation of ternary diagrams. R

package version 1.0.3.2.

Henny RJ, Ascher PD. 1976. The inheritance of pseudo-self-compatibility (PSC) in

Nemesia strumosa Benth. Theoretical and Applied Genetics 48: 185–195.

Heuch I. 1979a. Equilibrium populations of heterostylous plants. Theoretical Population

Biology 15: 43–57.

Heuch I. 1979b. The effect of partial self-fertilization on type frequencies in heterostylous

plants. Annals of Botany 44: 611–616.

Heuch I. 1980. Loss of incompatibility types in finite populations of the heterostylous

LITERATURE CITED 221

plant Lythrum salicaria. Hereditas 92: 53–57.

Heuch I, Lie RT. 1985. Genotype frequencies associated with incompatibility systems in

tristylous plants. Theoretical Population Biology 27: 318–336.

Hickey M, King C. 1981. 100 families of flowering plant. New York: Cambridge

University Press.

Hiscock SJ, Allen AM. 2008. Diverse cell signalling pathways regulate pollen-stigma

interactions: the search for consensus. New Phytologist 179: 286–317.

Hodgins KA, Barrett SCH. 2006. Female reproductive success and the evolution of

mating-type frequencies in tristylous populations. New Phytologist 171: 569–580.

Hodgins KA, Barrett SCH. 2008. Asymmetrical mating patterns and the evolution of

biased style-morph ratios in a tristylous daffodil. Genetic Research 90: 3–15.

Høeg. 1944. The ratio of different forms of flowers in Lythrum salicaria L. in Norway.

Blyttia 2: 71–74.

Hopper JE, Ascher PD, Peloquin SJ. 1967. Inactivation of self-incompatibility following

temperature pretreatments of styles in Lillium longiflorum. Euphytica 16: 215–220.

Hunt R. 1981. The fitted curve in plant growth studies In: Rose DA, Charles-Edwards DA,

eds. Mathematics and plant physiology. London: Academic Press, 283–298.

Husband BC, Barrett SCH. 1992a. Genetic drift and the maintenance of the style length

polymorphism in tristylous populations of Eichhornia paniculata (Pontederiaceae).

LITERATURE CITED 222

Heredity 69: 440–449.

Husband BC, Barrett SCH. 1992b. Effective population size and genetic drift in tristylous

Eichhornia paniculata (Pontederiaceae). Evolution 46: 1875–1890.

Husband BC, Schemske DW. 1996. Evolution of the magnitude and timing of inbreeding

depression in plants. Evolution 50: 54–70.

Husband BC, Schemske DW. 1997. The effect of inbreeding in diploid and tetraploid

populations of Epilobium angustifolium (Onagraceae): implications for the genetic

basis of inbreeding depression. Evolution 51: 737–746.

Igić B, Busch JW. 2013. Is self-fertilization an evolutionary dead end? New Phytologist

198: 386–397.

Igić B, Kohn JR. 2001. Evolutionary relationships among self-incompatibility RNases.

Proceedings of the National Academy of Sciences USA 98: 13167–13171.

Igić B, Lande R, Kohn JR. 2008. Loss of self-incompatibility and its evolutionary

consequences. International Journal of Plant Sciences 169: 93–104.

Ishihama F, Nakano C, Ueno S, Ajima M, Tsumura Y, Washitani I. 2003. Seed set and

gene flow patterns in an experimental population of an endangered heterostylous

herb with controlled local opposite-morph density. Functional Ecology 17: 680–

689.

Ison JL, Wagenius S, Reitz D, Ashley MV. 2014. Mating between Echinacea angustifolia

LITERATURE CITED 223

(Asteraceae) individuals increases with their flowering synchrony and spatial

proximity. American Journal of Botany 101: 180–189.

Iwaizumi MG, Takahashi M, Isoda K, Austerlitz F. 2013. Consecutive five-year analysis

of paternal and maternal gene flow and contributions of gametic heterogeneities to

overall genetic composition of dispersed seeds of Pinus densiflora (Pinaceae).

American Journal of Botany 100: 1896–1904.

Jennersten O. 1988. Pollination in Dianthus deltoides (Caryophyllaceae): effects of habitat

fragmentation on visitation and seed set. Conservation Biology 2: 359–366.

Johnston MO, Schoen DJ. 1994. On the measurement of inbreeding depression. Evolution

48: 1735–1741.

Johnson MTJ, Campbell SA, Barrett SCH. 2015. Evolutionary interactions between plant

reproduction and defense against herbivores. Annual Review of Ecology,

Evolution, and Systematics 46: 191–213.

Kahle D, Wickham H. 2013. ggmap: Spatial visualization with ggplot2. The R Journal 5:

144–161.

Kariyat RR, Balogh CM, Moraski RP, De Moraes CM, Mescher MC, Stephenson AG.

2013. Constitutive and herbivore-induced structural defenses are compromised by

inbreeding in Solanum carolinense (Solanaceae). American Journal of Botany 100:

1014–1021.

Kariyat RR, Mauck KE, Balogh CM, Stephenson AG, Moraes CM De, Mescher MC.

LITERATURE CITED 224

2013. Inbreeding in horsenettle (Solanum carolinense) alters night-time volatile

emissions that guide oviposition by Manduca sexta moths. Proceedings of the

Royal Society of London Ser. B 280: 1–8.

Kariyat RR, Scanlon SR, Mescher MC, De Moraes CM, Stephenson AG. 2011. Inbreeding

depression in Solanum carolinense (Solanaceae) under field conditions and

implications for mating system evolution. PloS One 6: e28459.

Keller SR, Fields PD, Berardi AE, Taylor DR. 2014. Recent admixture generates

heterozygosity-fitness correlations during the range expansion of an invading

species. Journal of Evolutionary Biology 27: 616–627.

Keller LF, Waller DM. 2002. Inbreeding effects in wild populations. Trends in Ecology &

Evolution 17: 230–241.

Kendall WA, Taylor NL. 1969. Effect of temperature on pseudo-self-compatibility in

Trifolium pratense L. Theoretical and Applied Genetics 26: 123–126.

King VM, Sargent RD. 2012. Presence of an invasive plant species alters pollinator

visitation to a native. Biological Invasions 14: 1809–1818.

Kirkpatrick M, Jarne P. 2000. The effects of a bottleneck on inbreeding depression and the

genetic load. American Naturalist 155: 154–167. van Kleunen M, Johnson SD. 2007. Effects of self-compatibility on the distribution range

of invasive European plants in North America. Conservation Biology 21: 1537–

1544.

LITERATURE CITED 225

van Kleunen M, Manning JC, Pasqualetto V, Johnson SD. 2008. Phylogenetically

independent associations between autonomous self-fertilization and plant

invasiveness. American Naturalist 171: 195–201.

Knight ME, Martin AP, Bishop S, Osborne JL, Hale RJ, Sanderson RA, Goulson D. 2005.

An interspecific comparison of foraging range and nest density of four bumblebee

(Bombus) species. Molecular Ecology 14: 1811–1820.

Koelling VA, Monnahan PJ, Kelly JK. 2012. A Bayesian method for the joint estimation

of outcrossing rate and inbreeding depression. Heredity 109: 393–400.

Kohn JR. 1988. Why be female? Nature 335: 431–433.

Kubátová B, Trávníček P, Bastlová D, Čurn V, Jarolímová V, Suda J. 2008. DNA ploidy-

level variation in native and invasive populations of Lythrum salicaria at a large

geographical scale. Journal of Biogeography 35: 167–176.

Lande R, Schemske DW. 1985. The evolution of self-fertilization and inbreeding

depression in plants. I. Genetic models. Evolution 39: 24–40.

Lande R, Schemske DW, Schultz ST. 1994. High inbreeding depression, selective

interference among loci, and the threshold selfing rate for purging recessive lethal

mutations. Evolution 48: 965–978.

Lane MD, Lawrence MJ. 1993. The population genetics of the self-incompatibility

polymorphism in Papaver rhoeas. VII. The number of S-alleles in the species.

Heredity 71: 596–602.

LITERATURE CITED 226

Larson BMH, Barrett SCH. 1999. The ecology of pollen limitation in buzz-pollinated

Rhexia virginica (Melastomataceae). Journal of Ecology 86: 502–511.

Larson BMH, Barrett SCH. 2000. A comparative analysis of pollen limitation in flowering

plants. Biological Journal of the Linnean Society 69: 503–520.

Lenth R. 2018. emmeans: Estimated marginal means, aka least-squares means. R package

version 1.1. https://CRAN.R-project.org/package=emmeans.

Levin DA. 1995. Plant outliers: an ecogenetic perspective. American Naturalist 145: 109–

118.

Levin DA. 1996. The evolutionary significance of pseudo-self-fertility. American

Naturalist 148: 321–332.

Lewis D, Jones D. 1992. The genetics of heterostyly In: Barrett SCH, ed. Evolution and

function of heterostyly. Monographs on theoretical and applied genetics. Berlin:

Springer-Verlag, 129–150.

Lewontin R. 1965. Selection for colonizing ability In: Baker HG, Stebbins GL, eds. The

genetics of colonizing species. New York: Academic Press, 77–94.

Lloyd DG. 1965. Evolution of self-compatibility and racial differentiation in

Leavenworthia (Cruciferae). Contributions from the Gray Herbarium of Harvard

University 195: 3–134.

Lloyd DG. 1979. Some reproductive factors affecting the selection of self-fertilization in

LITERATURE CITED 227

plants. American Naturalist 113: 67–79.

Lloyd DG. 1980. Demographic factors and mating patterns in angiosperms In: Solbrig OT,

ed. Demography and evolution in plant populations. Oxford: Blackwell Scientific

Publications, 67–88.

Lloyd DG. 1992. Self- and cross-fertilization in plants. II. The selection of self-

fertilization. International Journal of Plant Sciences 153: 370–380.

Lloyd DG, Schoen DJ. 1992. Self- and cross-fertilization in plants. I. Functional

dimensions. International Journal of Plant Sciences 153: 358–369.

Lunau K. 1996. Signalling functions of floral colour patterns for insect flower visitors.

Zoologischer Anzeiger 235: 11–30.

Lundqvist A. 1968. The mode of origin of self-fertility in grasses. Hereditas 59: 413–426.

Lynch M, Walsh B. 1998. Genetics and analysis of quantitative traits. Sunderland, MA:

Sinauer Associates Inc.

Mable BK. 2008. Genetic causes and consequences of the breakdown of self-

incompatibility: case studies in the Brassicaceae. Genetics Research 90: 47–60.

Mable BK, Robertson A V, Dart S, Di Berardo C, Witham L. 2005. Breakdown of self-

incompatibility in the perennial Arabidopsis lyrata (Brassicaceae) and its genetic

consequences. Evolution 59: 1437–1448.

Mal TK, Lovett-Doust J, Lovett-Doust L. 1999. Maternal and paternal success among

LITERATURE CITED 228

flower morphs in tristylous Lythrum salicaria. Aquatic Botany 63: 229–239.

Mal TK, Lovett-Doust J, Lovett-Doust L, Mulligan GA. 1992. The biology of Canadian

weeds. 100. Lythrum salicaria. Canadian Journal of Plant Science 72: 1305–1330.

Marchini GL, Sherlock NC, Ramakrishnan AP, Rosenthal DM, Cruzan MB. 2016. Rapid

purging of genetic load in a metapopulation and consequences for range expansion

in an invasive plant. Biological Invasions 18: 183–196.

Maron JL, Vila M, Bommarco R, Elmendorf S, Beardsley P. 2004. Rapid evolution of an

invasive plant. Ecological Monographs 74: 261–280.

McCommas SA, Bryant EH. 1990. Loss of electrophoretic variation in serially

bottlenecked populations. Heredity 64: 315–321.

McCubbin AG, Kao T. 2000. Molecular recognition and response in pollen and pistil

interactions. Annual Reviews of Cell Developmental Biology 16: 333–364.

Mena-Ali JI, Stephenson AG. 2007. Segregation analyses of partial self-incompatibility in

self and cross progeny of Solanum carolinense reveal a leaky S-allele. Genetics

177: 501–510.

Moeller DA, Geber MA. 2005. Ecological context of the evolution of self-pollination in

Clarkia xantiana: population size, plant communities, and reproductive assurance.

Evolution 59: 786–799.

Montague JL, Barrett SCH, Eckert CG. 2008. Re-establishment of clinal variation in

LITERATURE CITED 229

flowering time among introduced populations of purple loosestrife (Lythrum

salicaria, Lythraceae). Journal of Evolutionary Biology 21: 234–245.

Moran PPA. 1962. The statistical process of evolutionary theory. Oxford: Clarendon

Press.

Morgan MT, Barrett SCH. 1988. Historical factors and anisoplethic population structure in

tristylous Pontederia cordata: a reassessment. Evolution 42: 496–504.

Mulcahy DL, Caporello D. 1970. Pollen flow within a tristylous species: Lythrum

salicaria. American Journal of Botany 57: 1027–1030.

Mullarkey AA, Byers DL, Anderson RC. 2013. Inbreeding depression and partitioning of

genetic load in the invasive biennial Alliaria petiolata (Brassicaceae). American

Journal of Botany 100: 509–518.

Muller-Landau HC, Condit RS, Chave J, Thomas SC, Bohlman SA, Bunyavejchewin S,

Davies S, Foster R, Gunatilleke S, Gunatilleke N, Harms KE, Hart T, Hubbell SP,

Itoh A, Kassim AR, LaFrankie J V., Lee HS, Losos E, Makana JR, Ohkubo T,

Sukumar R, Sun IF, Nur Supardi MN, Tan S, Thompson J, Valencia R, Muñoz

GV, Wills C, Yamakura T, Chuyong G, Dattaraja HS, Esufali S, Hall P, Hernandez

C, Kenfack D, Kiratiprayoon S, Suresh HS, Thomas D, Vallejo MI, Ashton P.

2006. Testing metabolic ecology theory for allometric scaling of tree size, growth

and mortality in tropical forests. Ecology Letters 9: 575–588.

Nason JD., Ellstrand NC. 1995. Lifetime estimates of biparental inbreeding depression in

LITERATURE CITED 230

the self-incompatible annual plant Raphanus sativus. Evolution 49: 307–316.

Nason JD, Herre EA, Hamrick JL. 1998. The breeding structure of a tropical keystone

plant resource. Nature 391: 685–687.

Nei M. 1973. Analysis of gene diversity in subdivided populations. Proceedings of the

National Academy of Sciences USA 70: 3321–3323. de Nettancourt D. 1977. Incompatibility in angiosperms. Berlin: Springer-Verlag.

Novak SJ, Mack RN. 2005. Genetic bottlenecks in alien plant species: influence of mating

systems and introduction dynamics In: Sax DF, Stachowicz JJ, Gaines SD, eds.

Species invasions: insights into ecology, evolution, and biogeography. Sunderland,

MA: Sinauer Associates Inc., 201–228.

O’Neil P. 1994. Genetic incompatibility and offspring quality in the tristylous plant

Lythrum salicaria (Lythraceae). American Journal of Botany 81: 76–84.

Olsson K, Ågren J. 2002. Latitudinal population differentiation in phenology, life history

and flower morphology in the perennial herb Lythrum salicaria. Journal of

Evolutionary Biology 15: 983–996.

Oostermeijer JGB, Dennijs JCM, Raijmann LEL, Menken SBJ. 1992. Population biology

and management of the marsh gentian (Gentiana neumonanthe L.), a rare species

in The Netherlands. Botanical Journal of the Linnean Society 108: 117–130.

Ornduff R. 1972. The breakdown of trimorphic incompatibility in Oxalis section

LITERATURE CITED 231

Corniculatae. Evolution 26: 52–65.

Osborne JL, Clark SJ, Morris RJ, Williams IH, Riley JR, Smith AD, Reynolds DR,

Edwards AS. 1999. A landscape-scale of bumble bee foraging study range and

constancy, using harmonic radar. Journal of Applied Ecology 36: 519–533.

Owen F V. 1942. Inheritance of cross- and self-sterility and self-fertility in Beta vulgaris.

Journal of Agricultural Research 64: 679–698.

Paine CET, Marthews TR, Vogt DR, Purves D, Rees M, Hector A, Turnbull LA. 2012.

How to fit nonlinear plant growth models and calculate growth rates: an update for

ecologists. Methods in Ecology and Evolution 3: 245–256.

Pannell JR. 2015. Evolution of the mating system in colonizing plants. Molecular Ecology

24: 2018–2037.

Pannell JR, Auld JR, Brandvain Y, Burd M, Busch JW, Cheptou P-O, Conner JK,

Goldberg EE, Grant A, Grossenbacher DL, Hovick SM, Igic B, Kalisz S,

Petanidou T, Randle AM, Casas RR De, Pauw A, Vamosi JC. 2015. The scope of

Baker’s law. New Phytologist 208: 656–667.

Pannell JR, Barrett SCH. 1998. Baker’s law revisited: reproductive assurance in a

metapopulation. Evolution 52: 657–668.

Parisod C, Trippi C, Galland N. 2005. Genetic variability and founder effect in the pitcher

plant Sarracenia purpurea (Sarraceniaceae) in populations introduced into

Switzerland: from inbreeding to invasion. Annals of Botany 95: 277–286.

LITERATURE CITED 232

Paul-Victor C, Züst T, Rees M, Kliebenstein DJ, Turnbull LA. 2010. A new method for

measuring relative growth rate can uncover the costs of defensive compounds in

Arabidopsis thaliana. New Phytologist 187: 1102–1111.

Peischl S, Dupanloup I, Kirkpatrick M, Excoffier L. 2013. On the accumulation of

deleterious mutations during range expansions. Molecular Ecology 22: 5972–5982.

Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team. 2018. nlme: Linear and nonlinear

mixed effects models.

Puentes A, Cole WW, Barrett SCH. 2013. Trimorphic incompatibility in Pontederia

subovata (Pontederiaceae): an aquatic macrophyte from lowland South America.

International Journal of Plant Sciences 174: 47–56.

Pujol B, Zhou S-R, Sanchez Vilas J, Pannell JR. 2009. Reduced inbreeding depression

after species range expansion. Proceedings of the National Academy of Sciences

USA 106: 15379–15383.

Raduski AR, Haney EB, Igić B. 2012. The expression of self-incompatibility in

angiosperms is bimodal. Evolution 66: 1275–1283.

Rambuda TD, Johnson SD. 2004. Breeding systems of invasive alien plants in South

Africa: does Baker’s rule apply? Diversity and Distributions 10: 409–416.

R Core Team. 2014. R: A language and enviroment for statistical computing (V 3.1.1

‘Sock it to me’). Vienna, Austria: R Foundation for Statistical Computing.

LITERATURE CITED 233

R Core Team. 2016. R: A language and enviroment for statistical computing (V. 3.3.2

‘Sincere Pumpkin Patch’). Vienna, Austria: R Foundation for Statistical

Computing.

Renner SS, Ricklefs RE. 1995. Dioecy and its correlates in the flowering plants. American

Journal of Botany 82: 596–606.

Richards FJ. 1959. A flexible growth function for empirical use. Journal of Experimental

Botany 10: 290–301.

Richards R, Thurling N. 1973. The genetics of self-incompatibility in Brassica

campenstris L. ssp Oleifera Metzg. I. Characteristics of S-locus control of self-

incompatibility. Genetica 44: 439–453.

Ritland K. 1990. Inferences about inbreeding depression based on changes of the

inbreeding coefficient. Evolution 44: 1230–1241.

Robacker CD, Ascher PD. 1978. Restoration of pseudo-self-compatibility (PSC) in

derivatives of a high-PSC x no-PSC cross in Nemesia strumosa Benth. Theoretical

and Applied Genetics 53: 135–141.

Roman J, Darling JA. 2007. Paradox lost: genetic diversity and the success of aquatic

invasions. Trends in Ecology and Evolution 22: 454–464.

Rosche C, Hensen I, Mráz P, Durka W, Hartmann M, Lachmuth S. 2017. Invasion success

in polyploids: the role of inbreeding in the contrasting colonization abilities of

diploid versus tetraploid populations of Centaurea stoebe s.l. Journal of Ecology

LITERATURE CITED 234

105: 425–435.

Rose KE, Atkinson RL, Turnbull LA, Rees M. 2009. The costs and benefits of fast living.

Ecology Letters 12: 1379–1384. van Rossum F, Stiers I, van Geert A, Triest L, Hardy OJ. 2011. Fluorescent dye particles

as pollen analogues for measuring pollen dispersal in an insect-pollinated forest

herb. Oecologia 165: 663–674.

Sakai AK, Karoly K, Weller SG. 1989. Inbreeding depression in Schiedea globosa and S.

salicaria (Caryophyllaceae), subdioecious and gynodioecious Hawaiian species.

American Journal of Botany 76: 437–444.

Sakai AK, Allendorf FW, Holt JS, Lodge DM, Molofsky J, With KA, Baughman S, Cabin

RJ, Cohen JE, Ellstrand NC, McCauley DE, O’Neil P, Parker IM, Thompson JN,

Weller SG. 2001. The population biology of invasive species. Annual Review of

Ecology, Evolution, and Systematics 32: 305–332.

Sandner TM, Matthies D. 2016. The effects of stress intensity and stress type on

inbreeding depression in Silene vulgaris. Evolution 70: 1225–1238.

Schmitt J, Eccleston J, Ehrhardt D. 1987. Dominance and suppression, size-dependent

growth and self-thinning in a natural Impatiens capensis population. Journal of

Ecology 75: 651–665.

Schmitt J, Ehrhardt D. 1990. Enhancement of inbreeding depression by dominance and

suppression in Impatiens capensis. Evolution 44: 269–278.

LITERATURE CITED 235

Schoch-Bodmer H. 1938. The proportion of long-, mid- and short-styled plants in natural

populations of Lythrum salicaria L. Journal of Genetics 36: 39–43.

Schooler SS, McEvoy PB, Coombs EM. 2006. Negative per capita effects of purple

loosestrife and reed canary grass on plant diversity of wetland communities.

Diversity and Distributions 12: 351–363.

Schrieber K, Lachmuth S. 2017. The genetic paradox of invasions revisited: the potential

role of inbreeding × environment interactions in invasion success. Biological

Reviews 941: 939–952.

Scribailo RW, Barrett SCH. 1991. Pollen-stigma interactions in tristylous Pontederia

sagittata (Pontederiaceae). II. Patterns of pollen tube growth. American Journal of

Botany 78: 1662–1682.

Shamsi SRA, Whitehead FH. 1974. Comparative eco-physiology of Epilobium hirsutum

L. and Lythrum salicaria L. I. General biology, distribution, and germination.

Journal of Ecology 65: 71–84.

Shigesada N, Kawasaki K. 1997. Biological invasions: theory and practice. Oxford:

Oxford University Press.

Shore JS, Barrett SCH. 1986. Genetic modifications of dimorphic incompatibility in the

Turnera ulmifolia L. complex (Turneraceae). Canadian Journal of Genetics and

Cytology 28: 796–807.

Simpson EH. 1949. Measurement of diversity. Nature 163: 688–688.

LITERATURE CITED 236

Skogsmyr I. 1994. Gene dispersal from transgenic potatoes to conspecifics - a field trial.

Theoretical and Applied Genetics 88: 770–774.

Smyth CA, Hamrick JL. 1984. Variation in estimates of outcrossing in musk thistle

populations. Journal of Heredity 75: 303–307.

Spieth PT. 1971. A necessary condition for equilibrium in systems exhibiting self-

incompatible mating. Theoretical Population Biology 2: 404–418.

Spieth PT, Novitski E. 1969. Remarks on equilibrium conditions in certain trimorphic,

self-incompatible systems In: Morton NE, ed. Computer applications in genetics.

Honolulu: University of Hawaii Press, 161–167.

Stebbins GL. 1957. Self fertilization and population variability in the higher plants.

American Naturalist 91: 337–354.

Stebbins GL. 1974. Flowering plants: evolution above the species level. Cambridge, MA:

Belknap Press.

Steffan-Dewenter AI, Tscharntke T. 1999. Effects of habitat isolation on pollinator

communities and seed set. Oecologia 121: 432–440.

Stehlik I, Caspersen JP, Barrett SCH. 2006. Spatial ecology of mating success in a

sexually polymorphic plant. Proceedings of the Royal Society of London Ser. B

273: 387–394.

Steinbachs JE, Holsinger KE. 2002. S-RNase-mediated gametophytic self-incompatibility

LITERATURE CITED 237

is ancestral in eudicots. Molecular Biology and Evolution 19: 825–829.

Stephenson AG, Good-Avila SV, Vogler DW. 2000. Interrelationships among inbreeding

depression, plasticity in the self-incompatibility system, and the breeding system of

Campanula rapunculoides L. (Campanulaceae). Annals of Botany 85: 211–219.

Stephenson AG, Travers SE, Mena-Ali JI, Winsor JA. 2003. Pollen performance before

and during the autotrophic-heterotrophic transition of pollen tube growth.

Philosophical Transactions of the Royal Society of London. Ser. B 358: 1009–

1018.

Stone JL. 2002. Molecular mechanisms underlying the breakdown of gametophytic self-

incompatibility. Quarterly Review of Biology 77: 17–32.

Stone JL, Sasuclark MA, Blomberg CP. 2006. Variation in the self-incompatibility

response within and among populations of the tropical shrub Witheringia

solanacea (Solanaceae). American Journal of Botany 93: 592–598.

Stout AB. 1923. Studies of Lythrum salicaria I. The efficiency of self-pollination.

American Journal of Botany 10: 440–449.

Stout AB. 1925. Studies of Lythrum salicaria-II a new form of flower in this species.

Bulletin of the Torrey Botanical Club 52: 81–85.

Sutherland S. 2004. What makes a weed a weed: life history traits of native and exotic

plants in the USA. Oecologia 141: 24–39.

LITERATURE CITED 238

Terrie K, Arriola PE, Ellstrand NC. 1992. Crop-weed hybridization in radish (Raphanus

sativus): effects of distance and population size. American Journal of Botany 79:

1431–1435.

Thompson DQ, Stuckey RL, Thompson EB. 1987. Spread, impact, and control of purple

loosestrife (Lythrum salicaria) in North American wetlands. U.S. Fish and Wildlife

Service.

Travers SE, Mena-Ali J, Stephenson AG. 2004. Plasticity in the self-incompatibility

system of Solanum carolinense. Plant Species Biology 19: 127–136.

Tsuchimatsu T, Suwabe K, Shimizu-Inatsugi R, Isokawa S, Pavlidis P, Städler T, Suzuki

G, Takayama S, Watanabe M, Shimizu KK. 2010. Evolution of self-compatibility

in Arabidopsis by a mutation in the male specificity gene. Nature 464: 1342–1347. von Ubisch, G. 1925. Genetisch-pysiologisch analyse der heterostylie. Biobliographia genetica 2:287-340.

Uyenoyama MK, Holsinger KE, Waller DM. 1993. Ecological and genetic factors

directing the evolution of self-fertilization. In: Futuyma D, Antonovics J, eds.

Oxford surveys in evolutionary biology.Volume 9. Oxford: Oxford University

Press, 327–381.

Vekemans X, Schierup MH, Christiansen FB. 1998. Mate availability and fecundity

selection in multi-allelic self-incompatibility systems in plants. Evolution 52: 19–

29.

LITERATURE CITED 239

Verhoeven KJF, Macel M, Wolfe LM, Biere A. 2011. Population admixture, biological

invasions and the balance between local adaptation and inbreeding depression.

Proceedings of the Royal Society of London Ser. B 278: 2–8.

Vogler DW, Das C, Stephenson AG. 1998. Phenotypic plasticity in the expression of self-

incompatibility in Campanula rapunculoides. Heredity 81: 546–555.

Vogler DW, Filmore K, Stephenson AG. 1999. Inbreeding depression in Campanula

rapunculoides L. I. A comparison of inbreeding depression in plants derived from

strong and weak self-incompatibility phenotypes. Journal of Evolutionary Biology

12: 483–494.

Waites AR, Ågren J. 2004. Pollinator visitation, stigmatic pollen loads and among-

population variation in seed set in Lythrum salicaria. Journal of Ecology 95: 512–

526.

Waites AR, Ågren J. 2006. Stigma receptivity and effects of prior self-pollination on seed

set in tristylous Lythrum salicaria (Lythraceae). American Journal of Botany 93:

142–147.

Walther-Hellwig K, Frankl R. 2000. Foraging distances of Bombus muscorum, Bombus

lapidarius, and Bombus terrestris (Hymenoptera, Apidae). Journal of Insect

Behavior 13: 239–246.

Wang Y, Wang QF, Guo YH, Barrett SCH. 2005. Reproductive consequences of

interactions between clonal growth and sexual reproduction in Nymphoides peltata:

LITERATURE CITED 240

a distylous aquatic plant. New Phytologist 165: 329–336.

Weber JJ, Weller SG, Sakai AK, Tsyusko O V., Glenn TC, Domínguez CA, Molina-

Freaner FE, Fornoni J, Tran M, Nguyen N, Nguyen K, Tran LK, Joice G, Harding

E. 2013. The role of inbreeding depression and mating system in the evolution of

heterostyly. Evolution 67: 2309–2322.

Wei T, Simko V. 2017. R package ‘corrplot’: visualization of a correlation matrix.

Weller SG. 1976. The genetic control of tristyly in Oxalis section Ionoxalis. Heredity 37:

387–393.

Weller SG. 1986. Factors influencing frequency of the mid-styled morph in tristylous

populations of Oxalis alpina. Evolution 40: 279–289.

Weller SG. 1992. Evolutionary modification of tristylous breeding systems In: Barrett

SCH, ed. Evolution and function of heterostyly. Monographs on theoretical and

applied genetics. Berlin: Springer-Verlag, 247–272.

Weller SG. 2009. The different forms of flowers - what have we learned since Darwin?

Botanical Journal of the Linnean Society 160: 249–261.

Weller SG, Ornduff R. 1977. Cryptic self-incompatibility in Amsinckia grandiflora.

Evolution 31: 47–51.

Weller SG, Sakai AK, Gray T, Weber JJ, Tsyusko O V., Domínguez CA, Fornoni J,

Molina-Freaner FE. 2016. Variation in heterostylous breeding systems in

LITERATURE CITED 241

neighbouring populations of Oxalis alpina (Oxalidaceae). Plant Biology 18: 104–

110.

White GM, Boshier DH, 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.

Winsor CP. 1932. The Gompertz curve as a growth curve. Proceedings of the National

Academy of Sciences USA 18: 1–8.

Wolfe LM. 1993. Inbreeding depression in Hydrophyllum appendiculatum: role of

maternal effects, crowding, and parental mating history. Evolution 47: 374–386.

Wright S. 1969. The theory of gene frequencies. Chicago: University of Chicago Press.

Wright SI, Kalisz S, Slotte T. 2013. Evolutionary consequences of self-fertilization in

plants. Proceedings of the Royal Society London Ser. B 280: 20130133.

Yakimowski SB, Hager HA, Eckert CG. 2005. Limits and effects of invasion by the

nonindigenous wetland plant Lythrum salicaria (purple loosestrife): a seed bank

analysis. Biological Invasions 7: 687–698.

Young AG, Pickup M. 2010. Low S-allele numbers limit mate availability, reduce seed set

and skew fitness in small populations of a self-incompatible plant. Journal of

Applied Ecology 47: 541–548.

Young HJ, Young TP. 1992. Alternative outcomes of natural and experimental high pollen

LITERATURE CITED 242

loads. Ecology 73: 639–647.

Yun L, Agrawal AF. 2014. Variation in the strength of inbreeding depression across

environments: effects of stress and density dependence. Evolution 68: 3599–3606.

Zeide B. 1993. Analysis of growth equations. Forest Science 39: 594–616.

Zhou W, Barrett SCH, Wang H, Li DZ. 2015. Reciprocal herkogamy promotes

disassortative mating in a distylous species with intramorph compatibility. New

Phytologist 206: 1503–1512.

Züst T, Joseph B, Shimizu KK, Kliebenstein DJ, Turnbull LA. 2011. Using knockout

mutants to reveal the growth costs of defensive traits. Proceedings of the Royal

Society of London Ser. B 278: 2598–2603.

APPENDIX TO CHAPTER 2

Appendix 2 Table 1. Effects of model terms on the number of generations to allele loss in simulated tristylous populations.

Optimized model terms describing the mean number of generations before the loss of each allele at the tristyly loci in populations of different size and with different partially selfing morphs. Terms are separated by selfing morph, population size, and allele lost. The models were optimized using the “step” function until an optimized Bayesian Information Criterion (BIC) was obtained. The saturated model BIC and the BIC of the optimized model are presented.

Selfing morph N Allele lost Optimal model BIC Saturated model BIC Significant coefficients

L-morph 9 m-allele 8.29 21.71 Intercept, s2, ploidy

L-morph 9 M-allele 774.22 802.83 Intercept, s, s3

L-morph 9 S-allele 2443.95 2471.05 Intercept, s, ploidy

M-morph 9 M-allele 391.51 423.05 Intercept, s2

M-morph 9 m-allele 49.45 71.31 Intercept, s2, ploidy

M-morph 9 S-allele 3138.07 3170.89 Intercept, s2, s3

S-morph 9 m-allele 153.83 179.41 Intercept, s2, ploidy

S-morph 9 M-allele 1763.16 1781.69 Intercept, s, s2, ploidy, s2:ploidy

243 APPENDIX TO CHAPTER 2 244

S-morph 9 s-allele -215.44 -203.22 Intercept, s2, ploidy, s2:ploidy

S-morph 9 S-allele 1469.82 1496.57 Intercept, s, s2, ploidy

L-morph 24 m-allele 23.02 24.32 Intercept

L-morph 24 M-allele 1035.57 1071.07 Intercept, s

L-morph 24 S-allele 2001.69 2030.62 Intercept, s, s2, s3

M-morph 24 M-allele 296.92 312.16 Intercept, s2

M-morph 24 m-allele 138.2 156.29 Intercept, s2, ploidy

M-morph 24 S-allele 3232.1 3267.28 Intercept, s2, s3

S-morph 24 M-allele 2832.42 2863.23 Intercept, s2, ploidy

S-morph 24 m-allele 301.7 322.35 Intercept, s, s2, ploidy

S-morph 24 S-allele 578.99 598.05 Intercept, s2

S-morph 24 s-allele -69.34 -67 Intercept, s, s2, ploidy

L-morph 42 M-allele 773.78 806.44 Intercept, s2, s3

L-morph 42 S-allele 1367.42 1392.36 Intercept, s, s2, s3

M-morph 42 M-allele 12.06 14.4 Intercept

M-morph 42 m-allele 108.29 129.4 Intercept, s3, ploidy

APPENDIX TO CHAPTER 2 245

M-morph 42 S-allele 2483.99 2513.52 Intercept, s, s2, s3

S-morph 42 M-allele 1819.7 1856.92 Intercept, s3

S-morph 42 m-allele 136.09 157.93 Intercept, s3, ploidy

S-morph 42 s-allele -262.23 -262.23 Intercept, s, ploidy, s:ploidy

S-morph 42 S-allele 70.9 70.9 Intercept, s, ploidy, s:ploidy

L-morph 72 M-allele 543.6 565.17 Intercept, s, s2, s3

L-morph 72 S-allele 1045.99 1065.15 Intercept, s, s2, s3

M-morph 72 m-allele 39.9 50.98 Intercept

M-morph 72 S-allele 1470.67 1491.51 Intercept, s, s2, s3

S-morph 72 m-allele 59.76 70.54 Intercept

S-morph 72 M-allele 1060.47 1090.6 Intercept, s3

S-morph 72 s-allele -350.74 -350.74 Intercept, ploidy

L-morph 102 M-allele 377.57 390.17 Intercept, s, s2, s3

L-morph 102 S-allele 674.3 699.13 Intercept, s, s2, s3

M-morph 102 m-allele 34.62 34.62 Intercept, s, ploidy

M-morph 102 S-allele 1221.33 1252.58 Intercept, s2

APPENDIX TO CHAPTER 2 246

S-morph 102 M-allele 644.42 667.09 Intercept, s2, s3

S-morph 102 m-allele 0.86 6.64 Intercept, s, s2

S-morph 102 s-allele -379.93 -379.93 Intercept, ploidy

L-morph 201 M-allele 245.32 260.53 Intercept, s, s2

L-morph 201 S-allele 400.04 417.91 Intercept, s, s2

M-morph 201 m-allele 9.33 9.67 Intercept

M-morph 201 S-allele 617.92 624.32 Intercept, s, s2, ploidy, s2:ploidy

S-morph 201 M-allele 150.93 160.39 Intercept, s, s2

S-morph 201 s-allele -550.03 -550.03 Intercept, ploidy

APPENDIX TO CHAPTER 2 247

Appendix 2 Table 2. Optimized models of the frequency of allele loss in simulated tristylous populations. Optimized model terms describing the frequency of allele loss at the tristyly loci in populations of different size and with different partially selfing morphs. Terms are separated by selfing morph, population size, and allele lost. The models were optimized using the

“step” function until an optimized Bayesian Information Criterion (BIC) was obtained. The saturated model BIC and the BIC of the optimized model are presented.

Selfing Optimal model Saturated model N Allele lost Significant coefficients morph BIC BIC

L-morph 9 M-allele 4507.23 4547.11 Intercept, s2

L-morph 9 m-allele 543.56 582.31 Intercept, s, ploidy

No alleles L-morph 9 8.19 65.51 Intercept lost

L-morph 9 s-allele 8.19 65.51 Intercept

APPENDIX TO CHAPTER 2 248

L-morph 9 S-allele 4564.33 4607.62 Intercept, s2

M-morph 9 m-allele 1310.78 1340.14 Intercept, s3, ploidy

M-morph 9 M-allele 2219.97 2262.44 Intercept, s

No alleles M-morph 9 8.19 65.51 Intercept lost

M-morph 9 s-allele 8.19 65.51 Intercept

M-morph 9 S-allele 2782.16 2813.53 Intercept, s, ploidy

S-morph 9 m-allele 2241.9 2271.06 Intercept, s, s2, ploidy

S-morph 9 M-allele 4714.79 4734.09 Intercept, s, s2, ploidy, s2:ploidy

No alleles S-morph 9 8.19 65.51 Intercept lost

APPENDIX TO CHAPTER 2 249

S-morph 9 S-allele 3392.3 3423.93 Intercept, s, ploidy

S-morph 9 s-allele 1591.63 1620.29 Intercept, s, s2, ploidy

L-morph 24 M-allele 4246.61 4286.39 Intercept, s2, s3

L-morph 24 m-allele 140.16 188.03 Intercept, s

No alleles L-morph 24 950.71 994.82 Intercept, s2 lost

L-morph 24 s-allele 8.13 65.07 Intercept

L-morph 24 S-allele 4260.96 4300.33 Intercept, s2, s3

M-morph 24 m-allele 1254.4 1283.13 Intercept, s2, s3, ploidy

M-morph 24 M-allele 872.68 914.46 Intercept, s

3 M-morph 24 No alleles 1415.85 1450.85 Intercept, s, s

APPENDIX TO CHAPTER 2 250

lost

M-morph 24 s-allele 8.11 64.86 Intercept

M-morph 24 S-allele 1900.79 1916.84 Intercept, s, s2, s3, ploidy, s:ploidy

S-morph 24 m-allele 1862.95 1890.98 Intercept, s2, s3, ploidy

S-morph 24 M-allele 3229.5 3245.76 Intercept, s, s2, ploidy, s:ploidy

No alleles S-morph 24 2220.72 2249.6 Intercept, s2 lost

S-morph 24 S-allele 1008.27 1053.43 Intercept, s

S-morph 24 s-allele 1024.76 1061.47 Intercept, s, ploidy

L-morph 42 m-allele 7.87 62.99 Intercept

L-morph 42 M-allele 3187.93 3216.39 Intercept, s2, s3

APPENDIX TO CHAPTER 2 251

No alleles L-morph 42 1635.27 1667.42 Intercept, s, s3 lost

L-morph 42 s-allele 7.87 62.99 Intercept

L-morph 42 S-allele 3187.93 3216.39 Intercept, s2, s3

M-morph 42 M-allele 94.18 138.82 Intercept, s2

M-morph 42 m-allele 866.01 889.36 Intercept, s, s2, s3, ploidy

No alleles M-morph 42 2066.48 2095.32 Intercept, s, s3 lost

M-morph 42 s-allele 7.8 62.43 Intercept

M-morph 42 S-allele 933.83 947.72 Intercept, s, s2, s3, ploidy, s:ploidy

S-morph 42 m-allele 1063.46 1084.4 Intercept, s2, s3, ploidy, s2:ploidy

APPENDIX TO CHAPTER 2 252

S-morph 42 M-allele 1831.38 1852.64 Intercept, s, s2, ploidy

No alleles S-morph 42 2861.06 2889.2 Intercept, s, s3 lost

S-morph 42 S-allele 123.13 167.46 Intercept, s2

S-morph 42 s-allele 762.67 799.09 Intercept, s3, ploidy

L-morph 72 m-allele 7.58 60.62 Intercept

L-morph 72 M-allele 2133.74 2166.05 Intercept, s2, s3

No alleles L-morph 72 1375.34 1410.84 Intercept, s lost

L-morph 72 s-allele 7.58 60.62 Intercept

L-morph 72 S-allele 2133.74 2166.05 Intercept, s2, s3

APPENDIX TO CHAPTER 2 253

M-morph 72 m-allele 410 437.81 Intercept, s2, s3, ploidy

M-morph 72 M-allele 18.74 63.51 Intercept, s

No alleles Intercept, s, s2, s3, ploidy, s:ploidy, s2: M-morph 72 1638.6 1646.66 lost ploidy

M-morph 72 s-allele 7.46 59.69 Intercept

M-morph 72 S-allele 419.22 440.63 Intercept, s, s2, s3, ploidy

S-morph 72 m-allele 510.84 531.47 Intercept, s2, s3, ploidy

S-morph 72 M-allele 1096.91 1114.5 Intercept, s2, s3, ploidy

No alleles S-morph 72 2018.96 2048.05 Intercept, s, s2, s3 lost

S-morph 72 S-allele 7.25 57.97 Intercept

APPENDIX TO CHAPTER 2 254

S-morph 72 s-allele 622.34 658.57 Intercept, s3, ploidy

L-morph 102 m-allele 7.38 59.04 Intercept

Intercept, s2, s3, ploidy, s2:ploidy, s3: L-morph 102 M-allele 1845.88 1855.7 ploidy

No alleles L-morph 102 1163.86 1185.82 Intercept, s, s2, s3 lost

L-morph 102 s-allele 7.38 59.04 Intercept

Intercept, s2, s3, ploidy, s2:ploidy, s3: L-morph 102 S-allele 1845.88 1855.7 ploidy

M-morph 102 M-allele 7.28 58.25 Intercept

M-morph 102 m-allele 371.32 398.31 Intercept, s2, s3, ploidy

APPENDIX TO CHAPTER 2 255

No alleles M-morph 102 1354.62 1385.94 Intercept, s, s3 lost

M-morph 102 s-allele 7.28 58.25 Intercept

M-morph 102 S-allele 371.32 398.31 Intercept, s2, s3, ploidy

S-morph 102 m-allele 307.5 324.82 Intercept, s2, s3, ploidy

S-morph 102 M-allele 788.49 810.66 Intercept, s2, s3, ploidy

No alleles S-morph 102 1614.9 1636.96 Intercept, s, s2, s3 lost

S-morph 102 S-allele 7.02 56.17 Intercept

S-morph 102 s-allele 520.78 555.88 Intercept, s3, ploidy

L-morph 201 m-allele 7.06 42.36 Intercept

APPENDIX TO CHAPTER 2 256

L-morph 201 M-allele 1205.76 1219.63 Intercept, s2

No alleles L-morph 201 701.81 748.21 Intercept, s lost

L-morph 201 s-allele 7.06 42.36 Intercept

L-morph 201 S-allele 1205.76 1219.63 Intercept, s2

M-morph 201 M-allele 6.95 41.67 Intercept

M-morph 201 m-allele 115.17 127.26 Intercept, s2

No alleles M-morph 201 926.45 958.79 Intercept, s3 lost

M-morph 201 s-allele 6.95 41.67 Intercept

M-morph 201 S-allele 115.17 127.26 Intercept, s2

APPENDIX TO CHAPTER 2 257

S-morph 201 m-allele 21.91 50.07 Intercept

S-morph 201 M-allele 430.35 450.19 Intercept, s2, ploidy

No alleles S-morph 201 772.78 804.72 Intercept, s3, ploidy lost

S-morph 201 S-allele 6.64 39.83 Intercept

S-morph 201 s-allele 420.04 439.95 Intercept, s2, ploidy

APPENDIX TO CHAPTER 2 258

Appendix 2 Table 3. Optimized models for the frequency in which trimorphism is lost from populations resulting in dimorphism. Optimized model terms describing the proportion of tristylous populations of different size and with different selfing morphs which became dimorphic at generations t = 100 and t = 200. Terms are separated by selfing morph, population size and generation. The models were optimized using the

“step” function until an optimized Bayesian Information Criterion (BIC) was obtained.

The saturated model BIC and the BIC of the optimized model are presented.

Optimized Selfing time Saturated Significant coefficients after N model morph (t) model BIC the model is analysed BIC

L-morph 9 100 63.96 107.94 Intercept

L-morph 9 200 7.07 56.57 Intercept

L-morph 24 100 1811.53 1849.85 Intercept, s2

L-morph 24 200 944.46 988.39 Intercept, s2

L-morph 42 100 1913.8 1946.92 Intercept, s, s3

L-morph 42 200 1636.39 1668.58 Intercept, s

L-morph 72 100 1539.24 1559.65 Intercept, s, s2, s3

L-morph 72 200 1375.34 1410.84 Intercept, s

L-morph 102 100 1179.21 1206.65 Intercept, s2

L-morph 102 200 1163.86 1185.82 Intercept, s, s2, s3

L-morph 201 100 927.15 973.6 Intercept, s3

L-morph 201 200 701.81 748.21 Intercept, s

APPENDIX TO CHAPTER 2 259

M-morph 9 100 51.92 96.8 Intercept

M-morph 9 200 7.06 56.52 Intercept

M-morph 24 100 2341.7 2378.47 Intercept, s, s3

M-morph 24 200 1353.4 1390.18 Intercept, s, s3

M-morph 42 100 2232.01 2265.65 Intercept, s, s3

M-morph 42 200 2028.71 2061.9 Intercept, s, s3

M-morph 72 100 1621.82 1648.55 Intercept, s, s3

Intercept, s, s2, s3, ploidy, M-morph 72 200 1598.5 1606.67 s:ploidy, s2:ploidy

M-morph 102 100 1367.04 1400.12 Intercept, s

M-morph 102 200 1318.98 1350.5 Intercept, s, s3

M-morph 201 100 949.04 984.9 Intercept, s3, ploidy

M-morph 201 200 919.49 951.84 Intercept, s3

S-morph 9 100 132.35 169.78 Intercept, s3

S-morph 9 200 7.15 57.17 Intercept

S-morph 24 100 3215.64 3261.47 Intercept, s2

S-morph 24 200 2075.37 2110.21 Intercept, s2

S-morph 42 100 2722.58 2754.03 Intercept, s, s3

S-morph 42 200 2695.34 2732.28 Intercept, s, s3

S-morph 72 100 1739.84 1765.89 Intercept, s, s2, s3

S-morph 72 200 1950.18 1980.93 Intercept, s, s2, s3

S-morph 102 100 1295.34 1317.69 Intercept, s, s2, s3

S-morph 102 200 1606.06 1628.35 Intercept, s, s2, s3

APPENDIX TO CHAPTER 2 260

S-morph 201 100 438.4 468.09 Intercept, s3, ploidy

S-morph 201 200 823.35 857.01 Intercept, s3, ploidy

APPENDIX TO CHAPTER 2 261

Appendix 2 Table 4. Frequency of dimorphic population structures evolving from trimorphism under different simulation parameters. Optimized model terms describing the frequency in which each dimorphic population structure (LM-, LS- and MS-) evolves from trimorphism for populations of different size and with different partially selfing morphs.

Terms are separated by selfing morph, population size, and dimorphic population structure. The models were optimized using the “step” function until an optimized

Bayesian Information Criterion (BIC) was obtained. The saturated model BIC and the BIC of the optimized model are presented.

Dimorphic Optimal Saturated Selfing time N population model model Significant coefficients morph (t) structure BIC BIC

L-morph 9 100 LM- 1471.82 1510.61 Intercept

L-morph 9 100 LS- 1430.23 1468.62 Intercept

L-morph 9 100 MS- 402.27 424.58 Intercept, s2, s3, ploidy

Intercept, s, s2, ploidy, M-morph 9 100 LM- 1482.66 1499.42 s2:ploidy

Intercept, s, s2, ploidy, M-morph 9 100 LS- 1357.83 1370.26 s2:ploidy

M-morph 9 100 MS- 640.69 665.7 Intercept, ploidy

Intercept, s, s3, ploidy, S-morph 9 100 LM- 1668.17 1688.69 s3:ploidy

S-morph 9 100 LS- 1631.57 1650.58 Intercept, s, s3, ploidy,

APPENDIX TO CHAPTER 2 262

s3:ploidy

Intercept, s, s2, ploidy, S-morph 9 100 MS- 794.66 814.86 s:ploidy

L-morph 24 100 LM- 2735.58 2782.26 Intercept

L-morph 24 100 LS- 2722.61 2767.88 Intercept

L-morph 24 100 MS- 125.79 171.27 Intercept, s

Intercept, s, s2, s3, ploidy, M-morph 24 100 LM- 1484.35 1496.27 s:ploidy

M-morph 24 100 LS- 586.52 624.5 Intercept, s

M-morph 24 100 MS- 1097.06 1121.85 Intercept, s2, s3, ploidy

S-morph 24 100 LM- 606.56 645.71 Intercept, s

Intercept, s, s2, ploidy, S-morph 24 100 LS- 1857.55 1875.68 s:ploidy

S-morph 24 100 MS- 1502.99 1533.89 Intercept, s, ploidy

L-morph 42 100 LM- 2196.87 2237.3 Intercept

L-morph 42 100 LS- 2196.87 2237.3 Intercept

L-morph 42 100 MS- 7.4 59.23 Intercept

Intercept, s, ploidy, M-morph 42 100 LM- 780.75 803.6 s:ploidy

M-morph 42 100 LS- 33.13 72.68 Intercept, s

M-morph 42 100 MS- 759.85 794.59 Intercept, ploidy

S-morph 42 100 LM- 65.86 101.71 Intercept, s

S-morph 42 100 LS- 899.67 922.96 Intercept, s, s2

APPENDIX TO CHAPTER 2 263

S-morph 42 100 MS- 858.68 879.51 Intercept, s, s2, s3

L-morph 72 100 LM- 1394.27 1438.3 Intercept

L-morph 72 100 LS- 1394.27 1438.3 Intercept

L-morph 72 100 MS- 6.96 55.71 Intercept

M-morph 72 100 LM- 325.29 357.66 Intercept, ploidy

M-morph 72 100 LS- 13.61 54.45 Intercept, s

M-morph 72 100 MS- 319.92 356.99 Intercept, ploidy

S-morph 72 100 LM- 6.22 49.8 Intercept

S-morph 72 100 LS- 359.75 393.61 Intercept, ploidy

S-morph 72 100 MS- 359.75 393.61 Intercept, ploidy

L-morph 102 100 LM- 1030.55 1043.13 Intercept, s, ploidy

L-morph 102 100 LS- 1030.55 1043.13 Intercept, s, ploidy

L-morph 102 100 MS- 6.65 39.87 Intercept

M-morph 102 100 LM- 207.2 237.58 Intercept, s, ploidy

M-morph 102 100 LS- 6.49 51.89 Intercept

M-morph 102 100 MS- 207.2 237.58 Intercept, s, ploidy

S-morph 102 100 LM- 5.71 39.95 Intercept

S-morph 102 100 LS- 225.21 249.95 Intercept, ploidy

S-morph 102 100 MS- 225.21 249.95 Intercept, ploidy

L-morph 201 100 LM- 454.42 470.61 Intercept

L-morph 201 100 LS- 454.42 470.61 Intercept

L-morph 201 100 MS- 5.84 23.36 Intercept

M-morph 201 100 LM- 54.55 77.44 Intercept, ploidy

APPENDIX TO CHAPTER 2 264

M-morph 201 100 LS- 5.82 29.09 Intercept

M-morph 201 100 MS- 54.55 77.44 Intercept, ploidy

S-morph 201 100 LM- 4.32 17.27 Intercept

S-morph 201 100 LS- 14.94 26.86 Intercept

S-morph 201 100 MS- 14.94 26.86 Intercept

APPENDIX TO CHAPTER 2 265

Appendix 2 Table 5. Comparison between simulation results in this study and that of

Heuch (1980) for allele loss from tristylous populations. The frequency of allele loss in simulated diploid and autotetraploid populations in the present study and in Heuch (1980) for the autotetraploid case a population sizes of N = 9 and 24. There were no significant differences in allele loss between the two studies.

m - M - S - Alleles N allele allele allele Χ2 df P persist lost lost lost

diploid 9 85 327 488 0 3.52 2 0.17

tetraploid 9 37 302 561 0 5.16 2 0.08

Heuch 9 5 43 49 0 1980

2 30 231 400 239 4.44 3 0.22 diploid 4

2 8 225 391 276 1.02 3 0.8 tetraploid 4

Heuch 2 0 26 42 32 1980 4

APPENDIX TO CHAPTER 2 266

Appendix 2 Table 6. Comparison of model genotype frequencies against deterministically predicted genotype frequencies at equilibrium in tristylous populations. The equilibrium mean and 95% confidence intervals (CI) of genotype frequencies sampled from 300 simulated populations of size N = 201 after t = 200 generations compared with deterministic mean genotype frequencies at population equilibrium, as predicted by Heuch and Lie (1985). A: diploid genotype frequencies; B: autotetraploid genotype.

Genotype Ssmm SsMm SsMM Ssmm SsMm SsMM SSmm SSMm SSMM

Observed

mean 0.333 0.307 0.025 0.180 0.130 0.025 0 0 0

Upper CI 0.336 0.31 0.026 0.183 0.132 0.026 0 0 0

Lower CI 0.330 0.305 0.024 0.178 0.128 0.024 0 0 0

Deterministic 0.333 0.309 0.024 0.179 0.131 0.024 0 0 0 mean

Genotype s4m4 s4Mm s4M2m2 s4M3m s4M4 Ss3m4 Ss3M Ss3 Ss3M3 Ss

APPENDIX TO CHAPTER 2 267

3 m3 M2m2 m 3

Observed 4.

* mean 0.336 0.293 0.036 0.003 4.42 0.181 0.121 0.028 0.002 98

8.* Upper CI 0.338 0.296 0.037 0.003 7.54* 0.184 0.123 0.029 0.003 28

1.* Lower CI 0.333 0.291 0.034 0.002 1.31* 0.179 0.119 0.027 0.002 67

Deterministic 6.* 0.333 0.296 0.035 0.002 6.00* 0.18 0.123 0.028 0.002 mean 00

*Cell value multiplied by 10-05 *

APPENDIX TO CHAPTER 2 268

Appendix 2 Table 7. Optimized model terms describing the frequency of each morph in populations of different size and with different partially selfing morphs. Terms are separated by selfing morph, population size, and morph. The models were optimized using the “step” function until an optimized Bayesian Information Criterion (BIC) was obtained.

The saturated model BIC and BIC of the optimized model are presented.

Optimum Saturated

Selfing Morph- model model

morph N frequency BIC BIC Significant coefficients

L-morph 9 L-morph -1333.43 -1305.92 Intercept, s, s2

L-morph 9 M-morph -438.3 -410.06 Intercept, s

L-morph 9 S-morph -607.55 -577.77 Intercept, s

Intercept, s, s2, ploidy, s:ploidy,

M-morph 9 L-morph -1036.76 -1023.87 s2:ploidy

M-morph 9 M-morph 136.39 166.86 Intercept, s, s3

M-morph 9 S-morph -460.42 -428.38 Intercept, s

S-morph 9 L-morph -840.1 -810.19 Intercept, s, s3

S-morph 9 M-morph -183.53 -151.78 Intercept, s3

S-morph 9 S-morph 592.27 617.52 Intercept, s, s3

L-morph 24 L-morph -5436.48 -5405.97 Intercept, s, s2, s3

L-morph 24 M-morph -2781.97 -2743.2 Intercept, s, s2

L-morph 24 S-morph -2892.1 -2853.11 Intercept, s

M-morph 24 L-morph -4448.92 -4437.28 Intercept, s2, s3, ploidy, s2:ploidy,

APPENDIX TO CHAPTER 2 269

s3:ploidy

M-morph 24 M-morph -4256.54 -4232.81 Intercept, s, s2, s3

M-morph 24 S-morph -4061.09 -4034.95 Intercept, s, s2

S-morph 24 L-morph -4229.74 -4203.71 Intercept, s, s3, ploidy

S-morph 24 M-morph -3902.3 -3869.89 Intercept, s, s2

S-morph 24 S-morph -4039.09 -4002.2 Intercept, s

L-morph 42 L-morph -7895.22 -7865.84 Intercept, s, s2, s3

L-morph 42 M-morph -5032.23 -4993.1 Intercept, s, s3

L-morph 42 S-morph -4814.09 -4778.94 Intercept, s, s2

M-morph 42 L-morph -6173.35 -6149.33 Intercept, s, s2, s3

M-morph 42 M-morph -7518.11 -7485.09 Intercept, s, s3

M-morph 42 S-morph -6034.03 -6014.03 Intercept, s, s2, s3

S-morph 42 L-morph -6327.7 -6301.88 Intercept, s, s3

S-morph 42 M-morph -5976.58 -5945.15 Intercept, s, s3

S-morph 42 S-morph -7158.51 -7118.89 Intercept, s, s3

L-morph 72 L-morph -9686.34 -9647.83 Intercept, s, s2

L-morph 72 M-morph -7240.65 -7212.11 Intercept, s, s2, s3

L-morph 72 S-morph -7275.5 -7247.48 Intercept, s, s2, s3

Intercept, s, s2, s3, ploidy, s:

M-morph 72 L-morph -8487.37 -8479.4 ploidy, s2:ploidy

M-morph 72 M-morph -9348.49 -9311.88 Intercept, s, s2

Intercept, s, s2, s3, ploidy, s:

M-morph 72 S-morph -8194.51 -8186.38 ploidy, s3:ploidy

APPENDIX TO CHAPTER 2 270

S-morph 72 L-morph -8370.87 -8342.39 Intercept, s, s2, s3

S-morph 72 M-morph -8316.01 -8289.12 Intercept, s, s2, s3

S-morph 72 S-morph -9574.93 -9537.95 Intercept, s, s2

L-morph 102 L-morph -11084.6 -11046.2 Intercept, s, s2

L-morph 102 M-morph -8813.89 -8787.1 Intercept, s, s3

L-morph 102 S-morph -8711.12 -8682.93 Intercept, s, s2, s3

M-morph 102 L-morph -9658.17 -9627.43 Intercept, s, s2, s3

M-morph 102 M-morph -10640 -10602.9 Intercept, s, s2

M-morph 102 S-morph -9445.43 -9407.08 Intercept, s2, s3

S-morph 102 L-morph -9705.23 -9687.45 Intercept, s, s2, s3

S-morph 102 M-morph -9468.07 -9447.19 Intercept, s, s2, s3

S-morph 102 S-morph -10749.2 -10714 Intercept, s, s2

L-morph 201 L-morph -13608.5 -13575.5 Intercept, s, s2

L-morph 201 M-morph -11550.9 -11512.7 Intercept, s, s3

L-morph 201 S-morph -11703.3 -11669.6 Intercept, s, s2

M-morph 201 L-morph -12162.2 -12139.9 Intercept, s, s2, s3

M-morph 201 M-morph -13056.8 -13023.5 Intercept, s, s2

M-morph 201 S-morph -11912.1 -11885.3 Intercept, s, s2, s3

S-morph 201 L-morph -12433.6 -12405.5 Intercept, s, s2, s3

S-morph 201 M-morph -12272.7 -12243.6 Intercept, s, s2, s3

S-morph 201 S-morph -13313.2 -13276.7 Intercept, s, s3

APPENDIX TO CHAPTER 2 271

Appendix 2 Table 8. Morph frequencies in trimorphic populations of different size, selfing rates and with different selfing morphs. Optimized model terms describing the frequency of each morph in populations of different sizes and with different selfing morphs. Terms are separated by selfing morph, population size, and morph. The models were optimized using the “step” function until an optimized Bayesian Information Criterion (BIC) was obtained. The saturated model BIC and the BIC of the optimized model are presented.

Selfing Morph which is being Optimal model Saturated model Significant model N morph measured BIC BIC coefficients

L-morph 9 L-morph -9.23 -9 Intercept, s

L-morph 9 M-morph -4.95 -3.45 Intercept

L-morph 9 S-morph -0.45 -0.21 Intercept, s

S-morph 9 L-morph -3.67 1.51 Intercept, s2, s3

S-morph 9 M-morph -7.31 1.51 Intercept

S-morph 9 S-morph -2.06 2.38 Intercept, s2, s3

L-morph 24 L-morph -673.71 -646.33 Intercept, s

L-morph 24 M-morph -674.55 -647.4 Intercept, s

APPENDIX TO CHAPTER 2 272

L-morph 24 S-morph -643.43 -614.48 Intercept, s

M-morph 24 L-morph -927.73 -893.29 Intercept, s

M-morph 24 M-morph -811.67 -776.05 Intercept, s

M-morph 24 S-morph -894.76 -865.6 Intercept, s

S-morph 24 L-morph -1514.95 -1475.64 Intercept, s

S-morph 24 M-morph -1377.01 -1340.29 Intercept, s

S-morph 24 S-morph -1413.96 -1379.56 Intercept, s, s2

L-morph 42 L-morph -2561.04 -2521.61 Intercept, s

L-morph 42 M-morph -2278.66 -2236.45 Intercept, s

L-morph 42 S-morph -2389.6 -2347.94 Intercept, s

M-morph 42 L-morph -2716.65 -2678.44 Intercept, s

M-morph 42 M-morph -2788.27 -2756.46 Intercept, s, s3

M-morph 42 S-morph -2815.85 -2778.01 Intercept, s

S-morph 42 L-morph -3608.08 -3575.58 Intercept, s, s2

S-morph 42 M-morph -3492.26 -3454.73 Intercept, s

S-morph 42 S-morph -3713.72 -3681.09 Intercept, s, s2

APPENDIX TO CHAPTER 2 273

L-morph 72 L-morph -4671.1 -4635.43 Intercept, s, s3

L-morph 72 M-morph -4396.74 -4361.93 Intercept, s

L-morph 72 S-morph -4474.11 -4443.27 Intercept, s, s3

M-morph 72 L-morph -4931.19 -4898.77 Intercept, s, s2

M-morph 72 M-morph -4869.35 -4834.94 Intercept, s, s2

M-morph 72 S-morph -4820.44 -4780.42 Intercept, s

S-morph 72 L-morph -5683.27 -5646.4 Intercept, s, s2

S-morph 72 M-morph -5715.65 -5670.95 Intercept, s

S-morph 72 S-morph -6114.48 -6076.7 Intercept, s, s2

L-morph 102 L-morph -6222.83 -6186.32 Intercept, s, s2

L-morph 102 M-morph -5801.11 -5761.79 Intercept, s

L-morph 102 S-morph -5834.23 -5800.31 Intercept, s, s2

M-morph 102 L-morph -6242.93 -6207.9 Intercept, s, s3

M-morph 102 M-morph -6366.83 -6332.45 Intercept, s, s2

M-morph 102 S-morph -6184.62 -6159.11 Intercept, s, s2, s3

S-morph 102 L-morph -7097.4 -7061.8 Intercept, s, s2

APPENDIX TO CHAPTER 2 274

S-morph 102 M-morph -6967.59 -6929.21 Intercept, s, s2

S-morph 102 S-morph -7567.63 -7530.07 Intercept, s, s2

L-morph 201 L-morph -9436.04 -9399.85 Intercept, s, s2

L-morph 201 M-morph -8719.35 -8682.27 Intercept, s, s2

L-morph 201 S-morph -8833.1 -8795.11 Intercept, s, s3

M-morph 201 L-morph -8939.83 -8903.31 Intercept, s, s3

M-morph 201 M-morph -9136.42 -9099.25 Intercept, s, s3

M-morph 201 S-morph -8821.87 -8784.36 Intercept, s, s3

S-morph 201 L-morph -9710.2 -9672.08 Intercept, s, s3

S-morph 201 M-morph -9603.82 -9565.79 Intercept, s, s3

S-morph 201 S-morph -10312.1 -10273.5 Intercept, s, s3

APPENDIX TO CHAPTER 2 275

Appendix 2 Fig. 1. Frequency of floral morphs in tristylous populations with morph- specific partial selfing. The frequency of the L-, M-, and S-morph (respective to row) in populations of different sizes (defined by column) after morph-specific selfing. The morph that is selfing increases in frequency in all cases.

APPENDIX TO CHAPTER 5

Appendix 5 Table 1. The identification number, genotype, frequency at equilibrium and floral morph for individuals of Lythrum salicaria. When I represented populations in data vectors in

R, I filled each vector element with the frequency of the corresponding genotype defined by the number in this table. The equilibrium morph frequency assumes equal morph fecundity and complete inter-morph mating, after Heuch and Lie (1985).

Frequency at Number Genotype Morph equilibrium

1 ssssmmmm 0.333 L

2 ssssMmmm 0.296

3 ssssMMmm 0.035 M 4 ssssMMMm 0.002

5 ssssMMMM 6.00 x 10-5

6 Ssssmmmm 0.18

7 SsssMmmm 0.123

8 SsssMMmm 0.028 S 9 SsssMMMm 0.002

10 SsssMMMM 6.00 x 10-5

11 SSssmmmm 0

276 APPENDIX TO CHAPTER 5 277

12 SSssMmmm 0

13 SSssMMmm 0

14 SSssMMMm 0

15 SSssMMMM 0

16 SSSsmmmm 0

17 SSSsMmmm 0

18 SSSsMMmm 0

19 SSSsMMMm 0

20 SSSsMMMM 0

21 SSSSmmmm 0

22 SSSSMmmm 0

23 SSSSMMmm 0

24 SSSSMMMm 0

25 SSSSMMMM 0

Appendix 5 Table 2. The number of maternal families of each genotype sampled from populations and the progeny of floral morphs and total progeny per genotype. This table is an expanded version of Table 5.2 which details progeny morph ratios from each of the genotypes in populations. The numbers after the genotypes are the values of i, j, or k which represent those genotypes in equations 5.1 and 5.2.

Family- specific progeny ratios Progeny morph ratios

Sampled

Morph genotype Sampled

Population Structure and number morph Families L M S n

ssssmmmm

153 L (1) L 4 48 1 1 50

ssssmmmm

68 L, M (1) L 16 194 190 0 384

278 APPENDIX TO CHAPTER 5 279

ssssMmmm

68 L, M (2) M 17 199 177 1 377

ssssMMmm

68 L, M (3) M 1 3 18 0 21

ssssmmmm

100 L, M (1) L 14 108 169 0 277

ssssMmmm

100 L, M (2) M 9 73 66 1 140

ssssMMmm

100 L, M (3) M 3 15 40 0 55

ssssmmmm

84 L, S (1) L 12 68 14 37 119

84 L, S S 14 40 5 40 85 Ssssmmmm

APPENDIX TO CHAPTER 5 280

(6)

SsssMmmm

84 L, S (7) S 3 13 6 11 30

ssssmmmm

92 L, M, S (1) L 8 44 37 22 103

ssssMmmm

92 L, M, S (2) M 3 12 13 15 40

ssssMMmm

92 L, M, S (3) M 5 7 31 35 73

Ssssmmmm

92 L, M, S (6) S 5 22 10 25 57

SsssMmmm

92 L, M, S (7) S 12 40 62 80 182

APPENDIX TO CHAPTER 5 281

ssssmmmm

135 L, M, S (1) L 10 107 19 62 188

ssssMmmm

135 L, M, S (2) M 1 12 8 0 20

Ssssmmmm

135 L, M, S (6) S 10 86 8 99 193

APPENDIX TO CHAPTER 5 282

Appendix 5 Table 3. Progeny morph ratios obtained from maternal parents in six natural populations of Lythrum salicaria of varying floral morph structure. The population, floral morph and observed progeny morph ratios are provided with P-values of

Χ2 tests for different expectations based on genotype, and the predicted genotypes. Expectations based on genotypes were split into simple genotypes (simplex M-allele for the M-morph, nulliplex M-allele for the S-morph) or complex genotypes (duplex

M-allele for the M-morph, simplex M-allele in the S-morph). The table identifies as “yes” families which did not significantly fit either expectation, or in which family morph ratios fit both expectations, according to Χ2 P-values being < 0.05 or > 0.05 for each case, respectively.

Family identification Observed progeny P-values after Χ2 tests Notes on Χ2 test

Plant Population Morph L- M- S- Simple Complex Neither fits Both fit Genotype

37 68 L 21 11 0 NA NA 1

38 68 L 11 20 0 NA NA 1

40 68 L 12 20 0 NA NA 1

41 68 L 17 12 0 NA NA 1

43 68 L 17 13 0 NA NA 1

APPENDIX TO CHAPTER 5 283

44 68 L 8 20 0 NA NA 1

47 68 L 15 11 0 NA NA 1

50 68 L 8 11 0 NA NA 1

51 68 L 7 7 0 NA NA 1

57 68 L 14 6 0 NA NA 1

61 68 L 9 7 0 NA NA 1

65 68 L 9 7 0 NA NA 1

69 68 L 14 5 0 NA NA 1

79 68 L 6 7 0 NA NA 1

80 68 L 13 15 0 NA NA 1

81 68 L 13 18 0 NA NA 1

39 68 M 17 13 0 9.0x10-1 8.7x10-48 2

42 68 M 13 12 0 9.9x10-1 2.6x10-33 2

APPENDIX TO CHAPTER 5 284

45 68 M 11 11 0 9.7x10-1 4.7x10-27 2

46 68 M 11 9 0 9.7x10-1 3.8x10-30 2

48 68 M 14 13 0 9.9x10-1 1.0x10-35 2

49 68 M 6 9 0 6.2x10-1 1.0x10-11 2

52 68 M 12 7 0 6.4x10-1 2.5x10-38 2

58 68 M 7 11 0 5.1x10-1 4.0x10-13 2

59 68 M 11 7 0 7.6x10-1 6.3x10-34 2

60 68 M 13 13 0 9.6x10-1 7.7x10-32 2

62 68 M 8 8 0 9.8x10-1 7.1x10-20 2

63 68 M 3 18 0 2.1x10-3 5.5x10-2 3

64 68 M 6 4 0 8.9x10-1 1.8x10-18 2

66 68 M 14 17 0 7.1x10-1 1.9x10-30 2

67 68 M 12 8 1 7.8x10-1 1.6x10-34 2

APPENDIX TO CHAPTER 5 285

68 68 M 14 10 0 8.4x10-1 8.6x10-41 2

71 68 M 19 11 0 4.9x10-1 1.8x10-60 2

73 68 M 11 14 0 6.9x10-1 2.4x10-23 2

86 84 L 10 10 0 NA NA 1

92 84 L 9 0 5 NA NA 1

95 84 L 2 3 0 NA NA 1

97 84 L 7 0 9 NA NA 1

100 84 L 11 0 7 NA NA 1

102 84 L 7 0 4 NA NA 1

110 84 L 3 0 2 NA NA 1

112 84 L 2 0 0 NA NA 1

123 84 L 3 0 5 NA NA 1

124 84 L 3 0 3 NA NA 1

APPENDIX TO CHAPTER 5 286

266 84 L 5 0 2 NA NA 1

268 84 L 6 1 0 NA NA 1

85 84 S 1 0 1 1.0x100 6.4x10-1 yes 6

87 84 S 2 0 3 8.9x10-1 4.3x10-1 yes 6

89 84 S 2 0 1 8.5x10-1 2.6x10-1 yes 6

90 84 S 6 1 6 9.6x10-1 1.8x10-1 yes 6

91 84 S 2 0 6 3.5x10-1 2.2x10-1 yes 6

93 84 S 6 0 6 9.9x10-1 6.8x10-2 yes 6

94 84 S 1 0 3 5.9x10-1 4.7x10-1 yes 6

98 84 S 3 0 3 1.0x100 2.6x10-1 yes 6

99 84 S 8 1 5 7.2x10-1 2.8x10-2 7

101 84 S 2 0 1 8.5x10-1 2.6x10-1 yes 6

108 84 S 3 0 1 6.2x10-1 8.3x10-2 yes 6

APPENDIX TO CHAPTER 5 287

109 84 S 3 1 0 2.9x10-1 6.3x10-2 yes 7

116 84 S 4 1 3 8.9x10-1 3.1x10-1 yes 6

260 84 S 5 3 3 5.6x10-1 2.7x10-1 yes 6

261 84 S 2 0 2 1.0x100 4.0x10-1 yes 6

262 84 S 1 0 1 1.0x100 6.4x10-1 yes 6

263 84 S 2 4 6 2.5x10-1 6.4x10-1 yes 7

127 92 L 4 3 1 NA NA 1

128 92 L 4 6 6 NA NA 1

130 92 L 7 5 0 NA NA 1

132 92 L 11 3 3 NA NA 1

140 92 L 4 5 3 NA NA 1

142 92 L 4 5 2 NA NA 1

146 92 L 5 8 2 NA NA 1

APPENDIX TO CHAPTER 5 288

160 92 L 5 2 5 NA NA 1

136 92 M 3 4 6 5.6x10-1 1.4x10-1 yes 2

137 92 M 1 6 7 1.4x10-1 3.6x10-1 yes 3

143 92 M 1 9 10 4.2x10-2 2.1x10-1 3

144 92 M 4 3 5 6.3x10-1 2.2x10-2 2

149 92 M 2 7 7 3.0x10-1 5.6x10-1 yes 3

152 92 M 5 6 4 8.9x10-1 2.6x10-2 2

156 92 M 2 6 7 3.1x10-1 4.2x10-1 yes 3

157 92 M 1 3 4 4.5x10-1 5.2x10-1 yes 3

129 92 S 6 7 6 2.5x10-2 2.1x10-1 7

134 92 S 7 4 9 8.1x10-1 1.4x10-1 yes 6

138 92 S 3 4 8 2.9x10-1 9.0x10-1 yes 7

139 92 S 4 8 9 9.7x10-3 8.0x10-1 7

APPENDIX TO CHAPTER 5 289

141 92 S 1 3 3 1.0x10-1 8.2x10-1 yes 7

147 92 S 3 5 10 1.4x10-1 8.7x10-1 yes 7

150 92 S 1 0 4 3.6x10-1 2.7x10-1 yes 6

151 92 S 3 6 11 6.3x10-2 8.6x10-1 yes 7

154 92 S 2 1 4 9.1x10-1 5.6x10-1 yes 6

155 92 S 1 4 2 7.4x10-3 3.5x10-1 7

158 92 S 6 4 4 2.0x10-1 5.9x10-2 yes 6

159 92 S 3 6 5 1.3x10-2 5.7x10-1 7

161 92 S 6 1 4 4.1x10-1 7.7x10-3 6

163 92 S 4 6 14 1.1x10-1 6.6x10-1 yes 7

164 92 S 3 3 4 4.1x10-1 6.4x10-1 yes 7

165 92 S 7 7 4 8.2x10-3 3.2x10-2 yes 7

166 92 S 2 3 4 2.8x10-1 9.4x10-1 yes 7

APPENDIX TO CHAPTER 5 290

179 100 L 4 11 0 NA NA 1

186 100 L 8 7 0 NA NA 1

188 100 L 5 12 0 NA NA 1

189 100 L 7 14 0 NA NA 1

190 100 L 6 10 0 NA NA 1

191 100 L 6 9 0 NA NA 1

192 100 L 7 12 0 NA NA 1

193 100 L 11 16 0 NA NA 1

195 100 L 8 15 0 NA NA 1

197 100 L 13 24 0 NA NA 1

199 100 L 11 7 0 NA NA 1

200 100 L 5 10 0 NA NA 1

201 100 L 12 12 0 NA NA 1

APPENDIX TO CHAPTER 5 291

202 100 L 5 10 0 NA NA 1

176 100 M 11 4 0 2.7x10-1 1.6x10-6 2

178 100 M 7 9 0 7.8x10-1 5.9x10-2 yes 2

180 100 M 10 13 0 6.8x10-1 1.9x10-2 2

182 100 M 3 3 0 9.9x10-1 1.8x10-1 yes 2

183 100 M 6 4 0 8.9x10-1 6.7x10-3 2

184 100 M 10 10 0 9.7x10-1 3.6x10-3 2

185 100 M 2 13 0 9.9x10-3 8.1x10-1 3

187 100 M 5 11 0 2.3x10-1 5.3x10-1 yes 3

194 100 M 13 8 0 6.8x10-1 9.8x10-6 2

196 100 M 8 16 0 1.7x10-1 2.6x10-1 yes 3

198 100 M 9 12 1 6.7x10-1 3.7x10-2 2

203 100 M 4 3 0 8.0x10-1 1.6x10-2 2

APPENDIX TO CHAPTER 5 292

209 135 L 10 2 6 NA NA 1

211 135 L 5 1 3 NA NA 1

212 135 L 16 5 2 NA NA 1

214 135 L 11 4 3 NA NA 1

215 135 L 10 3 5 NA NA 1

273 135 L 10 0 19 NA NA 1

275 135 L 8 1 5 NA NA 1

276 135 L 20 2 6 NA NA 1

278 135 L 8 1 6 NA NA 1

281 135 L 9 0 7 NA NA 1

213 135 M 12 8 0 2.6x10-2 1.2x10-10 yes 2

204 135 S 10 0 7 2.4x10-1 5.8x10-16 6

207 135 S 12 2 13 9.5x10-1 2.9x10-13 6

APPENDIX TO CHAPTER 5 293

208 135 S 10 1 11 7.9x10-1 1.4x10-11 6

210 135 S 9 1 9 8.2x10-1 5.1x10-11 6

216 135 S 8 3 15 4.8x10-1 1.4x10-5 6

274 135 S 7 0 5 3.8x10-1 2.7x10-11 6

277 135 S 17 0 12 9.3x10-2 1.6x10-26 6

280 135 S 3 0 6 4.7x10-1 5.1x10-3 6

283 135 S 6 1 14 2.8x10-1 3.2x10-4 6

285 135 S 4 0 7 4.7x10-1 4.1x10-4 6

167 153 L 13 0 0 NA NA 1

169 153 L 18 0 0 NA NA 1

170 153 L 6 0 0 NA NA 1

171 153 L 11 1 1 NA NA 1

APPENDIX TO CHAPTER 5 294

Appendix 5 Table 4. Progeny ratios of isolated plants of Lythrum salicaria in a field experiment at the Koffler Scientific

Reserve. Provided is the plant identification, floral morph, observed and expected progeny morph ratios after inter-morph and intra-morph mating, and the percent M- and S-morph progeny from M- and S-morph plants. I detected S-morph progeny in all but one L- and one M-morph maternal family and M-morph progeny in all L-morph maternal families demonstrating that inter- morph mating frequently occurred in the experiment. For location of plants in the field see Appendix 5 Fig 1.

Plant Progeny expected Progeny expected Percentage of Observed Progeny information inter-morph intra-morph progeny

Plant Morph L M S n L M S L M S M S

11 L 6 3 13 22 11 5 5 22 0 0 0.14 0.59

13 L 9 6 4 19 10 5 5 19 0 0 0.32 0.21

28 L 19 2 0 21 11 5 5 21 0 0 0.10 0.00

9 L 14 5 5 24 12 6 6 24 0 0 0.21 0.21

20 L 6 3 5 14 7 3 3 14 0 0 0.21 0.36

APPENDIX TO CHAPTER 5 295

30 L 8 8 3 19 10 5 5 19 0 0 0.42 0.16

6 L 16 11 3 30 15 7 7 30 0 0 0.37 0.10

32 L 14 3 7 24 12 6 6 24 0 0 0.13 0.29

57 L 15 6 6 27 14 6 7 27 0 0 0.22 0.22

10 M 8 10 1 19 8 7 5 5 14 0 0.53 0.05

1 M 5 6 0 11 4 4 3 3 8 0 0.55 0.00

22 M 6 3 3 12 5 4 3 3 9 0 0.25 0.25

27 M 4 7 0 11 4 4 3 3 8 0 0.64 0.00

29 M 9 5 1 15 6 5 4 4 11 0 0.33 0.07

7 M 1 3 4 8 3 3 2 2 6 0 0.38 0.50

33 M 10 10 0 20 8 7 5 6 14 0 0.50 0.00

APPENDIX TO CHAPTER 5 296

49 M 5 5 0 10 4 4 2 3 7 0 0.50 0.00

48 M 9 7 5 21 8 8 5 6 15 0 0.33 0.24

Progeny expected Progeny expected Percentage of Observed Progeny inter-morph intra-morph progeny

Totals Morph L M S n L M S L M S M S

L 107 47 46 200 103 48 49 200 0 0 0.24 0.23

M 57 56 14 127 50 45 31 35 92 0 0.44 0.11

APPENDIX TO CHAPTER 5 297

Appendix 5 Fig. 1. Locations and floral morphs of 57 isolated plants in two adjacent fields of Lythrum salicaria at the Koffler Scientific Reserve. Each plant and location is identified with a number (1 – 57) and a shape indicating floral morph. Numbers correspond to plant identities in Appendix 5 Table 3. Note that progenies were not grown from all plants in the field (see methods for specific details). The grey circles represent ponds on the reserve property; the space between the sections contains forests and inaccessible old fields. Map tile from Stamen (Stamen Design, Available from http://maps.stamen.com [accessed 11

February 2018]) via GGMAPS in R v 3.3.2 (Kahle and Wickam, 2013).

APPENDIX TO CHAPTER 6

Appendix 6 Table 1. The likelihood-ratio significance values for harvest data collected from the glasshouse experiment on

Lythrum salicaria in 2014 including germination, survival, flowering, flowering time and final inflorescence mass. All traits were tested by breeding treatment and competitive treatment and if a significant competitor effect for a trait was detected, I provide the significance of the interaction between breeding treatment and competitive environment.

Likelihood- Trait Distribution Treatment df P-value Time ratio Χ2

early Germination percent Binomial Breeding 10.99 1 < 0.001 life

Survival before early Binomial Breeding 1.71 1 > 0.15 transplant life

Survival at harvest Binomial Breeding 0.01 1 > 0.90 2014

Flowering percent Binomial Breeding 2.05 1 > 0.15 2014

298 APPENDIX TO CHAPTER 6 299

Continuous (log- Flowering time Breeding 11.01 1 < 0.001 2014 transformed)

Mass of inflorescence Continuous Breeding 19.21 1 < 1.2x10-5 2014

Survival at harvest Binomial Breeding 0.51 1 > 0.47 2015

Flowering percent Binomial Breeding 3.17 1 > 0.05 2015

Continuous (log- Flowering time Breeding 0.23 1 > 0.63 2015 transformed)

Continuous (log- Mass of inflorescence Breeding 1.34 1 > 0.24 2015 transformed)

Survival at harvest Binomial Breeding 0.45 1 > 0.50 2016

Flowering percent Binomial Breeding 8.21 1 < 0.01 2016

Continuous (log- Flowering time Breeding 2.22 1 > 0.13 2016 transformed)

APPENDIX TO CHAPTER 6 300

Continuous (log- Mass of inflorescence Breeding 5 1 < 0.05 2016 transformed)

Survival at harvest Binomial Breeding 3.33 1 > 0.05 2017

Flowering percent Binomial Breeding 3.64 1 > 0.05 2017

Continuous (log- Flowering time Breeding 1.49 1 > 0.22 2017 transformed)

Continuous (log- Mass of inflorescence Breeding 7.05 1 < 0.05 2017 transformed)

Survival at harvest Binomial Competition 2.99 2 > 0.22 2014

Flowering percent Binomial Competition 3.07 2 > 0.22 2014

Continuous (log- Flowering time Competition 0.67 2 > 0.72 2014 transformed)

Mass of inflorescence Continuous Competition 10.86 2 < 0.01 2014

APPENDIX TO CHAPTER 6 301

Binomial Competition 0.83 2 > 0.65 2015 Survival at harvest

Binomial Competition 0.59 2 > 0.70 2015 Flowering percent

Continuous (log- Flowering time Competition 1.37 2 > 0.50 2015 transformed)

Continuous (log- Mass of inflorescence Competition 1.57 2 > 0.45 2015 transformed)

Survival at harvest Binomial Competition 0.97 2 > 0.60 2016

Flowering percent Binomial Competition 1.55 2 > 0.45 2016

Continuous (log- Flowering time Competition 0.53 2 > 0.75 2016 transformed)

Continuous (log- Mass of inflorescence Competition 1.96 2 > 0.35 2016 transformed)

Survival at harvest Binomial Competition 8.2 2 < 0.05 2017

APPENDIX TO CHAPTER 6 302

Flowering percent Binomial Competition 2.87 2 > 0.20 2017

Continuous (log- Flowering time Competition 3.05 2 > 0.22 2017 transformed)

Continuous (log- Mass of inflorescence Competition 5.04 2 > 0.05 2017 transformed)

Breeding 6.67 1 < 0.01

Mass of inflorescence Continuous Competition 5.3 2 > 0.05 2014

Interaction 1.09 2 > 0.57

Breeding 1.75 1 > 0.18

Survival at harvest Binomial Competition 2.99 2 > 0.22 2017

Interaction 0.28 2 > 0.86

APPENDIX TO CHAPTER 6 303

Appendix 6 Table 2. The AIC values for nonlinear model types to which each year of growth data for Lythrum salicaria was fit. The Gompertz model possesses the lowest AIC in 2014 whereas the Logistic model possessed the lowest AIC in 2015 and 2016.

Model AIC values Year

Four-part logistic 12128.9 2014

Gompertz 10774.12 2014

Logistic 11015.83 2014

monomolecular 12443.63 2014

Four-part logistic Failed 2015

Gompertz 5200.62 2015

Logistic 5191.99 2015

Monomolecular 5346.67 2015

Four-part logistic Failed 2016

APPENDIX TO CHAPTER 6 304

Gompertz 5881.16 2016

Logistic 5872.04 2016

Monomolecular 5921.76 2016

APPENDIX TO CHAPTER 6 305

Appendix 6 Fig. 1. Correlation plots depicting the level of covariance between each measured trait in each of the six treatments and across each year observed in the inbreeding depression experiment on Lythrum salicaria. No correlations were consistently expressed between years and between treatments within years, which prohibited use of a single easily-measured trait or multivariate test in the analysis of inbreeding depression.

APPENDIX TO CHAPTER 6 306

Appendix 6 Fig. 2. The multiplicative depiction of relative performance (RP) of plants of

Lythrum salicaria in the different competitive environments of the study. The relative performance was calculated as 1 – ARG1/AGR2 if AGR2 > AGR1 or AGR2/AGR1 – 1 if

AGR2 < AGR1 with AGR equal to the mean multiplicative performance of plants with no competitor, a selfed competitor (S), or an outcrossed competitor (X). This measure could only be calculated from the resampling method due to unequal survival within families. In all but one case, the 95% confidence intervals of relative performance overlapped with zero. The exception occurred for cumulative performance of plants with a self-fertilized competitor, which performed slightly worse than those with no competitor